From 5304543ebd60778ad46123cd63142e27627fa150 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Parm=C3=A9nides=20GV?= Date: Tue, 4 Nov 2014 20:45:42 +0100 Subject: Update ics-openvpn to rev 906. --- .../main/openssl/crypto/bn/asm/README | 27 + .../main/openssl/crypto/bn/asm/alpha-mont.pl | 321 +++ .../main/openssl/crypto/bn/asm/armv4-gf2m.S | 201 ++ .../main/openssl/crypto/bn/asm/armv4-gf2m.pl | 281 +++ .../main/openssl/crypto/bn/asm/armv4-mont.S | 579 +++++ .../main/openssl/crypto/bn/asm/armv4-mont.pl | 675 +++++ .../main/openssl/crypto/bn/asm/bn-586.S | 1533 ++++++++++++ .../main/openssl/crypto/bn/asm/bn-586.pl | 774 ++++++ .../main/openssl/crypto/bn/asm/bn-mips.S | 2175 ++++++++++++++++ .../main/openssl/crypto/bn/asm/co-586.S | 1254 ++++++++++ .../main/openssl/crypto/bn/asm/co-586.pl | 287 +++ .../main/openssl/crypto/bn/asm/ia64-mont.pl | 851 +++++++ .../main/openssl/crypto/bn/asm/ia64.S | 1555 ++++++++++++ .../main/openssl/crypto/bn/asm/mips-mont.S | 284 +++ .../main/openssl/crypto/bn/asm/mips-mont.pl | 426 ++++ .../main/openssl/crypto/bn/asm/mips.pl | 2583 ++++++++++++++++++++ .../main/openssl/crypto/bn/asm/mips3-mont.pl | 327 +++ .../main/openssl/crypto/bn/asm/mips3.S | 2201 +++++++++++++++++ .../main/openssl/crypto/bn/asm/modexp512-x86_64.S | 1773 ++++++++++++++ .../main/openssl/crypto/bn/asm/modexp512-x86_64.pl | 1497 ++++++++++++ .../main/openssl/crypto/bn/asm/pa-risc2.S | 1618 ++++++++++++ .../main/openssl/crypto/bn/asm/pa-risc2W.S | 1605 ++++++++++++ .../main/openssl/crypto/bn/asm/parisc-mont.pl | 995 ++++++++ .../main/openssl/crypto/bn/asm/ppc-mont.pl | 334 +++ .../main/openssl/crypto/bn/asm/ppc.pl | 1998 +++++++++++++++ .../main/openssl/crypto/bn/asm/ppc64-mont.pl | 1088 +++++++++ .../main/openssl/crypto/bn/asm/s390x-gf2m.pl | 221 ++ .../main/openssl/crypto/bn/asm/s390x-mont.pl | 277 +++ .../main/openssl/crypto/bn/asm/s390x.S | 678 +++++ .../main/openssl/crypto/bn/asm/sparcv8.S | 1458 +++++++++++ .../main/openssl/crypto/bn/asm/sparcv8plus.S | 1558 ++++++++++++ .../main/openssl/crypto/bn/asm/sparcv9-mont.pl | 606 +++++ .../main/openssl/crypto/bn/asm/sparcv9a-mont.pl | 882 +++++++ .../main/openssl/crypto/bn/asm/via-mont.pl | 242 ++ .../main/openssl/crypto/bn/asm/x86-gf2m.S | 347 +++ .../main/openssl/crypto/bn/asm/x86-gf2m.pl | 313 +++ .../main/openssl/crypto/bn/asm/x86-mont.S | 460 ++++ .../main/openssl/crypto/bn/asm/x86-mont.pl | 593 +++++ .../main/openssl/crypto/bn/asm/x86.pl | 28 + .../main/openssl/crypto/bn/asm/x86/add.pl | 76 + .../main/openssl/crypto/bn/asm/x86/comba.pl | 277 +++ .../main/openssl/crypto/bn/asm/x86/div.pl | 15 + .../main/openssl/crypto/bn/asm/x86/f | 3 + .../main/openssl/crypto/bn/asm/x86/mul.pl | 77 + .../main/openssl/crypto/bn/asm/x86/mul_add.pl | 87 + .../main/openssl/crypto/bn/asm/x86/sqr.pl | 60 + .../main/openssl/crypto/bn/asm/x86/sub.pl | 76 + .../main/openssl/crypto/bn/asm/x86_64-gcc.c | 606 +++++ .../main/openssl/crypto/bn/asm/x86_64-gf2m.S | 291 +++ .../main/openssl/crypto/bn/asm/x86_64-gf2m.pl | 390 +++ .../main/openssl/crypto/bn/asm/x86_64-mont.S | 1374 +++++++++++ .../main/openssl/crypto/bn/asm/x86_64-mont.pl | 1681 +++++++++++++ .../main/openssl/crypto/bn/asm/x86_64-mont5.S | 784 ++++++ .../main/openssl/crypto/bn/asm/x86_64-mont5.pl | 1071 ++++++++ 54 files changed, 41773 insertions(+) create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/README create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/alpha-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-gf2m.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-gf2m.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-mont.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-586.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-586.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-mips.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/co-586.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/co-586.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/ia64-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/ia64.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips-mont.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips3-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips3.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/modexp512-x86_64.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/modexp512-x86_64.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/pa-risc2.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/pa-risc2W.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/parisc-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc64-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x-gf2m.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x-mont.pl create mode 100755 ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv8.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv8plus.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv9-mont.pl create mode 100755 ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv9a-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/via-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-gf2m.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-gf2m.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-mont.S create mode 100755 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/add.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/comba.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/div.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/f create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/mul.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/mul_add.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/sqr.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/sub.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-gcc.c create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-gf2m.S create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-gf2m.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont.S create mode 100755 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont.pl create mode 100644 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont5.S create mode 100755 ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont5.pl (limited to 'ics-openvpn-stripped/main/openssl/crypto/bn/asm') diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/README b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/README new file mode 100644 index 00000000..b0f3a68a --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/README @@ -0,0 +1,27 @@ + + +All assember in this directory are just version of the file +crypto/bn/bn_asm.c. + +Quite a few of these files are just the assember output from gcc since on +quite a few machines they are 2 times faster than the system compiler. + +For the x86, I have hand written assember because of the bad job all +compilers seem to do on it. This normally gives a 2 time speed up in the RSA +routines. + +For the DEC alpha, I also hand wrote the assember (except the division which +is just the output from the C compiler pasted on the end of the file). +On the 2 alpha C compilers I had access to, it was not possible to do +64b x 64b -> 128b calculations (both long and the long long data types +were 64 bits). So the hand assember gives access to the 128 bit result and +a 2 times speedup :-). + +There are 3 versions of assember for the HP PA-RISC. + +pa-risc.s is the origional one which works fine and generated using gcc :-) + +pa-risc2W.s and pa-risc2.s are 64 and 32-bit PA-RISC 2.0 implementations +by Chris Ruemmler from HP (with some help from the HP C compiler). + + diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/alpha-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/alpha-mont.pl new file mode 100644 index 00000000..03596e20 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/alpha-mont.pl @@ -0,0 +1,321 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# On 21264 RSA sign performance improves by 70/35/20/15 percent for +# 512/1024/2048/4096 bit key lengths. This is against vendor compiler +# instructed to '-tune host' code with in-line assembler. Other +# benchmarks improve by 15-20%. To anchor it to something else, the +# code provides approximately the same performance per GHz as AMD64. +# I.e. if you compare 1GHz 21264 and 2GHz Opteron, you'll observe ~2x +# difference. + +# int bn_mul_mont( +$rp="a0"; # BN_ULONG *rp, +$ap="a1"; # const BN_ULONG *ap, +$bp="a2"; # const BN_ULONG *bp, +$np="a3"; # const BN_ULONG *np, +$n0="a4"; # const BN_ULONG *n0, +$num="a5"; # int num); + +$lo0="t0"; +$hi0="t1"; +$lo1="t2"; +$hi1="t3"; +$aj="t4"; +$bi="t5"; +$nj="t6"; +$tp="t7"; +$alo="t8"; +$ahi="t9"; +$nlo="t10"; +$nhi="t11"; +$tj="t12"; +$i="s3"; +$j="s4"; +$m1="s5"; + +$code=<<___; +#ifdef __linux__ +#include +#else +#include +#include +#endif + +.text + +.set noat +.set noreorder + +.globl bn_mul_mont +.align 5 +.ent bn_mul_mont +bn_mul_mont: + lda sp,-48(sp) + stq ra,0(sp) + stq s3,8(sp) + stq s4,16(sp) + stq s5,24(sp) + stq fp,32(sp) + mov sp,fp + .mask 0x0400f000,-48 + .frame fp,48,ra + .prologue 0 + + .align 4 + .set reorder + sextl $num,$num + mov 0,v0 + cmplt $num,4,AT + bne AT,.Lexit + + ldq $hi0,0($ap) # ap[0] + s8addq $num,16,AT + ldq $aj,8($ap) + subq sp,AT,sp + ldq $bi,0($bp) # bp[0] + lda AT,-4096(zero) # mov -4096,AT + ldq $n0,0($n0) + and sp,AT,sp + + mulq $hi0,$bi,$lo0 + ldq $hi1,0($np) # np[0] + umulh $hi0,$bi,$hi0 + ldq $nj,8($np) + + mulq $lo0,$n0,$m1 + + mulq $hi1,$m1,$lo1 + umulh $hi1,$m1,$hi1 + + addq $lo1,$lo0,$lo1 + cmpult $lo1,$lo0,AT + addq $hi1,AT,$hi1 + + mulq $aj,$bi,$alo + mov 2,$j + umulh $aj,$bi,$ahi + mov sp,$tp + + mulq $nj,$m1,$nlo + s8addq $j,$ap,$aj + umulh $nj,$m1,$nhi + s8addq $j,$np,$nj +.align 4 +.L1st: + .set noreorder + ldq $aj,0($aj) + addl $j,1,$j + ldq $nj,0($nj) + lda $tp,8($tp) + + addq $alo,$hi0,$lo0 + mulq $aj,$bi,$alo + cmpult $lo0,$hi0,AT + addq $nlo,$hi1,$lo1 + + mulq $nj,$m1,$nlo + addq $ahi,AT,$hi0 + cmpult $lo1,$hi1,v0 + cmplt $j,$num,$tj + + umulh $aj,$bi,$ahi + addq $nhi,v0,$hi1 + addq $lo1,$lo0,$lo1 + s8addq $j,$ap,$aj + + umulh $nj,$m1,$nhi + cmpult $lo1,$lo0,v0 + addq $hi1,v0,$hi1 + s8addq $j,$np,$nj + + stq $lo1,-8($tp) + nop + unop + bne $tj,.L1st + .set reorder + + addq $alo,$hi0,$lo0 + addq $nlo,$hi1,$lo1 + cmpult $lo0,$hi0,AT + cmpult $lo1,$hi1,v0 + addq $ahi,AT,$hi0 + addq $nhi,v0,$hi1 + + addq $lo1,$lo0,$lo1 + cmpult $lo1,$lo0,v0 + addq $hi1,v0,$hi1 + + stq $lo1,0($tp) + + addq $hi1,$hi0,$hi1 + cmpult $hi1,$hi0,AT + stq $hi1,8($tp) + stq AT,16($tp) + + mov 1,$i +.align 4 +.Louter: + s8addq $i,$bp,$bi + ldq $hi0,0($ap) + ldq $aj,8($ap) + ldq $bi,0($bi) + ldq $hi1,0($np) + ldq $nj,8($np) + ldq $tj,0(sp) + + mulq $hi0,$bi,$lo0 + umulh $hi0,$bi,$hi0 + + addq $lo0,$tj,$lo0 + cmpult $lo0,$tj,AT + addq $hi0,AT,$hi0 + + mulq $lo0,$n0,$m1 + + mulq $hi1,$m1,$lo1 + umulh $hi1,$m1,$hi1 + + addq $lo1,$lo0,$lo1 + cmpult $lo1,$lo0,AT + mov 2,$j + addq $hi1,AT,$hi1 + + mulq $aj,$bi,$alo + mov sp,$tp + umulh $aj,$bi,$ahi + + mulq $nj,$m1,$nlo + s8addq $j,$ap,$aj + umulh $nj,$m1,$nhi +.align 4 +.Linner: + .set noreorder + ldq $tj,8($tp) #L0 + nop #U1 + ldq $aj,0($aj) #L1 + s8addq $j,$np,$nj #U0 + + ldq $nj,0($nj) #L0 + nop #U1 + addq $alo,$hi0,$lo0 #L1 + lda $tp,8($tp) + + mulq $aj,$bi,$alo #U1 + cmpult $lo0,$hi0,AT #L0 + addq $nlo,$hi1,$lo1 #L1 + addl $j,1,$j + + mulq $nj,$m1,$nlo #U1 + addq $ahi,AT,$hi0 #L0 + addq $lo0,$tj,$lo0 #L1 + cmpult $lo1,$hi1,v0 #U0 + + umulh $aj,$bi,$ahi #U1 + cmpult $lo0,$tj,AT #L0 + addq $lo1,$lo0,$lo1 #L1 + addq $nhi,v0,$hi1 #U0 + + umulh $nj,$m1,$nhi #U1 + s8addq $j,$ap,$aj #L0 + cmpult $lo1,$lo0,v0 #L1 + cmplt $j,$num,$tj #U0 # borrow $tj + + addq $hi0,AT,$hi0 #L0 + addq $hi1,v0,$hi1 #U1 + stq $lo1,-8($tp) #L1 + bne $tj,.Linner #U0 + .set reorder + + ldq $tj,8($tp) + addq $alo,$hi0,$lo0 + addq $nlo,$hi1,$lo1 + cmpult $lo0,$hi0,AT + cmpult $lo1,$hi1,v0 + addq $ahi,AT,$hi0 + addq $nhi,v0,$hi1 + + addq $lo0,$tj,$lo0 + cmpult $lo0,$tj,AT + addq $hi0,AT,$hi0 + + ldq $tj,16($tp) + addq $lo1,$lo0,$j + cmpult $j,$lo0,v0 + addq $hi1,v0,$hi1 + + addq $hi1,$hi0,$lo1 + stq $j,0($tp) + cmpult $lo1,$hi0,$hi1 + addq $lo1,$tj,$lo1 + cmpult $lo1,$tj,AT + addl $i,1,$i + addq $hi1,AT,$hi1 + stq $lo1,8($tp) + cmplt $i,$num,$tj # borrow $tj + stq $hi1,16($tp) + bne $tj,.Louter + + s8addq $num,sp,$tj # &tp[num] + mov $rp,$bp # put rp aside + mov sp,$tp + mov sp,$ap + mov 0,$hi0 # clear borrow bit + +.align 4 +.Lsub: ldq $lo0,0($tp) + ldq $lo1,0($np) + lda $tp,8($tp) + lda $np,8($np) + subq $lo0,$lo1,$lo1 # tp[i]-np[i] + cmpult $lo0,$lo1,AT + subq $lo1,$hi0,$lo0 + cmpult $lo1,$lo0,$hi0 + or $hi0,AT,$hi0 + stq $lo0,0($rp) + cmpult $tp,$tj,v0 + lda $rp,8($rp) + bne v0,.Lsub + + subq $hi1,$hi0,$hi0 # handle upmost overflow bit + mov sp,$tp + mov $bp,$rp # restore rp + + and sp,$hi0,$ap + bic $bp,$hi0,$bp + bis $bp,$ap,$ap # ap=borrow?tp:rp + +.align 4 +.Lcopy: ldq $aj,0($ap) # copy or in-place refresh + lda $tp,8($tp) + lda $rp,8($rp) + lda $ap,8($ap) + stq zero,-8($tp) # zap tp + cmpult $tp,$tj,AT + stq $aj,-8($rp) + bne AT,.Lcopy + mov 1,v0 + +.Lexit: + .set noreorder + mov fp,sp + /*ldq ra,0(sp)*/ + ldq s3,8(sp) + ldq s4,16(sp) + ldq s5,24(sp) + ldq fp,32(sp) + lda sp,48(sp) + ret (ra) +.end bn_mul_mont +.ascii "Montgomery Multiplication for Alpha, CRYPTOGAMS by " +.align 2 +___ + +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-gf2m.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-gf2m.S new file mode 100644 index 00000000..0fa25b26 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-gf2m.S @@ -0,0 +1,201 @@ +#include "arm_arch.h" + +.text +.code 32 + +#if __ARM_ARCH__>=7 +.fpu neon +#endif +.type mul_1x1_ialu,%function +.align 5 +mul_1x1_ialu: + mov r4,#0 + bic r5,r1,#3<<30 @ a1=a&0x3fffffff + str r4,[sp,#0] @ tab[0]=0 + add r6,r5,r5 @ a2=a1<<1 + str r5,[sp,#4] @ tab[1]=a1 + eor r7,r5,r6 @ a1^a2 + str r6,[sp,#8] @ tab[2]=a2 + mov r8,r5,lsl#2 @ a4=a1<<2 + str r7,[sp,#12] @ tab[3]=a1^a2 + eor r9,r5,r8 @ a1^a4 + str r8,[sp,#16] @ tab[4]=a4 + eor r4,r6,r8 @ a2^a4 + str r9,[sp,#20] @ tab[5]=a1^a4 + eor r7,r7,r8 @ a1^a2^a4 + str r4,[sp,#24] @ tab[6]=a2^a4 + and r8,r12,r0,lsl#2 + str r7,[sp,#28] @ tab[7]=a1^a2^a4 + + and r9,r12,r0,lsr#1 + ldr r5,[sp,r8] @ tab[b & 0x7] + and r8,r12,r0,lsr#4 + ldr r7,[sp,r9] @ tab[b >> 3 & 0x7] + and r9,r12,r0,lsr#7 + ldr r6,[sp,r8] @ tab[b >> 6 & 0x7] + eor r5,r5,r7,lsl#3 @ stall + mov r4,r7,lsr#29 + ldr r7,[sp,r9] @ tab[b >> 9 & 0x7] + + and r8,r12,r0,lsr#10 + eor r5,r5,r6,lsl#6 + eor r4,r4,r6,lsr#26 + ldr r6,[sp,r8] @ tab[b >> 12 & 0x7] + + and r9,r12,r0,lsr#13 + eor r5,r5,r7,lsl#9 + eor r4,r4,r7,lsr#23 + ldr r7,[sp,r9] @ tab[b >> 15 & 0x7] + + and r8,r12,r0,lsr#16 + eor r5,r5,r6,lsl#12 + eor r4,r4,r6,lsr#20 + ldr r6,[sp,r8] @ tab[b >> 18 & 0x7] + + and r9,r12,r0,lsr#19 + eor r5,r5,r7,lsl#15 + eor r4,r4,r7,lsr#17 + ldr r7,[sp,r9] @ tab[b >> 21 & 0x7] + + and r8,r12,r0,lsr#22 + eor r5,r5,r6,lsl#18 + eor r4,r4,r6,lsr#14 + ldr r6,[sp,r8] @ tab[b >> 24 & 0x7] + + and r9,r12,r0,lsr#25 + eor r5,r5,r7,lsl#21 + eor r4,r4,r7,lsr#11 + ldr r7,[sp,r9] @ tab[b >> 27 & 0x7] + + tst r1,#1<<30 + and r8,r12,r0,lsr#28 + eor r5,r5,r6,lsl#24 + eor r4,r4,r6,lsr#8 + ldr r6,[sp,r8] @ tab[b >> 30 ] + + eorne r5,r5,r0,lsl#30 + eorne r4,r4,r0,lsr#2 + tst r1,#1<<31 + eor r5,r5,r7,lsl#27 + eor r4,r4,r7,lsr#5 + eorne r5,r5,r0,lsl#31 + eorne r4,r4,r0,lsr#1 + eor r5,r5,r6,lsl#30 + eor r4,r4,r6,lsr#2 + + mov pc,lr +.size mul_1x1_ialu,.-mul_1x1_ialu +.global bn_GF2m_mul_2x2 +.type bn_GF2m_mul_2x2,%function +.align 5 +bn_GF2m_mul_2x2: +#if __ARM_ARCH__>=7 + ldr r12,.LOPENSSL_armcap +.Lpic: ldr r12,[pc,r12] + tst r12,#1 + beq .Lialu + + ldr r12, [sp] @ 5th argument + vmov.32 d26, r2, r1 + vmov.32 d27, r12, r3 + vmov.i64 d28, #0x0000ffffffffffff + vmov.i64 d29, #0x00000000ffffffff + vmov.i64 d30, #0x000000000000ffff + + vext.8 d2, d26, d26, #1 @ A1 + vmull.p8 q1, d2, d27 @ F = A1*B + vext.8 d0, d27, d27, #1 @ B1 + vmull.p8 q0, d26, d0 @ E = A*B1 + vext.8 d4, d26, d26, #2 @ A2 + vmull.p8 q2, d4, d27 @ H = A2*B + vext.8 d16, d27, d27, #2 @ B2 + vmull.p8 q8, d26, d16 @ G = A*B2 + vext.8 d6, d26, d26, #3 @ A3 + veor q1, q1, q0 @ L = E + F + vmull.p8 q3, d6, d27 @ J = A3*B + vext.8 d0, d27, d27, #3 @ B3 + veor q2, q2, q8 @ M = G + H + vmull.p8 q0, d26, d0 @ I = A*B3 + veor d2, d2, d3 @ t0 = (L) (P0 + P1) << 8 + vand d3, d3, d28 + vext.8 d16, d27, d27, #4 @ B4 + veor d4, d4, d5 @ t1 = (M) (P2 + P3) << 16 + vand d5, d5, d29 + vmull.p8 q8, d26, d16 @ K = A*B4 + veor q3, q3, q0 @ N = I + J + veor d2, d2, d3 + veor d4, d4, d5 + veor d6, d6, d7 @ t2 = (N) (P4 + P5) << 24 + vand d7, d7, d30 + vext.8 q1, q1, q1, #15 + veor d16, d16, d17 @ t3 = (K) (P6 + P7) << 32 + vmov.i64 d17, #0 + vext.8 q2, q2, q2, #14 + veor d6, d6, d7 + vmull.p8 q0, d26, d27 @ D = A*B + vext.8 q8, q8, q8, #12 + vext.8 q3, q3, q3, #13 + veor q1, q1, q2 + veor q3, q3, q8 + veor q0, q0, q1 + veor q0, q0, q3 + + vst1.32 {q0}, [r0] + bx lr @ bx lr +.align 4 +.Lialu: +#endif + stmdb sp!,{r4-r10,lr} + mov r10,r0 @ reassign 1st argument + mov r0,r3 @ r0=b1 + ldr r3,[sp,#32] @ load b0 + mov r12,#7<<2 + sub sp,sp,#32 @ allocate tab[8] + + bl mul_1x1_ialu @ a1·b1 + str r5,[r10,#8] + str r4,[r10,#12] + + eor r0,r0,r3 @ flip b0 and b1 + eor r1,r1,r2 @ flip a0 and a1 + eor r3,r3,r0 + eor r2,r2,r1 + eor r0,r0,r3 + eor r1,r1,r2 + bl mul_1x1_ialu @ a0·b0 + str r5,[r10] + str r4,[r10,#4] + + eor r1,r1,r2 + eor r0,r0,r3 + bl mul_1x1_ialu @ (a1+a0)·(b1+b0) + ldmia r10,{r6-r9} + eor r5,r5,r4 + eor r4,r4,r7 + eor r5,r5,r6 + eor r4,r4,r8 + eor r5,r5,r9 + eor r4,r4,r9 + str r4,[r10,#8] + eor r5,r5,r4 + add sp,sp,#32 @ destroy tab[8] + str r5,[r10,#4] + +#if __ARM_ARCH__>=5 + ldmia sp!,{r4-r10,pc} +#else + ldmia sp!,{r4-r10,lr} + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet + .word 0xe12fff1e @ interoperable with Thumb ISA:-) +#endif +.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2 +#if __ARM_ARCH__>=7 +.align 5 +.LOPENSSL_armcap: +.word OPENSSL_armcap_P-(.Lpic+8) +#endif +.asciz "GF(2^m) Multiplication for ARMv4/NEON, CRYPTOGAMS by " +.align 5 + +.comm OPENSSL_armcap_P,4,4 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-gf2m.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-gf2m.pl new file mode 100644 index 00000000..3f1f4f67 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-gf2m.pl @@ -0,0 +1,281 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# May 2011 +# +# The module implements bn_GF2m_mul_2x2 polynomial multiplication +# used in bn_gf2m.c. It's kind of low-hanging mechanical port from +# C for the time being... Except that it has two code paths: pure +# integer code suitable for any ARMv4 and later CPU and NEON code +# suitable for ARMv7. Pure integer 1x1 multiplication subroutine runs +# in ~45 cycles on dual-issue core such as Cortex A8, which is ~50% +# faster than compiler-generated code. For ECDH and ECDSA verify (but +# not for ECDSA sign) it means 25%-45% improvement depending on key +# length, more for longer keys. Even though NEON 1x1 multiplication +# runs in even less cycles, ~30, improvement is measurable only on +# longer keys. One has to optimize code elsewhere to get NEON glow... +# +# April 2014 +# +# Double bn_GF2m_mul_2x2 performance by using algorithm from paper +# referred below, which improves ECDH and ECDSA verify benchmarks +# by 18-40%. +# +# Câmara, D.; Gouvêa, C. P. L.; López, J. & Dahab, R.: Fast Software +# Polynomial Multiplication on ARM Processors using the NEON Engine. +# +# http://conradoplg.cryptoland.net/files/2010/12/mocrysen13.pdf + +while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +$code=<<___; +#include "arm_arch.h" + +.text +.code 32 + +#if __ARM_ARCH__>=7 +.fpu neon +#endif +___ +################ +# private interface to mul_1x1_ialu +# +$a="r1"; +$b="r0"; + +($a0,$a1,$a2,$a12,$a4,$a14)= +($hi,$lo,$t0,$t1, $i0,$i1 )=map("r$_",(4..9),12); + +$mask="r12"; + +$code.=<<___; +.type mul_1x1_ialu,%function +.align 5 +mul_1x1_ialu: + mov $a0,#0 + bic $a1,$a,#3<<30 @ a1=a&0x3fffffff + str $a0,[sp,#0] @ tab[0]=0 + add $a2,$a1,$a1 @ a2=a1<<1 + str $a1,[sp,#4] @ tab[1]=a1 + eor $a12,$a1,$a2 @ a1^a2 + str $a2,[sp,#8] @ tab[2]=a2 + mov $a4,$a1,lsl#2 @ a4=a1<<2 + str $a12,[sp,#12] @ tab[3]=a1^a2 + eor $a14,$a1,$a4 @ a1^a4 + str $a4,[sp,#16] @ tab[4]=a4 + eor $a0,$a2,$a4 @ a2^a4 + str $a14,[sp,#20] @ tab[5]=a1^a4 + eor $a12,$a12,$a4 @ a1^a2^a4 + str $a0,[sp,#24] @ tab[6]=a2^a4 + and $i0,$mask,$b,lsl#2 + str $a12,[sp,#28] @ tab[7]=a1^a2^a4 + + and $i1,$mask,$b,lsr#1 + ldr $lo,[sp,$i0] @ tab[b & 0x7] + and $i0,$mask,$b,lsr#4 + ldr $t1,[sp,$i1] @ tab[b >> 3 & 0x7] + and $i1,$mask,$b,lsr#7 + ldr $t0,[sp,$i0] @ tab[b >> 6 & 0x7] + eor $lo,$lo,$t1,lsl#3 @ stall + mov $hi,$t1,lsr#29 + ldr $t1,[sp,$i1] @ tab[b >> 9 & 0x7] + + and $i0,$mask,$b,lsr#10 + eor $lo,$lo,$t0,lsl#6 + eor $hi,$hi,$t0,lsr#26 + ldr $t0,[sp,$i0] @ tab[b >> 12 & 0x7] + + and $i1,$mask,$b,lsr#13 + eor $lo,$lo,$t1,lsl#9 + eor $hi,$hi,$t1,lsr#23 + ldr $t1,[sp,$i1] @ tab[b >> 15 & 0x7] + + and $i0,$mask,$b,lsr#16 + eor $lo,$lo,$t0,lsl#12 + eor $hi,$hi,$t0,lsr#20 + ldr $t0,[sp,$i0] @ tab[b >> 18 & 0x7] + + and $i1,$mask,$b,lsr#19 + eor $lo,$lo,$t1,lsl#15 + eor $hi,$hi,$t1,lsr#17 + ldr $t1,[sp,$i1] @ tab[b >> 21 & 0x7] + + and $i0,$mask,$b,lsr#22 + eor $lo,$lo,$t0,lsl#18 + eor $hi,$hi,$t0,lsr#14 + ldr $t0,[sp,$i0] @ tab[b >> 24 & 0x7] + + and $i1,$mask,$b,lsr#25 + eor $lo,$lo,$t1,lsl#21 + eor $hi,$hi,$t1,lsr#11 + ldr $t1,[sp,$i1] @ tab[b >> 27 & 0x7] + + tst $a,#1<<30 + and $i0,$mask,$b,lsr#28 + eor $lo,$lo,$t0,lsl#24 + eor $hi,$hi,$t0,lsr#8 + ldr $t0,[sp,$i0] @ tab[b >> 30 ] + + eorne $lo,$lo,$b,lsl#30 + eorne $hi,$hi,$b,lsr#2 + tst $a,#1<<31 + eor $lo,$lo,$t1,lsl#27 + eor $hi,$hi,$t1,lsr#5 + eorne $lo,$lo,$b,lsl#31 + eorne $hi,$hi,$b,lsr#1 + eor $lo,$lo,$t0,lsl#30 + eor $hi,$hi,$t0,lsr#2 + + mov pc,lr +.size mul_1x1_ialu,.-mul_1x1_ialu +___ +################ +# void bn_GF2m_mul_2x2(BN_ULONG *r, +# BN_ULONG a1,BN_ULONG a0, +# BN_ULONG b1,BN_ULONG b0); # r[3..0]=a1a0·b1b0 +{ +my ($r,$t0,$t1,$t2,$t3)=map("q$_",(0..3,8..12)); +my ($a,$b,$k48,$k32,$k16)=map("d$_",(26..31)); + +$code.=<<___; +.global bn_GF2m_mul_2x2 +.type bn_GF2m_mul_2x2,%function +.align 5 +bn_GF2m_mul_2x2: +#if __ARM_ARCH__>=7 + ldr r12,.LOPENSSL_armcap +.Lpic: ldr r12,[pc,r12] + tst r12,#1 + beq .Lialu + + ldr r12, [sp] @ 5th argument + vmov.32 $a, r2, r1 + vmov.32 $b, r12, r3 + vmov.i64 $k48, #0x0000ffffffffffff + vmov.i64 $k32, #0x00000000ffffffff + vmov.i64 $k16, #0x000000000000ffff + + vext.8 $t0#lo, $a, $a, #1 @ A1 + vmull.p8 $t0, $t0#lo, $b @ F = A1*B + vext.8 $r#lo, $b, $b, #1 @ B1 + vmull.p8 $r, $a, $r#lo @ E = A*B1 + vext.8 $t1#lo, $a, $a, #2 @ A2 + vmull.p8 $t1, $t1#lo, $b @ H = A2*B + vext.8 $t3#lo, $b, $b, #2 @ B2 + vmull.p8 $t3, $a, $t3#lo @ G = A*B2 + vext.8 $t2#lo, $a, $a, #3 @ A3 + veor $t0, $t0, $r @ L = E + F + vmull.p8 $t2, $t2#lo, $b @ J = A3*B + vext.8 $r#lo, $b, $b, #3 @ B3 + veor $t1, $t1, $t3 @ M = G + H + vmull.p8 $r, $a, $r#lo @ I = A*B3 + veor $t0#lo, $t0#lo, $t0#hi @ t0 = (L) (P0 + P1) << 8 + vand $t0#hi, $t0#hi, $k48 + vext.8 $t3#lo, $b, $b, #4 @ B4 + veor $t1#lo, $t1#lo, $t1#hi @ t1 = (M) (P2 + P3) << 16 + vand $t1#hi, $t1#hi, $k32 + vmull.p8 $t3, $a, $t3#lo @ K = A*B4 + veor $t2, $t2, $r @ N = I + J + veor $t0#lo, $t0#lo, $t0#hi + veor $t1#lo, $t1#lo, $t1#hi + veor $t2#lo, $t2#lo, $t2#hi @ t2 = (N) (P4 + P5) << 24 + vand $t2#hi, $t2#hi, $k16 + vext.8 $t0, $t0, $t0, #15 + veor $t3#lo, $t3#lo, $t3#hi @ t3 = (K) (P6 + P7) << 32 + vmov.i64 $t3#hi, #0 + vext.8 $t1, $t1, $t1, #14 + veor $t2#lo, $t2#lo, $t2#hi + vmull.p8 $r, $a, $b @ D = A*B + vext.8 $t3, $t3, $t3, #12 + vext.8 $t2, $t2, $t2, #13 + veor $t0, $t0, $t1 + veor $t2, $t2, $t3 + veor $r, $r, $t0 + veor $r, $r, $t2 + + vst1.32 {$r}, [r0] + ret @ bx lr +.align 4 +.Lialu: +#endif +___ +} +$ret="r10"; # reassigned 1st argument +$code.=<<___; + stmdb sp!,{r4-r10,lr} + mov $ret,r0 @ reassign 1st argument + mov $b,r3 @ $b=b1 + ldr r3,[sp,#32] @ load b0 + mov $mask,#7<<2 + sub sp,sp,#32 @ allocate tab[8] + + bl mul_1x1_ialu @ a1·b1 + str $lo,[$ret,#8] + str $hi,[$ret,#12] + + eor $b,$b,r3 @ flip b0 and b1 + eor $a,$a,r2 @ flip a0 and a1 + eor r3,r3,$b + eor r2,r2,$a + eor $b,$b,r3 + eor $a,$a,r2 + bl mul_1x1_ialu @ a0·b0 + str $lo,[$ret] + str $hi,[$ret,#4] + + eor $a,$a,r2 + eor $b,$b,r3 + bl mul_1x1_ialu @ (a1+a0)·(b1+b0) +___ +@r=map("r$_",(6..9)); +$code.=<<___; + ldmia $ret,{@r[0]-@r[3]} + eor $lo,$lo,$hi + eor $hi,$hi,@r[1] + eor $lo,$lo,@r[0] + eor $hi,$hi,@r[2] + eor $lo,$lo,@r[3] + eor $hi,$hi,@r[3] + str $hi,[$ret,#8] + eor $lo,$lo,$hi + add sp,sp,#32 @ destroy tab[8] + str $lo,[$ret,#4] + +#if __ARM_ARCH__>=5 + ldmia sp!,{r4-r10,pc} +#else + ldmia sp!,{r4-r10,lr} + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet + bx lr @ interoperable with Thumb ISA:-) +#endif +.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2 +#if __ARM_ARCH__>=7 +.align 5 +.LOPENSSL_armcap: +.word OPENSSL_armcap_P-(.Lpic+8) +#endif +.asciz "GF(2^m) Multiplication for ARMv4/NEON, CRYPTOGAMS by " +.align 5 + +.comm OPENSSL_armcap_P,4,4 +___ + +foreach (split("\n",$code)) { + s/\`([^\`]*)\`/eval $1/geo; + + s/\bq([0-9]+)#(lo|hi)/sprintf "d%d",2*$1+($2 eq "hi")/geo or + s/\bret\b/bx lr/go or + s/\bbx\s+lr\b/.word\t0xe12fff1e/go; # make it possible to compile with -march=armv4 + + print $_,"\n"; +} +close STDOUT; # enforce flush diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-mont.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-mont.S new file mode 100644 index 00000000..fecae15e --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-mont.S @@ -0,0 +1,579 @@ +#include "arm_arch.h" + +.text +.code 32 + +#if __ARM_ARCH__>=7 +.align 5 +.LOPENSSL_armcap: +.word OPENSSL_armcap_P-bn_mul_mont +#endif + +.global bn_mul_mont +.type bn_mul_mont,%function + +.align 5 +bn_mul_mont: + ldr ip,[sp,#4] @ load num + stmdb sp!,{r0,r2} @ sp points at argument block +#if __ARM_ARCH__>=7 + tst ip,#7 + bne .Lialu + adr r0,bn_mul_mont + ldr r2,.LOPENSSL_armcap + ldr r0,[r0,r2] + tst r0,#1 @ NEON available? + ldmia sp, {r0,r2} + beq .Lialu + add sp,sp,#8 + b bn_mul8x_mont_neon +.align 4 +.Lialu: +#endif + cmp ip,#2 + mov r0,ip @ load num + movlt r0,#0 + addlt sp,sp,#2*4 + blt .Labrt + + stmdb sp!,{r4-r12,lr} @ save 10 registers + + mov r0,r0,lsl#2 @ rescale r0 for byte count + sub sp,sp,r0 @ alloca(4*num) + sub sp,sp,#4 @ +extra dword + sub r0,r0,#4 @ "num=num-1" + add r4,r2,r0 @ &bp[num-1] + + add r0,sp,r0 @ r0 to point at &tp[num-1] + ldr r8,[r0,#14*4] @ &n0 + ldr r2,[r2] @ bp[0] + ldr r5,[r1],#4 @ ap[0],ap++ + ldr r6,[r3],#4 @ np[0],np++ + ldr r8,[r8] @ *n0 + str r4,[r0,#15*4] @ save &bp[num] + + umull r10,r11,r5,r2 @ ap[0]*bp[0] + str r8,[r0,#14*4] @ save n0 value + mul r8,r10,r8 @ "tp[0]"*n0 + mov r12,#0 + umlal r10,r12,r6,r8 @ np[0]*n0+"t[0]" + mov r4,sp + +.L1st: + ldr r5,[r1],#4 @ ap[j],ap++ + mov r10,r11 + ldr r6,[r3],#4 @ np[j],np++ + mov r11,#0 + umlal r10,r11,r5,r2 @ ap[j]*bp[0] + mov r14,#0 + umlal r12,r14,r6,r8 @ np[j]*n0 + adds r12,r12,r10 + str r12,[r4],#4 @ tp[j-1]=,tp++ + adc r12,r14,#0 + cmp r4,r0 + bne .L1st + + adds r12,r12,r11 + ldr r4,[r0,#13*4] @ restore bp + mov r14,#0 + ldr r8,[r0,#14*4] @ restore n0 + adc r14,r14,#0 + str r12,[r0] @ tp[num-1]= + str r14,[r0,#4] @ tp[num]= + +.Louter: + sub r7,r0,sp @ "original" r0-1 value + sub r1,r1,r7 @ "rewind" ap to &ap[1] + ldr r2,[r4,#4]! @ *(++bp) + sub r3,r3,r7 @ "rewind" np to &np[1] + ldr r5,[r1,#-4] @ ap[0] + ldr r10,[sp] @ tp[0] + ldr r6,[r3,#-4] @ np[0] + ldr r7,[sp,#4] @ tp[1] + + mov r11,#0 + umlal r10,r11,r5,r2 @ ap[0]*bp[i]+tp[0] + str r4,[r0,#13*4] @ save bp + mul r8,r10,r8 + mov r12,#0 + umlal r10,r12,r6,r8 @ np[0]*n0+"tp[0]" + mov r4,sp + +.Linner: + ldr r5,[r1],#4 @ ap[j],ap++ + adds r10,r11,r7 @ +=tp[j] + ldr r6,[r3],#4 @ np[j],np++ + mov r11,#0 + umlal r10,r11,r5,r2 @ ap[j]*bp[i] + mov r14,#0 + umlal r12,r14,r6,r8 @ np[j]*n0 + adc r11,r11,#0 + ldr r7,[r4,#8] @ tp[j+1] + adds r12,r12,r10 + str r12,[r4],#4 @ tp[j-1]=,tp++ + adc r12,r14,#0 + cmp r4,r0 + bne .Linner + + adds r12,r12,r11 + mov r14,#0 + ldr r4,[r0,#13*4] @ restore bp + adc r14,r14,#0 + ldr r8,[r0,#14*4] @ restore n0 + adds r12,r12,r7 + ldr r7,[r0,#15*4] @ restore &bp[num] + adc r14,r14,#0 + str r12,[r0] @ tp[num-1]= + str r14,[r0,#4] @ tp[num]= + + cmp r4,r7 + bne .Louter + + ldr r2,[r0,#12*4] @ pull rp + add r0,r0,#4 @ r0 to point at &tp[num] + sub r5,r0,sp @ "original" num value + mov r4,sp @ "rewind" r4 + mov r1,r4 @ "borrow" r1 + sub r3,r3,r5 @ "rewind" r3 to &np[0] + + subs r7,r7,r7 @ "clear" carry flag +.Lsub: ldr r7,[r4],#4 + ldr r6,[r3],#4 + sbcs r7,r7,r6 @ tp[j]-np[j] + str r7,[r2],#4 @ rp[j]= + teq r4,r0 @ preserve carry + bne .Lsub + sbcs r14,r14,#0 @ upmost carry + mov r4,sp @ "rewind" r4 + sub r2,r2,r5 @ "rewind" r2 + + and r1,r4,r14 + bic r3,r2,r14 + orr r1,r1,r3 @ ap=borrow?tp:rp + +.Lcopy: ldr r7,[r1],#4 @ copy or in-place refresh + str sp,[r4],#4 @ zap tp + str r7,[r2],#4 + cmp r4,r0 + bne .Lcopy + + add sp,r0,#4 @ skip over tp[num+1] + ldmia sp!,{r4-r12,lr} @ restore registers + add sp,sp,#2*4 @ skip over {r0,r2} + mov r0,#1 +.Labrt: +#if __ARM_ARCH__>=5 + bx lr @ .word 0xe12fff1e +#else + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet + .word 0xe12fff1e @ interoperable with Thumb ISA:-) +#endif +.size bn_mul_mont,.-bn_mul_mont +#if __ARM_ARCH__>=7 +.fpu neon + +.type bn_mul8x_mont_neon,%function +.align 5 +bn_mul8x_mont_neon: + mov ip,sp + stmdb sp!,{r4-r11} + vstmdb sp!,{d8-d15} @ ABI specification says so + ldmia ip,{r4-r5} @ load rest of parameter block + + sub r7,sp,#16 + vld1.32 {d28[0]}, [r2,:32]! + sub r7,r7,r5,lsl#4 + vld1.32 {d0-d3}, [r1]! @ can't specify :32 :-( + and r7,r7,#-64 + vld1.32 {d30[0]}, [r4,:32] + mov sp,r7 @ alloca + veor d8,d8,d8 + subs r8,r5,#8 + vzip.16 d28,d8 + + vmull.u32 q6,d28,d0[0] + vmull.u32 q7,d28,d0[1] + vmull.u32 q8,d28,d1[0] + vshl.i64 d10,d13,#16 + vmull.u32 q9,d28,d1[1] + + vadd.u64 d10,d10,d12 + veor d8,d8,d8 + vmul.u32 d29,d10,d30 + + vmull.u32 q10,d28,d2[0] + vld1.32 {d4-d7}, [r3]! + vmull.u32 q11,d28,d2[1] + vmull.u32 q12,d28,d3[0] + vzip.16 d29,d8 + vmull.u32 q13,d28,d3[1] + + bne .LNEON_1st + + @ special case for num=8, everything is in register bank... + + vmlal.u32 q6,d29,d4[0] + sub r9,r5,#1 + vmlal.u32 q7,d29,d4[1] + vmlal.u32 q8,d29,d5[0] + vmlal.u32 q9,d29,d5[1] + + vmlal.u32 q10,d29,d6[0] + vmov q5,q6 + vmlal.u32 q11,d29,d6[1] + vmov q6,q7 + vmlal.u32 q12,d29,d7[0] + vmov q7,q8 + vmlal.u32 q13,d29,d7[1] + vmov q8,q9 + vmov q9,q10 + vshr.u64 d10,d10,#16 + vmov q10,q11 + vmov q11,q12 + vadd.u64 d10,d10,d11 + vmov q12,q13 + veor q13,q13 + vshr.u64 d10,d10,#16 + + b .LNEON_outer8 + +.align 4 +.LNEON_outer8: + vld1.32 {d28[0]}, [r2,:32]! + veor d8,d8,d8 + vzip.16 d28,d8 + vadd.u64 d12,d12,d10 + + vmlal.u32 q6,d28,d0[0] + vmlal.u32 q7,d28,d0[1] + vmlal.u32 q8,d28,d1[0] + vshl.i64 d10,d13,#16 + vmlal.u32 q9,d28,d1[1] + + vadd.u64 d10,d10,d12 + veor d8,d8,d8 + subs r9,r9,#1 + vmul.u32 d29,d10,d30 + + vmlal.u32 q10,d28,d2[0] + vmlal.u32 q11,d28,d2[1] + vmlal.u32 q12,d28,d3[0] + vzip.16 d29,d8 + vmlal.u32 q13,d28,d3[1] + + vmlal.u32 q6,d29,d4[0] + vmlal.u32 q7,d29,d4[1] + vmlal.u32 q8,d29,d5[0] + vmlal.u32 q9,d29,d5[1] + + vmlal.u32 q10,d29,d6[0] + vmov q5,q6 + vmlal.u32 q11,d29,d6[1] + vmov q6,q7 + vmlal.u32 q12,d29,d7[0] + vmov q7,q8 + vmlal.u32 q13,d29,d7[1] + vmov q8,q9 + vmov q9,q10 + vshr.u64 d10,d10,#16 + vmov q10,q11 + vmov q11,q12 + vadd.u64 d10,d10,d11 + vmov q12,q13 + veor q13,q13 + vshr.u64 d10,d10,#16 + + bne .LNEON_outer8 + + vadd.u64 d12,d12,d10 + mov r7,sp + vshr.u64 d10,d12,#16 + mov r8,r5 + vadd.u64 d13,d13,d10 + add r6,sp,#16 + vshr.u64 d10,d13,#16 + vzip.16 d12,d13 + + b .LNEON_tail2 + +.align 4 +.LNEON_1st: + vmlal.u32 q6,d29,d4[0] + vld1.32 {d0-d3}, [r1]! + vmlal.u32 q7,d29,d4[1] + subs r8,r8,#8 + vmlal.u32 q8,d29,d5[0] + vmlal.u32 q9,d29,d5[1] + + vmlal.u32 q10,d29,d6[0] + vld1.32 {d4-d5}, [r3]! + vmlal.u32 q11,d29,d6[1] + vst1.64 {q6-q7}, [r7,:256]! + vmlal.u32 q12,d29,d7[0] + vmlal.u32 q13,d29,d7[1] + vst1.64 {q8-q9}, [r7,:256]! + + vmull.u32 q6,d28,d0[0] + vld1.32 {d6-d7}, [r3]! + vmull.u32 q7,d28,d0[1] + vst1.64 {q10-q11}, [r7,:256]! + vmull.u32 q8,d28,d1[0] + vmull.u32 q9,d28,d1[1] + vst1.64 {q12-q13}, [r7,:256]! + + vmull.u32 q10,d28,d2[0] + vmull.u32 q11,d28,d2[1] + vmull.u32 q12,d28,d3[0] + vmull.u32 q13,d28,d3[1] + + bne .LNEON_1st + + vmlal.u32 q6,d29,d4[0] + add r6,sp,#16 + vmlal.u32 q7,d29,d4[1] + sub r1,r1,r5,lsl#2 @ rewind r1 + vmlal.u32 q8,d29,d5[0] + vld1.64 {q5}, [sp,:128] + vmlal.u32 q9,d29,d5[1] + sub r9,r5,#1 + + vmlal.u32 q10,d29,d6[0] + vst1.64 {q6-q7}, [r7,:256]! + vmlal.u32 q11,d29,d6[1] + vshr.u64 d10,d10,#16 + vld1.64 {q6}, [r6, :128]! + vmlal.u32 q12,d29,d7[0] + vst1.64 {q8-q9}, [r7,:256]! + vmlal.u32 q13,d29,d7[1] + + vst1.64 {q10-q11}, [r7,:256]! + vadd.u64 d10,d10,d11 + veor q4,q4,q4 + vst1.64 {q12-q13}, [r7,:256]! + vld1.64 {q7-q8}, [r6, :256]! + vst1.64 {q4}, [r7,:128] + vshr.u64 d10,d10,#16 + + b .LNEON_outer + +.align 4 +.LNEON_outer: + vld1.32 {d28[0]}, [r2,:32]! + sub r3,r3,r5,lsl#2 @ rewind r3 + vld1.32 {d0-d3}, [r1]! + veor d8,d8,d8 + mov r7,sp + vzip.16 d28,d8 + sub r8,r5,#8 + vadd.u64 d12,d12,d10 + + vmlal.u32 q6,d28,d0[0] + vld1.64 {q9-q10},[r6,:256]! + vmlal.u32 q7,d28,d0[1] + vmlal.u32 q8,d28,d1[0] + vld1.64 {q11-q12},[r6,:256]! + vmlal.u32 q9,d28,d1[1] + + vshl.i64 d10,d13,#16 + veor d8,d8,d8 + vadd.u64 d10,d10,d12 + vld1.64 {q13},[r6,:128]! + vmul.u32 d29,d10,d30 + + vmlal.u32 q10,d28,d2[0] + vld1.32 {d4-d7}, [r3]! + vmlal.u32 q11,d28,d2[1] + vmlal.u32 q12,d28,d3[0] + vzip.16 d29,d8 + vmlal.u32 q13,d28,d3[1] + +.LNEON_inner: + vmlal.u32 q6,d29,d4[0] + vld1.32 {d0-d3}, [r1]! + vmlal.u32 q7,d29,d4[1] + subs r8,r8,#8 + vmlal.u32 q8,d29,d5[0] + vmlal.u32 q9,d29,d5[1] + vst1.64 {q6-q7}, [r7,:256]! + + vmlal.u32 q10,d29,d6[0] + vld1.64 {q6}, [r6, :128]! + vmlal.u32 q11,d29,d6[1] + vst1.64 {q8-q9}, [r7,:256]! + vmlal.u32 q12,d29,d7[0] + vld1.64 {q7-q8}, [r6, :256]! + vmlal.u32 q13,d29,d7[1] + vst1.64 {q10-q11}, [r7,:256]! + + vmlal.u32 q6,d28,d0[0] + vld1.64 {q9-q10}, [r6, :256]! + vmlal.u32 q7,d28,d0[1] + vst1.64 {q12-q13}, [r7,:256]! + vmlal.u32 q8,d28,d1[0] + vld1.64 {q11-q12}, [r6, :256]! + vmlal.u32 q9,d28,d1[1] + vld1.32 {d4-d7}, [r3]! + + vmlal.u32 q10,d28,d2[0] + vld1.64 {q13}, [r6, :128]! + vmlal.u32 q11,d28,d2[1] + vmlal.u32 q12,d28,d3[0] + vmlal.u32 q13,d28,d3[1] + + bne .LNEON_inner + + vmlal.u32 q6,d29,d4[0] + add r6,sp,#16 + vmlal.u32 q7,d29,d4[1] + sub r1,r1,r5,lsl#2 @ rewind r1 + vmlal.u32 q8,d29,d5[0] + vld1.64 {q5}, [sp,:128] + vmlal.u32 q9,d29,d5[1] + subs r9,r9,#1 + + vmlal.u32 q10,d29,d6[0] + vst1.64 {q6-q7}, [r7,:256]! + vmlal.u32 q11,d29,d6[1] + vld1.64 {q6}, [r6, :128]! + vshr.u64 d10,d10,#16 + vst1.64 {q8-q9}, [r7,:256]! + vmlal.u32 q12,d29,d7[0] + vld1.64 {q7-q8}, [r6, :256]! + vmlal.u32 q13,d29,d7[1] + + vst1.64 {q10-q11}, [r7,:256]! + vadd.u64 d10,d10,d11 + vst1.64 {q12-q13}, [r7,:256]! + vshr.u64 d10,d10,#16 + + bne .LNEON_outer + + mov r7,sp + mov r8,r5 + +.LNEON_tail: + vadd.u64 d12,d12,d10 + vld1.64 {q9-q10}, [r6, :256]! + vshr.u64 d10,d12,#16 + vadd.u64 d13,d13,d10 + vld1.64 {q11-q12}, [r6, :256]! + vshr.u64 d10,d13,#16 + vld1.64 {q13}, [r6, :128]! + vzip.16 d12,d13 + +.LNEON_tail2: + vadd.u64 d14,d14,d10 + vst1.32 {d12[0]}, [r7, :32]! + vshr.u64 d10,d14,#16 + vadd.u64 d15,d15,d10 + vshr.u64 d10,d15,#16 + vzip.16 d14,d15 + + vadd.u64 d16,d16,d10 + vst1.32 {d14[0]}, [r7, :32]! + vshr.u64 d10,d16,#16 + vadd.u64 d17,d17,d10 + vshr.u64 d10,d17,#16 + vzip.16 d16,d17 + + vadd.u64 d18,d18,d10 + vst1.32 {d16[0]}, [r7, :32]! + vshr.u64 d10,d18,#16 + vadd.u64 d19,d19,d10 + vshr.u64 d10,d19,#16 + vzip.16 d18,d19 + + vadd.u64 d20,d20,d10 + vst1.32 {d18[0]}, [r7, :32]! + vshr.u64 d10,d20,#16 + vadd.u64 d21,d21,d10 + vshr.u64 d10,d21,#16 + vzip.16 d20,d21 + + vadd.u64 d22,d22,d10 + vst1.32 {d20[0]}, [r7, :32]! + vshr.u64 d10,d22,#16 + vadd.u64 d23,d23,d10 + vshr.u64 d10,d23,#16 + vzip.16 d22,d23 + + vadd.u64 d24,d24,d10 + vst1.32 {d22[0]}, [r7, :32]! + vshr.u64 d10,d24,#16 + vadd.u64 d25,d25,d10 + vld1.64 {q6}, [r6, :128]! + vshr.u64 d10,d25,#16 + vzip.16 d24,d25 + + vadd.u64 d26,d26,d10 + vst1.32 {d24[0]}, [r7, :32]! + vshr.u64 d10,d26,#16 + vadd.u64 d27,d27,d10 + vld1.64 {q7-q8}, [r6, :256]! + vshr.u64 d10,d27,#16 + vzip.16 d26,d27 + subs r8,r8,#8 + vst1.32 {d26[0]}, [r7, :32]! + + bne .LNEON_tail + + vst1.32 {d10[0]}, [r7, :32] @ top-most bit + sub r3,r3,r5,lsl#2 @ rewind r3 + subs r1,sp,#0 @ clear carry flag + add r2,sp,r5,lsl#2 + +.LNEON_sub: + ldmia r1!, {r4-r7} + ldmia r3!, {r8-r11} + sbcs r8, r4,r8 + sbcs r9, r5,r9 + sbcs r10,r6,r10 + sbcs r11,r7,r11 + teq r1,r2 @ preserves carry + stmia r0!, {r8-r11} + bne .LNEON_sub + + ldr r10, [r1] @ load top-most bit + veor q0,q0,q0 + sub r11,r2,sp @ this is num*4 + veor q1,q1,q1 + mov r1,sp + sub r0,r0,r11 @ rewind r0 + mov r3,r2 @ second 3/4th of frame + sbcs r10,r10,#0 @ result is carry flag + +.LNEON_copy_n_zap: + ldmia r1!, {r4-r7} + ldmia r0, {r8-r11} + movcc r8, r4 + vst1.64 {q0-q1}, [r3,:256]! @ wipe + movcc r9, r5 + movcc r10,r6 + vst1.64 {q0-q1}, [r3,:256]! @ wipe + movcc r11,r7 + ldmia r1, {r4-r7} + stmia r0!, {r8-r11} + sub r1,r1,#16 + ldmia r0, {r8-r11} + movcc r8, r4 + vst1.64 {q0-q1}, [r1,:256]! @ wipe + movcc r9, r5 + movcc r10,r6 + vst1.64 {q0-q1}, [r3,:256]! @ wipe + movcc r11,r7 + teq r1,r2 @ preserves carry + stmia r0!, {r8-r11} + bne .LNEON_copy_n_zap + + sub sp,ip,#96 + vldmia sp!,{d8-d15} + ldmia sp!,{r4-r11} + bx lr @ .word 0xe12fff1e +.size bn_mul8x_mont_neon,.-bn_mul8x_mont_neon +#endif +.asciz "Montgomery multiplication for ARMv4/NEON, CRYPTOGAMS by " +.align 2 +#if __ARM_ARCH__>=7 +.comm OPENSSL_armcap_P,4,4 +#endif diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-mont.pl new file mode 100644 index 00000000..72bad8e3 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/armv4-mont.pl @@ -0,0 +1,675 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# January 2007. + +# Montgomery multiplication for ARMv4. +# +# Performance improvement naturally varies among CPU implementations +# and compilers. The code was observed to provide +65-35% improvement +# [depending on key length, less for longer keys] on ARM920T, and +# +115-80% on Intel IXP425. This is compared to pre-bn_mul_mont code +# base and compiler generated code with in-lined umull and even umlal +# instructions. The latter means that this code didn't really have an +# "advantage" of utilizing some "secret" instruction. +# +# The code is interoperable with Thumb ISA and is rather compact, less +# than 1/2KB. Windows CE port would be trivial, as it's exclusively +# about decorations, ABI and instruction syntax are identical. + +# November 2013 +# +# Add NEON code path, which handles lengths divisible by 8. RSA/DSA +# performance improvement on Cortex-A8 is ~45-100% depending on key +# length, more for longer keys. On Cortex-A15 the span is ~10-105%. +# On Snapdragon S4 improvement was measured to vary from ~70% to +# incredible ~380%, yes, 4.8x faster, for RSA4096 sign. But this is +# rather because original integer-only code seems to perform +# suboptimally on S4. Situation on Cortex-A9 is unfortunately +# different. It's being looked into, but the trouble is that +# performance for vectors longer than 256 bits is actually couple +# of percent worse than for integer-only code. The code is chosen +# for execution on all NEON-capable processors, because gain on +# others outweighs the marginal loss on Cortex-A9. + +while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +$num="r0"; # starts as num argument, but holds &tp[num-1] +$ap="r1"; +$bp="r2"; $bi="r2"; $rp="r2"; +$np="r3"; +$tp="r4"; +$aj="r5"; +$nj="r6"; +$tj="r7"; +$n0="r8"; +########### # r9 is reserved by ELF as platform specific, e.g. TLS pointer +$alo="r10"; # sl, gcc uses it to keep @GOT +$ahi="r11"; # fp +$nlo="r12"; # ip +########### # r13 is stack pointer +$nhi="r14"; # lr +########### # r15 is program counter + +#### argument block layout relative to &tp[num-1], a.k.a. $num +$_rp="$num,#12*4"; +# ap permanently resides in r1 +$_bp="$num,#13*4"; +# np permanently resides in r3 +$_n0="$num,#14*4"; +$_num="$num,#15*4"; $_bpend=$_num; + +$code=<<___; +#include "arm_arch.h" + +.text +.code 32 + +#if __ARM_ARCH__>=7 +.align 5 +.LOPENSSL_armcap: +.word OPENSSL_armcap_P-bn_mul_mont +#endif + +.global bn_mul_mont +.type bn_mul_mont,%function + +.align 5 +bn_mul_mont: + ldr ip,[sp,#4] @ load num + stmdb sp!,{r0,r2} @ sp points at argument block +#if __ARM_ARCH__>=7 + tst ip,#7 + bne .Lialu + adr r0,bn_mul_mont + ldr r2,.LOPENSSL_armcap + ldr r0,[r0,r2] + tst r0,#1 @ NEON available? + ldmia sp, {r0,r2} + beq .Lialu + add sp,sp,#8 + b bn_mul8x_mont_neon +.align 4 +.Lialu: +#endif + cmp ip,#2 + mov $num,ip @ load num + movlt r0,#0 + addlt sp,sp,#2*4 + blt .Labrt + + stmdb sp!,{r4-r12,lr} @ save 10 registers + + mov $num,$num,lsl#2 @ rescale $num for byte count + sub sp,sp,$num @ alloca(4*num) + sub sp,sp,#4 @ +extra dword + sub $num,$num,#4 @ "num=num-1" + add $tp,$bp,$num @ &bp[num-1] + + add $num,sp,$num @ $num to point at &tp[num-1] + ldr $n0,[$_n0] @ &n0 + ldr $bi,[$bp] @ bp[0] + ldr $aj,[$ap],#4 @ ap[0],ap++ + ldr $nj,[$np],#4 @ np[0],np++ + ldr $n0,[$n0] @ *n0 + str $tp,[$_bpend] @ save &bp[num] + + umull $alo,$ahi,$aj,$bi @ ap[0]*bp[0] + str $n0,[$_n0] @ save n0 value + mul $n0,$alo,$n0 @ "tp[0]"*n0 + mov $nlo,#0 + umlal $alo,$nlo,$nj,$n0 @ np[0]*n0+"t[0]" + mov $tp,sp + +.L1st: + ldr $aj,[$ap],#4 @ ap[j],ap++ + mov $alo,$ahi + ldr $nj,[$np],#4 @ np[j],np++ + mov $ahi,#0 + umlal $alo,$ahi,$aj,$bi @ ap[j]*bp[0] + mov $nhi,#0 + umlal $nlo,$nhi,$nj,$n0 @ np[j]*n0 + adds $nlo,$nlo,$alo + str $nlo,[$tp],#4 @ tp[j-1]=,tp++ + adc $nlo,$nhi,#0 + cmp $tp,$num + bne .L1st + + adds $nlo,$nlo,$ahi + ldr $tp,[$_bp] @ restore bp + mov $nhi,#0 + ldr $n0,[$_n0] @ restore n0 + adc $nhi,$nhi,#0 + str $nlo,[$num] @ tp[num-1]= + str $nhi,[$num,#4] @ tp[num]= + +.Louter: + sub $tj,$num,sp @ "original" $num-1 value + sub $ap,$ap,$tj @ "rewind" ap to &ap[1] + ldr $bi,[$tp,#4]! @ *(++bp) + sub $np,$np,$tj @ "rewind" np to &np[1] + ldr $aj,[$ap,#-4] @ ap[0] + ldr $alo,[sp] @ tp[0] + ldr $nj,[$np,#-4] @ np[0] + ldr $tj,[sp,#4] @ tp[1] + + mov $ahi,#0 + umlal $alo,$ahi,$aj,$bi @ ap[0]*bp[i]+tp[0] + str $tp,[$_bp] @ save bp + mul $n0,$alo,$n0 + mov $nlo,#0 + umlal $alo,$nlo,$nj,$n0 @ np[0]*n0+"tp[0]" + mov $tp,sp + +.Linner: + ldr $aj,[$ap],#4 @ ap[j],ap++ + adds $alo,$ahi,$tj @ +=tp[j] + ldr $nj,[$np],#4 @ np[j],np++ + mov $ahi,#0 + umlal $alo,$ahi,$aj,$bi @ ap[j]*bp[i] + mov $nhi,#0 + umlal $nlo,$nhi,$nj,$n0 @ np[j]*n0 + adc $ahi,$ahi,#0 + ldr $tj,[$tp,#8] @ tp[j+1] + adds $nlo,$nlo,$alo + str $nlo,[$tp],#4 @ tp[j-1]=,tp++ + adc $nlo,$nhi,#0 + cmp $tp,$num + bne .Linner + + adds $nlo,$nlo,$ahi + mov $nhi,#0 + ldr $tp,[$_bp] @ restore bp + adc $nhi,$nhi,#0 + ldr $n0,[$_n0] @ restore n0 + adds $nlo,$nlo,$tj + ldr $tj,[$_bpend] @ restore &bp[num] + adc $nhi,$nhi,#0 + str $nlo,[$num] @ tp[num-1]= + str $nhi,[$num,#4] @ tp[num]= + + cmp $tp,$tj + bne .Louter + + ldr $rp,[$_rp] @ pull rp + add $num,$num,#4 @ $num to point at &tp[num] + sub $aj,$num,sp @ "original" num value + mov $tp,sp @ "rewind" $tp + mov $ap,$tp @ "borrow" $ap + sub $np,$np,$aj @ "rewind" $np to &np[0] + + subs $tj,$tj,$tj @ "clear" carry flag +.Lsub: ldr $tj,[$tp],#4 + ldr $nj,[$np],#4 + sbcs $tj,$tj,$nj @ tp[j]-np[j] + str $tj,[$rp],#4 @ rp[j]= + teq $tp,$num @ preserve carry + bne .Lsub + sbcs $nhi,$nhi,#0 @ upmost carry + mov $tp,sp @ "rewind" $tp + sub $rp,$rp,$aj @ "rewind" $rp + + and $ap,$tp,$nhi + bic $np,$rp,$nhi + orr $ap,$ap,$np @ ap=borrow?tp:rp + +.Lcopy: ldr $tj,[$ap],#4 @ copy or in-place refresh + str sp,[$tp],#4 @ zap tp + str $tj,[$rp],#4 + cmp $tp,$num + bne .Lcopy + + add sp,$num,#4 @ skip over tp[num+1] + ldmia sp!,{r4-r12,lr} @ restore registers + add sp,sp,#2*4 @ skip over {r0,r2} + mov r0,#1 +.Labrt: +#if __ARM_ARCH__>=5 + ret @ bx lr +#else + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet + bx lr @ interoperable with Thumb ISA:-) +#endif +.size bn_mul_mont,.-bn_mul_mont +___ +{ +sub Dlo() { shift=~m|q([1]?[0-9])|?"d".($1*2):""; } +sub Dhi() { shift=~m|q([1]?[0-9])|?"d".($1*2+1):""; } + +my ($A0,$A1,$A2,$A3)=map("d$_",(0..3)); +my ($N0,$N1,$N2,$N3)=map("d$_",(4..7)); +my ($Z,$Temp)=("q4","q5"); +my ($A0xB,$A1xB,$A2xB,$A3xB,$A4xB,$A5xB,$A6xB,$A7xB)=map("q$_",(6..13)); +my ($Bi,$Ni,$M0)=map("d$_",(28..31)); +my $zero=&Dlo($Z); +my $temp=&Dlo($Temp); + +my ($rptr,$aptr,$bptr,$nptr,$n0,$num)=map("r$_",(0..5)); +my ($tinptr,$toutptr,$inner,$outer)=map("r$_",(6..9)); + +$code.=<<___; +#if __ARM_ARCH__>=7 +.fpu neon + +.type bn_mul8x_mont_neon,%function +.align 5 +bn_mul8x_mont_neon: + mov ip,sp + stmdb sp!,{r4-r11} + vstmdb sp!,{d8-d15} @ ABI specification says so + ldmia ip,{r4-r5} @ load rest of parameter block + + sub $toutptr,sp,#16 + vld1.32 {${Bi}[0]}, [$bptr,:32]! + sub $toutptr,$toutptr,$num,lsl#4 + vld1.32 {$A0-$A3}, [$aptr]! @ can't specify :32 :-( + and $toutptr,$toutptr,#-64 + vld1.32 {${M0}[0]}, [$n0,:32] + mov sp,$toutptr @ alloca + veor $zero,$zero,$zero + subs $inner,$num,#8 + vzip.16 $Bi,$zero + + vmull.u32 $A0xB,$Bi,${A0}[0] + vmull.u32 $A1xB,$Bi,${A0}[1] + vmull.u32 $A2xB,$Bi,${A1}[0] + vshl.i64 $temp,`&Dhi("$A0xB")`,#16 + vmull.u32 $A3xB,$Bi,${A1}[1] + + vadd.u64 $temp,$temp,`&Dlo("$A0xB")` + veor $zero,$zero,$zero + vmul.u32 $Ni,$temp,$M0 + + vmull.u32 $A4xB,$Bi,${A2}[0] + vld1.32 {$N0-$N3}, [$nptr]! + vmull.u32 $A5xB,$Bi,${A2}[1] + vmull.u32 $A6xB,$Bi,${A3}[0] + vzip.16 $Ni,$zero + vmull.u32 $A7xB,$Bi,${A3}[1] + + bne .LNEON_1st + + @ special case for num=8, everything is in register bank... + + vmlal.u32 $A0xB,$Ni,${N0}[0] + sub $outer,$num,#1 + vmlal.u32 $A1xB,$Ni,${N0}[1] + vmlal.u32 $A2xB,$Ni,${N1}[0] + vmlal.u32 $A3xB,$Ni,${N1}[1] + + vmlal.u32 $A4xB,$Ni,${N2}[0] + vmov $Temp,$A0xB + vmlal.u32 $A5xB,$Ni,${N2}[1] + vmov $A0xB,$A1xB + vmlal.u32 $A6xB,$Ni,${N3}[0] + vmov $A1xB,$A2xB + vmlal.u32 $A7xB,$Ni,${N3}[1] + vmov $A2xB,$A3xB + vmov $A3xB,$A4xB + vshr.u64 $temp,$temp,#16 + vmov $A4xB,$A5xB + vmov $A5xB,$A6xB + vadd.u64 $temp,$temp,`&Dhi("$Temp")` + vmov $A6xB,$A7xB + veor $A7xB,$A7xB + vshr.u64 $temp,$temp,#16 + + b .LNEON_outer8 + +.align 4 +.LNEON_outer8: + vld1.32 {${Bi}[0]}, [$bptr,:32]! + veor $zero,$zero,$zero + vzip.16 $Bi,$zero + vadd.u64 `&Dlo("$A0xB")`,`&Dlo("$A0xB")`,$temp + + vmlal.u32 $A0xB,$Bi,${A0}[0] + vmlal.u32 $A1xB,$Bi,${A0}[1] + vmlal.u32 $A2xB,$Bi,${A1}[0] + vshl.i64 $temp,`&Dhi("$A0xB")`,#16 + vmlal.u32 $A3xB,$Bi,${A1}[1] + + vadd.u64 $temp,$temp,`&Dlo("$A0xB")` + veor $zero,$zero,$zero + subs $outer,$outer,#1 + vmul.u32 $Ni,$temp,$M0 + + vmlal.u32 $A4xB,$Bi,${A2}[0] + vmlal.u32 $A5xB,$Bi,${A2}[1] + vmlal.u32 $A6xB,$Bi,${A3}[0] + vzip.16 $Ni,$zero + vmlal.u32 $A7xB,$Bi,${A3}[1] + + vmlal.u32 $A0xB,$Ni,${N0}[0] + vmlal.u32 $A1xB,$Ni,${N0}[1] + vmlal.u32 $A2xB,$Ni,${N1}[0] + vmlal.u32 $A3xB,$Ni,${N1}[1] + + vmlal.u32 $A4xB,$Ni,${N2}[0] + vmov $Temp,$A0xB + vmlal.u32 $A5xB,$Ni,${N2}[1] + vmov $A0xB,$A1xB + vmlal.u32 $A6xB,$Ni,${N3}[0] + vmov $A1xB,$A2xB + vmlal.u32 $A7xB,$Ni,${N3}[1] + vmov $A2xB,$A3xB + vmov $A3xB,$A4xB + vshr.u64 $temp,$temp,#16 + vmov $A4xB,$A5xB + vmov $A5xB,$A6xB + vadd.u64 $temp,$temp,`&Dhi("$Temp")` + vmov $A6xB,$A7xB + veor $A7xB,$A7xB + vshr.u64 $temp,$temp,#16 + + bne .LNEON_outer8 + + vadd.u64 `&Dlo("$A0xB")`,`&Dlo("$A0xB")`,$temp + mov $toutptr,sp + vshr.u64 $temp,`&Dlo("$A0xB")`,#16 + mov $inner,$num + vadd.u64 `&Dhi("$A0xB")`,`&Dhi("$A0xB")`,$temp + add $tinptr,sp,#16 + vshr.u64 $temp,`&Dhi("$A0xB")`,#16 + vzip.16 `&Dlo("$A0xB")`,`&Dhi("$A0xB")` + + b .LNEON_tail2 + +.align 4 +.LNEON_1st: + vmlal.u32 $A0xB,$Ni,${N0}[0] + vld1.32 {$A0-$A3}, [$aptr]! + vmlal.u32 $A1xB,$Ni,${N0}[1] + subs $inner,$inner,#8 + vmlal.u32 $A2xB,$Ni,${N1}[0] + vmlal.u32 $A3xB,$Ni,${N1}[1] + + vmlal.u32 $A4xB,$Ni,${N2}[0] + vld1.32 {$N0-$N1}, [$nptr]! + vmlal.u32 $A5xB,$Ni,${N2}[1] + vst1.64 {$A0xB-$A1xB}, [$toutptr,:256]! + vmlal.u32 $A6xB,$Ni,${N3}[0] + vmlal.u32 $A7xB,$Ni,${N3}[1] + vst1.64 {$A2xB-$A3xB}, [$toutptr,:256]! + + vmull.u32 $A0xB,$Bi,${A0}[0] + vld1.32 {$N2-$N3}, [$nptr]! + vmull.u32 $A1xB,$Bi,${A0}[1] + vst1.64 {$A4xB-$A5xB}, [$toutptr,:256]! + vmull.u32 $A2xB,$Bi,${A1}[0] + vmull.u32 $A3xB,$Bi,${A1}[1] + vst1.64 {$A6xB-$A7xB}, [$toutptr,:256]! + + vmull.u32 $A4xB,$Bi,${A2}[0] + vmull.u32 $A5xB,$Bi,${A2}[1] + vmull.u32 $A6xB,$Bi,${A3}[0] + vmull.u32 $A7xB,$Bi,${A3}[1] + + bne .LNEON_1st + + vmlal.u32 $A0xB,$Ni,${N0}[0] + add $tinptr,sp,#16 + vmlal.u32 $A1xB,$Ni,${N0}[1] + sub $aptr,$aptr,$num,lsl#2 @ rewind $aptr + vmlal.u32 $A2xB,$Ni,${N1}[0] + vld1.64 {$Temp}, [sp,:128] + vmlal.u32 $A3xB,$Ni,${N1}[1] + sub $outer,$num,#1 + + vmlal.u32 $A4xB,$Ni,${N2}[0] + vst1.64 {$A0xB-$A1xB}, [$toutptr,:256]! + vmlal.u32 $A5xB,$Ni,${N2}[1] + vshr.u64 $temp,$temp,#16 + vld1.64 {$A0xB}, [$tinptr, :128]! + vmlal.u32 $A6xB,$Ni,${N3}[0] + vst1.64 {$A2xB-$A3xB}, [$toutptr,:256]! + vmlal.u32 $A7xB,$Ni,${N3}[1] + + vst1.64 {$A4xB-$A5xB}, [$toutptr,:256]! + vadd.u64 $temp,$temp,`&Dhi("$Temp")` + veor $Z,$Z,$Z + vst1.64 {$A6xB-$A7xB}, [$toutptr,:256]! + vld1.64 {$A1xB-$A2xB}, [$tinptr, :256]! + vst1.64 {$Z}, [$toutptr,:128] + vshr.u64 $temp,$temp,#16 + + b .LNEON_outer + +.align 4 +.LNEON_outer: + vld1.32 {${Bi}[0]}, [$bptr,:32]! + sub $nptr,$nptr,$num,lsl#2 @ rewind $nptr + vld1.32 {$A0-$A3}, [$aptr]! + veor $zero,$zero,$zero + mov $toutptr,sp + vzip.16 $Bi,$zero + sub $inner,$num,#8 + vadd.u64 `&Dlo("$A0xB")`,`&Dlo("$A0xB")`,$temp + + vmlal.u32 $A0xB,$Bi,${A0}[0] + vld1.64 {$A3xB-$A4xB},[$tinptr,:256]! + vmlal.u32 $A1xB,$Bi,${A0}[1] + vmlal.u32 $A2xB,$Bi,${A1}[0] + vld1.64 {$A5xB-$A6xB},[$tinptr,:256]! + vmlal.u32 $A3xB,$Bi,${A1}[1] + + vshl.i64 $temp,`&Dhi("$A0xB")`,#16 + veor $zero,$zero,$zero + vadd.u64 $temp,$temp,`&Dlo("$A0xB")` + vld1.64 {$A7xB},[$tinptr,:128]! + vmul.u32 $Ni,$temp,$M0 + + vmlal.u32 $A4xB,$Bi,${A2}[0] + vld1.32 {$N0-$N3}, [$nptr]! + vmlal.u32 $A5xB,$Bi,${A2}[1] + vmlal.u32 $A6xB,$Bi,${A3}[0] + vzip.16 $Ni,$zero + vmlal.u32 $A7xB,$Bi,${A3}[1] + +.LNEON_inner: + vmlal.u32 $A0xB,$Ni,${N0}[0] + vld1.32 {$A0-$A3}, [$aptr]! + vmlal.u32 $A1xB,$Ni,${N0}[1] + subs $inner,$inner,#8 + vmlal.u32 $A2xB,$Ni,${N1}[0] + vmlal.u32 $A3xB,$Ni,${N1}[1] + vst1.64 {$A0xB-$A1xB}, [$toutptr,:256]! + + vmlal.u32 $A4xB,$Ni,${N2}[0] + vld1.64 {$A0xB}, [$tinptr, :128]! + vmlal.u32 $A5xB,$Ni,${N2}[1] + vst1.64 {$A2xB-$A3xB}, [$toutptr,:256]! + vmlal.u32 $A6xB,$Ni,${N3}[0] + vld1.64 {$A1xB-$A2xB}, [$tinptr, :256]! + vmlal.u32 $A7xB,$Ni,${N3}[1] + vst1.64 {$A4xB-$A5xB}, [$toutptr,:256]! + + vmlal.u32 $A0xB,$Bi,${A0}[0] + vld1.64 {$A3xB-$A4xB}, [$tinptr, :256]! + vmlal.u32 $A1xB,$Bi,${A0}[1] + vst1.64 {$A6xB-$A7xB}, [$toutptr,:256]! + vmlal.u32 $A2xB,$Bi,${A1}[0] + vld1.64 {$A5xB-$A6xB}, [$tinptr, :256]! + vmlal.u32 $A3xB,$Bi,${A1}[1] + vld1.32 {$N0-$N3}, [$nptr]! + + vmlal.u32 $A4xB,$Bi,${A2}[0] + vld1.64 {$A7xB}, [$tinptr, :128]! + vmlal.u32 $A5xB,$Bi,${A2}[1] + vmlal.u32 $A6xB,$Bi,${A3}[0] + vmlal.u32 $A7xB,$Bi,${A3}[1] + + bne .LNEON_inner + + vmlal.u32 $A0xB,$Ni,${N0}[0] + add $tinptr,sp,#16 + vmlal.u32 $A1xB,$Ni,${N0}[1] + sub $aptr,$aptr,$num,lsl#2 @ rewind $aptr + vmlal.u32 $A2xB,$Ni,${N1}[0] + vld1.64 {$Temp}, [sp,:128] + vmlal.u32 $A3xB,$Ni,${N1}[1] + subs $outer,$outer,#1 + + vmlal.u32 $A4xB,$Ni,${N2}[0] + vst1.64 {$A0xB-$A1xB}, [$toutptr,:256]! + vmlal.u32 $A5xB,$Ni,${N2}[1] + vld1.64 {$A0xB}, [$tinptr, :128]! + vshr.u64 $temp,$temp,#16 + vst1.64 {$A2xB-$A3xB}, [$toutptr,:256]! + vmlal.u32 $A6xB,$Ni,${N3}[0] + vld1.64 {$A1xB-$A2xB}, [$tinptr, :256]! + vmlal.u32 $A7xB,$Ni,${N3}[1] + + vst1.64 {$A4xB-$A5xB}, [$toutptr,:256]! + vadd.u64 $temp,$temp,`&Dhi("$Temp")` + vst1.64 {$A6xB-$A7xB}, [$toutptr,:256]! + vshr.u64 $temp,$temp,#16 + + bne .LNEON_outer + + mov $toutptr,sp + mov $inner,$num + +.LNEON_tail: + vadd.u64 `&Dlo("$A0xB")`,`&Dlo("$A0xB")`,$temp + vld1.64 {$A3xB-$A4xB}, [$tinptr, :256]! + vshr.u64 $temp,`&Dlo("$A0xB")`,#16 + vadd.u64 `&Dhi("$A0xB")`,`&Dhi("$A0xB")`,$temp + vld1.64 {$A5xB-$A6xB}, [$tinptr, :256]! + vshr.u64 $temp,`&Dhi("$A0xB")`,#16 + vld1.64 {$A7xB}, [$tinptr, :128]! + vzip.16 `&Dlo("$A0xB")`,`&Dhi("$A0xB")` + +.LNEON_tail2: + vadd.u64 `&Dlo("$A1xB")`,`&Dlo("$A1xB")`,$temp + vst1.32 {`&Dlo("$A0xB")`[0]}, [$toutptr, :32]! + vshr.u64 $temp,`&Dlo("$A1xB")`,#16 + vadd.u64 `&Dhi("$A1xB")`,`&Dhi("$A1xB")`,$temp + vshr.u64 $temp,`&Dhi("$A1xB")`,#16 + vzip.16 `&Dlo("$A1xB")`,`&Dhi("$A1xB")` + + vadd.u64 `&Dlo("$A2xB")`,`&Dlo("$A2xB")`,$temp + vst1.32 {`&Dlo("$A1xB")`[0]}, [$toutptr, :32]! + vshr.u64 $temp,`&Dlo("$A2xB")`,#16 + vadd.u64 `&Dhi("$A2xB")`,`&Dhi("$A2xB")`,$temp + vshr.u64 $temp,`&Dhi("$A2xB")`,#16 + vzip.16 `&Dlo("$A2xB")`,`&Dhi("$A2xB")` + + vadd.u64 `&Dlo("$A3xB")`,`&Dlo("$A3xB")`,$temp + vst1.32 {`&Dlo("$A2xB")`[0]}, [$toutptr, :32]! + vshr.u64 $temp,`&Dlo("$A3xB")`,#16 + vadd.u64 `&Dhi("$A3xB")`,`&Dhi("$A3xB")`,$temp + vshr.u64 $temp,`&Dhi("$A3xB")`,#16 + vzip.16 `&Dlo("$A3xB")`,`&Dhi("$A3xB")` + + vadd.u64 `&Dlo("$A4xB")`,`&Dlo("$A4xB")`,$temp + vst1.32 {`&Dlo("$A3xB")`[0]}, [$toutptr, :32]! + vshr.u64 $temp,`&Dlo("$A4xB")`,#16 + vadd.u64 `&Dhi("$A4xB")`,`&Dhi("$A4xB")`,$temp + vshr.u64 $temp,`&Dhi("$A4xB")`,#16 + vzip.16 `&Dlo("$A4xB")`,`&Dhi("$A4xB")` + + vadd.u64 `&Dlo("$A5xB")`,`&Dlo("$A5xB")`,$temp + vst1.32 {`&Dlo("$A4xB")`[0]}, [$toutptr, :32]! + vshr.u64 $temp,`&Dlo("$A5xB")`,#16 + vadd.u64 `&Dhi("$A5xB")`,`&Dhi("$A5xB")`,$temp + vshr.u64 $temp,`&Dhi("$A5xB")`,#16 + vzip.16 `&Dlo("$A5xB")`,`&Dhi("$A5xB")` + + vadd.u64 `&Dlo("$A6xB")`,`&Dlo("$A6xB")`,$temp + vst1.32 {`&Dlo("$A5xB")`[0]}, [$toutptr, :32]! + vshr.u64 $temp,`&Dlo("$A6xB")`,#16 + vadd.u64 `&Dhi("$A6xB")`,`&Dhi("$A6xB")`,$temp + vld1.64 {$A0xB}, [$tinptr, :128]! + vshr.u64 $temp,`&Dhi("$A6xB")`,#16 + vzip.16 `&Dlo("$A6xB")`,`&Dhi("$A6xB")` + + vadd.u64 `&Dlo("$A7xB")`,`&Dlo("$A7xB")`,$temp + vst1.32 {`&Dlo("$A6xB")`[0]}, [$toutptr, :32]! + vshr.u64 $temp,`&Dlo("$A7xB")`,#16 + vadd.u64 `&Dhi("$A7xB")`,`&Dhi("$A7xB")`,$temp + vld1.64 {$A1xB-$A2xB}, [$tinptr, :256]! + vshr.u64 $temp,`&Dhi("$A7xB")`,#16 + vzip.16 `&Dlo("$A7xB")`,`&Dhi("$A7xB")` + subs $inner,$inner,#8 + vst1.32 {`&Dlo("$A7xB")`[0]}, [$toutptr, :32]! + + bne .LNEON_tail + + vst1.32 {${temp}[0]}, [$toutptr, :32] @ top-most bit + sub $nptr,$nptr,$num,lsl#2 @ rewind $nptr + subs $aptr,sp,#0 @ clear carry flag + add $bptr,sp,$num,lsl#2 + +.LNEON_sub: + ldmia $aptr!, {r4-r7} + ldmia $nptr!, {r8-r11} + sbcs r8, r4,r8 + sbcs r9, r5,r9 + sbcs r10,r6,r10 + sbcs r11,r7,r11 + teq $aptr,$bptr @ preserves carry + stmia $rptr!, {r8-r11} + bne .LNEON_sub + + ldr r10, [$aptr] @ load top-most bit + veor q0,q0,q0 + sub r11,$bptr,sp @ this is num*4 + veor q1,q1,q1 + mov $aptr,sp + sub $rptr,$rptr,r11 @ rewind $rptr + mov $nptr,$bptr @ second 3/4th of frame + sbcs r10,r10,#0 @ result is carry flag + +.LNEON_copy_n_zap: + ldmia $aptr!, {r4-r7} + ldmia $rptr, {r8-r11} + movcc r8, r4 + vst1.64 {q0-q1}, [$nptr,:256]! @ wipe + movcc r9, r5 + movcc r10,r6 + vst1.64 {q0-q1}, [$nptr,:256]! @ wipe + movcc r11,r7 + ldmia $aptr, {r4-r7} + stmia $rptr!, {r8-r11} + sub $aptr,$aptr,#16 + ldmia $rptr, {r8-r11} + movcc r8, r4 + vst1.64 {q0-q1}, [$aptr,:256]! @ wipe + movcc r9, r5 + movcc r10,r6 + vst1.64 {q0-q1}, [$nptr,:256]! @ wipe + movcc r11,r7 + teq $aptr,$bptr @ preserves carry + stmia $rptr!, {r8-r11} + bne .LNEON_copy_n_zap + + sub sp,ip,#96 + vldmia sp!,{d8-d15} + ldmia sp!,{r4-r11} + ret @ bx lr +.size bn_mul8x_mont_neon,.-bn_mul8x_mont_neon +#endif +___ +} +$code.=<<___; +.asciz "Montgomery multiplication for ARMv4/NEON, CRYPTOGAMS by " +.align 2 +#if __ARM_ARCH__>=7 +.comm OPENSSL_armcap_P,4,4 +#endif +___ + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +$code =~ s/\bbx\s+lr\b/.word\t0xe12fff1e/gm; # make it possible to compile with -march=armv4 +$code =~ s/\bret\b/bx lr/gm; +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-586.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-586.S new file mode 100644 index 00000000..66695e26 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-586.S @@ -0,0 +1,1533 @@ +.file "crypto/bn/asm/bn-586.s" +.text +.globl bn_mul_add_words +.type bn_mul_add_words,@function +.align 16 +bn_mul_add_words: +.L_bn_mul_add_words_begin: + call .L000PIC_me_up +.L000PIC_me_up: + popl %eax + leal _GLOBAL_OFFSET_TABLE_+[.-.L000PIC_me_up](%eax),%eax + movl OPENSSL_ia32cap_P@GOT(%eax),%eax + btl $26,(%eax) + jnc .L001maw_non_sse2 + movl 4(%esp),%eax + movl 8(%esp),%edx + movl 12(%esp),%ecx + movd 16(%esp),%mm0 + pxor %mm1,%mm1 + jmp .L002maw_sse2_entry +.align 16 +.L003maw_sse2_unrolled: + movd (%eax),%mm3 + paddq %mm3,%mm1 + movd (%edx),%mm2 + pmuludq %mm0,%mm2 + movd 4(%edx),%mm4 + pmuludq %mm0,%mm4 + movd 8(%edx),%mm6 + pmuludq %mm0,%mm6 + movd 12(%edx),%mm7 + pmuludq %mm0,%mm7 + paddq %mm2,%mm1 + movd 4(%eax),%mm3 + paddq %mm4,%mm3 + movd 8(%eax),%mm5 + paddq %mm6,%mm5 + movd 12(%eax),%mm4 + paddq %mm4,%mm7 + movd %mm1,(%eax) + movd 16(%edx),%mm2 + pmuludq %mm0,%mm2 + psrlq $32,%mm1 + movd 20(%edx),%mm4 + pmuludq %mm0,%mm4 + paddq %mm3,%mm1 + movd 24(%edx),%mm6 + pmuludq %mm0,%mm6 + movd %mm1,4(%eax) + psrlq $32,%mm1 + movd 28(%edx),%mm3 + addl $32,%edx + pmuludq %mm0,%mm3 + paddq %mm5,%mm1 + movd 16(%eax),%mm5 + paddq %mm5,%mm2 + movd %mm1,8(%eax) + psrlq $32,%mm1 + paddq %mm7,%mm1 + movd 20(%eax),%mm5 + paddq %mm5,%mm4 + movd %mm1,12(%eax) + psrlq $32,%mm1 + paddq %mm2,%mm1 + movd 24(%eax),%mm5 + paddq %mm5,%mm6 + movd %mm1,16(%eax) + psrlq $32,%mm1 + paddq %mm4,%mm1 + movd 28(%eax),%mm5 + paddq %mm5,%mm3 + movd %mm1,20(%eax) + psrlq $32,%mm1 + paddq %mm6,%mm1 + movd %mm1,24(%eax) + psrlq $32,%mm1 + paddq %mm3,%mm1 + movd %mm1,28(%eax) + leal 32(%eax),%eax + psrlq $32,%mm1 + subl $8,%ecx + jz .L004maw_sse2_exit +.L002maw_sse2_entry: + testl $4294967288,%ecx + jnz .L003maw_sse2_unrolled +.align 4 +.L005maw_sse2_loop: + movd (%edx),%mm2 + movd (%eax),%mm3 + pmuludq %mm0,%mm2 + leal 4(%edx),%edx + paddq %mm3,%mm1 + paddq %mm2,%mm1 + movd %mm1,(%eax) + subl $1,%ecx + psrlq $32,%mm1 + leal 4(%eax),%eax + jnz .L005maw_sse2_loop +.L004maw_sse2_exit: + movd %mm1,%eax + emms + ret +.align 16 +.L001maw_non_sse2: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + + xorl %esi,%esi + movl 20(%esp),%edi + movl 28(%esp),%ecx + movl 24(%esp),%ebx + andl $4294967288,%ecx + movl 32(%esp),%ebp + pushl %ecx + jz .L006maw_finish +.align 16 +.L007maw_loop: + + movl (%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl (%edi),%eax + adcl $0,%edx + movl %eax,(%edi) + movl %edx,%esi + + movl 4(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 4(%edi),%eax + adcl $0,%edx + movl %eax,4(%edi) + movl %edx,%esi + + movl 8(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 8(%edi),%eax + adcl $0,%edx + movl %eax,8(%edi) + movl %edx,%esi + + movl 12(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 12(%edi),%eax + adcl $0,%edx + movl %eax,12(%edi) + movl %edx,%esi + + movl 16(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 16(%edi),%eax + adcl $0,%edx + movl %eax,16(%edi) + movl %edx,%esi + + movl 20(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 20(%edi),%eax + adcl $0,%edx + movl %eax,20(%edi) + movl %edx,%esi + + movl 24(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 24(%edi),%eax + adcl $0,%edx + movl %eax,24(%edi) + movl %edx,%esi + + movl 28(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 28(%edi),%eax + adcl $0,%edx + movl %eax,28(%edi) + movl %edx,%esi + + subl $8,%ecx + leal 32(%ebx),%ebx + leal 32(%edi),%edi + jnz .L007maw_loop +.L006maw_finish: + movl 32(%esp),%ecx + andl $7,%ecx + jnz .L008maw_finish2 + jmp .L009maw_end +.L008maw_finish2: + + movl (%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl (%edi),%eax + adcl $0,%edx + decl %ecx + movl %eax,(%edi) + movl %edx,%esi + jz .L009maw_end + + movl 4(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 4(%edi),%eax + adcl $0,%edx + decl %ecx + movl %eax,4(%edi) + movl %edx,%esi + jz .L009maw_end + + movl 8(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 8(%edi),%eax + adcl $0,%edx + decl %ecx + movl %eax,8(%edi) + movl %edx,%esi + jz .L009maw_end + + movl 12(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 12(%edi),%eax + adcl $0,%edx + decl %ecx + movl %eax,12(%edi) + movl %edx,%esi + jz .L009maw_end + + movl 16(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 16(%edi),%eax + adcl $0,%edx + decl %ecx + movl %eax,16(%edi) + movl %edx,%esi + jz .L009maw_end + + movl 20(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 20(%edi),%eax + adcl $0,%edx + decl %ecx + movl %eax,20(%edi) + movl %edx,%esi + jz .L009maw_end + + movl 24(%ebx),%eax + mull %ebp + addl %esi,%eax + adcl $0,%edx + addl 24(%edi),%eax + adcl $0,%edx + movl %eax,24(%edi) + movl %edx,%esi +.L009maw_end: + movl %esi,%eax + popl %ecx + popl %edi + popl %esi + popl %ebx + popl %ebp + ret +.size bn_mul_add_words,.-.L_bn_mul_add_words_begin +.globl bn_mul_words +.type bn_mul_words,@function +.align 16 +bn_mul_words: +.L_bn_mul_words_begin: + call .L010PIC_me_up +.L010PIC_me_up: + popl %eax + leal _GLOBAL_OFFSET_TABLE_+[.-.L010PIC_me_up](%eax),%eax + movl OPENSSL_ia32cap_P@GOT(%eax),%eax + btl $26,(%eax) + jnc .L011mw_non_sse2 + movl 4(%esp),%eax + movl 8(%esp),%edx + movl 12(%esp),%ecx + movd 16(%esp),%mm0 + pxor %mm1,%mm1 +.align 16 +.L012mw_sse2_loop: + movd (%edx),%mm2 + pmuludq %mm0,%mm2 + leal 4(%edx),%edx + paddq %mm2,%mm1 + movd %mm1,(%eax) + subl $1,%ecx + psrlq $32,%mm1 + leal 4(%eax),%eax + jnz .L012mw_sse2_loop + movd %mm1,%eax + emms + ret +.align 16 +.L011mw_non_sse2: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + + xorl %esi,%esi + movl 20(%esp),%edi + movl 24(%esp),%ebx + movl 28(%esp),%ebp + movl 32(%esp),%ecx + andl $4294967288,%ebp + jz .L013mw_finish +.L014mw_loop: + + movl (%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,(%edi) + movl %edx,%esi + + movl 4(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,4(%edi) + movl %edx,%esi + + movl 8(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,8(%edi) + movl %edx,%esi + + movl 12(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,12(%edi) + movl %edx,%esi + + movl 16(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,16(%edi) + movl %edx,%esi + + movl 20(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,20(%edi) + movl %edx,%esi + + movl 24(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,24(%edi) + movl %edx,%esi + + movl 28(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,28(%edi) + movl %edx,%esi + + addl $32,%ebx + addl $32,%edi + subl $8,%ebp + jz .L013mw_finish + jmp .L014mw_loop +.L013mw_finish: + movl 28(%esp),%ebp + andl $7,%ebp + jnz .L015mw_finish2 + jmp .L016mw_end +.L015mw_finish2: + + movl (%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,(%edi) + movl %edx,%esi + decl %ebp + jz .L016mw_end + + movl 4(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,4(%edi) + movl %edx,%esi + decl %ebp + jz .L016mw_end + + movl 8(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,8(%edi) + movl %edx,%esi + decl %ebp + jz .L016mw_end + + movl 12(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,12(%edi) + movl %edx,%esi + decl %ebp + jz .L016mw_end + + movl 16(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,16(%edi) + movl %edx,%esi + decl %ebp + jz .L016mw_end + + movl 20(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,20(%edi) + movl %edx,%esi + decl %ebp + jz .L016mw_end + + movl 24(%ebx),%eax + mull %ecx + addl %esi,%eax + adcl $0,%edx + movl %eax,24(%edi) + movl %edx,%esi +.L016mw_end: + movl %esi,%eax + popl %edi + popl %esi + popl %ebx + popl %ebp + ret +.size bn_mul_words,.-.L_bn_mul_words_begin +.globl bn_sqr_words +.type bn_sqr_words,@function +.align 16 +bn_sqr_words: +.L_bn_sqr_words_begin: + call .L017PIC_me_up +.L017PIC_me_up: + popl %eax + leal _GLOBAL_OFFSET_TABLE_+[.-.L017PIC_me_up](%eax),%eax + movl OPENSSL_ia32cap_P@GOT(%eax),%eax + btl $26,(%eax) + jnc .L018sqr_non_sse2 + movl 4(%esp),%eax + movl 8(%esp),%edx + movl 12(%esp),%ecx +.align 16 +.L019sqr_sse2_loop: + movd (%edx),%mm0 + pmuludq %mm0,%mm0 + leal 4(%edx),%edx + movq %mm0,(%eax) + subl $1,%ecx + leal 8(%eax),%eax + jnz .L019sqr_sse2_loop + emms + ret +.align 16 +.L018sqr_non_sse2: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + + movl 20(%esp),%esi + movl 24(%esp),%edi + movl 28(%esp),%ebx + andl $4294967288,%ebx + jz .L020sw_finish +.L021sw_loop: + + movl (%edi),%eax + mull %eax + movl %eax,(%esi) + movl %edx,4(%esi) + + movl 4(%edi),%eax + mull %eax + movl %eax,8(%esi) + movl %edx,12(%esi) + + movl 8(%edi),%eax + mull %eax + movl %eax,16(%esi) + movl %edx,20(%esi) + + movl 12(%edi),%eax + mull %eax + movl %eax,24(%esi) + movl %edx,28(%esi) + + movl 16(%edi),%eax + mull %eax + movl %eax,32(%esi) + movl %edx,36(%esi) + + movl 20(%edi),%eax + mull %eax + movl %eax,40(%esi) + movl %edx,44(%esi) + + movl 24(%edi),%eax + mull %eax + movl %eax,48(%esi) + movl %edx,52(%esi) + + movl 28(%edi),%eax + mull %eax + movl %eax,56(%esi) + movl %edx,60(%esi) + + addl $32,%edi + addl $64,%esi + subl $8,%ebx + jnz .L021sw_loop +.L020sw_finish: + movl 28(%esp),%ebx + andl $7,%ebx + jz .L022sw_end + + movl (%edi),%eax + mull %eax + movl %eax,(%esi) + decl %ebx + movl %edx,4(%esi) + jz .L022sw_end + + movl 4(%edi),%eax + mull %eax + movl %eax,8(%esi) + decl %ebx + movl %edx,12(%esi) + jz .L022sw_end + + movl 8(%edi),%eax + mull %eax + movl %eax,16(%esi) + decl %ebx + movl %edx,20(%esi) + jz .L022sw_end + + movl 12(%edi),%eax + mull %eax + movl %eax,24(%esi) + decl %ebx + movl %edx,28(%esi) + jz .L022sw_end + + movl 16(%edi),%eax + mull %eax + movl %eax,32(%esi) + decl %ebx + movl %edx,36(%esi) + jz .L022sw_end + + movl 20(%edi),%eax + mull %eax + movl %eax,40(%esi) + decl %ebx + movl %edx,44(%esi) + jz .L022sw_end + + movl 24(%edi),%eax + mull %eax + movl %eax,48(%esi) + movl %edx,52(%esi) +.L022sw_end: + popl %edi + popl %esi + popl %ebx + popl %ebp + ret +.size bn_sqr_words,.-.L_bn_sqr_words_begin +.globl bn_div_words +.type bn_div_words,@function +.align 16 +bn_div_words: +.L_bn_div_words_begin: + movl 4(%esp),%edx + movl 8(%esp),%eax + movl 12(%esp),%ecx + divl %ecx + ret +.size bn_div_words,.-.L_bn_div_words_begin +.globl bn_add_words +.type bn_add_words,@function +.align 16 +bn_add_words: +.L_bn_add_words_begin: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + + movl 20(%esp),%ebx + movl 24(%esp),%esi + movl 28(%esp),%edi + movl 32(%esp),%ebp + xorl %eax,%eax + andl $4294967288,%ebp + jz .L023aw_finish +.L024aw_loop: + + movl (%esi),%ecx + movl (%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + + movl 4(%esi),%ecx + movl 4(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,4(%ebx) + + movl 8(%esi),%ecx + movl 8(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,8(%ebx) + + movl 12(%esi),%ecx + movl 12(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,12(%ebx) + + movl 16(%esi),%ecx + movl 16(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,16(%ebx) + + movl 20(%esi),%ecx + movl 20(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,20(%ebx) + + movl 24(%esi),%ecx + movl 24(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,24(%ebx) + + movl 28(%esi),%ecx + movl 28(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,28(%ebx) + + addl $32,%esi + addl $32,%edi + addl $32,%ebx + subl $8,%ebp + jnz .L024aw_loop +.L023aw_finish: + movl 32(%esp),%ebp + andl $7,%ebp + jz .L025aw_end + + movl (%esi),%ecx + movl (%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,(%ebx) + jz .L025aw_end + + movl 4(%esi),%ecx + movl 4(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,4(%ebx) + jz .L025aw_end + + movl 8(%esi),%ecx + movl 8(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,8(%ebx) + jz .L025aw_end + + movl 12(%esi),%ecx + movl 12(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,12(%ebx) + jz .L025aw_end + + movl 16(%esi),%ecx + movl 16(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,16(%ebx) + jz .L025aw_end + + movl 20(%esi),%ecx + movl 20(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,20(%ebx) + jz .L025aw_end + + movl 24(%esi),%ecx + movl 24(%edi),%edx + addl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + addl %edx,%ecx + adcl $0,%eax + movl %ecx,24(%ebx) +.L025aw_end: + popl %edi + popl %esi + popl %ebx + popl %ebp + ret +.size bn_add_words,.-.L_bn_add_words_begin +.globl bn_sub_words +.type bn_sub_words,@function +.align 16 +bn_sub_words: +.L_bn_sub_words_begin: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + + movl 20(%esp),%ebx + movl 24(%esp),%esi + movl 28(%esp),%edi + movl 32(%esp),%ebp + xorl %eax,%eax + andl $4294967288,%ebp + jz .L026aw_finish +.L027aw_loop: + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + + movl 4(%esi),%ecx + movl 4(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,4(%ebx) + + movl 8(%esi),%ecx + movl 8(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,8(%ebx) + + movl 12(%esi),%ecx + movl 12(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,12(%ebx) + + movl 16(%esi),%ecx + movl 16(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,16(%ebx) + + movl 20(%esi),%ecx + movl 20(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,20(%ebx) + + movl 24(%esi),%ecx + movl 24(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,24(%ebx) + + movl 28(%esi),%ecx + movl 28(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,28(%ebx) + + addl $32,%esi + addl $32,%edi + addl $32,%ebx + subl $8,%ebp + jnz .L027aw_loop +.L026aw_finish: + movl 32(%esp),%ebp + andl $7,%ebp + jz .L028aw_end + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,(%ebx) + jz .L028aw_end + + movl 4(%esi),%ecx + movl 4(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,4(%ebx) + jz .L028aw_end + + movl 8(%esi),%ecx + movl 8(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,8(%ebx) + jz .L028aw_end + + movl 12(%esi),%ecx + movl 12(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,12(%ebx) + jz .L028aw_end + + movl 16(%esi),%ecx + movl 16(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,16(%ebx) + jz .L028aw_end + + movl 20(%esi),%ecx + movl 20(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,20(%ebx) + jz .L028aw_end + + movl 24(%esi),%ecx + movl 24(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,24(%ebx) +.L028aw_end: + popl %edi + popl %esi + popl %ebx + popl %ebp + ret +.size bn_sub_words,.-.L_bn_sub_words_begin +.globl bn_sub_part_words +.type bn_sub_part_words,@function +.align 16 +bn_sub_part_words: +.L_bn_sub_part_words_begin: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + + movl 20(%esp),%ebx + movl 24(%esp),%esi + movl 28(%esp),%edi + movl 32(%esp),%ebp + xorl %eax,%eax + andl $4294967288,%ebp + jz .L029aw_finish +.L030aw_loop: + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + + movl 4(%esi),%ecx + movl 4(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,4(%ebx) + + movl 8(%esi),%ecx + movl 8(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,8(%ebx) + + movl 12(%esi),%ecx + movl 12(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,12(%ebx) + + movl 16(%esi),%ecx + movl 16(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,16(%ebx) + + movl 20(%esi),%ecx + movl 20(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,20(%ebx) + + movl 24(%esi),%ecx + movl 24(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,24(%ebx) + + movl 28(%esi),%ecx + movl 28(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,28(%ebx) + + addl $32,%esi + addl $32,%edi + addl $32,%ebx + subl $8,%ebp + jnz .L030aw_loop +.L029aw_finish: + movl 32(%esp),%ebp + andl $7,%ebp + jz .L031aw_end + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + addl $4,%esi + addl $4,%edi + addl $4,%ebx + decl %ebp + jz .L031aw_end + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + addl $4,%esi + addl $4,%edi + addl $4,%ebx + decl %ebp + jz .L031aw_end + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + addl $4,%esi + addl $4,%edi + addl $4,%ebx + decl %ebp + jz .L031aw_end + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + addl $4,%esi + addl $4,%edi + addl $4,%ebx + decl %ebp + jz .L031aw_end + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + addl $4,%esi + addl $4,%edi + addl $4,%ebx + decl %ebp + jz .L031aw_end + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + addl $4,%esi + addl $4,%edi + addl $4,%ebx + decl %ebp + jz .L031aw_end + + movl (%esi),%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + addl $4,%esi + addl $4,%edi + addl $4,%ebx +.L031aw_end: + cmpl $0,36(%esp) + je .L032pw_end + movl 36(%esp),%ebp + cmpl $0,%ebp + je .L032pw_end + jge .L033pw_pos + + movl $0,%edx + subl %ebp,%edx + movl %edx,%ebp + andl $4294967288,%ebp + jz .L034pw_neg_finish +.L035pw_neg_loop: + + movl $0,%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,(%ebx) + + movl $0,%ecx + movl 4(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,4(%ebx) + + movl $0,%ecx + movl 8(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,8(%ebx) + + movl $0,%ecx + movl 12(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,12(%ebx) + + movl $0,%ecx + movl 16(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,16(%ebx) + + movl $0,%ecx + movl 20(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,20(%ebx) + + movl $0,%ecx + movl 24(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,24(%ebx) + + movl $0,%ecx + movl 28(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,28(%ebx) + + addl $32,%edi + addl $32,%ebx + subl $8,%ebp + jnz .L035pw_neg_loop +.L034pw_neg_finish: + movl 36(%esp),%edx + movl $0,%ebp + subl %edx,%ebp + andl $7,%ebp + jz .L032pw_end + + movl $0,%ecx + movl (%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,(%ebx) + jz .L032pw_end + + movl $0,%ecx + movl 4(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,4(%ebx) + jz .L032pw_end + + movl $0,%ecx + movl 8(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,8(%ebx) + jz .L032pw_end + + movl $0,%ecx + movl 12(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,12(%ebx) + jz .L032pw_end + + movl $0,%ecx + movl 16(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,16(%ebx) + jz .L032pw_end + + movl $0,%ecx + movl 20(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + decl %ebp + movl %ecx,20(%ebx) + jz .L032pw_end + + movl $0,%ecx + movl 24(%edi),%edx + subl %eax,%ecx + movl $0,%eax + adcl %eax,%eax + subl %edx,%ecx + adcl $0,%eax + movl %ecx,24(%ebx) + jmp .L032pw_end +.L033pw_pos: + andl $4294967288,%ebp + jz .L036pw_pos_finish +.L037pw_pos_loop: + + movl (%esi),%ecx + subl %eax,%ecx + movl %ecx,(%ebx) + jnc .L038pw_nc0 + + movl 4(%esi),%ecx + subl %eax,%ecx + movl %ecx,4(%ebx) + jnc .L039pw_nc1 + + movl 8(%esi),%ecx + subl %eax,%ecx + movl %ecx,8(%ebx) + jnc .L040pw_nc2 + + movl 12(%esi),%ecx + subl %eax,%ecx + movl %ecx,12(%ebx) + jnc .L041pw_nc3 + + movl 16(%esi),%ecx + subl %eax,%ecx + movl %ecx,16(%ebx) + jnc .L042pw_nc4 + + movl 20(%esi),%ecx + subl %eax,%ecx + movl %ecx,20(%ebx) + jnc .L043pw_nc5 + + movl 24(%esi),%ecx + subl %eax,%ecx + movl %ecx,24(%ebx) + jnc .L044pw_nc6 + + movl 28(%esi),%ecx + subl %eax,%ecx + movl %ecx,28(%ebx) + jnc .L045pw_nc7 + + addl $32,%esi + addl $32,%ebx + subl $8,%ebp + jnz .L037pw_pos_loop +.L036pw_pos_finish: + movl 36(%esp),%ebp + andl $7,%ebp + jz .L032pw_end + + movl (%esi),%ecx + subl %eax,%ecx + movl %ecx,(%ebx) + jnc .L046pw_tail_nc0 + decl %ebp + jz .L032pw_end + + movl 4(%esi),%ecx + subl %eax,%ecx + movl %ecx,4(%ebx) + jnc .L047pw_tail_nc1 + decl %ebp + jz .L032pw_end + + movl 8(%esi),%ecx + subl %eax,%ecx + movl %ecx,8(%ebx) + jnc .L048pw_tail_nc2 + decl %ebp + jz .L032pw_end + + movl 12(%esi),%ecx + subl %eax,%ecx + movl %ecx,12(%ebx) + jnc .L049pw_tail_nc3 + decl %ebp + jz .L032pw_end + + movl 16(%esi),%ecx + subl %eax,%ecx + movl %ecx,16(%ebx) + jnc .L050pw_tail_nc4 + decl %ebp + jz .L032pw_end + + movl 20(%esi),%ecx + subl %eax,%ecx + movl %ecx,20(%ebx) + jnc .L051pw_tail_nc5 + decl %ebp + jz .L032pw_end + + movl 24(%esi),%ecx + subl %eax,%ecx + movl %ecx,24(%ebx) + jnc .L052pw_tail_nc6 + movl $1,%eax + jmp .L032pw_end +.L053pw_nc_loop: + movl (%esi),%ecx + movl %ecx,(%ebx) +.L038pw_nc0: + movl 4(%esi),%ecx + movl %ecx,4(%ebx) +.L039pw_nc1: + movl 8(%esi),%ecx + movl %ecx,8(%ebx) +.L040pw_nc2: + movl 12(%esi),%ecx + movl %ecx,12(%ebx) +.L041pw_nc3: + movl 16(%esi),%ecx + movl %ecx,16(%ebx) +.L042pw_nc4: + movl 20(%esi),%ecx + movl %ecx,20(%ebx) +.L043pw_nc5: + movl 24(%esi),%ecx + movl %ecx,24(%ebx) +.L044pw_nc6: + movl 28(%esi),%ecx + movl %ecx,28(%ebx) +.L045pw_nc7: + + addl $32,%esi + addl $32,%ebx + subl $8,%ebp + jnz .L053pw_nc_loop + movl 36(%esp),%ebp + andl $7,%ebp + jz .L054pw_nc_end + movl (%esi),%ecx + movl %ecx,(%ebx) +.L046pw_tail_nc0: + decl %ebp + jz .L054pw_nc_end + movl 4(%esi),%ecx + movl %ecx,4(%ebx) +.L047pw_tail_nc1: + decl %ebp + jz .L054pw_nc_end + movl 8(%esi),%ecx + movl %ecx,8(%ebx) +.L048pw_tail_nc2: + decl %ebp + jz .L054pw_nc_end + movl 12(%esi),%ecx + movl %ecx,12(%ebx) +.L049pw_tail_nc3: + decl %ebp + jz .L054pw_nc_end + movl 16(%esi),%ecx + movl %ecx,16(%ebx) +.L050pw_tail_nc4: + decl %ebp + jz .L054pw_nc_end + movl 20(%esi),%ecx + movl %ecx,20(%ebx) +.L051pw_tail_nc5: + decl %ebp + jz .L054pw_nc_end + movl 24(%esi),%ecx + movl %ecx,24(%ebx) +.L052pw_tail_nc6: +.L054pw_nc_end: + movl $0,%eax +.L032pw_end: + popl %edi + popl %esi + popl %ebx + popl %ebp + ret +.size bn_sub_part_words,.-.L_bn_sub_part_words_begin +.comm OPENSSL_ia32cap_P,8,4 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-586.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-586.pl new file mode 100644 index 00000000..332ef3e9 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-586.pl @@ -0,0 +1,774 @@ +#!/usr/local/bin/perl + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "x86asm.pl"; + +&asm_init($ARGV[0],$0); + +$sse2=0; +for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } + +&external_label("OPENSSL_ia32cap_P") if ($sse2); + +&bn_mul_add_words("bn_mul_add_words"); +&bn_mul_words("bn_mul_words"); +&bn_sqr_words("bn_sqr_words"); +&bn_div_words("bn_div_words"); +&bn_add_words("bn_add_words"); +&bn_sub_words("bn_sub_words"); +&bn_sub_part_words("bn_sub_part_words"); + +&asm_finish(); + +sub bn_mul_add_words + { + local($name)=@_; + + &function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":""); + + $r="eax"; + $a="edx"; + $c="ecx"; + + if ($sse2) { + &picmeup("eax","OPENSSL_ia32cap_P"); + &bt(&DWP(0,"eax"),26); + &jnc(&label("maw_non_sse2")); + + &mov($r,&wparam(0)); + &mov($a,&wparam(1)); + &mov($c,&wparam(2)); + &movd("mm0",&wparam(3)); # mm0 = w + &pxor("mm1","mm1"); # mm1 = carry_in + &jmp(&label("maw_sse2_entry")); + + &set_label("maw_sse2_unrolled",16); + &movd("mm3",&DWP(0,$r,"",0)); # mm3 = r[0] + &paddq("mm1","mm3"); # mm1 = carry_in + r[0] + &movd("mm2",&DWP(0,$a,"",0)); # mm2 = a[0] + &pmuludq("mm2","mm0"); # mm2 = w*a[0] + &movd("mm4",&DWP(4,$a,"",0)); # mm4 = a[1] + &pmuludq("mm4","mm0"); # mm4 = w*a[1] + &movd("mm6",&DWP(8,$a,"",0)); # mm6 = a[2] + &pmuludq("mm6","mm0"); # mm6 = w*a[2] + &movd("mm7",&DWP(12,$a,"",0)); # mm7 = a[3] + &pmuludq("mm7","mm0"); # mm7 = w*a[3] + &paddq("mm1","mm2"); # mm1 = carry_in + r[0] + w*a[0] + &movd("mm3",&DWP(4,$r,"",0)); # mm3 = r[1] + &paddq("mm3","mm4"); # mm3 = r[1] + w*a[1] + &movd("mm5",&DWP(8,$r,"",0)); # mm5 = r[2] + &paddq("mm5","mm6"); # mm5 = r[2] + w*a[2] + &movd("mm4",&DWP(12,$r,"",0)); # mm4 = r[3] + &paddq("mm7","mm4"); # mm7 = r[3] + w*a[3] + &movd(&DWP(0,$r,"",0),"mm1"); + &movd("mm2",&DWP(16,$a,"",0)); # mm2 = a[4] + &pmuludq("mm2","mm0"); # mm2 = w*a[4] + &psrlq("mm1",32); # mm1 = carry0 + &movd("mm4",&DWP(20,$a,"",0)); # mm4 = a[5] + &pmuludq("mm4","mm0"); # mm4 = w*a[5] + &paddq("mm1","mm3"); # mm1 = carry0 + r[1] + w*a[1] + &movd("mm6",&DWP(24,$a,"",0)); # mm6 = a[6] + &pmuludq("mm6","mm0"); # mm6 = w*a[6] + &movd(&DWP(4,$r,"",0),"mm1"); + &psrlq("mm1",32); # mm1 = carry1 + &movd("mm3",&DWP(28,$a,"",0)); # mm3 = a[7] + &add($a,32); + &pmuludq("mm3","mm0"); # mm3 = w*a[7] + &paddq("mm1","mm5"); # mm1 = carry1 + r[2] + w*a[2] + &movd("mm5",&DWP(16,$r,"",0)); # mm5 = r[4] + &paddq("mm2","mm5"); # mm2 = r[4] + w*a[4] + &movd(&DWP(8,$r,"",0),"mm1"); + &psrlq("mm1",32); # mm1 = carry2 + &paddq("mm1","mm7"); # mm1 = carry2 + r[3] + w*a[3] + &movd("mm5",&DWP(20,$r,"",0)); # mm5 = r[5] + &paddq("mm4","mm5"); # mm4 = r[5] + w*a[5] + &movd(&DWP(12,$r,"",0),"mm1"); + &psrlq("mm1",32); # mm1 = carry3 + &paddq("mm1","mm2"); # mm1 = carry3 + r[4] + w*a[4] + &movd("mm5",&DWP(24,$r,"",0)); # mm5 = r[6] + &paddq("mm6","mm5"); # mm6 = r[6] + w*a[6] + &movd(&DWP(16,$r,"",0),"mm1"); + &psrlq("mm1",32); # mm1 = carry4 + &paddq("mm1","mm4"); # mm1 = carry4 + r[5] + w*a[5] + &movd("mm5",&DWP(28,$r,"",0)); # mm5 = r[7] + &paddq("mm3","mm5"); # mm3 = r[7] + w*a[7] + &movd(&DWP(20,$r,"",0),"mm1"); + &psrlq("mm1",32); # mm1 = carry5 + &paddq("mm1","mm6"); # mm1 = carry5 + r[6] + w*a[6] + &movd(&DWP(24,$r,"",0),"mm1"); + &psrlq("mm1",32); # mm1 = carry6 + &paddq("mm1","mm3"); # mm1 = carry6 + r[7] + w*a[7] + &movd(&DWP(28,$r,"",0),"mm1"); + &lea($r,&DWP(32,$r)); + &psrlq("mm1",32); # mm1 = carry_out + + &sub($c,8); + &jz(&label("maw_sse2_exit")); + &set_label("maw_sse2_entry"); + &test($c,0xfffffff8); + &jnz(&label("maw_sse2_unrolled")); + + &set_label("maw_sse2_loop",4); + &movd("mm2",&DWP(0,$a)); # mm2 = a[i] + &movd("mm3",&DWP(0,$r)); # mm3 = r[i] + &pmuludq("mm2","mm0"); # a[i] *= w + &lea($a,&DWP(4,$a)); + &paddq("mm1","mm3"); # carry += r[i] + &paddq("mm1","mm2"); # carry += a[i]*w + &movd(&DWP(0,$r),"mm1"); # r[i] = carry_low + &sub($c,1); + &psrlq("mm1",32); # carry = carry_high + &lea($r,&DWP(4,$r)); + &jnz(&label("maw_sse2_loop")); + &set_label("maw_sse2_exit"); + &movd("eax","mm1"); # c = carry_out + &emms(); + &ret(); + + &set_label("maw_non_sse2",16); + } + + # function_begin prologue + &push("ebp"); + &push("ebx"); + &push("esi"); + &push("edi"); + + &comment(""); + $Low="eax"; + $High="edx"; + $a="ebx"; + $w="ebp"; + $r="edi"; + $c="esi"; + + &xor($c,$c); # clear carry + &mov($r,&wparam(0)); # + + &mov("ecx",&wparam(2)); # + &mov($a,&wparam(1)); # + + &and("ecx",0xfffffff8); # num / 8 + &mov($w,&wparam(3)); # + + &push("ecx"); # Up the stack for a tmp variable + + &jz(&label("maw_finish")); + + &set_label("maw_loop",16); + + for ($i=0; $i<32; $i+=4) + { + &comment("Round $i"); + + &mov("eax",&DWP($i,$a)); # *a + &mul($w); # *a * w + &add("eax",$c); # L(t)+= c + &adc("edx",0); # H(t)+=carry + &add("eax",&DWP($i,$r)); # L(t)+= *r + &adc("edx",0); # H(t)+=carry + &mov(&DWP($i,$r),"eax"); # *r= L(t); + &mov($c,"edx"); # c= H(t); + } + + &comment(""); + &sub("ecx",8); + &lea($a,&DWP(32,$a)); + &lea($r,&DWP(32,$r)); + &jnz(&label("maw_loop")); + + &set_label("maw_finish",0); + &mov("ecx",&wparam(2)); # get num + &and("ecx",7); + &jnz(&label("maw_finish2")); # helps branch prediction + &jmp(&label("maw_end")); + + &set_label("maw_finish2",1); + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov("eax",&DWP($i*4,$a)); # *a + &mul($w); # *a * w + &add("eax",$c); # L(t)+=c + &adc("edx",0); # H(t)+=carry + &add("eax",&DWP($i*4,$r)); # L(t)+= *r + &adc("edx",0); # H(t)+=carry + &dec("ecx") if ($i != 7-1); + &mov(&DWP($i*4,$r),"eax"); # *r= L(t); + &mov($c,"edx"); # c= H(t); + &jz(&label("maw_end")) if ($i != 7-1); + } + &set_label("maw_end",0); + &mov("eax",$c); + + &pop("ecx"); # clear variable from + + &function_end($name); + } + +sub bn_mul_words + { + local($name)=@_; + + &function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":""); + + $r="eax"; + $a="edx"; + $c="ecx"; + + if ($sse2) { + &picmeup("eax","OPENSSL_ia32cap_P"); + &bt(&DWP(0,"eax"),26); + &jnc(&label("mw_non_sse2")); + + &mov($r,&wparam(0)); + &mov($a,&wparam(1)); + &mov($c,&wparam(2)); + &movd("mm0",&wparam(3)); # mm0 = w + &pxor("mm1","mm1"); # mm1 = carry = 0 + + &set_label("mw_sse2_loop",16); + &movd("mm2",&DWP(0,$a)); # mm2 = a[i] + &pmuludq("mm2","mm0"); # a[i] *= w + &lea($a,&DWP(4,$a)); + &paddq("mm1","mm2"); # carry += a[i]*w + &movd(&DWP(0,$r),"mm1"); # r[i] = carry_low + &sub($c,1); + &psrlq("mm1",32); # carry = carry_high + &lea($r,&DWP(4,$r)); + &jnz(&label("mw_sse2_loop")); + + &movd("eax","mm1"); # return carry + &emms(); + &ret(); + &set_label("mw_non_sse2",16); + } + + # function_begin prologue + &push("ebp"); + &push("ebx"); + &push("esi"); + &push("edi"); + + &comment(""); + $Low="eax"; + $High="edx"; + $a="ebx"; + $w="ecx"; + $r="edi"; + $c="esi"; + $num="ebp"; + + &xor($c,$c); # clear carry + &mov($r,&wparam(0)); # + &mov($a,&wparam(1)); # + &mov($num,&wparam(2)); # + &mov($w,&wparam(3)); # + + &and($num,0xfffffff8); # num / 8 + &jz(&label("mw_finish")); + + &set_label("mw_loop",0); + for ($i=0; $i<32; $i+=4) + { + &comment("Round $i"); + + &mov("eax",&DWP($i,$a,"",0)); # *a + &mul($w); # *a * w + &add("eax",$c); # L(t)+=c + # XXX + + &adc("edx",0); # H(t)+=carry + &mov(&DWP($i,$r,"",0),"eax"); # *r= L(t); + + &mov($c,"edx"); # c= H(t); + } + + &comment(""); + &add($a,32); + &add($r,32); + &sub($num,8); + &jz(&label("mw_finish")); + &jmp(&label("mw_loop")); + + &set_label("mw_finish",0); + &mov($num,&wparam(2)); # get num + &and($num,7); + &jnz(&label("mw_finish2")); + &jmp(&label("mw_end")); + + &set_label("mw_finish2",1); + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov("eax",&DWP($i*4,$a,"",0));# *a + &mul($w); # *a * w + &add("eax",$c); # L(t)+=c + # XXX + &adc("edx",0); # H(t)+=carry + &mov(&DWP($i*4,$r,"",0),"eax");# *r= L(t); + &mov($c,"edx"); # c= H(t); + &dec($num) if ($i != 7-1); + &jz(&label("mw_end")) if ($i != 7-1); + } + &set_label("mw_end",0); + &mov("eax",$c); + + &function_end($name); + } + +sub bn_sqr_words + { + local($name)=@_; + + &function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":""); + + $r="eax"; + $a="edx"; + $c="ecx"; + + if ($sse2) { + &picmeup("eax","OPENSSL_ia32cap_P"); + &bt(&DWP(0,"eax"),26); + &jnc(&label("sqr_non_sse2")); + + &mov($r,&wparam(0)); + &mov($a,&wparam(1)); + &mov($c,&wparam(2)); + + &set_label("sqr_sse2_loop",16); + &movd("mm0",&DWP(0,$a)); # mm0 = a[i] + &pmuludq("mm0","mm0"); # a[i] *= a[i] + &lea($a,&DWP(4,$a)); # a++ + &movq(&QWP(0,$r),"mm0"); # r[i] = a[i]*a[i] + &sub($c,1); + &lea($r,&DWP(8,$r)); # r += 2 + &jnz(&label("sqr_sse2_loop")); + + &emms(); + &ret(); + &set_label("sqr_non_sse2",16); + } + + # function_begin prologue + &push("ebp"); + &push("ebx"); + &push("esi"); + &push("edi"); + + &comment(""); + $r="esi"; + $a="edi"; + $num="ebx"; + + &mov($r,&wparam(0)); # + &mov($a,&wparam(1)); # + &mov($num,&wparam(2)); # + + &and($num,0xfffffff8); # num / 8 + &jz(&label("sw_finish")); + + &set_label("sw_loop",0); + for ($i=0; $i<32; $i+=4) + { + &comment("Round $i"); + &mov("eax",&DWP($i,$a,"",0)); # *a + # XXX + &mul("eax"); # *a * *a + &mov(&DWP($i*2,$r,"",0),"eax"); # + &mov(&DWP($i*2+4,$r,"",0),"edx");# + } + + &comment(""); + &add($a,32); + &add($r,64); + &sub($num,8); + &jnz(&label("sw_loop")); + + &set_label("sw_finish",0); + &mov($num,&wparam(2)); # get num + &and($num,7); + &jz(&label("sw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov("eax",&DWP($i*4,$a,"",0)); # *a + # XXX + &mul("eax"); # *a * *a + &mov(&DWP($i*8,$r,"",0),"eax"); # + &dec($num) if ($i != 7-1); + &mov(&DWP($i*8+4,$r,"",0),"edx"); + &jz(&label("sw_end")) if ($i != 7-1); + } + &set_label("sw_end",0); + + &function_end($name); + } + +sub bn_div_words + { + local($name)=@_; + + &function_begin_B($name,""); + &mov("edx",&wparam(0)); # + &mov("eax",&wparam(1)); # + &mov("ecx",&wparam(2)); # + &div("ecx"); + &ret(); + &function_end_B($name); + } + +sub bn_add_words + { + local($name)=@_; + + &function_begin($name,""); + + &comment(""); + $a="esi"; + $b="edi"; + $c="eax"; + $r="ebx"; + $tmp1="ecx"; + $tmp2="edx"; + $num="ebp"; + + &mov($r,&wparam(0)); # get r + &mov($a,&wparam(1)); # get a + &mov($b,&wparam(2)); # get b + &mov($num,&wparam(3)); # get num + &xor($c,$c); # clear carry + &and($num,0xfffffff8); # num / 8 + + &jz(&label("aw_finish")); + + &set_label("aw_loop",0); + for ($i=0; $i<8; $i++) + { + &comment("Round $i"); + + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0)); # *b + &add($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &add($tmp1,$tmp2); + &adc($c,0); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + } + + &comment(""); + &add($a,32); + &add($b,32); + &add($r,32); + &sub($num,8); + &jnz(&label("aw_loop")); + + &set_label("aw_finish",0); + &mov($num,&wparam(3)); # get num + &and($num,7); + &jz(&label("aw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0));# *b + &add($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &add($tmp1,$tmp2); + &adc($c,0); + &dec($num) if ($i != 6); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + &jz(&label("aw_end")) if ($i != 6); + } + &set_label("aw_end",0); + +# &mov("eax",$c); # $c is "eax" + + &function_end($name); + } + +sub bn_sub_words + { + local($name)=@_; + + &function_begin($name,""); + + &comment(""); + $a="esi"; + $b="edi"; + $c="eax"; + $r="ebx"; + $tmp1="ecx"; + $tmp2="edx"; + $num="ebp"; + + &mov($r,&wparam(0)); # get r + &mov($a,&wparam(1)); # get a + &mov($b,&wparam(2)); # get b + &mov($num,&wparam(3)); # get num + &xor($c,$c); # clear carry + &and($num,0xfffffff8); # num / 8 + + &jz(&label("aw_finish")); + + &set_label("aw_loop",0); + for ($i=0; $i<8; $i++) + { + &comment("Round $i"); + + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0)); # *b + &sub($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &sub($tmp1,$tmp2); + &adc($c,0); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + } + + &comment(""); + &add($a,32); + &add($b,32); + &add($r,32); + &sub($num,8); + &jnz(&label("aw_loop")); + + &set_label("aw_finish",0); + &mov($num,&wparam(3)); # get num + &and($num,7); + &jz(&label("aw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0));# *b + &sub($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &sub($tmp1,$tmp2); + &adc($c,0); + &dec($num) if ($i != 6); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + &jz(&label("aw_end")) if ($i != 6); + } + &set_label("aw_end",0); + +# &mov("eax",$c); # $c is "eax" + + &function_end($name); + } + +sub bn_sub_part_words + { + local($name)=@_; + + &function_begin($name,""); + + &comment(""); + $a="esi"; + $b="edi"; + $c="eax"; + $r="ebx"; + $tmp1="ecx"; + $tmp2="edx"; + $num="ebp"; + + &mov($r,&wparam(0)); # get r + &mov($a,&wparam(1)); # get a + &mov($b,&wparam(2)); # get b + &mov($num,&wparam(3)); # get num + &xor($c,$c); # clear carry + &and($num,0xfffffff8); # num / 8 + + &jz(&label("aw_finish")); + + &set_label("aw_loop",0); + for ($i=0; $i<8; $i++) + { + &comment("Round $i"); + + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0)); # *b + &sub($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &sub($tmp1,$tmp2); + &adc($c,0); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + } + + &comment(""); + &add($a,32); + &add($b,32); + &add($r,32); + &sub($num,8); + &jnz(&label("aw_loop")); + + &set_label("aw_finish",0); + &mov($num,&wparam(3)); # get num + &and($num,7); + &jz(&label("aw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov($tmp1,&DWP(0,$a,"",0)); # *a + &mov($tmp2,&DWP(0,$b,"",0));# *b + &sub($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &sub($tmp1,$tmp2); + &adc($c,0); + &mov(&DWP(0,$r,"",0),$tmp1); # *r + &add($a, 4); + &add($b, 4); + &add($r, 4); + &dec($num) if ($i != 6); + &jz(&label("aw_end")) if ($i != 6); + } + &set_label("aw_end",0); + + &cmp(&wparam(4),0); + &je(&label("pw_end")); + + &mov($num,&wparam(4)); # get dl + &cmp($num,0); + &je(&label("pw_end")); + &jge(&label("pw_pos")); + + &comment("pw_neg"); + &mov($tmp2,0); + &sub($tmp2,$num); + &mov($num,$tmp2); + &and($num,0xfffffff8); # num / 8 + &jz(&label("pw_neg_finish")); + + &set_label("pw_neg_loop",0); + for ($i=0; $i<8; $i++) + { + &comment("dl<0 Round $i"); + + &mov($tmp1,0); + &mov($tmp2,&DWP($i*4,$b,"",0)); # *b + &sub($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &sub($tmp1,$tmp2); + &adc($c,0); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + } + + &comment(""); + &add($b,32); + &add($r,32); + &sub($num,8); + &jnz(&label("pw_neg_loop")); + + &set_label("pw_neg_finish",0); + &mov($tmp2,&wparam(4)); # get dl + &mov($num,0); + &sub($num,$tmp2); + &and($num,7); + &jz(&label("pw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("dl<0 Tail Round $i"); + &mov($tmp1,0); + &mov($tmp2,&DWP($i*4,$b,"",0));# *b + &sub($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &sub($tmp1,$tmp2); + &adc($c,0); + &dec($num) if ($i != 6); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + &jz(&label("pw_end")) if ($i != 6); + } + + &jmp(&label("pw_end")); + + &set_label("pw_pos",0); + + &and($num,0xfffffff8); # num / 8 + &jz(&label("pw_pos_finish")); + + &set_label("pw_pos_loop",0); + + for ($i=0; $i<8; $i++) + { + &comment("dl>0 Round $i"); + + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &sub($tmp1,$c); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + &jnc(&label("pw_nc".$i)); + } + + &comment(""); + &add($a,32); + &add($r,32); + &sub($num,8); + &jnz(&label("pw_pos_loop")); + + &set_label("pw_pos_finish",0); + &mov($num,&wparam(4)); # get dl + &and($num,7); + &jz(&label("pw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("dl>0 Tail Round $i"); + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &sub($tmp1,$c); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + &jnc(&label("pw_tail_nc".$i)); + &dec($num) if ($i != 6); + &jz(&label("pw_end")) if ($i != 6); + } + &mov($c,1); + &jmp(&label("pw_end")); + + &set_label("pw_nc_loop",0); + for ($i=0; $i<8; $i++) + { + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + &set_label("pw_nc".$i,0); + } + + &comment(""); + &add($a,32); + &add($r,32); + &sub($num,8); + &jnz(&label("pw_nc_loop")); + + &mov($num,&wparam(4)); # get dl + &and($num,7); + &jz(&label("pw_nc_end")); + + for ($i=0; $i<7; $i++) + { + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + &set_label("pw_tail_nc".$i,0); + &dec($num) if ($i != 6); + &jz(&label("pw_nc_end")) if ($i != 6); + } + + &set_label("pw_nc_end",0); + &mov($c,0); + + &set_label("pw_end",0); + +# &mov("eax",$c); # $c is "eax" + + &function_end($name); + } + diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-mips.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-mips.S new file mode 100644 index 00000000..2e7cccb7 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/bn-mips.S @@ -0,0 +1,2175 @@ +.set mips2 +.rdata +.asciiz "mips3.s, Version 1.2" +.asciiz "MIPS II/III/IV ISA artwork by Andy Polyakov " + +.text +.set noat + +.align 5 +.globl bn_mul_add_words +.ent bn_mul_add_words +bn_mul_add_words: + .set noreorder + bgtz $6,bn_mul_add_words_internal + move $2,$0 + jr $31 + move $4,$2 +.end bn_mul_add_words + +.align 5 +.ent bn_mul_add_words_internal +bn_mul_add_words_internal: + .set reorder + li $3,-4 + and $8,$6,$3 + beqz $8,.L_bn_mul_add_words_tail + +.L_bn_mul_add_words_loop: + lw $12,0($5) + multu $12,$7 + lw $13,0($4) + lw $14,4($5) + lw $15,4($4) + lw $8,2*4($5) + lw $9,2*4($4) + addu $13,$2 + sltu $2,$13,$2 # All manuals say it "compares 32-bit + # values", but it seems to work fine + # even on 64-bit registers. + mflo $1 + mfhi $12 + addu $13,$1 + addu $2,$12 + multu $14,$7 + sltu $1,$13,$1 + sw $13,0($4) + addu $2,$1 + + lw $10,3*4($5) + lw $11,3*4($4) + addu $15,$2 + sltu $2,$15,$2 + mflo $1 + mfhi $14 + addu $15,$1 + addu $2,$14 + multu $8,$7 + sltu $1,$15,$1 + sw $15,4($4) + addu $2,$1 + + subu $6,4 + addu $4,4*4 + addu $5,4*4 + addu $9,$2 + sltu $2,$9,$2 + mflo $1 + mfhi $8 + addu $9,$1 + addu $2,$8 + multu $10,$7 + sltu $1,$9,$1 + sw $9,-2*4($4) + addu $2,$1 + + + and $8,$6,$3 + addu $11,$2 + sltu $2,$11,$2 + mflo $1 + mfhi $10 + addu $11,$1 + addu $2,$10 + sltu $1,$11,$1 + sw $11,-4($4) + .set noreorder + bgtz $8,.L_bn_mul_add_words_loop + addu $2,$1 + + beqz $6,.L_bn_mul_add_words_return + nop + +.L_bn_mul_add_words_tail: + .set reorder + lw $12,0($5) + multu $12,$7 + lw $13,0($4) + subu $6,1 + addu $13,$2 + sltu $2,$13,$2 + mflo $1 + mfhi $12 + addu $13,$1 + addu $2,$12 + sltu $1,$13,$1 + sw $13,0($4) + addu $2,$1 + beqz $6,.L_bn_mul_add_words_return + + lw $12,4($5) + multu $12,$7 + lw $13,4($4) + subu $6,1 + addu $13,$2 + sltu $2,$13,$2 + mflo $1 + mfhi $12 + addu $13,$1 + addu $2,$12 + sltu $1,$13,$1 + sw $13,4($4) + addu $2,$1 + beqz $6,.L_bn_mul_add_words_return + + lw $12,2*4($5) + multu $12,$7 + lw $13,2*4($4) + addu $13,$2 + sltu $2,$13,$2 + mflo $1 + mfhi $12 + addu $13,$1 + addu $2,$12 + sltu $1,$13,$1 + sw $13,2*4($4) + addu $2,$1 + +.L_bn_mul_add_words_return: + .set noreorder + jr $31 + move $4,$2 +.end bn_mul_add_words_internal + +.align 5 +.globl bn_mul_words +.ent bn_mul_words +bn_mul_words: + .set noreorder + bgtz $6,bn_mul_words_internal + move $2,$0 + jr $31 + move $4,$2 +.end bn_mul_words + +.align 5 +.ent bn_mul_words_internal +bn_mul_words_internal: + .set reorder + li $3,-4 + and $8,$6,$3 + beqz $8,.L_bn_mul_words_tail + +.L_bn_mul_words_loop: + lw $12,0($5) + multu $12,$7 + lw $14,4($5) + lw $8,2*4($5) + lw $10,3*4($5) + mflo $1 + mfhi $12 + addu $2,$1 + sltu $13,$2,$1 + multu $14,$7 + sw $2,0($4) + addu $2,$13,$12 + + subu $6,4 + addu $4,4*4 + addu $5,4*4 + mflo $1 + mfhi $14 + addu $2,$1 + sltu $15,$2,$1 + multu $8,$7 + sw $2,-3*4($4) + addu $2,$15,$14 + + mflo $1 + mfhi $8 + addu $2,$1 + sltu $9,$2,$1 + multu $10,$7 + sw $2,-2*4($4) + addu $2,$9,$8 + + and $8,$6,$3 + mflo $1 + mfhi $10 + addu $2,$1 + sltu $11,$2,$1 + sw $2,-4($4) + .set noreorder + bgtz $8,.L_bn_mul_words_loop + addu $2,$11,$10 + + beqz $6,.L_bn_mul_words_return + nop + +.L_bn_mul_words_tail: + .set reorder + lw $12,0($5) + multu $12,$7 + subu $6,1 + mflo $1 + mfhi $12 + addu $2,$1 + sltu $13,$2,$1 + sw $2,0($4) + addu $2,$13,$12 + beqz $6,.L_bn_mul_words_return + + lw $12,4($5) + multu $12,$7 + subu $6,1 + mflo $1 + mfhi $12 + addu $2,$1 + sltu $13,$2,$1 + sw $2,4($4) + addu $2,$13,$12 + beqz $6,.L_bn_mul_words_return + + lw $12,2*4($5) + multu $12,$7 + mflo $1 + mfhi $12 + addu $2,$1 + sltu $13,$2,$1 + sw $2,2*4($4) + addu $2,$13,$12 + +.L_bn_mul_words_return: + .set noreorder + jr $31 + move $4,$2 +.end bn_mul_words_internal + +.align 5 +.globl bn_sqr_words +.ent bn_sqr_words +bn_sqr_words: + .set noreorder + bgtz $6,bn_sqr_words_internal + move $2,$0 + jr $31 + move $4,$2 +.end bn_sqr_words + +.align 5 +.ent bn_sqr_words_internal +bn_sqr_words_internal: + .set reorder + li $3,-4 + and $8,$6,$3 + beqz $8,.L_bn_sqr_words_tail + +.L_bn_sqr_words_loop: + lw $12,0($5) + multu $12,$12 + lw $14,4($5) + lw $8,2*4($5) + lw $10,3*4($5) + mflo $13 + mfhi $12 + sw $13,0($4) + sw $12,4($4) + + multu $14,$14 + subu $6,4 + addu $4,8*4 + addu $5,4*4 + mflo $15 + mfhi $14 + sw $15,-6*4($4) + sw $14,-5*4($4) + + multu $8,$8 + mflo $9 + mfhi $8 + sw $9,-4*4($4) + sw $8,-3*4($4) + + + multu $10,$10 + and $8,$6,$3 + mflo $11 + mfhi $10 + sw $11,-2*4($4) + + .set noreorder + bgtz $8,.L_bn_sqr_words_loop + sw $10,-4($4) + + beqz $6,.L_bn_sqr_words_return + nop + +.L_bn_sqr_words_tail: + .set reorder + lw $12,0($5) + multu $12,$12 + subu $6,1 + mflo $13 + mfhi $12 + sw $13,0($4) + sw $12,4($4) + beqz $6,.L_bn_sqr_words_return + + lw $12,4($5) + multu $12,$12 + subu $6,1 + mflo $13 + mfhi $12 + sw $13,2*4($4) + sw $12,3*4($4) + beqz $6,.L_bn_sqr_words_return + + lw $12,2*4($5) + multu $12,$12 + mflo $13 + mfhi $12 + sw $13,4*4($4) + sw $12,5*4($4) + +.L_bn_sqr_words_return: + .set noreorder + jr $31 + move $4,$2 + +.end bn_sqr_words_internal + +.align 5 +.globl bn_add_words +.ent bn_add_words +bn_add_words: + .set noreorder + bgtz $7,bn_add_words_internal + move $2,$0 + jr $31 + move $4,$2 +.end bn_add_words + +.align 5 +.ent bn_add_words_internal +bn_add_words_internal: + .set reorder + li $3,-4 + and $1,$7,$3 + beqz $1,.L_bn_add_words_tail + +.L_bn_add_words_loop: + lw $12,0($5) + lw $8,0($6) + subu $7,4 + lw $13,4($5) + and $1,$7,$3 + lw $14,2*4($5) + addu $6,4*4 + lw $15,3*4($5) + addu $4,4*4 + lw $9,-3*4($6) + addu $5,4*4 + lw $10,-2*4($6) + lw $11,-4($6) + addu $8,$12 + sltu $24,$8,$12 + addu $12,$8,$2 + sltu $2,$12,$8 + sw $12,-4*4($4) + addu $2,$24 + + addu $9,$13 + sltu $25,$9,$13 + addu $13,$9,$2 + sltu $2,$13,$9 + sw $13,-3*4($4) + addu $2,$25 + + addu $10,$14 + sltu $24,$10,$14 + addu $14,$10,$2 + sltu $2,$14,$10 + sw $14,-2*4($4) + addu $2,$24 + + addu $11,$15 + sltu $25,$11,$15 + addu $15,$11,$2 + sltu $2,$15,$11 + sw $15,-4($4) + + .set noreorder + bgtz $1,.L_bn_add_words_loop + addu $2,$25 + + beqz $7,.L_bn_add_words_return + nop + +.L_bn_add_words_tail: + .set reorder + lw $12,0($5) + lw $8,0($6) + addu $8,$12 + subu $7,1 + sltu $24,$8,$12 + addu $12,$8,$2 + sltu $2,$12,$8 + sw $12,0($4) + addu $2,$24 + beqz $7,.L_bn_add_words_return + + lw $13,4($5) + lw $9,4($6) + addu $9,$13 + subu $7,1 + sltu $25,$9,$13 + addu $13,$9,$2 + sltu $2,$13,$9 + sw $13,4($4) + addu $2,$25 + beqz $7,.L_bn_add_words_return + + lw $14,2*4($5) + lw $10,2*4($6) + addu $10,$14 + sltu $24,$10,$14 + addu $14,$10,$2 + sltu $2,$14,$10 + sw $14,2*4($4) + addu $2,$24 + +.L_bn_add_words_return: + .set noreorder + jr $31 + move $4,$2 + +.end bn_add_words_internal + +.align 5 +.globl bn_sub_words +.ent bn_sub_words +bn_sub_words: + .set noreorder + bgtz $7,bn_sub_words_internal + move $2,$0 + jr $31 + move $4,$0 +.end bn_sub_words + +.align 5 +.ent bn_sub_words_internal +bn_sub_words_internal: + .set reorder + li $3,-4 + and $1,$7,$3 + beqz $1,.L_bn_sub_words_tail + +.L_bn_sub_words_loop: + lw $12,0($5) + lw $8,0($6) + subu $7,4 + lw $13,4($5) + and $1,$7,$3 + lw $14,2*4($5) + addu $6,4*4 + lw $15,3*4($5) + addu $4,4*4 + lw $9,-3*4($6) + addu $5,4*4 + lw $10,-2*4($6) + lw $11,-4($6) + sltu $24,$12,$8 + subu $8,$12,$8 + subu $12,$8,$2 + sgtu $2,$12,$8 + sw $12,-4*4($4) + addu $2,$24 + + sltu $25,$13,$9 + subu $9,$13,$9 + subu $13,$9,$2 + sgtu $2,$13,$9 + sw $13,-3*4($4) + addu $2,$25 + + + sltu $24,$14,$10 + subu $10,$14,$10 + subu $14,$10,$2 + sgtu $2,$14,$10 + sw $14,-2*4($4) + addu $2,$24 + + sltu $25,$15,$11 + subu $11,$15,$11 + subu $15,$11,$2 + sgtu $2,$15,$11 + sw $15,-4($4) + + .set noreorder + bgtz $1,.L_bn_sub_words_loop + addu $2,$25 + + beqz $7,.L_bn_sub_words_return + nop + +.L_bn_sub_words_tail: + .set reorder + lw $12,0($5) + lw $8,0($6) + subu $7,1 + sltu $24,$12,$8 + subu $8,$12,$8 + subu $12,$8,$2 + sgtu $2,$12,$8 + sw $12,0($4) + addu $2,$24 + beqz $7,.L_bn_sub_words_return + + lw $13,4($5) + subu $7,1 + lw $9,4($6) + sltu $25,$13,$9 + subu $9,$13,$9 + subu $13,$9,$2 + sgtu $2,$13,$9 + sw $13,4($4) + addu $2,$25 + beqz $7,.L_bn_sub_words_return + + lw $14,2*4($5) + lw $10,2*4($6) + sltu $24,$14,$10 + subu $10,$14,$10 + subu $14,$10,$2 + sgtu $2,$14,$10 + sw $14,2*4($4) + addu $2,$24 + +.L_bn_sub_words_return: + .set noreorder + jr $31 + move $4,$2 +.end bn_sub_words_internal + +.align 5 +.globl bn_div_3_words +.ent bn_div_3_words +bn_div_3_words: + .set noreorder + move $7,$4 # we know that bn_div_words does not + # touch $7, $10, $11 and preserves $6 + # so that we can save two arguments + # and return address in registers + # instead of stack:-) + + lw $4,($7) + move $10,$5 + bne $4,$6,bn_div_3_words_internal + lw $5,-4($7) + li $2,-1 + jr $31 + move $4,$2 +.end bn_div_3_words + +.align 5 +.ent bn_div_3_words_internal +bn_div_3_words_internal: + .set reorder + move $11,$31 + bal bn_div_words_internal + move $31,$11 + multu $10,$2 + lw $14,-2*4($7) + move $8,$0 + mfhi $13 + mflo $12 + sltu $24,$13,$5 +.L_bn_div_3_words_inner_loop: + bnez $24,.L_bn_div_3_words_inner_loop_done + sgeu $1,$14,$12 + seq $25,$13,$5 + and $1,$25 + sltu $15,$12,$10 + addu $5,$6 + subu $13,$15 + subu $12,$10 + sltu $24,$13,$5 + sltu $8,$5,$6 + or $24,$8 + .set noreorder + beqz $1,.L_bn_div_3_words_inner_loop + subu $2,1 + addu $2,1 + .set reorder +.L_bn_div_3_words_inner_loop_done: + .set noreorder + jr $31 + move $4,$2 +.end bn_div_3_words_internal + +.align 5 +.globl bn_div_words +.ent bn_div_words +bn_div_words: + .set noreorder + bnez $6,bn_div_words_internal + li $2,-1 # I would rather signal div-by-zero + # which can be done with 'break 7' + jr $31 + move $4,$2 +.end bn_div_words + +.align 5 +.ent bn_div_words_internal +bn_div_words_internal: + move $3,$0 + bltz $6,.L_bn_div_words_body + move $25,$3 + sll $6,1 + bgtz $6,.-4 + addu $25,1 + + .set reorder + negu $13,$25 + li $14,-1 + sll $14,$13 + and $14,$4 + srl $1,$5,$13 + .set noreorder + beqz $14,.+12 + nop + break 6 # signal overflow + .set reorder + sll $4,$25 + sll $5,$25 + or $4,$1 +.L_bn_div_words_body: + srl $3,$6,4*4 # bits + sgeu $1,$4,$6 + .set noreorder + beqz $1,.+12 + nop + subu $4,$6 + .set reorder + + li $8,-1 + srl $9,$4,4*4 # bits + srl $8,4*4 # q=0xffffffff + beq $3,$9,.L_bn_div_words_skip_div1 + divu $0,$4,$3 + mflo $8 +.L_bn_div_words_skip_div1: + multu $6,$8 + sll $15,$4,4*4 # bits + srl $1,$5,4*4 # bits + or $15,$1 + mflo $12 + mfhi $13 +.L_bn_div_words_inner_loop1: + sltu $14,$15,$12 + seq $24,$9,$13 + sltu $1,$9,$13 + and $14,$24 + sltu $2,$12,$6 + or $1,$14 + .set noreorder + beqz $1,.L_bn_div_words_inner_loop1_done + subu $13,$2 + subu $12,$6 + b .L_bn_div_words_inner_loop1 + subu $8,1 + .set reorder +.L_bn_div_words_inner_loop1_done: + + sll $5,4*4 # bits + subu $4,$15,$12 + sll $2,$8,4*4 # bits + + li $8,-1 + srl $9,$4,4*4 # bits + srl $8,4*4 # q=0xffffffff + beq $3,$9,.L_bn_div_words_skip_div2 + divu $0,$4,$3 + mflo $8 +.L_bn_div_words_skip_div2: + multu $6,$8 + sll $15,$4,4*4 # bits + srl $1,$5,4*4 # bits + or $15,$1 + mflo $12 + mfhi $13 +.L_bn_div_words_inner_loop2: + sltu $14,$15,$12 + seq $24,$9,$13 + sltu $1,$9,$13 + and $14,$24 + sltu $3,$12,$6 + or $1,$14 + .set noreorder + beqz $1,.L_bn_div_words_inner_loop2_done + subu $13,$3 + subu $12,$6 + b .L_bn_div_words_inner_loop2 + subu $8,1 + .set reorder +.L_bn_div_words_inner_loop2_done: + + subu $4,$15,$12 + or $2,$8 + srl $3,$4,$25 # $3 contains remainder if anybody wants it + srl $6,$25 # restore $6 + + .set noreorder + move $5,$3 + jr $31 + move $4,$2 +.end bn_div_words_internal + +.align 5 +.globl bn_mul_comba8 +.ent bn_mul_comba8 +bn_mul_comba8: + .set noreorder + .frame $29,6*4,$31 + .mask 0x003f0000,-4 + subu $29,6*4 + sw $21,5*4($29) + sw $20,4*4($29) + sw $19,3*4($29) + sw $18,2*4($29) + sw $17,1*4($29) + sw $16,0*4($29) + + .set reorder + lw $12,0($5) # If compiled with -mips3 option on + # R5000 box assembler barks on this + # 1ine with "should not have mult/div + # as last instruction in bb (R10K + # bug)" warning. If anybody out there + # has a clue about how to circumvent + # this do send me a note. + # + + lw $8,0($6) + lw $13,4($5) + lw $14,2*4($5) + multu $12,$8 # mul_add_c(a[0],b[0],c1,c2,c3); + lw $15,3*4($5) + lw $9,4($6) + lw $10,2*4($6) + lw $11,3*4($6) + mflo $2 + mfhi $3 + + lw $16,4*4($5) + lw $18,5*4($5) + multu $12,$9 # mul_add_c(a[0],b[1],c2,c3,c1); + lw $20,6*4($5) + lw $5,7*4($5) + lw $17,4*4($6) + lw $19,5*4($6) + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $13,$8 # mul_add_c(a[1],b[0],c2,c3,c1); + addu $7,$25,$1 + lw $21,6*4($6) + lw $6,7*4($6) + sw $2,0($4) # r[0]=c1; + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $14,$8 # mul_add_c(a[2],b[0],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $2,$7,$25 + sw $3,4($4) # r[1]=c2; + + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $13,$9 # mul_add_c(a[1],b[1],c3,c1,c2); + addu $25,$1 + addu $2,$25 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $12,$10 # mul_add_c(a[0],b[2],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $3,$2,$25 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $12,$11 # mul_add_c(a[0],b[3],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,2*4($4) # r[2]=c3; + + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $13,$10 # mul_add_c(a[1],b[2],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $7,$3,$25 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $14,$9 # mul_add_c(a[2],b[1],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $15,$8 # mul_add_c(a[3],b[0],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $16,$8 # mul_add_c(a[4],b[0],c2,c3,c1); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,3*4($4) # r[3]=c1; + + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $15,$9 # mul_add_c(a[3],b[1],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $2,$7,$25 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $14,$10 # mul_add_c(a[2],b[2],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $13,$11 # mul_add_c(a[1],b[3],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $12,$17 # mul_add_c(a[0],b[4],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $12,$19 # mul_add_c(a[0],b[5],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,4*4($4) # r[4]=c2; + + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $13,$17 # mul_add_c(a[1],b[4],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $3,$2,$25 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $14,$11 # mul_add_c(a[2],b[3],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $15,$10 # mul_add_c(a[3],b[2],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $16,$9 # mul_add_c(a[4],b[1],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $18,$8 # mul_add_c(a[5],b[0],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $20,$8 # mul_add_c(a[6],b[0],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,5*4($4) # r[5]=c3; + + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $18,$9 # mul_add_c(a[5],b[1],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $7,$3,$25 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $16,$10 # mul_add_c(a[4],b[2],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $15,$11 # mul_add_c(a[3],b[3],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $14,$17 # mul_add_c(a[2],b[4],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $13,$19 # mul_add_c(a[1],b[5],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $12,$21 # mul_add_c(a[0],b[6],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $12,$6 # mul_add_c(a[0],b[7],c2,c3,c1); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,6*4($4) # r[6]=c1; + + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $13,$21 # mul_add_c(a[1],b[6],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $2,$7,$25 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $14,$19 # mul_add_c(a[2],b[5],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $15,$17 # mul_add_c(a[3],b[4],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $16,$11 # mul_add_c(a[4],b[3],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $18,$10 # mul_add_c(a[5],b[2],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $20,$9 # mul_add_c(a[6],b[1],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $5,$8 # mul_add_c(a[7],b[0],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $5,$9 # mul_add_c(a[7],b[1],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,7*4($4) # r[7]=c2; + + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $20,$10 # mul_add_c(a[6],b[2],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $3,$2,$25 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $18,$11 # mul_add_c(a[5],b[3],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $16,$17 # mul_add_c(a[4],b[4],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $15,$19 # mul_add_c(a[3],b[5],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $14,$21 # mul_add_c(a[2],b[6],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $13,$6 # mul_add_c(a[1],b[7],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $14,$6 # mul_add_c(a[2],b[7],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,8*4($4) # r[8]=c3; + + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $15,$21 # mul_add_c(a[3],b[6],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $7,$3,$25 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $16,$19 # mul_add_c(a[4],b[5],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $18,$17 # mul_add_c(a[5],b[4],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $20,$11 # mul_add_c(a[6],b[3],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $5,$10 # mul_add_c(a[7],b[2],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $5,$11 # mul_add_c(a[7],b[3],c2,c3,c1); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,9*4($4) # r[9]=c1; + + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $20,$17 # mul_add_c(a[6],b[4],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $2,$7,$25 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $18,$19 # mul_add_c(a[5],b[5],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $16,$21 # mul_add_c(a[4],b[6],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $15,$6 # mul_add_c(a[3],b[7],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $16,$6 # mul_add_c(a[4],b[7],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,10*4($4) # r[10]=c2; + + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $18,$21 # mul_add_c(a[5],b[6],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $3,$2,$25 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $20,$19 # mul_add_c(a[6],b[5],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $5,$17 # mul_add_c(a[7],b[4],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $5,$19 # mul_add_c(a[7],b[5],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,11*4($4) # r[11]=c3; + + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $20,$21 # mul_add_c(a[6],b[6],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $7,$3,$25 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $18,$6 # mul_add_c(a[5],b[7],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $20,$6 # mul_add_c(a[6],b[7],c2,c3,c1); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,12*4($4) # r[12]=c1; + + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $5,$21 # mul_add_c(a[7],b[6],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $2,$7,$25 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $5,$6 # mul_add_c(a[7],b[7],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,13*4($4) # r[13]=c2; + + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sw $7,14*4($4) # r[14]=c3; + sw $2,15*4($4) # r[15]=c1; + + .set noreorder + lw $21,5*4($29) + lw $20,4*4($29) + lw $19,3*4($29) + lw $18,2*4($29) + lw $17,1*4($29) + lw $16,0*4($29) + jr $31 + addu $29,6*4 +.end bn_mul_comba8 + +.align 5 +.globl bn_mul_comba4 +.ent bn_mul_comba4 +bn_mul_comba4: + .set reorder + lw $12,0($5) + lw $8,0($6) + lw $13,4($5) + lw $14,2*4($5) + multu $12,$8 # mul_add_c(a[0],b[0],c1,c2,c3); + lw $15,3*4($5) + lw $9,4($6) + lw $10,2*4($6) + lw $11,3*4($6) + mflo $2 + mfhi $3 + sw $2,0($4) + + multu $12,$9 # mul_add_c(a[0],b[1],c2,c3,c1); + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $13,$8 # mul_add_c(a[1],b[0],c2,c3,c1); + addu $7,$25,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $14,$8 # mul_add_c(a[2],b[0],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $2,$7,$25 + sw $3,4($4) + + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $13,$9 # mul_add_c(a[1],b[1],c3,c1,c2); + addu $25,$1 + addu $2,$25 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $12,$10 # mul_add_c(a[0],b[2],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $3,$2,$25 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $12,$11 # mul_add_c(a[0],b[3],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,2*4($4) + + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $13,$10 # mul_add_c(a[1],b[2],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $7,$3,$25 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $14,$9 # mul_add_c(a[2],b[1],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $15,$8 # mul_add_c(a[3],b[0],c1,c2,c3); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $15,$9 # mul_add_c(a[3],b[1],c2,c3,c1); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,3*4($4) + + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $14,$10 # mul_add_c(a[2],b[2],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $2,$7,$25 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $13,$11 # mul_add_c(a[1],b[3],c2,c3,c1); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $14,$11 # mul_add_c(a[2],b[3],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,4*4($4) + + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $15,$10 # mul_add_c(a[3],b[2],c3,c1,c2); + addu $25,$1 + addu $2,$25 + sltu $3,$2,$25 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $15,$11 # mul_add_c(a[3],b[3],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,5*4($4) + + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sw $2,6*4($4) + sw $3,7*4($4) + + .set noreorder + jr $31 + nop +.end bn_mul_comba4 + +.align 5 +.globl bn_sqr_comba8 +.ent bn_sqr_comba8 +bn_sqr_comba8: + .set reorder + lw $12,0($5) + lw $13,4($5) + lw $14,2*4($5) + lw $15,3*4($5) + + multu $12,$12 # mul_add_c(a[0],b[0],c1,c2,c3); + lw $8,4*4($5) + lw $9,5*4($5) + lw $10,6*4($5) + lw $11,7*4($5) + mflo $2 + mfhi $3 + sw $2,0($4) + + multu $12,$13 # mul_add_c2(a[0],b[1],c2,c3,c1); + mflo $24 + mfhi $25 + slt $2,$25,$0 + sll $25,1 + multu $14,$12 # mul_add_c2(a[2],b[0],c3,c1,c2); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $7,$25,$1 + sw $3,4($4) + + mflo $24 + mfhi $25 + slt $3,$25,$0 + sll $25,1 + multu $13,$13 # mul_add_c(a[1],b[1],c3,c1,c2); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $12,$15 # mul_add_c2(a[0],b[3],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,2*4($4) + + mflo $24 + mfhi $25 + slt $7,$25,$0 + sll $25,1 + multu $13,$14 # mul_add_c2(a[1],b[2],c1,c2,c3); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $7,$1 + multu $8,$12 # mul_add_c2(a[4],b[0],c2,c3,c1); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,3*4($4) + + mflo $24 + mfhi $25 + slt $2,$25,$0 + sll $25,1 + multu $15,$13 # mul_add_c2(a[3],b[1],c2,c3,c1); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $2,$1 + multu $14,$14 # mul_add_c(a[2],b[2],c2,c3,c1); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $12,$9 # mul_add_c2(a[0],b[5],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,4*4($4) + + mflo $24 + mfhi $25 + slt $3,$25,$0 + sll $25,1 + multu $13,$8 # mul_add_c2(a[1],b[4],c3,c1,c2); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $3,$1 + multu $14,$15 # mul_add_c2(a[2],b[3],c3,c1,c2); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + multu $10,$12 # mul_add_c2(a[6],b[0],c1,c2,c3); + addu $3,$1 + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,5*4($4) + + mflo $24 + mfhi $25 + slt $7,$25,$0 + sll $25,1 + multu $9,$13 # mul_add_c2(a[5],b[1],c1,c2,c3); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $7,$1 + multu $8,$14 # mul_add_c2(a[4],b[2],c1,c2,c3); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $7,$1 + multu $15,$15 # mul_add_c(a[3],b[3],c1,c2,c3); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $12,$11 # mul_add_c2(a[0],b[7],c2,c3,c1); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,6*4($4) + + mflo $24 + mfhi $25 + slt $2,$25,$0 + sll $25,1 + multu $13,$10 # mul_add_c2(a[1],b[6],c2,c3,c1); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $2,$1 + multu $14,$9 # mul_add_c2(a[2],b[5],c2,c3,c1); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $2,$1 + multu $15,$8 # mul_add_c2(a[3],b[4],c2,c3,c1); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $2,$1 + multu $11,$13 # mul_add_c2(a[7],b[1],c3,c1,c2); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,7*4($4) + + mflo $24 + mfhi $25 + slt $3,$25,$0 + sll $25,1 + multu $10,$14 # mul_add_c2(a[6],b[2],c3,c1,c2); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $3,$1 + multu $9,$15 # mul_add_c2(a[5],b[3],c3,c1,c2); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $3,$1 + multu $8,$8 # mul_add_c(a[4],b[4],c3,c1,c2); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $14,$11 # mul_add_c2(a[2],b[7],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,8*4($4) + + mflo $24 + mfhi $25 + slt $7,$25,$0 + sll $25,1 + multu $15,$10 # mul_add_c2(a[3],b[6],c1,c2,c3); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $7,$1 + multu $8,$9 # mul_add_c2(a[4],b[5],c1,c2,c3); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $7,$1 + multu $11,$15 # mul_add_c2(a[7],b[3],c2,c3,c1); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,9*4($4) + + mflo $24 + mfhi $25 + slt $2,$25,$0 + sll $25,1 + multu $10,$8 # mul_add_c2(a[6],b[4],c2,c3,c1); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $2,$1 + multu $9,$9 # mul_add_c(a[5],b[5],c2,c3,c1); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $8,$11 # mul_add_c2(a[4],b[7],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,10*4($4) + + mflo $24 + mfhi $25 + slt $3,$25,$0 + sll $25,1 + multu $9,$10 # mul_add_c2(a[5],b[6],c3,c1,c2); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $3,$1 + multu $11,$9 # mul_add_c2(a[7],b[5],c1,c2,c3); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,11*4($4) + + mflo $24 + mfhi $25 + slt $7,$25,$0 + sll $25,1 + multu $10,$10 # mul_add_c(a[6],b[6],c1,c2,c3); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + multu $10,$11 # mul_add_c2(a[6],b[7],c2,c3,c1); + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,12*4($4) + + mflo $24 + mfhi $25 + slt $2,$25,$0 + sll $25,1 + multu $11,$11 # mul_add_c(a[7],b[7],c3,c1,c2); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,13*4($4) + + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sw $7,14*4($4) + sw $2,15*4($4) + + .set noreorder + jr $31 + nop +.end bn_sqr_comba8 + +.align 5 +.globl bn_sqr_comba4 +.ent bn_sqr_comba4 +bn_sqr_comba4: + .set reorder + lw $12,0($5) + lw $13,4($5) + multu $12,$12 # mul_add_c(a[0],b[0],c1,c2,c3); + lw $14,2*4($5) + lw $15,3*4($5) + mflo $2 + mfhi $3 + sw $2,0($4) + + multu $12,$13 # mul_add_c2(a[0],b[1],c2,c3,c1); + mflo $24 + mfhi $25 + slt $2,$25,$0 + sll $25,1 + multu $14,$12 # mul_add_c2(a[2],b[0],c3,c1,c2); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $7,$25,$1 + sw $3,4($4) + + mflo $24 + mfhi $25 + slt $3,$25,$0 + sll $25,1 + multu $13,$13 # mul_add_c(a[1],b[1],c3,c1,c2); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + mflo $24 + mfhi $25 + addu $7,$24 + sltu $1,$7,$24 + multu $12,$15 # mul_add_c2(a[0],b[3],c1,c2,c3); + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,2*4($4) + + mflo $24 + mfhi $25 + slt $7,$25,$0 + sll $25,1 + multu $13,$14 # mul_add_c(a2[1],b[2],c1,c2,c3); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + mflo $24 + mfhi $25 + slt $1,$25,$0 + addu $7,$1 + multu $15,$13 # mul_add_c2(a[3],b[1],c2,c3,c1); + sll $25,1 + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sltu $1,$3,$25 + addu $7,$1 + sw $2,3*4($4) + + mflo $24 + mfhi $25 + slt $2,$25,$0 + sll $25,1 + multu $14,$14 # mul_add_c(a[2],b[2],c2,c3,c1); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $3,$24 + sltu $1,$3,$24 + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + mflo $24 + mfhi $25 + addu $3,$24 + sltu $1,$3,$24 + multu $14,$15 # mul_add_c2(a[2],b[3],c3,c1,c2); + addu $25,$1 + addu $7,$25 + sltu $1,$7,$25 + addu $2,$1 + sw $3,4*4($4) + + mflo $24 + mfhi $25 + slt $3,$25,$0 + sll $25,1 + multu $15,$15 # mul_add_c(a[3],b[3],c1,c2,c3); + slt $6,$24,$0 + addu $25,$6 + sll $24,1 + addu $7,$24 + sltu $1,$7,$24 + addu $25,$1 + addu $2,$25 + sltu $1,$2,$25 + addu $3,$1 + sw $7,5*4($4) + + mflo $24 + mfhi $25 + addu $2,$24 + sltu $1,$2,$24 + addu $25,$1 + addu $3,$25 + sw $2,6*4($4) + sw $3,7*4($4) + + .set noreorder + jr $31 + nop +.end bn_sqr_comba4 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/co-586.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/co-586.S new file mode 100644 index 00000000..3cb80735 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/co-586.S @@ -0,0 +1,1254 @@ +.file "crypto/bn/asm/co-586.s" +.text +.globl bn_mul_comba8 +.type bn_mul_comba8,@function +.align 16 +bn_mul_comba8: +.L_bn_mul_comba8_begin: + pushl %esi + movl 12(%esp),%esi + pushl %edi + movl 20(%esp),%edi + pushl %ebp + pushl %ebx + xorl %ebx,%ebx + movl (%esi),%eax + xorl %ecx,%ecx + movl (%edi),%edx + + xorl %ebp,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esp),%eax + adcl %edx,%ecx + movl (%edi),%edx + adcl $0,%ebp + movl %ebx,(%eax) + movl 4(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%ecx + movl (%esi),%eax + adcl %edx,%ebp + movl 4(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esp),%eax + adcl %edx,%ebp + movl (%edi),%edx + adcl $0,%ebx + movl %ecx,4(%eax) + movl 8(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%ebp + movl 4(%esi),%eax + adcl %edx,%ebx + movl 4(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl (%esi),%eax + adcl %edx,%ebx + movl 8(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esp),%eax + adcl %edx,%ebx + movl (%edi),%edx + adcl $0,%ecx + movl %ebp,8(%eax) + movl 12(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%ebx + movl 8(%esi),%eax + adcl %edx,%ecx + movl 4(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 4(%esi),%eax + adcl %edx,%ecx + movl 8(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl (%esi),%eax + adcl %edx,%ecx + movl 12(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esp),%eax + adcl %edx,%ecx + movl (%edi),%edx + adcl $0,%ebp + movl %ebx,12(%eax) + movl 16(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%ecx + movl 12(%esi),%eax + adcl %edx,%ebp + movl 4(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 8(%esi),%eax + adcl %edx,%ebp + movl 8(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 4(%esi),%eax + adcl %edx,%ebp + movl 12(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl (%esi),%eax + adcl %edx,%ebp + movl 16(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esp),%eax + adcl %edx,%ebp + movl (%edi),%edx + adcl $0,%ebx + movl %ecx,16(%eax) + movl 20(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%ebp + movl 16(%esi),%eax + adcl %edx,%ebx + movl 4(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 12(%esi),%eax + adcl %edx,%ebx + movl 8(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 8(%esi),%eax + adcl %edx,%ebx + movl 12(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 4(%esi),%eax + adcl %edx,%ebx + movl 16(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl (%esi),%eax + adcl %edx,%ebx + movl 20(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esp),%eax + adcl %edx,%ebx + movl (%edi),%edx + adcl $0,%ecx + movl %ebp,20(%eax) + movl 24(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esi),%eax + adcl %edx,%ecx + movl 4(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 16(%esi),%eax + adcl %edx,%ecx + movl 8(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 12(%esi),%eax + adcl %edx,%ecx + movl 12(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 8(%esi),%eax + adcl %edx,%ecx + movl 16(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 4(%esi),%eax + adcl %edx,%ecx + movl 20(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl (%esi),%eax + adcl %edx,%ecx + movl 24(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esp),%eax + adcl %edx,%ecx + movl (%edi),%edx + adcl $0,%ebp + movl %ebx,24(%eax) + movl 28(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%ecx + movl 24(%esi),%eax + adcl %edx,%ebp + movl 4(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esi),%eax + adcl %edx,%ebp + movl 8(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 16(%esi),%eax + adcl %edx,%ebp + movl 12(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 12(%esi),%eax + adcl %edx,%ebp + movl 16(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 8(%esi),%eax + adcl %edx,%ebp + movl 20(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 4(%esi),%eax + adcl %edx,%ebp + movl 24(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl (%esi),%eax + adcl %edx,%ebp + movl 28(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esp),%eax + adcl %edx,%ebp + movl 4(%edi),%edx + adcl $0,%ebx + movl %ecx,28(%eax) + movl 28(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%ebp + movl 24(%esi),%eax + adcl %edx,%ebx + movl 8(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esi),%eax + adcl %edx,%ebx + movl 12(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 16(%esi),%eax + adcl %edx,%ebx + movl 16(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 12(%esi),%eax + adcl %edx,%ebx + movl 20(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 8(%esi),%eax + adcl %edx,%ebx + movl 24(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 4(%esi),%eax + adcl %edx,%ebx + movl 28(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esp),%eax + adcl %edx,%ebx + movl 8(%edi),%edx + adcl $0,%ecx + movl %ebp,32(%eax) + movl 28(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%ebx + movl 24(%esi),%eax + adcl %edx,%ecx + movl 12(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esi),%eax + adcl %edx,%ecx + movl 16(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 16(%esi),%eax + adcl %edx,%ecx + movl 20(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 12(%esi),%eax + adcl %edx,%ecx + movl 24(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 8(%esi),%eax + adcl %edx,%ecx + movl 28(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esp),%eax + adcl %edx,%ecx + movl 12(%edi),%edx + adcl $0,%ebp + movl %ebx,36(%eax) + movl 28(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%ecx + movl 24(%esi),%eax + adcl %edx,%ebp + movl 16(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esi),%eax + adcl %edx,%ebp + movl 20(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 16(%esi),%eax + adcl %edx,%ebp + movl 24(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 12(%esi),%eax + adcl %edx,%ebp + movl 28(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esp),%eax + adcl %edx,%ebp + movl 16(%edi),%edx + adcl $0,%ebx + movl %ecx,40(%eax) + movl 28(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%ebp + movl 24(%esi),%eax + adcl %edx,%ebx + movl 20(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esi),%eax + adcl %edx,%ebx + movl 24(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 16(%esi),%eax + adcl %edx,%ebx + movl 28(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esp),%eax + adcl %edx,%ebx + movl 20(%edi),%edx + adcl $0,%ecx + movl %ebp,44(%eax) + movl 28(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%ebx + movl 24(%esi),%eax + adcl %edx,%ecx + movl 24(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esi),%eax + adcl %edx,%ecx + movl 28(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esp),%eax + adcl %edx,%ecx + movl 24(%edi),%edx + adcl $0,%ebp + movl %ebx,48(%eax) + movl 28(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%ecx + movl 24(%esi),%eax + adcl %edx,%ebp + movl 28(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esp),%eax + adcl %edx,%ebp + movl 28(%edi),%edx + adcl $0,%ebx + movl %ecx,52(%eax) + movl 28(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esp),%eax + adcl %edx,%ebx + adcl $0,%ecx + movl %ebp,56(%eax) + + + movl %ebx,60(%eax) + popl %ebx + popl %ebp + popl %edi + popl %esi + ret +.size bn_mul_comba8,.-.L_bn_mul_comba8_begin +.globl bn_mul_comba4 +.type bn_mul_comba4,@function +.align 16 +bn_mul_comba4: +.L_bn_mul_comba4_begin: + pushl %esi + movl 12(%esp),%esi + pushl %edi + movl 20(%esp),%edi + pushl %ebp + pushl %ebx + xorl %ebx,%ebx + movl (%esi),%eax + xorl %ecx,%ecx + movl (%edi),%edx + + xorl %ebp,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esp),%eax + adcl %edx,%ecx + movl (%edi),%edx + adcl $0,%ebp + movl %ebx,(%eax) + movl 4(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%ecx + movl (%esi),%eax + adcl %edx,%ebp + movl 4(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esp),%eax + adcl %edx,%ebp + movl (%edi),%edx + adcl $0,%ebx + movl %ecx,4(%eax) + movl 8(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%ebp + movl 4(%esi),%eax + adcl %edx,%ebx + movl 4(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl (%esi),%eax + adcl %edx,%ebx + movl 8(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esp),%eax + adcl %edx,%ebx + movl (%edi),%edx + adcl $0,%ecx + movl %ebp,8(%eax) + movl 12(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%ebx + movl 8(%esi),%eax + adcl %edx,%ecx + movl 4(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 4(%esi),%eax + adcl %edx,%ecx + movl 8(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl (%esi),%eax + adcl %edx,%ecx + movl 12(%edi),%edx + adcl $0,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esp),%eax + adcl %edx,%ecx + movl 4(%edi),%edx + adcl $0,%ebp + movl %ebx,12(%eax) + movl 12(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%ecx + movl 8(%esi),%eax + adcl %edx,%ebp + movl 8(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 4(%esi),%eax + adcl %edx,%ebp + movl 12(%edi),%edx + adcl $0,%ebx + + mull %edx + addl %eax,%ecx + movl 20(%esp),%eax + adcl %edx,%ebp + movl 8(%edi),%edx + adcl $0,%ebx + movl %ecx,16(%eax) + movl 12(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%ebp + movl 8(%esi),%eax + adcl %edx,%ebx + movl 12(%edi),%edx + adcl $0,%ecx + + mull %edx + addl %eax,%ebp + movl 20(%esp),%eax + adcl %edx,%ebx + movl 12(%edi),%edx + adcl $0,%ecx + movl %ebp,20(%eax) + movl 12(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%ebx + movl 20(%esp),%eax + adcl %edx,%ecx + adcl $0,%ebp + movl %ebx,24(%eax) + + + movl %ecx,28(%eax) + popl %ebx + popl %ebp + popl %edi + popl %esi + ret +.size bn_mul_comba4,.-.L_bn_mul_comba4_begin +.globl bn_sqr_comba8 +.type bn_sqr_comba8,@function +.align 16 +bn_sqr_comba8: +.L_bn_sqr_comba8_begin: + pushl %esi + pushl %edi + pushl %ebp + pushl %ebx + movl 20(%esp),%edi + movl 24(%esp),%esi + xorl %ebx,%ebx + xorl %ecx,%ecx + movl (%esi),%eax + + xorl %ebp,%ebp + + mull %eax + addl %eax,%ebx + adcl %edx,%ecx + movl (%esi),%edx + adcl $0,%ebp + movl %ebx,(%edi) + movl 4(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 8(%esi),%eax + adcl $0,%ebx + movl %ecx,4(%edi) + movl (%esi),%edx + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 4(%esi),%eax + adcl $0,%ecx + + mull %eax + addl %eax,%ebp + adcl %edx,%ebx + movl (%esi),%edx + adcl $0,%ecx + movl %ebp,8(%edi) + movl 12(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 8(%esi),%eax + adcl $0,%ebp + movl 4(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 16(%esi),%eax + adcl $0,%ebp + movl %ebx,12(%edi) + movl (%esi),%edx + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 12(%esi),%eax + adcl $0,%ebx + movl 4(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 8(%esi),%eax + adcl $0,%ebx + + mull %eax + addl %eax,%ecx + adcl %edx,%ebp + movl (%esi),%edx + adcl $0,%ebx + movl %ecx,16(%edi) + movl 20(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 16(%esi),%eax + adcl $0,%ecx + movl 4(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 12(%esi),%eax + adcl $0,%ecx + movl 8(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 24(%esi),%eax + adcl $0,%ecx + movl %ebp,20(%edi) + movl (%esi),%edx + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 20(%esi),%eax + adcl $0,%ebp + movl 4(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 16(%esi),%eax + adcl $0,%ebp + movl 8(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 12(%esi),%eax + adcl $0,%ebp + + mull %eax + addl %eax,%ebx + adcl %edx,%ecx + movl (%esi),%edx + adcl $0,%ebp + movl %ebx,24(%edi) + movl 28(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 24(%esi),%eax + adcl $0,%ebx + movl 4(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 20(%esi),%eax + adcl $0,%ebx + movl 8(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 16(%esi),%eax + adcl $0,%ebx + movl 12(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 28(%esi),%eax + adcl $0,%ebx + movl %ecx,28(%edi) + movl 4(%esi),%edx + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 24(%esi),%eax + adcl $0,%ecx + movl 8(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 20(%esi),%eax + adcl $0,%ecx + movl 12(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 16(%esi),%eax + adcl $0,%ecx + + mull %eax + addl %eax,%ebp + adcl %edx,%ebx + movl 8(%esi),%edx + adcl $0,%ecx + movl %ebp,32(%edi) + movl 28(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 24(%esi),%eax + adcl $0,%ebp + movl 12(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 20(%esi),%eax + adcl $0,%ebp + movl 16(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 28(%esi),%eax + adcl $0,%ebp + movl %ebx,36(%edi) + movl 12(%esi),%edx + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 24(%esi),%eax + adcl $0,%ebx + movl 16(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 20(%esi),%eax + adcl $0,%ebx + + mull %eax + addl %eax,%ecx + adcl %edx,%ebp + movl 16(%esi),%edx + adcl $0,%ebx + movl %ecx,40(%edi) + movl 28(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 24(%esi),%eax + adcl $0,%ecx + movl 20(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 28(%esi),%eax + adcl $0,%ecx + movl %ebp,44(%edi) + movl 20(%esi),%edx + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 24(%esi),%eax + adcl $0,%ebp + + mull %eax + addl %eax,%ebx + adcl %edx,%ecx + movl 24(%esi),%edx + adcl $0,%ebp + movl %ebx,48(%edi) + movl 28(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 28(%esi),%eax + adcl $0,%ebx + movl %ecx,52(%edi) + + + xorl %ecx,%ecx + + mull %eax + addl %eax,%ebp + adcl %edx,%ebx + adcl $0,%ecx + movl %ebp,56(%edi) + + movl %ebx,60(%edi) + popl %ebx + popl %ebp + popl %edi + popl %esi + ret +.size bn_sqr_comba8,.-.L_bn_sqr_comba8_begin +.globl bn_sqr_comba4 +.type bn_sqr_comba4,@function +.align 16 +bn_sqr_comba4: +.L_bn_sqr_comba4_begin: + pushl %esi + pushl %edi + pushl %ebp + pushl %ebx + movl 20(%esp),%edi + movl 24(%esp),%esi + xorl %ebx,%ebx + xorl %ecx,%ecx + movl (%esi),%eax + + xorl %ebp,%ebp + + mull %eax + addl %eax,%ebx + adcl %edx,%ecx + movl (%esi),%edx + adcl $0,%ebp + movl %ebx,(%edi) + movl 4(%esi),%eax + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 8(%esi),%eax + adcl $0,%ebx + movl %ecx,4(%edi) + movl (%esi),%edx + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 4(%esi),%eax + adcl $0,%ecx + + mull %eax + addl %eax,%ebp + adcl %edx,%ebx + movl (%esi),%edx + adcl $0,%ecx + movl %ebp,8(%edi) + movl 12(%esi),%eax + + + xorl %ebp,%ebp + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 8(%esi),%eax + adcl $0,%ebp + movl 4(%esi),%edx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebp + addl %eax,%ebx + adcl %edx,%ecx + movl 12(%esi),%eax + adcl $0,%ebp + movl %ebx,12(%edi) + movl 4(%esi),%edx + + + xorl %ebx,%ebx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ebx + addl %eax,%ecx + adcl %edx,%ebp + movl 8(%esi),%eax + adcl $0,%ebx + + mull %eax + addl %eax,%ecx + adcl %edx,%ebp + movl 8(%esi),%edx + adcl $0,%ebx + movl %ecx,16(%edi) + movl 12(%esi),%eax + + + xorl %ecx,%ecx + + mull %edx + addl %eax,%eax + adcl %edx,%edx + adcl $0,%ecx + addl %eax,%ebp + adcl %edx,%ebx + movl 12(%esi),%eax + adcl $0,%ecx + movl %ebp,20(%edi) + + + xorl %ebp,%ebp + + mull %eax + addl %eax,%ebx + adcl %edx,%ecx + adcl $0,%ebp + movl %ebx,24(%edi) + + movl %ecx,28(%edi) + popl %ebx + popl %ebp + popl %edi + popl %esi + ret +.size bn_sqr_comba4,.-.L_bn_sqr_comba4_begin diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/co-586.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/co-586.pl new file mode 100644 index 00000000..57101a6b --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/co-586.pl @@ -0,0 +1,287 @@ +#!/usr/local/bin/perl + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "x86asm.pl"; + +&asm_init($ARGV[0],$0); + +&bn_mul_comba("bn_mul_comba8",8); +&bn_mul_comba("bn_mul_comba4",4); +&bn_sqr_comba("bn_sqr_comba8",8); +&bn_sqr_comba("bn_sqr_comba4",4); + +&asm_finish(); + +sub mul_add_c + { + local($a,$ai,$b,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_; + + # pos == -1 if eax and edx are pre-loaded, 0 to load from next + # words, and 1 if load return value + + &comment("mul a[$ai]*b[$bi]"); + + # "eax" and "edx" will always be pre-loaded. + # &mov("eax",&DWP($ai*4,$a,"",0)) ; + # &mov("edx",&DWP($bi*4,$b,"",0)); + + &mul("edx"); + &add($c0,"eax"); + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # laod next a + &mov("eax",&wparam(0)) if $pos > 0; # load r[] + ### + &adc($c1,"edx"); + &mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 0; # laod next b + &mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 1; # laod next b + ### + &adc($c2,0); + # is pos > 1, it means it is the last loop + &mov(&DWP($i*4,"eax","",0),$c0) if $pos > 0; # save r[]; + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # laod next a + } + +sub sqr_add_c + { + local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_; + + # pos == -1 if eax and edx are pre-loaded, 0 to load from next + # words, and 1 if load return value + + &comment("sqr a[$ai]*a[$bi]"); + + # "eax" and "edx" will always be pre-loaded. + # &mov("eax",&DWP($ai*4,$a,"",0)) ; + # &mov("edx",&DWP($bi*4,$b,"",0)); + + if ($ai == $bi) + { &mul("eax");} + else + { &mul("edx");} + &add($c0,"eax"); + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a + ### + &adc($c1,"edx"); + &mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos == 1) && ($na != $nb); + ### + &adc($c2,0); + # is pos > 1, it means it is the last loop + &mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[]; + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b + } + +sub sqr_add_c2 + { + local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_; + + # pos == -1 if eax and edx are pre-loaded, 0 to load from next + # words, and 1 if load return value + + &comment("sqr a[$ai]*a[$bi]"); + + # "eax" and "edx" will always be pre-loaded. + # &mov("eax",&DWP($ai*4,$a,"",0)) ; + # &mov("edx",&DWP($bi*4,$a,"",0)); + + if ($ai == $bi) + { &mul("eax");} + else + { &mul("edx");} + &add("eax","eax"); + ### + &adc("edx","edx"); + ### + &adc($c2,0); + &add($c0,"eax"); + &adc($c1,"edx"); + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b + &adc($c2,0); + &mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[]; + &mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos <= 1) && ($na != $nb); + ### + } + +sub bn_mul_comba + { + local($name,$num)=@_; + local($a,$b,$c0,$c1,$c2); + local($i,$as,$ae,$bs,$be,$ai,$bi); + local($tot,$end); + + &function_begin_B($name,""); + + $c0="ebx"; + $c1="ecx"; + $c2="ebp"; + $a="esi"; + $b="edi"; + + $as=0; + $ae=0; + $bs=0; + $be=0; + $tot=$num+$num-1; + + &push("esi"); + &mov($a,&wparam(1)); + &push("edi"); + &mov($b,&wparam(2)); + &push("ebp"); + &push("ebx"); + + &xor($c0,$c0); + &mov("eax",&DWP(0,$a,"",0)); # load the first word + &xor($c1,$c1); + &mov("edx",&DWP(0,$b,"",0)); # load the first second + + for ($i=0; $i<$tot; $i++) + { + $ai=$as; + $bi=$bs; + $end=$be+1; + + &comment("################## Calculate word $i"); + + for ($j=$bs; $j<$end; $j++) + { + &xor($c2,$c2) if ($j == $bs); + if (($j+1) == $end) + { + $v=1; + $v=2 if (($i+1) == $tot); + } + else + { $v=0; } + if (($j+1) != $end) + { + $na=($ai-1); + $nb=($bi+1); + } + else + { + $na=$as+($i < ($num-1)); + $nb=$bs+($i >= ($num-1)); + } +#printf STDERR "[$ai,$bi] -> [$na,$nb]\n"; + &mul_add_c($a,$ai,$b,$bi,$c0,$c1,$c2,$v,$i,$na,$nb); + if ($v) + { + &comment("saved r[$i]"); + # &mov("eax",&wparam(0)); + # &mov(&DWP($i*4,"eax","",0),$c0); + ($c0,$c1,$c2)=($c1,$c2,$c0); + } + $ai--; + $bi++; + } + $as++ if ($i < ($num-1)); + $ae++ if ($i >= ($num-1)); + + $bs++ if ($i >= ($num-1)); + $be++ if ($i < ($num-1)); + } + &comment("save r[$i]"); + # &mov("eax",&wparam(0)); + &mov(&DWP($i*4,"eax","",0),$c0); + + &pop("ebx"); + &pop("ebp"); + &pop("edi"); + &pop("esi"); + &ret(); + &function_end_B($name); + } + +sub bn_sqr_comba + { + local($name,$num)=@_; + local($r,$a,$c0,$c1,$c2)=@_; + local($i,$as,$ae,$bs,$be,$ai,$bi); + local($b,$tot,$end,$half); + + &function_begin_B($name,""); + + $c0="ebx"; + $c1="ecx"; + $c2="ebp"; + $a="esi"; + $r="edi"; + + &push("esi"); + &push("edi"); + &push("ebp"); + &push("ebx"); + &mov($r,&wparam(0)); + &mov($a,&wparam(1)); + &xor($c0,$c0); + &xor($c1,$c1); + &mov("eax",&DWP(0,$a,"",0)); # load the first word + + $as=0; + $ae=0; + $bs=0; + $be=0; + $tot=$num+$num-1; + + for ($i=0; $i<$tot; $i++) + { + $ai=$as; + $bi=$bs; + $end=$be+1; + + &comment("############### Calculate word $i"); + for ($j=$bs; $j<$end; $j++) + { + &xor($c2,$c2) if ($j == $bs); + if (($ai-1) < ($bi+1)) + { + $v=1; + $v=2 if ($i+1) == $tot; + } + else + { $v=0; } + if (!$v) + { + $na=$ai-1; + $nb=$bi+1; + } + else + { + $na=$as+($i < ($num-1)); + $nb=$bs+($i >= ($num-1)); + } + if ($ai == $bi) + { + &sqr_add_c($r,$a,$ai,$bi, + $c0,$c1,$c2,$v,$i,$na,$nb); + } + else + { + &sqr_add_c2($r,$a,$ai,$bi, + $c0,$c1,$c2,$v,$i,$na,$nb); + } + if ($v) + { + &comment("saved r[$i]"); + #&mov(&DWP($i*4,$r,"",0),$c0); + ($c0,$c1,$c2)=($c1,$c2,$c0); + last; + } + $ai--; + $bi++; + } + $as++ if ($i < ($num-1)); + $ae++ if ($i >= ($num-1)); + + $bs++ if ($i >= ($num-1)); + $be++ if ($i < ($num-1)); + } + &mov(&DWP($i*4,$r,"",0),$c0); + &pop("ebx"); + &pop("ebp"); + &pop("edi"); + &pop("esi"); + &ret(); + &function_end_B($name); + } diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ia64-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ia64-mont.pl new file mode 100644 index 00000000..e2586584 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ia64-mont.pl @@ -0,0 +1,851 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# January 2010 +# +# "Teaser" Montgomery multiplication module for IA-64. There are +# several possibilities for improvement: +# +# - modulo-scheduling outer loop would eliminate quite a number of +# stalls after ldf8, xma and getf.sig outside inner loop and +# improve shorter key performance; +# - shorter vector support [with input vectors being fetched only +# once] should be added; +# - 2x unroll with help of n0[1] would make the code scalable on +# "wider" IA-64, "wider" than Itanium 2 that is, which is not of +# acute interest, because upcoming Tukwila's individual cores are +# reportedly based on Itanium 2 design; +# - dedicated squaring procedure(?); +# +# January 2010 +# +# Shorter vector support is implemented by zero-padding ap and np +# vectors up to 8 elements, or 512 bits. This means that 256-bit +# inputs will be processed only 2 times faster than 512-bit inputs, +# not 4 [as one would expect, because algorithm complexity is n^2]. +# The reason for padding is that inputs shorter than 512 bits won't +# be processed faster anyway, because minimal critical path of the +# core loop happens to match 512-bit timing. Either way, it resulted +# in >100% improvement of 512-bit RSA sign benchmark and 50% - of +# 1024-bit one [in comparison to original version of *this* module]. +# +# So far 'openssl speed rsa dsa' output on 900MHz Itanium 2 *with* +# this module is: +# sign verify sign/s verify/s +# rsa 512 bits 0.000290s 0.000024s 3452.8 42031.4 +# rsa 1024 bits 0.000793s 0.000058s 1261.7 17172.0 +# rsa 2048 bits 0.005908s 0.000148s 169.3 6754.0 +# rsa 4096 bits 0.033456s 0.000469s 29.9 2133.6 +# dsa 512 bits 0.000253s 0.000198s 3949.9 5057.0 +# dsa 1024 bits 0.000585s 0.000607s 1708.4 1647.4 +# dsa 2048 bits 0.001453s 0.001703s 688.1 587.4 +# +# ... and *without* (but still with ia64.S): +# +# rsa 512 bits 0.000670s 0.000041s 1491.8 24145.5 +# rsa 1024 bits 0.001988s 0.000080s 502.9 12499.3 +# rsa 2048 bits 0.008702s 0.000189s 114.9 5293.9 +# rsa 4096 bits 0.043860s 0.000533s 22.8 1875.9 +# dsa 512 bits 0.000441s 0.000427s 2265.3 2340.6 +# dsa 1024 bits 0.000823s 0.000867s 1215.6 1153.2 +# dsa 2048 bits 0.001894s 0.002179s 528.1 458.9 +# +# As it can be seen, RSA sign performance improves by 130-30%, +# hereafter less for longer keys, while verify - by 74-13%. +# DSA performance improves by 115-30%. + +if ($^O eq "hpux") { + $ADDP="addp4"; + for (@ARGV) { $ADDP="add" if (/[\+DD|\-mlp]64/); } +} else { $ADDP="add"; } + +$code=<<___; +.explicit +.text + +// int bn_mul_mont (BN_ULONG *rp,const BN_ULONG *ap, +// const BN_ULONG *bp,const BN_ULONG *np, +// const BN_ULONG *n0p,int num); +.align 64 +.global bn_mul_mont# +.proc bn_mul_mont# +bn_mul_mont: + .prologue + .body +{ .mmi; cmp4.le p6,p7=2,r37;; +(p6) cmp4.lt.unc p8,p9=8,r37 + mov ret0=r0 };; +{ .bbb; +(p9) br.cond.dptk.many bn_mul_mont_8 +(p8) br.cond.dpnt.many bn_mul_mont_general +(p7) br.ret.spnt.many b0 };; +.endp bn_mul_mont# + +prevfs=r2; prevpr=r3; prevlc=r10; prevsp=r11; + +rptr=r8; aptr=r9; bptr=r14; nptr=r15; +tptr=r16; // &tp[0] +tp_1=r17; // &tp[-1] +num=r18; len=r19; lc=r20; +topbit=r21; // carry bit from tmp[num] + +n0=f6; +m0=f7; +bi=f8; + +.align 64 +.local bn_mul_mont_general# +.proc bn_mul_mont_general# +bn_mul_mont_general: + .prologue +{ .mmi; .save ar.pfs,prevfs + alloc prevfs=ar.pfs,6,2,0,8 + $ADDP aptr=0,in1 + .save ar.lc,prevlc + mov prevlc=ar.lc } +{ .mmi; .vframe prevsp + mov prevsp=sp + $ADDP bptr=0,in2 + .save pr,prevpr + mov prevpr=pr };; + + .body + .rotf alo[6],nlo[4],ahi[8],nhi[6] + .rotr a[3],n[3],t[2] + +{ .mmi; ldf8 bi=[bptr],8 // (*bp++) + ldf8 alo[4]=[aptr],16 // ap[0] + $ADDP r30=8,in1 };; +{ .mmi; ldf8 alo[3]=[r30],16 // ap[1] + ldf8 alo[2]=[aptr],16 // ap[2] + $ADDP in4=0,in4 };; +{ .mmi; ldf8 alo[1]=[r30] // ap[3] + ldf8 n0=[in4] // n0 + $ADDP rptr=0,in0 } +{ .mmi; $ADDP nptr=0,in3 + mov r31=16 + zxt4 num=in5 };; +{ .mmi; ldf8 nlo[2]=[nptr],8 // np[0] + shladd len=num,3,r0 + shladd r31=num,3,r31 };; +{ .mmi; ldf8 nlo[1]=[nptr],8 // np[1] + add lc=-5,num + sub r31=sp,r31 };; +{ .mfb; and sp=-16,r31 // alloca + xmpy.hu ahi[2]=alo[4],bi // ap[0]*bp[0] + nop.b 0 } +{ .mfb; nop.m 0 + xmpy.lu alo[4]=alo[4],bi + brp.loop.imp .L1st_ctop,.L1st_cend-16 + };; +{ .mfi; nop.m 0 + xma.hu ahi[1]=alo[3],bi,ahi[2] // ap[1]*bp[0] + add tp_1=8,sp } +{ .mfi; nop.m 0 + xma.lu alo[3]=alo[3],bi,ahi[2] + mov pr.rot=0x20001f<<16 + // ------^----- (p40) at first (p23) + // ----------^^ p[16:20]=1 + };; +{ .mfi; nop.m 0 + xmpy.lu m0=alo[4],n0 // (ap[0]*bp[0])*n0 + mov ar.lc=lc } +{ .mfi; nop.m 0 + fcvt.fxu.s1 nhi[1]=f0 + mov ar.ec=8 };; + +.align 32 +.L1st_ctop: +.pred.rel "mutex",p40,p42 +{ .mfi; (p16) ldf8 alo[0]=[aptr],8 // *(aptr++) + (p18) xma.hu ahi[0]=alo[2],bi,ahi[1] + (p40) add n[2]=n[2],a[2] } // (p23) } +{ .mfi; (p18) ldf8 nlo[0]=[nptr],8 // *(nptr++)(p16) + (p18) xma.lu alo[2]=alo[2],bi,ahi[1] + (p42) add n[2]=n[2],a[2],1 };; // (p23) +{ .mfi; (p21) getf.sig a[0]=alo[5] + (p20) xma.hu nhi[0]=nlo[2],m0,nhi[1] + (p42) cmp.leu p41,p39=n[2],a[2] } // (p23) +{ .mfi; (p23) st8 [tp_1]=n[2],8 + (p20) xma.lu nlo[2]=nlo[2],m0,nhi[1] + (p40) cmp.ltu p41,p39=n[2],a[2] } // (p23) +{ .mmb; (p21) getf.sig n[0]=nlo[3] + (p16) nop.m 0 + br.ctop.sptk .L1st_ctop };; +.L1st_cend: + +{ .mmi; getf.sig a[0]=ahi[6] // (p24) + getf.sig n[0]=nhi[4] + add num=-1,num };; // num-- +{ .mmi; .pred.rel "mutex",p40,p42 +(p40) add n[0]=n[0],a[0] +(p42) add n[0]=n[0],a[0],1 + sub aptr=aptr,len };; // rewind +{ .mmi; .pred.rel "mutex",p40,p42 +(p40) cmp.ltu p41,p39=n[0],a[0] +(p42) cmp.leu p41,p39=n[0],a[0] + sub nptr=nptr,len };; +{ .mmi; .pred.rel "mutex",p39,p41 +(p39) add topbit=r0,r0 +(p41) add topbit=r0,r0,1 + nop.i 0 } +{ .mmi; st8 [tp_1]=n[0] + add tptr=16,sp + add tp_1=8,sp };; + +.Louter: +{ .mmi; ldf8 bi=[bptr],8 // (*bp++) + ldf8 ahi[3]=[tptr] // tp[0] + add r30=8,aptr };; +{ .mmi; ldf8 alo[4]=[aptr],16 // ap[0] + ldf8 alo[3]=[r30],16 // ap[1] + add r31=8,nptr };; +{ .mfb; ldf8 alo[2]=[aptr],16 // ap[2] + xma.hu ahi[2]=alo[4],bi,ahi[3] // ap[0]*bp[i]+tp[0] + brp.loop.imp .Linner_ctop,.Linner_cend-16 + } +{ .mfb; ldf8 alo[1]=[r30] // ap[3] + xma.lu alo[4]=alo[4],bi,ahi[3] + clrrrb.pr };; +{ .mfi; ldf8 nlo[2]=[nptr],16 // np[0] + xma.hu ahi[1]=alo[3],bi,ahi[2] // ap[1]*bp[i] + nop.i 0 } +{ .mfi; ldf8 nlo[1]=[r31] // np[1] + xma.lu alo[3]=alo[3],bi,ahi[2] + mov pr.rot=0x20101f<<16 + // ------^----- (p40) at first (p23) + // --------^--- (p30) at first (p22) + // ----------^^ p[16:20]=1 + };; +{ .mfi; st8 [tptr]=r0 // tp[0] is already accounted + xmpy.lu m0=alo[4],n0 // (ap[0]*bp[i]+tp[0])*n0 + mov ar.lc=lc } +{ .mfi; + fcvt.fxu.s1 nhi[1]=f0 + mov ar.ec=8 };; + +// This loop spins in 4*(n+7) ticks on Itanium 2 and should spin in +// 7*(n+7) ticks on Itanium (the one codenamed Merced). Factor of 7 +// in latter case accounts for two-tick pipeline stall, which means +// that its performance would be ~20% lower than optimal one. No +// attempt was made to address this, because original Itanium is +// hardly represented out in the wild... +.align 32 +.Linner_ctop: +.pred.rel "mutex",p40,p42 +.pred.rel "mutex",p30,p32 +{ .mfi; (p16) ldf8 alo[0]=[aptr],8 // *(aptr++) + (p18) xma.hu ahi[0]=alo[2],bi,ahi[1] + (p40) add n[2]=n[2],a[2] } // (p23) +{ .mfi; (p16) nop.m 0 + (p18) xma.lu alo[2]=alo[2],bi,ahi[1] + (p42) add n[2]=n[2],a[2],1 };; // (p23) +{ .mfi; (p21) getf.sig a[0]=alo[5] + (p16) nop.f 0 + (p40) cmp.ltu p41,p39=n[2],a[2] } // (p23) +{ .mfi; (p21) ld8 t[0]=[tptr],8 + (p16) nop.f 0 + (p42) cmp.leu p41,p39=n[2],a[2] };; // (p23) +{ .mfi; (p18) ldf8 nlo[0]=[nptr],8 // *(nptr++) + (p20) xma.hu nhi[0]=nlo[2],m0,nhi[1] + (p30) add a[1]=a[1],t[1] } // (p22) +{ .mfi; (p16) nop.m 0 + (p20) xma.lu nlo[2]=nlo[2],m0,nhi[1] + (p32) add a[1]=a[1],t[1],1 };; // (p22) +{ .mmi; (p21) getf.sig n[0]=nlo[3] + (p16) nop.m 0 + (p30) cmp.ltu p31,p29=a[1],t[1] } // (p22) +{ .mmb; (p23) st8 [tp_1]=n[2],8 + (p32) cmp.leu p31,p29=a[1],t[1] // (p22) + br.ctop.sptk .Linner_ctop };; +.Linner_cend: + +{ .mmi; getf.sig a[0]=ahi[6] // (p24) + getf.sig n[0]=nhi[4] + nop.i 0 };; + +{ .mmi; .pred.rel "mutex",p31,p33 +(p31) add a[0]=a[0],topbit +(p33) add a[0]=a[0],topbit,1 + mov topbit=r0 };; +{ .mfi; .pred.rel "mutex",p31,p33 +(p31) cmp.ltu p32,p30=a[0],topbit +(p33) cmp.leu p32,p30=a[0],topbit + } +{ .mfi; .pred.rel "mutex",p40,p42 +(p40) add n[0]=n[0],a[0] +(p42) add n[0]=n[0],a[0],1 + };; +{ .mmi; .pred.rel "mutex",p44,p46 +(p40) cmp.ltu p41,p39=n[0],a[0] +(p42) cmp.leu p41,p39=n[0],a[0] +(p32) add topbit=r0,r0,1 } + +{ .mmi; st8 [tp_1]=n[0],8 + cmp4.ne p6,p0=1,num + sub aptr=aptr,len };; // rewind +{ .mmi; sub nptr=nptr,len +(p41) add topbit=r0,r0,1 + add tptr=16,sp } +{ .mmb; add tp_1=8,sp + add num=-1,num // num-- +(p6) br.cond.sptk.many .Louter };; + +{ .mbb; add lc=4,lc + brp.loop.imp .Lsub_ctop,.Lsub_cend-16 + clrrrb.pr };; +{ .mii; nop.m 0 + mov pr.rot=0x10001<<16 + // ------^---- (p33) at first (p17) + mov ar.lc=lc } +{ .mii; nop.m 0 + mov ar.ec=3 + nop.i 0 };; + +.Lsub_ctop: +.pred.rel "mutex",p33,p35 +{ .mfi; (p16) ld8 t[0]=[tptr],8 // t=*(tp++) + (p16) nop.f 0 + (p33) sub n[1]=t[1],n[1] } // (p17) +{ .mfi; (p16) ld8 n[0]=[nptr],8 // n=*(np++) + (p16) nop.f 0 + (p35) sub n[1]=t[1],n[1],1 };; // (p17) +{ .mib; (p18) st8 [rptr]=n[2],8 // *(rp++)=r + (p33) cmp.gtu p34,p32=n[1],t[1] // (p17) + (p18) nop.b 0 } +{ .mib; (p18) nop.m 0 + (p35) cmp.geu p34,p32=n[1],t[1] // (p17) + br.ctop.sptk .Lsub_ctop };; +.Lsub_cend: + +{ .mmb; .pred.rel "mutex",p34,p36 +(p34) sub topbit=topbit,r0 // (p19) +(p36) sub topbit=topbit,r0,1 + brp.loop.imp .Lcopy_ctop,.Lcopy_cend-16 + } +{ .mmb; sub rptr=rptr,len // rewind + sub tptr=tptr,len + clrrrb.pr };; +{ .mmi; and aptr=tptr,topbit + andcm bptr=rptr,topbit + mov pr.rot=1<<16 };; +{ .mii; or nptr=aptr,bptr + mov ar.lc=lc + mov ar.ec=3 };; + +.Lcopy_ctop: +{ .mmb; (p16) ld8 n[0]=[nptr],8 + (p18) st8 [tptr]=r0,8 + (p16) nop.b 0 } +{ .mmb; (p16) nop.m 0 + (p18) st8 [rptr]=n[2],8 + br.ctop.sptk .Lcopy_ctop };; +.Lcopy_cend: + +{ .mmi; mov ret0=1 // signal "handled" + rum 1<<5 // clear um.mfh + mov ar.lc=prevlc } +{ .mib; .restore sp + mov sp=prevsp + mov pr=prevpr,0x1ffff + br.ret.sptk.many b0 };; +.endp bn_mul_mont_general# + +a1=r16; a2=r17; a3=r18; a4=r19; a5=r20; a6=r21; a7=r22; a8=r23; +n1=r24; n2=r25; n3=r26; n4=r27; n5=r28; n6=r29; n7=r30; n8=r31; +t0=r15; + +ai0=f8; ai1=f9; ai2=f10; ai3=f11; ai4=f12; ai5=f13; ai6=f14; ai7=f15; +ni0=f16; ni1=f17; ni2=f18; ni3=f19; ni4=f20; ni5=f21; ni6=f22; ni7=f23; + +.align 64 +.skip 48 // aligns loop body +.local bn_mul_mont_8# +.proc bn_mul_mont_8# +bn_mul_mont_8: + .prologue +{ .mmi; .save ar.pfs,prevfs + alloc prevfs=ar.pfs,6,2,0,8 + .vframe prevsp + mov prevsp=sp + .save ar.lc,prevlc + mov prevlc=ar.lc } +{ .mmi; add r17=-6*16,sp + add sp=-7*16,sp + .save pr,prevpr + mov prevpr=pr };; + +{ .mmi; .save.gf 0,0x10 + stf.spill [sp]=f16,-16 + .save.gf 0,0x20 + stf.spill [r17]=f17,32 + add r16=-5*16,prevsp};; +{ .mmi; .save.gf 0,0x40 + stf.spill [r16]=f18,32 + .save.gf 0,0x80 + stf.spill [r17]=f19,32 + $ADDP aptr=0,in1 };; +{ .mmi; .save.gf 0,0x100 + stf.spill [r16]=f20,32 + .save.gf 0,0x200 + stf.spill [r17]=f21,32 + $ADDP r29=8,in1 };; +{ .mmi; .save.gf 0,0x400 + stf.spill [r16]=f22 + .save.gf 0,0x800 + stf.spill [r17]=f23 + $ADDP rptr=0,in0 };; + + .body + .rotf bj[8],mj[2],tf[2],alo[10],ahi[10],nlo[10],nhi[10] + .rotr t[8] + +// load input vectors padding them to 8 elements +{ .mmi; ldf8 ai0=[aptr],16 // ap[0] + ldf8 ai1=[r29],16 // ap[1] + $ADDP bptr=0,in2 } +{ .mmi; $ADDP r30=8,in2 + $ADDP nptr=0,in3 + $ADDP r31=8,in3 };; +{ .mmi; ldf8 bj[7]=[bptr],16 // bp[0] + ldf8 bj[6]=[r30],16 // bp[1] + cmp4.le p4,p5=3,in5 } +{ .mmi; ldf8 ni0=[nptr],16 // np[0] + ldf8 ni1=[r31],16 // np[1] + cmp4.le p6,p7=4,in5 };; + +{ .mfi; (p4)ldf8 ai2=[aptr],16 // ap[2] + (p5)fcvt.fxu ai2=f0 + cmp4.le p8,p9=5,in5 } +{ .mfi; (p6)ldf8 ai3=[r29],16 // ap[3] + (p7)fcvt.fxu ai3=f0 + cmp4.le p10,p11=6,in5 } +{ .mfi; (p4)ldf8 bj[5]=[bptr],16 // bp[2] + (p5)fcvt.fxu bj[5]=f0 + cmp4.le p12,p13=7,in5 } +{ .mfi; (p6)ldf8 bj[4]=[r30],16 // bp[3] + (p7)fcvt.fxu bj[4]=f0 + cmp4.le p14,p15=8,in5 } +{ .mfi; (p4)ldf8 ni2=[nptr],16 // np[2] + (p5)fcvt.fxu ni2=f0 + addp4 r28=-1,in5 } +{ .mfi; (p6)ldf8 ni3=[r31],16 // np[3] + (p7)fcvt.fxu ni3=f0 + $ADDP in4=0,in4 };; + +{ .mfi; ldf8 n0=[in4] + fcvt.fxu tf[1]=f0 + nop.i 0 } + +{ .mfi; (p8)ldf8 ai4=[aptr],16 // ap[4] + (p9)fcvt.fxu ai4=f0 + mov t[0]=r0 } +{ .mfi; (p10)ldf8 ai5=[r29],16 // ap[5] + (p11)fcvt.fxu ai5=f0 + mov t[1]=r0 } +{ .mfi; (p8)ldf8 bj[3]=[bptr],16 // bp[4] + (p9)fcvt.fxu bj[3]=f0 + mov t[2]=r0 } +{ .mfi; (p10)ldf8 bj[2]=[r30],16 // bp[5] + (p11)fcvt.fxu bj[2]=f0 + mov t[3]=r0 } +{ .mfi; (p8)ldf8 ni4=[nptr],16 // np[4] + (p9)fcvt.fxu ni4=f0 + mov t[4]=r0 } +{ .mfi; (p10)ldf8 ni5=[r31],16 // np[5] + (p11)fcvt.fxu ni5=f0 + mov t[5]=r0 };; + +{ .mfi; (p12)ldf8 ai6=[aptr],16 // ap[6] + (p13)fcvt.fxu ai6=f0 + mov t[6]=r0 } +{ .mfi; (p14)ldf8 ai7=[r29],16 // ap[7] + (p15)fcvt.fxu ai7=f0 + mov t[7]=r0 } +{ .mfi; (p12)ldf8 bj[1]=[bptr],16 // bp[6] + (p13)fcvt.fxu bj[1]=f0 + mov ar.lc=r28 } +{ .mfi; (p14)ldf8 bj[0]=[r30],16 // bp[7] + (p15)fcvt.fxu bj[0]=f0 + mov ar.ec=1 } +{ .mfi; (p12)ldf8 ni6=[nptr],16 // np[6] + (p13)fcvt.fxu ni6=f0 + mov pr.rot=1<<16 } +{ .mfb; (p14)ldf8 ni7=[r31],16 // np[7] + (p15)fcvt.fxu ni7=f0 + brp.loop.imp .Louter_8_ctop,.Louter_8_cend-16 + };; + +// The loop is scheduled for 32*n ticks on Itanium 2. Actual attempt +// to measure with help of Interval Time Counter indicated that the +// factor is a tad higher: 33 or 34, if not 35. Exact measurement and +// addressing the issue is problematic, because I don't have access +// to platform-specific instruction-level profiler. On Itanium it +// should run in 56*n ticks, because of higher xma latency... +.Louter_8_ctop: + .pred.rel "mutex",p40,p42 + .pred.rel "mutex",p48,p50 +{ .mfi; (p16) nop.m 0 // 0: + (p16) xma.hu ahi[0]=ai0,bj[7],tf[1] // ap[0]*b[i]+t[0] + (p40) add a3=a3,n3 } // (p17) a3+=n3 +{ .mfi; (p42) add a3=a3,n3,1 + (p16) xma.lu alo[0]=ai0,bj[7],tf[1] + (p16) nop.i 0 };; +{ .mii; (p17) getf.sig a7=alo[8] // 1: + (p48) add t[6]=t[6],a3 // (p17) t[6]+=a3 + (p50) add t[6]=t[6],a3,1 };; +{ .mfi; (p17) getf.sig a8=ahi[8] // 2: + (p17) xma.hu nhi[7]=ni6,mj[1],nhi[6] // np[6]*m0 + (p40) cmp.ltu p43,p41=a3,n3 } +{ .mfi; (p42) cmp.leu p43,p41=a3,n3 + (p17) xma.lu nlo[7]=ni6,mj[1],nhi[6] + (p16) nop.i 0 };; +{ .mii; (p17) getf.sig n5=nlo[6] // 3: + (p48) cmp.ltu p51,p49=t[6],a3 + (p50) cmp.leu p51,p49=t[6],a3 };; + .pred.rel "mutex",p41,p43 + .pred.rel "mutex",p49,p51 +{ .mfi; (p16) nop.m 0 // 4: + (p16) xma.hu ahi[1]=ai1,bj[7],ahi[0] // ap[1]*b[i] + (p41) add a4=a4,n4 } // (p17) a4+=n4 +{ .mfi; (p43) add a4=a4,n4,1 + (p16) xma.lu alo[1]=ai1,bj[7],ahi[0] + (p16) nop.i 0 };; +{ .mfi; (p49) add t[5]=t[5],a4 // 5: (p17) t[5]+=a4 + (p16) xmpy.lu mj[0]=alo[0],n0 // (ap[0]*b[i]+t[0])*n0 + (p51) add t[5]=t[5],a4,1 };; +{ .mfi; (p16) nop.m 0 // 6: + (p17) xma.hu nhi[8]=ni7,mj[1],nhi[7] // np[7]*m0 + (p41) cmp.ltu p42,p40=a4,n4 } +{ .mfi; (p43) cmp.leu p42,p40=a4,n4 + (p17) xma.lu nlo[8]=ni7,mj[1],nhi[7] + (p16) nop.i 0 };; +{ .mii; (p17) getf.sig n6=nlo[7] // 7: + (p49) cmp.ltu p50,p48=t[5],a4 + (p51) cmp.leu p50,p48=t[5],a4 };; + .pred.rel "mutex",p40,p42 + .pred.rel "mutex",p48,p50 +{ .mfi; (p16) nop.m 0 // 8: + (p16) xma.hu ahi[2]=ai2,bj[7],ahi[1] // ap[2]*b[i] + (p40) add a5=a5,n5 } // (p17) a5+=n5 +{ .mfi; (p42) add a5=a5,n5,1 + (p16) xma.lu alo[2]=ai2,bj[7],ahi[1] + (p16) nop.i 0 };; +{ .mii; (p16) getf.sig a1=alo[1] // 9: + (p48) add t[4]=t[4],a5 // p(17) t[4]+=a5 + (p50) add t[4]=t[4],a5,1 };; +{ .mfi; (p16) nop.m 0 // 10: + (p16) xma.hu nhi[0]=ni0,mj[0],alo[0] // np[0]*m0 + (p40) cmp.ltu p43,p41=a5,n5 } +{ .mfi; (p42) cmp.leu p43,p41=a5,n5 + (p16) xma.lu nlo[0]=ni0,mj[0],alo[0] + (p16) nop.i 0 };; +{ .mii; (p17) getf.sig n7=nlo[8] // 11: + (p48) cmp.ltu p51,p49=t[4],a5 + (p50) cmp.leu p51,p49=t[4],a5 };; + .pred.rel "mutex",p41,p43 + .pred.rel "mutex",p49,p51 +{ .mfi; (p17) getf.sig n8=nhi[8] // 12: + (p16) xma.hu ahi[3]=ai3,bj[7],ahi[2] // ap[3]*b[i] + (p41) add a6=a6,n6 } // (p17) a6+=n6 +{ .mfi; (p43) add a6=a6,n6,1 + (p16) xma.lu alo[3]=ai3,bj[7],ahi[2] + (p16) nop.i 0 };; +{ .mii; (p16) getf.sig a2=alo[2] // 13: + (p49) add t[3]=t[3],a6 // (p17) t[3]+=a6 + (p51) add t[3]=t[3],a6,1 };; +{ .mfi; (p16) nop.m 0 // 14: + (p16) xma.hu nhi[1]=ni1,mj[0],nhi[0] // np[1]*m0 + (p41) cmp.ltu p42,p40=a6,n6 } +{ .mfi; (p43) cmp.leu p42,p40=a6,n6 + (p16) xma.lu nlo[1]=ni1,mj[0],nhi[0] + (p16) nop.i 0 };; +{ .mii; (p16) nop.m 0 // 15: + (p49) cmp.ltu p50,p48=t[3],a6 + (p51) cmp.leu p50,p48=t[3],a6 };; + .pred.rel "mutex",p40,p42 + .pred.rel "mutex",p48,p50 +{ .mfi; (p16) nop.m 0 // 16: + (p16) xma.hu ahi[4]=ai4,bj[7],ahi[3] // ap[4]*b[i] + (p40) add a7=a7,n7 } // (p17) a7+=n7 +{ .mfi; (p42) add a7=a7,n7,1 + (p16) xma.lu alo[4]=ai4,bj[7],ahi[3] + (p16) nop.i 0 };; +{ .mii; (p16) getf.sig a3=alo[3] // 17: + (p48) add t[2]=t[2],a7 // (p17) t[2]+=a7 + (p50) add t[2]=t[2],a7,1 };; +{ .mfi; (p16) nop.m 0 // 18: + (p16) xma.hu nhi[2]=ni2,mj[0],nhi[1] // np[2]*m0 + (p40) cmp.ltu p43,p41=a7,n7 } +{ .mfi; (p42) cmp.leu p43,p41=a7,n7 + (p16) xma.lu nlo[2]=ni2,mj[0],nhi[1] + (p16) nop.i 0 };; +{ .mii; (p16) getf.sig n1=nlo[1] // 19: + (p48) cmp.ltu p51,p49=t[2],a7 + (p50) cmp.leu p51,p49=t[2],a7 };; + .pred.rel "mutex",p41,p43 + .pred.rel "mutex",p49,p51 +{ .mfi; (p16) nop.m 0 // 20: + (p16) xma.hu ahi[5]=ai5,bj[7],ahi[4] // ap[5]*b[i] + (p41) add a8=a8,n8 } // (p17) a8+=n8 +{ .mfi; (p43) add a8=a8,n8,1 + (p16) xma.lu alo[5]=ai5,bj[7],ahi[4] + (p16) nop.i 0 };; +{ .mii; (p16) getf.sig a4=alo[4] // 21: + (p49) add t[1]=t[1],a8 // (p17) t[1]+=a8 + (p51) add t[1]=t[1],a8,1 };; +{ .mfi; (p16) nop.m 0 // 22: + (p16) xma.hu nhi[3]=ni3,mj[0],nhi[2] // np[3]*m0 + (p41) cmp.ltu p42,p40=a8,n8 } +{ .mfi; (p43) cmp.leu p42,p40=a8,n8 + (p16) xma.lu nlo[3]=ni3,mj[0],nhi[2] + (p16) nop.i 0 };; +{ .mii; (p16) getf.sig n2=nlo[2] // 23: + (p49) cmp.ltu p50,p48=t[1],a8 + (p51) cmp.leu p50,p48=t[1],a8 };; +{ .mfi; (p16) nop.m 0 // 24: + (p16) xma.hu ahi[6]=ai6,bj[7],ahi[5] // ap[6]*b[i] + (p16) add a1=a1,n1 } // (p16) a1+=n1 +{ .mfi; (p16) nop.m 0 + (p16) xma.lu alo[6]=ai6,bj[7],ahi[5] + (p17) mov t[0]=r0 };; +{ .mii; (p16) getf.sig a5=alo[5] // 25: + (p16) add t0=t[7],a1 // (p16) t[7]+=a1 + (p42) add t[0]=t[0],r0,1 };; +{ .mfi; (p16) setf.sig tf[0]=t0 // 26: + (p16) xma.hu nhi[4]=ni4,mj[0],nhi[3] // np[4]*m0 + (p50) add t[0]=t[0],r0,1 } +{ .mfi; (p16) cmp.ltu.unc p42,p40=a1,n1 + (p16) xma.lu nlo[4]=ni4,mj[0],nhi[3] + (p16) nop.i 0 };; +{ .mii; (p16) getf.sig n3=nlo[3] // 27: + (p16) cmp.ltu.unc p50,p48=t0,a1 + (p16) nop.i 0 };; + .pred.rel "mutex",p40,p42 + .pred.rel "mutex",p48,p50 +{ .mfi; (p16) nop.m 0 // 28: + (p16) xma.hu ahi[7]=ai7,bj[7],ahi[6] // ap[7]*b[i] + (p40) add a2=a2,n2 } // (p16) a2+=n2 +{ .mfi; (p42) add a2=a2,n2,1 + (p16) xma.lu alo[7]=ai7,bj[7],ahi[6] + (p16) nop.i 0 };; +{ .mii; (p16) getf.sig a6=alo[6] // 29: + (p48) add t[6]=t[6],a2 // (p16) t[6]+=a2 + (p50) add t[6]=t[6],a2,1 };; +{ .mfi; (p16) nop.m 0 // 30: + (p16) xma.hu nhi[5]=ni5,mj[0],nhi[4] // np[5]*m0 + (p40) cmp.ltu p41,p39=a2,n2 } +{ .mfi; (p42) cmp.leu p41,p39=a2,n2 + (p16) xma.lu nlo[5]=ni5,mj[0],nhi[4] + (p16) nop.i 0 };; +{ .mfi; (p16) getf.sig n4=nlo[4] // 31: + (p16) nop.f 0 + (p48) cmp.ltu p49,p47=t[6],a2 } +{ .mfb; (p50) cmp.leu p49,p47=t[6],a2 + (p16) nop.f 0 + br.ctop.sptk.many .Louter_8_ctop };; +.Louter_8_cend: + +// above loop has to execute one more time, without (p16), which is +// replaced with merged move of np[8] to GPR bank + .pred.rel "mutex",p40,p42 + .pred.rel "mutex",p48,p50 +{ .mmi; (p0) getf.sig n1=ni0 // 0: + (p40) add a3=a3,n3 // (p17) a3+=n3 + (p42) add a3=a3,n3,1 };; +{ .mii; (p17) getf.sig a7=alo[8] // 1: + (p48) add t[6]=t[6],a3 // (p17) t[6]+=a3 + (p50) add t[6]=t[6],a3,1 };; +{ .mfi; (p17) getf.sig a8=ahi[8] // 2: + (p17) xma.hu nhi[7]=ni6,mj[1],nhi[6] // np[6]*m0 + (p40) cmp.ltu p43,p41=a3,n3 } +{ .mfi; (p42) cmp.leu p43,p41=a3,n3 + (p17) xma.lu nlo[7]=ni6,mj[1],nhi[6] + (p0) nop.i 0 };; +{ .mii; (p17) getf.sig n5=nlo[6] // 3: + (p48) cmp.ltu p51,p49=t[6],a3 + (p50) cmp.leu p51,p49=t[6],a3 };; + .pred.rel "mutex",p41,p43 + .pred.rel "mutex",p49,p51 +{ .mmi; (p0) getf.sig n2=ni1 // 4: + (p41) add a4=a4,n4 // (p17) a4+=n4 + (p43) add a4=a4,n4,1 };; +{ .mfi; (p49) add t[5]=t[5],a4 // 5: (p17) t[5]+=a4 + (p0) nop.f 0 + (p51) add t[5]=t[5],a4,1 };; +{ .mfi; (p0) getf.sig n3=ni2 // 6: + (p17) xma.hu nhi[8]=ni7,mj[1],nhi[7] // np[7]*m0 + (p41) cmp.ltu p42,p40=a4,n4 } +{ .mfi; (p43) cmp.leu p42,p40=a4,n4 + (p17) xma.lu nlo[8]=ni7,mj[1],nhi[7] + (p0) nop.i 0 };; +{ .mii; (p17) getf.sig n6=nlo[7] // 7: + (p49) cmp.ltu p50,p48=t[5],a4 + (p51) cmp.leu p50,p48=t[5],a4 };; + .pred.rel "mutex",p40,p42 + .pred.rel "mutex",p48,p50 +{ .mii; (p0) getf.sig n4=ni3 // 8: + (p40) add a5=a5,n5 // (p17) a5+=n5 + (p42) add a5=a5,n5,1 };; +{ .mii; (p0) nop.m 0 // 9: + (p48) add t[4]=t[4],a5 // p(17) t[4]+=a5 + (p50) add t[4]=t[4],a5,1 };; +{ .mii; (p0) nop.m 0 // 10: + (p40) cmp.ltu p43,p41=a5,n5 + (p42) cmp.leu p43,p41=a5,n5 };; +{ .mii; (p17) getf.sig n7=nlo[8] // 11: + (p48) cmp.ltu p51,p49=t[4],a5 + (p50) cmp.leu p51,p49=t[4],a5 };; + .pred.rel "mutex",p41,p43 + .pred.rel "mutex",p49,p51 +{ .mii; (p17) getf.sig n8=nhi[8] // 12: + (p41) add a6=a6,n6 // (p17) a6+=n6 + (p43) add a6=a6,n6,1 };; +{ .mii; (p0) getf.sig n5=ni4 // 13: + (p49) add t[3]=t[3],a6 // (p17) t[3]+=a6 + (p51) add t[3]=t[3],a6,1 };; +{ .mii; (p0) nop.m 0 // 14: + (p41) cmp.ltu p42,p40=a6,n6 + (p43) cmp.leu p42,p40=a6,n6 };; +{ .mii; (p0) getf.sig n6=ni5 // 15: + (p49) cmp.ltu p50,p48=t[3],a6 + (p51) cmp.leu p50,p48=t[3],a6 };; + .pred.rel "mutex",p40,p42 + .pred.rel "mutex",p48,p50 +{ .mii; (p0) nop.m 0 // 16: + (p40) add a7=a7,n7 // (p17) a7+=n7 + (p42) add a7=a7,n7,1 };; +{ .mii; (p0) nop.m 0 // 17: + (p48) add t[2]=t[2],a7 // (p17) t[2]+=a7 + (p50) add t[2]=t[2],a7,1 };; +{ .mii; (p0) nop.m 0 // 18: + (p40) cmp.ltu p43,p41=a7,n7 + (p42) cmp.leu p43,p41=a7,n7 };; +{ .mii; (p0) getf.sig n7=ni6 // 19: + (p48) cmp.ltu p51,p49=t[2],a7 + (p50) cmp.leu p51,p49=t[2],a7 };; + .pred.rel "mutex",p41,p43 + .pred.rel "mutex",p49,p51 +{ .mii; (p0) nop.m 0 // 20: + (p41) add a8=a8,n8 // (p17) a8+=n8 + (p43) add a8=a8,n8,1 };; +{ .mmi; (p0) nop.m 0 // 21: + (p49) add t[1]=t[1],a8 // (p17) t[1]+=a8 + (p51) add t[1]=t[1],a8,1 } +{ .mmi; (p17) mov t[0]=r0 + (p41) cmp.ltu p42,p40=a8,n8 + (p43) cmp.leu p42,p40=a8,n8 };; +{ .mmi; (p0) getf.sig n8=ni7 // 22: + (p49) cmp.ltu p50,p48=t[1],a8 + (p51) cmp.leu p50,p48=t[1],a8 } +{ .mmi; (p42) add t[0]=t[0],r0,1 + (p0) add r16=-7*16,prevsp + (p0) add r17=-6*16,prevsp };; + +// subtract np[8] from carrybit|tmp[8] +// carrybit|tmp[8] layout upon exit from above loop is: +// t[0]|t[1]|t[2]|t[3]|t[4]|t[5]|t[6]|t[7]|t0 (least significant) +{ .mmi; (p50)add t[0]=t[0],r0,1 + add r18=-5*16,prevsp + sub n1=t0,n1 };; +{ .mmi; cmp.gtu p34,p32=n1,t0;; + .pred.rel "mutex",p32,p34 + (p32)sub n2=t[7],n2 + (p34)sub n2=t[7],n2,1 };; +{ .mii; (p32)cmp.gtu p35,p33=n2,t[7] + (p34)cmp.geu p35,p33=n2,t[7];; + .pred.rel "mutex",p33,p35 + (p33)sub n3=t[6],n3 } +{ .mmi; (p35)sub n3=t[6],n3,1;; + (p33)cmp.gtu p34,p32=n3,t[6] + (p35)cmp.geu p34,p32=n3,t[6] };; + .pred.rel "mutex",p32,p34 +{ .mii; (p32)sub n4=t[5],n4 + (p34)sub n4=t[5],n4,1;; + (p32)cmp.gtu p35,p33=n4,t[5] } +{ .mmi; (p34)cmp.geu p35,p33=n4,t[5];; + .pred.rel "mutex",p33,p35 + (p33)sub n5=t[4],n5 + (p35)sub n5=t[4],n5,1 };; +{ .mii; (p33)cmp.gtu p34,p32=n5,t[4] + (p35)cmp.geu p34,p32=n5,t[4];; + .pred.rel "mutex",p32,p34 + (p32)sub n6=t[3],n6 } +{ .mmi; (p34)sub n6=t[3],n6,1;; + (p32)cmp.gtu p35,p33=n6,t[3] + (p34)cmp.geu p35,p33=n6,t[3] };; + .pred.rel "mutex",p33,p35 +{ .mii; (p33)sub n7=t[2],n7 + (p35)sub n7=t[2],n7,1;; + (p33)cmp.gtu p34,p32=n7,t[2] } +{ .mmi; (p35)cmp.geu p34,p32=n7,t[2];; + .pred.rel "mutex",p32,p34 + (p32)sub n8=t[1],n8 + (p34)sub n8=t[1],n8,1 };; +{ .mii; (p32)cmp.gtu p35,p33=n8,t[1] + (p34)cmp.geu p35,p33=n8,t[1];; + .pred.rel "mutex",p33,p35 + (p33)sub a8=t[0],r0 } +{ .mmi; (p35)sub a8=t[0],r0,1;; + (p33)cmp.gtu p34,p32=a8,t[0] + (p35)cmp.geu p34,p32=a8,t[0] };; + +// save the result, either tmp[num] or tmp[num]-np[num] + .pred.rel "mutex",p32,p34 +{ .mmi; (p32)st8 [rptr]=n1,8 + (p34)st8 [rptr]=t0,8 + add r19=-4*16,prevsp};; +{ .mmb; (p32)st8 [rptr]=n2,8 + (p34)st8 [rptr]=t[7],8 + (p5)br.cond.dpnt.few .Ldone };; +{ .mmb; (p32)st8 [rptr]=n3,8 + (p34)st8 [rptr]=t[6],8 + (p7)br.cond.dpnt.few .Ldone };; +{ .mmb; (p32)st8 [rptr]=n4,8 + (p34)st8 [rptr]=t[5],8 + (p9)br.cond.dpnt.few .Ldone };; +{ .mmb; (p32)st8 [rptr]=n5,8 + (p34)st8 [rptr]=t[4],8 + (p11)br.cond.dpnt.few .Ldone };; +{ .mmb; (p32)st8 [rptr]=n6,8 + (p34)st8 [rptr]=t[3],8 + (p13)br.cond.dpnt.few .Ldone };; +{ .mmb; (p32)st8 [rptr]=n7,8 + (p34)st8 [rptr]=t[2],8 + (p15)br.cond.dpnt.few .Ldone };; +{ .mmb; (p32)st8 [rptr]=n8,8 + (p34)st8 [rptr]=t[1],8 + nop.b 0 };; +.Ldone: // epilogue +{ .mmi; ldf.fill f16=[r16],64 + ldf.fill f17=[r17],64 + nop.i 0 } +{ .mmi; ldf.fill f18=[r18],64 + ldf.fill f19=[r19],64 + mov pr=prevpr,0x1ffff };; +{ .mmi; ldf.fill f20=[r16] + ldf.fill f21=[r17] + mov ar.lc=prevlc } +{ .mmi; ldf.fill f22=[r18] + ldf.fill f23=[r19] + mov ret0=1 } // signal "handled" +{ .mib; rum 1<<5 + .restore sp + mov sp=prevsp + br.ret.sptk.many b0 };; +.endp bn_mul_mont_8# + +.type copyright#,\@object +copyright: +stringz "Montgomery multiplication for IA-64, CRYPTOGAMS by " +___ + +$output=shift and open STDOUT,">$output"; +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ia64.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ia64.S new file mode 100644 index 00000000..c0cee821 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ia64.S @@ -0,0 +1,1555 @@ +.explicit +.text +.ident "ia64.S, Version 2.1" +.ident "IA-64 ISA artwork by Andy Polyakov " + +// +// ==================================================================== +// Written by Andy Polyakov for the OpenSSL +// project. +// +// Rights for redistribution and usage in source and binary forms are +// granted according to the OpenSSL license. Warranty of any kind is +// disclaimed. +// ==================================================================== +// +// Version 2.x is Itanium2 re-tune. Few words about how Itanum2 is +// different from Itanium to this module viewpoint. Most notably, is it +// "wider" than Itanium? Can you experience loop scalability as +// discussed in commentary sections? Not really:-( Itanium2 has 6 +// integer ALU ports, i.e. it's 2 ports wider, but it's not enough to +// spin twice as fast, as I need 8 IALU ports. Amount of floating point +// ports is the same, i.e. 2, while I need 4. In other words, to this +// module Itanium2 remains effectively as "wide" as Itanium. Yet it's +// essentially different in respect to this module, and a re-tune was +// required. Well, because some intruction latencies has changed. Most +// noticeably those intensively used: +// +// Itanium Itanium2 +// ldf8 9 6 L2 hit +// ld8 2 1 L1 hit +// getf 2 5 +// xma[->getf] 7[+1] 4[+0] +// add[->st8] 1[+1] 1[+0] +// +// What does it mean? You might ratiocinate that the original code +// should run just faster... Because sum of latencies is smaller... +// Wrong! Note that getf latency increased. This means that if a loop is +// scheduled for lower latency (as they were), then it will suffer from +// stall condition and the code will therefore turn anti-scalable, e.g. +// original bn_mul_words spun at 5*n or 2.5 times slower than expected +// on Itanium2! What to do? Reschedule loops for Itanium2? But then +// Itanium would exhibit anti-scalability. So I've chosen to reschedule +// for worst latency for every instruction aiming for best *all-round* +// performance. + +// Q. How much faster does it get? +// A. Here is the output from 'openssl speed rsa dsa' for vanilla +// 0.9.6a compiled with gcc version 2.96 20000731 (Red Hat +// Linux 7.1 2.96-81): +// +// sign verify sign/s verify/s +// rsa 512 bits 0.0036s 0.0003s 275.3 2999.2 +// rsa 1024 bits 0.0203s 0.0011s 49.3 894.1 +// rsa 2048 bits 0.1331s 0.0040s 7.5 250.9 +// rsa 4096 bits 0.9270s 0.0147s 1.1 68.1 +// sign verify sign/s verify/s +// dsa 512 bits 0.0035s 0.0043s 288.3 234.8 +// dsa 1024 bits 0.0111s 0.0135s 90.0 74.2 +// +// And here is similar output but for this assembler +// implementation:-) +// +// sign verify sign/s verify/s +// rsa 512 bits 0.0021s 0.0001s 549.4 9638.5 +// rsa 1024 bits 0.0055s 0.0002s 183.8 4481.1 +// rsa 2048 bits 0.0244s 0.0006s 41.4 1726.3 +// rsa 4096 bits 0.1295s 0.0018s 7.7 561.5 +// sign verify sign/s verify/s +// dsa 512 bits 0.0012s 0.0013s 891.9 756.6 +// dsa 1024 bits 0.0023s 0.0028s 440.4 376.2 +// +// Yes, you may argue that it's not fair comparison as it's +// possible to craft the C implementation with BN_UMULT_HIGH +// inline assembler macro. But of course! Here is the output +// with the macro: +// +// sign verify sign/s verify/s +// rsa 512 bits 0.0020s 0.0002s 495.0 6561.0 +// rsa 1024 bits 0.0086s 0.0004s 116.2 2235.7 +// rsa 2048 bits 0.0519s 0.0015s 19.3 667.3 +// rsa 4096 bits 0.3464s 0.0053s 2.9 187.7 +// sign verify sign/s verify/s +// dsa 512 bits 0.0016s 0.0020s 613.1 510.5 +// dsa 1024 bits 0.0045s 0.0054s 221.0 183.9 +// +// My code is still way faster, huh:-) And I believe that even +// higher performance can be achieved. Note that as keys get +// longer, performance gain is larger. Why? According to the +// profiler there is another player in the field, namely +// BN_from_montgomery consuming larger and larger portion of CPU +// time as keysize decreases. I therefore consider putting effort +// to assembler implementation of the following routine: +// +// void bn_mul_add_mont (BN_ULONG *rp,BN_ULONG *np,int nl,BN_ULONG n0) +// { +// int i,j; +// BN_ULONG v; +// +// for (i=0; i" diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips-mont.pl new file mode 100644 index 00000000..caae04ed --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips-mont.pl @@ -0,0 +1,426 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# This module doesn't present direct interest for OpenSSL, because it +# doesn't provide better performance for longer keys, at least not on +# in-order-execution cores. While 512-bit RSA sign operations can be +# 65% faster in 64-bit mode, 1024-bit ones are only 15% faster, and +# 4096-bit ones are up to 15% slower. In 32-bit mode it varies from +# 16% improvement for 512-bit RSA sign to -33% for 4096-bit RSA +# verify:-( All comparisons are against bn_mul_mont-free assembler. +# The module might be of interest to embedded system developers, as +# the code is smaller than 1KB, yet offers >3x improvement on MIPS64 +# and 75-30% [less for longer keys] on MIPS32 over compiler-generated +# code. + +###################################################################### +# There is a number of MIPS ABI in use, O32 and N32/64 are most +# widely used. Then there is a new contender: NUBI. It appears that if +# one picks the latter, it's possible to arrange code in ABI neutral +# manner. Therefore let's stick to NUBI register layout: +# +($zero,$at,$t0,$t1,$t2)=map("\$$_",(0..2,24,25)); +($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11)); +($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7,$s8,$s9,$s10,$s11)=map("\$$_",(12..23)); +($gp,$tp,$sp,$fp,$ra)=map("\$$_",(3,28..31)); +# +# The return value is placed in $a0. Following coding rules facilitate +# interoperability: +# +# - never ever touch $tp, "thread pointer", former $gp; +# - copy return value to $t0, former $v0 [or to $a0 if you're adapting +# old code]; +# - on O32 populate $a4-$a7 with 'lw $aN,4*N($sp)' if necessary; +# +# For reference here is register layout for N32/64 MIPS ABIs: +# +# ($zero,$at,$v0,$v1)=map("\$$_",(0..3)); +# ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11)); +# ($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25)); +# ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23)); +# ($gp,$sp,$fp,$ra)=map("\$$_",(28..31)); +# +$flavour = shift; # supported flavours are o32,n32,64,nubi32,nubi64 + +if ($flavour =~ /64|n32/i) { + $PTR_ADD="dadd"; # incidentally works even on n32 + $PTR_SUB="dsub"; # incidentally works even on n32 + $REG_S="sd"; + $REG_L="ld"; + $SZREG=8; +} else { + $PTR_ADD="add"; + $PTR_SUB="sub"; + $REG_S="sw"; + $REG_L="lw"; + $SZREG=4; +} +$SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0x00fff000 : 0x00ff0000; +# +# +# +###################################################################### + +while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +if ($flavour =~ /64|n32/i) { + $LD="ld"; + $ST="sd"; + $MULTU="dmultu"; + $ADDU="daddu"; + $SUBU="dsubu"; + $BNSZ=8; +} else { + $LD="lw"; + $ST="sw"; + $MULTU="multu"; + $ADDU="addu"; + $SUBU="subu"; + $BNSZ=4; +} + +# int bn_mul_mont( +$rp=$a0; # BN_ULONG *rp, +$ap=$a1; # const BN_ULONG *ap, +$bp=$a2; # const BN_ULONG *bp, +$np=$a3; # const BN_ULONG *np, +$n0=$a4; # const BN_ULONG *n0, +$num=$a5; # int num); + +$lo0=$a6; +$hi0=$a7; +$lo1=$t1; +$hi1=$t2; +$aj=$s0; +$bi=$s1; +$nj=$s2; +$tp=$s3; +$alo=$s4; +$ahi=$s5; +$nlo=$s6; +$nhi=$s7; +$tj=$s8; +$i=$s9; +$j=$s10; +$m1=$s11; + +$FRAMESIZE=14; + +$code=<<___; +.text + +.set noat +.set noreorder + +.align 5 +.globl bn_mul_mont +.ent bn_mul_mont +bn_mul_mont: +___ +$code.=<<___ if ($flavour =~ /o32/i); + lw $n0,16($sp) + lw $num,20($sp) +___ +$code.=<<___; + slt $at,$num,4 + bnez $at,1f + li $t0,0 + slt $at,$num,17 # on in-order CPU + bnez $at,bn_mul_mont_internal + nop +1: jr $ra + li $a0,0 +.end bn_mul_mont + +.align 5 +.ent bn_mul_mont_internal +bn_mul_mont_internal: + .frame $fp,$FRAMESIZE*$SZREG,$ra + .mask 0x40000000|$SAVED_REGS_MASK,-$SZREG + $PTR_SUB $sp,$FRAMESIZE*$SZREG + $REG_S $fp,($FRAMESIZE-1)*$SZREG($sp) + $REG_S $s11,($FRAMESIZE-2)*$SZREG($sp) + $REG_S $s10,($FRAMESIZE-3)*$SZREG($sp) + $REG_S $s9,($FRAMESIZE-4)*$SZREG($sp) + $REG_S $s8,($FRAMESIZE-5)*$SZREG($sp) + $REG_S $s7,($FRAMESIZE-6)*$SZREG($sp) + $REG_S $s6,($FRAMESIZE-7)*$SZREG($sp) + $REG_S $s5,($FRAMESIZE-8)*$SZREG($sp) + $REG_S $s4,($FRAMESIZE-9)*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_S $s3,($FRAMESIZE-10)*$SZREG($sp) + $REG_S $s2,($FRAMESIZE-11)*$SZREG($sp) + $REG_S $s1,($FRAMESIZE-12)*$SZREG($sp) + $REG_S $s0,($FRAMESIZE-13)*$SZREG($sp) +___ +$code.=<<___; + move $fp,$sp + + .set reorder + $LD $n0,0($n0) + $LD $bi,0($bp) # bp[0] + $LD $aj,0($ap) # ap[0] + $LD $nj,0($np) # np[0] + + $PTR_SUB $sp,2*$BNSZ # place for two extra words + sll $num,`log($BNSZ)/log(2)` + li $at,-4096 + $PTR_SUB $sp,$num + and $sp,$at + + $MULTU $aj,$bi + $LD $alo,$BNSZ($ap) + $LD $nlo,$BNSZ($np) + mflo $lo0 + mfhi $hi0 + $MULTU $lo0,$n0 + mflo $m1 + + $MULTU $alo,$bi + mflo $alo + mfhi $ahi + + $MULTU $nj,$m1 + mflo $lo1 + mfhi $hi1 + $MULTU $nlo,$m1 + $ADDU $lo1,$lo0 + sltu $at,$lo1,$lo0 + $ADDU $hi1,$at + mflo $nlo + mfhi $nhi + + move $tp,$sp + li $j,2*$BNSZ +.align 4 +.L1st: + .set noreorder + $PTR_ADD $aj,$ap,$j + $PTR_ADD $nj,$np,$j + $LD $aj,($aj) + $LD $nj,($nj) + + $MULTU $aj,$bi + $ADDU $lo0,$alo,$hi0 + $ADDU $lo1,$nlo,$hi1 + sltu $at,$lo0,$hi0 + sltu $t0,$lo1,$hi1 + $ADDU $hi0,$ahi,$at + $ADDU $hi1,$nhi,$t0 + mflo $alo + mfhi $ahi + + $ADDU $lo1,$lo0 + sltu $at,$lo1,$lo0 + $MULTU $nj,$m1 + $ADDU $hi1,$at + addu $j,$BNSZ + $ST $lo1,($tp) + sltu $t0,$j,$num + mflo $nlo + mfhi $nhi + + bnez $t0,.L1st + $PTR_ADD $tp,$BNSZ + .set reorder + + $ADDU $lo0,$alo,$hi0 + sltu $at,$lo0,$hi0 + $ADDU $hi0,$ahi,$at + + $ADDU $lo1,$nlo,$hi1 + sltu $t0,$lo1,$hi1 + $ADDU $hi1,$nhi,$t0 + $ADDU $lo1,$lo0 + sltu $at,$lo1,$lo0 + $ADDU $hi1,$at + + $ST $lo1,($tp) + + $ADDU $hi1,$hi0 + sltu $at,$hi1,$hi0 + $ST $hi1,$BNSZ($tp) + $ST $at,2*$BNSZ($tp) + + li $i,$BNSZ +.align 4 +.Louter: + $PTR_ADD $bi,$bp,$i + $LD $bi,($bi) + $LD $aj,($ap) + $LD $alo,$BNSZ($ap) + $LD $tj,($sp) + + $MULTU $aj,$bi + $LD $nj,($np) + $LD $nlo,$BNSZ($np) + mflo $lo0 + mfhi $hi0 + $ADDU $lo0,$tj + $MULTU $lo0,$n0 + sltu $at,$lo0,$tj + $ADDU $hi0,$at + mflo $m1 + + $MULTU $alo,$bi + mflo $alo + mfhi $ahi + + $MULTU $nj,$m1 + mflo $lo1 + mfhi $hi1 + + $MULTU $nlo,$m1 + $ADDU $lo1,$lo0 + sltu $at,$lo1,$lo0 + $ADDU $hi1,$at + mflo $nlo + mfhi $nhi + + move $tp,$sp + li $j,2*$BNSZ + $LD $tj,$BNSZ($tp) +.align 4 +.Linner: + .set noreorder + $PTR_ADD $aj,$ap,$j + $PTR_ADD $nj,$np,$j + $LD $aj,($aj) + $LD $nj,($nj) + + $MULTU $aj,$bi + $ADDU $lo0,$alo,$hi0 + $ADDU $lo1,$nlo,$hi1 + sltu $at,$lo0,$hi0 + sltu $t0,$lo1,$hi1 + $ADDU $hi0,$ahi,$at + $ADDU $hi1,$nhi,$t0 + mflo $alo + mfhi $ahi + + $ADDU $lo0,$tj + addu $j,$BNSZ + $MULTU $nj,$m1 + sltu $at,$lo0,$tj + $ADDU $lo1,$lo0 + $ADDU $hi0,$at + sltu $t0,$lo1,$lo0 + $LD $tj,2*$BNSZ($tp) + $ADDU $hi1,$t0 + sltu $at,$j,$num + mflo $nlo + mfhi $nhi + $ST $lo1,($tp) + bnez $at,.Linner + $PTR_ADD $tp,$BNSZ + .set reorder + + $ADDU $lo0,$alo,$hi0 + sltu $at,$lo0,$hi0 + $ADDU $hi0,$ahi,$at + $ADDU $lo0,$tj + sltu $t0,$lo0,$tj + $ADDU $hi0,$t0 + + $LD $tj,2*$BNSZ($tp) + $ADDU $lo1,$nlo,$hi1 + sltu $at,$lo1,$hi1 + $ADDU $hi1,$nhi,$at + $ADDU $lo1,$lo0 + sltu $t0,$lo1,$lo0 + $ADDU $hi1,$t0 + $ST $lo1,($tp) + + $ADDU $lo1,$hi1,$hi0 + sltu $hi1,$lo1,$hi0 + $ADDU $lo1,$tj + sltu $at,$lo1,$tj + $ADDU $hi1,$at + $ST $lo1,$BNSZ($tp) + $ST $hi1,2*$BNSZ($tp) + + addu $i,$BNSZ + sltu $t0,$i,$num + bnez $t0,.Louter + + .set noreorder + $PTR_ADD $tj,$sp,$num # &tp[num] + move $tp,$sp + move $ap,$sp + li $hi0,0 # clear borrow bit + +.align 4 +.Lsub: $LD $lo0,($tp) + $LD $lo1,($np) + $PTR_ADD $tp,$BNSZ + $PTR_ADD $np,$BNSZ + $SUBU $lo1,$lo0,$lo1 # tp[i]-np[i] + sgtu $at,$lo1,$lo0 + $SUBU $lo0,$lo1,$hi0 + sgtu $hi0,$lo0,$lo1 + $ST $lo0,($rp) + or $hi0,$at + sltu $at,$tp,$tj + bnez $at,.Lsub + $PTR_ADD $rp,$BNSZ + + $SUBU $hi0,$hi1,$hi0 # handle upmost overflow bit + move $tp,$sp + $PTR_SUB $rp,$num # restore rp + not $hi1,$hi0 + + and $ap,$hi0,$sp + and $bp,$hi1,$rp + or $ap,$ap,$bp # ap=borrow?tp:rp + +.align 4 +.Lcopy: $LD $aj,($ap) + $PTR_ADD $ap,$BNSZ + $ST $zero,($tp) + $PTR_ADD $tp,$BNSZ + sltu $at,$tp,$tj + $ST $aj,($rp) + bnez $at,.Lcopy + $PTR_ADD $rp,$BNSZ + + li $a0,1 + li $t0,1 + + .set noreorder + move $sp,$fp + $REG_L $fp,($FRAMESIZE-1)*$SZREG($sp) + $REG_L $s11,($FRAMESIZE-2)*$SZREG($sp) + $REG_L $s10,($FRAMESIZE-3)*$SZREG($sp) + $REG_L $s9,($FRAMESIZE-4)*$SZREG($sp) + $REG_L $s8,($FRAMESIZE-5)*$SZREG($sp) + $REG_L $s7,($FRAMESIZE-6)*$SZREG($sp) + $REG_L $s6,($FRAMESIZE-7)*$SZREG($sp) + $REG_L $s5,($FRAMESIZE-8)*$SZREG($sp) + $REG_L $s4,($FRAMESIZE-9)*$SZREG($sp) +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $s3,($FRAMESIZE-10)*$SZREG($sp) + $REG_L $s2,($FRAMESIZE-11)*$SZREG($sp) + $REG_L $s1,($FRAMESIZE-12)*$SZREG($sp) + $REG_L $s0,($FRAMESIZE-13)*$SZREG($sp) +___ +$code.=<<___; + jr $ra + $PTR_ADD $sp,$FRAMESIZE*$SZREG +.end bn_mul_mont_internal +.rdata +.asciiz "Montgomery Multiplication for MIPS, CRYPTOGAMS by " +___ + +$code =~ s/\`([^\`]*)\`/eval $1/gem; + +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips.pl new file mode 100644 index 00000000..d2f3ef7b --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips.pl @@ -0,0 +1,2583 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. +# +# Rights for redistribution and usage in source and binary forms are +# granted according to the OpenSSL license. Warranty of any kind is +# disclaimed. +# ==================================================================== + + +# July 1999 +# +# This is drop-in MIPS III/IV ISA replacement for crypto/bn/bn_asm.c. +# +# The module is designed to work with either of the "new" MIPS ABI(5), +# namely N32 or N64, offered by IRIX 6.x. It's not ment to work under +# IRIX 5.x not only because it doesn't support new ABIs but also +# because 5.x kernels put R4x00 CPU into 32-bit mode and all those +# 64-bit instructions (daddu, dmultu, etc.) found below gonna only +# cause illegal instruction exception:-( +# +# In addition the code depends on preprocessor flags set up by MIPSpro +# compiler driver (either as or cc) and therefore (probably?) can't be +# compiled by the GNU assembler. GNU C driver manages fine though... +# I mean as long as -mmips-as is specified or is the default option, +# because then it simply invokes /usr/bin/as which in turn takes +# perfect care of the preprocessor definitions. Another neat feature +# offered by the MIPSpro assembler is an optimization pass. This gave +# me the opportunity to have the code looking more regular as all those +# architecture dependent instruction rescheduling details were left to +# the assembler. Cool, huh? +# +# Performance improvement is astonishing! 'apps/openssl speed rsa dsa' +# goes way over 3 times faster! +# +# + +# October 2010 +# +# Adapt the module even for 32-bit ABIs and other OSes. The former was +# achieved by mechanical replacement of 64-bit arithmetic instructions +# such as dmultu, daddu, etc. with their 32-bit counterparts and +# adjusting offsets denoting multiples of BN_ULONG. Above mentioned +# >3x performance improvement naturally does not apply to 32-bit code +# [because there is no instruction 32-bit compiler can't use], one +# has to content with 40-85% improvement depending on benchmark and +# key length, more for longer keys. + +$flavour = shift; +while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +if ($flavour =~ /64|n32/i) { + $LD="ld"; + $ST="sd"; + $MULTU="dmultu"; + $DIVU="ddivu"; + $ADDU="daddu"; + $SUBU="dsubu"; + $SRL="dsrl"; + $SLL="dsll"; + $BNSZ=8; + $PTR_ADD="daddu"; + $PTR_SUB="dsubu"; + $SZREG=8; + $REG_S="sd"; + $REG_L="ld"; +} else { + $LD="lw"; + $ST="sw"; + $MULTU="multu"; + $DIVU="divu"; + $ADDU="addu"; + $SUBU="subu"; + $SRL="srl"; + $SLL="sll"; + $BNSZ=4; + $PTR_ADD="addu"; + $PTR_SUB="subu"; + $SZREG=4; + $REG_S="sw"; + $REG_L="lw"; + $code=".set mips2\n"; +} + +# Below is N32/64 register layout used in the original module. +# +($zero,$at,$v0,$v1)=map("\$$_",(0..3)); +($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11)); +($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25)); +($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23)); +($gp,$sp,$fp,$ra)=map("\$$_",(28..31)); +($ta0,$ta1,$ta2,$ta3)=($a4,$a5,$a6,$a7); +# +# No special adaptation is required for O32. NUBI on the other hand +# is treated by saving/restoring ($v1,$t0..$t3). + +$gp=$v1 if ($flavour =~ /nubi/i); + +$minus4=$v1; + +$code.=<<___; +.rdata +.asciiz "mips3.s, Version 1.2" +.asciiz "MIPS II/III/IV ISA artwork by Andy Polyakov " + +.text +.set noat + +.align 5 +.globl bn_mul_add_words +.ent bn_mul_add_words +bn_mul_add_words: + .set noreorder + bgtz $a2,bn_mul_add_words_internal + move $v0,$zero + jr $ra + move $a0,$v0 +.end bn_mul_add_words + +.align 5 +.ent bn_mul_add_words_internal +bn_mul_add_words_internal: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + li $minus4,-4 + and $ta0,$a2,$minus4 + beqz $ta0,.L_bn_mul_add_words_tail + +.L_bn_mul_add_words_loop: + $LD $t0,0($a1) + $MULTU $t0,$a3 + $LD $t1,0($a0) + $LD $t2,$BNSZ($a1) + $LD $t3,$BNSZ($a0) + $LD $ta0,2*$BNSZ($a1) + $LD $ta1,2*$BNSZ($a0) + $ADDU $t1,$v0 + sltu $v0,$t1,$v0 # All manuals say it "compares 32-bit + # values", but it seems to work fine + # even on 64-bit registers. + mflo $at + mfhi $t0 + $ADDU $t1,$at + $ADDU $v0,$t0 + $MULTU $t2,$a3 + sltu $at,$t1,$at + $ST $t1,0($a0) + $ADDU $v0,$at + + $LD $ta2,3*$BNSZ($a1) + $LD $ta3,3*$BNSZ($a0) + $ADDU $t3,$v0 + sltu $v0,$t3,$v0 + mflo $at + mfhi $t2 + $ADDU $t3,$at + $ADDU $v0,$t2 + $MULTU $ta0,$a3 + sltu $at,$t3,$at + $ST $t3,$BNSZ($a0) + $ADDU $v0,$at + + subu $a2,4 + $PTR_ADD $a0,4*$BNSZ + $PTR_ADD $a1,4*$BNSZ + $ADDU $ta1,$v0 + sltu $v0,$ta1,$v0 + mflo $at + mfhi $ta0 + $ADDU $ta1,$at + $ADDU $v0,$ta0 + $MULTU $ta2,$a3 + sltu $at,$ta1,$at + $ST $ta1,-2*$BNSZ($a0) + $ADDU $v0,$at + + + and $ta0,$a2,$minus4 + $ADDU $ta3,$v0 + sltu $v0,$ta3,$v0 + mflo $at + mfhi $ta2 + $ADDU $ta3,$at + $ADDU $v0,$ta2 + sltu $at,$ta3,$at + $ST $ta3,-$BNSZ($a0) + .set noreorder + bgtz $ta0,.L_bn_mul_add_words_loop + $ADDU $v0,$at + + beqz $a2,.L_bn_mul_add_words_return + nop + +.L_bn_mul_add_words_tail: + .set reorder + $LD $t0,0($a1) + $MULTU $t0,$a3 + $LD $t1,0($a0) + subu $a2,1 + $ADDU $t1,$v0 + sltu $v0,$t1,$v0 + mflo $at + mfhi $t0 + $ADDU $t1,$at + $ADDU $v0,$t0 + sltu $at,$t1,$at + $ST $t1,0($a0) + $ADDU $v0,$at + beqz $a2,.L_bn_mul_add_words_return + + $LD $t0,$BNSZ($a1) + $MULTU $t0,$a3 + $LD $t1,$BNSZ($a0) + subu $a2,1 + $ADDU $t1,$v0 + sltu $v0,$t1,$v0 + mflo $at + mfhi $t0 + $ADDU $t1,$at + $ADDU $v0,$t0 + sltu $at,$t1,$at + $ST $t1,$BNSZ($a0) + $ADDU $v0,$at + beqz $a2,.L_bn_mul_add_words_return + + $LD $t0,2*$BNSZ($a1) + $MULTU $t0,$a3 + $LD $t1,2*$BNSZ($a0) + $ADDU $t1,$v0 + sltu $v0,$t1,$v0 + mflo $at + mfhi $t0 + $ADDU $t1,$at + $ADDU $v0,$t0 + sltu $at,$t1,$at + $ST $t1,2*$BNSZ($a0) + $ADDU $v0,$at + +.L_bn_mul_add_words_return: + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + move $a0,$v0 +.end bn_mul_add_words_internal + +.align 5 +.globl bn_mul_words +.ent bn_mul_words +bn_mul_words: + .set noreorder + bgtz $a2,bn_mul_words_internal + move $v0,$zero + jr $ra + move $a0,$v0 +.end bn_mul_words + +.align 5 +.ent bn_mul_words_internal +bn_mul_words_internal: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + li $minus4,-4 + and $ta0,$a2,$minus4 + beqz $ta0,.L_bn_mul_words_tail + +.L_bn_mul_words_loop: + $LD $t0,0($a1) + $MULTU $t0,$a3 + $LD $t2,$BNSZ($a1) + $LD $ta0,2*$BNSZ($a1) + $LD $ta2,3*$BNSZ($a1) + mflo $at + mfhi $t0 + $ADDU $v0,$at + sltu $t1,$v0,$at + $MULTU $t2,$a3 + $ST $v0,0($a0) + $ADDU $v0,$t1,$t0 + + subu $a2,4 + $PTR_ADD $a0,4*$BNSZ + $PTR_ADD $a1,4*$BNSZ + mflo $at + mfhi $t2 + $ADDU $v0,$at + sltu $t3,$v0,$at + $MULTU $ta0,$a3 + $ST $v0,-3*$BNSZ($a0) + $ADDU $v0,$t3,$t2 + + mflo $at + mfhi $ta0 + $ADDU $v0,$at + sltu $ta1,$v0,$at + $MULTU $ta2,$a3 + $ST $v0,-2*$BNSZ($a0) + $ADDU $v0,$ta1,$ta0 + + and $ta0,$a2,$minus4 + mflo $at + mfhi $ta2 + $ADDU $v0,$at + sltu $ta3,$v0,$at + $ST $v0,-$BNSZ($a0) + .set noreorder + bgtz $ta0,.L_bn_mul_words_loop + $ADDU $v0,$ta3,$ta2 + + beqz $a2,.L_bn_mul_words_return + nop + +.L_bn_mul_words_tail: + .set reorder + $LD $t0,0($a1) + $MULTU $t0,$a3 + subu $a2,1 + mflo $at + mfhi $t0 + $ADDU $v0,$at + sltu $t1,$v0,$at + $ST $v0,0($a0) + $ADDU $v0,$t1,$t0 + beqz $a2,.L_bn_mul_words_return + + $LD $t0,$BNSZ($a1) + $MULTU $t0,$a3 + subu $a2,1 + mflo $at + mfhi $t0 + $ADDU $v0,$at + sltu $t1,$v0,$at + $ST $v0,$BNSZ($a0) + $ADDU $v0,$t1,$t0 + beqz $a2,.L_bn_mul_words_return + + $LD $t0,2*$BNSZ($a1) + $MULTU $t0,$a3 + mflo $at + mfhi $t0 + $ADDU $v0,$at + sltu $t1,$v0,$at + $ST $v0,2*$BNSZ($a0) + $ADDU $v0,$t1,$t0 + +.L_bn_mul_words_return: + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + move $a0,$v0 +.end bn_mul_words_internal + +.align 5 +.globl bn_sqr_words +.ent bn_sqr_words +bn_sqr_words: + .set noreorder + bgtz $a2,bn_sqr_words_internal + move $v0,$zero + jr $ra + move $a0,$v0 +.end bn_sqr_words + +.align 5 +.ent bn_sqr_words_internal +bn_sqr_words_internal: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + li $minus4,-4 + and $ta0,$a2,$minus4 + beqz $ta0,.L_bn_sqr_words_tail + +.L_bn_sqr_words_loop: + $LD $t0,0($a1) + $MULTU $t0,$t0 + $LD $t2,$BNSZ($a1) + $LD $ta0,2*$BNSZ($a1) + $LD $ta2,3*$BNSZ($a1) + mflo $t1 + mfhi $t0 + $ST $t1,0($a0) + $ST $t0,$BNSZ($a0) + + $MULTU $t2,$t2 + subu $a2,4 + $PTR_ADD $a0,8*$BNSZ + $PTR_ADD $a1,4*$BNSZ + mflo $t3 + mfhi $t2 + $ST $t3,-6*$BNSZ($a0) + $ST $t2,-5*$BNSZ($a0) + + $MULTU $ta0,$ta0 + mflo $ta1 + mfhi $ta0 + $ST $ta1,-4*$BNSZ($a0) + $ST $ta0,-3*$BNSZ($a0) + + + $MULTU $ta2,$ta2 + and $ta0,$a2,$minus4 + mflo $ta3 + mfhi $ta2 + $ST $ta3,-2*$BNSZ($a0) + + .set noreorder + bgtz $ta0,.L_bn_sqr_words_loop + $ST $ta2,-$BNSZ($a0) + + beqz $a2,.L_bn_sqr_words_return + nop + +.L_bn_sqr_words_tail: + .set reorder + $LD $t0,0($a1) + $MULTU $t0,$t0 + subu $a2,1 + mflo $t1 + mfhi $t0 + $ST $t1,0($a0) + $ST $t0,$BNSZ($a0) + beqz $a2,.L_bn_sqr_words_return + + $LD $t0,$BNSZ($a1) + $MULTU $t0,$t0 + subu $a2,1 + mflo $t1 + mfhi $t0 + $ST $t1,2*$BNSZ($a0) + $ST $t0,3*$BNSZ($a0) + beqz $a2,.L_bn_sqr_words_return + + $LD $t0,2*$BNSZ($a1) + $MULTU $t0,$t0 + mflo $t1 + mfhi $t0 + $ST $t1,4*$BNSZ($a0) + $ST $t0,5*$BNSZ($a0) + +.L_bn_sqr_words_return: + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + move $a0,$v0 + +.end bn_sqr_words_internal + +.align 5 +.globl bn_add_words +.ent bn_add_words +bn_add_words: + .set noreorder + bgtz $a3,bn_add_words_internal + move $v0,$zero + jr $ra + move $a0,$v0 +.end bn_add_words + +.align 5 +.ent bn_add_words_internal +bn_add_words_internal: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + li $minus4,-4 + and $at,$a3,$minus4 + beqz $at,.L_bn_add_words_tail + +.L_bn_add_words_loop: + $LD $t0,0($a1) + $LD $ta0,0($a2) + subu $a3,4 + $LD $t1,$BNSZ($a1) + and $at,$a3,$minus4 + $LD $t2,2*$BNSZ($a1) + $PTR_ADD $a2,4*$BNSZ + $LD $t3,3*$BNSZ($a1) + $PTR_ADD $a0,4*$BNSZ + $LD $ta1,-3*$BNSZ($a2) + $PTR_ADD $a1,4*$BNSZ + $LD $ta2,-2*$BNSZ($a2) + $LD $ta3,-$BNSZ($a2) + $ADDU $ta0,$t0 + sltu $t8,$ta0,$t0 + $ADDU $t0,$ta0,$v0 + sltu $v0,$t0,$ta0 + $ST $t0,-4*$BNSZ($a0) + $ADDU $v0,$t8 + + $ADDU $ta1,$t1 + sltu $t9,$ta1,$t1 + $ADDU $t1,$ta1,$v0 + sltu $v0,$t1,$ta1 + $ST $t1,-3*$BNSZ($a0) + $ADDU $v0,$t9 + + $ADDU $ta2,$t2 + sltu $t8,$ta2,$t2 + $ADDU $t2,$ta2,$v0 + sltu $v0,$t2,$ta2 + $ST $t2,-2*$BNSZ($a0) + $ADDU $v0,$t8 + + $ADDU $ta3,$t3 + sltu $t9,$ta3,$t3 + $ADDU $t3,$ta3,$v0 + sltu $v0,$t3,$ta3 + $ST $t3,-$BNSZ($a0) + + .set noreorder + bgtz $at,.L_bn_add_words_loop + $ADDU $v0,$t9 + + beqz $a3,.L_bn_add_words_return + nop + +.L_bn_add_words_tail: + .set reorder + $LD $t0,0($a1) + $LD $ta0,0($a2) + $ADDU $ta0,$t0 + subu $a3,1 + sltu $t8,$ta0,$t0 + $ADDU $t0,$ta0,$v0 + sltu $v0,$t0,$ta0 + $ST $t0,0($a0) + $ADDU $v0,$t8 + beqz $a3,.L_bn_add_words_return + + $LD $t1,$BNSZ($a1) + $LD $ta1,$BNSZ($a2) + $ADDU $ta1,$t1 + subu $a3,1 + sltu $t9,$ta1,$t1 + $ADDU $t1,$ta1,$v0 + sltu $v0,$t1,$ta1 + $ST $t1,$BNSZ($a0) + $ADDU $v0,$t9 + beqz $a3,.L_bn_add_words_return + + $LD $t2,2*$BNSZ($a1) + $LD $ta2,2*$BNSZ($a2) + $ADDU $ta2,$t2 + sltu $t8,$ta2,$t2 + $ADDU $t2,$ta2,$v0 + sltu $v0,$t2,$ta2 + $ST $t2,2*$BNSZ($a0) + $ADDU $v0,$t8 + +.L_bn_add_words_return: + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + move $a0,$v0 + +.end bn_add_words_internal + +.align 5 +.globl bn_sub_words +.ent bn_sub_words +bn_sub_words: + .set noreorder + bgtz $a3,bn_sub_words_internal + move $v0,$zero + jr $ra + move $a0,$zero +.end bn_sub_words + +.align 5 +.ent bn_sub_words_internal +bn_sub_words_internal: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + li $minus4,-4 + and $at,$a3,$minus4 + beqz $at,.L_bn_sub_words_tail + +.L_bn_sub_words_loop: + $LD $t0,0($a1) + $LD $ta0,0($a2) + subu $a3,4 + $LD $t1,$BNSZ($a1) + and $at,$a3,$minus4 + $LD $t2,2*$BNSZ($a1) + $PTR_ADD $a2,4*$BNSZ + $LD $t3,3*$BNSZ($a1) + $PTR_ADD $a0,4*$BNSZ + $LD $ta1,-3*$BNSZ($a2) + $PTR_ADD $a1,4*$BNSZ + $LD $ta2,-2*$BNSZ($a2) + $LD $ta3,-$BNSZ($a2) + sltu $t8,$t0,$ta0 + $SUBU $ta0,$t0,$ta0 + $SUBU $t0,$ta0,$v0 + sgtu $v0,$t0,$ta0 + $ST $t0,-4*$BNSZ($a0) + $ADDU $v0,$t8 + + sltu $t9,$t1,$ta1 + $SUBU $ta1,$t1,$ta1 + $SUBU $t1,$ta1,$v0 + sgtu $v0,$t1,$ta1 + $ST $t1,-3*$BNSZ($a0) + $ADDU $v0,$t9 + + + sltu $t8,$t2,$ta2 + $SUBU $ta2,$t2,$ta2 + $SUBU $t2,$ta2,$v0 + sgtu $v0,$t2,$ta2 + $ST $t2,-2*$BNSZ($a0) + $ADDU $v0,$t8 + + sltu $t9,$t3,$ta3 + $SUBU $ta3,$t3,$ta3 + $SUBU $t3,$ta3,$v0 + sgtu $v0,$t3,$ta3 + $ST $t3,-$BNSZ($a0) + + .set noreorder + bgtz $at,.L_bn_sub_words_loop + $ADDU $v0,$t9 + + beqz $a3,.L_bn_sub_words_return + nop + +.L_bn_sub_words_tail: + .set reorder + $LD $t0,0($a1) + $LD $ta0,0($a2) + subu $a3,1 + sltu $t8,$t0,$ta0 + $SUBU $ta0,$t0,$ta0 + $SUBU $t0,$ta0,$v0 + sgtu $v0,$t0,$ta0 + $ST $t0,0($a0) + $ADDU $v0,$t8 + beqz $a3,.L_bn_sub_words_return + + $LD $t1,$BNSZ($a1) + subu $a3,1 + $LD $ta1,$BNSZ($a2) + sltu $t9,$t1,$ta1 + $SUBU $ta1,$t1,$ta1 + $SUBU $t1,$ta1,$v0 + sgtu $v0,$t1,$ta1 + $ST $t1,$BNSZ($a0) + $ADDU $v0,$t9 + beqz $a3,.L_bn_sub_words_return + + $LD $t2,2*$BNSZ($a1) + $LD $ta2,2*$BNSZ($a2) + sltu $t8,$t2,$ta2 + $SUBU $ta2,$t2,$ta2 + $SUBU $t2,$ta2,$v0 + sgtu $v0,$t2,$ta2 + $ST $t2,2*$BNSZ($a0) + $ADDU $v0,$t8 + +.L_bn_sub_words_return: + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + move $a0,$v0 +.end bn_sub_words_internal + +.align 5 +.globl bn_div_3_words +.ent bn_div_3_words +bn_div_3_words: + .set noreorder + move $a3,$a0 # we know that bn_div_words does not + # touch $a3, $ta2, $ta3 and preserves $a2 + # so that we can save two arguments + # and return address in registers + # instead of stack:-) + + $LD $a0,($a3) + move $ta2,$a1 + bne $a0,$a2,bn_div_3_words_internal + $LD $a1,-$BNSZ($a3) + li $v0,-1 + jr $ra + move $a0,$v0 +.end bn_div_3_words + +.align 5 +.ent bn_div_3_words_internal +bn_div_3_words_internal: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + move $ta3,$ra + bal bn_div_words_internal + move $ra,$ta3 + $MULTU $ta2,$v0 + $LD $t2,-2*$BNSZ($a3) + move $ta0,$zero + mfhi $t1 + mflo $t0 + sltu $t8,$t1,$a1 +.L_bn_div_3_words_inner_loop: + bnez $t8,.L_bn_div_3_words_inner_loop_done + sgeu $at,$t2,$t0 + seq $t9,$t1,$a1 + and $at,$t9 + sltu $t3,$t0,$ta2 + $ADDU $a1,$a2 + $SUBU $t1,$t3 + $SUBU $t0,$ta2 + sltu $t8,$t1,$a1 + sltu $ta0,$a1,$a2 + or $t8,$ta0 + .set noreorder + beqz $at,.L_bn_div_3_words_inner_loop + $SUBU $v0,1 + $ADDU $v0,1 + .set reorder +.L_bn_div_3_words_inner_loop_done: + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + move $a0,$v0 +.end bn_div_3_words_internal + +.align 5 +.globl bn_div_words +.ent bn_div_words +bn_div_words: + .set noreorder + bnez $a2,bn_div_words_internal + li $v0,-1 # I would rather signal div-by-zero + # which can be done with 'break 7' + jr $ra + move $a0,$v0 +.end bn_div_words + +.align 5 +.ent bn_div_words_internal +bn_div_words_internal: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + move $v1,$zero + bltz $a2,.L_bn_div_words_body + move $t9,$v1 + $SLL $a2,1 + bgtz $a2,.-4 + addu $t9,1 + + .set reorder + negu $t1,$t9 + li $t2,-1 + $SLL $t2,$t1 + and $t2,$a0 + $SRL $at,$a1,$t1 + .set noreorder + beqz $t2,.+12 + nop + break 6 # signal overflow + .set reorder + $SLL $a0,$t9 + $SLL $a1,$t9 + or $a0,$at +___ +$QT=$ta0; +$HH=$ta1; +$DH=$v1; +$code.=<<___; +.L_bn_div_words_body: + $SRL $DH,$a2,4*$BNSZ # bits + sgeu $at,$a0,$a2 + .set noreorder + beqz $at,.+12 + nop + $SUBU $a0,$a2 + .set reorder + + li $QT,-1 + $SRL $HH,$a0,4*$BNSZ # bits + $SRL $QT,4*$BNSZ # q=0xffffffff + beq $DH,$HH,.L_bn_div_words_skip_div1 + $DIVU $zero,$a0,$DH + mflo $QT +.L_bn_div_words_skip_div1: + $MULTU $a2,$QT + $SLL $t3,$a0,4*$BNSZ # bits + $SRL $at,$a1,4*$BNSZ # bits + or $t3,$at + mflo $t0 + mfhi $t1 +.L_bn_div_words_inner_loop1: + sltu $t2,$t3,$t0 + seq $t8,$HH,$t1 + sltu $at,$HH,$t1 + and $t2,$t8 + sltu $v0,$t0,$a2 + or $at,$t2 + .set noreorder + beqz $at,.L_bn_div_words_inner_loop1_done + $SUBU $t1,$v0 + $SUBU $t0,$a2 + b .L_bn_div_words_inner_loop1 + $SUBU $QT,1 + .set reorder +.L_bn_div_words_inner_loop1_done: + + $SLL $a1,4*$BNSZ # bits + $SUBU $a0,$t3,$t0 + $SLL $v0,$QT,4*$BNSZ # bits + + li $QT,-1 + $SRL $HH,$a0,4*$BNSZ # bits + $SRL $QT,4*$BNSZ # q=0xffffffff + beq $DH,$HH,.L_bn_div_words_skip_div2 + $DIVU $zero,$a0,$DH + mflo $QT +.L_bn_div_words_skip_div2: + $MULTU $a2,$QT + $SLL $t3,$a0,4*$BNSZ # bits + $SRL $at,$a1,4*$BNSZ # bits + or $t3,$at + mflo $t0 + mfhi $t1 +.L_bn_div_words_inner_loop2: + sltu $t2,$t3,$t0 + seq $t8,$HH,$t1 + sltu $at,$HH,$t1 + and $t2,$t8 + sltu $v1,$t0,$a2 + or $at,$t2 + .set noreorder + beqz $at,.L_bn_div_words_inner_loop2_done + $SUBU $t1,$v1 + $SUBU $t0,$a2 + b .L_bn_div_words_inner_loop2 + $SUBU $QT,1 + .set reorder +.L_bn_div_words_inner_loop2_done: + + $SUBU $a0,$t3,$t0 + or $v0,$QT + $SRL $v1,$a0,$t9 # $v1 contains remainder if anybody wants it + $SRL $a2,$t9 # restore $a2 + + .set noreorder + move $a1,$v1 +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + move $a0,$v0 +.end bn_div_words_internal +___ +undef $HH; undef $QT; undef $DH; + +($a_0,$a_1,$a_2,$a_3)=($t0,$t1,$t2,$t3); +($b_0,$b_1,$b_2,$b_3)=($ta0,$ta1,$ta2,$ta3); + +($a_4,$a_5,$a_6,$a_7)=($s0,$s2,$s4,$a1); # once we load a[7], no use for $a1 +($b_4,$b_5,$b_6,$b_7)=($s1,$s3,$s5,$a2); # once we load b[7], no use for $a2 + +($t_1,$t_2,$c_1,$c_2,$c_3)=($t8,$t9,$v0,$v1,$a3); + +$code.=<<___; + +.align 5 +.globl bn_mul_comba8 +.ent bn_mul_comba8 +bn_mul_comba8: + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,12*$SZREG,$ra + .mask 0x803ff008,-$SZREG + $PTR_SUB $sp,12*$SZREG + $REG_S $ra,11*$SZREG($sp) + $REG_S $s5,10*$SZREG($sp) + $REG_S $s4,9*$SZREG($sp) + $REG_S $s3,8*$SZREG($sp) + $REG_S $s2,7*$SZREG($sp) + $REG_S $s1,6*$SZREG($sp) + $REG_S $s0,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___ if ($flavour !~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x003f0000,-$SZREG + $PTR_SUB $sp,6*$SZREG + $REG_S $s5,5*$SZREG($sp) + $REG_S $s4,4*$SZREG($sp) + $REG_S $s3,3*$SZREG($sp) + $REG_S $s2,2*$SZREG($sp) + $REG_S $s1,1*$SZREG($sp) + $REG_S $s0,0*$SZREG($sp) +___ +$code.=<<___; + + .set reorder + $LD $a_0,0($a1) # If compiled with -mips3 option on + # R5000 box assembler barks on this + # 1ine with "should not have mult/div + # as last instruction in bb (R10K + # bug)" warning. If anybody out there + # has a clue about how to circumvent + # this do send me a note. + # + + $LD $b_0,0($a2) + $LD $a_1,$BNSZ($a1) + $LD $a_2,2*$BNSZ($a1) + $MULTU $a_0,$b_0 # mul_add_c(a[0],b[0],c1,c2,c3); + $LD $a_3,3*$BNSZ($a1) + $LD $b_1,$BNSZ($a2) + $LD $b_2,2*$BNSZ($a2) + $LD $b_3,3*$BNSZ($a2) + mflo $c_1 + mfhi $c_2 + + $LD $a_4,4*$BNSZ($a1) + $LD $a_5,5*$BNSZ($a1) + $MULTU $a_0,$b_1 # mul_add_c(a[0],b[1],c2,c3,c1); + $LD $a_6,6*$BNSZ($a1) + $LD $a_7,7*$BNSZ($a1) + $LD $b_4,4*$BNSZ($a2) + $LD $b_5,5*$BNSZ($a2) + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_1,$b_0 # mul_add_c(a[1],b[0],c2,c3,c1); + $ADDU $c_3,$t_2,$at + $LD $b_6,6*$BNSZ($a2) + $LD $b_7,7*$BNSZ($a2) + $ST $c_1,0($a0) # r[0]=c1; + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_2,$b_0 # mul_add_c(a[2],b[0],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $c_1,$c_3,$t_2 + $ST $c_2,$BNSZ($a0) # r[1]=c2; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_1,$b_1 # mul_add_c(a[1],b[1],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_0,$b_2 # mul_add_c(a[0],b[2],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $c_2,$c_1,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_0,$b_3 # mul_add_c(a[0],b[3],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,2*$BNSZ($a0) # r[2]=c3; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_1,$b_2 # mul_add_c(a[1],b[2],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $c_3,$c_2,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_2,$b_1 # mul_add_c(a[2],b[1],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_3,$b_0 # mul_add_c(a[3],b[0],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_4,$b_0 # mul_add_c(a[4],b[0],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,3*$BNSZ($a0) # r[3]=c1; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_3,$b_1 # mul_add_c(a[3],b[1],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $c_1,$c_3,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_2,$b_2 # mul_add_c(a[2],b[2],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_1,$b_3 # mul_add_c(a[1],b[3],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_0,$b_4 # mul_add_c(a[0],b[4],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_0,$b_5 # mul_add_c(a[0],b[5],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,4*$BNSZ($a0) # r[4]=c2; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_1,$b_4 # mul_add_c(a[1],b[4],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $c_2,$c_1,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_2,$b_3 # mul_add_c(a[2],b[3],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_3,$b_2 # mul_add_c(a[3],b[2],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_4,$b_1 # mul_add_c(a[4],b[1],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_5,$b_0 # mul_add_c(a[5],b[0],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_6,$b_0 # mul_add_c(a[6],b[0],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,5*$BNSZ($a0) # r[5]=c3; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_5,$b_1 # mul_add_c(a[5],b[1],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $c_3,$c_2,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_4,$b_2 # mul_add_c(a[4],b[2],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_3,$b_3 # mul_add_c(a[3],b[3],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_2,$b_4 # mul_add_c(a[2],b[4],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_1,$b_5 # mul_add_c(a[1],b[5],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_0,$b_6 # mul_add_c(a[0],b[6],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_0,$b_7 # mul_add_c(a[0],b[7],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,6*$BNSZ($a0) # r[6]=c1; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_1,$b_6 # mul_add_c(a[1],b[6],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $c_1,$c_3,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_2,$b_5 # mul_add_c(a[2],b[5],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_3,$b_4 # mul_add_c(a[3],b[4],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_4,$b_3 # mul_add_c(a[4],b[3],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_5,$b_2 # mul_add_c(a[5],b[2],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_6,$b_1 # mul_add_c(a[6],b[1],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_7,$b_0 # mul_add_c(a[7],b[0],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_7,$b_1 # mul_add_c(a[7],b[1],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,7*$BNSZ($a0) # r[7]=c2; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_6,$b_2 # mul_add_c(a[6],b[2],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $c_2,$c_1,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_5,$b_3 # mul_add_c(a[5],b[3],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_4,$b_4 # mul_add_c(a[4],b[4],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_3,$b_5 # mul_add_c(a[3],b[5],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_2,$b_6 # mul_add_c(a[2],b[6],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_1,$b_7 # mul_add_c(a[1],b[7],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_2,$b_7 # mul_add_c(a[2],b[7],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,8*$BNSZ($a0) # r[8]=c3; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_3,$b_6 # mul_add_c(a[3],b[6],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $c_3,$c_2,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_4,$b_5 # mul_add_c(a[4],b[5],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_5,$b_4 # mul_add_c(a[5],b[4],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_6,$b_3 # mul_add_c(a[6],b[3],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_7,$b_2 # mul_add_c(a[7],b[2],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_7,$b_3 # mul_add_c(a[7],b[3],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,9*$BNSZ($a0) # r[9]=c1; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_6,$b_4 # mul_add_c(a[6],b[4],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $c_1,$c_3,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_5,$b_5 # mul_add_c(a[5],b[5],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_4,$b_6 # mul_add_c(a[4],b[6],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_3,$b_7 # mul_add_c(a[3],b[7],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_4,$b_7 # mul_add_c(a[4],b[7],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,10*$BNSZ($a0) # r[10]=c2; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_5,$b_6 # mul_add_c(a[5],b[6],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $c_2,$c_1,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_6,$b_5 # mul_add_c(a[6],b[5],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_7,$b_4 # mul_add_c(a[7],b[4],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_7,$b_5 # mul_add_c(a[7],b[5],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,11*$BNSZ($a0) # r[11]=c3; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_6,$b_6 # mul_add_c(a[6],b[6],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $c_3,$c_2,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_5,$b_7 # mul_add_c(a[5],b[7],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_6,$b_7 # mul_add_c(a[6],b[7],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,12*$BNSZ($a0) # r[12]=c1; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_7,$b_6 # mul_add_c(a[7],b[6],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $c_1,$c_3,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_7,$b_7 # mul_add_c(a[7],b[7],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,13*$BNSZ($a0) # r[13]=c2; + + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + $ST $c_3,14*$BNSZ($a0) # r[14]=c3; + $ST $c_1,15*$BNSZ($a0) # r[15]=c1; + + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $s5,10*$SZREG($sp) + $REG_L $s4,9*$SZREG($sp) + $REG_L $s3,8*$SZREG($sp) + $REG_L $s2,7*$SZREG($sp) + $REG_L $s1,6*$SZREG($sp) + $REG_L $s0,5*$SZREG($sp) + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + jr $ra + $PTR_ADD $sp,12*$SZREG +___ +$code.=<<___ if ($flavour !~ /nubi/i); + $REG_L $s5,5*$SZREG($sp) + $REG_L $s4,4*$SZREG($sp) + $REG_L $s3,3*$SZREG($sp) + $REG_L $s2,2*$SZREG($sp) + $REG_L $s1,1*$SZREG($sp) + $REG_L $s0,0*$SZREG($sp) + jr $ra + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; +.end bn_mul_comba8 + +.align 5 +.globl bn_mul_comba4 +.ent bn_mul_comba4 +bn_mul_comba4: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + $LD $a_0,0($a1) + $LD $b_0,0($a2) + $LD $a_1,$BNSZ($a1) + $LD $a_2,2*$BNSZ($a1) + $MULTU $a_0,$b_0 # mul_add_c(a[0],b[0],c1,c2,c3); + $LD $a_3,3*$BNSZ($a1) + $LD $b_1,$BNSZ($a2) + $LD $b_2,2*$BNSZ($a2) + $LD $b_3,3*$BNSZ($a2) + mflo $c_1 + mfhi $c_2 + $ST $c_1,0($a0) + + $MULTU $a_0,$b_1 # mul_add_c(a[0],b[1],c2,c3,c1); + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_1,$b_0 # mul_add_c(a[1],b[0],c2,c3,c1); + $ADDU $c_3,$t_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_2,$b_0 # mul_add_c(a[2],b[0],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $c_1,$c_3,$t_2 + $ST $c_2,$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_1,$b_1 # mul_add_c(a[1],b[1],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_0,$b_2 # mul_add_c(a[0],b[2],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $c_2,$c_1,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_0,$b_3 # mul_add_c(a[0],b[3],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,2*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_1,$b_2 # mul_add_c(a[1],b[2],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $c_3,$c_2,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_2,$b_1 # mul_add_c(a[2],b[1],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_3,$b_0 # mul_add_c(a[3],b[0],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_3,$b_1 # mul_add_c(a[3],b[1],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,3*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_2,$b_2 # mul_add_c(a[2],b[2],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $c_1,$c_3,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_1,$b_3 # mul_add_c(a[1],b[3],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_2,$b_3 # mul_add_c(a[2],b[3],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,4*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_3,$b_2 # mul_add_c(a[3],b[2],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $c_2,$c_1,$t_2 + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_3,$b_3 # mul_add_c(a[3],b[3],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,5*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + $ST $c_1,6*$BNSZ($a0) + $ST $c_2,7*$BNSZ($a0) + + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + nop +.end bn_mul_comba4 +___ + +($a_4,$a_5,$a_6,$a_7)=($b_0,$b_1,$b_2,$b_3); + +$code.=<<___; + +.align 5 +.globl bn_sqr_comba8 +.ent bn_sqr_comba8 +bn_sqr_comba8: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + $LD $a_0,0($a1) + $LD $a_1,$BNSZ($a1) + $LD $a_2,2*$BNSZ($a1) + $LD $a_3,3*$BNSZ($a1) + + $MULTU $a_0,$a_0 # mul_add_c(a[0],b[0],c1,c2,c3); + $LD $a_4,4*$BNSZ($a1) + $LD $a_5,5*$BNSZ($a1) + $LD $a_6,6*$BNSZ($a1) + $LD $a_7,7*$BNSZ($a1) + mflo $c_1 + mfhi $c_2 + $ST $c_1,0($a0) + + $MULTU $a_0,$a_1 # mul_add_c2(a[0],b[1],c2,c3,c1); + mflo $t_1 + mfhi $t_2 + slt $c_1,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_2,$a_0 # mul_add_c2(a[2],b[0],c3,c1,c2); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $c_3,$t_2,$at + $ST $c_2,$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_2,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_1,$a_1 # mul_add_c(a[1],b[1],c3,c1,c2); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_0,$a_3 # mul_add_c2(a[0],b[3],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,2*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_3,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_1,$a_2 # mul_add_c2(a[1],b[2],c1,c2,c3); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_3,$at + $MULTU $a_4,$a_0 # mul_add_c2(a[4],b[0],c2,c3,c1); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,3*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_1,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_3,$a_1 # mul_add_c2(a[3],b[1],c2,c3,c1); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_1,$at + $MULTU $a_2,$a_2 # mul_add_c(a[2],b[2],c2,c3,c1); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_0,$a_5 # mul_add_c2(a[0],b[5],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,4*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_2,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_1,$a_4 # mul_add_c2(a[1],b[4],c3,c1,c2); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_2,$at + $MULTU $a_2,$a_3 # mul_add_c2(a[2],b[3],c3,c1,c2); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $MULTU $a_6,$a_0 # mul_add_c2(a[6],b[0],c1,c2,c3); + $ADDU $c_2,$at + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,5*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_3,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_5,$a_1 # mul_add_c2(a[5],b[1],c1,c2,c3); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_3,$at + $MULTU $a_4,$a_2 # mul_add_c2(a[4],b[2],c1,c2,c3); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_3,$at + $MULTU $a_3,$a_3 # mul_add_c(a[3],b[3],c1,c2,c3); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_0,$a_7 # mul_add_c2(a[0],b[7],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,6*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_1,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_1,$a_6 # mul_add_c2(a[1],b[6],c2,c3,c1); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_1,$at + $MULTU $a_2,$a_5 # mul_add_c2(a[2],b[5],c2,c3,c1); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_1,$at + $MULTU $a_3,$a_4 # mul_add_c2(a[3],b[4],c2,c3,c1); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_1,$at + $MULTU $a_7,$a_1 # mul_add_c2(a[7],b[1],c3,c1,c2); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,7*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_2,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_6,$a_2 # mul_add_c2(a[6],b[2],c3,c1,c2); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_2,$at + $MULTU $a_5,$a_3 # mul_add_c2(a[5],b[3],c3,c1,c2); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_2,$at + $MULTU $a_4,$a_4 # mul_add_c(a[4],b[4],c3,c1,c2); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_2,$a_7 # mul_add_c2(a[2],b[7],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,8*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_3,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_3,$a_6 # mul_add_c2(a[3],b[6],c1,c2,c3); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_3,$at + $MULTU $a_4,$a_5 # mul_add_c2(a[4],b[5],c1,c2,c3); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_3,$at + $MULTU $a_7,$a_3 # mul_add_c2(a[7],b[3],c2,c3,c1); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,9*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_1,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_6,$a_4 # mul_add_c2(a[6],b[4],c2,c3,c1); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_1,$at + $MULTU $a_5,$a_5 # mul_add_c(a[5],b[5],c2,c3,c1); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_4,$a_7 # mul_add_c2(a[4],b[7],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,10*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_2,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_5,$a_6 # mul_add_c2(a[5],b[6],c3,c1,c2); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_2,$at + $MULTU $a_7,$a_5 # mul_add_c2(a[7],b[5],c1,c2,c3); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,11*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_3,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_6,$a_6 # mul_add_c(a[6],b[6],c1,c2,c3); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $MULTU $a_6,$a_7 # mul_add_c2(a[6],b[7],c2,c3,c1); + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,12*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_1,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_7,$a_7 # mul_add_c(a[7],b[7],c3,c1,c2); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,13*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + $ST $c_3,14*$BNSZ($a0) + $ST $c_1,15*$BNSZ($a0) + + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + nop +.end bn_sqr_comba8 + +.align 5 +.globl bn_sqr_comba4 +.ent bn_sqr_comba4 +bn_sqr_comba4: +___ +$code.=<<___ if ($flavour =~ /nubi/i); + .frame $sp,6*$SZREG,$ra + .mask 0x8000f008,-$SZREG + .set noreorder + $PTR_SUB $sp,6*$SZREG + $REG_S $ra,5*$SZREG($sp) + $REG_S $t3,4*$SZREG($sp) + $REG_S $t2,3*$SZREG($sp) + $REG_S $t1,2*$SZREG($sp) + $REG_S $t0,1*$SZREG($sp) + $REG_S $gp,0*$SZREG($sp) +___ +$code.=<<___; + .set reorder + $LD $a_0,0($a1) + $LD $a_1,$BNSZ($a1) + $MULTU $a_0,$a_0 # mul_add_c(a[0],b[0],c1,c2,c3); + $LD $a_2,2*$BNSZ($a1) + $LD $a_3,3*$BNSZ($a1) + mflo $c_1 + mfhi $c_2 + $ST $c_1,0($a0) + + $MULTU $a_0,$a_1 # mul_add_c2(a[0],b[1],c2,c3,c1); + mflo $t_1 + mfhi $t_2 + slt $c_1,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_2,$a_0 # mul_add_c2(a[2],b[0],c3,c1,c2); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $c_3,$t_2,$at + $ST $c_2,$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_2,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_1,$a_1 # mul_add_c(a[1],b[1],c3,c1,c2); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $MULTU $a_0,$a_3 # mul_add_c2(a[0],b[3],c1,c2,c3); + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,2*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_3,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_1,$a_2 # mul_add_c(a2[1],b[2],c1,c2,c3); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + mflo $t_1 + mfhi $t_2 + slt $at,$t_2,$zero + $ADDU $c_3,$at + $MULTU $a_3,$a_1 # mul_add_c2(a[3],b[1],c2,c3,c1); + $SLL $t_2,1 + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + sltu $at,$c_2,$t_2 + $ADDU $c_3,$at + $ST $c_1,3*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_1,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_2,$a_2 # mul_add_c(a[2],b[2],c2,c3,c1); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + mflo $t_1 + mfhi $t_2 + $ADDU $c_2,$t_1 + sltu $at,$c_2,$t_1 + $MULTU $a_2,$a_3 # mul_add_c2(a[2],b[3],c3,c1,c2); + $ADDU $t_2,$at + $ADDU $c_3,$t_2 + sltu $at,$c_3,$t_2 + $ADDU $c_1,$at + $ST $c_2,4*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + slt $c_2,$t_2,$zero + $SLL $t_2,1 + $MULTU $a_3,$a_3 # mul_add_c(a[3],b[3],c1,c2,c3); + slt $a2,$t_1,$zero + $ADDU $t_2,$a2 + $SLL $t_1,1 + $ADDU $c_3,$t_1 + sltu $at,$c_3,$t_1 + $ADDU $t_2,$at + $ADDU $c_1,$t_2 + sltu $at,$c_1,$t_2 + $ADDU $c_2,$at + $ST $c_3,5*$BNSZ($a0) + + mflo $t_1 + mfhi $t_2 + $ADDU $c_1,$t_1 + sltu $at,$c_1,$t_1 + $ADDU $t_2,$at + $ADDU $c_2,$t_2 + $ST $c_1,6*$BNSZ($a0) + $ST $c_2,7*$BNSZ($a0) + + .set noreorder +___ +$code.=<<___ if ($flavour =~ /nubi/i); + $REG_L $t3,4*$SZREG($sp) + $REG_L $t2,3*$SZREG($sp) + $REG_L $t1,2*$SZREG($sp) + $REG_L $t0,1*$SZREG($sp) + $REG_L $gp,0*$SZREG($sp) + $PTR_ADD $sp,6*$SZREG +___ +$code.=<<___; + jr $ra + nop +.end bn_sqr_comba4 +___ +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips3-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips3-mont.pl new file mode 100644 index 00000000..8f9156e0 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips3-mont.pl @@ -0,0 +1,327 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# This module doesn't present direct interest for OpenSSL, because it +# doesn't provide better performance for longer keys. While 512-bit +# RSA private key operations are 40% faster, 1024-bit ones are hardly +# faster at all, while longer key operations are slower by up to 20%. +# It might be of interest to embedded system developers though, as +# it's smaller than 1KB, yet offers ~3x improvement over compiler +# generated code. +# +# The module targets N32 and N64 MIPS ABIs and currently is a bit +# IRIX-centric, i.e. is likely to require adaptation for other OSes. + +# int bn_mul_mont( +$rp="a0"; # BN_ULONG *rp, +$ap="a1"; # const BN_ULONG *ap, +$bp="a2"; # const BN_ULONG *bp, +$np="a3"; # const BN_ULONG *np, +$n0="a4"; # const BN_ULONG *n0, +$num="a5"; # int num); + +$lo0="a6"; +$hi0="a7"; +$lo1="v0"; +$hi1="v1"; +$aj="t0"; +$bi="t1"; +$nj="t2"; +$tp="t3"; +$alo="s0"; +$ahi="s1"; +$nlo="s2"; +$nhi="s3"; +$tj="s4"; +$i="s5"; +$j="s6"; +$fp="t8"; +$m1="t9"; + +$FRAME=8*(2+8); + +$code=<<___; +#include +#include + +.text + +.set noat +.set reorder + +.align 5 +.globl bn_mul_mont +.ent bn_mul_mont +bn_mul_mont: + .set noreorder + PTR_SUB sp,64 + move $fp,sp + .frame $fp,64,ra + slt AT,$num,4 + li v0,0 + beqzl AT,.Lproceed + nop + jr ra + PTR_ADD sp,$fp,64 + .set reorder +.align 5 +.Lproceed: + ld $n0,0($n0) + ld $bi,0($bp) # bp[0] + ld $aj,0($ap) # ap[0] + ld $nj,0($np) # np[0] + PTR_SUB sp,16 # place for two extra words + sll $num,3 + li AT,-4096 + PTR_SUB sp,$num + and sp,AT + + sd s0,0($fp) + sd s1,8($fp) + sd s2,16($fp) + sd s3,24($fp) + sd s4,32($fp) + sd s5,40($fp) + sd s6,48($fp) + sd s7,56($fp) + + dmultu $aj,$bi + ld $alo,8($ap) + ld $nlo,8($np) + mflo $lo0 + mfhi $hi0 + dmultu $lo0,$n0 + mflo $m1 + + dmultu $alo,$bi + mflo $alo + mfhi $ahi + + dmultu $nj,$m1 + mflo $lo1 + mfhi $hi1 + dmultu $nlo,$m1 + daddu $lo1,$lo0 + sltu AT,$lo1,$lo0 + daddu $hi1,AT + mflo $nlo + mfhi $nhi + + move $tp,sp + li $j,16 +.align 4 +.L1st: + .set noreorder + PTR_ADD $aj,$ap,$j + ld $aj,($aj) + PTR_ADD $nj,$np,$j + ld $nj,($nj) + + dmultu $aj,$bi + daddu $lo0,$alo,$hi0 + daddu $lo1,$nlo,$hi1 + sltu AT,$lo0,$hi0 + sltu s7,$lo1,$hi1 + daddu $hi0,$ahi,AT + daddu $hi1,$nhi,s7 + mflo $alo + mfhi $ahi + + daddu $lo1,$lo0 + sltu AT,$lo1,$lo0 + dmultu $nj,$m1 + daddu $hi1,AT + addu $j,8 + sd $lo1,($tp) + sltu s7,$j,$num + mflo $nlo + mfhi $nhi + + bnez s7,.L1st + PTR_ADD $tp,8 + .set reorder + + daddu $lo0,$alo,$hi0 + sltu AT,$lo0,$hi0 + daddu $hi0,$ahi,AT + + daddu $lo1,$nlo,$hi1 + sltu s7,$lo1,$hi1 + daddu $hi1,$nhi,s7 + daddu $lo1,$lo0 + sltu AT,$lo1,$lo0 + daddu $hi1,AT + + sd $lo1,($tp) + + daddu $hi1,$hi0 + sltu AT,$hi1,$hi0 + sd $hi1,8($tp) + sd AT,16($tp) + + li $i,8 +.align 4 +.Louter: + PTR_ADD $bi,$bp,$i + ld $bi,($bi) + ld $aj,($ap) + ld $alo,8($ap) + ld $tj,(sp) + + dmultu $aj,$bi + ld $nj,($np) + ld $nlo,8($np) + mflo $lo0 + mfhi $hi0 + daddu $lo0,$tj + dmultu $lo0,$n0 + sltu AT,$lo0,$tj + daddu $hi0,AT + mflo $m1 + + dmultu $alo,$bi + mflo $alo + mfhi $ahi + + dmultu $nj,$m1 + mflo $lo1 + mfhi $hi1 + + dmultu $nlo,$m1 + daddu $lo1,$lo0 + sltu AT,$lo1,$lo0 + daddu $hi1,AT + mflo $nlo + mfhi $nhi + + move $tp,sp + li $j,16 + ld $tj,8($tp) +.align 4 +.Linner: + .set noreorder + PTR_ADD $aj,$ap,$j + ld $aj,($aj) + PTR_ADD $nj,$np,$j + ld $nj,($nj) + + dmultu $aj,$bi + daddu $lo0,$alo,$hi0 + daddu $lo1,$nlo,$hi1 + sltu AT,$lo0,$hi0 + sltu s7,$lo1,$hi1 + daddu $hi0,$ahi,AT + daddu $hi1,$nhi,s7 + mflo $alo + mfhi $ahi + + daddu $lo0,$tj + addu $j,8 + dmultu $nj,$m1 + sltu AT,$lo0,$tj + daddu $lo1,$lo0 + daddu $hi0,AT + sltu s7,$lo1,$lo0 + ld $tj,16($tp) + daddu $hi1,s7 + sltu AT,$j,$num + mflo $nlo + mfhi $nhi + sd $lo1,($tp) + bnez AT,.Linner + PTR_ADD $tp,8 + .set reorder + + daddu $lo0,$alo,$hi0 + sltu AT,$lo0,$hi0 + daddu $hi0,$ahi,AT + daddu $lo0,$tj + sltu s7,$lo0,$tj + daddu $hi0,s7 + + ld $tj,16($tp) + daddu $lo1,$nlo,$hi1 + sltu AT,$lo1,$hi1 + daddu $hi1,$nhi,AT + daddu $lo1,$lo0 + sltu s7,$lo1,$lo0 + daddu $hi1,s7 + sd $lo1,($tp) + + daddu $lo1,$hi1,$hi0 + sltu $hi1,$lo1,$hi0 + daddu $lo1,$tj + sltu AT,$lo1,$tj + daddu $hi1,AT + sd $lo1,8($tp) + sd $hi1,16($tp) + + addu $i,8 + sltu s7,$i,$num + bnez s7,.Louter + + .set noreorder + PTR_ADD $tj,sp,$num # &tp[num] + move $tp,sp + move $ap,sp + li $hi0,0 # clear borrow bit + +.align 4 +.Lsub: ld $lo0,($tp) + ld $lo1,($np) + PTR_ADD $tp,8 + PTR_ADD $np,8 + dsubu $lo1,$lo0,$lo1 # tp[i]-np[i] + sgtu AT,$lo1,$lo0 + dsubu $lo0,$lo1,$hi0 + sgtu $hi0,$lo0,$lo1 + sd $lo0,($rp) + or $hi0,AT + sltu AT,$tp,$tj + bnez AT,.Lsub + PTR_ADD $rp,8 + + dsubu $hi0,$hi1,$hi0 # handle upmost overflow bit + move $tp,sp + PTR_SUB $rp,$num # restore rp + not $hi1,$hi0 + + and $ap,$hi0,sp + and $bp,$hi1,$rp + or $ap,$ap,$bp # ap=borrow?tp:rp + +.align 4 +.Lcopy: ld $aj,($ap) + PTR_ADD $ap,8 + PTR_ADD $tp,8 + sd zero,-8($tp) + sltu AT,$tp,$tj + sd $aj,($rp) + bnez AT,.Lcopy + PTR_ADD $rp,8 + + ld s0,0($fp) + ld s1,8($fp) + ld s2,16($fp) + ld s3,24($fp) + ld s4,32($fp) + ld s5,40($fp) + ld s6,48($fp) + ld s7,56($fp) + li v0,1 + jr ra + PTR_ADD sp,$fp,64 + .set reorder +END(bn_mul_mont) +.rdata +.asciiz "Montgomery Multiplication for MIPS III/IV, CRYPTOGAMS by " +___ + +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips3.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips3.S new file mode 100644 index 00000000..dca4105c --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/mips3.S @@ -0,0 +1,2201 @@ +.rdata +.asciiz "mips3.s, Version 1.1" +.asciiz "MIPS III/IV ISA artwork by Andy Polyakov " + +/* + * ==================================================================== + * Written by Andy Polyakov for the OpenSSL + * project. + * + * Rights for redistribution and usage in source and binary forms are + * granted according to the OpenSSL license. Warranty of any kind is + * disclaimed. + * ==================================================================== + */ + +/* + * This is my modest contributon to the OpenSSL project (see + * http://www.openssl.org/ for more information about it) and is + * a drop-in MIPS III/IV ISA replacement for crypto/bn/bn_asm.c + * module. For updates see http://fy.chalmers.se/~appro/hpe/. + * + * The module is designed to work with either of the "new" MIPS ABI(5), + * namely N32 or N64, offered by IRIX 6.x. It's not ment to work under + * IRIX 5.x not only because it doesn't support new ABIs but also + * because 5.x kernels put R4x00 CPU into 32-bit mode and all those + * 64-bit instructions (daddu, dmultu, etc.) found below gonna only + * cause illegal instruction exception:-( + * + * In addition the code depends on preprocessor flags set up by MIPSpro + * compiler driver (either as or cc) and therefore (probably?) can't be + * compiled by the GNU assembler. GNU C driver manages fine though... + * I mean as long as -mmips-as is specified or is the default option, + * because then it simply invokes /usr/bin/as which in turn takes + * perfect care of the preprocessor definitions. Another neat feature + * offered by the MIPSpro assembler is an optimization pass. This gave + * me the opportunity to have the code looking more regular as all those + * architecture dependent instruction rescheduling details were left to + * the assembler. Cool, huh? + * + * Performance improvement is astonishing! 'apps/openssl speed rsa dsa' + * goes way over 3 times faster! + * + * + */ +#include +#include + +#if _MIPS_ISA>=4 +#define MOVNZ(cond,dst,src) \ + movn dst,src,cond +#else +#define MOVNZ(cond,dst,src) \ + .set noreorder; \ + bnezl cond,.+8; \ + move dst,src; \ + .set reorder +#endif + +.text + +.set noat +.set reorder + +#define MINUS4 v1 + +.align 5 +LEAF(bn_mul_add_words) + .set noreorder + bgtzl a2,.L_bn_mul_add_words_proceed + ld t0,0(a1) + jr ra + move v0,zero + .set reorder + +.L_bn_mul_add_words_proceed: + li MINUS4,-4 + and ta0,a2,MINUS4 + move v0,zero + beqz ta0,.L_bn_mul_add_words_tail + +.L_bn_mul_add_words_loop: + dmultu t0,a3 + ld t1,0(a0) + ld t2,8(a1) + ld t3,8(a0) + ld ta0,16(a1) + ld ta1,16(a0) + daddu t1,v0 + sltu v0,t1,v0 /* All manuals say it "compares 32-bit + * values", but it seems to work fine + * even on 64-bit registers. */ + mflo AT + mfhi t0 + daddu t1,AT + daddu v0,t0 + sltu AT,t1,AT + sd t1,0(a0) + daddu v0,AT + + dmultu t2,a3 + ld ta2,24(a1) + ld ta3,24(a0) + daddu t3,v0 + sltu v0,t3,v0 + mflo AT + mfhi t2 + daddu t3,AT + daddu v0,t2 + sltu AT,t3,AT + sd t3,8(a0) + daddu v0,AT + + dmultu ta0,a3 + subu a2,4 + PTR_ADD a0,32 + PTR_ADD a1,32 + daddu ta1,v0 + sltu v0,ta1,v0 + mflo AT + mfhi ta0 + daddu ta1,AT + daddu v0,ta0 + sltu AT,ta1,AT + sd ta1,-16(a0) + daddu v0,AT + + + dmultu ta2,a3 + and ta0,a2,MINUS4 + daddu ta3,v0 + sltu v0,ta3,v0 + mflo AT + mfhi ta2 + daddu ta3,AT + daddu v0,ta2 + sltu AT,ta3,AT + sd ta3,-8(a0) + daddu v0,AT + .set noreorder + bgtzl ta0,.L_bn_mul_add_words_loop + ld t0,0(a1) + + bnezl a2,.L_bn_mul_add_words_tail + ld t0,0(a1) + .set reorder + +.L_bn_mul_add_words_return: + jr ra + +.L_bn_mul_add_words_tail: + dmultu t0,a3 + ld t1,0(a0) + subu a2,1 + daddu t1,v0 + sltu v0,t1,v0 + mflo AT + mfhi t0 + daddu t1,AT + daddu v0,t0 + sltu AT,t1,AT + sd t1,0(a0) + daddu v0,AT + beqz a2,.L_bn_mul_add_words_return + + ld t0,8(a1) + dmultu t0,a3 + ld t1,8(a0) + subu a2,1 + daddu t1,v0 + sltu v0,t1,v0 + mflo AT + mfhi t0 + daddu t1,AT + daddu v0,t0 + sltu AT,t1,AT + sd t1,8(a0) + daddu v0,AT + beqz a2,.L_bn_mul_add_words_return + + ld t0,16(a1) + dmultu t0,a3 + ld t1,16(a0) + daddu t1,v0 + sltu v0,t1,v0 + mflo AT + mfhi t0 + daddu t1,AT + daddu v0,t0 + sltu AT,t1,AT + sd t1,16(a0) + daddu v0,AT + jr ra +END(bn_mul_add_words) + +.align 5 +LEAF(bn_mul_words) + .set noreorder + bgtzl a2,.L_bn_mul_words_proceed + ld t0,0(a1) + jr ra + move v0,zero + .set reorder + +.L_bn_mul_words_proceed: + li MINUS4,-4 + and ta0,a2,MINUS4 + move v0,zero + beqz ta0,.L_bn_mul_words_tail + +.L_bn_mul_words_loop: + dmultu t0,a3 + ld t2,8(a1) + ld ta0,16(a1) + ld ta2,24(a1) + mflo AT + mfhi t0 + daddu v0,AT + sltu t1,v0,AT + sd v0,0(a0) + daddu v0,t1,t0 + + dmultu t2,a3 + subu a2,4 + PTR_ADD a0,32 + PTR_ADD a1,32 + mflo AT + mfhi t2 + daddu v0,AT + sltu t3,v0,AT + sd v0,-24(a0) + daddu v0,t3,t2 + + dmultu ta0,a3 + mflo AT + mfhi ta0 + daddu v0,AT + sltu ta1,v0,AT + sd v0,-16(a0) + daddu v0,ta1,ta0 + + + dmultu ta2,a3 + and ta0,a2,MINUS4 + mflo AT + mfhi ta2 + daddu v0,AT + sltu ta3,v0,AT + sd v0,-8(a0) + daddu v0,ta3,ta2 + .set noreorder + bgtzl ta0,.L_bn_mul_words_loop + ld t0,0(a1) + + bnezl a2,.L_bn_mul_words_tail + ld t0,0(a1) + .set reorder + +.L_bn_mul_words_return: + jr ra + +.L_bn_mul_words_tail: + dmultu t0,a3 + subu a2,1 + mflo AT + mfhi t0 + daddu v0,AT + sltu t1,v0,AT + sd v0,0(a0) + daddu v0,t1,t0 + beqz a2,.L_bn_mul_words_return + + ld t0,8(a1) + dmultu t0,a3 + subu a2,1 + mflo AT + mfhi t0 + daddu v0,AT + sltu t1,v0,AT + sd v0,8(a0) + daddu v0,t1,t0 + beqz a2,.L_bn_mul_words_return + + ld t0,16(a1) + dmultu t0,a3 + mflo AT + mfhi t0 + daddu v0,AT + sltu t1,v0,AT + sd v0,16(a0) + daddu v0,t1,t0 + jr ra +END(bn_mul_words) + +.align 5 +LEAF(bn_sqr_words) + .set noreorder + bgtzl a2,.L_bn_sqr_words_proceed + ld t0,0(a1) + jr ra + move v0,zero + .set reorder + +.L_bn_sqr_words_proceed: + li MINUS4,-4 + and ta0,a2,MINUS4 + move v0,zero + beqz ta0,.L_bn_sqr_words_tail + +.L_bn_sqr_words_loop: + dmultu t0,t0 + ld t2,8(a1) + ld ta0,16(a1) + ld ta2,24(a1) + mflo t1 + mfhi t0 + sd t1,0(a0) + sd t0,8(a0) + + dmultu t2,t2 + subu a2,4 + PTR_ADD a0,64 + PTR_ADD a1,32 + mflo t3 + mfhi t2 + sd t3,-48(a0) + sd t2,-40(a0) + + dmultu ta0,ta0 + mflo ta1 + mfhi ta0 + sd ta1,-32(a0) + sd ta0,-24(a0) + + + dmultu ta2,ta2 + and ta0,a2,MINUS4 + mflo ta3 + mfhi ta2 + sd ta3,-16(a0) + sd ta2,-8(a0) + + .set noreorder + bgtzl ta0,.L_bn_sqr_words_loop + ld t0,0(a1) + + bnezl a2,.L_bn_sqr_words_tail + ld t0,0(a1) + .set reorder + +.L_bn_sqr_words_return: + move v0,zero + jr ra + +.L_bn_sqr_words_tail: + dmultu t0,t0 + subu a2,1 + mflo t1 + mfhi t0 + sd t1,0(a0) + sd t0,8(a0) + beqz a2,.L_bn_sqr_words_return + + ld t0,8(a1) + dmultu t0,t0 + subu a2,1 + mflo t1 + mfhi t0 + sd t1,16(a0) + sd t0,24(a0) + beqz a2,.L_bn_sqr_words_return + + ld t0,16(a1) + dmultu t0,t0 + mflo t1 + mfhi t0 + sd t1,32(a0) + sd t0,40(a0) + jr ra +END(bn_sqr_words) + +.align 5 +LEAF(bn_add_words) + .set noreorder + bgtzl a3,.L_bn_add_words_proceed + ld t0,0(a1) + jr ra + move v0,zero + .set reorder + +.L_bn_add_words_proceed: + li MINUS4,-4 + and AT,a3,MINUS4 + move v0,zero + beqz AT,.L_bn_add_words_tail + +.L_bn_add_words_loop: + ld ta0,0(a2) + subu a3,4 + ld t1,8(a1) + and AT,a3,MINUS4 + ld t2,16(a1) + PTR_ADD a2,32 + ld t3,24(a1) + PTR_ADD a0,32 + ld ta1,-24(a2) + PTR_ADD a1,32 + ld ta2,-16(a2) + ld ta3,-8(a2) + daddu ta0,t0 + sltu t8,ta0,t0 + daddu t0,ta0,v0 + sltu v0,t0,ta0 + sd t0,-32(a0) + daddu v0,t8 + + daddu ta1,t1 + sltu t9,ta1,t1 + daddu t1,ta1,v0 + sltu v0,t1,ta1 + sd t1,-24(a0) + daddu v0,t9 + + daddu ta2,t2 + sltu t8,ta2,t2 + daddu t2,ta2,v0 + sltu v0,t2,ta2 + sd t2,-16(a0) + daddu v0,t8 + + daddu ta3,t3 + sltu t9,ta3,t3 + daddu t3,ta3,v0 + sltu v0,t3,ta3 + sd t3,-8(a0) + daddu v0,t9 + + .set noreorder + bgtzl AT,.L_bn_add_words_loop + ld t0,0(a1) + + bnezl a3,.L_bn_add_words_tail + ld t0,0(a1) + .set reorder + +.L_bn_add_words_return: + jr ra + +.L_bn_add_words_tail: + ld ta0,0(a2) + daddu ta0,t0 + subu a3,1 + sltu t8,ta0,t0 + daddu t0,ta0,v0 + sltu v0,t0,ta0 + sd t0,0(a0) + daddu v0,t8 + beqz a3,.L_bn_add_words_return + + ld t1,8(a1) + ld ta1,8(a2) + daddu ta1,t1 + subu a3,1 + sltu t9,ta1,t1 + daddu t1,ta1,v0 + sltu v0,t1,ta1 + sd t1,8(a0) + daddu v0,t9 + beqz a3,.L_bn_add_words_return + + ld t2,16(a1) + ld ta2,16(a2) + daddu ta2,t2 + sltu t8,ta2,t2 + daddu t2,ta2,v0 + sltu v0,t2,ta2 + sd t2,16(a0) + daddu v0,t8 + jr ra +END(bn_add_words) + +.align 5 +LEAF(bn_sub_words) + .set noreorder + bgtzl a3,.L_bn_sub_words_proceed + ld t0,0(a1) + jr ra + move v0,zero + .set reorder + +.L_bn_sub_words_proceed: + li MINUS4,-4 + and AT,a3,MINUS4 + move v0,zero + beqz AT,.L_bn_sub_words_tail + +.L_bn_sub_words_loop: + ld ta0,0(a2) + subu a3,4 + ld t1,8(a1) + and AT,a3,MINUS4 + ld t2,16(a1) + PTR_ADD a2,32 + ld t3,24(a1) + PTR_ADD a0,32 + ld ta1,-24(a2) + PTR_ADD a1,32 + ld ta2,-16(a2) + ld ta3,-8(a2) + sltu t8,t0,ta0 + dsubu t0,ta0 + dsubu ta0,t0,v0 + sd ta0,-32(a0) + MOVNZ (t0,v0,t8) + + sltu t9,t1,ta1 + dsubu t1,ta1 + dsubu ta1,t1,v0 + sd ta1,-24(a0) + MOVNZ (t1,v0,t9) + + + sltu t8,t2,ta2 + dsubu t2,ta2 + dsubu ta2,t2,v0 + sd ta2,-16(a0) + MOVNZ (t2,v0,t8) + + sltu t9,t3,ta3 + dsubu t3,ta3 + dsubu ta3,t3,v0 + sd ta3,-8(a0) + MOVNZ (t3,v0,t9) + + .set noreorder + bgtzl AT,.L_bn_sub_words_loop + ld t0,0(a1) + + bnezl a3,.L_bn_sub_words_tail + ld t0,0(a1) + .set reorder + +.L_bn_sub_words_return: + jr ra + +.L_bn_sub_words_tail: + ld ta0,0(a2) + subu a3,1 + sltu t8,t0,ta0 + dsubu t0,ta0 + dsubu ta0,t0,v0 + MOVNZ (t0,v0,t8) + sd ta0,0(a0) + beqz a3,.L_bn_sub_words_return + + ld t1,8(a1) + subu a3,1 + ld ta1,8(a2) + sltu t9,t1,ta1 + dsubu t1,ta1 + dsubu ta1,t1,v0 + MOVNZ (t1,v0,t9) + sd ta1,8(a0) + beqz a3,.L_bn_sub_words_return + + ld t2,16(a1) + ld ta2,16(a2) + sltu t8,t2,ta2 + dsubu t2,ta2 + dsubu ta2,t2,v0 + MOVNZ (t2,v0,t8) + sd ta2,16(a0) + jr ra +END(bn_sub_words) + +#undef MINUS4 + +.align 5 +LEAF(bn_div_3_words) + .set reorder + move a3,a0 /* we know that bn_div_words doesn't + * touch a3, ta2, ta3 and preserves a2 + * so that we can save two arguments + * and return address in registers + * instead of stack:-) + */ + ld a0,(a3) + move ta2,a1 + ld a1,-8(a3) + bne a0,a2,.L_bn_div_3_words_proceed + li v0,-1 + jr ra +.L_bn_div_3_words_proceed: + move ta3,ra + bal bn_div_words + move ra,ta3 + dmultu ta2,v0 + ld t2,-16(a3) + move ta0,zero + mfhi t1 + mflo t0 + sltu t8,t1,v1 +.L_bn_div_3_words_inner_loop: + bnez t8,.L_bn_div_3_words_inner_loop_done + sgeu AT,t2,t0 + seq t9,t1,v1 + and AT,t9 + sltu t3,t0,ta2 + daddu v1,a2 + dsubu t1,t3 + dsubu t0,ta2 + sltu t8,t1,v1 + sltu ta0,v1,a2 + or t8,ta0 + .set noreorder + beqzl AT,.L_bn_div_3_words_inner_loop + dsubu v0,1 + .set reorder +.L_bn_div_3_words_inner_loop_done: + jr ra +END(bn_div_3_words) + +.align 5 +LEAF(bn_div_words) + .set noreorder + bnezl a2,.L_bn_div_words_proceed + move v1,zero + jr ra + li v0,-1 /* I'd rather signal div-by-zero + * which can be done with 'break 7' */ + +.L_bn_div_words_proceed: + bltz a2,.L_bn_div_words_body + move t9,v1 + dsll a2,1 + bgtz a2,.-4 + addu t9,1 + + .set reorder + negu t1,t9 + li t2,-1 + dsll t2,t1 + and t2,a0 + dsrl AT,a1,t1 + .set noreorder + bnezl t2,.+8 + break 6 /* signal overflow */ + .set reorder + dsll a0,t9 + dsll a1,t9 + or a0,AT + +#define QT ta0 +#define HH ta1 +#define DH v1 +.L_bn_div_words_body: + dsrl DH,a2,32 + sgeu AT,a0,a2 + .set noreorder + bnezl AT,.+8 + dsubu a0,a2 + .set reorder + + li QT,-1 + dsrl HH,a0,32 + dsrl QT,32 /* q=0xffffffff */ + beq DH,HH,.L_bn_div_words_skip_div1 + ddivu zero,a0,DH + mflo QT +.L_bn_div_words_skip_div1: + dmultu a2,QT + dsll t3,a0,32 + dsrl AT,a1,32 + or t3,AT + mflo t0 + mfhi t1 +.L_bn_div_words_inner_loop1: + sltu t2,t3,t0 + seq t8,HH,t1 + sltu AT,HH,t1 + and t2,t8 + sltu v0,t0,a2 + or AT,t2 + .set noreorder + beqz AT,.L_bn_div_words_inner_loop1_done + dsubu t1,v0 + dsubu t0,a2 + b .L_bn_div_words_inner_loop1 + dsubu QT,1 + .set reorder +.L_bn_div_words_inner_loop1_done: + + dsll a1,32 + dsubu a0,t3,t0 + dsll v0,QT,32 + + li QT,-1 + dsrl HH,a0,32 + dsrl QT,32 /* q=0xffffffff */ + beq DH,HH,.L_bn_div_words_skip_div2 + ddivu zero,a0,DH + mflo QT +.L_bn_div_words_skip_div2: +#undef DH + dmultu a2,QT + dsll t3,a0,32 + dsrl AT,a1,32 + or t3,AT + mflo t0 + mfhi t1 +.L_bn_div_words_inner_loop2: + sltu t2,t3,t0 + seq t8,HH,t1 + sltu AT,HH,t1 + and t2,t8 + sltu v1,t0,a2 + or AT,t2 + .set noreorder + beqz AT,.L_bn_div_words_inner_loop2_done + dsubu t1,v1 + dsubu t0,a2 + b .L_bn_div_words_inner_loop2 + dsubu QT,1 + .set reorder +.L_bn_div_words_inner_loop2_done: +#undef HH + + dsubu a0,t3,t0 + or v0,QT + dsrl v1,a0,t9 /* v1 contains remainder if anybody wants it */ + dsrl a2,t9 /* restore a2 */ + jr ra +#undef QT +END(bn_div_words) + +#define a_0 t0 +#define a_1 t1 +#define a_2 t2 +#define a_3 t3 +#define b_0 ta0 +#define b_1 ta1 +#define b_2 ta2 +#define b_3 ta3 + +#define a_4 s0 +#define a_5 s2 +#define a_6 s4 +#define a_7 a1 /* once we load a[7] we don't need a anymore */ +#define b_4 s1 +#define b_5 s3 +#define b_6 s5 +#define b_7 a2 /* once we load b[7] we don't need b anymore */ + +#define t_1 t8 +#define t_2 t9 + +#define c_1 v0 +#define c_2 v1 +#define c_3 a3 + +#define FRAME_SIZE 48 + +.align 5 +LEAF(bn_mul_comba8) + .set noreorder + PTR_SUB sp,FRAME_SIZE + .frame sp,64,ra + .set reorder + ld a_0,0(a1) /* If compiled with -mips3 option on + * R5000 box assembler barks on this + * line with "shouldn't have mult/div + * as last instruction in bb (R10K + * bug)" warning. If anybody out there + * has a clue about how to circumvent + * this do send me a note. + * + */ + ld b_0,0(a2) + ld a_1,8(a1) + ld a_2,16(a1) + ld a_3,24(a1) + ld b_1,8(a2) + ld b_2,16(a2) + ld b_3,24(a2) + dmultu a_0,b_0 /* mul_add_c(a[0],b[0],c1,c2,c3); */ + sd s0,0(sp) + sd s1,8(sp) + sd s2,16(sp) + sd s3,24(sp) + sd s4,32(sp) + sd s5,40(sp) + mflo c_1 + mfhi c_2 + + dmultu a_0,b_1 /* mul_add_c(a[0],b[1],c2,c3,c1); */ + ld a_4,32(a1) + ld a_5,40(a1) + ld a_6,48(a1) + ld a_7,56(a1) + ld b_4,32(a2) + ld b_5,40(a2) + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu c_3,t_2,AT + dmultu a_1,b_0 /* mul_add_c(a[1],b[0],c2,c3,c1); */ + ld b_6,48(a2) + ld b_7,56(a2) + sd c_1,0(a0) /* r[0]=c1; */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu c_1,c_3,t_2 + sd c_2,8(a0) /* r[1]=c2; */ + + dmultu a_2,b_0 /* mul_add_c(a[2],b[0],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + dmultu a_1,b_1 /* mul_add_c(a[1],b[1],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu c_2,c_1,t_2 + dmultu a_0,b_2 /* mul_add_c(a[0],b[2],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,16(a0) /* r[2]=c3; */ + + dmultu a_0,b_3 /* mul_add_c(a[0],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu c_3,c_2,t_2 + dmultu a_1,b_2 /* mul_add_c(a[1],b[2],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_2,b_1 /* mul_add_c(a[2],b[1],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_3,b_0 /* mul_add_c(a[3],b[0],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,24(a0) /* r[3]=c1; */ + + dmultu a_4,b_0 /* mul_add_c(a[4],b[0],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu c_1,c_3,t_2 + dmultu a_3,b_1 /* mul_add_c(a[3],b[1],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_2,b_2 /* mul_add_c(a[2],b[2],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_1,b_3 /* mul_add_c(a[1],b[3],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_0,b_4 /* mul_add_c(a[0],b[4],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,32(a0) /* r[4]=c2; */ + + dmultu a_0,b_5 /* mul_add_c(a[0],b[5],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu c_2,c_1,t_2 + dmultu a_1,b_4 /* mul_add_c(a[1],b[4],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_2,b_3 /* mul_add_c(a[2],b[3],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_3,b_2 /* mul_add_c(a[3],b[2],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_4,b_1 /* mul_add_c(a[4],b[1],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_5,b_0 /* mul_add_c(a[5],b[0],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,40(a0) /* r[5]=c3; */ + + dmultu a_6,b_0 /* mul_add_c(a[6],b[0],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu c_3,c_2,t_2 + dmultu a_5,b_1 /* mul_add_c(a[5],b[1],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_4,b_2 /* mul_add_c(a[4],b[2],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_3,b_3 /* mul_add_c(a[3],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_2,b_4 /* mul_add_c(a[2],b[4],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_1,b_5 /* mul_add_c(a[1],b[5],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_0,b_6 /* mul_add_c(a[0],b[6],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,48(a0) /* r[6]=c1; */ + + dmultu a_0,b_7 /* mul_add_c(a[0],b[7],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu c_1,c_3,t_2 + dmultu a_1,b_6 /* mul_add_c(a[1],b[6],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_2,b_5 /* mul_add_c(a[2],b[5],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_3,b_4 /* mul_add_c(a[3],b[4],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_4,b_3 /* mul_add_c(a[4],b[3],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_5,b_2 /* mul_add_c(a[5],b[2],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_6,b_1 /* mul_add_c(a[6],b[1],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_7,b_0 /* mul_add_c(a[7],b[0],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,56(a0) /* r[7]=c2; */ + + dmultu a_7,b_1 /* mul_add_c(a[7],b[1],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu c_2,c_1,t_2 + dmultu a_6,b_2 /* mul_add_c(a[6],b[2],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_5,b_3 /* mul_add_c(a[5],b[3],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_4,b_4 /* mul_add_c(a[4],b[4],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_3,b_5 /* mul_add_c(a[3],b[5],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_2,b_6 /* mul_add_c(a[2],b[6],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_1,b_7 /* mul_add_c(a[1],b[7],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,64(a0) /* r[8]=c3; */ + + dmultu a_2,b_7 /* mul_add_c(a[2],b[7],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu c_3,c_2,t_2 + dmultu a_3,b_6 /* mul_add_c(a[3],b[6],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_4,b_5 /* mul_add_c(a[4],b[5],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_5,b_4 /* mul_add_c(a[5],b[4],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_6,b_3 /* mul_add_c(a[6],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_7,b_2 /* mul_add_c(a[7],b[2],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,72(a0) /* r[9]=c1; */ + + dmultu a_7,b_3 /* mul_add_c(a[7],b[3],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu c_1,c_3,t_2 + dmultu a_6,b_4 /* mul_add_c(a[6],b[4],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_5,b_5 /* mul_add_c(a[5],b[5],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_4,b_6 /* mul_add_c(a[4],b[6],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_3,b_7 /* mul_add_c(a[3],b[7],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,80(a0) /* r[10]=c2; */ + + dmultu a_4,b_7 /* mul_add_c(a[4],b[7],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu c_2,c_1,t_2 + dmultu a_5,b_6 /* mul_add_c(a[5],b[6],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_6,b_5 /* mul_add_c(a[6],b[5],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_7,b_4 /* mul_add_c(a[7],b[4],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,88(a0) /* r[11]=c3; */ + + dmultu a_7,b_5 /* mul_add_c(a[7],b[5],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu c_3,c_2,t_2 + dmultu a_6,b_6 /* mul_add_c(a[6],b[6],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_5,b_7 /* mul_add_c(a[5],b[7],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,96(a0) /* r[12]=c1; */ + + dmultu a_6,b_7 /* mul_add_c(a[6],b[7],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu c_1,c_3,t_2 + dmultu a_7,b_6 /* mul_add_c(a[7],b[6],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,104(a0) /* r[13]=c2; */ + + dmultu a_7,b_7 /* mul_add_c(a[7],b[7],c3,c1,c2); */ + ld s0,0(sp) + ld s1,8(sp) + ld s2,16(sp) + ld s3,24(sp) + ld s4,32(sp) + ld s5,40(sp) + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sd c_3,112(a0) /* r[14]=c3; */ + sd c_1,120(a0) /* r[15]=c1; */ + + PTR_ADD sp,FRAME_SIZE + + jr ra +END(bn_mul_comba8) + +.align 5 +LEAF(bn_mul_comba4) + .set reorder + ld a_0,0(a1) + ld b_0,0(a2) + ld a_1,8(a1) + ld a_2,16(a1) + dmultu a_0,b_0 /* mul_add_c(a[0],b[0],c1,c2,c3); */ + ld a_3,24(a1) + ld b_1,8(a2) + ld b_2,16(a2) + ld b_3,24(a2) + mflo c_1 + mfhi c_2 + sd c_1,0(a0) + + dmultu a_0,b_1 /* mul_add_c(a[0],b[1],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu c_3,t_2,AT + dmultu a_1,b_0 /* mul_add_c(a[1],b[0],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu c_1,c_3,t_2 + sd c_2,8(a0) + + dmultu a_2,b_0 /* mul_add_c(a[2],b[0],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + dmultu a_1,b_1 /* mul_add_c(a[1],b[1],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu c_2,c_1,t_2 + dmultu a_0,b_2 /* mul_add_c(a[0],b[2],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,16(a0) + + dmultu a_0,b_3 /* mul_add_c(a[0],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu c_3,c_2,t_2 + dmultu a_1,b_2 /* mul_add_c(a[1],b[2],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_2,b_1 /* mul_add_c(a[2],b[1],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_3,b_0 /* mul_add_c(a[3],b[0],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,24(a0) + + dmultu a_3,b_1 /* mul_add_c(a[3],b[1],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu c_1,c_3,t_2 + dmultu a_2,b_2 /* mul_add_c(a[2],b[2],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_1,b_3 /* mul_add_c(a[1],b[3],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,32(a0) + + dmultu a_2,b_3 /* mul_add_c(a[2],b[3],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu c_2,c_1,t_2 + dmultu a_3,b_2 /* mul_add_c(a[3],b[2],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,40(a0) + + dmultu a_3,b_3 /* mul_add_c(a[3],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sd c_1,48(a0) + sd c_2,56(a0) + + jr ra +END(bn_mul_comba4) + +#undef a_4 +#undef a_5 +#undef a_6 +#undef a_7 +#define a_4 b_0 +#define a_5 b_1 +#define a_6 b_2 +#define a_7 b_3 + +.align 5 +LEAF(bn_sqr_comba8) + .set reorder + ld a_0,0(a1) + ld a_1,8(a1) + ld a_2,16(a1) + ld a_3,24(a1) + + dmultu a_0,a_0 /* mul_add_c(a[0],b[0],c1,c2,c3); */ + ld a_4,32(a1) + ld a_5,40(a1) + ld a_6,48(a1) + ld a_7,56(a1) + mflo c_1 + mfhi c_2 + sd c_1,0(a0) + + dmultu a_0,a_1 /* mul_add_c2(a[0],b[1],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt c_1,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu c_3,t_2,AT + sd c_2,8(a0) + + dmultu a_2,a_0 /* mul_add_c2(a[2],b[0],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt c_2,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_1,a_1 /* mul_add_c(a[1],b[1],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,16(a0) + + dmultu a_0,a_3 /* mul_add_c2(a[0],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt c_3,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_1,a_2 /* mul_add_c2(a[1],b[2],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_3,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,24(a0) + + dmultu a_4,a_0 /* mul_add_c2(a[4],b[0],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt c_1,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_3,a_1 /* mul_add_c2(a[3],b[1],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_1,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_2,a_2 /* mul_add_c(a[2],b[2],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,32(a0) + + dmultu a_0,a_5 /* mul_add_c2(a[0],b[5],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt c_2,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_1,a_4 /* mul_add_c2(a[1],b[4],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_2,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_2,a_3 /* mul_add_c2(a[2],b[3],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_2,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,40(a0) + + dmultu a_6,a_0 /* mul_add_c2(a[6],b[0],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt c_3,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_5,a_1 /* mul_add_c2(a[5],b[1],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_3,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_4,a_2 /* mul_add_c2(a[4],b[2],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_3,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_3,a_3 /* mul_add_c(a[3],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,48(a0) + + dmultu a_0,a_7 /* mul_add_c2(a[0],b[7],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt c_1,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_1,a_6 /* mul_add_c2(a[1],b[6],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_1,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_2,a_5 /* mul_add_c2(a[2],b[5],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_1,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_3,a_4 /* mul_add_c2(a[3],b[4],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_1,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,56(a0) + + dmultu a_7,a_1 /* mul_add_c2(a[7],b[1],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt c_2,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_6,a_2 /* mul_add_c2(a[6],b[2],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_2,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_5,a_3 /* mul_add_c2(a[5],b[3],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_2,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_4,a_4 /* mul_add_c(a[4],b[4],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,64(a0) + + dmultu a_2,a_7 /* mul_add_c2(a[2],b[7],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt c_3,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_3,a_6 /* mul_add_c2(a[3],b[6],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_3,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_4,a_5 /* mul_add_c2(a[4],b[5],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_3,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,72(a0) + + dmultu a_7,a_3 /* mul_add_c2(a[7],b[3],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt c_1,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_6,a_4 /* mul_add_c2(a[6],b[4],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_1,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_5,a_5 /* mul_add_c(a[5],b[5],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,80(a0) + + dmultu a_4,a_7 /* mul_add_c2(a[4],b[7],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt c_2,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_5,a_6 /* mul_add_c2(a[5],b[6],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_2,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,88(a0) + + dmultu a_7,a_5 /* mul_add_c2(a[7],b[5],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt c_3,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_6,a_6 /* mul_add_c(a[6],b[6],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,96(a0) + + dmultu a_6,a_7 /* mul_add_c2(a[6],b[7],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt c_1,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,104(a0) + + dmultu a_7,a_7 /* mul_add_c(a[7],b[7],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sd c_3,112(a0) + sd c_1,120(a0) + + jr ra +END(bn_sqr_comba8) + +.align 5 +LEAF(bn_sqr_comba4) + .set reorder + ld a_0,0(a1) + ld a_1,8(a1) + ld a_2,16(a1) + ld a_3,24(a1) + dmultu a_0,a_0 /* mul_add_c(a[0],b[0],c1,c2,c3); */ + mflo c_1 + mfhi c_2 + sd c_1,0(a0) + + dmultu a_0,a_1 /* mul_add_c2(a[0],b[1],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt c_1,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu c_3,t_2,AT + sd c_2,8(a0) + + dmultu a_2,a_0 /* mul_add_c2(a[2],b[0],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt c_2,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + dmultu a_1,a_1 /* mul_add_c(a[1],b[1],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,16(a0) + + dmultu a_0,a_3 /* mul_add_c2(a[0],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt c_3,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + dmultu a_1,a_2 /* mul_add_c(a2[1],b[2],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + slt AT,t_2,zero + daddu c_3,AT + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sltu AT,c_2,t_2 + daddu c_3,AT + sd c_1,24(a0) + + dmultu a_3,a_1 /* mul_add_c2(a[3],b[1],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + slt c_1,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + dmultu a_2,a_2 /* mul_add_c(a[2],b[2],c2,c3,c1); */ + mflo t_1 + mfhi t_2 + daddu c_2,t_1 + sltu AT,c_2,t_1 + daddu t_2,AT + daddu c_3,t_2 + sltu AT,c_3,t_2 + daddu c_1,AT + sd c_2,32(a0) + + dmultu a_2,a_3 /* mul_add_c2(a[2],b[3],c3,c1,c2); */ + mflo t_1 + mfhi t_2 + slt c_2,t_2,zero + dsll t_2,1 + slt a2,t_1,zero + daddu t_2,a2 + dsll t_1,1 + daddu c_3,t_1 + sltu AT,c_3,t_1 + daddu t_2,AT + daddu c_1,t_2 + sltu AT,c_1,t_2 + daddu c_2,AT + sd c_3,40(a0) + + dmultu a_3,a_3 /* mul_add_c(a[3],b[3],c1,c2,c3); */ + mflo t_1 + mfhi t_2 + daddu c_1,t_1 + sltu AT,c_1,t_1 + daddu t_2,AT + daddu c_2,t_2 + sd c_1,48(a0) + sd c_2,56(a0) + + jr ra +END(bn_sqr_comba4) diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/modexp512-x86_64.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/modexp512-x86_64.S new file mode 100644 index 00000000..6cccafb8 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/modexp512-x86_64.S @@ -0,0 +1,1773 @@ +.text + +.type MULADD_128x512,@function +.align 16 +MULADD_128x512: + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + movq %r8,0(%rcx) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%r8 + movq 8(%rdi),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + movq %r9,8(%rcx) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%r9 + .byte 0xf3,0xc3 +.size MULADD_128x512,.-MULADD_128x512 +.type mont_reduce,@function +.align 16 +mont_reduce: + leaq 192(%rsp),%rdi + movq 32(%rsp),%rsi + addq $576,%rsi + leaq 520(%rsp),%rcx + + movq 96(%rcx),%rbp + movq 0(%rsi),%rax + mulq %rbp + movq (%rcx),%r8 + addq %rax,%r8 + adcq $0,%rdx + movq %r8,0(%rdi) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + movq 8(%rcx),%r9 + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + movq 16(%rcx),%r10 + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + movq 24(%rcx),%r11 + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + movq 32(%rcx),%r12 + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + movq 40(%rcx),%r13 + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + movq 48(%rcx),%r14 + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + movq 56(%rcx),%r15 + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%r8 + movq 104(%rcx),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + movq %r9,8(%rdi) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%r9 + movq 112(%rcx),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + movq %r10,16(%rdi) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%r10 + movq 120(%rcx),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + movq %r11,24(%rdi) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%r11 + xorq %rax,%rax + + addq 64(%rcx),%r8 + adcq 72(%rcx),%r9 + adcq 80(%rcx),%r10 + adcq 88(%rcx),%r11 + adcq $0,%rax + + + + + movq %r8,64(%rdi) + movq %r9,72(%rdi) + movq %r10,%rbp + movq %r11,88(%rdi) + + movq %rax,384(%rsp) + + movq 0(%rdi),%r8 + movq 8(%rdi),%r9 + movq 16(%rdi),%r10 + movq 24(%rdi),%r11 + + + + + + + + + addq $80,%rdi + + addq $64,%rsi + leaq 296(%rsp),%rcx + + call MULADD_128x512 + + movq 384(%rsp),%rax + + + addq -16(%rdi),%r8 + adcq -8(%rdi),%r9 + movq %r8,64(%rcx) + movq %r9,72(%rcx) + + adcq %rax,%rax + movq %rax,384(%rsp) + + leaq 192(%rsp),%rdi + addq $64,%rsi + + + + + + movq (%rsi),%r8 + movq 8(%rsi),%rbx + + movq (%rcx),%rax + mulq %r8 + movq %rax,%rbp + movq %rdx,%r9 + + movq 8(%rcx),%rax + mulq %r8 + addq %rax,%r9 + + movq (%rcx),%rax + mulq %rbx + addq %rax,%r9 + + movq %r9,8(%rdi) + + + subq $192,%rsi + + movq (%rcx),%r8 + movq 8(%rcx),%r9 + + call MULADD_128x512 + + + + + movq 0(%rsi),%rax + movq 8(%rsi),%rbx + movq 16(%rsi),%rdi + movq 24(%rsi),%rdx + + + movq 384(%rsp),%rbp + + addq 64(%rcx),%r8 + adcq 72(%rcx),%r9 + + + adcq %rbp,%rbp + + + + shlq $3,%rbp + movq 32(%rsp),%rcx + addq %rcx,%rbp + + + xorq %rsi,%rsi + + addq 0(%rbp),%r10 + adcq 64(%rbp),%r11 + adcq 128(%rbp),%r12 + adcq 192(%rbp),%r13 + adcq 256(%rbp),%r14 + adcq 320(%rbp),%r15 + adcq 384(%rbp),%r8 + adcq 448(%rbp),%r9 + + + + sbbq $0,%rsi + + + andq %rsi,%rax + andq %rsi,%rbx + andq %rsi,%rdi + andq %rsi,%rdx + + movq $1,%rbp + subq %rax,%r10 + sbbq %rbx,%r11 + sbbq %rdi,%r12 + sbbq %rdx,%r13 + + + + + sbbq $0,%rbp + + + + addq $512,%rcx + movq 32(%rcx),%rax + movq 40(%rcx),%rbx + movq 48(%rcx),%rdi + movq 56(%rcx),%rdx + + + + andq %rsi,%rax + andq %rsi,%rbx + andq %rsi,%rdi + andq %rsi,%rdx + + + + subq $1,%rbp + + sbbq %rax,%r14 + sbbq %rbx,%r15 + sbbq %rdi,%r8 + sbbq %rdx,%r9 + + + + movq 144(%rsp),%rsi + movq %r10,0(%rsi) + movq %r11,8(%rsi) + movq %r12,16(%rsi) + movq %r13,24(%rsi) + movq %r14,32(%rsi) + movq %r15,40(%rsi) + movq %r8,48(%rsi) + movq %r9,56(%rsi) + + .byte 0xf3,0xc3 +.size mont_reduce,.-mont_reduce +.type mont_mul_a3b,@function +.align 16 +mont_mul_a3b: + + + + + movq 0(%rdi),%rbp + + movq %r10,%rax + mulq %rbp + movq %rax,520(%rsp) + movq %rdx,%r10 + movq %r11,%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + movq %rdx,%r11 + movq %r12,%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + movq %rdx,%r12 + movq %r13,%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + movq %rdx,%r13 + movq %r14,%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + movq %rdx,%r14 + movq %r15,%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + movq %rdx,%r15 + movq %r8,%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + movq %rdx,%r8 + movq %r9,%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + movq %rdx,%r9 + movq 8(%rdi),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + movq %r10,528(%rsp) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%r10 + movq 16(%rdi),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + movq %r11,536(%rsp) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%r11 + movq 24(%rdi),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + movq %r12,544(%rsp) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%r12 + movq 32(%rdi),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + movq %r13,552(%rsp) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%r13 + movq 40(%rdi),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + movq %r14,560(%rsp) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%r14 + movq 48(%rdi),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + movq %r15,568(%rsp) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + addq %rbx,%r8 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%r15 + movq 56(%rdi),%rbp + movq 0(%rsi),%rax + mulq %rbp + addq %rax,%r8 + adcq $0,%rdx + movq %r8,576(%rsp) + movq %rdx,%rbx + + movq 8(%rsi),%rax + mulq %rbp + addq %rax,%r9 + adcq $0,%rdx + addq %rbx,%r9 + adcq $0,%rdx + movq %rdx,%rbx + + movq 16(%rsi),%rax + mulq %rbp + addq %rax,%r10 + adcq $0,%rdx + addq %rbx,%r10 + adcq $0,%rdx + movq %rdx,%rbx + + movq 24(%rsi),%rax + mulq %rbp + addq %rax,%r11 + adcq $0,%rdx + addq %rbx,%r11 + adcq $0,%rdx + movq %rdx,%rbx + + movq 32(%rsi),%rax + mulq %rbp + addq %rax,%r12 + adcq $0,%rdx + addq %rbx,%r12 + adcq $0,%rdx + movq %rdx,%rbx + + movq 40(%rsi),%rax + mulq %rbp + addq %rax,%r13 + adcq $0,%rdx + addq %rbx,%r13 + adcq $0,%rdx + movq %rdx,%rbx + + movq 48(%rsi),%rax + mulq %rbp + addq %rax,%r14 + adcq $0,%rdx + addq %rbx,%r14 + adcq $0,%rdx + movq %rdx,%rbx + + movq 56(%rsi),%rax + mulq %rbp + addq %rax,%r15 + adcq $0,%rdx + addq %rbx,%r15 + adcq $0,%rdx + movq %rdx,%r8 + movq %r9,584(%rsp) + movq %r10,592(%rsp) + movq %r11,600(%rsp) + movq %r12,608(%rsp) + movq %r13,616(%rsp) + movq %r14,624(%rsp) + movq %r15,632(%rsp) + movq %r8,640(%rsp) + + + + + + jmp mont_reduce + + +.size mont_mul_a3b,.-mont_mul_a3b +.type sqr_reduce,@function +.align 16 +sqr_reduce: + movq 16(%rsp),%rcx + + + + movq %r10,%rbx + + movq %r11,%rax + mulq %rbx + movq %rax,528(%rsp) + movq %rdx,%r10 + movq %r12,%rax + mulq %rbx + addq %rax,%r10 + adcq $0,%rdx + movq %rdx,%r11 + movq %r13,%rax + mulq %rbx + addq %rax,%r11 + adcq $0,%rdx + movq %rdx,%r12 + movq %r14,%rax + mulq %rbx + addq %rax,%r12 + adcq $0,%rdx + movq %rdx,%r13 + movq %r15,%rax + mulq %rbx + addq %rax,%r13 + adcq $0,%rdx + movq %rdx,%r14 + movq %r8,%rax + mulq %rbx + addq %rax,%r14 + adcq $0,%rdx + movq %rdx,%r15 + movq %r9,%rax + mulq %rbx + addq %rax,%r15 + adcq $0,%rdx + movq %rdx,%rsi + + movq %r10,536(%rsp) + + + + + + movq 8(%rcx),%rbx + + movq 16(%rcx),%rax + mulq %rbx + addq %rax,%r11 + adcq $0,%rdx + movq %r11,544(%rsp) + + movq %rdx,%r10 + movq 24(%rcx),%rax + mulq %rbx + addq %rax,%r12 + adcq $0,%rdx + addq %r10,%r12 + adcq $0,%rdx + movq %r12,552(%rsp) + + movq %rdx,%r10 + movq 32(%rcx),%rax + mulq %rbx + addq %rax,%r13 + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + + movq %rdx,%r10 + movq 40(%rcx),%rax + mulq %rbx + addq %rax,%r14 + adcq $0,%rdx + addq %r10,%r14 + adcq $0,%rdx + + movq %rdx,%r10 + movq %r8,%rax + mulq %rbx + addq %rax,%r15 + adcq $0,%rdx + addq %r10,%r15 + adcq $0,%rdx + + movq %rdx,%r10 + movq %r9,%rax + mulq %rbx + addq %rax,%rsi + adcq $0,%rdx + addq %r10,%rsi + adcq $0,%rdx + + movq %rdx,%r11 + + + + + movq 16(%rcx),%rbx + + movq 24(%rcx),%rax + mulq %rbx + addq %rax,%r13 + adcq $0,%rdx + movq %r13,560(%rsp) + + movq %rdx,%r10 + movq 32(%rcx),%rax + mulq %rbx + addq %rax,%r14 + adcq $0,%rdx + addq %r10,%r14 + adcq $0,%rdx + movq %r14,568(%rsp) + + movq %rdx,%r10 + movq 40(%rcx),%rax + mulq %rbx + addq %rax,%r15 + adcq $0,%rdx + addq %r10,%r15 + adcq $0,%rdx + + movq %rdx,%r10 + movq %r8,%rax + mulq %rbx + addq %rax,%rsi + adcq $0,%rdx + addq %r10,%rsi + adcq $0,%rdx + + movq %rdx,%r10 + movq %r9,%rax + mulq %rbx + addq %rax,%r11 + adcq $0,%rdx + addq %r10,%r11 + adcq $0,%rdx + + movq %rdx,%r12 + + + + + + movq 24(%rcx),%rbx + + movq 32(%rcx),%rax + mulq %rbx + addq %rax,%r15 + adcq $0,%rdx + movq %r15,576(%rsp) + + movq %rdx,%r10 + movq 40(%rcx),%rax + mulq %rbx + addq %rax,%rsi + adcq $0,%rdx + addq %r10,%rsi + adcq $0,%rdx + movq %rsi,584(%rsp) + + movq %rdx,%r10 + movq %r8,%rax + mulq %rbx + addq %rax,%r11 + adcq $0,%rdx + addq %r10,%r11 + adcq $0,%rdx + + movq %rdx,%r10 + movq %r9,%rax + mulq %rbx + addq %rax,%r12 + adcq $0,%rdx + addq %r10,%r12 + adcq $0,%rdx + + movq %rdx,%r15 + + + + + movq 32(%rcx),%rbx + + movq 40(%rcx),%rax + mulq %rbx + addq %rax,%r11 + adcq $0,%rdx + movq %r11,592(%rsp) + + movq %rdx,%r10 + movq %r8,%rax + mulq %rbx + addq %rax,%r12 + adcq $0,%rdx + addq %r10,%r12 + adcq $0,%rdx + movq %r12,600(%rsp) + + movq %rdx,%r10 + movq %r9,%rax + mulq %rbx + addq %rax,%r15 + adcq $0,%rdx + addq %r10,%r15 + adcq $0,%rdx + + movq %rdx,%r11 + + + + + movq 40(%rcx),%rbx + + movq %r8,%rax + mulq %rbx + addq %rax,%r15 + adcq $0,%rdx + movq %r15,608(%rsp) + + movq %rdx,%r10 + movq %r9,%rax + mulq %rbx + addq %rax,%r11 + adcq $0,%rdx + addq %r10,%r11 + adcq $0,%rdx + movq %r11,616(%rsp) + + movq %rdx,%r12 + + + + + movq %r8,%rbx + + movq %r9,%rax + mulq %rbx + addq %rax,%r12 + adcq $0,%rdx + movq %r12,624(%rsp) + + movq %rdx,632(%rsp) + + + movq 528(%rsp),%r10 + movq 536(%rsp),%r11 + movq 544(%rsp),%r12 + movq 552(%rsp),%r13 + movq 560(%rsp),%r14 + movq 568(%rsp),%r15 + + movq 24(%rcx),%rax + mulq %rax + movq %rax,%rdi + movq %rdx,%r8 + + addq %r10,%r10 + adcq %r11,%r11 + adcq %r12,%r12 + adcq %r13,%r13 + adcq %r14,%r14 + adcq %r15,%r15 + adcq $0,%r8 + + movq 0(%rcx),%rax + mulq %rax + movq %rax,520(%rsp) + movq %rdx,%rbx + + movq 8(%rcx),%rax + mulq %rax + + addq %rbx,%r10 + adcq %rax,%r11 + adcq $0,%rdx + + movq %rdx,%rbx + movq %r10,528(%rsp) + movq %r11,536(%rsp) + + movq 16(%rcx),%rax + mulq %rax + + addq %rbx,%r12 + adcq %rax,%r13 + adcq $0,%rdx + + movq %rdx,%rbx + + movq %r12,544(%rsp) + movq %r13,552(%rsp) + + xorq %rbp,%rbp + addq %rbx,%r14 + adcq %rdi,%r15 + adcq $0,%rbp + + movq %r14,560(%rsp) + movq %r15,568(%rsp) + + + + + movq 576(%rsp),%r10 + movq 584(%rsp),%r11 + movq 592(%rsp),%r12 + movq 600(%rsp),%r13 + movq 608(%rsp),%r14 + movq 616(%rsp),%r15 + movq 624(%rsp),%rdi + movq 632(%rsp),%rsi + + movq %r9,%rax + mulq %rax + movq %rax,%r9 + movq %rdx,%rbx + + addq %r10,%r10 + adcq %r11,%r11 + adcq %r12,%r12 + adcq %r13,%r13 + adcq %r14,%r14 + adcq %r15,%r15 + adcq %rdi,%rdi + adcq %rsi,%rsi + adcq $0,%rbx + + addq %rbp,%r10 + + movq 32(%rcx),%rax + mulq %rax + + addq %r8,%r10 + adcq %rax,%r11 + adcq $0,%rdx + + movq %rdx,%rbp + + movq %r10,576(%rsp) + movq %r11,584(%rsp) + + movq 40(%rcx),%rax + mulq %rax + + addq %rbp,%r12 + adcq %rax,%r13 + adcq $0,%rdx + + movq %rdx,%rbp + + movq %r12,592(%rsp) + movq %r13,600(%rsp) + + movq 48(%rcx),%rax + mulq %rax + + addq %rbp,%r14 + adcq %rax,%r15 + adcq $0,%rdx + + movq %r14,608(%rsp) + movq %r15,616(%rsp) + + addq %rdx,%rdi + adcq %r9,%rsi + adcq $0,%rbx + + movq %rdi,624(%rsp) + movq %rsi,632(%rsp) + movq %rbx,640(%rsp) + + jmp mont_reduce + + +.size sqr_reduce,.-sqr_reduce +.globl mod_exp_512 +.type mod_exp_512,@function +mod_exp_512: + pushq %rbp + pushq %rbx + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + + + movq %rsp,%r8 + subq $2688,%rsp + andq $-64,%rsp + + + movq %r8,0(%rsp) + movq %rdi,8(%rsp) + movq %rsi,16(%rsp) + movq %rcx,24(%rsp) +.Lbody: + + + + pxor %xmm4,%xmm4 + movdqu 0(%rsi),%xmm0 + movdqu 16(%rsi),%xmm1 + movdqu 32(%rsi),%xmm2 + movdqu 48(%rsi),%xmm3 + movdqa %xmm4,512(%rsp) + movdqa %xmm4,528(%rsp) + movdqa %xmm4,608(%rsp) + movdqa %xmm4,624(%rsp) + movdqa %xmm0,544(%rsp) + movdqa %xmm1,560(%rsp) + movdqa %xmm2,576(%rsp) + movdqa %xmm3,592(%rsp) + + + movdqu 0(%rdx),%xmm0 + movdqu 16(%rdx),%xmm1 + movdqu 32(%rdx),%xmm2 + movdqu 48(%rdx),%xmm3 + + leaq 384(%rsp),%rbx + movq %rbx,136(%rsp) + call mont_reduce + + + leaq 448(%rsp),%rcx + xorq %rax,%rax + movq %rax,0(%rcx) + movq %rax,8(%rcx) + movq %rax,24(%rcx) + movq %rax,32(%rcx) + movq %rax,40(%rcx) + movq %rax,48(%rcx) + movq %rax,56(%rcx) + movq %rax,128(%rsp) + movq $1,16(%rcx) + + leaq 640(%rsp),%rbp + movq %rcx,%rsi + movq %rbp,%rdi + movq $8,%rax +loop_0: + movq (%rcx),%rbx + movw %bx,(%rdi) + shrq $16,%rbx + movw %bx,64(%rdi) + shrq $16,%rbx + movw %bx,128(%rdi) + shrq $16,%rbx + movw %bx,192(%rdi) + leaq 8(%rcx),%rcx + leaq 256(%rdi),%rdi + decq %rax + jnz loop_0 + movq $31,%rax + movq %rax,32(%rsp) + movq %rbp,40(%rsp) + + movq %rsi,136(%rsp) + movq 0(%rsi),%r10 + movq 8(%rsi),%r11 + movq 16(%rsi),%r12 + movq 24(%rsi),%r13 + movq 32(%rsi),%r14 + movq 40(%rsi),%r15 + movq 48(%rsi),%r8 + movq 56(%rsi),%r9 +init_loop: + leaq 384(%rsp),%rdi + call mont_mul_a3b + leaq 448(%rsp),%rsi + movq 40(%rsp),%rbp + addq $2,%rbp + movq %rbp,40(%rsp) + movq %rsi,%rcx + movq $8,%rax +loop_1: + movq (%rcx),%rbx + movw %bx,(%rbp) + shrq $16,%rbx + movw %bx,64(%rbp) + shrq $16,%rbx + movw %bx,128(%rbp) + shrq $16,%rbx + movw %bx,192(%rbp) + leaq 8(%rcx),%rcx + leaq 256(%rbp),%rbp + decq %rax + jnz loop_1 + movq 32(%rsp),%rax + subq $1,%rax + movq %rax,32(%rsp) + jne init_loop + + + + movdqa %xmm0,64(%rsp) + movdqa %xmm1,80(%rsp) + movdqa %xmm2,96(%rsp) + movdqa %xmm3,112(%rsp) + + + + + + movl 126(%rsp),%eax + movq %rax,%rdx + shrq $11,%rax + andl $2047,%edx + movl %edx,126(%rsp) + leaq 640(%rsp,%rax,2),%rsi + movq 8(%rsp),%rdx + movq $4,%rbp +loop_2: + movzwq 192(%rsi),%rbx + movzwq 448(%rsi),%rax + shlq $16,%rbx + shlq $16,%rax + movw 128(%rsi),%bx + movw 384(%rsi),%ax + shlq $16,%rbx + shlq $16,%rax + movw 64(%rsi),%bx + movw 320(%rsi),%ax + shlq $16,%rbx + shlq $16,%rax + movw 0(%rsi),%bx + movw 256(%rsi),%ax + movq %rbx,0(%rdx) + movq %rax,8(%rdx) + leaq 512(%rsi),%rsi + leaq 16(%rdx),%rdx + subq $1,%rbp + jnz loop_2 + movq $505,48(%rsp) + + movq 8(%rsp),%rcx + movq %rcx,136(%rsp) + movq 0(%rcx),%r10 + movq 8(%rcx),%r11 + movq 16(%rcx),%r12 + movq 24(%rcx),%r13 + movq 32(%rcx),%r14 + movq 40(%rcx),%r15 + movq 48(%rcx),%r8 + movq 56(%rcx),%r9 + jmp sqr_2 + +main_loop_a3b: + call sqr_reduce + call sqr_reduce + call sqr_reduce +sqr_2: + call sqr_reduce + call sqr_reduce + + + + movq 48(%rsp),%rcx + movq %rcx,%rax + shrq $4,%rax + movl 64(%rsp,%rax,2),%edx + andq $15,%rcx + shrq %cl,%rdx + andq $31,%rdx + + leaq 640(%rsp,%rdx,2),%rsi + leaq 448(%rsp),%rdx + movq %rdx,%rdi + movq $4,%rbp +loop_3: + movzwq 192(%rsi),%rbx + movzwq 448(%rsi),%rax + shlq $16,%rbx + shlq $16,%rax + movw 128(%rsi),%bx + movw 384(%rsi),%ax + shlq $16,%rbx + shlq $16,%rax + movw 64(%rsi),%bx + movw 320(%rsi),%ax + shlq $16,%rbx + shlq $16,%rax + movw 0(%rsi),%bx + movw 256(%rsi),%ax + movq %rbx,0(%rdx) + movq %rax,8(%rdx) + leaq 512(%rsi),%rsi + leaq 16(%rdx),%rdx + subq $1,%rbp + jnz loop_3 + movq 8(%rsp),%rsi + call mont_mul_a3b + + + + movq 48(%rsp),%rcx + subq $5,%rcx + movq %rcx,48(%rsp) + jge main_loop_a3b + + + +end_main_loop_a3b: + + + movq 8(%rsp),%rdx + pxor %xmm4,%xmm4 + movdqu 0(%rdx),%xmm0 + movdqu 16(%rdx),%xmm1 + movdqu 32(%rdx),%xmm2 + movdqu 48(%rdx),%xmm3 + movdqa %xmm4,576(%rsp) + movdqa %xmm4,592(%rsp) + movdqa %xmm4,608(%rsp) + movdqa %xmm4,624(%rsp) + movdqa %xmm0,512(%rsp) + movdqa %xmm1,528(%rsp) + movdqa %xmm2,544(%rsp) + movdqa %xmm3,560(%rsp) + call mont_reduce + + + + movq 8(%rsp),%rax + movq 0(%rax),%r8 + movq 8(%rax),%r9 + movq 16(%rax),%r10 + movq 24(%rax),%r11 + movq 32(%rax),%r12 + movq 40(%rax),%r13 + movq 48(%rax),%r14 + movq 56(%rax),%r15 + + + movq 24(%rsp),%rbx + addq $512,%rbx + + subq 0(%rbx),%r8 + sbbq 8(%rbx),%r9 + sbbq 16(%rbx),%r10 + sbbq 24(%rbx),%r11 + sbbq 32(%rbx),%r12 + sbbq 40(%rbx),%r13 + sbbq 48(%rbx),%r14 + sbbq 56(%rbx),%r15 + + + movq 0(%rax),%rsi + movq 8(%rax),%rdi + movq 16(%rax),%rcx + movq 24(%rax),%rdx + cmovncq %r8,%rsi + cmovncq %r9,%rdi + cmovncq %r10,%rcx + cmovncq %r11,%rdx + movq %rsi,0(%rax) + movq %rdi,8(%rax) + movq %rcx,16(%rax) + movq %rdx,24(%rax) + + movq 32(%rax),%rsi + movq 40(%rax),%rdi + movq 48(%rax),%rcx + movq 56(%rax),%rdx + cmovncq %r12,%rsi + cmovncq %r13,%rdi + cmovncq %r14,%rcx + cmovncq %r15,%rdx + movq %rsi,32(%rax) + movq %rdi,40(%rax) + movq %rcx,48(%rax) + movq %rdx,56(%rax) + + movq 0(%rsp),%rsi + movq 0(%rsi),%r15 + movq 8(%rsi),%r14 + movq 16(%rsi),%r13 + movq 24(%rsi),%r12 + movq 32(%rsi),%rbx + movq 40(%rsi),%rbp + leaq 48(%rsi),%rsp +.Lepilogue: + .byte 0xf3,0xc3 +.size mod_exp_512, . - mod_exp_512 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/modexp512-x86_64.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/modexp512-x86_64.pl new file mode 100644 index 00000000..bfd6e975 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/modexp512-x86_64.pl @@ -0,0 +1,1497 @@ +#!/usr/bin/env perl +# +# Copyright (c) 2010-2011 Intel Corp. +# Author: Vinodh.Gopal@intel.com +# Jim Guilford +# Erdinc.Ozturk@intel.com +# Maxim.Perminov@intel.com +# +# More information about algorithm used can be found at: +# http://www.cse.buffalo.edu/srds2009/escs2009_submission_Gopal.pdf +# +# ==================================================================== +# Copyright (c) 2011 The OpenSSL Project. All rights reserved. +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions +# are met: +# +# 1. Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# 2. Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in +# the documentation and/or other materials provided with the +# distribution. +# +# 3. All advertising materials mentioning features or use of this +# software must display the following acknowledgment: +# "This product includes software developed by the OpenSSL Project +# for use in the OpenSSL Toolkit. (http://www.OpenSSL.org/)" +# +# 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to +# endorse or promote products derived from this software without +# prior written permission. For written permission, please contact +# licensing@OpenSSL.org. +# +# 5. Products derived from this software may not be called "OpenSSL" +# nor may "OpenSSL" appear in their names without prior written +# permission of the OpenSSL Project. +# +# 6. Redistributions of any form whatsoever must retain the following +# acknowledgment: +# "This product includes software developed by the OpenSSL Project +# for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)" +# +# THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY +# EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR +# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OpenSSL PROJECT OR +# ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT +# NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +# HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +# STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED +# OF THE POSSIBILITY OF SUCH DAMAGE. +# ==================================================================== + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +my $win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open OUT,"| \"$^X\" $xlate $flavour $output"; +*STDOUT=*OUT; + +use strict; +my $code=".text\n\n"; +my $m=0; + +# +# Define x512 macros +# + +#MULSTEP_512_ADD MACRO x7, x6, x5, x4, x3, x2, x1, x0, dst, src1, src2, add_src, tmp1, tmp2 +# +# uses rax, rdx, and args +sub MULSTEP_512_ADD +{ + my ($x, $DST, $SRC2, $ASRC, $OP, $TMP)=@_; + my @X=@$x; # make a copy +$code.=<<___; + mov (+8*0)($SRC2), %rax + mul $OP # rdx:rax = %OP * [0] + mov ($ASRC), $X[0] + add %rax, $X[0] + adc \$0, %rdx + mov $X[0], $DST +___ +for(my $i=1;$i<8;$i++) { +$code.=<<___; + mov %rdx, $TMP + + mov (+8*$i)($SRC2), %rax + mul $OP # rdx:rax = %OP * [$i] + mov (+8*$i)($ASRC), $X[$i] + add %rax, $X[$i] + adc \$0, %rdx + add $TMP, $X[$i] + adc \$0, %rdx +___ +} +$code.=<<___; + mov %rdx, $X[0] +___ +} + +#MULSTEP_512 MACRO x7, x6, x5, x4, x3, x2, x1, x0, dst, src2, src1_val, tmp +# +# uses rax, rdx, and args +sub MULSTEP_512 +{ + my ($x, $DST, $SRC2, $OP, $TMP)=@_; + my @X=@$x; # make a copy +$code.=<<___; + mov (+8*0)($SRC2), %rax + mul $OP # rdx:rax = %OP * [0] + add %rax, $X[0] + adc \$0, %rdx + mov $X[0], $DST +___ +for(my $i=1;$i<8;$i++) { +$code.=<<___; + mov %rdx, $TMP + + mov (+8*$i)($SRC2), %rax + mul $OP # rdx:rax = %OP * [$i] + add %rax, $X[$i] + adc \$0, %rdx + add $TMP, $X[$i] + adc \$0, %rdx +___ +} +$code.=<<___; + mov %rdx, $X[0] +___ +} + +# +# Swizzle Macros +# + +# macro to copy data from flat space to swizzled table +#MACRO swizzle pDst, pSrc, tmp1, tmp2 +# pDst and pSrc are modified +sub swizzle +{ + my ($pDst, $pSrc, $cnt, $d0)=@_; +$code.=<<___; + mov \$8, $cnt +loop_$m: + mov ($pSrc), $d0 + mov $d0#w, ($pDst) + shr \$16, $d0 + mov $d0#w, (+64*1)($pDst) + shr \$16, $d0 + mov $d0#w, (+64*2)($pDst) + shr \$16, $d0 + mov $d0#w, (+64*3)($pDst) + lea 8($pSrc), $pSrc + lea 64*4($pDst), $pDst + dec $cnt + jnz loop_$m +___ + + $m++; +} + +# macro to copy data from swizzled table to flat space +#MACRO unswizzle pDst, pSrc, tmp*3 +sub unswizzle +{ + my ($pDst, $pSrc, $cnt, $d0, $d1)=@_; +$code.=<<___; + mov \$4, $cnt +loop_$m: + movzxw (+64*3+256*0)($pSrc), $d0 + movzxw (+64*3+256*1)($pSrc), $d1 + shl \$16, $d0 + shl \$16, $d1 + mov (+64*2+256*0)($pSrc), $d0#w + mov (+64*2+256*1)($pSrc), $d1#w + shl \$16, $d0 + shl \$16, $d1 + mov (+64*1+256*0)($pSrc), $d0#w + mov (+64*1+256*1)($pSrc), $d1#w + shl \$16, $d0 + shl \$16, $d1 + mov (+64*0+256*0)($pSrc), $d0#w + mov (+64*0+256*1)($pSrc), $d1#w + mov $d0, (+8*0)($pDst) + mov $d1, (+8*1)($pDst) + lea 256*2($pSrc), $pSrc + lea 8*2($pDst), $pDst + sub \$1, $cnt + jnz loop_$m +___ + + $m++; +} + +# +# Data Structures +# + +# Reduce Data +# +# +# Offset Value +# 0C0 Carries +# 0B8 X2[10] +# 0B0 X2[9] +# 0A8 X2[8] +# 0A0 X2[7] +# 098 X2[6] +# 090 X2[5] +# 088 X2[4] +# 080 X2[3] +# 078 X2[2] +# 070 X2[1] +# 068 X2[0] +# 060 X1[12] P[10] +# 058 X1[11] P[9] Z[8] +# 050 X1[10] P[8] Z[7] +# 048 X1[9] P[7] Z[6] +# 040 X1[8] P[6] Z[5] +# 038 X1[7] P[5] Z[4] +# 030 X1[6] P[4] Z[3] +# 028 X1[5] P[3] Z[2] +# 020 X1[4] P[2] Z[1] +# 018 X1[3] P[1] Z[0] +# 010 X1[2] P[0] Y[2] +# 008 X1[1] Q[1] Y[1] +# 000 X1[0] Q[0] Y[0] + +my $X1_offset = 0; # 13 qwords +my $X2_offset = $X1_offset + 13*8; # 11 qwords +my $Carries_offset = $X2_offset + 11*8; # 1 qword +my $Q_offset = 0; # 2 qwords +my $P_offset = $Q_offset + 2*8; # 11 qwords +my $Y_offset = 0; # 3 qwords +my $Z_offset = $Y_offset + 3*8; # 9 qwords + +my $Red_Data_Size = $Carries_offset + 1*8; # (25 qwords) + +# +# Stack Frame +# +# +# offset value +# ... +# ... +# 280 Garray + +# 278 tmp16[15] +# ... ... +# 200 tmp16[0] + +# 1F8 tmp[7] +# ... ... +# 1C0 tmp[0] + +# 1B8 GT[7] +# ... ... +# 180 GT[0] + +# 178 Reduce Data +# ... ... +# 0B8 Reduce Data +# 0B0 reserved +# 0A8 reserved +# 0A0 reserved +# 098 reserved +# 090 reserved +# 088 reduce result addr +# 080 exp[8] + +# ... +# 048 exp[1] +# 040 exp[0] + +# 038 reserved +# 030 loop_idx +# 028 pg +# 020 i +# 018 pData ; arg 4 +# 010 pG ; arg 2 +# 008 pResult ; arg 1 +# 000 rsp ; stack pointer before subtract + +my $rsp_offset = 0; +my $pResult_offset = 8*1 + $rsp_offset; +my $pG_offset = 8*1 + $pResult_offset; +my $pData_offset = 8*1 + $pG_offset; +my $i_offset = 8*1 + $pData_offset; +my $pg_offset = 8*1 + $i_offset; +my $loop_idx_offset = 8*1 + $pg_offset; +my $reserved1_offset = 8*1 + $loop_idx_offset; +my $exp_offset = 8*1 + $reserved1_offset; +my $red_result_addr_offset= 8*9 + $exp_offset; +my $reserved2_offset = 8*1 + $red_result_addr_offset; +my $Reduce_Data_offset = 8*5 + $reserved2_offset; +my $GT_offset = $Red_Data_Size + $Reduce_Data_offset; +my $tmp_offset = 8*8 + $GT_offset; +my $tmp16_offset = 8*8 + $tmp_offset; +my $garray_offset = 8*16 + $tmp16_offset; +my $mem_size = 8*8*32 + $garray_offset; + +# +# Offsets within Reduce Data +# +# +# struct MODF_2FOLD_MONT_512_C1_DATA { +# UINT64 t[8][8]; +# UINT64 m[8]; +# UINT64 m1[8]; /* 2^768 % m */ +# UINT64 m2[8]; /* 2^640 % m */ +# UINT64 k1[2]; /* (- 1/m) % 2^128 */ +# }; + +my $T = 0; +my $M = 512; # = 8 * 8 * 8 +my $M1 = 576; # = 8 * 8 * 9 /* += 8 * 8 */ +my $M2 = 640; # = 8 * 8 * 10 /* += 8 * 8 */ +my $K1 = 704; # = 8 * 8 * 11 /* += 8 * 8 */ + +# +# FUNCTIONS +# + +{{{ +# +# MULADD_128x512 : Function to multiply 128-bits (2 qwords) by 512-bits (8 qwords) +# and add 512-bits (8 qwords) +# to get 640 bits (10 qwords) +# Input: 128-bit mul source: [rdi+8*1], rbp +# 512-bit mul source: [rsi+8*n] +# 512-bit add source: r15, r14, ..., r9, r8 +# Output: r9, r8, r15, r14, r13, r12, r11, r10, [rcx+8*1], [rcx+8*0] +# Clobbers all regs except: rcx, rsi, rdi +$code.=<<___; +.type MULADD_128x512,\@abi-omnipotent +.align 16 +MULADD_128x512: +___ + &MULSTEP_512([map("%r$_",(8..15))], "(+8*0)(%rcx)", "%rsi", "%rbp", "%rbx"); +$code.=<<___; + mov (+8*1)(%rdi), %rbp +___ + &MULSTEP_512([map("%r$_",(9..15,8))], "(+8*1)(%rcx)", "%rsi", "%rbp", "%rbx"); +$code.=<<___; + ret +.size MULADD_128x512,.-MULADD_128x512 +___ +}}} + +{{{ +#MULADD_256x512 MACRO pDst, pA, pB, OP, TMP, X7, X6, X5, X4, X3, X2, X1, X0 +# +# Inputs: pDst: Destination (768 bits, 12 qwords) +# pA: Multiplicand (1024 bits, 16 qwords) +# pB: Multiplicand (512 bits, 8 qwords) +# Dst = Ah * B + Al +# where Ah is (in qwords) A[15:12] (256 bits) and Al is A[7:0] (512 bits) +# Results in X3 X2 X1 X0 X7 X6 X5 X4 Dst[3:0] +# Uses registers: arguments, RAX, RDX +sub MULADD_256x512 +{ + my ($pDst, $pA, $pB, $OP, $TMP, $X)=@_; +$code.=<<___; + mov (+8*12)($pA), $OP +___ + &MULSTEP_512_ADD($X, "(+8*0)($pDst)", $pB, $pA, $OP, $TMP); + push(@$X,shift(@$X)); + +$code.=<<___; + mov (+8*13)($pA), $OP +___ + &MULSTEP_512($X, "(+8*1)($pDst)", $pB, $OP, $TMP); + push(@$X,shift(@$X)); + +$code.=<<___; + mov (+8*14)($pA), $OP +___ + &MULSTEP_512($X, "(+8*2)($pDst)", $pB, $OP, $TMP); + push(@$X,shift(@$X)); + +$code.=<<___; + mov (+8*15)($pA), $OP +___ + &MULSTEP_512($X, "(+8*3)($pDst)", $pB, $OP, $TMP); + push(@$X,shift(@$X)); +} + +# +# mont_reduce(UINT64 *x, /* 1024 bits, 16 qwords */ +# UINT64 *m, /* 512 bits, 8 qwords */ +# MODF_2FOLD_MONT_512_C1_DATA *data, +# UINT64 *r) /* 512 bits, 8 qwords */ +# Input: x (number to be reduced): tmp16 (Implicit) +# m (modulus): [pM] (Implicit) +# data (reduce data): [pData] (Implicit) +# Output: r (result): Address in [red_res_addr] +# result also in: r9, r8, r15, r14, r13, r12, r11, r10 + +my @X=map("%r$_",(8..15)); + +$code.=<<___; +.type mont_reduce,\@abi-omnipotent +.align 16 +mont_reduce: +___ + +my $STACK_DEPTH = 8; + # + # X1 = Xh * M1 + Xl +$code.=<<___; + lea (+$Reduce_Data_offset+$X1_offset+$STACK_DEPTH)(%rsp), %rdi # pX1 (Dst) 769 bits, 13 qwords + mov (+$pData_offset+$STACK_DEPTH)(%rsp), %rsi # pM1 (Bsrc) 512 bits, 8 qwords + add \$$M1, %rsi + lea (+$tmp16_offset+$STACK_DEPTH)(%rsp), %rcx # X (Asrc) 1024 bits, 16 qwords + +___ + + &MULADD_256x512("%rdi", "%rcx", "%rsi", "%rbp", "%rbx", \@X); # rotates @X 4 times + # results in r11, r10, r9, r8, r15, r14, r13, r12, X1[3:0] + +$code.=<<___; + xor %rax, %rax + # X1 += xl + add (+8*8)(%rcx), $X[4] + adc (+8*9)(%rcx), $X[5] + adc (+8*10)(%rcx), $X[6] + adc (+8*11)(%rcx), $X[7] + adc \$0, %rax + # X1 is now rax, r11-r8, r15-r12, tmp16[3:0] + + # + # check for carry ;; carry stored in rax + mov $X[4], (+8*8)(%rdi) # rdi points to X1 + mov $X[5], (+8*9)(%rdi) + mov $X[6], %rbp + mov $X[7], (+8*11)(%rdi) + + mov %rax, (+$Reduce_Data_offset+$Carries_offset+$STACK_DEPTH)(%rsp) + + mov (+8*0)(%rdi), $X[4] + mov (+8*1)(%rdi), $X[5] + mov (+8*2)(%rdi), $X[6] + mov (+8*3)(%rdi), $X[7] + + # X1 is now stored in: X1[11], rbp, X1[9:8], r15-r8 + # rdi -> X1 + # rsi -> M1 + + # + # X2 = Xh * M2 + Xl + # do first part (X2 = Xh * M2) + add \$8*10, %rdi # rdi -> pXh ; 128 bits, 2 qwords + # Xh is actually { [rdi+8*1], rbp } + add \$`$M2-$M1`, %rsi # rsi -> M2 + lea (+$Reduce_Data_offset+$X2_offset+$STACK_DEPTH)(%rsp), %rcx # rcx -> pX2 ; 641 bits, 11 qwords +___ + unshift(@X,pop(@X)); unshift(@X,pop(@X)); +$code.=<<___; + + call MULADD_128x512 # args in rcx, rdi / rbp, rsi, r15-r8 + # result in r9, r8, r15, r14, r13, r12, r11, r10, X2[1:0] + mov (+$Reduce_Data_offset+$Carries_offset+$STACK_DEPTH)(%rsp), %rax + + # X2 += Xl + add (+8*8-8*10)(%rdi), $X[6] # (-8*10) is to adjust rdi -> Xh to Xl + adc (+8*9-8*10)(%rdi), $X[7] + mov $X[6], (+8*8)(%rcx) + mov $X[7], (+8*9)(%rcx) + + adc %rax, %rax + mov %rax, (+$Reduce_Data_offset+$Carries_offset+$STACK_DEPTH)(%rsp) + + lea (+$Reduce_Data_offset+$Q_offset+$STACK_DEPTH)(%rsp), %rdi # rdi -> pQ ; 128 bits, 2 qwords + add \$`$K1-$M2`, %rsi # rsi -> pK1 ; 128 bits, 2 qwords + + # MUL_128x128t128 rdi, rcx, rsi ; Q = X2 * K1 (bottom half) + # B1:B0 = rsi[1:0] = K1[1:0] + # A1:A0 = rcx[1:0] = X2[1:0] + # Result = rdi[1],rbp = Q[1],rbp + mov (%rsi), %r8 # B0 + mov (+8*1)(%rsi), %rbx # B1 + + mov (%rcx), %rax # A0 + mul %r8 # B0 + mov %rax, %rbp + mov %rdx, %r9 + + mov (+8*1)(%rcx), %rax # A1 + mul %r8 # B0 + add %rax, %r9 + + mov (%rcx), %rax # A0 + mul %rbx # B1 + add %rax, %r9 + + mov %r9, (+8*1)(%rdi) + # end MUL_128x128t128 + + sub \$`$K1-$M`, %rsi + + mov (%rcx), $X[6] + mov (+8*1)(%rcx), $X[7] # r9:r8 = X2[1:0] + + call MULADD_128x512 # args in rcx, rdi / rbp, rsi, r15-r8 + # result in r9, r8, r15, r14, r13, r12, r11, r10, X2[1:0] + + # load first half of m to rdx, rdi, rbx, rax + # moved this here for efficiency + mov (+8*0)(%rsi), %rax + mov (+8*1)(%rsi), %rbx + mov (+8*2)(%rsi), %rdi + mov (+8*3)(%rsi), %rdx + + # continue with reduction + mov (+$Reduce_Data_offset+$Carries_offset+$STACK_DEPTH)(%rsp), %rbp + + add (+8*8)(%rcx), $X[6] + adc (+8*9)(%rcx), $X[7] + + #accumulate the final carry to rbp + adc %rbp, %rbp + + # Add in overflow corrections: R = (X2>>128) += T[overflow] + # R = {r9, r8, r15, r14, ..., r10} + shl \$3, %rbp + mov (+$pData_offset+$STACK_DEPTH)(%rsp), %rcx # rsi -> Data (and points to T) + add %rcx, %rbp # pT ; 512 bits, 8 qwords, spread out + + # rsi will be used to generate a mask after the addition + xor %rsi, %rsi + + add (+8*8*0)(%rbp), $X[0] + adc (+8*8*1)(%rbp), $X[1] + adc (+8*8*2)(%rbp), $X[2] + adc (+8*8*3)(%rbp), $X[3] + adc (+8*8*4)(%rbp), $X[4] + adc (+8*8*5)(%rbp), $X[5] + adc (+8*8*6)(%rbp), $X[6] + adc (+8*8*7)(%rbp), $X[7] + + # if there is a carry: rsi = 0xFFFFFFFFFFFFFFFF + # if carry is clear: rsi = 0x0000000000000000 + sbb \$0, %rsi + + # if carry is clear, subtract 0. Otherwise, subtract 256 bits of m + and %rsi, %rax + and %rsi, %rbx + and %rsi, %rdi + and %rsi, %rdx + + mov \$1, %rbp + sub %rax, $X[0] + sbb %rbx, $X[1] + sbb %rdi, $X[2] + sbb %rdx, $X[3] + + # if there is a borrow: rbp = 0 + # if there is no borrow: rbp = 1 + # this is used to save the borrows in between the first half and the 2nd half of the subtraction of m + sbb \$0, %rbp + + #load second half of m to rdx, rdi, rbx, rax + + add \$$M, %rcx + mov (+8*4)(%rcx), %rax + mov (+8*5)(%rcx), %rbx + mov (+8*6)(%rcx), %rdi + mov (+8*7)(%rcx), %rdx + + # use the rsi mask as before + # if carry is clear, subtract 0. Otherwise, subtract 256 bits of m + and %rsi, %rax + and %rsi, %rbx + and %rsi, %rdi + and %rsi, %rdx + + # if rbp = 0, there was a borrow before, it is moved to the carry flag + # if rbp = 1, there was not a borrow before, carry flag is cleared + sub \$1, %rbp + + sbb %rax, $X[4] + sbb %rbx, $X[5] + sbb %rdi, $X[6] + sbb %rdx, $X[7] + + # write R back to memory + + mov (+$red_result_addr_offset+$STACK_DEPTH)(%rsp), %rsi + mov $X[0], (+8*0)(%rsi) + mov $X[1], (+8*1)(%rsi) + mov $X[2], (+8*2)(%rsi) + mov $X[3], (+8*3)(%rsi) + mov $X[4], (+8*4)(%rsi) + mov $X[5], (+8*5)(%rsi) + mov $X[6], (+8*6)(%rsi) + mov $X[7], (+8*7)(%rsi) + + ret +.size mont_reduce,.-mont_reduce +___ +}}} + +{{{ +#MUL_512x512 MACRO pDst, pA, pB, x7, x6, x5, x4, x3, x2, x1, x0, tmp*2 +# +# Inputs: pDst: Destination (1024 bits, 16 qwords) +# pA: Multiplicand (512 bits, 8 qwords) +# pB: Multiplicand (512 bits, 8 qwords) +# Uses registers rax, rdx, args +# B operand in [pB] and also in x7...x0 +sub MUL_512x512 +{ + my ($pDst, $pA, $pB, $x, $OP, $TMP, $pDst_o)=@_; + my ($pDst, $pDst_o) = ($pDst =~ m/([^+]*)\+?(.*)?/); + my @X=@$x; # make a copy + +$code.=<<___; + mov (+8*0)($pA), $OP + + mov $X[0], %rax + mul $OP # rdx:rax = %OP * [0] + mov %rax, (+$pDst_o+8*0)($pDst) + mov %rdx, $X[0] +___ +for(my $i=1;$i<8;$i++) { +$code.=<<___; + mov $X[$i], %rax + mul $OP # rdx:rax = %OP * [$i] + add %rax, $X[$i-1] + adc \$0, %rdx + mov %rdx, $X[$i] +___ +} + +for(my $i=1;$i<8;$i++) { +$code.=<<___; + mov (+8*$i)($pA), $OP +___ + + &MULSTEP_512(\@X, "(+$pDst_o+8*$i)($pDst)", $pB, $OP, $TMP); + push(@X,shift(@X)); +} + +$code.=<<___; + mov $X[0], (+$pDst_o+8*8)($pDst) + mov $X[1], (+$pDst_o+8*9)($pDst) + mov $X[2], (+$pDst_o+8*10)($pDst) + mov $X[3], (+$pDst_o+8*11)($pDst) + mov $X[4], (+$pDst_o+8*12)($pDst) + mov $X[5], (+$pDst_o+8*13)($pDst) + mov $X[6], (+$pDst_o+8*14)($pDst) + mov $X[7], (+$pDst_o+8*15)($pDst) +___ +} + +# +# mont_mul_a3b : subroutine to compute (Src1 * Src2) % M (all 512-bits) +# Input: src1: Address of source 1: rdi +# src2: Address of source 2: rsi +# Output: dst: Address of destination: [red_res_addr] +# src2 and result also in: r9, r8, r15, r14, r13, r12, r11, r10 +# Temp: Clobbers [tmp16], all registers +$code.=<<___; +.type mont_mul_a3b,\@abi-omnipotent +.align 16 +mont_mul_a3b: + # + # multiply tmp = src1 * src2 + # For multiply: dst = rcx, src1 = rdi, src2 = rsi + # stack depth is extra 8 from call +___ + &MUL_512x512("%rsp+$tmp16_offset+8", "%rdi", "%rsi", [map("%r$_",(10..15,8..9))], "%rbp", "%rbx"); +$code.=<<___; + # + # Dst = tmp % m + # Call reduce(tmp, m, data, dst) + + # tail recursion optimization: jmp to mont_reduce and return from there + jmp mont_reduce + # call mont_reduce + # ret +.size mont_mul_a3b,.-mont_mul_a3b +___ +}}} + +{{{ +#SQR_512 MACRO pDest, pA, x7, x6, x5, x4, x3, x2, x1, x0, tmp*4 +# +# Input in memory [pA] and also in x7...x0 +# Uses all argument registers plus rax and rdx +# +# This version computes all of the off-diagonal terms into memory, +# and then it adds in the diagonal terms + +sub SQR_512 +{ + my ($pDst, $pA, $x, $A, $tmp, $x7, $x6, $pDst_o)=@_; + my ($pDst, $pDst_o) = ($pDst =~ m/([^+]*)\+?(.*)?/); + my @X=@$x; # make a copy +$code.=<<___; + # ------------------ + # first pass 01...07 + # ------------------ + mov $X[0], $A + + mov $X[1],%rax + mul $A + mov %rax, (+$pDst_o+8*1)($pDst) +___ +for(my $i=2;$i<8;$i++) { +$code.=<<___; + mov %rdx, $X[$i-2] + mov $X[$i],%rax + mul $A + add %rax, $X[$i-2] + adc \$0, %rdx +___ +} +$code.=<<___; + mov %rdx, $x7 + + mov $X[0], (+$pDst_o+8*2)($pDst) + + # ------------------ + # second pass 12...17 + # ------------------ + + mov (+8*1)($pA), $A + + mov (+8*2)($pA),%rax + mul $A + add %rax, $X[1] + adc \$0, %rdx + mov $X[1], (+$pDst_o+8*3)($pDst) + + mov %rdx, $X[0] + mov (+8*3)($pA),%rax + mul $A + add %rax, $X[2] + adc \$0, %rdx + add $X[0], $X[2] + adc \$0, %rdx + mov $X[2], (+$pDst_o+8*4)($pDst) + + mov %rdx, $X[0] + mov (+8*4)($pA),%rax + mul $A + add %rax, $X[3] + adc \$0, %rdx + add $X[0], $X[3] + adc \$0, %rdx + + mov %rdx, $X[0] + mov (+8*5)($pA),%rax + mul $A + add %rax, $X[4] + adc \$0, %rdx + add $X[0], $X[4] + adc \$0, %rdx + + mov %rdx, $X[0] + mov $X[6],%rax + mul $A + add %rax, $X[5] + adc \$0, %rdx + add $X[0], $X[5] + adc \$0, %rdx + + mov %rdx, $X[0] + mov $X[7],%rax + mul $A + add %rax, $x7 + adc \$0, %rdx + add $X[0], $x7 + adc \$0, %rdx + + mov %rdx, $X[1] + + # ------------------ + # third pass 23...27 + # ------------------ + mov (+8*2)($pA), $A + + mov (+8*3)($pA),%rax + mul $A + add %rax, $X[3] + adc \$0, %rdx + mov $X[3], (+$pDst_o+8*5)($pDst) + + mov %rdx, $X[0] + mov (+8*4)($pA),%rax + mul $A + add %rax, $X[4] + adc \$0, %rdx + add $X[0], $X[4] + adc \$0, %rdx + mov $X[4], (+$pDst_o+8*6)($pDst) + + mov %rdx, $X[0] + mov (+8*5)($pA),%rax + mul $A + add %rax, $X[5] + adc \$0, %rdx + add $X[0], $X[5] + adc \$0, %rdx + + mov %rdx, $X[0] + mov $X[6],%rax + mul $A + add %rax, $x7 + adc \$0, %rdx + add $X[0], $x7 + adc \$0, %rdx + + mov %rdx, $X[0] + mov $X[7],%rax + mul $A + add %rax, $X[1] + adc \$0, %rdx + add $X[0], $X[1] + adc \$0, %rdx + + mov %rdx, $X[2] + + # ------------------ + # fourth pass 34...37 + # ------------------ + + mov (+8*3)($pA), $A + + mov (+8*4)($pA),%rax + mul $A + add %rax, $X[5] + adc \$0, %rdx + mov $X[5], (+$pDst_o+8*7)($pDst) + + mov %rdx, $X[0] + mov (+8*5)($pA),%rax + mul $A + add %rax, $x7 + adc \$0, %rdx + add $X[0], $x7 + adc \$0, %rdx + mov $x7, (+$pDst_o+8*8)($pDst) + + mov %rdx, $X[0] + mov $X[6],%rax + mul $A + add %rax, $X[1] + adc \$0, %rdx + add $X[0], $X[1] + adc \$0, %rdx + + mov %rdx, $X[0] + mov $X[7],%rax + mul $A + add %rax, $X[2] + adc \$0, %rdx + add $X[0], $X[2] + adc \$0, %rdx + + mov %rdx, $X[5] + + # ------------------ + # fifth pass 45...47 + # ------------------ + mov (+8*4)($pA), $A + + mov (+8*5)($pA),%rax + mul $A + add %rax, $X[1] + adc \$0, %rdx + mov $X[1], (+$pDst_o+8*9)($pDst) + + mov %rdx, $X[0] + mov $X[6],%rax + mul $A + add %rax, $X[2] + adc \$0, %rdx + add $X[0], $X[2] + adc \$0, %rdx + mov $X[2], (+$pDst_o+8*10)($pDst) + + mov %rdx, $X[0] + mov $X[7],%rax + mul $A + add %rax, $X[5] + adc \$0, %rdx + add $X[0], $X[5] + adc \$0, %rdx + + mov %rdx, $X[1] + + # ------------------ + # sixth pass 56...57 + # ------------------ + mov (+8*5)($pA), $A + + mov $X[6],%rax + mul $A + add %rax, $X[5] + adc \$0, %rdx + mov $X[5], (+$pDst_o+8*11)($pDst) + + mov %rdx, $X[0] + mov $X[7],%rax + mul $A + add %rax, $X[1] + adc \$0, %rdx + add $X[0], $X[1] + adc \$0, %rdx + mov $X[1], (+$pDst_o+8*12)($pDst) + + mov %rdx, $X[2] + + # ------------------ + # seventh pass 67 + # ------------------ + mov $X[6], $A + + mov $X[7],%rax + mul $A + add %rax, $X[2] + adc \$0, %rdx + mov $X[2], (+$pDst_o+8*13)($pDst) + + mov %rdx, (+$pDst_o+8*14)($pDst) + + # start finalize (add in squares, and double off-terms) + mov (+$pDst_o+8*1)($pDst), $X[0] + mov (+$pDst_o+8*2)($pDst), $X[1] + mov (+$pDst_o+8*3)($pDst), $X[2] + mov (+$pDst_o+8*4)($pDst), $X[3] + mov (+$pDst_o+8*5)($pDst), $X[4] + mov (+$pDst_o+8*6)($pDst), $X[5] + + mov (+8*3)($pA), %rax + mul %rax + mov %rax, $x6 + mov %rdx, $X[6] + + add $X[0], $X[0] + adc $X[1], $X[1] + adc $X[2], $X[2] + adc $X[3], $X[3] + adc $X[4], $X[4] + adc $X[5], $X[5] + adc \$0, $X[6] + + mov (+8*0)($pA), %rax + mul %rax + mov %rax, (+$pDst_o+8*0)($pDst) + mov %rdx, $A + + mov (+8*1)($pA), %rax + mul %rax + + add $A, $X[0] + adc %rax, $X[1] + adc \$0, %rdx + + mov %rdx, $A + mov $X[0], (+$pDst_o+8*1)($pDst) + mov $X[1], (+$pDst_o+8*2)($pDst) + + mov (+8*2)($pA), %rax + mul %rax + + add $A, $X[2] + adc %rax, $X[3] + adc \$0, %rdx + + mov %rdx, $A + + mov $X[2], (+$pDst_o+8*3)($pDst) + mov $X[3], (+$pDst_o+8*4)($pDst) + + xor $tmp, $tmp + add $A, $X[4] + adc $x6, $X[5] + adc \$0, $tmp + + mov $X[4], (+$pDst_o+8*5)($pDst) + mov $X[5], (+$pDst_o+8*6)($pDst) + + # %%tmp has 0/1 in column 7 + # %%A6 has a full value in column 7 + + mov (+$pDst_o+8*7)($pDst), $X[0] + mov (+$pDst_o+8*8)($pDst), $X[1] + mov (+$pDst_o+8*9)($pDst), $X[2] + mov (+$pDst_o+8*10)($pDst), $X[3] + mov (+$pDst_o+8*11)($pDst), $X[4] + mov (+$pDst_o+8*12)($pDst), $X[5] + mov (+$pDst_o+8*13)($pDst), $x6 + mov (+$pDst_o+8*14)($pDst), $x7 + + mov $X[7], %rax + mul %rax + mov %rax, $X[7] + mov %rdx, $A + + add $X[0], $X[0] + adc $X[1], $X[1] + adc $X[2], $X[2] + adc $X[3], $X[3] + adc $X[4], $X[4] + adc $X[5], $X[5] + adc $x6, $x6 + adc $x7, $x7 + adc \$0, $A + + add $tmp, $X[0] + + mov (+8*4)($pA), %rax + mul %rax + + add $X[6], $X[0] + adc %rax, $X[1] + adc \$0, %rdx + + mov %rdx, $tmp + + mov $X[0], (+$pDst_o+8*7)($pDst) + mov $X[1], (+$pDst_o+8*8)($pDst) + + mov (+8*5)($pA), %rax + mul %rax + + add $tmp, $X[2] + adc %rax, $X[3] + adc \$0, %rdx + + mov %rdx, $tmp + + mov $X[2], (+$pDst_o+8*9)($pDst) + mov $X[3], (+$pDst_o+8*10)($pDst) + + mov (+8*6)($pA), %rax + mul %rax + + add $tmp, $X[4] + adc %rax, $X[5] + adc \$0, %rdx + + mov $X[4], (+$pDst_o+8*11)($pDst) + mov $X[5], (+$pDst_o+8*12)($pDst) + + add %rdx, $x6 + adc $X[7], $x7 + adc \$0, $A + + mov $x6, (+$pDst_o+8*13)($pDst) + mov $x7, (+$pDst_o+8*14)($pDst) + mov $A, (+$pDst_o+8*15)($pDst) +___ +} + +# +# sqr_reduce: subroutine to compute Result = reduce(Result * Result) +# +# input and result also in: r9, r8, r15, r14, r13, r12, r11, r10 +# +$code.=<<___; +.type sqr_reduce,\@abi-omnipotent +.align 16 +sqr_reduce: + mov (+$pResult_offset+8)(%rsp), %rcx +___ + &SQR_512("%rsp+$tmp16_offset+8", "%rcx", [map("%r$_",(10..15,8..9))], "%rbx", "%rbp", "%rsi", "%rdi"); +$code.=<<___; + # tail recursion optimization: jmp to mont_reduce and return from there + jmp mont_reduce + # call mont_reduce + # ret +.size sqr_reduce,.-sqr_reduce +___ +}}} + +# +# MAIN FUNCTION +# + +#mod_exp_512(UINT64 *result, /* 512 bits, 8 qwords */ +# UINT64 *g, /* 512 bits, 8 qwords */ +# UINT64 *exp, /* 512 bits, 8 qwords */ +# struct mod_ctx_512 *data) + +# window size = 5 +# table size = 2^5 = 32 +#table_entries equ 32 +#table_size equ table_entries * 8 +$code.=<<___; +.globl mod_exp_512 +.type mod_exp_512,\@function,4 +mod_exp_512: + push %rbp + push %rbx + push %r12 + push %r13 + push %r14 + push %r15 + + # adjust stack down and then align it with cache boundary + mov %rsp, %r8 + sub \$$mem_size, %rsp + and \$-64, %rsp + + # store previous stack pointer and arguments + mov %r8, (+$rsp_offset)(%rsp) + mov %rdi, (+$pResult_offset)(%rsp) + mov %rsi, (+$pG_offset)(%rsp) + mov %rcx, (+$pData_offset)(%rsp) +.Lbody: + # transform g into montgomery space + # GT = reduce(g * C2) = reduce(g * (2^256)) + # reduce expects to have the input in [tmp16] + pxor %xmm4, %xmm4 + movdqu (+16*0)(%rsi), %xmm0 + movdqu (+16*1)(%rsi), %xmm1 + movdqu (+16*2)(%rsi), %xmm2 + movdqu (+16*3)(%rsi), %xmm3 + movdqa %xmm4, (+$tmp16_offset+16*0)(%rsp) + movdqa %xmm4, (+$tmp16_offset+16*1)(%rsp) + movdqa %xmm4, (+$tmp16_offset+16*6)(%rsp) + movdqa %xmm4, (+$tmp16_offset+16*7)(%rsp) + movdqa %xmm0, (+$tmp16_offset+16*2)(%rsp) + movdqa %xmm1, (+$tmp16_offset+16*3)(%rsp) + movdqa %xmm2, (+$tmp16_offset+16*4)(%rsp) + movdqa %xmm3, (+$tmp16_offset+16*5)(%rsp) + + # load pExp before rdx gets blown away + movdqu (+16*0)(%rdx), %xmm0 + movdqu (+16*1)(%rdx), %xmm1 + movdqu (+16*2)(%rdx), %xmm2 + movdqu (+16*3)(%rdx), %xmm3 + + lea (+$GT_offset)(%rsp), %rbx + mov %rbx, (+$red_result_addr_offset)(%rsp) + call mont_reduce + + # Initialize tmp = C + lea (+$tmp_offset)(%rsp), %rcx + xor %rax, %rax + mov %rax, (+8*0)(%rcx) + mov %rax, (+8*1)(%rcx) + mov %rax, (+8*3)(%rcx) + mov %rax, (+8*4)(%rcx) + mov %rax, (+8*5)(%rcx) + mov %rax, (+8*6)(%rcx) + mov %rax, (+8*7)(%rcx) + mov %rax, (+$exp_offset+8*8)(%rsp) + movq \$1, (+8*2)(%rcx) + + lea (+$garray_offset)(%rsp), %rbp + mov %rcx, %rsi # pTmp + mov %rbp, %rdi # Garray[][0] +___ + + &swizzle("%rdi", "%rcx", "%rax", "%rbx"); + + # for (rax = 31; rax != 0; rax--) { + # tmp = reduce(tmp * G) + # swizzle(pg, tmp); + # pg += 2; } +$code.=<<___; + mov \$31, %rax + mov %rax, (+$i_offset)(%rsp) + mov %rbp, (+$pg_offset)(%rsp) + # rsi -> pTmp + mov %rsi, (+$red_result_addr_offset)(%rsp) + mov (+8*0)(%rsi), %r10 + mov (+8*1)(%rsi), %r11 + mov (+8*2)(%rsi), %r12 + mov (+8*3)(%rsi), %r13 + mov (+8*4)(%rsi), %r14 + mov (+8*5)(%rsi), %r15 + mov (+8*6)(%rsi), %r8 + mov (+8*7)(%rsi), %r9 +init_loop: + lea (+$GT_offset)(%rsp), %rdi + call mont_mul_a3b + lea (+$tmp_offset)(%rsp), %rsi + mov (+$pg_offset)(%rsp), %rbp + add \$2, %rbp + mov %rbp, (+$pg_offset)(%rsp) + mov %rsi, %rcx # rcx = rsi = addr of tmp +___ + + &swizzle("%rbp", "%rcx", "%rax", "%rbx"); +$code.=<<___; + mov (+$i_offset)(%rsp), %rax + sub \$1, %rax + mov %rax, (+$i_offset)(%rsp) + jne init_loop + + # + # Copy exponent onto stack + movdqa %xmm0, (+$exp_offset+16*0)(%rsp) + movdqa %xmm1, (+$exp_offset+16*1)(%rsp) + movdqa %xmm2, (+$exp_offset+16*2)(%rsp) + movdqa %xmm3, (+$exp_offset+16*3)(%rsp) + + + # + # Do exponentiation + # Initialize result to G[exp{511:507}] + mov (+$exp_offset+62)(%rsp), %eax + mov %rax, %rdx + shr \$11, %rax + and \$0x07FF, %edx + mov %edx, (+$exp_offset+62)(%rsp) + lea (+$garray_offset)(%rsp,%rax,2), %rsi + mov (+$pResult_offset)(%rsp), %rdx +___ + + &unswizzle("%rdx", "%rsi", "%rbp", "%rbx", "%rax"); + + # + # Loop variables + # rcx = [loop_idx] = index: 510-5 to 0 by 5 +$code.=<<___; + movq \$505, (+$loop_idx_offset)(%rsp) + + mov (+$pResult_offset)(%rsp), %rcx + mov %rcx, (+$red_result_addr_offset)(%rsp) + mov (+8*0)(%rcx), %r10 + mov (+8*1)(%rcx), %r11 + mov (+8*2)(%rcx), %r12 + mov (+8*3)(%rcx), %r13 + mov (+8*4)(%rcx), %r14 + mov (+8*5)(%rcx), %r15 + mov (+8*6)(%rcx), %r8 + mov (+8*7)(%rcx), %r9 + jmp sqr_2 + +main_loop_a3b: + call sqr_reduce + call sqr_reduce + call sqr_reduce +sqr_2: + call sqr_reduce + call sqr_reduce + + # + # Do multiply, first look up proper value in Garray + mov (+$loop_idx_offset)(%rsp), %rcx # bit index + mov %rcx, %rax + shr \$4, %rax # rax is word pointer + mov (+$exp_offset)(%rsp,%rax,2), %edx + and \$15, %rcx + shrq %cl, %rdx + and \$0x1F, %rdx + + lea (+$garray_offset)(%rsp,%rdx,2), %rsi + lea (+$tmp_offset)(%rsp), %rdx + mov %rdx, %rdi +___ + + &unswizzle("%rdx", "%rsi", "%rbp", "%rbx", "%rax"); + # rdi = tmp = pG + + # + # Call mod_mul_a1(pDst, pSrc1, pSrc2, pM, pData) + # result result pG M Data +$code.=<<___; + mov (+$pResult_offset)(%rsp), %rsi + call mont_mul_a3b + + # + # finish loop + mov (+$loop_idx_offset)(%rsp), %rcx + sub \$5, %rcx + mov %rcx, (+$loop_idx_offset)(%rsp) + jge main_loop_a3b + + # + +end_main_loop_a3b: + # transform result out of Montgomery space + # result = reduce(result) + mov (+$pResult_offset)(%rsp), %rdx + pxor %xmm4, %xmm4 + movdqu (+16*0)(%rdx), %xmm0 + movdqu (+16*1)(%rdx), %xmm1 + movdqu (+16*2)(%rdx), %xmm2 + movdqu (+16*3)(%rdx), %xmm3 + movdqa %xmm4, (+$tmp16_offset+16*4)(%rsp) + movdqa %xmm4, (+$tmp16_offset+16*5)(%rsp) + movdqa %xmm4, (+$tmp16_offset+16*6)(%rsp) + movdqa %xmm4, (+$tmp16_offset+16*7)(%rsp) + movdqa %xmm0, (+$tmp16_offset+16*0)(%rsp) + movdqa %xmm1, (+$tmp16_offset+16*1)(%rsp) + movdqa %xmm2, (+$tmp16_offset+16*2)(%rsp) + movdqa %xmm3, (+$tmp16_offset+16*3)(%rsp) + call mont_reduce + + # If result > m, subract m + # load result into r15:r8 + mov (+$pResult_offset)(%rsp), %rax + mov (+8*0)(%rax), %r8 + mov (+8*1)(%rax), %r9 + mov (+8*2)(%rax), %r10 + mov (+8*3)(%rax), %r11 + mov (+8*4)(%rax), %r12 + mov (+8*5)(%rax), %r13 + mov (+8*6)(%rax), %r14 + mov (+8*7)(%rax), %r15 + + # subtract m + mov (+$pData_offset)(%rsp), %rbx + add \$$M, %rbx + + sub (+8*0)(%rbx), %r8 + sbb (+8*1)(%rbx), %r9 + sbb (+8*2)(%rbx), %r10 + sbb (+8*3)(%rbx), %r11 + sbb (+8*4)(%rbx), %r12 + sbb (+8*5)(%rbx), %r13 + sbb (+8*6)(%rbx), %r14 + sbb (+8*7)(%rbx), %r15 + + # if Carry is clear, replace result with difference + mov (+8*0)(%rax), %rsi + mov (+8*1)(%rax), %rdi + mov (+8*2)(%rax), %rcx + mov (+8*3)(%rax), %rdx + cmovnc %r8, %rsi + cmovnc %r9, %rdi + cmovnc %r10, %rcx + cmovnc %r11, %rdx + mov %rsi, (+8*0)(%rax) + mov %rdi, (+8*1)(%rax) + mov %rcx, (+8*2)(%rax) + mov %rdx, (+8*3)(%rax) + + mov (+8*4)(%rax), %rsi + mov (+8*5)(%rax), %rdi + mov (+8*6)(%rax), %rcx + mov (+8*7)(%rax), %rdx + cmovnc %r12, %rsi + cmovnc %r13, %rdi + cmovnc %r14, %rcx + cmovnc %r15, %rdx + mov %rsi, (+8*4)(%rax) + mov %rdi, (+8*5)(%rax) + mov %rcx, (+8*6)(%rax) + mov %rdx, (+8*7)(%rax) + + mov (+$rsp_offset)(%rsp), %rsi + mov 0(%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbx + mov 40(%rsi),%rbp + lea 48(%rsi),%rsp +.Lepilogue: + ret +.size mod_exp_512, . - mod_exp_512 +___ + +if ($win64) { +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +my $rec="%rcx"; +my $frame="%rdx"; +my $context="%r8"; +my $disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type mod_exp_512_se_handler,\@abi-omnipotent +.align 16 +mod_exp_512_se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + lea .Lbody(%rip),%r10 + cmp %r10,%rbx # context->RipRsp + + lea .Lepilogue(%rip),%r10 + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lin_prologue + + mov $rsp_offset(%rax),%rax # pull saved Rsp + + mov 32(%rax),%rbx + mov 40(%rax),%rbp + mov 24(%rax),%r12 + mov 16(%rax),%r13 + mov 8(%rax),%r14 + mov 0(%rax),%r15 + lea 48(%rax),%rax + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + +.Lin_prologue: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$154,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size mod_exp_512_se_handler,.-mod_exp_512_se_handler + +.section .pdata +.align 4 + .rva .LSEH_begin_mod_exp_512 + .rva .LSEH_end_mod_exp_512 + .rva .LSEH_info_mod_exp_512 + +.section .xdata +.align 8 +.LSEH_info_mod_exp_512: + .byte 9,0,0,0 + .rva mod_exp_512_se_handler +___ +} + +sub reg_part { +my ($reg,$conv)=@_; + if ($reg =~ /%r[0-9]+/) { $reg .= $conv; } + elsif ($conv eq "b") { $reg =~ s/%[er]([^x]+)x?/%$1l/; } + elsif ($conv eq "w") { $reg =~ s/%[er](.+)/%$1/; } + elsif ($conv eq "d") { $reg =~ s/%[er](.+)/%e$1/; } + return $reg; +} + +$code =~ s/(%[a-z0-9]+)#([bwd])/reg_part($1,$2)/gem; +$code =~ s/\`([^\`]*)\`/eval $1/gem; +$code =~ s/(\(\+[^)]+\))/eval $1/gem; +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/pa-risc2.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/pa-risc2.S new file mode 100644 index 00000000..f3b16290 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/pa-risc2.S @@ -0,0 +1,1618 @@ +; +; PA-RISC 2.0 implementation of bn_asm code, based on the +; 64-bit version of the code. This code is effectively the +; same as the 64-bit version except the register model is +; slightly different given all values must be 32-bit between +; function calls. Thus the 64-bit return values are returned +; in %ret0 and %ret1 vs just %ret0 as is done in 64-bit +; +; +; This code is approximately 2x faster than the C version +; for RSA/DSA. +; +; See http://devresource.hp.com/ for more details on the PA-RISC +; architecture. Also see the book "PA-RISC 2.0 Architecture" +; by Gerry Kane for information on the instruction set architecture. +; +; Code written by Chris Ruemmler (with some help from the HP C +; compiler). +; +; The code compiles with HP's assembler +; + + .level 2.0N + .space $TEXT$ + .subspa $CODE$,QUAD=0,ALIGN=8,ACCESS=0x2c,CODE_ONLY + +; +; Global Register definitions used for the routines. +; +; Some information about HP's runtime architecture for 32-bits. +; +; "Caller save" means the calling function must save the register +; if it wants the register to be preserved. +; "Callee save" means if a function uses the register, it must save +; the value before using it. +; +; For the floating point registers +; +; "caller save" registers: fr4-fr11, fr22-fr31 +; "callee save" registers: fr12-fr21 +; "special" registers: fr0-fr3 (status and exception registers) +; +; For the integer registers +; value zero : r0 +; "caller save" registers: r1,r19-r26 +; "callee save" registers: r3-r18 +; return register : r2 (rp) +; return values ; r28,r29 (ret0,ret1) +; Stack pointer ; r30 (sp) +; millicode return ptr ; r31 (also a caller save register) + + +; +; Arguments to the routines +; +r_ptr .reg %r26 +a_ptr .reg %r25 +b_ptr .reg %r24 +num .reg %r24 +n .reg %r23 + +; +; Note that the "w" argument for bn_mul_add_words and bn_mul_words +; is passed on the stack at a delta of -56 from the top of stack +; as the routine is entered. +; + +; +; Globals used in some routines +; + +top_overflow .reg %r23 +high_mask .reg %r22 ; value 0xffffffff80000000L + + +;------------------------------------------------------------------------------ +; +; bn_mul_add_words +; +;BN_ULONG bn_mul_add_words(BN_ULONG *r_ptr, BN_ULONG *a_ptr, +; int num, BN_ULONG w) +; +; arg0 = r_ptr +; arg1 = a_ptr +; arg3 = num +; -56(sp) = w +; +; Local register definitions +; + +fm1 .reg %fr22 +fm .reg %fr23 +ht_temp .reg %fr24 +ht_temp_1 .reg %fr25 +lt_temp .reg %fr26 +lt_temp_1 .reg %fr27 +fm1_1 .reg %fr28 +fm_1 .reg %fr29 + +fw_h .reg %fr7L +fw_l .reg %fr7R +fw .reg %fr7 + +fht_0 .reg %fr8L +flt_0 .reg %fr8R +t_float_0 .reg %fr8 + +fht_1 .reg %fr9L +flt_1 .reg %fr9R +t_float_1 .reg %fr9 + +tmp_0 .reg %r31 +tmp_1 .reg %r21 +m_0 .reg %r20 +m_1 .reg %r19 +ht_0 .reg %r1 +ht_1 .reg %r3 +lt_0 .reg %r4 +lt_1 .reg %r5 +m1_0 .reg %r6 +m1_1 .reg %r7 +rp_val .reg %r8 +rp_val_1 .reg %r9 + +bn_mul_add_words + .export bn_mul_add_words,entry,NO_RELOCATION,LONG_RETURN + .proc + .callinfo frame=128 + .entry + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + NOP ; Needed to make the loop 16-byte aligned + NOP ; needed to make the loop 16-byte aligned + + STD %r5,16(%sp) ; save r5 + NOP + STD %r6,24(%sp) ; save r6 + STD %r7,32(%sp) ; save r7 + + STD %r8,40(%sp) ; save r8 + STD %r9,48(%sp) ; save r9 + COPY %r0,%ret1 ; return 0 by default + DEPDI,Z 1,31,1,top_overflow ; top_overflow = 1 << 32 + + CMPIB,>= 0,num,bn_mul_add_words_exit ; if (num <= 0) then exit + LDO 128(%sp),%sp ; bump stack + + ; + ; The loop is unrolled twice, so if there is only 1 number + ; then go straight to the cleanup code. + ; + CMPIB,= 1,num,bn_mul_add_words_single_top + FLDD -184(%sp),fw ; (-56-128) load up w into fw (fw_h/fw_l) + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; + ; PA-RISC 2.0 chips have two fully pipelined multipliers, thus + ; two 32-bit mutiplies can be issued per cycle. + ; +bn_mul_add_words_unroll2 + + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + FLDD 8(a_ptr),t_float_1 ; load up 64-bit value (fr8L) ht(L)/lt(R) + LDD 0(r_ptr),rp_val ; rp[0] + LDD 8(r_ptr),rp_val_1 ; rp[1] + + XMPYU fht_0,fw_l,fm1 ; m1[0] = fht_0*fw_l + XMPYU fht_1,fw_l,fm1_1 ; m1[1] = fht_1*fw_l + FSTD fm1,-16(%sp) ; -16(sp) = m1[0] + FSTD fm1_1,-48(%sp) ; -48(sp) = m1[1] + + XMPYU flt_0,fw_h,fm ; m[0] = flt_0*fw_h + XMPYU flt_1,fw_h,fm_1 ; m[1] = flt_1*fw_h + FSTD fm,-8(%sp) ; -8(sp) = m[0] + FSTD fm_1,-40(%sp) ; -40(sp) = m[1] + + XMPYU fht_0,fw_h,ht_temp ; ht_temp = fht_0*fw_h + XMPYU fht_1,fw_h,ht_temp_1 ; ht_temp_1 = fht_1*fw_h + FSTD ht_temp,-24(%sp) ; -24(sp) = ht_temp + FSTD ht_temp_1,-56(%sp) ; -56(sp) = ht_temp_1 + + XMPYU flt_0,fw_l,lt_temp ; lt_temp = lt*fw_l + XMPYU flt_1,fw_l,lt_temp_1 ; lt_temp = lt*fw_l + FSTD lt_temp,-32(%sp) ; -32(sp) = lt_temp + FSTD lt_temp_1,-64(%sp) ; -64(sp) = lt_temp_1 + + LDD -8(%sp),m_0 ; m[0] + LDD -40(%sp),m_1 ; m[1] + LDD -16(%sp),m1_0 ; m1[0] + LDD -48(%sp),m1_1 ; m1[1] + + LDD -24(%sp),ht_0 ; ht[0] + LDD -56(%sp),ht_1 ; ht[1] + ADD,L m1_0,m_0,tmp_0 ; tmp_0 = m[0] + m1[0]; + ADD,L m1_1,m_1,tmp_1 ; tmp_1 = m[1] + m1[1]; + + LDD -32(%sp),lt_0 + LDD -64(%sp),lt_1 + CMPCLR,*>>= tmp_0,m1_0, %r0 ; if (m[0] < m1[0]) + ADD,L ht_0,top_overflow,ht_0 ; ht[0] += (1<<32) + + CMPCLR,*>>= tmp_1,m1_1,%r0 ; if (m[1] < m1[1]) + ADD,L ht_1,top_overflow,ht_1 ; ht[1] += (1<<32) + EXTRD,U tmp_0,31,32,m_0 ; m[0]>>32 + DEPD,Z tmp_0,31,32,m1_0 ; m1[0] = m[0]<<32 + + EXTRD,U tmp_1,31,32,m_1 ; m[1]>>32 + DEPD,Z tmp_1,31,32,m1_1 ; m1[1] = m[1]<<32 + ADD,L ht_0,m_0,ht_0 ; ht[0]+= (m[0]>>32) + ADD,L ht_1,m_1,ht_1 ; ht[1]+= (m[1]>>32) + + ADD lt_0,m1_0,lt_0 ; lt[0] = lt[0]+m1[0]; + ADD,DC ht_0,%r0,ht_0 ; ht[0]++ + ADD lt_1,m1_1,lt_1 ; lt[1] = lt[1]+m1[1]; + ADD,DC ht_1,%r0,ht_1 ; ht[1]++ + + ADD %ret1,lt_0,lt_0 ; lt[0] = lt[0] + c; + ADD,DC ht_0,%r0,ht_0 ; ht[0]++ + ADD lt_0,rp_val,lt_0 ; lt[0] = lt[0]+rp[0] + ADD,DC ht_0,%r0,ht_0 ; ht[0]++ + + LDO -2(num),num ; num = num - 2; + ADD ht_0,lt_1,lt_1 ; lt[1] = lt[1] + ht_0 (c); + ADD,DC ht_1,%r0,ht_1 ; ht[1]++ + STD lt_0,0(r_ptr) ; rp[0] = lt[0] + + ADD lt_1,rp_val_1,lt_1 ; lt[1] = lt[1]+rp[1] + ADD,DC ht_1,%r0,%ret1 ; ht[1]++ + LDO 16(a_ptr),a_ptr ; a_ptr += 2 + + STD lt_1,8(r_ptr) ; rp[1] = lt[1] + CMPIB,<= 2,num,bn_mul_add_words_unroll2 ; go again if more to do + LDO 16(r_ptr),r_ptr ; r_ptr += 2 + + CMPIB,=,N 0,num,bn_mul_add_words_exit ; are we done, or cleanup last one + + ; + ; Top of loop aligned on 64-byte boundary + ; +bn_mul_add_words_single_top + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + LDD 0(r_ptr),rp_val ; rp[0] + LDO 8(a_ptr),a_ptr ; a_ptr++ + XMPYU fht_0,fw_l,fm1 ; m1 = ht*fw_l + FSTD fm1,-16(%sp) ; -16(sp) = m1 + XMPYU flt_0,fw_h,fm ; m = lt*fw_h + FSTD fm,-8(%sp) ; -8(sp) = m + XMPYU fht_0,fw_h,ht_temp ; ht_temp = ht*fw_h + FSTD ht_temp,-24(%sp) ; -24(sp) = ht + XMPYU flt_0,fw_l,lt_temp ; lt_temp = lt*fw_l + FSTD lt_temp,-32(%sp) ; -32(sp) = lt + + LDD -8(%sp),m_0 + LDD -16(%sp),m1_0 ; m1 = temp1 + ADD,L m_0,m1_0,tmp_0 ; tmp_0 = m + m1; + LDD -24(%sp),ht_0 + LDD -32(%sp),lt_0 + + CMPCLR,*>>= tmp_0,m1_0,%r0 ; if (m < m1) + ADD,L ht_0,top_overflow,ht_0 ; ht += (1<<32) + + EXTRD,U tmp_0,31,32,m_0 ; m>>32 + DEPD,Z tmp_0,31,32,m1_0 ; m1 = m<<32 + + ADD,L ht_0,m_0,ht_0 ; ht+= (m>>32) + ADD lt_0,m1_0,tmp_0 ; tmp_0 = lt+m1; + ADD,DC ht_0,%r0,ht_0 ; ht++ + ADD %ret1,tmp_0,lt_0 ; lt = lt + c; + ADD,DC ht_0,%r0,ht_0 ; ht++ + ADD lt_0,rp_val,lt_0 ; lt = lt+rp[0] + ADD,DC ht_0,%r0,%ret1 ; ht++ + STD lt_0,0(r_ptr) ; rp[0] = lt + +bn_mul_add_words_exit + .EXIT + + EXTRD,U %ret1,31,32,%ret0 ; for 32-bit, return in ret0/ret1 + LDD -80(%sp),%r9 ; restore r9 + LDD -88(%sp),%r8 ; restore r8 + LDD -96(%sp),%r7 ; restore r7 + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 ; restore r3 + .PROCEND ;in=23,24,25,26,29;out=28; + +;---------------------------------------------------------------------------- +; +;BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w) +; +; arg0 = rp +; arg1 = ap +; arg3 = num +; w on stack at -56(sp) + +bn_mul_words + .proc + .callinfo frame=128 + .entry + .EXPORT bn_mul_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + NOP + STD %r5,16(%sp) ; save r5 + + STD %r6,24(%sp) ; save r6 + STD %r7,32(%sp) ; save r7 + COPY %r0,%ret1 ; return 0 by default + DEPDI,Z 1,31,1,top_overflow ; top_overflow = 1 << 32 + + CMPIB,>= 0,num,bn_mul_words_exit + LDO 128(%sp),%sp ; bump stack + + ; + ; See if only 1 word to do, thus just do cleanup + ; + CMPIB,= 1,num,bn_mul_words_single_top + FLDD -184(%sp),fw ; (-56-128) load up w into fw (fw_h/fw_l) + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; + ; PA-RISC 2.0 chips have two fully pipelined multipliers, thus + ; two 32-bit mutiplies can be issued per cycle. + ; +bn_mul_words_unroll2 + + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + FLDD 8(a_ptr),t_float_1 ; load up 64-bit value (fr8L) ht(L)/lt(R) + XMPYU fht_0,fw_l,fm1 ; m1[0] = fht_0*fw_l + XMPYU fht_1,fw_l,fm1_1 ; m1[1] = ht*fw_l + + FSTD fm1,-16(%sp) ; -16(sp) = m1 + FSTD fm1_1,-48(%sp) ; -48(sp) = m1 + XMPYU flt_0,fw_h,fm ; m = lt*fw_h + XMPYU flt_1,fw_h,fm_1 ; m = lt*fw_h + + FSTD fm,-8(%sp) ; -8(sp) = m + FSTD fm_1,-40(%sp) ; -40(sp) = m + XMPYU fht_0,fw_h,ht_temp ; ht_temp = fht_0*fw_h + XMPYU fht_1,fw_h,ht_temp_1 ; ht_temp = ht*fw_h + + FSTD ht_temp,-24(%sp) ; -24(sp) = ht + FSTD ht_temp_1,-56(%sp) ; -56(sp) = ht + XMPYU flt_0,fw_l,lt_temp ; lt_temp = lt*fw_l + XMPYU flt_1,fw_l,lt_temp_1 ; lt_temp = lt*fw_l + + FSTD lt_temp,-32(%sp) ; -32(sp) = lt + FSTD lt_temp_1,-64(%sp) ; -64(sp) = lt + LDD -8(%sp),m_0 + LDD -40(%sp),m_1 + + LDD -16(%sp),m1_0 + LDD -48(%sp),m1_1 + LDD -24(%sp),ht_0 + LDD -56(%sp),ht_1 + + ADD,L m1_0,m_0,tmp_0 ; tmp_0 = m + m1; + ADD,L m1_1,m_1,tmp_1 ; tmp_1 = m + m1; + LDD -32(%sp),lt_0 + LDD -64(%sp),lt_1 + + CMPCLR,*>>= tmp_0,m1_0, %r0 ; if (m < m1) + ADD,L ht_0,top_overflow,ht_0 ; ht += (1<<32) + CMPCLR,*>>= tmp_1,m1_1,%r0 ; if (m < m1) + ADD,L ht_1,top_overflow,ht_1 ; ht += (1<<32) + + EXTRD,U tmp_0,31,32,m_0 ; m>>32 + DEPD,Z tmp_0,31,32,m1_0 ; m1 = m<<32 + EXTRD,U tmp_1,31,32,m_1 ; m>>32 + DEPD,Z tmp_1,31,32,m1_1 ; m1 = m<<32 + + ADD,L ht_0,m_0,ht_0 ; ht+= (m>>32) + ADD,L ht_1,m_1,ht_1 ; ht+= (m>>32) + ADD lt_0,m1_0,lt_0 ; lt = lt+m1; + ADD,DC ht_0,%r0,ht_0 ; ht++ + + ADD lt_1,m1_1,lt_1 ; lt = lt+m1; + ADD,DC ht_1,%r0,ht_1 ; ht++ + ADD %ret1,lt_0,lt_0 ; lt = lt + c (ret1); + ADD,DC ht_0,%r0,ht_0 ; ht++ + + ADD ht_0,lt_1,lt_1 ; lt = lt + c (ht_0) + ADD,DC ht_1,%r0,ht_1 ; ht++ + STD lt_0,0(r_ptr) ; rp[0] = lt + STD lt_1,8(r_ptr) ; rp[1] = lt + + COPY ht_1,%ret1 ; carry = ht + LDO -2(num),num ; num = num - 2; + LDO 16(a_ptr),a_ptr ; ap += 2 + CMPIB,<= 2,num,bn_mul_words_unroll2 + LDO 16(r_ptr),r_ptr ; rp++ + + CMPIB,=,N 0,num,bn_mul_words_exit ; are we done? + + ; + ; Top of loop aligned on 64-byte boundary + ; +bn_mul_words_single_top + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + + XMPYU fht_0,fw_l,fm1 ; m1 = ht*fw_l + FSTD fm1,-16(%sp) ; -16(sp) = m1 + XMPYU flt_0,fw_h,fm ; m = lt*fw_h + FSTD fm,-8(%sp) ; -8(sp) = m + XMPYU fht_0,fw_h,ht_temp ; ht_temp = ht*fw_h + FSTD ht_temp,-24(%sp) ; -24(sp) = ht + XMPYU flt_0,fw_l,lt_temp ; lt_temp = lt*fw_l + FSTD lt_temp,-32(%sp) ; -32(sp) = lt + + LDD -8(%sp),m_0 + LDD -16(%sp),m1_0 + ADD,L m_0,m1_0,tmp_0 ; tmp_0 = m + m1; + LDD -24(%sp),ht_0 + LDD -32(%sp),lt_0 + + CMPCLR,*>>= tmp_0,m1_0,%r0 ; if (m < m1) + ADD,L ht_0,top_overflow,ht_0 ; ht += (1<<32) + + EXTRD,U tmp_0,31,32,m_0 ; m>>32 + DEPD,Z tmp_0,31,32,m1_0 ; m1 = m<<32 + + ADD,L ht_0,m_0,ht_0 ; ht+= (m>>32) + ADD lt_0,m1_0,lt_0 ; lt= lt+m1; + ADD,DC ht_0,%r0,ht_0 ; ht++ + + ADD %ret1,lt_0,lt_0 ; lt = lt + c; + ADD,DC ht_0,%r0,ht_0 ; ht++ + + COPY ht_0,%ret1 ; copy carry + STD lt_0,0(r_ptr) ; rp[0] = lt + +bn_mul_words_exit + .EXIT + EXTRD,U %ret1,31,32,%ret0 ; for 32-bit, return in ret0/ret1 + LDD -96(%sp),%r7 ; restore r7 + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 ; restore r3 + .PROCEND + +;---------------------------------------------------------------------------- +; +;void bn_sqr_words(BN_ULONG *rp, BN_ULONG *ap, int num) +; +; arg0 = rp +; arg1 = ap +; arg2 = num +; + +bn_sqr_words + .proc + .callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_sqr_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + NOP + STD %r5,16(%sp) ; save r5 + + CMPIB,>= 0,num,bn_sqr_words_exit + LDO 128(%sp),%sp ; bump stack + + ; + ; If only 1, the goto straight to cleanup + ; + CMPIB,= 1,num,bn_sqr_words_single_top + DEPDI,Z -1,32,33,high_mask ; Create Mask 0xffffffff80000000L + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; + +bn_sqr_words_unroll2 + FLDD 0(a_ptr),t_float_0 ; a[0] + FLDD 8(a_ptr),t_float_1 ; a[1] + XMPYU fht_0,flt_0,fm ; m[0] + XMPYU fht_1,flt_1,fm_1 ; m[1] + + FSTD fm,-24(%sp) ; store m[0] + FSTD fm_1,-56(%sp) ; store m[1] + XMPYU flt_0,flt_0,lt_temp ; lt[0] + XMPYU flt_1,flt_1,lt_temp_1 ; lt[1] + + FSTD lt_temp,-16(%sp) ; store lt[0] + FSTD lt_temp_1,-48(%sp) ; store lt[1] + XMPYU fht_0,fht_0,ht_temp ; ht[0] + XMPYU fht_1,fht_1,ht_temp_1 ; ht[1] + + FSTD ht_temp,-8(%sp) ; store ht[0] + FSTD ht_temp_1,-40(%sp) ; store ht[1] + LDD -24(%sp),m_0 + LDD -56(%sp),m_1 + + AND m_0,high_mask,tmp_0 ; m[0] & Mask + AND m_1,high_mask,tmp_1 ; m[1] & Mask + DEPD,Z m_0,30,31,m_0 ; m[0] << 32+1 + DEPD,Z m_1,30,31,m_1 ; m[1] << 32+1 + + LDD -16(%sp),lt_0 + LDD -48(%sp),lt_1 + EXTRD,U tmp_0,32,33,tmp_0 ; tmp_0 = m[0]&Mask >> 32-1 + EXTRD,U tmp_1,32,33,tmp_1 ; tmp_1 = m[1]&Mask >> 32-1 + + LDD -8(%sp),ht_0 + LDD -40(%sp),ht_1 + ADD,L ht_0,tmp_0,ht_0 ; ht[0] += tmp_0 + ADD,L ht_1,tmp_1,ht_1 ; ht[1] += tmp_1 + + ADD lt_0,m_0,lt_0 ; lt = lt+m + ADD,DC ht_0,%r0,ht_0 ; ht[0]++ + STD lt_0,0(r_ptr) ; rp[0] = lt[0] + STD ht_0,8(r_ptr) ; rp[1] = ht[1] + + ADD lt_1,m_1,lt_1 ; lt = lt+m + ADD,DC ht_1,%r0,ht_1 ; ht[1]++ + STD lt_1,16(r_ptr) ; rp[2] = lt[1] + STD ht_1,24(r_ptr) ; rp[3] = ht[1] + + LDO -2(num),num ; num = num - 2; + LDO 16(a_ptr),a_ptr ; ap += 2 + CMPIB,<= 2,num,bn_sqr_words_unroll2 + LDO 32(r_ptr),r_ptr ; rp += 4 + + CMPIB,=,N 0,num,bn_sqr_words_exit ; are we done? + + ; + ; Top of loop aligned on 64-byte boundary + ; +bn_sqr_words_single_top + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + + XMPYU fht_0,flt_0,fm ; m + FSTD fm,-24(%sp) ; store m + + XMPYU flt_0,flt_0,lt_temp ; lt + FSTD lt_temp,-16(%sp) ; store lt + + XMPYU fht_0,fht_0,ht_temp ; ht + FSTD ht_temp,-8(%sp) ; store ht + + LDD -24(%sp),m_0 ; load m + AND m_0,high_mask,tmp_0 ; m & Mask + DEPD,Z m_0,30,31,m_0 ; m << 32+1 + LDD -16(%sp),lt_0 ; lt + + LDD -8(%sp),ht_0 ; ht + EXTRD,U tmp_0,32,33,tmp_0 ; tmp_0 = m&Mask >> 32-1 + ADD m_0,lt_0,lt_0 ; lt = lt+m + ADD,L ht_0,tmp_0,ht_0 ; ht += tmp_0 + ADD,DC ht_0,%r0,ht_0 ; ht++ + + STD lt_0,0(r_ptr) ; rp[0] = lt + STD ht_0,8(r_ptr) ; rp[1] = ht + +bn_sqr_words_exit + .EXIT + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + .PROCEND ;in=23,24,25,26,29;out=28; + + +;---------------------------------------------------------------------------- +; +;BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) +; +; arg0 = rp +; arg1 = ap +; arg2 = bp +; arg3 = n + +t .reg %r22 +b .reg %r21 +l .reg %r20 + +bn_add_words + .proc + .entry + .callinfo + .EXPORT bn_add_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .align 64 + + CMPIB,>= 0,n,bn_add_words_exit + COPY %r0,%ret1 ; return 0 by default + + ; + ; If 2 or more numbers do the loop + ; + CMPIB,= 1,n,bn_add_words_single_top + NOP + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; +bn_add_words_unroll2 + LDD 0(a_ptr),t + LDD 0(b_ptr),b + ADD t,%ret1,t ; t = t+c; + ADD,DC %r0,%r0,%ret1 ; set c to carry + ADD t,b,l ; l = t + b[0] + ADD,DC %ret1,%r0,%ret1 ; c+= carry + STD l,0(r_ptr) + + LDD 8(a_ptr),t + LDD 8(b_ptr),b + ADD t,%ret1,t ; t = t+c; + ADD,DC %r0,%r0,%ret1 ; set c to carry + ADD t,b,l ; l = t + b[0] + ADD,DC %ret1,%r0,%ret1 ; c+= carry + STD l,8(r_ptr) + + LDO -2(n),n + LDO 16(a_ptr),a_ptr + LDO 16(b_ptr),b_ptr + + CMPIB,<= 2,n,bn_add_words_unroll2 + LDO 16(r_ptr),r_ptr + + CMPIB,=,N 0,n,bn_add_words_exit ; are we done? + +bn_add_words_single_top + LDD 0(a_ptr),t + LDD 0(b_ptr),b + + ADD t,%ret1,t ; t = t+c; + ADD,DC %r0,%r0,%ret1 ; set c to carry (could use CMPCLR??) + ADD t,b,l ; l = t + b[0] + ADD,DC %ret1,%r0,%ret1 ; c+= carry + STD l,0(r_ptr) + +bn_add_words_exit + .EXIT + BVE (%rp) + EXTRD,U %ret1,31,32,%ret0 ; for 32-bit, return in ret0/ret1 + .PROCEND ;in=23,24,25,26,29;out=28; + +;---------------------------------------------------------------------------- +; +;BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) +; +; arg0 = rp +; arg1 = ap +; arg2 = bp +; arg3 = n + +t1 .reg %r22 +t2 .reg %r21 +sub_tmp1 .reg %r20 +sub_tmp2 .reg %r19 + + +bn_sub_words + .proc + .callinfo + .EXPORT bn_sub_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + + CMPIB,>= 0,n,bn_sub_words_exit + COPY %r0,%ret1 ; return 0 by default + + ; + ; If 2 or more numbers do the loop + ; + CMPIB,= 1,n,bn_sub_words_single_top + NOP + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; +bn_sub_words_unroll2 + LDD 0(a_ptr),t1 + LDD 0(b_ptr),t2 + SUB t1,t2,sub_tmp1 ; t3 = t1-t2; + SUB sub_tmp1,%ret1,sub_tmp1 ; t3 = t3- c; + + CMPCLR,*>> t1,t2,sub_tmp2 ; clear if t1 > t2 + LDO 1(%r0),sub_tmp2 + + CMPCLR,*= t1,t2,%r0 + COPY sub_tmp2,%ret1 + STD sub_tmp1,0(r_ptr) + + LDD 8(a_ptr),t1 + LDD 8(b_ptr),t2 + SUB t1,t2,sub_tmp1 ; t3 = t1-t2; + SUB sub_tmp1,%ret1,sub_tmp1 ; t3 = t3- c; + CMPCLR,*>> t1,t2,sub_tmp2 ; clear if t1 > t2 + LDO 1(%r0),sub_tmp2 + + CMPCLR,*= t1,t2,%r0 + COPY sub_tmp2,%ret1 + STD sub_tmp1,8(r_ptr) + + LDO -2(n),n + LDO 16(a_ptr),a_ptr + LDO 16(b_ptr),b_ptr + + CMPIB,<= 2,n,bn_sub_words_unroll2 + LDO 16(r_ptr),r_ptr + + CMPIB,=,N 0,n,bn_sub_words_exit ; are we done? + +bn_sub_words_single_top + LDD 0(a_ptr),t1 + LDD 0(b_ptr),t2 + SUB t1,t2,sub_tmp1 ; t3 = t1-t2; + SUB sub_tmp1,%ret1,sub_tmp1 ; t3 = t3- c; + CMPCLR,*>> t1,t2,sub_tmp2 ; clear if t1 > t2 + LDO 1(%r0),sub_tmp2 + + CMPCLR,*= t1,t2,%r0 + COPY sub_tmp2,%ret1 + + STD sub_tmp1,0(r_ptr) + +bn_sub_words_exit + .EXIT + BVE (%rp) + EXTRD,U %ret1,31,32,%ret0 ; for 32-bit, return in ret0/ret1 + .PROCEND ;in=23,24,25,26,29;out=28; + +;------------------------------------------------------------------------------ +; +; unsigned long bn_div_words(unsigned long h, unsigned long l, unsigned long d) +; +; arg0 = h +; arg1 = l +; arg2 = d +; +; This is mainly just output from the HP C compiler. +; +;------------------------------------------------------------------------------ +bn_div_words + .PROC + .EXPORT bn_div_words,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR,RTNVAL=GR,LONG_RETURN + .IMPORT BN_num_bits_word,CODE + ;--- not PIC .IMPORT __iob,DATA + ;--- not PIC .IMPORT fprintf,CODE + .IMPORT abort,CODE + .IMPORT $$div2U,MILLICODE + .CALLINFO CALLER,FRAME=144,ENTRY_GR=%r9,SAVE_RP,ARGS_SAVED,ORDERING_AWARE + .ENTRY + STW %r2,-20(%r30) ;offset 0x8ec + STW,MA %r3,192(%r30) ;offset 0x8f0 + STW %r4,-188(%r30) ;offset 0x8f4 + DEPD %r5,31,32,%r6 ;offset 0x8f8 + STD %r6,-184(%r30) ;offset 0x8fc + DEPD %r7,31,32,%r8 ;offset 0x900 + STD %r8,-176(%r30) ;offset 0x904 + STW %r9,-168(%r30) ;offset 0x908 + LDD -248(%r30),%r3 ;offset 0x90c + COPY %r26,%r4 ;offset 0x910 + COPY %r24,%r5 ;offset 0x914 + DEPD %r25,31,32,%r4 ;offset 0x918 + CMPB,*<> %r3,%r0,$0006000C ;offset 0x91c + DEPD %r23,31,32,%r5 ;offset 0x920 + MOVIB,TR -1,%r29,$00060002 ;offset 0x924 + EXTRD,U %r29,31,32,%r28 ;offset 0x928 +$0006002A + LDO -1(%r29),%r29 ;offset 0x92c + SUB %r23,%r7,%r23 ;offset 0x930 +$00060024 + SUB %r4,%r31,%r25 ;offset 0x934 + AND %r25,%r19,%r26 ;offset 0x938 + CMPB,*<>,N %r0,%r26,$00060046 ;offset 0x93c + DEPD,Z %r25,31,32,%r20 ;offset 0x940 + OR %r20,%r24,%r21 ;offset 0x944 + CMPB,*<<,N %r21,%r23,$0006002A ;offset 0x948 + SUB %r31,%r2,%r31 ;offset 0x94c +$00060046 +$0006002E + DEPD,Z %r23,31,32,%r25 ;offset 0x950 + EXTRD,U %r23,31,32,%r26 ;offset 0x954 + AND %r25,%r19,%r24 ;offset 0x958 + ADD,L %r31,%r26,%r31 ;offset 0x95c + CMPCLR,*>>= %r5,%r24,%r0 ;offset 0x960 + LDO 1(%r31),%r31 ;offset 0x964 +$00060032 + CMPB,*<<=,N %r31,%r4,$00060036 ;offset 0x968 + LDO -1(%r29),%r29 ;offset 0x96c + ADD,L %r4,%r3,%r4 ;offset 0x970 +$00060036 + ADDIB,=,N -1,%r8,$D0 ;offset 0x974 + SUB %r5,%r24,%r28 ;offset 0x978 +$0006003A + SUB %r4,%r31,%r24 ;offset 0x97c + SHRPD %r24,%r28,32,%r4 ;offset 0x980 + DEPD,Z %r29,31,32,%r9 ;offset 0x984 + DEPD,Z %r28,31,32,%r5 ;offset 0x988 +$0006001C + EXTRD,U %r4,31,32,%r31 ;offset 0x98c + CMPB,*<>,N %r31,%r2,$00060020 ;offset 0x990 + MOVB,TR %r6,%r29,$D1 ;offset 0x994 + STD %r29,-152(%r30) ;offset 0x998 +$0006000C + EXTRD,U %r3,31,32,%r25 ;offset 0x99c + COPY %r3,%r26 ;offset 0x9a0 + EXTRD,U %r3,31,32,%r9 ;offset 0x9a4 + EXTRD,U %r4,31,32,%r8 ;offset 0x9a8 + .CALL ARGW0=GR,ARGW1=GR,RTNVAL=GR ;in=25,26;out=28; + B,L BN_num_bits_word,%r2 ;offset 0x9ac + EXTRD,U %r5,31,32,%r7 ;offset 0x9b0 + LDI 64,%r20 ;offset 0x9b4 + DEPD %r7,31,32,%r5 ;offset 0x9b8 + DEPD %r8,31,32,%r4 ;offset 0x9bc + DEPD %r9,31,32,%r3 ;offset 0x9c0 + CMPB,= %r28,%r20,$00060012 ;offset 0x9c4 + COPY %r28,%r24 ;offset 0x9c8 + MTSARCM %r24 ;offset 0x9cc + DEPDI,Z -1,%sar,1,%r19 ;offset 0x9d0 + CMPB,*>>,N %r4,%r19,$D2 ;offset 0x9d4 +$00060012 + SUBI 64,%r24,%r31 ;offset 0x9d8 + CMPCLR,*<< %r4,%r3,%r0 ;offset 0x9dc + SUB %r4,%r3,%r4 ;offset 0x9e0 +$00060016 + CMPB,= %r31,%r0,$0006001A ;offset 0x9e4 + COPY %r0,%r9 ;offset 0x9e8 + MTSARCM %r31 ;offset 0x9ec + DEPD,Z %r3,%sar,64,%r3 ;offset 0x9f0 + SUBI 64,%r31,%r26 ;offset 0x9f4 + MTSAR %r26 ;offset 0x9f8 + SHRPD %r4,%r5,%sar,%r4 ;offset 0x9fc + MTSARCM %r31 ;offset 0xa00 + DEPD,Z %r5,%sar,64,%r5 ;offset 0xa04 +$0006001A + DEPDI,Z -1,31,32,%r19 ;offset 0xa08 + AND %r3,%r19,%r29 ;offset 0xa0c + EXTRD,U %r29,31,32,%r2 ;offset 0xa10 + DEPDI,Z -1,63,32,%r6 ;offset 0xa14 + MOVIB,TR 2,%r8,$0006001C ;offset 0xa18 + EXTRD,U %r3,63,32,%r7 ;offset 0xa1c +$D2 + ;--- not PIC ADDIL LR'__iob-$global$,%r27,%r1 ;offset 0xa20 + ;--- not PIC LDIL LR'C$7,%r21 ;offset 0xa24 + ;--- not PIC LDO RR'__iob-$global$+32(%r1),%r26 ;offset 0xa28 + ;--- not PIC .CALL ARGW0=GR,ARGW1=GR,ARGW2=GR,RTNVAL=GR ;in=24,25,26;out=28; + ;--- not PIC B,L fprintf,%r2 ;offset 0xa2c + ;--- not PIC LDO RR'C$7(%r21),%r25 ;offset 0xa30 + .CALL ; + B,L abort,%r2 ;offset 0xa34 + NOP ;offset 0xa38 + B $D3 ;offset 0xa3c + LDW -212(%r30),%r2 ;offset 0xa40 +$00060020 + COPY %r4,%r26 ;offset 0xa44 + EXTRD,U %r4,31,32,%r25 ;offset 0xa48 + COPY %r2,%r24 ;offset 0xa4c + .CALL ;in=23,24,25,26;out=20,21,22,28,29; (MILLICALL) + B,L $$div2U,%r31 ;offset 0xa50 + EXTRD,U %r2,31,32,%r23 ;offset 0xa54 + DEPD %r28,31,32,%r29 ;offset 0xa58 +$00060022 + STD %r29,-152(%r30) ;offset 0xa5c +$D1 + AND %r5,%r19,%r24 ;offset 0xa60 + EXTRD,U %r24,31,32,%r24 ;offset 0xa64 + STW %r2,-160(%r30) ;offset 0xa68 + STW %r7,-128(%r30) ;offset 0xa6c + FLDD -152(%r30),%fr4 ;offset 0xa70 + FLDD -152(%r30),%fr7 ;offset 0xa74 + FLDW -160(%r30),%fr8L ;offset 0xa78 + FLDW -128(%r30),%fr5L ;offset 0xa7c + XMPYU %fr8L,%fr7L,%fr10 ;offset 0xa80 + FSTD %fr10,-136(%r30) ;offset 0xa84 + XMPYU %fr8L,%fr7R,%fr22 ;offset 0xa88 + FSTD %fr22,-144(%r30) ;offset 0xa8c + XMPYU %fr5L,%fr4L,%fr11 ;offset 0xa90 + XMPYU %fr5L,%fr4R,%fr23 ;offset 0xa94 + FSTD %fr11,-112(%r30) ;offset 0xa98 + FSTD %fr23,-120(%r30) ;offset 0xa9c + LDD -136(%r30),%r28 ;offset 0xaa0 + DEPD,Z %r28,31,32,%r31 ;offset 0xaa4 + LDD -144(%r30),%r20 ;offset 0xaa8 + ADD,L %r20,%r31,%r31 ;offset 0xaac + LDD -112(%r30),%r22 ;offset 0xab0 + DEPD,Z %r22,31,32,%r22 ;offset 0xab4 + LDD -120(%r30),%r21 ;offset 0xab8 + B $00060024 ;offset 0xabc + ADD,L %r21,%r22,%r23 ;offset 0xac0 +$D0 + OR %r9,%r29,%r29 ;offset 0xac4 +$00060040 + EXTRD,U %r29,31,32,%r28 ;offset 0xac8 +$00060002 +$L2 + LDW -212(%r30),%r2 ;offset 0xacc +$D3 + LDW -168(%r30),%r9 ;offset 0xad0 + LDD -176(%r30),%r8 ;offset 0xad4 + EXTRD,U %r8,31,32,%r7 ;offset 0xad8 + LDD -184(%r30),%r6 ;offset 0xadc + EXTRD,U %r6,31,32,%r5 ;offset 0xae0 + LDW -188(%r30),%r4 ;offset 0xae4 + BVE (%r2) ;offset 0xae8 + .EXIT + LDW,MB -192(%r30),%r3 ;offset 0xaec + .PROCEND ;in=23,25;out=28,29;fpin=105,107; + + + + +;---------------------------------------------------------------------------- +; +; Registers to hold 64-bit values to manipulate. The "L" part +; of the register corresponds to the upper 32-bits, while the "R" +; part corresponds to the lower 32-bits +; +; Note, that when using b6 and b7, the code must save these before +; using them because they are callee save registers +; +; +; Floating point registers to use to save values that +; are manipulated. These don't collide with ftemp1-6 and +; are all caller save registers +; +a0 .reg %fr22 +a0L .reg %fr22L +a0R .reg %fr22R + +a1 .reg %fr23 +a1L .reg %fr23L +a1R .reg %fr23R + +a2 .reg %fr24 +a2L .reg %fr24L +a2R .reg %fr24R + +a3 .reg %fr25 +a3L .reg %fr25L +a3R .reg %fr25R + +a4 .reg %fr26 +a4L .reg %fr26L +a4R .reg %fr26R + +a5 .reg %fr27 +a5L .reg %fr27L +a5R .reg %fr27R + +a6 .reg %fr28 +a6L .reg %fr28L +a6R .reg %fr28R + +a7 .reg %fr29 +a7L .reg %fr29L +a7R .reg %fr29R + +b0 .reg %fr30 +b0L .reg %fr30L +b0R .reg %fr30R + +b1 .reg %fr31 +b1L .reg %fr31L +b1R .reg %fr31R + +; +; Temporary floating point variables, these are all caller save +; registers +; +ftemp1 .reg %fr4 +ftemp2 .reg %fr5 +ftemp3 .reg %fr6 +ftemp4 .reg %fr7 + +; +; The B set of registers when used. +; + +b2 .reg %fr8 +b2L .reg %fr8L +b2R .reg %fr8R + +b3 .reg %fr9 +b3L .reg %fr9L +b3R .reg %fr9R + +b4 .reg %fr10 +b4L .reg %fr10L +b4R .reg %fr10R + +b5 .reg %fr11 +b5L .reg %fr11L +b5R .reg %fr11R + +b6 .reg %fr12 +b6L .reg %fr12L +b6R .reg %fr12R + +b7 .reg %fr13 +b7L .reg %fr13L +b7R .reg %fr13R + +c1 .reg %r21 ; only reg +temp1 .reg %r20 ; only reg +temp2 .reg %r19 ; only reg +temp3 .reg %r31 ; only reg + +m1 .reg %r28 +c2 .reg %r23 +high_one .reg %r1 +ht .reg %r6 +lt .reg %r5 +m .reg %r4 +c3 .reg %r3 + +SQR_ADD_C .macro A0L,A0R,C1,C2,C3 + XMPYU A0L,A0R,ftemp1 ; m + FSTD ftemp1,-24(%sp) ; store m + + XMPYU A0R,A0R,ftemp2 ; lt + FSTD ftemp2,-16(%sp) ; store lt + + XMPYU A0L,A0L,ftemp3 ; ht + FSTD ftemp3,-8(%sp) ; store ht + + LDD -24(%sp),m ; load m + AND m,high_mask,temp2 ; m & Mask + DEPD,Z m,30,31,temp3 ; m << 32+1 + LDD -16(%sp),lt ; lt + + LDD -8(%sp),ht ; ht + EXTRD,U temp2,32,33,temp1 ; temp1 = m&Mask >> 32-1 + ADD temp3,lt,lt ; lt = lt+m + ADD,L ht,temp1,ht ; ht += temp1 + ADD,DC ht,%r0,ht ; ht++ + + ADD C1,lt,C1 ; c1=c1+lt + ADD,DC ht,%r0,ht ; ht++ + + ADD C2,ht,C2 ; c2=c2+ht + ADD,DC C3,%r0,C3 ; c3++ +.endm + +SQR_ADD_C2 .macro A0L,A0R,A1L,A1R,C1,C2,C3 + XMPYU A0L,A1R,ftemp1 ; m1 = bl*ht + FSTD ftemp1,-16(%sp) ; + XMPYU A0R,A1L,ftemp2 ; m = bh*lt + FSTD ftemp2,-8(%sp) ; + XMPYU A0R,A1R,ftemp3 ; lt = bl*lt + FSTD ftemp3,-32(%sp) + XMPYU A0L,A1L,ftemp4 ; ht = bh*ht + FSTD ftemp4,-24(%sp) ; + + LDD -8(%sp),m ; r21 = m + LDD -16(%sp),m1 ; r19 = m1 + ADD,L m,m1,m ; m+m1 + + DEPD,Z m,31,32,temp3 ; (m+m1<<32) + LDD -24(%sp),ht ; r24 = ht + + CMPCLR,*>>= m,m1,%r0 ; if (m < m1) + ADD,L ht,high_one,ht ; ht+=high_one + + EXTRD,U m,31,32,temp1 ; m >> 32 + LDD -32(%sp),lt ; lt + ADD,L ht,temp1,ht ; ht+= m>>32 + ADD lt,temp3,lt ; lt = lt+m1 + ADD,DC ht,%r0,ht ; ht++ + + ADD ht,ht,ht ; ht=ht+ht; + ADD,DC C3,%r0,C3 ; add in carry (c3++) + + ADD lt,lt,lt ; lt=lt+lt; + ADD,DC ht,%r0,ht ; add in carry (ht++) + + ADD C1,lt,C1 ; c1=c1+lt + ADD,DC,*NUV ht,%r0,ht ; add in carry (ht++) + LDO 1(C3),C3 ; bump c3 if overflow,nullify otherwise + + ADD C2,ht,C2 ; c2 = c2 + ht + ADD,DC C3,%r0,C3 ; add in carry (c3++) +.endm + +; +;void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a) +; arg0 = r_ptr +; arg1 = a_ptr +; + +bn_sqr_comba8 + .PROC + .CALLINFO FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_sqr_comba8,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .ENTRY + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + + ; + ; Zero out carries + ; + COPY %r0,c1 + COPY %r0,c2 + COPY %r0,c3 + + LDO 128(%sp),%sp ; bump stack + DEPDI,Z -1,32,33,high_mask ; Create Mask 0xffffffff80000000L + DEPDI,Z 1,31,1,high_one ; Create Value 1 << 32 + + ; + ; Load up all of the values we are going to use + ; + FLDD 0(a_ptr),a0 + FLDD 8(a_ptr),a1 + FLDD 16(a_ptr),a2 + FLDD 24(a_ptr),a3 + FLDD 32(a_ptr),a4 + FLDD 40(a_ptr),a5 + FLDD 48(a_ptr),a6 + FLDD 56(a_ptr),a7 + + SQR_ADD_C a0L,a0R,c1,c2,c3 + STD c1,0(r_ptr) ; r[0] = c1; + COPY %r0,c1 + + SQR_ADD_C2 a1L,a1R,a0L,a0R,c2,c3,c1 + STD c2,8(r_ptr) ; r[1] = c2; + COPY %r0,c2 + + SQR_ADD_C a1L,a1R,c3,c1,c2 + SQR_ADD_C2 a2L,a2R,a0L,a0R,c3,c1,c2 + STD c3,16(r_ptr) ; r[2] = c3; + COPY %r0,c3 + + SQR_ADD_C2 a3L,a3R,a0L,a0R,c1,c2,c3 + SQR_ADD_C2 a2L,a2R,a1L,a1R,c1,c2,c3 + STD c1,24(r_ptr) ; r[3] = c1; + COPY %r0,c1 + + SQR_ADD_C a2L,a2R,c2,c3,c1 + SQR_ADD_C2 a3L,a3R,a1L,a1R,c2,c3,c1 + SQR_ADD_C2 a4L,a4R,a0L,a0R,c2,c3,c1 + STD c2,32(r_ptr) ; r[4] = c2; + COPY %r0,c2 + + SQR_ADD_C2 a5L,a5R,a0L,a0R,c3,c1,c2 + SQR_ADD_C2 a4L,a4R,a1L,a1R,c3,c1,c2 + SQR_ADD_C2 a3L,a3R,a2L,a2R,c3,c1,c2 + STD c3,40(r_ptr) ; r[5] = c3; + COPY %r0,c3 + + SQR_ADD_C a3L,a3R,c1,c2,c3 + SQR_ADD_C2 a4L,a4R,a2L,a2R,c1,c2,c3 + SQR_ADD_C2 a5L,a5R,a1L,a1R,c1,c2,c3 + SQR_ADD_C2 a6L,a6R,a0L,a0R,c1,c2,c3 + STD c1,48(r_ptr) ; r[6] = c1; + COPY %r0,c1 + + SQR_ADD_C2 a7L,a7R,a0L,a0R,c2,c3,c1 + SQR_ADD_C2 a6L,a6R,a1L,a1R,c2,c3,c1 + SQR_ADD_C2 a5L,a5R,a2L,a2R,c2,c3,c1 + SQR_ADD_C2 a4L,a4R,a3L,a3R,c2,c3,c1 + STD c2,56(r_ptr) ; r[7] = c2; + COPY %r0,c2 + + SQR_ADD_C a4L,a4R,c3,c1,c2 + SQR_ADD_C2 a5L,a5R,a3L,a3R,c3,c1,c2 + SQR_ADD_C2 a6L,a6R,a2L,a2R,c3,c1,c2 + SQR_ADD_C2 a7L,a7R,a1L,a1R,c3,c1,c2 + STD c3,64(r_ptr) ; r[8] = c3; + COPY %r0,c3 + + SQR_ADD_C2 a7L,a7R,a2L,a2R,c1,c2,c3 + SQR_ADD_C2 a6L,a6R,a3L,a3R,c1,c2,c3 + SQR_ADD_C2 a5L,a5R,a4L,a4R,c1,c2,c3 + STD c1,72(r_ptr) ; r[9] = c1; + COPY %r0,c1 + + SQR_ADD_C a5L,a5R,c2,c3,c1 + SQR_ADD_C2 a6L,a6R,a4L,a4R,c2,c3,c1 + SQR_ADD_C2 a7L,a7R,a3L,a3R,c2,c3,c1 + STD c2,80(r_ptr) ; r[10] = c2; + COPY %r0,c2 + + SQR_ADD_C2 a7L,a7R,a4L,a4R,c3,c1,c2 + SQR_ADD_C2 a6L,a6R,a5L,a5R,c3,c1,c2 + STD c3,88(r_ptr) ; r[11] = c3; + COPY %r0,c3 + + SQR_ADD_C a6L,a6R,c1,c2,c3 + SQR_ADD_C2 a7L,a7R,a5L,a5R,c1,c2,c3 + STD c1,96(r_ptr) ; r[12] = c1; + COPY %r0,c1 + + SQR_ADD_C2 a7L,a7R,a6L,a6R,c2,c3,c1 + STD c2,104(r_ptr) ; r[13] = c2; + COPY %r0,c2 + + SQR_ADD_C a7L,a7R,c3,c1,c2 + STD c3, 112(r_ptr) ; r[14] = c3 + STD c1, 120(r_ptr) ; r[15] = c1 + + .EXIT + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + + .PROCEND + +;----------------------------------------------------------------------------- +; +;void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a) +; arg0 = r_ptr +; arg1 = a_ptr +; + +bn_sqr_comba4 + .proc + .callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_sqr_comba4,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + + ; + ; Zero out carries + ; + COPY %r0,c1 + COPY %r0,c2 + COPY %r0,c3 + + LDO 128(%sp),%sp ; bump stack + DEPDI,Z -1,32,33,high_mask ; Create Mask 0xffffffff80000000L + DEPDI,Z 1,31,1,high_one ; Create Value 1 << 32 + + ; + ; Load up all of the values we are going to use + ; + FLDD 0(a_ptr),a0 + FLDD 8(a_ptr),a1 + FLDD 16(a_ptr),a2 + FLDD 24(a_ptr),a3 + FLDD 32(a_ptr),a4 + FLDD 40(a_ptr),a5 + FLDD 48(a_ptr),a6 + FLDD 56(a_ptr),a7 + + SQR_ADD_C a0L,a0R,c1,c2,c3 + + STD c1,0(r_ptr) ; r[0] = c1; + COPY %r0,c1 + + SQR_ADD_C2 a1L,a1R,a0L,a0R,c2,c3,c1 + + STD c2,8(r_ptr) ; r[1] = c2; + COPY %r0,c2 + + SQR_ADD_C a1L,a1R,c3,c1,c2 + SQR_ADD_C2 a2L,a2R,a0L,a0R,c3,c1,c2 + + STD c3,16(r_ptr) ; r[2] = c3; + COPY %r0,c3 + + SQR_ADD_C2 a3L,a3R,a0L,a0R,c1,c2,c3 + SQR_ADD_C2 a2L,a2R,a1L,a1R,c1,c2,c3 + + STD c1,24(r_ptr) ; r[3] = c1; + COPY %r0,c1 + + SQR_ADD_C a2L,a2R,c2,c3,c1 + SQR_ADD_C2 a3L,a3R,a1L,a1R,c2,c3,c1 + + STD c2,32(r_ptr) ; r[4] = c2; + COPY %r0,c2 + + SQR_ADD_C2 a3L,a3R,a2L,a2R,c3,c1,c2 + STD c3,40(r_ptr) ; r[5] = c3; + COPY %r0,c3 + + SQR_ADD_C a3L,a3R,c1,c2,c3 + STD c1,48(r_ptr) ; r[6] = c1; + STD c2,56(r_ptr) ; r[7] = c2; + + .EXIT + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + + .PROCEND + + +;--------------------------------------------------------------------------- + +MUL_ADD_C .macro A0L,A0R,B0L,B0R,C1,C2,C3 + XMPYU A0L,B0R,ftemp1 ; m1 = bl*ht + FSTD ftemp1,-16(%sp) ; + XMPYU A0R,B0L,ftemp2 ; m = bh*lt + FSTD ftemp2,-8(%sp) ; + XMPYU A0R,B0R,ftemp3 ; lt = bl*lt + FSTD ftemp3,-32(%sp) + XMPYU A0L,B0L,ftemp4 ; ht = bh*ht + FSTD ftemp4,-24(%sp) ; + + LDD -8(%sp),m ; r21 = m + LDD -16(%sp),m1 ; r19 = m1 + ADD,L m,m1,m ; m+m1 + + DEPD,Z m,31,32,temp3 ; (m+m1<<32) + LDD -24(%sp),ht ; r24 = ht + + CMPCLR,*>>= m,m1,%r0 ; if (m < m1) + ADD,L ht,high_one,ht ; ht+=high_one + + EXTRD,U m,31,32,temp1 ; m >> 32 + LDD -32(%sp),lt ; lt + ADD,L ht,temp1,ht ; ht+= m>>32 + ADD lt,temp3,lt ; lt = lt+m1 + ADD,DC ht,%r0,ht ; ht++ + + ADD C1,lt,C1 ; c1=c1+lt + ADD,DC ht,%r0,ht ; bump c3 if overflow,nullify otherwise + + ADD C2,ht,C2 ; c2 = c2 + ht + ADD,DC C3,%r0,C3 ; add in carry (c3++) +.endm + + +; +;void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b) +; arg0 = r_ptr +; arg1 = a_ptr +; arg2 = b_ptr +; + +bn_mul_comba8 + .proc + .callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_mul_comba8,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + FSTD %fr12,32(%sp) ; save r6 + FSTD %fr13,40(%sp) ; save r7 + + ; + ; Zero out carries + ; + COPY %r0,c1 + COPY %r0,c2 + COPY %r0,c3 + + LDO 128(%sp),%sp ; bump stack + DEPDI,Z 1,31,1,high_one ; Create Value 1 << 32 + + ; + ; Load up all of the values we are going to use + ; + FLDD 0(a_ptr),a0 + FLDD 8(a_ptr),a1 + FLDD 16(a_ptr),a2 + FLDD 24(a_ptr),a3 + FLDD 32(a_ptr),a4 + FLDD 40(a_ptr),a5 + FLDD 48(a_ptr),a6 + FLDD 56(a_ptr),a7 + + FLDD 0(b_ptr),b0 + FLDD 8(b_ptr),b1 + FLDD 16(b_ptr),b2 + FLDD 24(b_ptr),b3 + FLDD 32(b_ptr),b4 + FLDD 40(b_ptr),b5 + FLDD 48(b_ptr),b6 + FLDD 56(b_ptr),b7 + + MUL_ADD_C a0L,a0R,b0L,b0R,c1,c2,c3 + STD c1,0(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a0L,a0R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b0L,b0R,c2,c3,c1 + STD c2,8(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a2L,a2R,b0L,b0R,c3,c1,c2 + MUL_ADD_C a1L,a1R,b1L,b1R,c3,c1,c2 + MUL_ADD_C a0L,a0R,b2L,b2R,c3,c1,c2 + STD c3,16(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a0L,a0R,b3L,b3R,c1,c2,c3 + MUL_ADD_C a1L,a1R,b2L,b2R,c1,c2,c3 + MUL_ADD_C a2L,a2R,b1L,b1R,c1,c2,c3 + MUL_ADD_C a3L,a3R,b0L,b0R,c1,c2,c3 + STD c1,24(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a4L,a4R,b0L,b0R,c2,c3,c1 + MUL_ADD_C a3L,a3R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a2L,a2R,b2L,b2R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b3L,b3R,c2,c3,c1 + MUL_ADD_C a0L,a0R,b4L,b4R,c2,c3,c1 + STD c2,32(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a0L,a0R,b5L,b5R,c3,c1,c2 + MUL_ADD_C a1L,a1R,b4L,b4R,c3,c1,c2 + MUL_ADD_C a2L,a2R,b3L,b3R,c3,c1,c2 + MUL_ADD_C a3L,a3R,b2L,b2R,c3,c1,c2 + MUL_ADD_C a4L,a4R,b1L,b1R,c3,c1,c2 + MUL_ADD_C a5L,a5R,b0L,b0R,c3,c1,c2 + STD c3,40(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a6L,a6R,b0L,b0R,c1,c2,c3 + MUL_ADD_C a5L,a5R,b1L,b1R,c1,c2,c3 + MUL_ADD_C a4L,a4R,b2L,b2R,c1,c2,c3 + MUL_ADD_C a3L,a3R,b3L,b3R,c1,c2,c3 + MUL_ADD_C a2L,a2R,b4L,b4R,c1,c2,c3 + MUL_ADD_C a1L,a1R,b5L,b5R,c1,c2,c3 + MUL_ADD_C a0L,a0R,b6L,b6R,c1,c2,c3 + STD c1,48(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a0L,a0R,b7L,b7R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b6L,b6R,c2,c3,c1 + MUL_ADD_C a2L,a2R,b5L,b5R,c2,c3,c1 + MUL_ADD_C a3L,a3R,b4L,b4R,c2,c3,c1 + MUL_ADD_C a4L,a4R,b3L,b3R,c2,c3,c1 + MUL_ADD_C a5L,a5R,b2L,b2R,c2,c3,c1 + MUL_ADD_C a6L,a6R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a7L,a7R,b0L,b0R,c2,c3,c1 + STD c2,56(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a7L,a7R,b1L,b1R,c3,c1,c2 + MUL_ADD_C a6L,a6R,b2L,b2R,c3,c1,c2 + MUL_ADD_C a5L,a5R,b3L,b3R,c3,c1,c2 + MUL_ADD_C a4L,a4R,b4L,b4R,c3,c1,c2 + MUL_ADD_C a3L,a3R,b5L,b5R,c3,c1,c2 + MUL_ADD_C a2L,a2R,b6L,b6R,c3,c1,c2 + MUL_ADD_C a1L,a1R,b7L,b7R,c3,c1,c2 + STD c3,64(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a2L,a2R,b7L,b7R,c1,c2,c3 + MUL_ADD_C a3L,a3R,b6L,b6R,c1,c2,c3 + MUL_ADD_C a4L,a4R,b5L,b5R,c1,c2,c3 + MUL_ADD_C a5L,a5R,b4L,b4R,c1,c2,c3 + MUL_ADD_C a6L,a6R,b3L,b3R,c1,c2,c3 + MUL_ADD_C a7L,a7R,b2L,b2R,c1,c2,c3 + STD c1,72(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a7L,a7R,b3L,b3R,c2,c3,c1 + MUL_ADD_C a6L,a6R,b4L,b4R,c2,c3,c1 + MUL_ADD_C a5L,a5R,b5L,b5R,c2,c3,c1 + MUL_ADD_C a4L,a4R,b6L,b6R,c2,c3,c1 + MUL_ADD_C a3L,a3R,b7L,b7R,c2,c3,c1 + STD c2,80(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a4L,a4R,b7L,b7R,c3,c1,c2 + MUL_ADD_C a5L,a5R,b6L,b6R,c3,c1,c2 + MUL_ADD_C a6L,a6R,b5L,b5R,c3,c1,c2 + MUL_ADD_C a7L,a7R,b4L,b4R,c3,c1,c2 + STD c3,88(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a7L,a7R,b5L,b5R,c1,c2,c3 + MUL_ADD_C a6L,a6R,b6L,b6R,c1,c2,c3 + MUL_ADD_C a5L,a5R,b7L,b7R,c1,c2,c3 + STD c1,96(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a6L,a6R,b7L,b7R,c2,c3,c1 + MUL_ADD_C a7L,a7R,b6L,b6R,c2,c3,c1 + STD c2,104(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a7L,a7R,b7L,b7R,c3,c1,c2 + STD c3,112(r_ptr) + STD c1,120(r_ptr) + + .EXIT + FLDD -88(%sp),%fr13 + FLDD -96(%sp),%fr12 + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + + .PROCEND + +;----------------------------------------------------------------------------- +; +;void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b) +; arg0 = r_ptr +; arg1 = a_ptr +; arg2 = b_ptr +; + +bn_mul_comba4 + .proc + .callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_mul_comba4,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + FSTD %fr12,32(%sp) ; save r6 + FSTD %fr13,40(%sp) ; save r7 + + ; + ; Zero out carries + ; + COPY %r0,c1 + COPY %r0,c2 + COPY %r0,c3 + + LDO 128(%sp),%sp ; bump stack + DEPDI,Z 1,31,1,high_one ; Create Value 1 << 32 + + ; + ; Load up all of the values we are going to use + ; + FLDD 0(a_ptr),a0 + FLDD 8(a_ptr),a1 + FLDD 16(a_ptr),a2 + FLDD 24(a_ptr),a3 + + FLDD 0(b_ptr),b0 + FLDD 8(b_ptr),b1 + FLDD 16(b_ptr),b2 + FLDD 24(b_ptr),b3 + + MUL_ADD_C a0L,a0R,b0L,b0R,c1,c2,c3 + STD c1,0(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a0L,a0R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b0L,b0R,c2,c3,c1 + STD c2,8(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a2L,a2R,b0L,b0R,c3,c1,c2 + MUL_ADD_C a1L,a1R,b1L,b1R,c3,c1,c2 + MUL_ADD_C a0L,a0R,b2L,b2R,c3,c1,c2 + STD c3,16(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a0L,a0R,b3L,b3R,c1,c2,c3 + MUL_ADD_C a1L,a1R,b2L,b2R,c1,c2,c3 + MUL_ADD_C a2L,a2R,b1L,b1R,c1,c2,c3 + MUL_ADD_C a3L,a3R,b0L,b0R,c1,c2,c3 + STD c1,24(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a3L,a3R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a2L,a2R,b2L,b2R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b3L,b3R,c2,c3,c1 + STD c2,32(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a2L,a2R,b3L,b3R,c3,c1,c2 + MUL_ADD_C a3L,a3R,b2L,b2R,c3,c1,c2 + STD c3,40(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a3L,a3R,b3L,b3R,c1,c2,c3 + STD c1,48(r_ptr) + STD c2,56(r_ptr) + + .EXIT + FLDD -88(%sp),%fr13 + FLDD -96(%sp),%fr12 + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + + .PROCEND + + +;--- not PIC .SPACE $TEXT$ +;--- not PIC .SUBSPA $CODE$ +;--- not PIC .SPACE $PRIVATE$,SORT=16 +;--- not PIC .IMPORT $global$,DATA +;--- not PIC .SPACE $TEXT$ +;--- not PIC .SUBSPA $CODE$ +;--- not PIC .SUBSPA $LIT$,ACCESS=0x2c +;--- not PIC C$7 +;--- not PIC .ALIGN 8 +;--- not PIC .STRINGZ "Division would overflow (%d)\n" + .END diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/pa-risc2W.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/pa-risc2W.S new file mode 100644 index 00000000..a9954575 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/pa-risc2W.S @@ -0,0 +1,1605 @@ +; +; PA-RISC 64-bit implementation of bn_asm code +; +; This code is approximately 2x faster than the C version +; for RSA/DSA. +; +; See http://devresource.hp.com/ for more details on the PA-RISC +; architecture. Also see the book "PA-RISC 2.0 Architecture" +; by Gerry Kane for information on the instruction set architecture. +; +; Code written by Chris Ruemmler (with some help from the HP C +; compiler). +; +; The code compiles with HP's assembler +; + + .level 2.0W + .space $TEXT$ + .subspa $CODE$,QUAD=0,ALIGN=8,ACCESS=0x2c,CODE_ONLY + +; +; Global Register definitions used for the routines. +; +; Some information about HP's runtime architecture for 64-bits. +; +; "Caller save" means the calling function must save the register +; if it wants the register to be preserved. +; "Callee save" means if a function uses the register, it must save +; the value before using it. +; +; For the floating point registers +; +; "caller save" registers: fr4-fr11, fr22-fr31 +; "callee save" registers: fr12-fr21 +; "special" registers: fr0-fr3 (status and exception registers) +; +; For the integer registers +; value zero : r0 +; "caller save" registers: r1,r19-r26 +; "callee save" registers: r3-r18 +; return register : r2 (rp) +; return values ; r28 (ret0,ret1) +; Stack pointer ; r30 (sp) +; global data pointer ; r27 (dp) +; argument pointer ; r29 (ap) +; millicode return ptr ; r31 (also a caller save register) + + +; +; Arguments to the routines +; +r_ptr .reg %r26 +a_ptr .reg %r25 +b_ptr .reg %r24 +num .reg %r24 +w .reg %r23 +n .reg %r23 + + +; +; Globals used in some routines +; + +top_overflow .reg %r29 +high_mask .reg %r22 ; value 0xffffffff80000000L + + +;------------------------------------------------------------------------------ +; +; bn_mul_add_words +; +;BN_ULONG bn_mul_add_words(BN_ULONG *r_ptr, BN_ULONG *a_ptr, +; int num, BN_ULONG w) +; +; arg0 = r_ptr +; arg1 = a_ptr +; arg2 = num +; arg3 = w +; +; Local register definitions +; + +fm1 .reg %fr22 +fm .reg %fr23 +ht_temp .reg %fr24 +ht_temp_1 .reg %fr25 +lt_temp .reg %fr26 +lt_temp_1 .reg %fr27 +fm1_1 .reg %fr28 +fm_1 .reg %fr29 + +fw_h .reg %fr7L +fw_l .reg %fr7R +fw .reg %fr7 + +fht_0 .reg %fr8L +flt_0 .reg %fr8R +t_float_0 .reg %fr8 + +fht_1 .reg %fr9L +flt_1 .reg %fr9R +t_float_1 .reg %fr9 + +tmp_0 .reg %r31 +tmp_1 .reg %r21 +m_0 .reg %r20 +m_1 .reg %r19 +ht_0 .reg %r1 +ht_1 .reg %r3 +lt_0 .reg %r4 +lt_1 .reg %r5 +m1_0 .reg %r6 +m1_1 .reg %r7 +rp_val .reg %r8 +rp_val_1 .reg %r9 + +bn_mul_add_words + .export bn_mul_add_words,entry,NO_RELOCATION,LONG_RETURN + .proc + .callinfo frame=128 + .entry + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + NOP ; Needed to make the loop 16-byte aligned + NOP ; Needed to make the loop 16-byte aligned + + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + STD %r7,32(%sp) ; save r7 + STD %r8,40(%sp) ; save r8 + + STD %r9,48(%sp) ; save r9 + COPY %r0,%ret0 ; return 0 by default + DEPDI,Z 1,31,1,top_overflow ; top_overflow = 1 << 32 + STD w,56(%sp) ; store w on stack + + CMPIB,>= 0,num,bn_mul_add_words_exit ; if (num <= 0) then exit + LDO 128(%sp),%sp ; bump stack + + ; + ; The loop is unrolled twice, so if there is only 1 number + ; then go straight to the cleanup code. + ; + CMPIB,= 1,num,bn_mul_add_words_single_top + FLDD -72(%sp),fw ; load up w into fp register fw (fw_h/fw_l) + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; + ; PA-RISC 2.0 chips have two fully pipelined multipliers, thus + ; two 32-bit mutiplies can be issued per cycle. + ; +bn_mul_add_words_unroll2 + + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + FLDD 8(a_ptr),t_float_1 ; load up 64-bit value (fr8L) ht(L)/lt(R) + LDD 0(r_ptr),rp_val ; rp[0] + LDD 8(r_ptr),rp_val_1 ; rp[1] + + XMPYU fht_0,fw_l,fm1 ; m1[0] = fht_0*fw_l + XMPYU fht_1,fw_l,fm1_1 ; m1[1] = fht_1*fw_l + FSTD fm1,-16(%sp) ; -16(sp) = m1[0] + FSTD fm1_1,-48(%sp) ; -48(sp) = m1[1] + + XMPYU flt_0,fw_h,fm ; m[0] = flt_0*fw_h + XMPYU flt_1,fw_h,fm_1 ; m[1] = flt_1*fw_h + FSTD fm,-8(%sp) ; -8(sp) = m[0] + FSTD fm_1,-40(%sp) ; -40(sp) = m[1] + + XMPYU fht_0,fw_h,ht_temp ; ht_temp = fht_0*fw_h + XMPYU fht_1,fw_h,ht_temp_1 ; ht_temp_1 = fht_1*fw_h + FSTD ht_temp,-24(%sp) ; -24(sp) = ht_temp + FSTD ht_temp_1,-56(%sp) ; -56(sp) = ht_temp_1 + + XMPYU flt_0,fw_l,lt_temp ; lt_temp = lt*fw_l + XMPYU flt_1,fw_l,lt_temp_1 ; lt_temp = lt*fw_l + FSTD lt_temp,-32(%sp) ; -32(sp) = lt_temp + FSTD lt_temp_1,-64(%sp) ; -64(sp) = lt_temp_1 + + LDD -8(%sp),m_0 ; m[0] + LDD -40(%sp),m_1 ; m[1] + LDD -16(%sp),m1_0 ; m1[0] + LDD -48(%sp),m1_1 ; m1[1] + + LDD -24(%sp),ht_0 ; ht[0] + LDD -56(%sp),ht_1 ; ht[1] + ADD,L m1_0,m_0,tmp_0 ; tmp_0 = m[0] + m1[0]; + ADD,L m1_1,m_1,tmp_1 ; tmp_1 = m[1] + m1[1]; + + LDD -32(%sp),lt_0 + LDD -64(%sp),lt_1 + CMPCLR,*>>= tmp_0,m1_0, %r0 ; if (m[0] < m1[0]) + ADD,L ht_0,top_overflow,ht_0 ; ht[0] += (1<<32) + + CMPCLR,*>>= tmp_1,m1_1,%r0 ; if (m[1] < m1[1]) + ADD,L ht_1,top_overflow,ht_1 ; ht[1] += (1<<32) + EXTRD,U tmp_0,31,32,m_0 ; m[0]>>32 + DEPD,Z tmp_0,31,32,m1_0 ; m1[0] = m[0]<<32 + + EXTRD,U tmp_1,31,32,m_1 ; m[1]>>32 + DEPD,Z tmp_1,31,32,m1_1 ; m1[1] = m[1]<<32 + ADD,L ht_0,m_0,ht_0 ; ht[0]+= (m[0]>>32) + ADD,L ht_1,m_1,ht_1 ; ht[1]+= (m[1]>>32) + + ADD lt_0,m1_0,lt_0 ; lt[0] = lt[0]+m1[0]; + ADD,DC ht_0,%r0,ht_0 ; ht[0]++ + ADD lt_1,m1_1,lt_1 ; lt[1] = lt[1]+m1[1]; + ADD,DC ht_1,%r0,ht_1 ; ht[1]++ + + ADD %ret0,lt_0,lt_0 ; lt[0] = lt[0] + c; + ADD,DC ht_0,%r0,ht_0 ; ht[0]++ + ADD lt_0,rp_val,lt_0 ; lt[0] = lt[0]+rp[0] + ADD,DC ht_0,%r0,ht_0 ; ht[0]++ + + LDO -2(num),num ; num = num - 2; + ADD ht_0,lt_1,lt_1 ; lt[1] = lt[1] + ht_0 (c); + ADD,DC ht_1,%r0,ht_1 ; ht[1]++ + STD lt_0,0(r_ptr) ; rp[0] = lt[0] + + ADD lt_1,rp_val_1,lt_1 ; lt[1] = lt[1]+rp[1] + ADD,DC ht_1,%r0,%ret0 ; ht[1]++ + LDO 16(a_ptr),a_ptr ; a_ptr += 2 + + STD lt_1,8(r_ptr) ; rp[1] = lt[1] + CMPIB,<= 2,num,bn_mul_add_words_unroll2 ; go again if more to do + LDO 16(r_ptr),r_ptr ; r_ptr += 2 + + CMPIB,=,N 0,num,bn_mul_add_words_exit ; are we done, or cleanup last one + + ; + ; Top of loop aligned on 64-byte boundary + ; +bn_mul_add_words_single_top + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + LDD 0(r_ptr),rp_val ; rp[0] + LDO 8(a_ptr),a_ptr ; a_ptr++ + XMPYU fht_0,fw_l,fm1 ; m1 = ht*fw_l + FSTD fm1,-16(%sp) ; -16(sp) = m1 + XMPYU flt_0,fw_h,fm ; m = lt*fw_h + FSTD fm,-8(%sp) ; -8(sp) = m + XMPYU fht_0,fw_h,ht_temp ; ht_temp = ht*fw_h + FSTD ht_temp,-24(%sp) ; -24(sp) = ht + XMPYU flt_0,fw_l,lt_temp ; lt_temp = lt*fw_l + FSTD lt_temp,-32(%sp) ; -32(sp) = lt + + LDD -8(%sp),m_0 + LDD -16(%sp),m1_0 ; m1 = temp1 + ADD,L m_0,m1_0,tmp_0 ; tmp_0 = m + m1; + LDD -24(%sp),ht_0 + LDD -32(%sp),lt_0 + + CMPCLR,*>>= tmp_0,m1_0,%r0 ; if (m < m1) + ADD,L ht_0,top_overflow,ht_0 ; ht += (1<<32) + + EXTRD,U tmp_0,31,32,m_0 ; m>>32 + DEPD,Z tmp_0,31,32,m1_0 ; m1 = m<<32 + + ADD,L ht_0,m_0,ht_0 ; ht+= (m>>32) + ADD lt_0,m1_0,tmp_0 ; tmp_0 = lt+m1; + ADD,DC ht_0,%r0,ht_0 ; ht++ + ADD %ret0,tmp_0,lt_0 ; lt = lt + c; + ADD,DC ht_0,%r0,ht_0 ; ht++ + ADD lt_0,rp_val,lt_0 ; lt = lt+rp[0] + ADD,DC ht_0,%r0,%ret0 ; ht++ + STD lt_0,0(r_ptr) ; rp[0] = lt + +bn_mul_add_words_exit + .EXIT + LDD -80(%sp),%r9 ; restore r9 + LDD -88(%sp),%r8 ; restore r8 + LDD -96(%sp),%r7 ; restore r7 + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 ; restore r3 + .PROCEND ;in=23,24,25,26,29;out=28; + +;---------------------------------------------------------------------------- +; +;BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w) +; +; arg0 = rp +; arg1 = ap +; arg2 = num +; arg3 = w + +bn_mul_words + .proc + .callinfo frame=128 + .entry + .EXPORT bn_mul_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + + STD %r7,32(%sp) ; save r7 + COPY %r0,%ret0 ; return 0 by default + DEPDI,Z 1,31,1,top_overflow ; top_overflow = 1 << 32 + STD w,56(%sp) ; w on stack + + CMPIB,>= 0,num,bn_mul_words_exit + LDO 128(%sp),%sp ; bump stack + + ; + ; See if only 1 word to do, thus just do cleanup + ; + CMPIB,= 1,num,bn_mul_words_single_top + FLDD -72(%sp),fw ; load up w into fp register fw (fw_h/fw_l) + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; + ; PA-RISC 2.0 chips have two fully pipelined multipliers, thus + ; two 32-bit mutiplies can be issued per cycle. + ; +bn_mul_words_unroll2 + + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + FLDD 8(a_ptr),t_float_1 ; load up 64-bit value (fr8L) ht(L)/lt(R) + XMPYU fht_0,fw_l,fm1 ; m1[0] = fht_0*fw_l + XMPYU fht_1,fw_l,fm1_1 ; m1[1] = ht*fw_l + + FSTD fm1,-16(%sp) ; -16(sp) = m1 + FSTD fm1_1,-48(%sp) ; -48(sp) = m1 + XMPYU flt_0,fw_h,fm ; m = lt*fw_h + XMPYU flt_1,fw_h,fm_1 ; m = lt*fw_h + + FSTD fm,-8(%sp) ; -8(sp) = m + FSTD fm_1,-40(%sp) ; -40(sp) = m + XMPYU fht_0,fw_h,ht_temp ; ht_temp = fht_0*fw_h + XMPYU fht_1,fw_h,ht_temp_1 ; ht_temp = ht*fw_h + + FSTD ht_temp,-24(%sp) ; -24(sp) = ht + FSTD ht_temp_1,-56(%sp) ; -56(sp) = ht + XMPYU flt_0,fw_l,lt_temp ; lt_temp = lt*fw_l + XMPYU flt_1,fw_l,lt_temp_1 ; lt_temp = lt*fw_l + + FSTD lt_temp,-32(%sp) ; -32(sp) = lt + FSTD lt_temp_1,-64(%sp) ; -64(sp) = lt + LDD -8(%sp),m_0 + LDD -40(%sp),m_1 + + LDD -16(%sp),m1_0 + LDD -48(%sp),m1_1 + LDD -24(%sp),ht_0 + LDD -56(%sp),ht_1 + + ADD,L m1_0,m_0,tmp_0 ; tmp_0 = m + m1; + ADD,L m1_1,m_1,tmp_1 ; tmp_1 = m + m1; + LDD -32(%sp),lt_0 + LDD -64(%sp),lt_1 + + CMPCLR,*>>= tmp_0,m1_0, %r0 ; if (m < m1) + ADD,L ht_0,top_overflow,ht_0 ; ht += (1<<32) + CMPCLR,*>>= tmp_1,m1_1,%r0 ; if (m < m1) + ADD,L ht_1,top_overflow,ht_1 ; ht += (1<<32) + + EXTRD,U tmp_0,31,32,m_0 ; m>>32 + DEPD,Z tmp_0,31,32,m1_0 ; m1 = m<<32 + EXTRD,U tmp_1,31,32,m_1 ; m>>32 + DEPD,Z tmp_1,31,32,m1_1 ; m1 = m<<32 + + ADD,L ht_0,m_0,ht_0 ; ht+= (m>>32) + ADD,L ht_1,m_1,ht_1 ; ht+= (m>>32) + ADD lt_0,m1_0,lt_0 ; lt = lt+m1; + ADD,DC ht_0,%r0,ht_0 ; ht++ + + ADD lt_1,m1_1,lt_1 ; lt = lt+m1; + ADD,DC ht_1,%r0,ht_1 ; ht++ + ADD %ret0,lt_0,lt_0 ; lt = lt + c (ret0); + ADD,DC ht_0,%r0,ht_0 ; ht++ + + ADD ht_0,lt_1,lt_1 ; lt = lt + c (ht_0) + ADD,DC ht_1,%r0,ht_1 ; ht++ + STD lt_0,0(r_ptr) ; rp[0] = lt + STD lt_1,8(r_ptr) ; rp[1] = lt + + COPY ht_1,%ret0 ; carry = ht + LDO -2(num),num ; num = num - 2; + LDO 16(a_ptr),a_ptr ; ap += 2 + CMPIB,<= 2,num,bn_mul_words_unroll2 + LDO 16(r_ptr),r_ptr ; rp++ + + CMPIB,=,N 0,num,bn_mul_words_exit ; are we done? + + ; + ; Top of loop aligned on 64-byte boundary + ; +bn_mul_words_single_top + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + + XMPYU fht_0,fw_l,fm1 ; m1 = ht*fw_l + FSTD fm1,-16(%sp) ; -16(sp) = m1 + XMPYU flt_0,fw_h,fm ; m = lt*fw_h + FSTD fm,-8(%sp) ; -8(sp) = m + XMPYU fht_0,fw_h,ht_temp ; ht_temp = ht*fw_h + FSTD ht_temp,-24(%sp) ; -24(sp) = ht + XMPYU flt_0,fw_l,lt_temp ; lt_temp = lt*fw_l + FSTD lt_temp,-32(%sp) ; -32(sp) = lt + + LDD -8(%sp),m_0 + LDD -16(%sp),m1_0 + ADD,L m_0,m1_0,tmp_0 ; tmp_0 = m + m1; + LDD -24(%sp),ht_0 + LDD -32(%sp),lt_0 + + CMPCLR,*>>= tmp_0,m1_0,%r0 ; if (m < m1) + ADD,L ht_0,top_overflow,ht_0 ; ht += (1<<32) + + EXTRD,U tmp_0,31,32,m_0 ; m>>32 + DEPD,Z tmp_0,31,32,m1_0 ; m1 = m<<32 + + ADD,L ht_0,m_0,ht_0 ; ht+= (m>>32) + ADD lt_0,m1_0,lt_0 ; lt= lt+m1; + ADD,DC ht_0,%r0,ht_0 ; ht++ + + ADD %ret0,lt_0,lt_0 ; lt = lt + c; + ADD,DC ht_0,%r0,ht_0 ; ht++ + + COPY ht_0,%ret0 ; copy carry + STD lt_0,0(r_ptr) ; rp[0] = lt + +bn_mul_words_exit + .EXIT + LDD -96(%sp),%r7 ; restore r7 + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 ; restore r3 + .PROCEND ;in=23,24,25,26,29;out=28; + +;---------------------------------------------------------------------------- +; +;void bn_sqr_words(BN_ULONG *rp, BN_ULONG *ap, int num) +; +; arg0 = rp +; arg1 = ap +; arg2 = num +; + +bn_sqr_words + .proc + .callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_sqr_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + NOP + STD %r5,16(%sp) ; save r5 + + CMPIB,>= 0,num,bn_sqr_words_exit + LDO 128(%sp),%sp ; bump stack + + ; + ; If only 1, the goto straight to cleanup + ; + CMPIB,= 1,num,bn_sqr_words_single_top + DEPDI,Z -1,32,33,high_mask ; Create Mask 0xffffffff80000000L + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; + +bn_sqr_words_unroll2 + FLDD 0(a_ptr),t_float_0 ; a[0] + FLDD 8(a_ptr),t_float_1 ; a[1] + XMPYU fht_0,flt_0,fm ; m[0] + XMPYU fht_1,flt_1,fm_1 ; m[1] + + FSTD fm,-24(%sp) ; store m[0] + FSTD fm_1,-56(%sp) ; store m[1] + XMPYU flt_0,flt_0,lt_temp ; lt[0] + XMPYU flt_1,flt_1,lt_temp_1 ; lt[1] + + FSTD lt_temp,-16(%sp) ; store lt[0] + FSTD lt_temp_1,-48(%sp) ; store lt[1] + XMPYU fht_0,fht_0,ht_temp ; ht[0] + XMPYU fht_1,fht_1,ht_temp_1 ; ht[1] + + FSTD ht_temp,-8(%sp) ; store ht[0] + FSTD ht_temp_1,-40(%sp) ; store ht[1] + LDD -24(%sp),m_0 + LDD -56(%sp),m_1 + + AND m_0,high_mask,tmp_0 ; m[0] & Mask + AND m_1,high_mask,tmp_1 ; m[1] & Mask + DEPD,Z m_0,30,31,m_0 ; m[0] << 32+1 + DEPD,Z m_1,30,31,m_1 ; m[1] << 32+1 + + LDD -16(%sp),lt_0 + LDD -48(%sp),lt_1 + EXTRD,U tmp_0,32,33,tmp_0 ; tmp_0 = m[0]&Mask >> 32-1 + EXTRD,U tmp_1,32,33,tmp_1 ; tmp_1 = m[1]&Mask >> 32-1 + + LDD -8(%sp),ht_0 + LDD -40(%sp),ht_1 + ADD,L ht_0,tmp_0,ht_0 ; ht[0] += tmp_0 + ADD,L ht_1,tmp_1,ht_1 ; ht[1] += tmp_1 + + ADD lt_0,m_0,lt_0 ; lt = lt+m + ADD,DC ht_0,%r0,ht_0 ; ht[0]++ + STD lt_0,0(r_ptr) ; rp[0] = lt[0] + STD ht_0,8(r_ptr) ; rp[1] = ht[1] + + ADD lt_1,m_1,lt_1 ; lt = lt+m + ADD,DC ht_1,%r0,ht_1 ; ht[1]++ + STD lt_1,16(r_ptr) ; rp[2] = lt[1] + STD ht_1,24(r_ptr) ; rp[3] = ht[1] + + LDO -2(num),num ; num = num - 2; + LDO 16(a_ptr),a_ptr ; ap += 2 + CMPIB,<= 2,num,bn_sqr_words_unroll2 + LDO 32(r_ptr),r_ptr ; rp += 4 + + CMPIB,=,N 0,num,bn_sqr_words_exit ; are we done? + + ; + ; Top of loop aligned on 64-byte boundary + ; +bn_sqr_words_single_top + FLDD 0(a_ptr),t_float_0 ; load up 64-bit value (fr8L) ht(L)/lt(R) + + XMPYU fht_0,flt_0,fm ; m + FSTD fm,-24(%sp) ; store m + + XMPYU flt_0,flt_0,lt_temp ; lt + FSTD lt_temp,-16(%sp) ; store lt + + XMPYU fht_0,fht_0,ht_temp ; ht + FSTD ht_temp,-8(%sp) ; store ht + + LDD -24(%sp),m_0 ; load m + AND m_0,high_mask,tmp_0 ; m & Mask + DEPD,Z m_0,30,31,m_0 ; m << 32+1 + LDD -16(%sp),lt_0 ; lt + + LDD -8(%sp),ht_0 ; ht + EXTRD,U tmp_0,32,33,tmp_0 ; tmp_0 = m&Mask >> 32-1 + ADD m_0,lt_0,lt_0 ; lt = lt+m + ADD,L ht_0,tmp_0,ht_0 ; ht += tmp_0 + ADD,DC ht_0,%r0,ht_0 ; ht++ + + STD lt_0,0(r_ptr) ; rp[0] = lt + STD ht_0,8(r_ptr) ; rp[1] = ht + +bn_sqr_words_exit + .EXIT + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + .PROCEND ;in=23,24,25,26,29;out=28; + + +;---------------------------------------------------------------------------- +; +;BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) +; +; arg0 = rp +; arg1 = ap +; arg2 = bp +; arg3 = n + +t .reg %r22 +b .reg %r21 +l .reg %r20 + +bn_add_words + .proc + .entry + .callinfo + .EXPORT bn_add_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .align 64 + + CMPIB,>= 0,n,bn_add_words_exit + COPY %r0,%ret0 ; return 0 by default + + ; + ; If 2 or more numbers do the loop + ; + CMPIB,= 1,n,bn_add_words_single_top + NOP + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; +bn_add_words_unroll2 + LDD 0(a_ptr),t + LDD 0(b_ptr),b + ADD t,%ret0,t ; t = t+c; + ADD,DC %r0,%r0,%ret0 ; set c to carry + ADD t,b,l ; l = t + b[0] + ADD,DC %ret0,%r0,%ret0 ; c+= carry + STD l,0(r_ptr) + + LDD 8(a_ptr),t + LDD 8(b_ptr),b + ADD t,%ret0,t ; t = t+c; + ADD,DC %r0,%r0,%ret0 ; set c to carry + ADD t,b,l ; l = t + b[0] + ADD,DC %ret0,%r0,%ret0 ; c+= carry + STD l,8(r_ptr) + + LDO -2(n),n + LDO 16(a_ptr),a_ptr + LDO 16(b_ptr),b_ptr + + CMPIB,<= 2,n,bn_add_words_unroll2 + LDO 16(r_ptr),r_ptr + + CMPIB,=,N 0,n,bn_add_words_exit ; are we done? + +bn_add_words_single_top + LDD 0(a_ptr),t + LDD 0(b_ptr),b + + ADD t,%ret0,t ; t = t+c; + ADD,DC %r0,%r0,%ret0 ; set c to carry (could use CMPCLR??) + ADD t,b,l ; l = t + b[0] + ADD,DC %ret0,%r0,%ret0 ; c+= carry + STD l,0(r_ptr) + +bn_add_words_exit + .EXIT + BVE (%rp) + NOP + .PROCEND ;in=23,24,25,26,29;out=28; + +;---------------------------------------------------------------------------- +; +;BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) +; +; arg0 = rp +; arg1 = ap +; arg2 = bp +; arg3 = n + +t1 .reg %r22 +t2 .reg %r21 +sub_tmp1 .reg %r20 +sub_tmp2 .reg %r19 + + +bn_sub_words + .proc + .callinfo + .EXPORT bn_sub_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + + CMPIB,>= 0,n,bn_sub_words_exit + COPY %r0,%ret0 ; return 0 by default + + ; + ; If 2 or more numbers do the loop + ; + CMPIB,= 1,n,bn_sub_words_single_top + NOP + + ; + ; This loop is unrolled 2 times (64-byte aligned as well) + ; +bn_sub_words_unroll2 + LDD 0(a_ptr),t1 + LDD 0(b_ptr),t2 + SUB t1,t2,sub_tmp1 ; t3 = t1-t2; + SUB sub_tmp1,%ret0,sub_tmp1 ; t3 = t3- c; + + CMPCLR,*>> t1,t2,sub_tmp2 ; clear if t1 > t2 + LDO 1(%r0),sub_tmp2 + + CMPCLR,*= t1,t2,%r0 + COPY sub_tmp2,%ret0 + STD sub_tmp1,0(r_ptr) + + LDD 8(a_ptr),t1 + LDD 8(b_ptr),t2 + SUB t1,t2,sub_tmp1 ; t3 = t1-t2; + SUB sub_tmp1,%ret0,sub_tmp1 ; t3 = t3- c; + CMPCLR,*>> t1,t2,sub_tmp2 ; clear if t1 > t2 + LDO 1(%r0),sub_tmp2 + + CMPCLR,*= t1,t2,%r0 + COPY sub_tmp2,%ret0 + STD sub_tmp1,8(r_ptr) + + LDO -2(n),n + LDO 16(a_ptr),a_ptr + LDO 16(b_ptr),b_ptr + + CMPIB,<= 2,n,bn_sub_words_unroll2 + LDO 16(r_ptr),r_ptr + + CMPIB,=,N 0,n,bn_sub_words_exit ; are we done? + +bn_sub_words_single_top + LDD 0(a_ptr),t1 + LDD 0(b_ptr),t2 + SUB t1,t2,sub_tmp1 ; t3 = t1-t2; + SUB sub_tmp1,%ret0,sub_tmp1 ; t3 = t3- c; + CMPCLR,*>> t1,t2,sub_tmp2 ; clear if t1 > t2 + LDO 1(%r0),sub_tmp2 + + CMPCLR,*= t1,t2,%r0 + COPY sub_tmp2,%ret0 + + STD sub_tmp1,0(r_ptr) + +bn_sub_words_exit + .EXIT + BVE (%rp) + NOP + .PROCEND ;in=23,24,25,26,29;out=28; + +;------------------------------------------------------------------------------ +; +; unsigned long bn_div_words(unsigned long h, unsigned long l, unsigned long d) +; +; arg0 = h +; arg1 = l +; arg2 = d +; +; This is mainly just modified assembly from the compiler, thus the +; lack of variable names. +; +;------------------------------------------------------------------------------ +bn_div_words + .proc + .callinfo CALLER,FRAME=272,ENTRY_GR=%r10,SAVE_RP,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_div_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .IMPORT BN_num_bits_word,CODE,NO_RELOCATION + .IMPORT __iob,DATA + .IMPORT fprintf,CODE,NO_RELOCATION + .IMPORT abort,CODE,NO_RELOCATION + .IMPORT $$div2U,MILLICODE + .entry + STD %r2,-16(%r30) + STD,MA %r3,352(%r30) + STD %r4,-344(%r30) + STD %r5,-336(%r30) + STD %r6,-328(%r30) + STD %r7,-320(%r30) + STD %r8,-312(%r30) + STD %r9,-304(%r30) + STD %r10,-296(%r30) + + STD %r27,-288(%r30) ; save gp + + COPY %r24,%r3 ; save d + COPY %r26,%r4 ; save h (high 64-bits) + LDO -1(%r0),%ret0 ; return -1 by default + + CMPB,*= %r0,%arg2,$D3 ; if (d == 0) + COPY %r25,%r5 ; save l (low 64-bits) + + LDO -48(%r30),%r29 ; create ap + .CALL ;in=26,29;out=28; + B,L BN_num_bits_word,%r2 + COPY %r3,%r26 + LDD -288(%r30),%r27 ; restore gp + LDI 64,%r21 + + CMPB,= %r21,%ret0,$00000012 ;if (i == 64) (forward) + COPY %ret0,%r24 ; i + MTSARCM %r24 + DEPDI,Z -1,%sar,1,%r29 + CMPB,*<<,N %r29,%r4,bn_div_err_case ; if (h > 1<= d) + SUB %r4,%r3,%r4 ; h -= d + CMPB,= %r31,%r0,$0000001A ; if (i) + COPY %r0,%r10 ; ret = 0 + MTSARCM %r31 ; i to shift + DEPD,Z %r3,%sar,64,%r3 ; d <<= i; + SUBI 64,%r31,%r19 ; 64 - i; redundent + MTSAR %r19 ; (64 -i) to shift + SHRPD %r4,%r5,%sar,%r4 ; l>> (64-i) + MTSARCM %r31 ; i to shift + DEPD,Z %r5,%sar,64,%r5 ; l <<= i; + +$0000001A + DEPDI,Z -1,31,32,%r19 + EXTRD,U %r3,31,32,%r6 ; dh=(d&0xfff)>>32 + EXTRD,U %r3,63,32,%r8 ; dl = d&0xffffff + LDO 2(%r0),%r9 + STD %r3,-280(%r30) ; "d" to stack + +$0000001C + DEPDI,Z -1,63,32,%r29 ; + EXTRD,U %r4,31,32,%r31 ; h >> 32 + CMPB,*=,N %r31,%r6,$D2 ; if ((h>>32) != dh)(forward) div + COPY %r4,%r26 + EXTRD,U %r4,31,32,%r25 + COPY %r6,%r24 + .CALL ;in=23,24,25,26;out=20,21,22,28,29; (MILLICALL) + B,L $$div2U,%r2 + EXTRD,U %r6,31,32,%r23 + DEPD %r28,31,32,%r29 +$D2 + STD %r29,-272(%r30) ; q + AND %r5,%r19,%r24 ; t & 0xffffffff00000000; + EXTRD,U %r24,31,32,%r24 ; ??? + FLDD -272(%r30),%fr7 ; q + FLDD -280(%r30),%fr8 ; d + XMPYU %fr8L,%fr7L,%fr10 + FSTD %fr10,-256(%r30) + XMPYU %fr8L,%fr7R,%fr22 + FSTD %fr22,-264(%r30) + XMPYU %fr8R,%fr7L,%fr11 + XMPYU %fr8R,%fr7R,%fr23 + FSTD %fr11,-232(%r30) + FSTD %fr23,-240(%r30) + LDD -256(%r30),%r28 + DEPD,Z %r28,31,32,%r2 + LDD -264(%r30),%r20 + ADD,L %r20,%r2,%r31 + LDD -232(%r30),%r22 + DEPD,Z %r22,31,32,%r22 + LDD -240(%r30),%r21 + B $00000024 ; enter loop + ADD,L %r21,%r22,%r23 + +$0000002A + LDO -1(%r29),%r29 + SUB %r23,%r8,%r23 +$00000024 + SUB %r4,%r31,%r25 + AND %r25,%r19,%r26 + CMPB,*<>,N %r0,%r26,$00000046 ; (forward) + DEPD,Z %r25,31,32,%r20 + OR %r20,%r24,%r21 + CMPB,*<<,N %r21,%r23,$0000002A ;(backward) + SUB %r31,%r6,%r31 +;-------------Break path--------------------- + +$00000046 + DEPD,Z %r23,31,32,%r25 ;tl + EXTRD,U %r23,31,32,%r26 ;t + AND %r25,%r19,%r24 ;tl = (tl<<32)&0xfffffff0000000L + ADD,L %r31,%r26,%r31 ;th += t; + CMPCLR,*>>= %r5,%r24,%r0 ;if (l>32)); + DEPD,Z %r29,31,32,%r10 ; ret = q<<32 + b $0000001C + DEPD,Z %r28,31,32,%r5 ; l = l << 32 + +$D1 + OR %r10,%r29,%r28 ; ret |= q +$D3 + LDD -368(%r30),%r2 +$D0 + LDD -296(%r30),%r10 + LDD -304(%r30),%r9 + LDD -312(%r30),%r8 + LDD -320(%r30),%r7 + LDD -328(%r30),%r6 + LDD -336(%r30),%r5 + LDD -344(%r30),%r4 + BVE (%r2) + .EXIT + LDD,MB -352(%r30),%r3 + +bn_div_err_case + MFIA %r6 + ADDIL L'bn_div_words-bn_div_err_case,%r6,%r1 + LDO R'bn_div_words-bn_div_err_case(%r1),%r6 + ADDIL LT'__iob,%r27,%r1 + LDD RT'__iob(%r1),%r26 + ADDIL L'C$4-bn_div_words,%r6,%r1 + LDO R'C$4-bn_div_words(%r1),%r25 + LDO 64(%r26),%r26 + .CALL ;in=24,25,26,29;out=28; + B,L fprintf,%r2 + LDO -48(%r30),%r29 + LDD -288(%r30),%r27 + .CALL ;in=29; + B,L abort,%r2 + LDO -48(%r30),%r29 + LDD -288(%r30),%r27 + B $D0 + LDD -368(%r30),%r2 + .PROCEND ;in=24,25,26,29;out=28; + +;---------------------------------------------------------------------------- +; +; Registers to hold 64-bit values to manipulate. The "L" part +; of the register corresponds to the upper 32-bits, while the "R" +; part corresponds to the lower 32-bits +; +; Note, that when using b6 and b7, the code must save these before +; using them because they are callee save registers +; +; +; Floating point registers to use to save values that +; are manipulated. These don't collide with ftemp1-6 and +; are all caller save registers +; +a0 .reg %fr22 +a0L .reg %fr22L +a0R .reg %fr22R + +a1 .reg %fr23 +a1L .reg %fr23L +a1R .reg %fr23R + +a2 .reg %fr24 +a2L .reg %fr24L +a2R .reg %fr24R + +a3 .reg %fr25 +a3L .reg %fr25L +a3R .reg %fr25R + +a4 .reg %fr26 +a4L .reg %fr26L +a4R .reg %fr26R + +a5 .reg %fr27 +a5L .reg %fr27L +a5R .reg %fr27R + +a6 .reg %fr28 +a6L .reg %fr28L +a6R .reg %fr28R + +a7 .reg %fr29 +a7L .reg %fr29L +a7R .reg %fr29R + +b0 .reg %fr30 +b0L .reg %fr30L +b0R .reg %fr30R + +b1 .reg %fr31 +b1L .reg %fr31L +b1R .reg %fr31R + +; +; Temporary floating point variables, these are all caller save +; registers +; +ftemp1 .reg %fr4 +ftemp2 .reg %fr5 +ftemp3 .reg %fr6 +ftemp4 .reg %fr7 + +; +; The B set of registers when used. +; + +b2 .reg %fr8 +b2L .reg %fr8L +b2R .reg %fr8R + +b3 .reg %fr9 +b3L .reg %fr9L +b3R .reg %fr9R + +b4 .reg %fr10 +b4L .reg %fr10L +b4R .reg %fr10R + +b5 .reg %fr11 +b5L .reg %fr11L +b5R .reg %fr11R + +b6 .reg %fr12 +b6L .reg %fr12L +b6R .reg %fr12R + +b7 .reg %fr13 +b7L .reg %fr13L +b7R .reg %fr13R + +c1 .reg %r21 ; only reg +temp1 .reg %r20 ; only reg +temp2 .reg %r19 ; only reg +temp3 .reg %r31 ; only reg + +m1 .reg %r28 +c2 .reg %r23 +high_one .reg %r1 +ht .reg %r6 +lt .reg %r5 +m .reg %r4 +c3 .reg %r3 + +SQR_ADD_C .macro A0L,A0R,C1,C2,C3 + XMPYU A0L,A0R,ftemp1 ; m + FSTD ftemp1,-24(%sp) ; store m + + XMPYU A0R,A0R,ftemp2 ; lt + FSTD ftemp2,-16(%sp) ; store lt + + XMPYU A0L,A0L,ftemp3 ; ht + FSTD ftemp3,-8(%sp) ; store ht + + LDD -24(%sp),m ; load m + AND m,high_mask,temp2 ; m & Mask + DEPD,Z m,30,31,temp3 ; m << 32+1 + LDD -16(%sp),lt ; lt + + LDD -8(%sp),ht ; ht + EXTRD,U temp2,32,33,temp1 ; temp1 = m&Mask >> 32-1 + ADD temp3,lt,lt ; lt = lt+m + ADD,L ht,temp1,ht ; ht += temp1 + ADD,DC ht,%r0,ht ; ht++ + + ADD C1,lt,C1 ; c1=c1+lt + ADD,DC ht,%r0,ht ; ht++ + + ADD C2,ht,C2 ; c2=c2+ht + ADD,DC C3,%r0,C3 ; c3++ +.endm + +SQR_ADD_C2 .macro A0L,A0R,A1L,A1R,C1,C2,C3 + XMPYU A0L,A1R,ftemp1 ; m1 = bl*ht + FSTD ftemp1,-16(%sp) ; + XMPYU A0R,A1L,ftemp2 ; m = bh*lt + FSTD ftemp2,-8(%sp) ; + XMPYU A0R,A1R,ftemp3 ; lt = bl*lt + FSTD ftemp3,-32(%sp) + XMPYU A0L,A1L,ftemp4 ; ht = bh*ht + FSTD ftemp4,-24(%sp) ; + + LDD -8(%sp),m ; r21 = m + LDD -16(%sp),m1 ; r19 = m1 + ADD,L m,m1,m ; m+m1 + + DEPD,Z m,31,32,temp3 ; (m+m1<<32) + LDD -24(%sp),ht ; r24 = ht + + CMPCLR,*>>= m,m1,%r0 ; if (m < m1) + ADD,L ht,high_one,ht ; ht+=high_one + + EXTRD,U m,31,32,temp1 ; m >> 32 + LDD -32(%sp),lt ; lt + ADD,L ht,temp1,ht ; ht+= m>>32 + ADD lt,temp3,lt ; lt = lt+m1 + ADD,DC ht,%r0,ht ; ht++ + + ADD ht,ht,ht ; ht=ht+ht; + ADD,DC C3,%r0,C3 ; add in carry (c3++) + + ADD lt,lt,lt ; lt=lt+lt; + ADD,DC ht,%r0,ht ; add in carry (ht++) + + ADD C1,lt,C1 ; c1=c1+lt + ADD,DC,*NUV ht,%r0,ht ; add in carry (ht++) + LDO 1(C3),C3 ; bump c3 if overflow,nullify otherwise + + ADD C2,ht,C2 ; c2 = c2 + ht + ADD,DC C3,%r0,C3 ; add in carry (c3++) +.endm + +; +;void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a) +; arg0 = r_ptr +; arg1 = a_ptr +; + +bn_sqr_comba8 + .PROC + .CALLINFO FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_sqr_comba8,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .ENTRY + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + + ; + ; Zero out carries + ; + COPY %r0,c1 + COPY %r0,c2 + COPY %r0,c3 + + LDO 128(%sp),%sp ; bump stack + DEPDI,Z -1,32,33,high_mask ; Create Mask 0xffffffff80000000L + DEPDI,Z 1,31,1,high_one ; Create Value 1 << 32 + + ; + ; Load up all of the values we are going to use + ; + FLDD 0(a_ptr),a0 + FLDD 8(a_ptr),a1 + FLDD 16(a_ptr),a2 + FLDD 24(a_ptr),a3 + FLDD 32(a_ptr),a4 + FLDD 40(a_ptr),a5 + FLDD 48(a_ptr),a6 + FLDD 56(a_ptr),a7 + + SQR_ADD_C a0L,a0R,c1,c2,c3 + STD c1,0(r_ptr) ; r[0] = c1; + COPY %r0,c1 + + SQR_ADD_C2 a1L,a1R,a0L,a0R,c2,c3,c1 + STD c2,8(r_ptr) ; r[1] = c2; + COPY %r0,c2 + + SQR_ADD_C a1L,a1R,c3,c1,c2 + SQR_ADD_C2 a2L,a2R,a0L,a0R,c3,c1,c2 + STD c3,16(r_ptr) ; r[2] = c3; + COPY %r0,c3 + + SQR_ADD_C2 a3L,a3R,a0L,a0R,c1,c2,c3 + SQR_ADD_C2 a2L,a2R,a1L,a1R,c1,c2,c3 + STD c1,24(r_ptr) ; r[3] = c1; + COPY %r0,c1 + + SQR_ADD_C a2L,a2R,c2,c3,c1 + SQR_ADD_C2 a3L,a3R,a1L,a1R,c2,c3,c1 + SQR_ADD_C2 a4L,a4R,a0L,a0R,c2,c3,c1 + STD c2,32(r_ptr) ; r[4] = c2; + COPY %r0,c2 + + SQR_ADD_C2 a5L,a5R,a0L,a0R,c3,c1,c2 + SQR_ADD_C2 a4L,a4R,a1L,a1R,c3,c1,c2 + SQR_ADD_C2 a3L,a3R,a2L,a2R,c3,c1,c2 + STD c3,40(r_ptr) ; r[5] = c3; + COPY %r0,c3 + + SQR_ADD_C a3L,a3R,c1,c2,c3 + SQR_ADD_C2 a4L,a4R,a2L,a2R,c1,c2,c3 + SQR_ADD_C2 a5L,a5R,a1L,a1R,c1,c2,c3 + SQR_ADD_C2 a6L,a6R,a0L,a0R,c1,c2,c3 + STD c1,48(r_ptr) ; r[6] = c1; + COPY %r0,c1 + + SQR_ADD_C2 a7L,a7R,a0L,a0R,c2,c3,c1 + SQR_ADD_C2 a6L,a6R,a1L,a1R,c2,c3,c1 + SQR_ADD_C2 a5L,a5R,a2L,a2R,c2,c3,c1 + SQR_ADD_C2 a4L,a4R,a3L,a3R,c2,c3,c1 + STD c2,56(r_ptr) ; r[7] = c2; + COPY %r0,c2 + + SQR_ADD_C a4L,a4R,c3,c1,c2 + SQR_ADD_C2 a5L,a5R,a3L,a3R,c3,c1,c2 + SQR_ADD_C2 a6L,a6R,a2L,a2R,c3,c1,c2 + SQR_ADD_C2 a7L,a7R,a1L,a1R,c3,c1,c2 + STD c3,64(r_ptr) ; r[8] = c3; + COPY %r0,c3 + + SQR_ADD_C2 a7L,a7R,a2L,a2R,c1,c2,c3 + SQR_ADD_C2 a6L,a6R,a3L,a3R,c1,c2,c3 + SQR_ADD_C2 a5L,a5R,a4L,a4R,c1,c2,c3 + STD c1,72(r_ptr) ; r[9] = c1; + COPY %r0,c1 + + SQR_ADD_C a5L,a5R,c2,c3,c1 + SQR_ADD_C2 a6L,a6R,a4L,a4R,c2,c3,c1 + SQR_ADD_C2 a7L,a7R,a3L,a3R,c2,c3,c1 + STD c2,80(r_ptr) ; r[10] = c2; + COPY %r0,c2 + + SQR_ADD_C2 a7L,a7R,a4L,a4R,c3,c1,c2 + SQR_ADD_C2 a6L,a6R,a5L,a5R,c3,c1,c2 + STD c3,88(r_ptr) ; r[11] = c3; + COPY %r0,c3 + + SQR_ADD_C a6L,a6R,c1,c2,c3 + SQR_ADD_C2 a7L,a7R,a5L,a5R,c1,c2,c3 + STD c1,96(r_ptr) ; r[12] = c1; + COPY %r0,c1 + + SQR_ADD_C2 a7L,a7R,a6L,a6R,c2,c3,c1 + STD c2,104(r_ptr) ; r[13] = c2; + COPY %r0,c2 + + SQR_ADD_C a7L,a7R,c3,c1,c2 + STD c3, 112(r_ptr) ; r[14] = c3 + STD c1, 120(r_ptr) ; r[15] = c1 + + .EXIT + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + + .PROCEND + +;----------------------------------------------------------------------------- +; +;void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a) +; arg0 = r_ptr +; arg1 = a_ptr +; + +bn_sqr_comba4 + .proc + .callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_sqr_comba4,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + + ; + ; Zero out carries + ; + COPY %r0,c1 + COPY %r0,c2 + COPY %r0,c3 + + LDO 128(%sp),%sp ; bump stack + DEPDI,Z -1,32,33,high_mask ; Create Mask 0xffffffff80000000L + DEPDI,Z 1,31,1,high_one ; Create Value 1 << 32 + + ; + ; Load up all of the values we are going to use + ; + FLDD 0(a_ptr),a0 + FLDD 8(a_ptr),a1 + FLDD 16(a_ptr),a2 + FLDD 24(a_ptr),a3 + FLDD 32(a_ptr),a4 + FLDD 40(a_ptr),a5 + FLDD 48(a_ptr),a6 + FLDD 56(a_ptr),a7 + + SQR_ADD_C a0L,a0R,c1,c2,c3 + + STD c1,0(r_ptr) ; r[0] = c1; + COPY %r0,c1 + + SQR_ADD_C2 a1L,a1R,a0L,a0R,c2,c3,c1 + + STD c2,8(r_ptr) ; r[1] = c2; + COPY %r0,c2 + + SQR_ADD_C a1L,a1R,c3,c1,c2 + SQR_ADD_C2 a2L,a2R,a0L,a0R,c3,c1,c2 + + STD c3,16(r_ptr) ; r[2] = c3; + COPY %r0,c3 + + SQR_ADD_C2 a3L,a3R,a0L,a0R,c1,c2,c3 + SQR_ADD_C2 a2L,a2R,a1L,a1R,c1,c2,c3 + + STD c1,24(r_ptr) ; r[3] = c1; + COPY %r0,c1 + + SQR_ADD_C a2L,a2R,c2,c3,c1 + SQR_ADD_C2 a3L,a3R,a1L,a1R,c2,c3,c1 + + STD c2,32(r_ptr) ; r[4] = c2; + COPY %r0,c2 + + SQR_ADD_C2 a3L,a3R,a2L,a2R,c3,c1,c2 + STD c3,40(r_ptr) ; r[5] = c3; + COPY %r0,c3 + + SQR_ADD_C a3L,a3R,c1,c2,c3 + STD c1,48(r_ptr) ; r[6] = c1; + STD c2,56(r_ptr) ; r[7] = c2; + + .EXIT + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + + .PROCEND + + +;--------------------------------------------------------------------------- + +MUL_ADD_C .macro A0L,A0R,B0L,B0R,C1,C2,C3 + XMPYU A0L,B0R,ftemp1 ; m1 = bl*ht + FSTD ftemp1,-16(%sp) ; + XMPYU A0R,B0L,ftemp2 ; m = bh*lt + FSTD ftemp2,-8(%sp) ; + XMPYU A0R,B0R,ftemp3 ; lt = bl*lt + FSTD ftemp3,-32(%sp) + XMPYU A0L,B0L,ftemp4 ; ht = bh*ht + FSTD ftemp4,-24(%sp) ; + + LDD -8(%sp),m ; r21 = m + LDD -16(%sp),m1 ; r19 = m1 + ADD,L m,m1,m ; m+m1 + + DEPD,Z m,31,32,temp3 ; (m+m1<<32) + LDD -24(%sp),ht ; r24 = ht + + CMPCLR,*>>= m,m1,%r0 ; if (m < m1) + ADD,L ht,high_one,ht ; ht+=high_one + + EXTRD,U m,31,32,temp1 ; m >> 32 + LDD -32(%sp),lt ; lt + ADD,L ht,temp1,ht ; ht+= m>>32 + ADD lt,temp3,lt ; lt = lt+m1 + ADD,DC ht,%r0,ht ; ht++ + + ADD C1,lt,C1 ; c1=c1+lt + ADD,DC ht,%r0,ht ; bump c3 if overflow,nullify otherwise + + ADD C2,ht,C2 ; c2 = c2 + ht + ADD,DC C3,%r0,C3 ; add in carry (c3++) +.endm + + +; +;void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b) +; arg0 = r_ptr +; arg1 = a_ptr +; arg2 = b_ptr +; + +bn_mul_comba8 + .proc + .callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_mul_comba8,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + FSTD %fr12,32(%sp) ; save r6 + FSTD %fr13,40(%sp) ; save r7 + + ; + ; Zero out carries + ; + COPY %r0,c1 + COPY %r0,c2 + COPY %r0,c3 + + LDO 128(%sp),%sp ; bump stack + DEPDI,Z 1,31,1,high_one ; Create Value 1 << 32 + + ; + ; Load up all of the values we are going to use + ; + FLDD 0(a_ptr),a0 + FLDD 8(a_ptr),a1 + FLDD 16(a_ptr),a2 + FLDD 24(a_ptr),a3 + FLDD 32(a_ptr),a4 + FLDD 40(a_ptr),a5 + FLDD 48(a_ptr),a6 + FLDD 56(a_ptr),a7 + + FLDD 0(b_ptr),b0 + FLDD 8(b_ptr),b1 + FLDD 16(b_ptr),b2 + FLDD 24(b_ptr),b3 + FLDD 32(b_ptr),b4 + FLDD 40(b_ptr),b5 + FLDD 48(b_ptr),b6 + FLDD 56(b_ptr),b7 + + MUL_ADD_C a0L,a0R,b0L,b0R,c1,c2,c3 + STD c1,0(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a0L,a0R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b0L,b0R,c2,c3,c1 + STD c2,8(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a2L,a2R,b0L,b0R,c3,c1,c2 + MUL_ADD_C a1L,a1R,b1L,b1R,c3,c1,c2 + MUL_ADD_C a0L,a0R,b2L,b2R,c3,c1,c2 + STD c3,16(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a0L,a0R,b3L,b3R,c1,c2,c3 + MUL_ADD_C a1L,a1R,b2L,b2R,c1,c2,c3 + MUL_ADD_C a2L,a2R,b1L,b1R,c1,c2,c3 + MUL_ADD_C a3L,a3R,b0L,b0R,c1,c2,c3 + STD c1,24(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a4L,a4R,b0L,b0R,c2,c3,c1 + MUL_ADD_C a3L,a3R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a2L,a2R,b2L,b2R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b3L,b3R,c2,c3,c1 + MUL_ADD_C a0L,a0R,b4L,b4R,c2,c3,c1 + STD c2,32(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a0L,a0R,b5L,b5R,c3,c1,c2 + MUL_ADD_C a1L,a1R,b4L,b4R,c3,c1,c2 + MUL_ADD_C a2L,a2R,b3L,b3R,c3,c1,c2 + MUL_ADD_C a3L,a3R,b2L,b2R,c3,c1,c2 + MUL_ADD_C a4L,a4R,b1L,b1R,c3,c1,c2 + MUL_ADD_C a5L,a5R,b0L,b0R,c3,c1,c2 + STD c3,40(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a6L,a6R,b0L,b0R,c1,c2,c3 + MUL_ADD_C a5L,a5R,b1L,b1R,c1,c2,c3 + MUL_ADD_C a4L,a4R,b2L,b2R,c1,c2,c3 + MUL_ADD_C a3L,a3R,b3L,b3R,c1,c2,c3 + MUL_ADD_C a2L,a2R,b4L,b4R,c1,c2,c3 + MUL_ADD_C a1L,a1R,b5L,b5R,c1,c2,c3 + MUL_ADD_C a0L,a0R,b6L,b6R,c1,c2,c3 + STD c1,48(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a0L,a0R,b7L,b7R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b6L,b6R,c2,c3,c1 + MUL_ADD_C a2L,a2R,b5L,b5R,c2,c3,c1 + MUL_ADD_C a3L,a3R,b4L,b4R,c2,c3,c1 + MUL_ADD_C a4L,a4R,b3L,b3R,c2,c3,c1 + MUL_ADD_C a5L,a5R,b2L,b2R,c2,c3,c1 + MUL_ADD_C a6L,a6R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a7L,a7R,b0L,b0R,c2,c3,c1 + STD c2,56(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a7L,a7R,b1L,b1R,c3,c1,c2 + MUL_ADD_C a6L,a6R,b2L,b2R,c3,c1,c2 + MUL_ADD_C a5L,a5R,b3L,b3R,c3,c1,c2 + MUL_ADD_C a4L,a4R,b4L,b4R,c3,c1,c2 + MUL_ADD_C a3L,a3R,b5L,b5R,c3,c1,c2 + MUL_ADD_C a2L,a2R,b6L,b6R,c3,c1,c2 + MUL_ADD_C a1L,a1R,b7L,b7R,c3,c1,c2 + STD c3,64(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a2L,a2R,b7L,b7R,c1,c2,c3 + MUL_ADD_C a3L,a3R,b6L,b6R,c1,c2,c3 + MUL_ADD_C a4L,a4R,b5L,b5R,c1,c2,c3 + MUL_ADD_C a5L,a5R,b4L,b4R,c1,c2,c3 + MUL_ADD_C a6L,a6R,b3L,b3R,c1,c2,c3 + MUL_ADD_C a7L,a7R,b2L,b2R,c1,c2,c3 + STD c1,72(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a7L,a7R,b3L,b3R,c2,c3,c1 + MUL_ADD_C a6L,a6R,b4L,b4R,c2,c3,c1 + MUL_ADD_C a5L,a5R,b5L,b5R,c2,c3,c1 + MUL_ADD_C a4L,a4R,b6L,b6R,c2,c3,c1 + MUL_ADD_C a3L,a3R,b7L,b7R,c2,c3,c1 + STD c2,80(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a4L,a4R,b7L,b7R,c3,c1,c2 + MUL_ADD_C a5L,a5R,b6L,b6R,c3,c1,c2 + MUL_ADD_C a6L,a6R,b5L,b5R,c3,c1,c2 + MUL_ADD_C a7L,a7R,b4L,b4R,c3,c1,c2 + STD c3,88(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a7L,a7R,b5L,b5R,c1,c2,c3 + MUL_ADD_C a6L,a6R,b6L,b6R,c1,c2,c3 + MUL_ADD_C a5L,a5R,b7L,b7R,c1,c2,c3 + STD c1,96(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a6L,a6R,b7L,b7R,c2,c3,c1 + MUL_ADD_C a7L,a7R,b6L,b6R,c2,c3,c1 + STD c2,104(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a7L,a7R,b7L,b7R,c3,c1,c2 + STD c3,112(r_ptr) + STD c1,120(r_ptr) + + .EXIT + FLDD -88(%sp),%fr13 + FLDD -96(%sp),%fr12 + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + + .PROCEND + +;----------------------------------------------------------------------------- +; +;void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b) +; arg0 = r_ptr +; arg1 = a_ptr +; arg2 = b_ptr +; + +bn_mul_comba4 + .proc + .callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE + .EXPORT bn_mul_comba4,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN + .entry + .align 64 + + STD %r3,0(%sp) ; save r3 + STD %r4,8(%sp) ; save r4 + STD %r5,16(%sp) ; save r5 + STD %r6,24(%sp) ; save r6 + FSTD %fr12,32(%sp) ; save r6 + FSTD %fr13,40(%sp) ; save r7 + + ; + ; Zero out carries + ; + COPY %r0,c1 + COPY %r0,c2 + COPY %r0,c3 + + LDO 128(%sp),%sp ; bump stack + DEPDI,Z 1,31,1,high_one ; Create Value 1 << 32 + + ; + ; Load up all of the values we are going to use + ; + FLDD 0(a_ptr),a0 + FLDD 8(a_ptr),a1 + FLDD 16(a_ptr),a2 + FLDD 24(a_ptr),a3 + + FLDD 0(b_ptr),b0 + FLDD 8(b_ptr),b1 + FLDD 16(b_ptr),b2 + FLDD 24(b_ptr),b3 + + MUL_ADD_C a0L,a0R,b0L,b0R,c1,c2,c3 + STD c1,0(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a0L,a0R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b0L,b0R,c2,c3,c1 + STD c2,8(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a2L,a2R,b0L,b0R,c3,c1,c2 + MUL_ADD_C a1L,a1R,b1L,b1R,c3,c1,c2 + MUL_ADD_C a0L,a0R,b2L,b2R,c3,c1,c2 + STD c3,16(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a0L,a0R,b3L,b3R,c1,c2,c3 + MUL_ADD_C a1L,a1R,b2L,b2R,c1,c2,c3 + MUL_ADD_C a2L,a2R,b1L,b1R,c1,c2,c3 + MUL_ADD_C a3L,a3R,b0L,b0R,c1,c2,c3 + STD c1,24(r_ptr) + COPY %r0,c1 + + MUL_ADD_C a3L,a3R,b1L,b1R,c2,c3,c1 + MUL_ADD_C a2L,a2R,b2L,b2R,c2,c3,c1 + MUL_ADD_C a1L,a1R,b3L,b3R,c2,c3,c1 + STD c2,32(r_ptr) + COPY %r0,c2 + + MUL_ADD_C a2L,a2R,b3L,b3R,c3,c1,c2 + MUL_ADD_C a3L,a3R,b2L,b2R,c3,c1,c2 + STD c3,40(r_ptr) + COPY %r0,c3 + + MUL_ADD_C a3L,a3R,b3L,b3R,c1,c2,c3 + STD c1,48(r_ptr) + STD c2,56(r_ptr) + + .EXIT + FLDD -88(%sp),%fr13 + FLDD -96(%sp),%fr12 + LDD -104(%sp),%r6 ; restore r6 + LDD -112(%sp),%r5 ; restore r5 + LDD -120(%sp),%r4 ; restore r4 + BVE (%rp) + LDD,MB -128(%sp),%r3 + + .PROCEND + + + .SPACE $TEXT$ + .SUBSPA $CODE$ + .SPACE $PRIVATE$,SORT=16 + .IMPORT $global$,DATA + .SPACE $TEXT$ + .SUBSPA $CODE$ + .SUBSPA $LIT$,ACCESS=0x2c +C$4 + .ALIGN 8 + .STRINGZ "Division would overflow (%d)\n" + .END diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/parisc-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/parisc-mont.pl new file mode 100644 index 00000000..c02ef6f0 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/parisc-mont.pl @@ -0,0 +1,995 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# On PA-7100LC this module performs ~90-50% better, less for longer +# keys, than code generated by gcc 3.2 for PA-RISC 1.1. Latter means +# that compiler utilized xmpyu instruction to perform 32x32=64-bit +# multiplication, which in turn means that "baseline" performance was +# optimal in respect to instruction set capabilities. Fair comparison +# with vendor compiler is problematic, because OpenSSL doesn't define +# BN_LLONG [presumably] for historical reasons, which drives compiler +# toward 4 times 16x16=32-bit multiplicatons [plus complementary +# shifts and additions] instead. This means that you should observe +# several times improvement over code generated by vendor compiler +# for PA-RISC 1.1, but the "baseline" is far from optimal. The actual +# improvement coefficient was never collected on PA-7100LC, or any +# other 1.1 CPU, because I don't have access to such machine with +# vendor compiler. But to give you a taste, PA-RISC 1.1 code path +# reportedly outperformed code generated by cc +DA1.1 +O3 by factor +# of ~5x on PA-8600. +# +# On PA-RISC 2.0 it has to compete with pa-risc2[W].s, which is +# reportedly ~2x faster than vendor compiler generated code [according +# to comment in pa-risc2[W].s]. Here comes a catch. Execution core of +# this implementation is actually 32-bit one, in the sense that it +# operates on 32-bit values. But pa-risc2[W].s operates on arrays of +# 64-bit BN_LONGs... How do they interoperate then? No problem. This +# module picks halves of 64-bit values in reverse order and pretends +# they were 32-bit BN_LONGs. But can 32-bit core compete with "pure" +# 64-bit code such as pa-risc2[W].s then? Well, the thing is that +# 32x32=64-bit multiplication is the best even PA-RISC 2.0 can do, +# i.e. there is no "wider" multiplication like on most other 64-bit +# platforms. This means that even being effectively 32-bit, this +# implementation performs "64-bit" computational task in same amount +# of arithmetic operations, most notably multiplications. It requires +# more memory references, most notably to tp[num], but this doesn't +# seem to exhaust memory port capacity. And indeed, dedicated PA-RISC +# 2.0 code path provides virtually same performance as pa-risc2[W].s: +# it's ~10% better for shortest key length and ~10% worse for longest +# one. +# +# In case it wasn't clear. The module has two distinct code paths: +# PA-RISC 1.1 and PA-RISC 2.0 ones. Latter features carry-free 64-bit +# additions and 64-bit integer loads, not to mention specific +# instruction scheduling. In 64-bit build naturally only 2.0 code path +# is assembled. In 32-bit application context both code paths are +# assembled, PA-RISC 2.0 CPU is detected at run-time and proper path +# is taken automatically. Also, in 32-bit build the module imposes +# couple of limitations: vector lengths has to be even and vector +# addresses has to be 64-bit aligned. Normally neither is a problem: +# most common key lengths are even and vectors are commonly malloc-ed, +# which ensures alignment. +# +# Special thanks to polarhome.com for providing HP-UX account on +# PA-RISC 1.1 machine, and to correspondent who chose to remain +# anonymous for testing the code on PA-RISC 2.0 machine. + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; + +$flavour = shift; +$output = shift; + +open STDOUT,">$output"; + +if ($flavour =~ /64/) { + $LEVEL ="2.0W"; + $SIZE_T =8; + $FRAME_MARKER =80; + $SAVED_RP =16; + $PUSH ="std"; + $PUSHMA ="std,ma"; + $POP ="ldd"; + $POPMB ="ldd,mb"; + $BN_SZ =$SIZE_T; +} else { + $LEVEL ="1.1"; #$LEVEL.="\n\t.ALLOW\t2.0"; + $SIZE_T =4; + $FRAME_MARKER =48; + $SAVED_RP =20; + $PUSH ="stw"; + $PUSHMA ="stwm"; + $POP ="ldw"; + $POPMB ="ldwm"; + $BN_SZ =$SIZE_T; + if (open CONF,"<${dir}../../opensslconf.h") { + while() { + if (m/#\s*define\s+SIXTY_FOUR_BIT/) { + $BN_SZ=8; + $LEVEL="2.0"; + last; + } + } + close CONF; + } +} + +$FRAME=8*$SIZE_T+$FRAME_MARKER; # 8 saved regs + frame marker + # [+ argument transfer] +$LOCALS=$FRAME-$FRAME_MARKER; +$FRAME+=32; # local variables + +$tp="%r31"; +$ti1="%r29"; +$ti0="%r28"; + +$rp="%r26"; +$ap="%r25"; +$bp="%r24"; +$np="%r23"; +$n0="%r22"; # passed through stack in 32-bit +$num="%r21"; # passed through stack in 32-bit +$idx="%r20"; +$arrsz="%r19"; + +$nm1="%r7"; +$nm0="%r6"; +$ab1="%r5"; +$ab0="%r4"; + +$fp="%r3"; +$hi1="%r2"; +$hi0="%r1"; + +$xfer=$n0; # accomodates [-16..15] offset in fld[dw]s + +$fm0="%fr4"; $fti=$fm0; +$fbi="%fr5L"; +$fn0="%fr5R"; +$fai="%fr6"; $fab0="%fr7"; $fab1="%fr8"; +$fni="%fr9"; $fnm0="%fr10"; $fnm1="%fr11"; + +$code=<<___; + .LEVEL $LEVEL + .SPACE \$TEXT\$ + .SUBSPA \$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY + + .EXPORT bn_mul_mont,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR + .ALIGN 64 +bn_mul_mont + .PROC + .CALLINFO FRAME=`$FRAME-8*$SIZE_T`,NO_CALLS,SAVE_RP,SAVE_SP,ENTRY_GR=6 + .ENTRY + $PUSH %r2,-$SAVED_RP(%sp) ; standard prologue + $PUSHMA %r3,$FRAME(%sp) + $PUSH %r4,`-$FRAME+1*$SIZE_T`(%sp) + $PUSH %r5,`-$FRAME+2*$SIZE_T`(%sp) + $PUSH %r6,`-$FRAME+3*$SIZE_T`(%sp) + $PUSH %r7,`-$FRAME+4*$SIZE_T`(%sp) + $PUSH %r8,`-$FRAME+5*$SIZE_T`(%sp) + $PUSH %r9,`-$FRAME+6*$SIZE_T`(%sp) + $PUSH %r10,`-$FRAME+7*$SIZE_T`(%sp) + ldo -$FRAME(%sp),$fp +___ +$code.=<<___ if ($SIZE_T==4); + ldw `-$FRAME_MARKER-4`($fp),$n0 + ldw `-$FRAME_MARKER-8`($fp),$num + nop + nop ; alignment +___ +$code.=<<___ if ($BN_SZ==4); + comiclr,<= 6,$num,%r0 ; are vectors long enough? + b L\$abort + ldi 0,%r28 ; signal "unhandled" + add,ev %r0,$num,$num ; is $num even? + b L\$abort + nop + or $ap,$np,$ti1 + extru,= $ti1,31,3,%r0 ; are ap and np 64-bit aligned? + b L\$abort + nop + nop ; alignment + nop + + fldws 0($n0),${fn0} + fldws,ma 4($bp),${fbi} ; bp[0] +___ +$code.=<<___ if ($BN_SZ==8); + comib,> 3,$num,L\$abort ; are vectors long enough? + ldi 0,%r28 ; signal "unhandled" + addl $num,$num,$num ; I operate on 32-bit values + + fldws 4($n0),${fn0} ; only low part of n0 + fldws 4($bp),${fbi} ; bp[0] in flipped word order +___ +$code.=<<___; + fldds 0($ap),${fai} ; ap[0,1] + fldds 0($np),${fni} ; np[0,1] + + sh2addl $num,%r0,$arrsz + ldi 31,$hi0 + ldo 36($arrsz),$hi1 ; space for tp[num+1] + andcm $hi1,$hi0,$hi1 ; align + addl $hi1,%sp,%sp + $PUSH $fp,-$SIZE_T(%sp) + + ldo `$LOCALS+16`($fp),$xfer + ldo `$LOCALS+32+4`($fp),$tp + + xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[0] + xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[0] + xmpyu ${fn0},${fab0}R,${fm0} + + addl $arrsz,$ap,$ap ; point at the end + addl $arrsz,$np,$np + subi 0,$arrsz,$idx ; j=0 + ldo 8($idx),$idx ; j++++ + + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m + fstds ${fab0},-16($xfer) + fstds ${fnm0},-8($xfer) + fstds ${fab1},0($xfer) + fstds ${fnm1},8($xfer) + flddx $idx($ap),${fai} ; ap[2,3] + flddx $idx($np),${fni} ; np[2,3] +___ +$code.=<<___ if ($BN_SZ==4); + mtctl $hi0,%cr11 ; $hi0 still holds 31 + extrd,u,*= $hi0,%sar,1,$hi0 ; executes on PA-RISC 1.0 + b L\$parisc11 + nop +___ +$code.=<<___; # PA-RISC 2.0 code-path + xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0] + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m + ldd -16($xfer),$ab0 + fstds ${fab0},-16($xfer) + + extrd,u $ab0,31,32,$hi0 + extrd,u $ab0,63,32,$ab0 + ldd -8($xfer),$nm0 + fstds ${fnm0},-8($xfer) + ldo 8($idx),$idx ; j++++ + addl $ab0,$nm0,$nm0 ; low part is discarded + extrd,u $nm0,31,32,$hi1 + +L\$1st + xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[0] + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m + ldd 0($xfer),$ab1 + fstds ${fab1},0($xfer) + addl $hi0,$ab1,$ab1 + extrd,u $ab1,31,32,$hi0 + ldd 8($xfer),$nm1 + fstds ${fnm1},8($xfer) + extrd,u $ab1,63,32,$ab1 + addl $hi1,$nm1,$nm1 + flddx $idx($ap),${fai} ; ap[j,j+1] + flddx $idx($np),${fni} ; np[j,j+1] + addl $ab1,$nm1,$nm1 + extrd,u $nm1,31,32,$hi1 + + xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0] + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m + ldd -16($xfer),$ab0 + fstds ${fab0},-16($xfer) + addl $hi0,$ab0,$ab0 + extrd,u $ab0,31,32,$hi0 + ldd -8($xfer),$nm0 + fstds ${fnm0},-8($xfer) + extrd,u $ab0,63,32,$ab0 + addl $hi1,$nm0,$nm0 + stw $nm1,-4($tp) ; tp[j-1] + addl $ab0,$nm0,$nm0 + stw,ma $nm0,8($tp) ; tp[j-1] + addib,<> 8,$idx,L\$1st ; j++++ + extrd,u $nm0,31,32,$hi1 + + xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[0] + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m + ldd 0($xfer),$ab1 + fstds ${fab1},0($xfer) + addl $hi0,$ab1,$ab1 + extrd,u $ab1,31,32,$hi0 + ldd 8($xfer),$nm1 + fstds ${fnm1},8($xfer) + extrd,u $ab1,63,32,$ab1 + addl $hi1,$nm1,$nm1 + ldd -16($xfer),$ab0 + addl $ab1,$nm1,$nm1 + ldd -8($xfer),$nm0 + extrd,u $nm1,31,32,$hi1 + + addl $hi0,$ab0,$ab0 + extrd,u $ab0,31,32,$hi0 + stw $nm1,-4($tp) ; tp[j-1] + extrd,u $ab0,63,32,$ab0 + addl $hi1,$nm0,$nm0 + ldd 0($xfer),$ab1 + addl $ab0,$nm0,$nm0 + ldd,mb 8($xfer),$nm1 + extrd,u $nm0,31,32,$hi1 + stw,ma $nm0,8($tp) ; tp[j-1] + + ldo -1($num),$num ; i-- + subi 0,$arrsz,$idx ; j=0 +___ +$code.=<<___ if ($BN_SZ==4); + fldws,ma 4($bp),${fbi} ; bp[1] +___ +$code.=<<___ if ($BN_SZ==8); + fldws 0($bp),${fbi} ; bp[1] in flipped word order +___ +$code.=<<___; + flddx $idx($ap),${fai} ; ap[0,1] + flddx $idx($np),${fni} ; np[0,1] + fldws 8($xfer),${fti}R ; tp[0] + addl $hi0,$ab1,$ab1 + extrd,u $ab1,31,32,$hi0 + extrd,u $ab1,63,32,$ab1 + ldo 8($idx),$idx ; j++++ + xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[1] + xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[1] + addl $hi1,$nm1,$nm1 + addl $ab1,$nm1,$nm1 + extrd,u $nm1,31,32,$hi1 + fstws,mb ${fab0}L,-8($xfer) ; save high part + stw $nm1,-4($tp) ; tp[j-1] + + fcpy,sgl %fr0,${fti}L ; zero high part + fcpy,sgl %fr0,${fab0}L + addl $hi1,$hi0,$hi0 + extrd,u $hi0,31,32,$hi1 + fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double + fcnvxf,dbl,dbl ${fab0},${fab0} + stw $hi0,0($tp) + stw $hi1,4($tp) + + fadd,dbl ${fti},${fab0},${fab0} ; add tp[0] + fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int + xmpyu ${fn0},${fab0}R,${fm0} + ldo `$LOCALS+32+4`($fp),$tp +L\$outer + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m + fstds ${fab0},-16($xfer) ; 33-bit value + fstds ${fnm0},-8($xfer) + flddx $idx($ap),${fai} ; ap[2] + flddx $idx($np),${fni} ; np[2] + ldo 8($idx),$idx ; j++++ + ldd -16($xfer),$ab0 ; 33-bit value + ldd -8($xfer),$nm0 + ldw 0($xfer),$hi0 ; high part + + xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i] + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m + extrd,u $ab0,31,32,$ti0 ; carry bit + extrd,u $ab0,63,32,$ab0 + fstds ${fab1},0($xfer) + addl $ti0,$hi0,$hi0 ; account carry bit + fstds ${fnm1},8($xfer) + addl $ab0,$nm0,$nm0 ; low part is discarded + ldw 0($tp),$ti1 ; tp[1] + extrd,u $nm0,31,32,$hi1 + fstds ${fab0},-16($xfer) + fstds ${fnm0},-8($xfer) + +L\$inner + xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[i] + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m + ldd 0($xfer),$ab1 + fstds ${fab1},0($xfer) + addl $hi0,$ti1,$ti1 + addl $ti1,$ab1,$ab1 + ldd 8($xfer),$nm1 + fstds ${fnm1},8($xfer) + extrd,u $ab1,31,32,$hi0 + extrd,u $ab1,63,32,$ab1 + flddx $idx($ap),${fai} ; ap[j,j+1] + flddx $idx($np),${fni} ; np[j,j+1] + addl $hi1,$nm1,$nm1 + addl $ab1,$nm1,$nm1 + ldw 4($tp),$ti0 ; tp[j] + stw $nm1,-4($tp) ; tp[j-1] + + xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i] + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m + ldd -16($xfer),$ab0 + fstds ${fab0},-16($xfer) + addl $hi0,$ti0,$ti0 + addl $ti0,$ab0,$ab0 + ldd -8($xfer),$nm0 + fstds ${fnm0},-8($xfer) + extrd,u $ab0,31,32,$hi0 + extrd,u $nm1,31,32,$hi1 + ldw 8($tp),$ti1 ; tp[j] + extrd,u $ab0,63,32,$ab0 + addl $hi1,$nm0,$nm0 + addl $ab0,$nm0,$nm0 + stw,ma $nm0,8($tp) ; tp[j-1] + addib,<> 8,$idx,L\$inner ; j++++ + extrd,u $nm0,31,32,$hi1 + + xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[i] + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m + ldd 0($xfer),$ab1 + fstds ${fab1},0($xfer) + addl $hi0,$ti1,$ti1 + addl $ti1,$ab1,$ab1 + ldd 8($xfer),$nm1 + fstds ${fnm1},8($xfer) + extrd,u $ab1,31,32,$hi0 + extrd,u $ab1,63,32,$ab1 + ldw 4($tp),$ti0 ; tp[j] + addl $hi1,$nm1,$nm1 + addl $ab1,$nm1,$nm1 + ldd -16($xfer),$ab0 + ldd -8($xfer),$nm0 + extrd,u $nm1,31,32,$hi1 + + addl $hi0,$ab0,$ab0 + addl $ti0,$ab0,$ab0 + stw $nm1,-4($tp) ; tp[j-1] + extrd,u $ab0,31,32,$hi0 + ldw 8($tp),$ti1 ; tp[j] + extrd,u $ab0,63,32,$ab0 + addl $hi1,$nm0,$nm0 + ldd 0($xfer),$ab1 + addl $ab0,$nm0,$nm0 + ldd,mb 8($xfer),$nm1 + extrd,u $nm0,31,32,$hi1 + stw,ma $nm0,8($tp) ; tp[j-1] + + addib,= -1,$num,L\$outerdone ; i-- + subi 0,$arrsz,$idx ; j=0 +___ +$code.=<<___ if ($BN_SZ==4); + fldws,ma 4($bp),${fbi} ; bp[i] +___ +$code.=<<___ if ($BN_SZ==8); + ldi 12,$ti0 ; bp[i] in flipped word order + addl,ev %r0,$num,$num + ldi -4,$ti0 + addl $ti0,$bp,$bp + fldws 0($bp),${fbi} +___ +$code.=<<___; + flddx $idx($ap),${fai} ; ap[0] + addl $hi0,$ab1,$ab1 + flddx $idx($np),${fni} ; np[0] + fldws 8($xfer),${fti}R ; tp[0] + addl $ti1,$ab1,$ab1 + extrd,u $ab1,31,32,$hi0 + extrd,u $ab1,63,32,$ab1 + + ldo 8($idx),$idx ; j++++ + xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[i] + xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[i] + ldw 4($tp),$ti0 ; tp[j] + + addl $hi1,$nm1,$nm1 + fstws,mb ${fab0}L,-8($xfer) ; save high part + addl $ab1,$nm1,$nm1 + extrd,u $nm1,31,32,$hi1 + fcpy,sgl %fr0,${fti}L ; zero high part + fcpy,sgl %fr0,${fab0}L + stw $nm1,-4($tp) ; tp[j-1] + + fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double + fcnvxf,dbl,dbl ${fab0},${fab0} + addl $hi1,$hi0,$hi0 + fadd,dbl ${fti},${fab0},${fab0} ; add tp[0] + addl $ti0,$hi0,$hi0 + extrd,u $hi0,31,32,$hi1 + fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int + stw $hi0,0($tp) + stw $hi1,4($tp) + xmpyu ${fn0},${fab0}R,${fm0} + + b L\$outer + ldo `$LOCALS+32+4`($fp),$tp + +L\$outerdone + addl $hi0,$ab1,$ab1 + addl $ti1,$ab1,$ab1 + extrd,u $ab1,31,32,$hi0 + extrd,u $ab1,63,32,$ab1 + + ldw 4($tp),$ti0 ; tp[j] + + addl $hi1,$nm1,$nm1 + addl $ab1,$nm1,$nm1 + extrd,u $nm1,31,32,$hi1 + stw $nm1,-4($tp) ; tp[j-1] + + addl $hi1,$hi0,$hi0 + addl $ti0,$hi0,$hi0 + extrd,u $hi0,31,32,$hi1 + stw $hi0,0($tp) + stw $hi1,4($tp) + + ldo `$LOCALS+32`($fp),$tp + sub %r0,%r0,%r0 ; clear borrow +___ +$code.=<<___ if ($BN_SZ==4); + ldws,ma 4($tp),$ti0 + extru,= $rp,31,3,%r0 ; is rp 64-bit aligned? + b L\$sub_pa11 + addl $tp,$arrsz,$tp +L\$sub + ldwx $idx($np),$hi0 + subb $ti0,$hi0,$hi1 + ldwx $idx($tp),$ti0 + addib,<> 4,$idx,L\$sub + stws,ma $hi1,4($rp) + + subb $ti0,%r0,$hi1 + ldo -4($tp),$tp +___ +$code.=<<___ if ($BN_SZ==8); + ldd,ma 8($tp),$ti0 +L\$sub + ldd $idx($np),$hi0 + shrpd $ti0,$ti0,32,$ti0 ; flip word order + std $ti0,-8($tp) ; save flipped value + sub,db $ti0,$hi0,$hi1 + ldd,ma 8($tp),$ti0 + addib,<> 8,$idx,L\$sub + std,ma $hi1,8($rp) + + extrd,u $ti0,31,32,$ti0 ; carry in flipped word order + sub,db $ti0,%r0,$hi1 + ldo -8($tp),$tp +___ +$code.=<<___; + and $tp,$hi1,$ap + andcm $rp,$hi1,$bp + or $ap,$bp,$np + + sub $rp,$arrsz,$rp ; rewind rp + subi 0,$arrsz,$idx + ldo `$LOCALS+32`($fp),$tp +L\$copy + ldd $idx($np),$hi0 + std,ma %r0,8($tp) + addib,<> 8,$idx,.-8 ; L\$copy + std,ma $hi0,8($rp) +___ + +if ($BN_SZ==4) { # PA-RISC 1.1 code-path +$ablo=$ab0; +$abhi=$ab1; +$nmlo0=$nm0; +$nmhi0=$nm1; +$nmlo1="%r9"; +$nmhi1="%r8"; + +$code.=<<___; + b L\$done + nop + + .ALIGN 8 +L\$parisc11 + xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0] + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m + ldw -12($xfer),$ablo + ldw -16($xfer),$hi0 + ldw -4($xfer),$nmlo0 + ldw -8($xfer),$nmhi0 + fstds ${fab0},-16($xfer) + fstds ${fnm0},-8($xfer) + + ldo 8($idx),$idx ; j++++ + add $ablo,$nmlo0,$nmlo0 ; discarded + addc %r0,$nmhi0,$hi1 + ldw 4($xfer),$ablo + ldw 0($xfer),$abhi + nop + +L\$1st_pa11 + xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[0] + flddx $idx($ap),${fai} ; ap[j,j+1] + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m + flddx $idx($np),${fni} ; np[j,j+1] + add $hi0,$ablo,$ablo + ldw 12($xfer),$nmlo1 + addc %r0,$abhi,$hi0 + ldw 8($xfer),$nmhi1 + add $ablo,$nmlo1,$nmlo1 + fstds ${fab1},0($xfer) + addc %r0,$nmhi1,$nmhi1 + fstds ${fnm1},8($xfer) + add $hi1,$nmlo1,$nmlo1 + ldw -12($xfer),$ablo + addc %r0,$nmhi1,$hi1 + ldw -16($xfer),$abhi + + xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0] + ldw -4($xfer),$nmlo0 + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m + ldw -8($xfer),$nmhi0 + add $hi0,$ablo,$ablo + stw $nmlo1,-4($tp) ; tp[j-1] + addc %r0,$abhi,$hi0 + fstds ${fab0},-16($xfer) + add $ablo,$nmlo0,$nmlo0 + fstds ${fnm0},-8($xfer) + addc %r0,$nmhi0,$nmhi0 + ldw 0($xfer),$abhi + add $hi1,$nmlo0,$nmlo0 + ldw 4($xfer),$ablo + stws,ma $nmlo0,8($tp) ; tp[j-1] + addib,<> 8,$idx,L\$1st_pa11 ; j++++ + addc %r0,$nmhi0,$hi1 + + ldw 8($xfer),$nmhi1 + ldw 12($xfer),$nmlo1 + xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[0] + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m + add $hi0,$ablo,$ablo + fstds ${fab1},0($xfer) + addc %r0,$abhi,$hi0 + fstds ${fnm1},8($xfer) + add $ablo,$nmlo1,$nmlo1 + ldw -16($xfer),$abhi + addc %r0,$nmhi1,$nmhi1 + ldw -12($xfer),$ablo + add $hi1,$nmlo1,$nmlo1 + ldw -8($xfer),$nmhi0 + addc %r0,$nmhi1,$hi1 + ldw -4($xfer),$nmlo0 + + add $hi0,$ablo,$ablo + stw $nmlo1,-4($tp) ; tp[j-1] + addc %r0,$abhi,$hi0 + ldw 0($xfer),$abhi + add $ablo,$nmlo0,$nmlo0 + ldw 4($xfer),$ablo + addc %r0,$nmhi0,$nmhi0 + ldws,mb 8($xfer),$nmhi1 + add $hi1,$nmlo0,$nmlo0 + ldw 4($xfer),$nmlo1 + addc %r0,$nmhi0,$hi1 + stws,ma $nmlo0,8($tp) ; tp[j-1] + + ldo -1($num),$num ; i-- + subi 0,$arrsz,$idx ; j=0 + + fldws,ma 4($bp),${fbi} ; bp[1] + flddx $idx($ap),${fai} ; ap[0,1] + flddx $idx($np),${fni} ; np[0,1] + fldws 8($xfer),${fti}R ; tp[0] + add $hi0,$ablo,$ablo + addc %r0,$abhi,$hi0 + ldo 8($idx),$idx ; j++++ + xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[1] + xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[1] + add $hi1,$nmlo1,$nmlo1 + addc %r0,$nmhi1,$nmhi1 + add $ablo,$nmlo1,$nmlo1 + addc %r0,$nmhi1,$hi1 + fstws,mb ${fab0}L,-8($xfer) ; save high part + stw $nmlo1,-4($tp) ; tp[j-1] + + fcpy,sgl %fr0,${fti}L ; zero high part + fcpy,sgl %fr0,${fab0}L + add $hi1,$hi0,$hi0 + addc %r0,%r0,$hi1 + fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double + fcnvxf,dbl,dbl ${fab0},${fab0} + stw $hi0,0($tp) + stw $hi1,4($tp) + + fadd,dbl ${fti},${fab0},${fab0} ; add tp[0] + fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int + xmpyu ${fn0},${fab0}R,${fm0} + ldo `$LOCALS+32+4`($fp),$tp +L\$outer_pa11 + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m + fstds ${fab0},-16($xfer) ; 33-bit value + fstds ${fnm0},-8($xfer) + flddx $idx($ap),${fai} ; ap[2,3] + flddx $idx($np),${fni} ; np[2,3] + ldw -16($xfer),$abhi ; carry bit actually + ldo 8($idx),$idx ; j++++ + ldw -12($xfer),$ablo + ldw -8($xfer),$nmhi0 + ldw -4($xfer),$nmlo0 + ldw 0($xfer),$hi0 ; high part + + xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i] + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m + fstds ${fab1},0($xfer) + addl $abhi,$hi0,$hi0 ; account carry bit + fstds ${fnm1},8($xfer) + add $ablo,$nmlo0,$nmlo0 ; discarded + ldw 0($tp),$ti1 ; tp[1] + addc %r0,$nmhi0,$hi1 + fstds ${fab0},-16($xfer) + fstds ${fnm0},-8($xfer) + ldw 4($xfer),$ablo + ldw 0($xfer),$abhi + +L\$inner_pa11 + xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[i] + flddx $idx($ap),${fai} ; ap[j,j+1] + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m + flddx $idx($np),${fni} ; np[j,j+1] + add $hi0,$ablo,$ablo + ldw 4($tp),$ti0 ; tp[j] + addc %r0,$abhi,$abhi + ldw 12($xfer),$nmlo1 + add $ti1,$ablo,$ablo + ldw 8($xfer),$nmhi1 + addc %r0,$abhi,$hi0 + fstds ${fab1},0($xfer) + add $ablo,$nmlo1,$nmlo1 + fstds ${fnm1},8($xfer) + addc %r0,$nmhi1,$nmhi1 + ldw -12($xfer),$ablo + add $hi1,$nmlo1,$nmlo1 + ldw -16($xfer),$abhi + addc %r0,$nmhi1,$hi1 + + xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i] + ldw 8($tp),$ti1 ; tp[j] + xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m + ldw -4($xfer),$nmlo0 + add $hi0,$ablo,$ablo + ldw -8($xfer),$nmhi0 + addc %r0,$abhi,$abhi + stw $nmlo1,-4($tp) ; tp[j-1] + add $ti0,$ablo,$ablo + fstds ${fab0},-16($xfer) + addc %r0,$abhi,$hi0 + fstds ${fnm0},-8($xfer) + add $ablo,$nmlo0,$nmlo0 + ldw 4($xfer),$ablo + addc %r0,$nmhi0,$nmhi0 + ldw 0($xfer),$abhi + add $hi1,$nmlo0,$nmlo0 + stws,ma $nmlo0,8($tp) ; tp[j-1] + addib,<> 8,$idx,L\$inner_pa11 ; j++++ + addc %r0,$nmhi0,$hi1 + + xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[i] + ldw 12($xfer),$nmlo1 + xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m + ldw 8($xfer),$nmhi1 + add $hi0,$ablo,$ablo + ldw 4($tp),$ti0 ; tp[j] + addc %r0,$abhi,$abhi + fstds ${fab1},0($xfer) + add $ti1,$ablo,$ablo + fstds ${fnm1},8($xfer) + addc %r0,$abhi,$hi0 + ldw -16($xfer),$abhi + add $ablo,$nmlo1,$nmlo1 + ldw -12($xfer),$ablo + addc %r0,$nmhi1,$nmhi1 + ldw -8($xfer),$nmhi0 + add $hi1,$nmlo1,$nmlo1 + ldw -4($xfer),$nmlo0 + addc %r0,$nmhi1,$hi1 + + add $hi0,$ablo,$ablo + stw $nmlo1,-4($tp) ; tp[j-1] + addc %r0,$abhi,$abhi + add $ti0,$ablo,$ablo + ldw 8($tp),$ti1 ; tp[j] + addc %r0,$abhi,$hi0 + ldw 0($xfer),$abhi + add $ablo,$nmlo0,$nmlo0 + ldw 4($xfer),$ablo + addc %r0,$nmhi0,$nmhi0 + ldws,mb 8($xfer),$nmhi1 + add $hi1,$nmlo0,$nmlo0 + ldw 4($xfer),$nmlo1 + addc %r0,$nmhi0,$hi1 + stws,ma $nmlo0,8($tp) ; tp[j-1] + + addib,= -1,$num,L\$outerdone_pa11; i-- + subi 0,$arrsz,$idx ; j=0 + + fldws,ma 4($bp),${fbi} ; bp[i] + flddx $idx($ap),${fai} ; ap[0] + add $hi0,$ablo,$ablo + addc %r0,$abhi,$abhi + flddx $idx($np),${fni} ; np[0] + fldws 8($xfer),${fti}R ; tp[0] + add $ti1,$ablo,$ablo + addc %r0,$abhi,$hi0 + + ldo 8($idx),$idx ; j++++ + xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[i] + xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[i] + ldw 4($tp),$ti0 ; tp[j] + + add $hi1,$nmlo1,$nmlo1 + addc %r0,$nmhi1,$nmhi1 + fstws,mb ${fab0}L,-8($xfer) ; save high part + add $ablo,$nmlo1,$nmlo1 + addc %r0,$nmhi1,$hi1 + fcpy,sgl %fr0,${fti}L ; zero high part + fcpy,sgl %fr0,${fab0}L + stw $nmlo1,-4($tp) ; tp[j-1] + + fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double + fcnvxf,dbl,dbl ${fab0},${fab0} + add $hi1,$hi0,$hi0 + addc %r0,%r0,$hi1 + fadd,dbl ${fti},${fab0},${fab0} ; add tp[0] + add $ti0,$hi0,$hi0 + addc %r0,$hi1,$hi1 + fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int + stw $hi0,0($tp) + stw $hi1,4($tp) + xmpyu ${fn0},${fab0}R,${fm0} + + b L\$outer_pa11 + ldo `$LOCALS+32+4`($fp),$tp + +L\$outerdone_pa11 + add $hi0,$ablo,$ablo + addc %r0,$abhi,$abhi + add $ti1,$ablo,$ablo + addc %r0,$abhi,$hi0 + + ldw 4($tp),$ti0 ; tp[j] + + add $hi1,$nmlo1,$nmlo1 + addc %r0,$nmhi1,$nmhi1 + add $ablo,$nmlo1,$nmlo1 + addc %r0,$nmhi1,$hi1 + stw $nmlo1,-4($tp) ; tp[j-1] + + add $hi1,$hi0,$hi0 + addc %r0,%r0,$hi1 + add $ti0,$hi0,$hi0 + addc %r0,$hi1,$hi1 + stw $hi0,0($tp) + stw $hi1,4($tp) + + ldo `$LOCALS+32+4`($fp),$tp + sub %r0,%r0,%r0 ; clear borrow + ldw -4($tp),$ti0 + addl $tp,$arrsz,$tp +L\$sub_pa11 + ldwx $idx($np),$hi0 + subb $ti0,$hi0,$hi1 + ldwx $idx($tp),$ti0 + addib,<> 4,$idx,L\$sub_pa11 + stws,ma $hi1,4($rp) + + subb $ti0,%r0,$hi1 + ldo -4($tp),$tp + and $tp,$hi1,$ap + andcm $rp,$hi1,$bp + or $ap,$bp,$np + + sub $rp,$arrsz,$rp ; rewind rp + subi 0,$arrsz,$idx + ldo `$LOCALS+32`($fp),$tp +L\$copy_pa11 + ldwx $idx($np),$hi0 + stws,ma %r0,4($tp) + addib,<> 4,$idx,L\$copy_pa11 + stws,ma $hi0,4($rp) + + nop ; alignment +L\$done +___ +} + +$code.=<<___; + ldi 1,%r28 ; signal "handled" + ldo $FRAME($fp),%sp ; destroy tp[num+1] + + $POP `-$FRAME-$SAVED_RP`(%sp),%r2 ; standard epilogue + $POP `-$FRAME+1*$SIZE_T`(%sp),%r4 + $POP `-$FRAME+2*$SIZE_T`(%sp),%r5 + $POP `-$FRAME+3*$SIZE_T`(%sp),%r6 + $POP `-$FRAME+4*$SIZE_T`(%sp),%r7 + $POP `-$FRAME+5*$SIZE_T`(%sp),%r8 + $POP `-$FRAME+6*$SIZE_T`(%sp),%r9 + $POP `-$FRAME+7*$SIZE_T`(%sp),%r10 +L\$abort + bv (%r2) + .EXIT + $POPMB -$FRAME(%sp),%r3 + .PROCEND + .STRINGZ "Montgomery Multiplication for PA-RISC, CRYPTOGAMS by " +___ + +# Explicitly encode PA-RISC 2.0 instructions used in this module, so +# that it can be compiled with .LEVEL 1.0. It should be noted that I +# wouldn't have to do this, if GNU assembler understood .ALLOW 2.0 +# directive... + +my $ldd = sub { + my ($mod,$args) = @_; + my $orig = "ldd$mod\t$args"; + + if ($args =~ /%r([0-9]+)\(%r([0-9]+)\),%r([0-9]+)/) # format 4 + { my $opcode=(0x03<<26)|($2<<21)|($1<<16)|(3<<6)|$3; + sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig; + } + elsif ($args =~ /(\-?[0-9]+)\(%r([0-9]+)\),%r([0-9]+)/) # format 5 + { my $opcode=(0x03<<26)|($2<<21)|(1<<12)|(3<<6)|$3; + $opcode|=(($1&0xF)<<17)|(($1&0x10)<<12); # encode offset + $opcode|=(1<<5) if ($mod =~ /^,m/); + $opcode|=(1<<13) if ($mod =~ /^,mb/); + sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig; + } + else { "\t".$orig; } +}; + +my $std = sub { + my ($mod,$args) = @_; + my $orig = "std$mod\t$args"; + + if ($args =~ /%r([0-9]+),(\-?[0-9]+)\(%r([0-9]+)\)/) # format 6 + { my $opcode=(0x03<<26)|($3<<21)|($1<<16)|(1<<12)|(0xB<<6); + $opcode|=(($2&0xF)<<1)|(($2&0x10)>>4); # encode offset + $opcode|=(1<<5) if ($mod =~ /^,m/); + $opcode|=(1<<13) if ($mod =~ /^,mb/); + sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig; + } + else { "\t".$orig; } +}; + +my $extrd = sub { + my ($mod,$args) = @_; + my $orig = "extrd$mod\t$args"; + + # I only have ",u" completer, it's implicitly encoded... + if ($args =~ /%r([0-9]+),([0-9]+),([0-9]+),%r([0-9]+)/) # format 15 + { my $opcode=(0x36<<26)|($1<<21)|($4<<16); + my $len=32-$3; + $opcode |= (($2&0x20)<<6)|(($2&0x1f)<<5); # encode pos + $opcode |= (($len&0x20)<<7)|($len&0x1f); # encode len + sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig; + } + elsif ($args =~ /%r([0-9]+),%sar,([0-9]+),%r([0-9]+)/) # format 12 + { my $opcode=(0x34<<26)|($1<<21)|($3<<16)|(2<<11)|(1<<9); + my $len=32-$2; + $opcode |= (($len&0x20)<<3)|($len&0x1f); # encode len + $opcode |= (1<<13) if ($mod =~ /,\**=/); + sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig; + } + else { "\t".$orig; } +}; + +my $shrpd = sub { + my ($mod,$args) = @_; + my $orig = "shrpd$mod\t$args"; + + if ($args =~ /%r([0-9]+),%r([0-9]+),([0-9]+),%r([0-9]+)/) # format 14 + { my $opcode=(0x34<<26)|($2<<21)|($1<<16)|(1<<10)|$4; + my $cpos=63-$3; + $opcode |= (($cpos&0x20)<<6)|(($cpos&0x1f)<<5); # encode sa + sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig; + } + else { "\t".$orig; } +}; + +my $sub = sub { + my ($mod,$args) = @_; + my $orig = "sub$mod\t$args"; + + if ($mod eq ",db" && $args =~ /%r([0-9]+),%r([0-9]+),%r([0-9]+)/) { + my $opcode=(0x02<<26)|($2<<21)|($1<<16)|$3; + $opcode|=(1<<10); # e1 + $opcode|=(1<<8); # e2 + $opcode|=(1<<5); # d + sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig + } + else { "\t".$orig; } +}; + +sub assemble { + my ($mnemonic,$mod,$args)=@_; + my $opcode = eval("\$$mnemonic"); + + ref($opcode) eq 'CODE' ? &$opcode($mod,$args) : "\t$mnemonic$mod\t$args"; +} + +foreach (split("\n",$code)) { + s/\`([^\`]*)\`/eval $1/ge; + # flip word order in 64-bit mode... + s/(xmpyu\s+)($fai|$fni)([LR])/$1.$2.($3 eq "L"?"R":"L")/e if ($BN_SZ==8); + # assemble 2.0 instructions in 32-bit mode... + s/^\s+([a-z]+)([\S]*)\s+([\S]*)/&assemble($1,$2,$3)/e if ($BN_SZ==4); + + s/\bbv\b/bve/gm if ($SIZE_T==8); + + print $_,"\n"; +} +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc-mont.pl new file mode 100644 index 00000000..f9b6992c --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc-mont.pl @@ -0,0 +1,334 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# April 2006 + +# "Teaser" Montgomery multiplication module for PowerPC. It's possible +# to gain a bit more by modulo-scheduling outer loop, then dedicated +# squaring procedure should give further 20% and code can be adapted +# for 32-bit application running on 64-bit CPU. As for the latter. +# It won't be able to achieve "native" 64-bit performance, because in +# 32-bit application context every addc instruction will have to be +# expanded as addc, twice right shift by 32 and finally adde, etc. +# So far RSA *sign* performance improvement over pre-bn_mul_mont asm +# for 64-bit application running on PPC970/G5 is: +# +# 512-bit +65% +# 1024-bit +35% +# 2048-bit +18% +# 4096-bit +4% + +$flavour = shift; + +if ($flavour =~ /32/) { + $BITS= 32; + $BNSZ= $BITS/8; + $SIZE_T=4; + $RZONE= 224; + + $LD= "lwz"; # load + $LDU= "lwzu"; # load and update + $LDX= "lwzx"; # load indexed + $ST= "stw"; # store + $STU= "stwu"; # store and update + $STX= "stwx"; # store indexed + $STUX= "stwux"; # store indexed and update + $UMULL= "mullw"; # unsigned multiply low + $UMULH= "mulhwu"; # unsigned multiply high + $UCMP= "cmplw"; # unsigned compare + $SHRI= "srwi"; # unsigned shift right by immediate + $PUSH= $ST; + $POP= $LD; +} elsif ($flavour =~ /64/) { + $BITS= 64; + $BNSZ= $BITS/8; + $SIZE_T=8; + $RZONE= 288; + + # same as above, but 64-bit mnemonics... + $LD= "ld"; # load + $LDU= "ldu"; # load and update + $LDX= "ldx"; # load indexed + $ST= "std"; # store + $STU= "stdu"; # store and update + $STX= "stdx"; # store indexed + $STUX= "stdux"; # store indexed and update + $UMULL= "mulld"; # unsigned multiply low + $UMULH= "mulhdu"; # unsigned multiply high + $UCMP= "cmpld"; # unsigned compare + $SHRI= "srdi"; # unsigned shift right by immediate + $PUSH= $ST; + $POP= $LD; +} else { die "nonsense $flavour"; } + +$FRAME=8*$SIZE_T+$RZONE; +$LOCALS=8*$SIZE_T; + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or +die "can't locate ppc-xlate.pl"; + +open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!"; + +$sp="r1"; +$toc="r2"; +$rp="r3"; $ovf="r3"; +$ap="r4"; +$bp="r5"; +$np="r6"; +$n0="r7"; +$num="r8"; +$rp="r9"; # $rp is reassigned +$aj="r10"; +$nj="r11"; +$tj="r12"; +# non-volatile registers +$i="r20"; +$j="r21"; +$tp="r22"; +$m0="r23"; +$m1="r24"; +$lo0="r25"; +$hi0="r26"; +$lo1="r27"; +$hi1="r28"; +$alo="r29"; +$ahi="r30"; +$nlo="r31"; +# +$nhi="r0"; + +$code=<<___; +.machine "any" +.text + +.globl .bn_mul_mont_int +.align 4 +.bn_mul_mont_int: + cmpwi $num,4 + mr $rp,r3 ; $rp is reassigned + li r3,0 + bltlr +___ +$code.=<<___ if ($BNSZ==4); + cmpwi $num,32 ; longer key performance is not better + bgelr +___ +$code.=<<___; + slwi $num,$num,`log($BNSZ)/log(2)` + li $tj,-4096 + addi $ovf,$num,$FRAME + subf $ovf,$ovf,$sp ; $sp-$ovf + and $ovf,$ovf,$tj ; minimize TLB usage + subf $ovf,$sp,$ovf ; $ovf-$sp + mr $tj,$sp + srwi $num,$num,`log($BNSZ)/log(2)` + $STUX $sp,$sp,$ovf + + $PUSH r20,`-12*$SIZE_T`($tj) + $PUSH r21,`-11*$SIZE_T`($tj) + $PUSH r22,`-10*$SIZE_T`($tj) + $PUSH r23,`-9*$SIZE_T`($tj) + $PUSH r24,`-8*$SIZE_T`($tj) + $PUSH r25,`-7*$SIZE_T`($tj) + $PUSH r26,`-6*$SIZE_T`($tj) + $PUSH r27,`-5*$SIZE_T`($tj) + $PUSH r28,`-4*$SIZE_T`($tj) + $PUSH r29,`-3*$SIZE_T`($tj) + $PUSH r30,`-2*$SIZE_T`($tj) + $PUSH r31,`-1*$SIZE_T`($tj) + + $LD $n0,0($n0) ; pull n0[0] value + addi $num,$num,-2 ; adjust $num for counter register + + $LD $m0,0($bp) ; m0=bp[0] + $LD $aj,0($ap) ; ap[0] + addi $tp,$sp,$LOCALS + $UMULL $lo0,$aj,$m0 ; ap[0]*bp[0] + $UMULH $hi0,$aj,$m0 + + $LD $aj,$BNSZ($ap) ; ap[1] + $LD $nj,0($np) ; np[0] + + $UMULL $m1,$lo0,$n0 ; "tp[0]"*n0 + + $UMULL $alo,$aj,$m0 ; ap[1]*bp[0] + $UMULH $ahi,$aj,$m0 + + $UMULL $lo1,$nj,$m1 ; np[0]*m1 + $UMULH $hi1,$nj,$m1 + $LD $nj,$BNSZ($np) ; np[1] + addc $lo1,$lo1,$lo0 + addze $hi1,$hi1 + + $UMULL $nlo,$nj,$m1 ; np[1]*m1 + $UMULH $nhi,$nj,$m1 + + mtctr $num + li $j,`2*$BNSZ` +.align 4 +L1st: + $LDX $aj,$ap,$j ; ap[j] + addc $lo0,$alo,$hi0 + $LDX $nj,$np,$j ; np[j] + addze $hi0,$ahi + $UMULL $alo,$aj,$m0 ; ap[j]*bp[0] + addc $lo1,$nlo,$hi1 + $UMULH $ahi,$aj,$m0 + addze $hi1,$nhi + $UMULL $nlo,$nj,$m1 ; np[j]*m1 + addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[0] + $UMULH $nhi,$nj,$m1 + addze $hi1,$hi1 + $ST $lo1,0($tp) ; tp[j-1] + + addi $j,$j,$BNSZ ; j++ + addi $tp,$tp,$BNSZ ; tp++ + bdnz- L1st +;L1st + addc $lo0,$alo,$hi0 + addze $hi0,$ahi + + addc $lo1,$nlo,$hi1 + addze $hi1,$nhi + addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[0] + addze $hi1,$hi1 + $ST $lo1,0($tp) ; tp[j-1] + + li $ovf,0 + addc $hi1,$hi1,$hi0 + addze $ovf,$ovf ; upmost overflow bit + $ST $hi1,$BNSZ($tp) + + li $i,$BNSZ +.align 4 +Louter: + $LDX $m0,$bp,$i ; m0=bp[i] + $LD $aj,0($ap) ; ap[0] + addi $tp,$sp,$LOCALS + $LD $tj,$LOCALS($sp); tp[0] + $UMULL $lo0,$aj,$m0 ; ap[0]*bp[i] + $UMULH $hi0,$aj,$m0 + $LD $aj,$BNSZ($ap) ; ap[1] + $LD $nj,0($np) ; np[0] + addc $lo0,$lo0,$tj ; ap[0]*bp[i]+tp[0] + $UMULL $alo,$aj,$m0 ; ap[j]*bp[i] + addze $hi0,$hi0 + $UMULL $m1,$lo0,$n0 ; tp[0]*n0 + $UMULH $ahi,$aj,$m0 + $UMULL $lo1,$nj,$m1 ; np[0]*m1 + $UMULH $hi1,$nj,$m1 + $LD $nj,$BNSZ($np) ; np[1] + addc $lo1,$lo1,$lo0 + $UMULL $nlo,$nj,$m1 ; np[1]*m1 + addze $hi1,$hi1 + $UMULH $nhi,$nj,$m1 + + mtctr $num + li $j,`2*$BNSZ` +.align 4 +Linner: + $LDX $aj,$ap,$j ; ap[j] + addc $lo0,$alo,$hi0 + $LD $tj,$BNSZ($tp) ; tp[j] + addze $hi0,$ahi + $LDX $nj,$np,$j ; np[j] + addc $lo1,$nlo,$hi1 + $UMULL $alo,$aj,$m0 ; ap[j]*bp[i] + addze $hi1,$nhi + $UMULH $ahi,$aj,$m0 + addc $lo0,$lo0,$tj ; ap[j]*bp[i]+tp[j] + $UMULL $nlo,$nj,$m1 ; np[j]*m1 + addze $hi0,$hi0 + $UMULH $nhi,$nj,$m1 + addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[i]+tp[j] + addi $j,$j,$BNSZ ; j++ + addze $hi1,$hi1 + $ST $lo1,0($tp) ; tp[j-1] + addi $tp,$tp,$BNSZ ; tp++ + bdnz- Linner +;Linner + $LD $tj,$BNSZ($tp) ; tp[j] + addc $lo0,$alo,$hi0 + addze $hi0,$ahi + addc $lo0,$lo0,$tj ; ap[j]*bp[i]+tp[j] + addze $hi0,$hi0 + + addc $lo1,$nlo,$hi1 + addze $hi1,$nhi + addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[i]+tp[j] + addze $hi1,$hi1 + $ST $lo1,0($tp) ; tp[j-1] + + addic $ovf,$ovf,-1 ; move upmost overflow to XER[CA] + li $ovf,0 + adde $hi1,$hi1,$hi0 + addze $ovf,$ovf + $ST $hi1,$BNSZ($tp) +; + slwi $tj,$num,`log($BNSZ)/log(2)` + $UCMP $i,$tj + addi $i,$i,$BNSZ + ble- Louter + + addi $num,$num,2 ; restore $num + subfc $j,$j,$j ; j=0 and "clear" XER[CA] + addi $tp,$sp,$LOCALS + mtctr $num + +.align 4 +Lsub: $LDX $tj,$tp,$j + $LDX $nj,$np,$j + subfe $aj,$nj,$tj ; tp[j]-np[j] + $STX $aj,$rp,$j + addi $j,$j,$BNSZ + bdnz- Lsub + + li $j,0 + mtctr $num + subfe $ovf,$j,$ovf ; handle upmost overflow bit + and $ap,$tp,$ovf + andc $np,$rp,$ovf + or $ap,$ap,$np ; ap=borrow?tp:rp + +.align 4 +Lcopy: ; copy or in-place refresh + $LDX $tj,$ap,$j + $STX $tj,$rp,$j + $STX $j,$tp,$j ; zap at once + addi $j,$j,$BNSZ + bdnz- Lcopy + + $POP $tj,0($sp) + li r3,1 + $POP r20,`-12*$SIZE_T`($tj) + $POP r21,`-11*$SIZE_T`($tj) + $POP r22,`-10*$SIZE_T`($tj) + $POP r23,`-9*$SIZE_T`($tj) + $POP r24,`-8*$SIZE_T`($tj) + $POP r25,`-7*$SIZE_T`($tj) + $POP r26,`-6*$SIZE_T`($tj) + $POP r27,`-5*$SIZE_T`($tj) + $POP r28,`-4*$SIZE_T`($tj) + $POP r29,`-3*$SIZE_T`($tj) + $POP r30,`-2*$SIZE_T`($tj) + $POP r31,`-1*$SIZE_T`($tj) + mr $sp,$tj + blr + .long 0 + .byte 0,12,4,0,0x80,12,6,0 + .long 0 + +.asciz "Montgomery Multiplication for PPC, CRYPTOGAMS by " +___ + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc.pl new file mode 100644 index 00000000..1249ce22 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc.pl @@ -0,0 +1,1998 @@ +#!/usr/bin/env perl +# +# Implemented as a Perl wrapper as we want to support several different +# architectures with single file. We pick up the target based on the +# file name we are asked to generate. +# +# It should be noted though that this perl code is nothing like +# /crypto/perlasm/x86*. In this case perl is used pretty much +# as pre-processor to cover for platform differences in name decoration, +# linker tables, 32-/64-bit instruction sets... +# +# As you might know there're several PowerPC ABI in use. Most notably +# Linux and AIX use different 32-bit ABIs. Good news are that these ABIs +# are similar enough to implement leaf(!) functions, which would be ABI +# neutral. And that's what you find here: ABI neutral leaf functions. +# In case you wonder what that is... +# +# AIX performance +# +# MEASUREMENTS WITH cc ON a 200 MhZ PowerPC 604e. +# +# The following is the performance of 32-bit compiler +# generated code: +# +# OpenSSL 0.9.6c 21 dec 2001 +# built on: Tue Jun 11 11:06:51 EDT 2002 +# options:bn(64,32) ... +#compiler: cc -DTHREADS -DAIX -DB_ENDIAN -DBN_LLONG -O3 +# sign verify sign/s verify/s +#rsa 512 bits 0.0098s 0.0009s 102.0 1170.6 +#rsa 1024 bits 0.0507s 0.0026s 19.7 387.5 +#rsa 2048 bits 0.3036s 0.0085s 3.3 117.1 +#rsa 4096 bits 2.0040s 0.0299s 0.5 33.4 +#dsa 512 bits 0.0087s 0.0106s 114.3 94.5 +#dsa 1024 bits 0.0256s 0.0313s 39.0 32.0 +# +# Same bechmark with this assembler code: +# +#rsa 512 bits 0.0056s 0.0005s 178.6 2049.2 +#rsa 1024 bits 0.0283s 0.0015s 35.3 674.1 +#rsa 2048 bits 0.1744s 0.0050s 5.7 201.2 +#rsa 4096 bits 1.1644s 0.0179s 0.9 55.7 +#dsa 512 bits 0.0052s 0.0062s 191.6 162.0 +#dsa 1024 bits 0.0149s 0.0180s 67.0 55.5 +# +# Number of operations increases by at almost 75% +# +# Here are performance numbers for 64-bit compiler +# generated code: +# +# OpenSSL 0.9.6g [engine] 9 Aug 2002 +# built on: Fri Apr 18 16:59:20 EDT 2003 +# options:bn(64,64) ... +# compiler: cc -DTHREADS -D_REENTRANT -q64 -DB_ENDIAN -O3 +# sign verify sign/s verify/s +#rsa 512 bits 0.0028s 0.0003s 357.1 3844.4 +#rsa 1024 bits 0.0148s 0.0008s 67.5 1239.7 +#rsa 2048 bits 0.0963s 0.0028s 10.4 353.0 +#rsa 4096 bits 0.6538s 0.0102s 1.5 98.1 +#dsa 512 bits 0.0026s 0.0032s 382.5 313.7 +#dsa 1024 bits 0.0081s 0.0099s 122.8 100.6 +# +# Same benchmark with this assembler code: +# +#rsa 512 bits 0.0020s 0.0002s 510.4 6273.7 +#rsa 1024 bits 0.0088s 0.0005s 114.1 2128.3 +#rsa 2048 bits 0.0540s 0.0016s 18.5 622.5 +#rsa 4096 bits 0.3700s 0.0058s 2.7 171.0 +#dsa 512 bits 0.0016s 0.0020s 610.7 507.1 +#dsa 1024 bits 0.0047s 0.0058s 212.5 173.2 +# +# Again, performance increases by at about 75% +# +# Mac OS X, Apple G5 1.8GHz (Note this is 32 bit code) +# OpenSSL 0.9.7c 30 Sep 2003 +# +# Original code. +# +#rsa 512 bits 0.0011s 0.0001s 906.1 11012.5 +#rsa 1024 bits 0.0060s 0.0003s 166.6 3363.1 +#rsa 2048 bits 0.0370s 0.0010s 27.1 982.4 +#rsa 4096 bits 0.2426s 0.0036s 4.1 280.4 +#dsa 512 bits 0.0010s 0.0012s 1038.1 841.5 +#dsa 1024 bits 0.0030s 0.0037s 329.6 269.7 +#dsa 2048 bits 0.0101s 0.0127s 98.9 78.6 +# +# Same benchmark with this assembler code: +# +#rsa 512 bits 0.0007s 0.0001s 1416.2 16645.9 +#rsa 1024 bits 0.0036s 0.0002s 274.4 5380.6 +#rsa 2048 bits 0.0222s 0.0006s 45.1 1589.5 +#rsa 4096 bits 0.1469s 0.0022s 6.8 449.6 +#dsa 512 bits 0.0006s 0.0007s 1664.2 1376.2 +#dsa 1024 bits 0.0018s 0.0023s 545.0 442.2 +#dsa 2048 bits 0.0061s 0.0075s 163.5 132.8 +# +# Performance increase of ~60% +# +# If you have comments or suggestions to improve code send +# me a note at schari@us.ibm.com +# + +$flavour = shift; + +if ($flavour =~ /32/) { + $BITS= 32; + $BNSZ= $BITS/8; + $ISA= "\"ppc\""; + + $LD= "lwz"; # load + $LDU= "lwzu"; # load and update + $ST= "stw"; # store + $STU= "stwu"; # store and update + $UMULL= "mullw"; # unsigned multiply low + $UMULH= "mulhwu"; # unsigned multiply high + $UDIV= "divwu"; # unsigned divide + $UCMPI= "cmplwi"; # unsigned compare with immediate + $UCMP= "cmplw"; # unsigned compare + $CNTLZ= "cntlzw"; # count leading zeros + $SHL= "slw"; # shift left + $SHR= "srw"; # unsigned shift right + $SHRI= "srwi"; # unsigned shift right by immediate + $SHLI= "slwi"; # shift left by immediate + $CLRU= "clrlwi"; # clear upper bits + $INSR= "insrwi"; # insert right + $ROTL= "rotlwi"; # rotate left by immediate + $TR= "tw"; # conditional trap +} elsif ($flavour =~ /64/) { + $BITS= 64; + $BNSZ= $BITS/8; + $ISA= "\"ppc64\""; + + # same as above, but 64-bit mnemonics... + $LD= "ld"; # load + $LDU= "ldu"; # load and update + $ST= "std"; # store + $STU= "stdu"; # store and update + $UMULL= "mulld"; # unsigned multiply low + $UMULH= "mulhdu"; # unsigned multiply high + $UDIV= "divdu"; # unsigned divide + $UCMPI= "cmpldi"; # unsigned compare with immediate + $UCMP= "cmpld"; # unsigned compare + $CNTLZ= "cntlzd"; # count leading zeros + $SHL= "sld"; # shift left + $SHR= "srd"; # unsigned shift right + $SHRI= "srdi"; # unsigned shift right by immediate + $SHLI= "sldi"; # shift left by immediate + $CLRU= "clrldi"; # clear upper bits + $INSR= "insrdi"; # insert right + $ROTL= "rotldi"; # rotate left by immediate + $TR= "td"; # conditional trap +} else { die "nonsense $flavour"; } + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or +die "can't locate ppc-xlate.pl"; + +open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!"; + +$data=< 0 then result !=0 + # In either case carry bit is set. + beq Lppcasm_sub_adios + addi r4,r4,-$BNSZ + addi r3,r3,-$BNSZ + addi r5,r5,-$BNSZ + mtctr r6 +Lppcasm_sub_mainloop: + $LDU r7,$BNSZ(r4) + $LDU r8,$BNSZ(r5) + subfe r6,r8,r7 # r6 = r7+carry bit + onescomplement(r8) + # if carry = 1 this is r7-r8. Else it + # is r7-r8 -1 as we need. + $STU r6,$BNSZ(r3) + bdnz- Lppcasm_sub_mainloop +Lppcasm_sub_adios: + subfze r3,r0 # if carry bit is set then r3 = 0 else -1 + andi. r3,r3,1 # keep only last bit. + blr + .long 0 + .byte 0,12,0x14,0,0,0,4,0 + .long 0 + +# +# NOTE: The following label name should be changed to +# "bn_add_words" i.e. remove the first dot +# for the gcc compiler. This should be automatically +# done in the build +# + +.align 4 +.bn_add_words: +# +# Handcoded version of bn_add_words +# +#BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) +# +# r3 = r +# r4 = a +# r5 = b +# r6 = n +# +# Note: No loop unrolling done since this is not a performance +# critical loop. + + xor r0,r0,r0 +# +# check for r6 = 0. Is this needed? +# + addic. r6,r6,0 #test r6 and clear carry bit. + beq Lppcasm_add_adios + addi r4,r4,-$BNSZ + addi r3,r3,-$BNSZ + addi r5,r5,-$BNSZ + mtctr r6 +Lppcasm_add_mainloop: + $LDU r7,$BNSZ(r4) + $LDU r8,$BNSZ(r5) + adde r8,r7,r8 + $STU r8,$BNSZ(r3) + bdnz- Lppcasm_add_mainloop +Lppcasm_add_adios: + addze r3,r0 #return carry bit. + blr + .long 0 + .byte 0,12,0x14,0,0,0,4,0 + .long 0 + +# +# NOTE: The following label name should be changed to +# "bn_div_words" i.e. remove the first dot +# for the gcc compiler. This should be automatically +# done in the build +# + +.align 4 +.bn_div_words: +# +# This is a cleaned up version of code generated by +# the AIX compiler. The only optimization is to use +# the PPC instruction to count leading zeros instead +# of call to num_bits_word. Since this was compiled +# only at level -O2 we can possibly squeeze it more? +# +# r3 = h +# r4 = l +# r5 = d + + $UCMPI 0,r5,0 # compare r5 and 0 + bne Lppcasm_div1 # proceed if d!=0 + li r3,-1 # d=0 return -1 + blr +Lppcasm_div1: + xor r0,r0,r0 #r0=0 + li r8,$BITS + $CNTLZ. r7,r5 #r7 = num leading 0s in d. + beq Lppcasm_div2 #proceed if no leading zeros + subf r8,r7,r8 #r8 = BN_num_bits_word(d) + $SHR. r9,r3,r8 #are there any bits above r8'th? + $TR 16,r9,r0 #if there're, signal to dump core... +Lppcasm_div2: + $UCMP 0,r3,r5 #h>=d? + blt Lppcasm_div3 #goto Lppcasm_div3 if not + subf r3,r5,r3 #h-=d ; +Lppcasm_div3: #r7 = BN_BITS2-i. so r7=i + cmpi 0,0,r7,0 # is (i == 0)? + beq Lppcasm_div4 + $SHL r3,r3,r7 # h = (h<< i) + $SHR r8,r4,r8 # r8 = (l >> BN_BITS2 -i) + $SHL r5,r5,r7 # d<<=i + or r3,r3,r8 # h = (h<>(BN_BITS2-i)) + $SHL r4,r4,r7 # l <<=i +Lppcasm_div4: + $SHRI r9,r5,`$BITS/2` # r9 = dh + # dl will be computed when needed + # as it saves registers. + li r6,2 #r6=2 + mtctr r6 #counter will be in count. +Lppcasm_divouterloop: + $SHRI r8,r3,`$BITS/2` #r8 = (h>>BN_BITS4) + $SHRI r11,r4,`$BITS/2` #r11= (l&BN_MASK2h)>>BN_BITS4 + # compute here for innerloop. + $UCMP 0,r8,r9 # is (h>>BN_BITS4)==dh + bne Lppcasm_div5 # goto Lppcasm_div5 if not + + li r8,-1 + $CLRU r8,r8,`$BITS/2` #q = BN_MASK2l + b Lppcasm_div6 +Lppcasm_div5: + $UDIV r8,r3,r9 #q = h/dh +Lppcasm_div6: + $UMULL r12,r9,r8 #th = q*dh + $CLRU r10,r5,`$BITS/2` #r10=dl + $UMULL r6,r8,r10 #tl = q*dl + +Lppcasm_divinnerloop: + subf r10,r12,r3 #t = h -th + $SHRI r7,r10,`$BITS/2` #r7= (t &BN_MASK2H), sort of... + addic. r7,r7,0 #test if r7 == 0. used below. + # now want to compute + # r7 = (t<>BN_BITS4) + # the following 2 instructions do that + $SHLI r7,r10,`$BITS/2` # r7 = (t<>BN_BITS4) + $UCMP cr1,r6,r7 # compare (tl <= r7) + bne Lppcasm_divinnerexit + ble cr1,Lppcasm_divinnerexit + addi r8,r8,-1 #q-- + subf r12,r9,r12 #th -=dh + $CLRU r10,r5,`$BITS/2` #r10=dl. t is no longer needed in loop. + subf r6,r10,r6 #tl -=dl + b Lppcasm_divinnerloop +Lppcasm_divinnerexit: + $SHRI r10,r6,`$BITS/2` #t=(tl>>BN_BITS4) + $SHLI r11,r6,`$BITS/2` #tl=(tl<=tl) goto Lppcasm_div7 + addi r12,r12,1 # th++ +Lppcasm_div7: + subf r11,r11,r4 #r11=l-tl + $UCMP cr1,r3,r12 #compare h and th + bge cr1,Lppcasm_div8 #if (h>=th) goto Lppcasm_div8 + addi r8,r8,-1 # q-- + add r3,r5,r3 # h+=d +Lppcasm_div8: + subf r12,r12,r3 #r12 = h-th + $SHLI r4,r11,`$BITS/2` #l=(l&BN_MASK2l)<>BN_BITS4))&BN_MASK2 + # the following 2 instructions will do this. + $INSR r11,r12,`$BITS/2`,`$BITS/2` # r11 is the value we want rotated $BITS/2. + $ROTL r3,r11,`$BITS/2` # rotate by $BITS/2 and store in r3 + bdz Lppcasm_div9 #if (count==0) break ; + $SHLI r0,r8,`$BITS/2` #ret =q<> 2 + beq Lppcasm_mw_REM + mtctr r7 +Lppcasm_mw_LOOP: + #mul(rp[0],ap[0],w,c1); + $LD r8,`0*$BNSZ`(r4) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + addc r9,r9,r12 + #addze r10,r10 #carry is NOT ignored. + #will be taken care of + #in second spin below + #using adde. + $ST r9,`0*$BNSZ`(r3) + #mul(rp[1],ap[1],w,c1); + $LD r8,`1*$BNSZ`(r4) + $UMULL r11,r6,r8 + $UMULH r12,r6,r8 + adde r11,r11,r10 + #addze r12,r12 + $ST r11,`1*$BNSZ`(r3) + #mul(rp[2],ap[2],w,c1); + $LD r8,`2*$BNSZ`(r4) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + adde r9,r9,r12 + #addze r10,r10 + $ST r9,`2*$BNSZ`(r3) + #mul_add(rp[3],ap[3],w,c1); + $LD r8,`3*$BNSZ`(r4) + $UMULL r11,r6,r8 + $UMULH r12,r6,r8 + adde r11,r11,r10 + addze r12,r12 #this spin we collect carry into + #r12 + $ST r11,`3*$BNSZ`(r3) + + addi r3,r3,`4*$BNSZ` + addi r4,r4,`4*$BNSZ` + bdnz- Lppcasm_mw_LOOP + +Lppcasm_mw_REM: + andi. r5,r5,0x3 + beq Lppcasm_mw_OVER + #mul(rp[0],ap[0],w,c1); + $LD r8,`0*$BNSZ`(r4) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + addc r9,r9,r12 + addze r10,r10 + $ST r9,`0*$BNSZ`(r3) + addi r12,r10,0 + + addi r5,r5,-1 + cmpli 0,0,r5,0 + beq Lppcasm_mw_OVER + + + #mul(rp[1],ap[1],w,c1); + $LD r8,`1*$BNSZ`(r4) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + addc r9,r9,r12 + addze r10,r10 + $ST r9,`1*$BNSZ`(r3) + addi r12,r10,0 + + addi r5,r5,-1 + cmpli 0,0,r5,0 + beq Lppcasm_mw_OVER + + #mul_add(rp[2],ap[2],w,c1); + $LD r8,`2*$BNSZ`(r4) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + addc r9,r9,r12 + addze r10,r10 + $ST r9,`2*$BNSZ`(r3) + addi r12,r10,0 + +Lppcasm_mw_OVER: + addi r3,r12,0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,4,0 + .long 0 + +# +# NOTE: The following label name should be changed to +# "bn_mul_add_words" i.e. remove the first dot +# for the gcc compiler. This should be automatically +# done in the build +# + +.align 4 +.bn_mul_add_words: +# +# BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w) +# +# r3 = rp +# r4 = ap +# r5 = num +# r6 = w +# +# empirical evidence suggests that unrolled version performs best!! +# + xor r0,r0,r0 #r0 = 0 + xor r12,r12,r12 #r12 = 0 . used for carry + rlwinm. r7,r5,30,2,31 # num >> 2 + beq Lppcasm_maw_leftover # if (num < 4) go LPPCASM_maw_leftover + mtctr r7 +Lppcasm_maw_mainloop: + #mul_add(rp[0],ap[0],w,c1); + $LD r8,`0*$BNSZ`(r4) + $LD r11,`0*$BNSZ`(r3) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + addc r9,r9,r12 #r12 is carry. + addze r10,r10 + addc r9,r9,r11 + #addze r10,r10 + #the above instruction addze + #is NOT needed. Carry will NOT + #be ignored. It's not affected + #by multiply and will be collected + #in the next spin + $ST r9,`0*$BNSZ`(r3) + + #mul_add(rp[1],ap[1],w,c1); + $LD r8,`1*$BNSZ`(r4) + $LD r9,`1*$BNSZ`(r3) + $UMULL r11,r6,r8 + $UMULH r12,r6,r8 + adde r11,r11,r10 #r10 is carry. + addze r12,r12 + addc r11,r11,r9 + #addze r12,r12 + $ST r11,`1*$BNSZ`(r3) + + #mul_add(rp[2],ap[2],w,c1); + $LD r8,`2*$BNSZ`(r4) + $UMULL r9,r6,r8 + $LD r11,`2*$BNSZ`(r3) + $UMULH r10,r6,r8 + adde r9,r9,r12 + addze r10,r10 + addc r9,r9,r11 + #addze r10,r10 + $ST r9,`2*$BNSZ`(r3) + + #mul_add(rp[3],ap[3],w,c1); + $LD r8,`3*$BNSZ`(r4) + $UMULL r11,r6,r8 + $LD r9,`3*$BNSZ`(r3) + $UMULH r12,r6,r8 + adde r11,r11,r10 + addze r12,r12 + addc r11,r11,r9 + addze r12,r12 + $ST r11,`3*$BNSZ`(r3) + addi r3,r3,`4*$BNSZ` + addi r4,r4,`4*$BNSZ` + bdnz- Lppcasm_maw_mainloop + +Lppcasm_maw_leftover: + andi. r5,r5,0x3 + beq Lppcasm_maw_adios + addi r3,r3,-$BNSZ + addi r4,r4,-$BNSZ + #mul_add(rp[0],ap[0],w,c1); + mtctr r5 + $LDU r8,$BNSZ(r4) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + $LDU r11,$BNSZ(r3) + addc r9,r9,r11 + addze r10,r10 + addc r9,r9,r12 + addze r12,r10 + $ST r9,0(r3) + + bdz Lppcasm_maw_adios + #mul_add(rp[1],ap[1],w,c1); + $LDU r8,$BNSZ(r4) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + $LDU r11,$BNSZ(r3) + addc r9,r9,r11 + addze r10,r10 + addc r9,r9,r12 + addze r12,r10 + $ST r9,0(r3) + + bdz Lppcasm_maw_adios + #mul_add(rp[2],ap[2],w,c1); + $LDU r8,$BNSZ(r4) + $UMULL r9,r6,r8 + $UMULH r10,r6,r8 + $LDU r11,$BNSZ(r3) + addc r9,r9,r11 + addze r10,r10 + addc r9,r9,r12 + addze r12,r10 + $ST r9,0(r3) + +Lppcasm_maw_adios: + addi r3,r12,0 + blr + .long 0 + .byte 0,12,0x14,0,0,0,4,0 + .long 0 + .align 4 +EOF +$data =~ s/\`([^\`]*)\`/eval $1/gem; +print $data; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc64-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc64-mont.pl new file mode 100644 index 00000000..a14e769a --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/ppc64-mont.pl @@ -0,0 +1,1088 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# December 2007 + +# The reason for undertaken effort is basically following. Even though +# Power 6 CPU operates at incredible 4.7GHz clock frequency, its PKI +# performance was observed to be less than impressive, essentially as +# fast as 1.8GHz PPC970, or 2.6 times(!) slower than one would hope. +# Well, it's not surprising that IBM had to make some sacrifices to +# boost the clock frequency that much, but no overall improvement? +# Having observed how much difference did switching to FPU make on +# UltraSPARC, playing same stunt on Power 6 appeared appropriate... +# Unfortunately the resulting performance improvement is not as +# impressive, ~30%, and in absolute terms is still very far from what +# one would expect from 4.7GHz CPU. There is a chance that I'm doing +# something wrong, but in the lack of assembler level micro-profiling +# data or at least decent platform guide I can't tell... Or better +# results might be achieved with VMX... Anyway, this module provides +# *worse* performance on other PowerPC implementations, ~40-15% slower +# on PPC970 depending on key length and ~40% slower on Power 5 for all +# key lengths. As it's obviously inappropriate as "best all-round" +# alternative, it has to be complemented with run-time CPU family +# detection. Oh! It should also be noted that unlike other PowerPC +# implementation IALU ppc-mont.pl module performs *suboptimaly* on +# >=1024-bit key lengths on Power 6. It should also be noted that +# *everything* said so far applies to 64-bit builds! As far as 32-bit +# application executed on 64-bit CPU goes, this module is likely to +# become preferred choice, because it's easy to adapt it for such +# case and *is* faster than 32-bit ppc-mont.pl on *all* processors. + +# February 2008 + +# Micro-profiling assisted optimization results in ~15% improvement +# over original ppc64-mont.pl version, or overall ~50% improvement +# over ppc.pl module on Power 6. If compared to ppc-mont.pl on same +# Power 6 CPU, this module is 5-150% faster depending on key length, +# [hereafter] more for longer keys. But if compared to ppc-mont.pl +# on 1.8GHz PPC970, it's only 5-55% faster. Still far from impressive +# in absolute terms, but it's apparently the way Power 6 is... + +# December 2009 + +# Adapted for 32-bit build this module delivers 25-120%, yes, more +# than *twice* for longer keys, performance improvement over 32-bit +# ppc-mont.pl on 1.8GHz PPC970. However! This implementation utilizes +# even 64-bit integer operations and the trouble is that most PPC +# operating systems don't preserve upper halves of general purpose +# registers upon 32-bit signal delivery. They do preserve them upon +# context switch, but not signalling:-( This means that asynchronous +# signals have to be blocked upon entry to this subroutine. Signal +# masking (and of course complementary unmasking) has quite an impact +# on performance, naturally larger for shorter keys. It's so severe +# that 512-bit key performance can be as low as 1/3 of expected one. +# This is why this routine can be engaged for longer key operations +# only on these OSes, see crypto/ppccap.c for further details. MacOS X +# is an exception from this and doesn't require signal masking, and +# that's where above improvement coefficients were collected. For +# others alternative would be to break dependence on upper halves of +# GPRs by sticking to 32-bit integer operations... + +$flavour = shift; + +if ($flavour =~ /32/) { + $SIZE_T=4; + $RZONE= 224; + $fname= "bn_mul_mont_fpu64"; + + $STUX= "stwux"; # store indexed and update + $PUSH= "stw"; + $POP= "lwz"; +} elsif ($flavour =~ /64/) { + $SIZE_T=8; + $RZONE= 288; + $fname= "bn_mul_mont_fpu64"; + + # same as above, but 64-bit mnemonics... + $STUX= "stdux"; # store indexed and update + $PUSH= "std"; + $POP= "ld"; +} else { die "nonsense $flavour"; } + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or +die "can't locate ppc-xlate.pl"; + +open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!"; + +$FRAME=64; # padded frame header +$TRANSFER=16*8; + +$carry="r0"; +$sp="r1"; +$toc="r2"; +$rp="r3"; $ovf="r3"; +$ap="r4"; +$bp="r5"; +$np="r6"; +$n0="r7"; +$num="r8"; +$rp="r9"; # $rp is reassigned +$tp="r10"; +$j="r11"; +$i="r12"; +# non-volatile registers +$nap_d="r22"; # interleaved ap and np in double format +$a0="r23"; # ap[0] +$t0="r24"; # temporary registers +$t1="r25"; +$t2="r26"; +$t3="r27"; +$t4="r28"; +$t5="r29"; +$t6="r30"; +$t7="r31"; + +# PPC offers enough register bank capacity to unroll inner loops twice +# +# ..A3A2A1A0 +# dcba +# ----------- +# A0a +# A0b +# A0c +# A0d +# A1a +# A1b +# A1c +# A1d +# A2a +# A2b +# A2c +# A2d +# A3a +# A3b +# A3c +# A3d +# ..a +# ..b +# +$ba="f0"; $bb="f1"; $bc="f2"; $bd="f3"; +$na="f4"; $nb="f5"; $nc="f6"; $nd="f7"; +$dota="f8"; $dotb="f9"; +$A0="f10"; $A1="f11"; $A2="f12"; $A3="f13"; +$N0="f20"; $N1="f21"; $N2="f22"; $N3="f23"; +$T0a="f24"; $T0b="f25"; +$T1a="f26"; $T1b="f27"; +$T2a="f28"; $T2b="f29"; +$T3a="f30"; $T3b="f31"; + +# sp----------->+-------------------------------+ +# | saved sp | +# +-------------------------------+ +# . . +# +64 +-------------------------------+ +# | 16 gpr<->fpr transfer zone | +# . . +# . . +# +16*8 +-------------------------------+ +# | __int64 tmp[-1] | +# +-------------------------------+ +# | __int64 tmp[num] | +# . . +# . . +# . . +# +(num+1)*8 +-------------------------------+ +# | padding to 64 byte boundary | +# . . +# +X +-------------------------------+ +# | double nap_d[4*num] | +# . . +# . . +# . . +# +-------------------------------+ +# . . +# -12*size_t +-------------------------------+ +# | 10 saved gpr, r22-r31 | +# . . +# . . +# -12*8 +-------------------------------+ +# | 12 saved fpr, f20-f31 | +# . . +# . . +# +-------------------------------+ + +$code=<<___; +.machine "any" +.text + +.globl .$fname +.align 5 +.$fname: + cmpwi $num,`3*8/$SIZE_T` + mr $rp,r3 ; $rp is reassigned + li r3,0 ; possible "not handled" return code + bltlr- + andi. r0,$num,`16/$SIZE_T-1` ; $num has to be "even" + bnelr- + + slwi $num,$num,`log($SIZE_T)/log(2)` ; num*=sizeof(BN_LONG) + li $i,-4096 + slwi $tp,$num,2 ; place for {an}p_{lh}[num], i.e. 4*num + add $tp,$tp,$num ; place for tp[num+1] + addi $tp,$tp,`$FRAME+$TRANSFER+8+64+$RZONE` + subf $tp,$tp,$sp ; $sp-$tp + and $tp,$tp,$i ; minimize TLB usage + subf $tp,$sp,$tp ; $tp-$sp + mr $i,$sp + $STUX $sp,$sp,$tp ; alloca + + $PUSH r22,`-12*8-10*$SIZE_T`($i) + $PUSH r23,`-12*8-9*$SIZE_T`($i) + $PUSH r24,`-12*8-8*$SIZE_T`($i) + $PUSH r25,`-12*8-7*$SIZE_T`($i) + $PUSH r26,`-12*8-6*$SIZE_T`($i) + $PUSH r27,`-12*8-5*$SIZE_T`($i) + $PUSH r28,`-12*8-4*$SIZE_T`($i) + $PUSH r29,`-12*8-3*$SIZE_T`($i) + $PUSH r30,`-12*8-2*$SIZE_T`($i) + $PUSH r31,`-12*8-1*$SIZE_T`($i) + stfd f20,`-12*8`($i) + stfd f21,`-11*8`($i) + stfd f22,`-10*8`($i) + stfd f23,`-9*8`($i) + stfd f24,`-8*8`($i) + stfd f25,`-7*8`($i) + stfd f26,`-6*8`($i) + stfd f27,`-5*8`($i) + stfd f28,`-4*8`($i) + stfd f29,`-3*8`($i) + stfd f30,`-2*8`($i) + stfd f31,`-1*8`($i) +___ +$code.=<<___ if ($SIZE_T==8); + ld $a0,0($ap) ; pull ap[0] value + ld $n0,0($n0) ; pull n0[0] value + ld $t3,0($bp) ; bp[0] +___ +$code.=<<___ if ($SIZE_T==4); + mr $t1,$n0 + lwz $a0,0($ap) ; pull ap[0,1] value + lwz $t0,4($ap) + lwz $n0,0($t1) ; pull n0[0,1] value + lwz $t1,4($t1) + lwz $t3,0($bp) ; bp[0,1] + lwz $t2,4($bp) + insrdi $a0,$t0,32,0 + insrdi $n0,$t1,32,0 + insrdi $t3,$t2,32,0 +___ +$code.=<<___; + addi $tp,$sp,`$FRAME+$TRANSFER+8+64` + li $i,-64 + add $nap_d,$tp,$num + and $nap_d,$nap_d,$i ; align to 64 bytes + + mulld $t7,$a0,$t3 ; ap[0]*bp[0] + ; nap_d is off by 1, because it's used with stfdu/lfdu + addi $nap_d,$nap_d,-8 + srwi $j,$num,`3+1` ; counter register, num/2 + mulld $t7,$t7,$n0 ; tp[0]*n0 + addi $j,$j,-1 + addi $tp,$sp,`$FRAME+$TRANSFER-8` + li $carry,0 + mtctr $j + + ; transfer bp[0] to FPU as 4x16-bit values + extrdi $t0,$t3,16,48 + extrdi $t1,$t3,16,32 + extrdi $t2,$t3,16,16 + extrdi $t3,$t3,16,0 + std $t0,`$FRAME+0`($sp) + std $t1,`$FRAME+8`($sp) + std $t2,`$FRAME+16`($sp) + std $t3,`$FRAME+24`($sp) + ; transfer (ap[0]*bp[0])*n0 to FPU as 4x16-bit values + extrdi $t4,$t7,16,48 + extrdi $t5,$t7,16,32 + extrdi $t6,$t7,16,16 + extrdi $t7,$t7,16,0 + std $t4,`$FRAME+32`($sp) + std $t5,`$FRAME+40`($sp) + std $t6,`$FRAME+48`($sp) + std $t7,`$FRAME+56`($sp) +___ +$code.=<<___ if ($SIZE_T==8); + lwz $t0,4($ap) ; load a[j] as 32-bit word pair + lwz $t1,0($ap) + lwz $t2,12($ap) ; load a[j+1] as 32-bit word pair + lwz $t3,8($ap) + lwz $t4,4($np) ; load n[j] as 32-bit word pair + lwz $t5,0($np) + lwz $t6,12($np) ; load n[j+1] as 32-bit word pair + lwz $t7,8($np) +___ +$code.=<<___ if ($SIZE_T==4); + lwz $t0,0($ap) ; load a[j..j+3] as 32-bit word pairs + lwz $t1,4($ap) + lwz $t2,8($ap) + lwz $t3,12($ap) + lwz $t4,0($np) ; load n[j..j+3] as 32-bit word pairs + lwz $t5,4($np) + lwz $t6,8($np) + lwz $t7,12($np) +___ +$code.=<<___; + lfd $ba,`$FRAME+0`($sp) + lfd $bb,`$FRAME+8`($sp) + lfd $bc,`$FRAME+16`($sp) + lfd $bd,`$FRAME+24`($sp) + lfd $na,`$FRAME+32`($sp) + lfd $nb,`$FRAME+40`($sp) + lfd $nc,`$FRAME+48`($sp) + lfd $nd,`$FRAME+56`($sp) + std $t0,`$FRAME+64`($sp) + std $t1,`$FRAME+72`($sp) + std $t2,`$FRAME+80`($sp) + std $t3,`$FRAME+88`($sp) + std $t4,`$FRAME+96`($sp) + std $t5,`$FRAME+104`($sp) + std $t6,`$FRAME+112`($sp) + std $t7,`$FRAME+120`($sp) + fcfid $ba,$ba + fcfid $bb,$bb + fcfid $bc,$bc + fcfid $bd,$bd + fcfid $na,$na + fcfid $nb,$nb + fcfid $nc,$nc + fcfid $nd,$nd + + lfd $A0,`$FRAME+64`($sp) + lfd $A1,`$FRAME+72`($sp) + lfd $A2,`$FRAME+80`($sp) + lfd $A3,`$FRAME+88`($sp) + lfd $N0,`$FRAME+96`($sp) + lfd $N1,`$FRAME+104`($sp) + lfd $N2,`$FRAME+112`($sp) + lfd $N3,`$FRAME+120`($sp) + fcfid $A0,$A0 + fcfid $A1,$A1 + fcfid $A2,$A2 + fcfid $A3,$A3 + fcfid $N0,$N0 + fcfid $N1,$N1 + fcfid $N2,$N2 + fcfid $N3,$N3 + addi $ap,$ap,16 + addi $np,$np,16 + + fmul $T1a,$A1,$ba + fmul $T1b,$A1,$bb + stfd $A0,8($nap_d) ; save a[j] in double format + stfd $A1,16($nap_d) + fmul $T2a,$A2,$ba + fmul $T2b,$A2,$bb + stfd $A2,24($nap_d) ; save a[j+1] in double format + stfd $A3,32($nap_d) + fmul $T3a,$A3,$ba + fmul $T3b,$A3,$bb + stfd $N0,40($nap_d) ; save n[j] in double format + stfd $N1,48($nap_d) + fmul $T0a,$A0,$ba + fmul $T0b,$A0,$bb + stfd $N2,56($nap_d) ; save n[j+1] in double format + stfdu $N3,64($nap_d) + + fmadd $T1a,$A0,$bc,$T1a + fmadd $T1b,$A0,$bd,$T1b + fmadd $T2a,$A1,$bc,$T2a + fmadd $T2b,$A1,$bd,$T2b + fmadd $T3a,$A2,$bc,$T3a + fmadd $T3b,$A2,$bd,$T3b + fmul $dota,$A3,$bc + fmul $dotb,$A3,$bd + + fmadd $T1a,$N1,$na,$T1a + fmadd $T1b,$N1,$nb,$T1b + fmadd $T2a,$N2,$na,$T2a + fmadd $T2b,$N2,$nb,$T2b + fmadd $T3a,$N3,$na,$T3a + fmadd $T3b,$N3,$nb,$T3b + fmadd $T0a,$N0,$na,$T0a + fmadd $T0b,$N0,$nb,$T0b + + fmadd $T1a,$N0,$nc,$T1a + fmadd $T1b,$N0,$nd,$T1b + fmadd $T2a,$N1,$nc,$T2a + fmadd $T2b,$N1,$nd,$T2b + fmadd $T3a,$N2,$nc,$T3a + fmadd $T3b,$N2,$nd,$T3b + fmadd $dota,$N3,$nc,$dota + fmadd $dotb,$N3,$nd,$dotb + + fctid $T0a,$T0a + fctid $T0b,$T0b + fctid $T1a,$T1a + fctid $T1b,$T1b + fctid $T2a,$T2a + fctid $T2b,$T2b + fctid $T3a,$T3a + fctid $T3b,$T3b + + stfd $T0a,`$FRAME+0`($sp) + stfd $T0b,`$FRAME+8`($sp) + stfd $T1a,`$FRAME+16`($sp) + stfd $T1b,`$FRAME+24`($sp) + stfd $T2a,`$FRAME+32`($sp) + stfd $T2b,`$FRAME+40`($sp) + stfd $T3a,`$FRAME+48`($sp) + stfd $T3b,`$FRAME+56`($sp) + +.align 5 +L1st: +___ +$code.=<<___ if ($SIZE_T==8); + lwz $t0,4($ap) ; load a[j] as 32-bit word pair + lwz $t1,0($ap) + lwz $t2,12($ap) ; load a[j+1] as 32-bit word pair + lwz $t3,8($ap) + lwz $t4,4($np) ; load n[j] as 32-bit word pair + lwz $t5,0($np) + lwz $t6,12($np) ; load n[j+1] as 32-bit word pair + lwz $t7,8($np) +___ +$code.=<<___ if ($SIZE_T==4); + lwz $t0,0($ap) ; load a[j..j+3] as 32-bit word pairs + lwz $t1,4($ap) + lwz $t2,8($ap) + lwz $t3,12($ap) + lwz $t4,0($np) ; load n[j..j+3] as 32-bit word pairs + lwz $t5,4($np) + lwz $t6,8($np) + lwz $t7,12($np) +___ +$code.=<<___; + std $t0,`$FRAME+64`($sp) + std $t1,`$FRAME+72`($sp) + std $t2,`$FRAME+80`($sp) + std $t3,`$FRAME+88`($sp) + std $t4,`$FRAME+96`($sp) + std $t5,`$FRAME+104`($sp) + std $t6,`$FRAME+112`($sp) + std $t7,`$FRAME+120`($sp) + ld $t0,`$FRAME+0`($sp) + ld $t1,`$FRAME+8`($sp) + ld $t2,`$FRAME+16`($sp) + ld $t3,`$FRAME+24`($sp) + ld $t4,`$FRAME+32`($sp) + ld $t5,`$FRAME+40`($sp) + ld $t6,`$FRAME+48`($sp) + ld $t7,`$FRAME+56`($sp) + lfd $A0,`$FRAME+64`($sp) + lfd $A1,`$FRAME+72`($sp) + lfd $A2,`$FRAME+80`($sp) + lfd $A3,`$FRAME+88`($sp) + lfd $N0,`$FRAME+96`($sp) + lfd $N1,`$FRAME+104`($sp) + lfd $N2,`$FRAME+112`($sp) + lfd $N3,`$FRAME+120`($sp) + fcfid $A0,$A0 + fcfid $A1,$A1 + fcfid $A2,$A2 + fcfid $A3,$A3 + fcfid $N0,$N0 + fcfid $N1,$N1 + fcfid $N2,$N2 + fcfid $N3,$N3 + addi $ap,$ap,16 + addi $np,$np,16 + + fmul $T1a,$A1,$ba + fmul $T1b,$A1,$bb + fmul $T2a,$A2,$ba + fmul $T2b,$A2,$bb + stfd $A0,8($nap_d) ; save a[j] in double format + stfd $A1,16($nap_d) + fmul $T3a,$A3,$ba + fmul $T3b,$A3,$bb + fmadd $T0a,$A0,$ba,$dota + fmadd $T0b,$A0,$bb,$dotb + stfd $A2,24($nap_d) ; save a[j+1] in double format + stfd $A3,32($nap_d) + + fmadd $T1a,$A0,$bc,$T1a + fmadd $T1b,$A0,$bd,$T1b + fmadd $T2a,$A1,$bc,$T2a + fmadd $T2b,$A1,$bd,$T2b + stfd $N0,40($nap_d) ; save n[j] in double format + stfd $N1,48($nap_d) + fmadd $T3a,$A2,$bc,$T3a + fmadd $T3b,$A2,$bd,$T3b + add $t0,$t0,$carry ; can not overflow + fmul $dota,$A3,$bc + fmul $dotb,$A3,$bd + stfd $N2,56($nap_d) ; save n[j+1] in double format + stfdu $N3,64($nap_d) + srdi $carry,$t0,16 + add $t1,$t1,$carry + srdi $carry,$t1,16 + + fmadd $T1a,$N1,$na,$T1a + fmadd $T1b,$N1,$nb,$T1b + insrdi $t0,$t1,16,32 + fmadd $T2a,$N2,$na,$T2a + fmadd $T2b,$N2,$nb,$T2b + add $t2,$t2,$carry + fmadd $T3a,$N3,$na,$T3a + fmadd $T3b,$N3,$nb,$T3b + srdi $carry,$t2,16 + fmadd $T0a,$N0,$na,$T0a + fmadd $T0b,$N0,$nb,$T0b + insrdi $t0,$t2,16,16 + add $t3,$t3,$carry + srdi $carry,$t3,16 + + fmadd $T1a,$N0,$nc,$T1a + fmadd $T1b,$N0,$nd,$T1b + insrdi $t0,$t3,16,0 ; 0..63 bits + fmadd $T2a,$N1,$nc,$T2a + fmadd $T2b,$N1,$nd,$T2b + add $t4,$t4,$carry + fmadd $T3a,$N2,$nc,$T3a + fmadd $T3b,$N2,$nd,$T3b + srdi $carry,$t4,16 + fmadd $dota,$N3,$nc,$dota + fmadd $dotb,$N3,$nd,$dotb + add $t5,$t5,$carry + srdi $carry,$t5,16 + insrdi $t4,$t5,16,32 + + fctid $T0a,$T0a + fctid $T0b,$T0b + add $t6,$t6,$carry + fctid $T1a,$T1a + fctid $T1b,$T1b + srdi $carry,$t6,16 + fctid $T2a,$T2a + fctid $T2b,$T2b + insrdi $t4,$t6,16,16 + fctid $T3a,$T3a + fctid $T3b,$T3b + add $t7,$t7,$carry + insrdi $t4,$t7,16,0 ; 64..127 bits + srdi $carry,$t7,16 ; upper 33 bits + + stfd $T0a,`$FRAME+0`($sp) + stfd $T0b,`$FRAME+8`($sp) + stfd $T1a,`$FRAME+16`($sp) + stfd $T1b,`$FRAME+24`($sp) + stfd $T2a,`$FRAME+32`($sp) + stfd $T2b,`$FRAME+40`($sp) + stfd $T3a,`$FRAME+48`($sp) + stfd $T3b,`$FRAME+56`($sp) + std $t0,8($tp) ; tp[j-1] + stdu $t4,16($tp) ; tp[j] + bdnz- L1st + + fctid $dota,$dota + fctid $dotb,$dotb + + ld $t0,`$FRAME+0`($sp) + ld $t1,`$FRAME+8`($sp) + ld $t2,`$FRAME+16`($sp) + ld $t3,`$FRAME+24`($sp) + ld $t4,`$FRAME+32`($sp) + ld $t5,`$FRAME+40`($sp) + ld $t6,`$FRAME+48`($sp) + ld $t7,`$FRAME+56`($sp) + stfd $dota,`$FRAME+64`($sp) + stfd $dotb,`$FRAME+72`($sp) + + add $t0,$t0,$carry ; can not overflow + srdi $carry,$t0,16 + add $t1,$t1,$carry + srdi $carry,$t1,16 + insrdi $t0,$t1,16,32 + add $t2,$t2,$carry + srdi $carry,$t2,16 + insrdi $t0,$t2,16,16 + add $t3,$t3,$carry + srdi $carry,$t3,16 + insrdi $t0,$t3,16,0 ; 0..63 bits + add $t4,$t4,$carry + srdi $carry,$t4,16 + add $t5,$t5,$carry + srdi $carry,$t5,16 + insrdi $t4,$t5,16,32 + add $t6,$t6,$carry + srdi $carry,$t6,16 + insrdi $t4,$t6,16,16 + add $t7,$t7,$carry + insrdi $t4,$t7,16,0 ; 64..127 bits + srdi $carry,$t7,16 ; upper 33 bits + ld $t6,`$FRAME+64`($sp) + ld $t7,`$FRAME+72`($sp) + + std $t0,8($tp) ; tp[j-1] + stdu $t4,16($tp) ; tp[j] + + add $t6,$t6,$carry ; can not overflow + srdi $carry,$t6,16 + add $t7,$t7,$carry + insrdi $t6,$t7,48,0 + srdi $ovf,$t7,48 + std $t6,8($tp) ; tp[num-1] + + slwi $t7,$num,2 + subf $nap_d,$t7,$nap_d ; rewind pointer + + li $i,8 ; i=1 +.align 5 +Louter: +___ +$code.=<<___ if ($SIZE_T==8); + ldx $t3,$bp,$i ; bp[i] +___ +$code.=<<___ if ($SIZE_T==4); + add $t0,$bp,$i + lwz $t3,0($t0) ; bp[i,i+1] + lwz $t0,4($t0) + insrdi $t3,$t0,32,0 +___ +$code.=<<___; + ld $t6,`$FRAME+$TRANSFER+8`($sp) ; tp[0] + mulld $t7,$a0,$t3 ; ap[0]*bp[i] + + addi $tp,$sp,`$FRAME+$TRANSFER` + add $t7,$t7,$t6 ; ap[0]*bp[i]+tp[0] + li $carry,0 + mulld $t7,$t7,$n0 ; tp[0]*n0 + mtctr $j + + ; transfer bp[i] to FPU as 4x16-bit values + extrdi $t0,$t3,16,48 + extrdi $t1,$t3,16,32 + extrdi $t2,$t3,16,16 + extrdi $t3,$t3,16,0 + std $t0,`$FRAME+0`($sp) + std $t1,`$FRAME+8`($sp) + std $t2,`$FRAME+16`($sp) + std $t3,`$FRAME+24`($sp) + ; transfer (ap[0]*bp[i]+tp[0])*n0 to FPU as 4x16-bit values + extrdi $t4,$t7,16,48 + extrdi $t5,$t7,16,32 + extrdi $t6,$t7,16,16 + extrdi $t7,$t7,16,0 + std $t4,`$FRAME+32`($sp) + std $t5,`$FRAME+40`($sp) + std $t6,`$FRAME+48`($sp) + std $t7,`$FRAME+56`($sp) + + lfd $A0,8($nap_d) ; load a[j] in double format + lfd $A1,16($nap_d) + lfd $A2,24($nap_d) ; load a[j+1] in double format + lfd $A3,32($nap_d) + lfd $N0,40($nap_d) ; load n[j] in double format + lfd $N1,48($nap_d) + lfd $N2,56($nap_d) ; load n[j+1] in double format + lfdu $N3,64($nap_d) + + lfd $ba,`$FRAME+0`($sp) + lfd $bb,`$FRAME+8`($sp) + lfd $bc,`$FRAME+16`($sp) + lfd $bd,`$FRAME+24`($sp) + lfd $na,`$FRAME+32`($sp) + lfd $nb,`$FRAME+40`($sp) + lfd $nc,`$FRAME+48`($sp) + lfd $nd,`$FRAME+56`($sp) + + fcfid $ba,$ba + fcfid $bb,$bb + fcfid $bc,$bc + fcfid $bd,$bd + fcfid $na,$na + fcfid $nb,$nb + fcfid $nc,$nc + fcfid $nd,$nd + + fmul $T1a,$A1,$ba + fmul $T1b,$A1,$bb + fmul $T2a,$A2,$ba + fmul $T2b,$A2,$bb + fmul $T3a,$A3,$ba + fmul $T3b,$A3,$bb + fmul $T0a,$A0,$ba + fmul $T0b,$A0,$bb + + fmadd $T1a,$A0,$bc,$T1a + fmadd $T1b,$A0,$bd,$T1b + fmadd $T2a,$A1,$bc,$T2a + fmadd $T2b,$A1,$bd,$T2b + fmadd $T3a,$A2,$bc,$T3a + fmadd $T3b,$A2,$bd,$T3b + fmul $dota,$A3,$bc + fmul $dotb,$A3,$bd + + fmadd $T1a,$N1,$na,$T1a + fmadd $T1b,$N1,$nb,$T1b + lfd $A0,8($nap_d) ; load a[j] in double format + lfd $A1,16($nap_d) + fmadd $T2a,$N2,$na,$T2a + fmadd $T2b,$N2,$nb,$T2b + lfd $A2,24($nap_d) ; load a[j+1] in double format + lfd $A3,32($nap_d) + fmadd $T3a,$N3,$na,$T3a + fmadd $T3b,$N3,$nb,$T3b + fmadd $T0a,$N0,$na,$T0a + fmadd $T0b,$N0,$nb,$T0b + + fmadd $T1a,$N0,$nc,$T1a + fmadd $T1b,$N0,$nd,$T1b + fmadd $T2a,$N1,$nc,$T2a + fmadd $T2b,$N1,$nd,$T2b + fmadd $T3a,$N2,$nc,$T3a + fmadd $T3b,$N2,$nd,$T3b + fmadd $dota,$N3,$nc,$dota + fmadd $dotb,$N3,$nd,$dotb + + fctid $T0a,$T0a + fctid $T0b,$T0b + fctid $T1a,$T1a + fctid $T1b,$T1b + fctid $T2a,$T2a + fctid $T2b,$T2b + fctid $T3a,$T3a + fctid $T3b,$T3b + + stfd $T0a,`$FRAME+0`($sp) + stfd $T0b,`$FRAME+8`($sp) + stfd $T1a,`$FRAME+16`($sp) + stfd $T1b,`$FRAME+24`($sp) + stfd $T2a,`$FRAME+32`($sp) + stfd $T2b,`$FRAME+40`($sp) + stfd $T3a,`$FRAME+48`($sp) + stfd $T3b,`$FRAME+56`($sp) + +.align 5 +Linner: + fmul $T1a,$A1,$ba + fmul $T1b,$A1,$bb + fmul $T2a,$A2,$ba + fmul $T2b,$A2,$bb + lfd $N0,40($nap_d) ; load n[j] in double format + lfd $N1,48($nap_d) + fmul $T3a,$A3,$ba + fmul $T3b,$A3,$bb + fmadd $T0a,$A0,$ba,$dota + fmadd $T0b,$A0,$bb,$dotb + lfd $N2,56($nap_d) ; load n[j+1] in double format + lfdu $N3,64($nap_d) + + fmadd $T1a,$A0,$bc,$T1a + fmadd $T1b,$A0,$bd,$T1b + fmadd $T2a,$A1,$bc,$T2a + fmadd $T2b,$A1,$bd,$T2b + lfd $A0,8($nap_d) ; load a[j] in double format + lfd $A1,16($nap_d) + fmadd $T3a,$A2,$bc,$T3a + fmadd $T3b,$A2,$bd,$T3b + fmul $dota,$A3,$bc + fmul $dotb,$A3,$bd + lfd $A2,24($nap_d) ; load a[j+1] in double format + lfd $A3,32($nap_d) + + fmadd $T1a,$N1,$na,$T1a + fmadd $T1b,$N1,$nb,$T1b + ld $t0,`$FRAME+0`($sp) + ld $t1,`$FRAME+8`($sp) + fmadd $T2a,$N2,$na,$T2a + fmadd $T2b,$N2,$nb,$T2b + ld $t2,`$FRAME+16`($sp) + ld $t3,`$FRAME+24`($sp) + fmadd $T3a,$N3,$na,$T3a + fmadd $T3b,$N3,$nb,$T3b + add $t0,$t0,$carry ; can not overflow + ld $t4,`$FRAME+32`($sp) + ld $t5,`$FRAME+40`($sp) + fmadd $T0a,$N0,$na,$T0a + fmadd $T0b,$N0,$nb,$T0b + srdi $carry,$t0,16 + add $t1,$t1,$carry + srdi $carry,$t1,16 + ld $t6,`$FRAME+48`($sp) + ld $t7,`$FRAME+56`($sp) + + fmadd $T1a,$N0,$nc,$T1a + fmadd $T1b,$N0,$nd,$T1b + insrdi $t0,$t1,16,32 + ld $t1,8($tp) ; tp[j] + fmadd $T2a,$N1,$nc,$T2a + fmadd $T2b,$N1,$nd,$T2b + add $t2,$t2,$carry + fmadd $T3a,$N2,$nc,$T3a + fmadd $T3b,$N2,$nd,$T3b + srdi $carry,$t2,16 + insrdi $t0,$t2,16,16 + fmadd $dota,$N3,$nc,$dota + fmadd $dotb,$N3,$nd,$dotb + add $t3,$t3,$carry + ldu $t2,16($tp) ; tp[j+1] + srdi $carry,$t3,16 + insrdi $t0,$t3,16,0 ; 0..63 bits + add $t4,$t4,$carry + + fctid $T0a,$T0a + fctid $T0b,$T0b + srdi $carry,$t4,16 + fctid $T1a,$T1a + fctid $T1b,$T1b + add $t5,$t5,$carry + fctid $T2a,$T2a + fctid $T2b,$T2b + srdi $carry,$t5,16 + insrdi $t4,$t5,16,32 + fctid $T3a,$T3a + fctid $T3b,$T3b + add $t6,$t6,$carry + srdi $carry,$t6,16 + insrdi $t4,$t6,16,16 + + stfd $T0a,`$FRAME+0`($sp) + stfd $T0b,`$FRAME+8`($sp) + add $t7,$t7,$carry + addc $t3,$t0,$t1 +___ +$code.=<<___ if ($SIZE_T==4); # adjust XER[CA] + extrdi $t0,$t0,32,0 + extrdi $t1,$t1,32,0 + adde $t0,$t0,$t1 +___ +$code.=<<___; + stfd $T1a,`$FRAME+16`($sp) + stfd $T1b,`$FRAME+24`($sp) + insrdi $t4,$t7,16,0 ; 64..127 bits + srdi $carry,$t7,16 ; upper 33 bits + stfd $T2a,`$FRAME+32`($sp) + stfd $T2b,`$FRAME+40`($sp) + adde $t5,$t4,$t2 +___ +$code.=<<___ if ($SIZE_T==4); # adjust XER[CA] + extrdi $t4,$t4,32,0 + extrdi $t2,$t2,32,0 + adde $t4,$t4,$t2 +___ +$code.=<<___; + stfd $T3a,`$FRAME+48`($sp) + stfd $T3b,`$FRAME+56`($sp) + addze $carry,$carry + std $t3,-16($tp) ; tp[j-1] + std $t5,-8($tp) ; tp[j] + bdnz- Linner + + fctid $dota,$dota + fctid $dotb,$dotb + ld $t0,`$FRAME+0`($sp) + ld $t1,`$FRAME+8`($sp) + ld $t2,`$FRAME+16`($sp) + ld $t3,`$FRAME+24`($sp) + ld $t4,`$FRAME+32`($sp) + ld $t5,`$FRAME+40`($sp) + ld $t6,`$FRAME+48`($sp) + ld $t7,`$FRAME+56`($sp) + stfd $dota,`$FRAME+64`($sp) + stfd $dotb,`$FRAME+72`($sp) + + add $t0,$t0,$carry ; can not overflow + srdi $carry,$t0,16 + add $t1,$t1,$carry + srdi $carry,$t1,16 + insrdi $t0,$t1,16,32 + add $t2,$t2,$carry + ld $t1,8($tp) ; tp[j] + srdi $carry,$t2,16 + insrdi $t0,$t2,16,16 + add $t3,$t3,$carry + ldu $t2,16($tp) ; tp[j+1] + srdi $carry,$t3,16 + insrdi $t0,$t3,16,0 ; 0..63 bits + add $t4,$t4,$carry + srdi $carry,$t4,16 + add $t5,$t5,$carry + srdi $carry,$t5,16 + insrdi $t4,$t5,16,32 + add $t6,$t6,$carry + srdi $carry,$t6,16 + insrdi $t4,$t6,16,16 + add $t7,$t7,$carry + insrdi $t4,$t7,16,0 ; 64..127 bits + srdi $carry,$t7,16 ; upper 33 bits + ld $t6,`$FRAME+64`($sp) + ld $t7,`$FRAME+72`($sp) + + addc $t3,$t0,$t1 +___ +$code.=<<___ if ($SIZE_T==4); # adjust XER[CA] + extrdi $t0,$t0,32,0 + extrdi $t1,$t1,32,0 + adde $t0,$t0,$t1 +___ +$code.=<<___; + adde $t5,$t4,$t2 +___ +$code.=<<___ if ($SIZE_T==4); # adjust XER[CA] + extrdi $t4,$t4,32,0 + extrdi $t2,$t2,32,0 + adde $t4,$t4,$t2 +___ +$code.=<<___; + addze $carry,$carry + + std $t3,-16($tp) ; tp[j-1] + std $t5,-8($tp) ; tp[j] + + add $carry,$carry,$ovf ; comsume upmost overflow + add $t6,$t6,$carry ; can not overflow + srdi $carry,$t6,16 + add $t7,$t7,$carry + insrdi $t6,$t7,48,0 + srdi $ovf,$t7,48 + std $t6,0($tp) ; tp[num-1] + + slwi $t7,$num,2 + addi $i,$i,8 + subf $nap_d,$t7,$nap_d ; rewind pointer + cmpw $i,$num + blt- Louter +___ + +$code.=<<___ if ($SIZE_T==8); + subf $np,$num,$np ; rewind np + addi $j,$j,1 ; restore counter + subfc $i,$i,$i ; j=0 and "clear" XER[CA] + addi $tp,$sp,`$FRAME+$TRANSFER+8` + addi $t4,$sp,`$FRAME+$TRANSFER+16` + addi $t5,$np,8 + addi $t6,$rp,8 + mtctr $j + +.align 4 +Lsub: ldx $t0,$tp,$i + ldx $t1,$np,$i + ldx $t2,$t4,$i + ldx $t3,$t5,$i + subfe $t0,$t1,$t0 ; tp[j]-np[j] + subfe $t2,$t3,$t2 ; tp[j+1]-np[j+1] + stdx $t0,$rp,$i + stdx $t2,$t6,$i + addi $i,$i,16 + bdnz- Lsub + + li $i,0 + subfe $ovf,$i,$ovf ; handle upmost overflow bit + and $ap,$tp,$ovf + andc $np,$rp,$ovf + or $ap,$ap,$np ; ap=borrow?tp:rp + addi $t7,$ap,8 + mtctr $j + +.align 4 +Lcopy: ; copy or in-place refresh + ldx $t0,$ap,$i + ldx $t1,$t7,$i + std $i,8($nap_d) ; zap nap_d + std $i,16($nap_d) + std $i,24($nap_d) + std $i,32($nap_d) + std $i,40($nap_d) + std $i,48($nap_d) + std $i,56($nap_d) + stdu $i,64($nap_d) + stdx $t0,$rp,$i + stdx $t1,$t6,$i + stdx $i,$tp,$i ; zap tp at once + stdx $i,$t4,$i + addi $i,$i,16 + bdnz- Lcopy +___ +$code.=<<___ if ($SIZE_T==4); + subf $np,$num,$np ; rewind np + addi $j,$j,1 ; restore counter + subfc $i,$i,$i ; j=0 and "clear" XER[CA] + addi $tp,$sp,`$FRAME+$TRANSFER` + addi $np,$np,-4 + addi $rp,$rp,-4 + addi $ap,$sp,`$FRAME+$TRANSFER+4` + mtctr $j + +.align 4 +Lsub: ld $t0,8($tp) ; load tp[j..j+3] in 64-bit word order + ldu $t2,16($tp) + lwz $t4,4($np) ; load np[j..j+3] in 32-bit word order + lwz $t5,8($np) + lwz $t6,12($np) + lwzu $t7,16($np) + extrdi $t1,$t0,32,0 + extrdi $t3,$t2,32,0 + subfe $t4,$t4,$t0 ; tp[j]-np[j] + stw $t0,4($ap) ; save tp[j..j+3] in 32-bit word order + subfe $t5,$t5,$t1 ; tp[j+1]-np[j+1] + stw $t1,8($ap) + subfe $t6,$t6,$t2 ; tp[j+2]-np[j+2] + stw $t2,12($ap) + subfe $t7,$t7,$t3 ; tp[j+3]-np[j+3] + stwu $t3,16($ap) + stw $t4,4($rp) + stw $t5,8($rp) + stw $t6,12($rp) + stwu $t7,16($rp) + bdnz- Lsub + + li $i,0 + subfe $ovf,$i,$ovf ; handle upmost overflow bit + addi $tp,$sp,`$FRAME+$TRANSFER+4` + subf $rp,$num,$rp ; rewind rp + and $ap,$tp,$ovf + andc $np,$rp,$ovf + or $ap,$ap,$np ; ap=borrow?tp:rp + addi $tp,$sp,`$FRAME+$TRANSFER` + mtctr $j + +.align 4 +Lcopy: ; copy or in-place refresh + lwz $t0,4($ap) + lwz $t1,8($ap) + lwz $t2,12($ap) + lwzu $t3,16($ap) + std $i,8($nap_d) ; zap nap_d + std $i,16($nap_d) + std $i,24($nap_d) + std $i,32($nap_d) + std $i,40($nap_d) + std $i,48($nap_d) + std $i,56($nap_d) + stdu $i,64($nap_d) + stw $t0,4($rp) + stw $t1,8($rp) + stw $t2,12($rp) + stwu $t3,16($rp) + std $i,8($tp) ; zap tp at once + stdu $i,16($tp) + bdnz- Lcopy +___ + +$code.=<<___; + $POP $i,0($sp) + li r3,1 ; signal "handled" + $POP r22,`-12*8-10*$SIZE_T`($i) + $POP r23,`-12*8-9*$SIZE_T`($i) + $POP r24,`-12*8-8*$SIZE_T`($i) + $POP r25,`-12*8-7*$SIZE_T`($i) + $POP r26,`-12*8-6*$SIZE_T`($i) + $POP r27,`-12*8-5*$SIZE_T`($i) + $POP r28,`-12*8-4*$SIZE_T`($i) + $POP r29,`-12*8-3*$SIZE_T`($i) + $POP r30,`-12*8-2*$SIZE_T`($i) + $POP r31,`-12*8-1*$SIZE_T`($i) + lfd f20,`-12*8`($i) + lfd f21,`-11*8`($i) + lfd f22,`-10*8`($i) + lfd f23,`-9*8`($i) + lfd f24,`-8*8`($i) + lfd f25,`-7*8`($i) + lfd f26,`-6*8`($i) + lfd f27,`-5*8`($i) + lfd f28,`-4*8`($i) + lfd f29,`-3*8`($i) + lfd f30,`-2*8`($i) + lfd f31,`-1*8`($i) + mr $sp,$i + blr + .long 0 + .byte 0,12,4,0,0x8c,10,6,0 + .long 0 + +.asciz "Montgomery Multiplication for PPC64, CRYPTOGAMS by " +___ + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x-gf2m.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x-gf2m.pl new file mode 100644 index 00000000..9d18d40e --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x-gf2m.pl @@ -0,0 +1,221 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# May 2011 +# +# The module implements bn_GF2m_mul_2x2 polynomial multiplication used +# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for +# the time being... gcc 4.3 appeared to generate poor code, therefore +# the effort. And indeed, the module delivers 55%-90%(*) improvement +# on haviest ECDSA verify and ECDH benchmarks for 163- and 571-bit +# key lengths on z990, 30%-55%(*) - on z10, and 70%-110%(*) - on z196. +# This is for 64-bit build. In 32-bit "highgprs" case improvement is +# even higher, for example on z990 it was measured 80%-150%. ECDSA +# sign is modest 9%-12% faster. Keep in mind that these coefficients +# are not ones for bn_GF2m_mul_2x2 itself, as not all CPU time is +# burnt in it... +# +# (*) gcc 4.1 was observed to deliver better results than gcc 4.3, +# so that improvement coefficients can vary from one specific +# setup to another. + +$flavour = shift; + +if ($flavour =~ /3[12]/) { + $SIZE_T=4; + $g=""; +} else { + $SIZE_T=8; + $g="g"; +} + +while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +$stdframe=16*$SIZE_T+4*8; + +$rp="%r2"; +$a1="%r3"; +$a0="%r4"; +$b1="%r5"; +$b0="%r6"; + +$ra="%r14"; +$sp="%r15"; + +@T=("%r0","%r1"); +@i=("%r12","%r13"); + +($a1,$a2,$a4,$a8,$a12,$a48)=map("%r$_",(6..11)); +($lo,$hi,$b)=map("%r$_",(3..5)); $a=$lo; $mask=$a8; + +$code.=<<___; +.text + +.type _mul_1x1,\@function +.align 16 +_mul_1x1: + lgr $a1,$a + sllg $a2,$a,1 + sllg $a4,$a,2 + sllg $a8,$a,3 + + srag $lo,$a1,63 # broadcast 63rd bit + nihh $a1,0x1fff + srag @i[0],$a2,63 # broadcast 62nd bit + nihh $a2,0x3fff + srag @i[1],$a4,63 # broadcast 61st bit + nihh $a4,0x7fff + ngr $lo,$b + ngr @i[0],$b + ngr @i[1],$b + + lghi @T[0],0 + lgr $a12,$a1 + stg @T[0],`$stdframe+0*8`($sp) # tab[0]=0 + xgr $a12,$a2 + stg $a1,`$stdframe+1*8`($sp) # tab[1]=a1 + lgr $a48,$a4 + stg $a2,`$stdframe+2*8`($sp) # tab[2]=a2 + xgr $a48,$a8 + stg $a12,`$stdframe+3*8`($sp) # tab[3]=a1^a2 + xgr $a1,$a4 + + stg $a4,`$stdframe+4*8`($sp) # tab[4]=a4 + xgr $a2,$a4 + stg $a1,`$stdframe+5*8`($sp) # tab[5]=a1^a4 + xgr $a12,$a4 + stg $a2,`$stdframe+6*8`($sp) # tab[6]=a2^a4 + xgr $a1,$a48 + stg $a12,`$stdframe+7*8`($sp) # tab[7]=a1^a2^a4 + xgr $a2,$a48 + + stg $a8,`$stdframe+8*8`($sp) # tab[8]=a8 + xgr $a12,$a48 + stg $a1,`$stdframe+9*8`($sp) # tab[9]=a1^a8 + xgr $a1,$a4 + stg $a2,`$stdframe+10*8`($sp) # tab[10]=a2^a8 + xgr $a2,$a4 + stg $a12,`$stdframe+11*8`($sp) # tab[11]=a1^a2^a8 + + xgr $a12,$a4 + stg $a48,`$stdframe+12*8`($sp) # tab[12]=a4^a8 + srlg $hi,$lo,1 + stg $a1,`$stdframe+13*8`($sp) # tab[13]=a1^a4^a8 + sllg $lo,$lo,63 + stg $a2,`$stdframe+14*8`($sp) # tab[14]=a2^a4^a8 + srlg @T[0],@i[0],2 + stg $a12,`$stdframe+15*8`($sp) # tab[15]=a1^a2^a4^a8 + + lghi $mask,`0xf<<3` + sllg $a1,@i[0],62 + sllg @i[0],$b,3 + srlg @T[1],@i[1],3 + ngr @i[0],$mask + sllg $a2,@i[1],61 + srlg @i[1],$b,4-3 + xgr $hi,@T[0] + ngr @i[1],$mask + xgr $lo,$a1 + xgr $hi,@T[1] + xgr $lo,$a2 + + xg $lo,$stdframe(@i[0],$sp) + srlg @i[0],$b,8-3 + ngr @i[0],$mask +___ +for($n=1;$n<14;$n++) { +$code.=<<___; + lg @T[1],$stdframe(@i[1],$sp) + srlg @i[1],$b,`($n+2)*4`-3 + sllg @T[0],@T[1],`$n*4` + ngr @i[1],$mask + srlg @T[1],@T[1],`64-$n*4` + xgr $lo,@T[0] + xgr $hi,@T[1] +___ + push(@i,shift(@i)); push(@T,shift(@T)); +} +$code.=<<___; + lg @T[1],$stdframe(@i[1],$sp) + sllg @T[0],@T[1],`$n*4` + srlg @T[1],@T[1],`64-$n*4` + xgr $lo,@T[0] + xgr $hi,@T[1] + + lg @T[0],$stdframe(@i[0],$sp) + sllg @T[1],@T[0],`($n+1)*4` + srlg @T[0],@T[0],`64-($n+1)*4` + xgr $lo,@T[1] + xgr $hi,@T[0] + + br $ra +.size _mul_1x1,.-_mul_1x1 + +.globl bn_GF2m_mul_2x2 +.type bn_GF2m_mul_2x2,\@function +.align 16 +bn_GF2m_mul_2x2: + stm${g} %r3,%r15,3*$SIZE_T($sp) + + lghi %r1,-$stdframe-128 + la %r0,0($sp) + la $sp,0(%r1,$sp) # alloca + st${g} %r0,0($sp) # back chain +___ +if ($SIZE_T==8) { +my @r=map("%r$_",(6..9)); +$code.=<<___; + bras $ra,_mul_1x1 # a1·b1 + stmg $lo,$hi,16($rp) + + lg $a,`$stdframe+128+4*$SIZE_T`($sp) + lg $b,`$stdframe+128+6*$SIZE_T`($sp) + bras $ra,_mul_1x1 # a0·b0 + stmg $lo,$hi,0($rp) + + lg $a,`$stdframe+128+3*$SIZE_T`($sp) + lg $b,`$stdframe+128+5*$SIZE_T`($sp) + xg $a,`$stdframe+128+4*$SIZE_T`($sp) + xg $b,`$stdframe+128+6*$SIZE_T`($sp) + bras $ra,_mul_1x1 # (a0+a1)·(b0+b1) + lmg @r[0],@r[3],0($rp) + + xgr $lo,$hi + xgr $hi,@r[1] + xgr $lo,@r[0] + xgr $hi,@r[2] + xgr $lo,@r[3] + xgr $hi,@r[3] + xgr $lo,$hi + stg $hi,16($rp) + stg $lo,8($rp) +___ +} else { +$code.=<<___; + sllg %r3,%r3,32 + sllg %r5,%r5,32 + or %r3,%r4 + or %r5,%r6 + bras $ra,_mul_1x1 + rllg $lo,$lo,32 + rllg $hi,$hi,32 + stmg $lo,$hi,0($rp) +___ +} +$code.=<<___; + lm${g} %r6,%r15,`$stdframe+128+6*$SIZE_T`($sp) + br $ra +.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2 +.string "GF(2^m) Multiplication for s390x, CRYPTOGAMS by " +___ + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x-mont.pl new file mode 100644 index 00000000..9fd64e81 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x-mont.pl @@ -0,0 +1,277 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# April 2007. +# +# Performance improvement over vanilla C code varies from 85% to 45% +# depending on key length and benchmark. Unfortunately in this context +# these are not very impressive results [for code that utilizes "wide" +# 64x64=128-bit multiplication, which is not commonly available to C +# programmers], at least hand-coded bn_asm.c replacement is known to +# provide 30-40% better results for longest keys. Well, on a second +# thought it's not very surprising, because z-CPUs are single-issue +# and _strictly_ in-order execution, while bn_mul_mont is more or less +# dependent on CPU ability to pipe-line instructions and have several +# of them "in-flight" at the same time. I mean while other methods, +# for example Karatsuba, aim to minimize amount of multiplications at +# the cost of other operations increase, bn_mul_mont aim to neatly +# "overlap" multiplications and the other operations [and on most +# platforms even minimize the amount of the other operations, in +# particular references to memory]. But it's possible to improve this +# module performance by implementing dedicated squaring code-path and +# possibly by unrolling loops... + +# January 2009. +# +# Reschedule to minimize/avoid Address Generation Interlock hazard, +# make inner loops counter-based. + +# November 2010. +# +# Adapt for -m31 build. If kernel supports what's called "highgprs" +# feature on Linux [see /proc/cpuinfo], it's possible to use 64-bit +# instructions and achieve "64-bit" performance even in 31-bit legacy +# application context. The feature is not specific to any particular +# processor, as long as it's "z-CPU". Latter implies that the code +# remains z/Architecture specific. Compatibility with 32-bit BN_ULONG +# is achieved by swapping words after 64-bit loads, follow _dswap-s. +# On z990 it was measured to perform 2.6-2.2 times better than +# compiler-generated code, less for longer keys... + +$flavour = shift; + +if ($flavour =~ /3[12]/) { + $SIZE_T=4; + $g=""; +} else { + $SIZE_T=8; + $g="g"; +} + +while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +$stdframe=16*$SIZE_T+4*8; + +$mn0="%r0"; +$num="%r1"; + +# int bn_mul_mont( +$rp="%r2"; # BN_ULONG *rp, +$ap="%r3"; # const BN_ULONG *ap, +$bp="%r4"; # const BN_ULONG *bp, +$np="%r5"; # const BN_ULONG *np, +$n0="%r6"; # const BN_ULONG *n0, +#$num="160(%r15)" # int num); + +$bi="%r2"; # zaps rp +$j="%r7"; + +$ahi="%r8"; +$alo="%r9"; +$nhi="%r10"; +$nlo="%r11"; +$AHI="%r12"; +$NHI="%r13"; +$count="%r14"; +$sp="%r15"; + +$code.=<<___; +.text +.globl bn_mul_mont +.type bn_mul_mont,\@function +bn_mul_mont: + lgf $num,`$stdframe+$SIZE_T-4`($sp) # pull $num + sla $num,`log($SIZE_T)/log(2)` # $num to enumerate bytes + la $bp,0($num,$bp) + + st${g} %r2,2*$SIZE_T($sp) + + cghi $num,16 # + lghi %r2,0 # + blr %r14 # if($num<16) return 0; +___ +$code.=<<___ if ($flavour =~ /3[12]/); + tmll $num,4 + bnzr %r14 # if ($num&1) return 0; +___ +$code.=<<___ if ($flavour !~ /3[12]/); + cghi $num,96 # + bhr %r14 # if($num>96) return 0; +___ +$code.=<<___; + stm${g} %r3,%r15,3*$SIZE_T($sp) + + lghi $rp,-$stdframe-8 # leave room for carry bit + lcgr $j,$num # -$num + lgr %r0,$sp + la $rp,0($rp,$sp) + la $sp,0($j,$rp) # alloca + st${g} %r0,0($sp) # back chain + + sra $num,3 # restore $num + la $bp,0($j,$bp) # restore $bp + ahi $num,-1 # adjust $num for inner loop + lg $n0,0($n0) # pull n0 + _dswap $n0 + + lg $bi,0($bp) + _dswap $bi + lg $alo,0($ap) + _dswap $alo + mlgr $ahi,$bi # ap[0]*bp[0] + lgr $AHI,$ahi + + lgr $mn0,$alo # "tp[0]"*n0 + msgr $mn0,$n0 + + lg $nlo,0($np) # + _dswap $nlo + mlgr $nhi,$mn0 # np[0]*m1 + algr $nlo,$alo # +="tp[0]" + lghi $NHI,0 + alcgr $NHI,$nhi + + la $j,8(%r0) # j=1 + lr $count,$num + +.align 16 +.L1st: + lg $alo,0($j,$ap) + _dswap $alo + mlgr $ahi,$bi # ap[j]*bp[0] + algr $alo,$AHI + lghi $AHI,0 + alcgr $AHI,$ahi + + lg $nlo,0($j,$np) + _dswap $nlo + mlgr $nhi,$mn0 # np[j]*m1 + algr $nlo,$NHI + lghi $NHI,0 + alcgr $nhi,$NHI # +="tp[j]" + algr $nlo,$alo + alcgr $NHI,$nhi + + stg $nlo,$stdframe-8($j,$sp) # tp[j-1]= + la $j,8($j) # j++ + brct $count,.L1st + + algr $NHI,$AHI + lghi $AHI,0 + alcgr $AHI,$AHI # upmost overflow bit + stg $NHI,$stdframe-8($j,$sp) + stg $AHI,$stdframe($j,$sp) + la $bp,8($bp) # bp++ + +.Louter: + lg $bi,0($bp) # bp[i] + _dswap $bi + lg $alo,0($ap) + _dswap $alo + mlgr $ahi,$bi # ap[0]*bp[i] + alg $alo,$stdframe($sp) # +=tp[0] + lghi $AHI,0 + alcgr $AHI,$ahi + + lgr $mn0,$alo + msgr $mn0,$n0 # tp[0]*n0 + + lg $nlo,0($np) # np[0] + _dswap $nlo + mlgr $nhi,$mn0 # np[0]*m1 + algr $nlo,$alo # +="tp[0]" + lghi $NHI,0 + alcgr $NHI,$nhi + + la $j,8(%r0) # j=1 + lr $count,$num + +.align 16 +.Linner: + lg $alo,0($j,$ap) + _dswap $alo + mlgr $ahi,$bi # ap[j]*bp[i] + algr $alo,$AHI + lghi $AHI,0 + alcgr $ahi,$AHI + alg $alo,$stdframe($j,$sp)# +=tp[j] + alcgr $AHI,$ahi + + lg $nlo,0($j,$np) + _dswap $nlo + mlgr $nhi,$mn0 # np[j]*m1 + algr $nlo,$NHI + lghi $NHI,0 + alcgr $nhi,$NHI + algr $nlo,$alo # +="tp[j]" + alcgr $NHI,$nhi + + stg $nlo,$stdframe-8($j,$sp) # tp[j-1]= + la $j,8($j) # j++ + brct $count,.Linner + + algr $NHI,$AHI + lghi $AHI,0 + alcgr $AHI,$AHI + alg $NHI,$stdframe($j,$sp)# accumulate previous upmost overflow bit + lghi $ahi,0 + alcgr $AHI,$ahi # new upmost overflow bit + stg $NHI,$stdframe-8($j,$sp) + stg $AHI,$stdframe($j,$sp) + + la $bp,8($bp) # bp++ + cl${g} $bp,`$stdframe+8+4*$SIZE_T`($j,$sp) # compare to &bp[num] + jne .Louter + + l${g} $rp,`$stdframe+8+2*$SIZE_T`($j,$sp) # reincarnate rp + la $ap,$stdframe($sp) + ahi $num,1 # restore $num, incidentally clears "borrow" + + la $j,0(%r0) + lr $count,$num +.Lsub: lg $alo,0($j,$ap) + lg $nlo,0($j,$np) + _dswap $nlo + slbgr $alo,$nlo + stg $alo,0($j,$rp) + la $j,8($j) + brct $count,.Lsub + lghi $ahi,0 + slbgr $AHI,$ahi # handle upmost carry + + ngr $ap,$AHI + lghi $np,-1 + xgr $np,$AHI + ngr $np,$rp + ogr $ap,$np # ap=borrow?tp:rp + + la $j,0(%r0) + lgr $count,$num +.Lcopy: lg $alo,0($j,$ap) # copy or in-place refresh + _dswap $alo + stg $j,$stdframe($j,$sp) # zap tp + stg $alo,0($j,$rp) + la $j,8($j) + brct $count,.Lcopy + + la %r1,`$stdframe+8+6*$SIZE_T`($j,$sp) + lm${g} %r6,%r15,0(%r1) + lghi %r2,1 # signal "processed" + br %r14 +.size bn_mul_mont,.-bn_mul_mont +.string "Montgomery Multiplication for s390x, CRYPTOGAMS by " +___ + +foreach (split("\n",$code)) { + s/\`([^\`]*)\`/eval $1/ge; + s/_dswap\s+(%r[0-9]+)/sprintf("rllg\t%s,%s,32",$1,$1) if($SIZE_T==4)/e; + print $_,"\n"; +} +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x.S new file mode 100755 index 00000000..43fcb79b --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/s390x.S @@ -0,0 +1,678 @@ +.ident "s390x.S, version 1.1" +// ==================================================================== +// Written by Andy Polyakov for the OpenSSL +// project. +// +// Rights for redistribution and usage in source and binary forms are +// granted according to the OpenSSL license. Warranty of any kind is +// disclaimed. +// ==================================================================== + +.text + +#define zero %r0 + +// BN_ULONG bn_mul_add_words(BN_ULONG *r2,BN_ULONG *r3,int r4,BN_ULONG r5); +.globl bn_mul_add_words +.type bn_mul_add_words,@function +.align 4 +bn_mul_add_words: + lghi zero,0 // zero = 0 + la %r1,0(%r2) // put rp aside + lghi %r2,0 // i=0; + ltgfr %r4,%r4 + bler %r14 // if (len<=0) return 0; + + stmg %r6,%r10,48(%r15) + lghi %r10,3 + lghi %r8,0 // carry = 0 + nr %r10,%r4 // len%4 + sra %r4,2 // cnt=len/4 + jz .Loop1_madd // carry is incidentally cleared if branch taken + algr zero,zero // clear carry + +.Loop4_madd: + lg %r7,0(%r2,%r3) // ap[i] + mlgr %r6,%r5 // *=w + alcgr %r7,%r8 // +=carry + alcgr %r6,zero + alg %r7,0(%r2,%r1) // +=rp[i] + stg %r7,0(%r2,%r1) // rp[i]= + + lg %r9,8(%r2,%r3) + mlgr %r8,%r5 + alcgr %r9,%r6 + alcgr %r8,zero + alg %r9,8(%r2,%r1) + stg %r9,8(%r2,%r1) + + lg %r7,16(%r2,%r3) + mlgr %r6,%r5 + alcgr %r7,%r8 + alcgr %r6,zero + alg %r7,16(%r2,%r1) + stg %r7,16(%r2,%r1) + + lg %r9,24(%r2,%r3) + mlgr %r8,%r5 + alcgr %r9,%r6 + alcgr %r8,zero + alg %r9,24(%r2,%r1) + stg %r9,24(%r2,%r1) + + la %r2,32(%r2) // i+=4 + brct %r4,.Loop4_madd + + la %r10,1(%r10) // see if len%4 is zero ... + brct %r10,.Loop1_madd // without touching condition code:-) + +.Lend_madd: + alcgr %r8,zero // collect carry bit + lgr %r2,%r8 + lmg %r6,%r10,48(%r15) + br %r14 + +.Loop1_madd: + lg %r7,0(%r2,%r3) // ap[i] + mlgr %r6,%r5 // *=w + alcgr %r7,%r8 // +=carry + alcgr %r6,zero + alg %r7,0(%r2,%r1) // +=rp[i] + stg %r7,0(%r2,%r1) // rp[i]= + + lgr %r8,%r6 + la %r2,8(%r2) // i++ + brct %r10,.Loop1_madd + + j .Lend_madd +.size bn_mul_add_words,.-bn_mul_add_words + +// BN_ULONG bn_mul_words(BN_ULONG *r2,BN_ULONG *r3,int r4,BN_ULONG r5); +.globl bn_mul_words +.type bn_mul_words,@function +.align 4 +bn_mul_words: + lghi zero,0 // zero = 0 + la %r1,0(%r2) // put rp aside + lghi %r2,0 // i=0; + ltgfr %r4,%r4 + bler %r14 // if (len<=0) return 0; + + stmg %r6,%r10,48(%r15) + lghi %r10,3 + lghi %r8,0 // carry = 0 + nr %r10,%r4 // len%4 + sra %r4,2 // cnt=len/4 + jz .Loop1_mul // carry is incidentally cleared if branch taken + algr zero,zero // clear carry + +.Loop4_mul: + lg %r7,0(%r2,%r3) // ap[i] + mlgr %r6,%r5 // *=w + alcgr %r7,%r8 // +=carry + stg %r7,0(%r2,%r1) // rp[i]= + + lg %r9,8(%r2,%r3) + mlgr %r8,%r5 + alcgr %r9,%r6 + stg %r9,8(%r2,%r1) + + lg %r7,16(%r2,%r3) + mlgr %r6,%r5 + alcgr %r7,%r8 + stg %r7,16(%r2,%r1) + + lg %r9,24(%r2,%r3) + mlgr %r8,%r5 + alcgr %r9,%r6 + stg %r9,24(%r2,%r1) + + la %r2,32(%r2) // i+=4 + brct %r4,.Loop4_mul + + la %r10,1(%r10) // see if len%4 is zero ... + brct %r10,.Loop1_mul // without touching condition code:-) + +.Lend_mul: + alcgr %r8,zero // collect carry bit + lgr %r2,%r8 + lmg %r6,%r10,48(%r15) + br %r14 + +.Loop1_mul: + lg %r7,0(%r2,%r3) // ap[i] + mlgr %r6,%r5 // *=w + alcgr %r7,%r8 // +=carry + stg %r7,0(%r2,%r1) // rp[i]= + + lgr %r8,%r6 + la %r2,8(%r2) // i++ + brct %r10,.Loop1_mul + + j .Lend_mul +.size bn_mul_words,.-bn_mul_words + +// void bn_sqr_words(BN_ULONG *r2,BN_ULONG *r2,int r4) +.globl bn_sqr_words +.type bn_sqr_words,@function +.align 4 +bn_sqr_words: + ltgfr %r4,%r4 + bler %r14 + + stmg %r6,%r7,48(%r15) + srag %r1,%r4,2 // cnt=len/4 + jz .Loop1_sqr + +.Loop4_sqr: + lg %r7,0(%r3) + mlgr %r6,%r7 + stg %r7,0(%r2) + stg %r6,8(%r2) + + lg %r7,8(%r3) + mlgr %r6,%r7 + stg %r7,16(%r2) + stg %r6,24(%r2) + + lg %r7,16(%r3) + mlgr %r6,%r7 + stg %r7,32(%r2) + stg %r6,40(%r2) + + lg %r7,24(%r3) + mlgr %r6,%r7 + stg %r7,48(%r2) + stg %r6,56(%r2) + + la %r3,32(%r3) + la %r2,64(%r2) + brct %r1,.Loop4_sqr + + lghi %r1,3 + nr %r4,%r1 // cnt=len%4 + jz .Lend_sqr + +.Loop1_sqr: + lg %r7,0(%r3) + mlgr %r6,%r7 + stg %r7,0(%r2) + stg %r6,8(%r2) + + la %r3,8(%r3) + la %r2,16(%r2) + brct %r4,.Loop1_sqr + +.Lend_sqr: + lmg %r6,%r7,48(%r15) + br %r14 +.size bn_sqr_words,.-bn_sqr_words + +// BN_ULONG bn_div_words(BN_ULONG h,BN_ULONG l,BN_ULONG d); +.globl bn_div_words +.type bn_div_words,@function +.align 4 +bn_div_words: + dlgr %r2,%r4 + lgr %r2,%r3 + br %r14 +.size bn_div_words,.-bn_div_words + +// BN_ULONG bn_add_words(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4,int r5); +.globl bn_add_words +.type bn_add_words,@function +.align 4 +bn_add_words: + la %r1,0(%r2) // put rp aside + lghi %r2,0 // i=0 + ltgfr %r5,%r5 + bler %r14 // if (len<=0) return 0; + + stg %r6,48(%r15) + lghi %r6,3 + nr %r6,%r5 // len%4 + sra %r5,2 // len/4, use sra because it sets condition code + jz .Loop1_add // carry is incidentally cleared if branch taken + algr %r2,%r2 // clear carry + +.Loop4_add: + lg %r0,0(%r2,%r3) + alcg %r0,0(%r2,%r4) + stg %r0,0(%r2,%r1) + lg %r0,8(%r2,%r3) + alcg %r0,8(%r2,%r4) + stg %r0,8(%r2,%r1) + lg %r0,16(%r2,%r3) + alcg %r0,16(%r2,%r4) + stg %r0,16(%r2,%r1) + lg %r0,24(%r2,%r3) + alcg %r0,24(%r2,%r4) + stg %r0,24(%r2,%r1) + + la %r2,32(%r2) // i+=4 + brct %r5,.Loop4_add + + la %r6,1(%r6) // see if len%4 is zero ... + brct %r6,.Loop1_add // without touching condition code:-) + +.Lexit_add: + lghi %r2,0 + alcgr %r2,%r2 + lg %r6,48(%r15) + br %r14 + +.Loop1_add: + lg %r0,0(%r2,%r3) + alcg %r0,0(%r2,%r4) + stg %r0,0(%r2,%r1) + + la %r2,8(%r2) // i++ + brct %r6,.Loop1_add + + j .Lexit_add +.size bn_add_words,.-bn_add_words + +// BN_ULONG bn_sub_words(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4,int r5); +.globl bn_sub_words +.type bn_sub_words,@function +.align 4 +bn_sub_words: + la %r1,0(%r2) // put rp aside + lghi %r2,0 // i=0 + ltgfr %r5,%r5 + bler %r14 // if (len<=0) return 0; + + stg %r6,48(%r15) + lghi %r6,3 + nr %r6,%r5 // len%4 + sra %r5,2 // len/4, use sra because it sets condition code + jnz .Loop4_sub // borrow is incidentally cleared if branch taken + slgr %r2,%r2 // clear borrow + +.Loop1_sub: + lg %r0,0(%r2,%r3) + slbg %r0,0(%r2,%r4) + stg %r0,0(%r2,%r1) + + la %r2,8(%r2) // i++ + brct %r6,.Loop1_sub + j .Lexit_sub + +.Loop4_sub: + lg %r0,0(%r2,%r3) + slbg %r0,0(%r2,%r4) + stg %r0,0(%r2,%r1) + lg %r0,8(%r2,%r3) + slbg %r0,8(%r2,%r4) + stg %r0,8(%r2,%r1) + lg %r0,16(%r2,%r3) + slbg %r0,16(%r2,%r4) + stg %r0,16(%r2,%r1) + lg %r0,24(%r2,%r3) + slbg %r0,24(%r2,%r4) + stg %r0,24(%r2,%r1) + + la %r2,32(%r2) // i+=4 + brct %r5,.Loop4_sub + + la %r6,1(%r6) // see if len%4 is zero ... + brct %r6,.Loop1_sub // without touching condition code:-) + +.Lexit_sub: + lghi %r2,0 + slbgr %r2,%r2 + lcgr %r2,%r2 + lg %r6,48(%r15) + br %r14 +.size bn_sub_words,.-bn_sub_words + +#define c1 %r1 +#define c2 %r5 +#define c3 %r8 + +#define mul_add_c(ai,bi,c1,c2,c3) \ + lg %r7,ai*8(%r3); \ + mlg %r6,bi*8(%r4); \ + algr c1,%r7; \ + alcgr c2,%r6; \ + alcgr c3,zero + +// void bn_mul_comba8(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4); +.globl bn_mul_comba8 +.type bn_mul_comba8,@function +.align 4 +bn_mul_comba8: + stmg %r6,%r8,48(%r15) + + lghi c1,0 + lghi c2,0 + lghi c3,0 + lghi zero,0 + + mul_add_c(0,0,c1,c2,c3); + stg c1,0*8(%r2) + lghi c1,0 + + mul_add_c(0,1,c2,c3,c1); + mul_add_c(1,0,c2,c3,c1); + stg c2,1*8(%r2) + lghi c2,0 + + mul_add_c(2,0,c3,c1,c2); + mul_add_c(1,1,c3,c1,c2); + mul_add_c(0,2,c3,c1,c2); + stg c3,2*8(%r2) + lghi c3,0 + + mul_add_c(0,3,c1,c2,c3); + mul_add_c(1,2,c1,c2,c3); + mul_add_c(2,1,c1,c2,c3); + mul_add_c(3,0,c1,c2,c3); + stg c1,3*8(%r2) + lghi c1,0 + + mul_add_c(4,0,c2,c3,c1); + mul_add_c(3,1,c2,c3,c1); + mul_add_c(2,2,c2,c3,c1); + mul_add_c(1,3,c2,c3,c1); + mul_add_c(0,4,c2,c3,c1); + stg c2,4*8(%r2) + lghi c2,0 + + mul_add_c(0,5,c3,c1,c2); + mul_add_c(1,4,c3,c1,c2); + mul_add_c(2,3,c3,c1,c2); + mul_add_c(3,2,c3,c1,c2); + mul_add_c(4,1,c3,c1,c2); + mul_add_c(5,0,c3,c1,c2); + stg c3,5*8(%r2) + lghi c3,0 + + mul_add_c(6,0,c1,c2,c3); + mul_add_c(5,1,c1,c2,c3); + mul_add_c(4,2,c1,c2,c3); + mul_add_c(3,3,c1,c2,c3); + mul_add_c(2,4,c1,c2,c3); + mul_add_c(1,5,c1,c2,c3); + mul_add_c(0,6,c1,c2,c3); + stg c1,6*8(%r2) + lghi c1,0 + + mul_add_c(0,7,c2,c3,c1); + mul_add_c(1,6,c2,c3,c1); + mul_add_c(2,5,c2,c3,c1); + mul_add_c(3,4,c2,c3,c1); + mul_add_c(4,3,c2,c3,c1); + mul_add_c(5,2,c2,c3,c1); + mul_add_c(6,1,c2,c3,c1); + mul_add_c(7,0,c2,c3,c1); + stg c2,7*8(%r2) + lghi c2,0 + + mul_add_c(7,1,c3,c1,c2); + mul_add_c(6,2,c3,c1,c2); + mul_add_c(5,3,c3,c1,c2); + mul_add_c(4,4,c3,c1,c2); + mul_add_c(3,5,c3,c1,c2); + mul_add_c(2,6,c3,c1,c2); + mul_add_c(1,7,c3,c1,c2); + stg c3,8*8(%r2) + lghi c3,0 + + mul_add_c(2,7,c1,c2,c3); + mul_add_c(3,6,c1,c2,c3); + mul_add_c(4,5,c1,c2,c3); + mul_add_c(5,4,c1,c2,c3); + mul_add_c(6,3,c1,c2,c3); + mul_add_c(7,2,c1,c2,c3); + stg c1,9*8(%r2) + lghi c1,0 + + mul_add_c(7,3,c2,c3,c1); + mul_add_c(6,4,c2,c3,c1); + mul_add_c(5,5,c2,c3,c1); + mul_add_c(4,6,c2,c3,c1); + mul_add_c(3,7,c2,c3,c1); + stg c2,10*8(%r2) + lghi c2,0 + + mul_add_c(4,7,c3,c1,c2); + mul_add_c(5,6,c3,c1,c2); + mul_add_c(6,5,c3,c1,c2); + mul_add_c(7,4,c3,c1,c2); + stg c3,11*8(%r2) + lghi c3,0 + + mul_add_c(7,5,c1,c2,c3); + mul_add_c(6,6,c1,c2,c3); + mul_add_c(5,7,c1,c2,c3); + stg c1,12*8(%r2) + lghi c1,0 + + + mul_add_c(6,7,c2,c3,c1); + mul_add_c(7,6,c2,c3,c1); + stg c2,13*8(%r2) + lghi c2,0 + + mul_add_c(7,7,c3,c1,c2); + stg c3,14*8(%r2) + stg c1,15*8(%r2) + + lmg %r6,%r8,48(%r15) + br %r14 +.size bn_mul_comba8,.-bn_mul_comba8 + +// void bn_mul_comba4(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4); +.globl bn_mul_comba4 +.type bn_mul_comba4,@function +.align 4 +bn_mul_comba4: + stmg %r6,%r8,48(%r15) + + lghi c1,0 + lghi c2,0 + lghi c3,0 + lghi zero,0 + + mul_add_c(0,0,c1,c2,c3); + stg c1,0*8(%r3) + lghi c1,0 + + mul_add_c(0,1,c2,c3,c1); + mul_add_c(1,0,c2,c3,c1); + stg c2,1*8(%r2) + lghi c2,0 + + mul_add_c(2,0,c3,c1,c2); + mul_add_c(1,1,c3,c1,c2); + mul_add_c(0,2,c3,c1,c2); + stg c3,2*8(%r2) + lghi c3,0 + + mul_add_c(0,3,c1,c2,c3); + mul_add_c(1,2,c1,c2,c3); + mul_add_c(2,1,c1,c2,c3); + mul_add_c(3,0,c1,c2,c3); + stg c1,3*8(%r2) + lghi c1,0 + + mul_add_c(3,1,c2,c3,c1); + mul_add_c(2,2,c2,c3,c1); + mul_add_c(1,3,c2,c3,c1); + stg c2,4*8(%r2) + lghi c2,0 + + mul_add_c(2,3,c3,c1,c2); + mul_add_c(3,2,c3,c1,c2); + stg c3,5*8(%r2) + lghi c3,0 + + mul_add_c(3,3,c1,c2,c3); + stg c1,6*8(%r2) + stg c2,7*8(%r2) + + stmg %r6,%r8,48(%r15) + br %r14 +.size bn_mul_comba4,.-bn_mul_comba4 + +#define sqr_add_c(ai,c1,c2,c3) \ + lg %r7,ai*8(%r3); \ + mlgr %r6,%r7; \ + algr c1,%r7; \ + alcgr c2,%r6; \ + alcgr c3,zero + +#define sqr_add_c2(ai,aj,c1,c2,c3) \ + lg %r7,ai*8(%r3); \ + mlg %r6,aj*8(%r3); \ + algr c1,%r7; \ + alcgr c2,%r6; \ + alcgr c3,zero; \ + algr c1,%r7; \ + alcgr c2,%r6; \ + alcgr c3,zero + +// void bn_sqr_comba8(BN_ULONG *r2,BN_ULONG *r3); +.globl bn_sqr_comba8 +.type bn_sqr_comba8,@function +.align 4 +bn_sqr_comba8: + stmg %r6,%r8,48(%r15) + + lghi c1,0 + lghi c2,0 + lghi c3,0 + lghi zero,0 + + sqr_add_c(0,c1,c2,c3); + stg c1,0*8(%r2) + lghi c1,0 + + sqr_add_c2(1,0,c2,c3,c1); + stg c2,1*8(%r2) + lghi c2,0 + + sqr_add_c(1,c3,c1,c2); + sqr_add_c2(2,0,c3,c1,c2); + stg c3,2*8(%r2) + lghi c3,0 + + sqr_add_c2(3,0,c1,c2,c3); + sqr_add_c2(2,1,c1,c2,c3); + stg c1,3*8(%r2) + lghi c1,0 + + sqr_add_c(2,c2,c3,c1); + sqr_add_c2(3,1,c2,c3,c1); + sqr_add_c2(4,0,c2,c3,c1); + stg c2,4*8(%r2) + lghi c2,0 + + sqr_add_c2(5,0,c3,c1,c2); + sqr_add_c2(4,1,c3,c1,c2); + sqr_add_c2(3,2,c3,c1,c2); + stg c3,5*8(%r2) + lghi c3,0 + + sqr_add_c(3,c1,c2,c3); + sqr_add_c2(4,2,c1,c2,c3); + sqr_add_c2(5,1,c1,c2,c3); + sqr_add_c2(6,0,c1,c2,c3); + stg c1,6*8(%r2) + lghi c1,0 + + sqr_add_c2(7,0,c2,c3,c1); + sqr_add_c2(6,1,c2,c3,c1); + sqr_add_c2(5,2,c2,c3,c1); + sqr_add_c2(4,3,c2,c3,c1); + stg c2,7*8(%r2) + lghi c2,0 + + sqr_add_c(4,c3,c1,c2); + sqr_add_c2(5,3,c3,c1,c2); + sqr_add_c2(6,2,c3,c1,c2); + sqr_add_c2(7,1,c3,c1,c2); + stg c3,8*8(%r2) + lghi c3,0 + + sqr_add_c2(7,2,c1,c2,c3); + sqr_add_c2(6,3,c1,c2,c3); + sqr_add_c2(5,4,c1,c2,c3); + stg c1,9*8(%r2) + lghi c1,0 + + sqr_add_c(5,c2,c3,c1); + sqr_add_c2(6,4,c2,c3,c1); + sqr_add_c2(7,3,c2,c3,c1); + stg c2,10*8(%r2) + lghi c2,0 + + sqr_add_c2(7,4,c3,c1,c2); + sqr_add_c2(6,5,c3,c1,c2); + stg c3,11*8(%r2) + lghi c3,0 + + sqr_add_c(6,c1,c2,c3); + sqr_add_c2(7,5,c1,c2,c3); + stg c1,12*8(%r2) + lghi c1,0 + + sqr_add_c2(7,6,c2,c3,c1); + stg c2,13*8(%r2) + lghi c2,0 + + sqr_add_c(7,c3,c1,c2); + stg c3,14*8(%r2) + stg c1,15*8(%r2) + + lmg %r6,%r8,48(%r15) + br %r14 +.size bn_sqr_comba8,.-bn_sqr_comba8 + +// void bn_sqr_comba4(BN_ULONG *r2,BN_ULONG *r3); +.globl bn_sqr_comba4 +.type bn_sqr_comba4,@function +.align 4 +bn_sqr_comba4: + stmg %r6,%r8,48(%r15) + + lghi c1,0 + lghi c2,0 + lghi c3,0 + lghi zero,0 + + sqr_add_c(0,c1,c2,c3); + stg c1,0*8(%r2) + lghi c1,0 + + sqr_add_c2(1,0,c2,c3,c1); + stg c2,1*8(%r2) + lghi c2,0 + + sqr_add_c(1,c3,c1,c2); + sqr_add_c2(2,0,c3,c1,c2); + stg c3,2*8(%r2) + lghi c3,0 + + sqr_add_c2(3,0,c1,c2,c3); + sqr_add_c2(2,1,c1,c2,c3); + stg c1,3*8(%r2) + lghi c1,0 + + sqr_add_c(2,c2,c3,c1); + sqr_add_c2(3,1,c2,c3,c1); + stg c2,4*8(%r2) + lghi c2,0 + + sqr_add_c2(3,2,c3,c1,c2); + stg c3,5*8(%r2) + lghi c3,0 + + sqr_add_c(3,c1,c2,c3); + stg c1,6*8(%r2) + stg c2,7*8(%r2) + + lmg %r6,%r8,48(%r15) + br %r14 +.size bn_sqr_comba4,.-bn_sqr_comba4 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv8.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv8.S new file mode 100644 index 00000000..88c5dc48 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv8.S @@ -0,0 +1,1458 @@ +.ident "sparcv8.s, Version 1.4" +.ident "SPARC v8 ISA artwork by Andy Polyakov " + +/* + * ==================================================================== + * Written by Andy Polyakov for the OpenSSL + * project. + * + * Rights for redistribution and usage in source and binary forms are + * granted according to the OpenSSL license. Warranty of any kind is + * disclaimed. + * ==================================================================== + */ + +/* + * This is my modest contributon to OpenSSL project (see + * http://www.openssl.org/ for more information about it) and is + * a drop-in SuperSPARC ISA replacement for crypto/bn/bn_asm.c + * module. For updates see http://fy.chalmers.se/~appro/hpe/. + * + * See bn_asm.sparc.v8plus.S for more details. + */ + +/* + * Revision history. + * + * 1.1 - new loop unrolling model(*); + * 1.2 - made gas friendly; + * 1.3 - fixed problem with /usr/ccs/lib/cpp; + * 1.4 - some retunes; + * + * (*) see bn_asm.sparc.v8plus.S for details + */ + +.section ".text",#alloc,#execinstr +.file "bn_asm.sparc.v8.S" + +.align 32 + +.global bn_mul_add_words +/* + * BN_ULONG bn_mul_add_words(rp,ap,num,w) + * BN_ULONG *rp,*ap; + * int num; + * BN_ULONG w; + */ +bn_mul_add_words: + cmp %o2,0 + bg,a .L_bn_mul_add_words_proceed + ld [%o1],%g2 + retl + clr %o0 + +.L_bn_mul_add_words_proceed: + andcc %o2,-4,%g0 + bz .L_bn_mul_add_words_tail + clr %o5 + +.L_bn_mul_add_words_loop: + ld [%o0],%o4 + ld [%o1+4],%g3 + umul %o3,%g2,%g2 + rd %y,%g1 + addcc %o4,%o5,%o4 + addx %g1,0,%g1 + addcc %o4,%g2,%o4 + st %o4,[%o0] + addx %g1,0,%o5 + + ld [%o0+4],%o4 + ld [%o1+8],%g2 + umul %o3,%g3,%g3 + dec 4,%o2 + rd %y,%g1 + addcc %o4,%o5,%o4 + addx %g1,0,%g1 + addcc %o4,%g3,%o4 + st %o4,[%o0+4] + addx %g1,0,%o5 + + ld [%o0+8],%o4 + ld [%o1+12],%g3 + umul %o3,%g2,%g2 + inc 16,%o1 + rd %y,%g1 + addcc %o4,%o5,%o4 + addx %g1,0,%g1 + addcc %o4,%g2,%o4 + st %o4,[%o0+8] + addx %g1,0,%o5 + + ld [%o0+12],%o4 + umul %o3,%g3,%g3 + inc 16,%o0 + rd %y,%g1 + addcc %o4,%o5,%o4 + addx %g1,0,%g1 + addcc %o4,%g3,%o4 + st %o4,[%o0-4] + addx %g1,0,%o5 + andcc %o2,-4,%g0 + bnz,a .L_bn_mul_add_words_loop + ld [%o1],%g2 + + tst %o2 + bnz,a .L_bn_mul_add_words_tail + ld [%o1],%g2 +.L_bn_mul_add_words_return: + retl + mov %o5,%o0 + nop + +.L_bn_mul_add_words_tail: + ld [%o0],%o4 + umul %o3,%g2,%g2 + addcc %o4,%o5,%o4 + rd %y,%g1 + addx %g1,0,%g1 + addcc %o4,%g2,%o4 + addx %g1,0,%o5 + deccc %o2 + bz .L_bn_mul_add_words_return + st %o4,[%o0] + + ld [%o1+4],%g2 + ld [%o0+4],%o4 + umul %o3,%g2,%g2 + rd %y,%g1 + addcc %o4,%o5,%o4 + addx %g1,0,%g1 + addcc %o4,%g2,%o4 + addx %g1,0,%o5 + deccc %o2 + bz .L_bn_mul_add_words_return + st %o4,[%o0+4] + + ld [%o1+8],%g2 + ld [%o0+8],%o4 + umul %o3,%g2,%g2 + rd %y,%g1 + addcc %o4,%o5,%o4 + addx %g1,0,%g1 + addcc %o4,%g2,%o4 + st %o4,[%o0+8] + retl + addx %g1,0,%o0 + +.type bn_mul_add_words,#function +.size bn_mul_add_words,(.-bn_mul_add_words) + +.align 32 + +.global bn_mul_words +/* + * BN_ULONG bn_mul_words(rp,ap,num,w) + * BN_ULONG *rp,*ap; + * int num; + * BN_ULONG w; + */ +bn_mul_words: + cmp %o2,0 + bg,a .L_bn_mul_words_proceeed + ld [%o1],%g2 + retl + clr %o0 + +.L_bn_mul_words_proceeed: + andcc %o2,-4,%g0 + bz .L_bn_mul_words_tail + clr %o5 + +.L_bn_mul_words_loop: + ld [%o1+4],%g3 + umul %o3,%g2,%g2 + addcc %g2,%o5,%g2 + rd %y,%g1 + addx %g1,0,%o5 + st %g2,[%o0] + + ld [%o1+8],%g2 + umul %o3,%g3,%g3 + addcc %g3,%o5,%g3 + rd %y,%g1 + dec 4,%o2 + addx %g1,0,%o5 + st %g3,[%o0+4] + + ld [%o1+12],%g3 + umul %o3,%g2,%g2 + addcc %g2,%o5,%g2 + rd %y,%g1 + inc 16,%o1 + st %g2,[%o0+8] + addx %g1,0,%o5 + + umul %o3,%g3,%g3 + addcc %g3,%o5,%g3 + rd %y,%g1 + inc 16,%o0 + addx %g1,0,%o5 + st %g3,[%o0-4] + andcc %o2,-4,%g0 + nop + bnz,a .L_bn_mul_words_loop + ld [%o1],%g2 + + tst %o2 + bnz,a .L_bn_mul_words_tail + ld [%o1],%g2 +.L_bn_mul_words_return: + retl + mov %o5,%o0 + nop + +.L_bn_mul_words_tail: + umul %o3,%g2,%g2 + addcc %g2,%o5,%g2 + rd %y,%g1 + addx %g1,0,%o5 + deccc %o2 + bz .L_bn_mul_words_return + st %g2,[%o0] + nop + + ld [%o1+4],%g2 + umul %o3,%g2,%g2 + addcc %g2,%o5,%g2 + rd %y,%g1 + addx %g1,0,%o5 + deccc %o2 + bz .L_bn_mul_words_return + st %g2,[%o0+4] + + ld [%o1+8],%g2 + umul %o3,%g2,%g2 + addcc %g2,%o5,%g2 + rd %y,%g1 + st %g2,[%o0+8] + retl + addx %g1,0,%o0 + +.type bn_mul_words,#function +.size bn_mul_words,(.-bn_mul_words) + +.align 32 +.global bn_sqr_words +/* + * void bn_sqr_words(r,a,n) + * BN_ULONG *r,*a; + * int n; + */ +bn_sqr_words: + cmp %o2,0 + bg,a .L_bn_sqr_words_proceeed + ld [%o1],%g2 + retl + clr %o0 + +.L_bn_sqr_words_proceeed: + andcc %o2,-4,%g0 + bz .L_bn_sqr_words_tail + clr %o5 + +.L_bn_sqr_words_loop: + ld [%o1+4],%g3 + umul %g2,%g2,%o4 + st %o4,[%o0] + rd %y,%o5 + st %o5,[%o0+4] + + ld [%o1+8],%g2 + umul %g3,%g3,%o4 + dec 4,%o2 + st %o4,[%o0+8] + rd %y,%o5 + st %o5,[%o0+12] + nop + + ld [%o1+12],%g3 + umul %g2,%g2,%o4 + st %o4,[%o0+16] + rd %y,%o5 + inc 16,%o1 + st %o5,[%o0+20] + + umul %g3,%g3,%o4 + inc 32,%o0 + st %o4,[%o0-8] + rd %y,%o5 + st %o5,[%o0-4] + andcc %o2,-4,%g2 + bnz,a .L_bn_sqr_words_loop + ld [%o1],%g2 + + tst %o2 + nop + bnz,a .L_bn_sqr_words_tail + ld [%o1],%g2 +.L_bn_sqr_words_return: + retl + clr %o0 + +.L_bn_sqr_words_tail: + umul %g2,%g2,%o4 + st %o4,[%o0] + deccc %o2 + rd %y,%o5 + bz .L_bn_sqr_words_return + st %o5,[%o0+4] + + ld [%o1+4],%g2 + umul %g2,%g2,%o4 + st %o4,[%o0+8] + deccc %o2 + rd %y,%o5 + nop + bz .L_bn_sqr_words_return + st %o5,[%o0+12] + + ld [%o1+8],%g2 + umul %g2,%g2,%o4 + st %o4,[%o0+16] + rd %y,%o5 + st %o5,[%o0+20] + retl + clr %o0 + +.type bn_sqr_words,#function +.size bn_sqr_words,(.-bn_sqr_words) + +.align 32 + +.global bn_div_words +/* + * BN_ULONG bn_div_words(h,l,d) + * BN_ULONG h,l,d; + */ +bn_div_words: + wr %o0,%y + udiv %o1,%o2,%o0 + retl + nop + +.type bn_div_words,#function +.size bn_div_words,(.-bn_div_words) + +.align 32 + +.global bn_add_words +/* + * BN_ULONG bn_add_words(rp,ap,bp,n) + * BN_ULONG *rp,*ap,*bp; + * int n; + */ +bn_add_words: + cmp %o3,0 + bg,a .L_bn_add_words_proceed + ld [%o1],%o4 + retl + clr %o0 + +.L_bn_add_words_proceed: + andcc %o3,-4,%g0 + bz .L_bn_add_words_tail + clr %g1 + ba .L_bn_add_words_warn_loop + addcc %g0,0,%g0 ! clear carry flag + +.L_bn_add_words_loop: + ld [%o1],%o4 +.L_bn_add_words_warn_loop: + ld [%o2],%o5 + ld [%o1+4],%g3 + ld [%o2+4],%g4 + dec 4,%o3 + addxcc %o5,%o4,%o5 + st %o5,[%o0] + + ld [%o1+8],%o4 + ld [%o2+8],%o5 + inc 16,%o1 + addxcc %g3,%g4,%g3 + st %g3,[%o0+4] + + ld [%o1-4],%g3 + ld [%o2+12],%g4 + inc 16,%o2 + addxcc %o5,%o4,%o5 + st %o5,[%o0+8] + + inc 16,%o0 + addxcc %g3,%g4,%g3 + st %g3,[%o0-4] + addx %g0,0,%g1 + andcc %o3,-4,%g0 + bnz,a .L_bn_add_words_loop + addcc %g1,-1,%g0 + + tst %o3 + bnz,a .L_bn_add_words_tail + ld [%o1],%o4 +.L_bn_add_words_return: + retl + mov %g1,%o0 + +.L_bn_add_words_tail: + addcc %g1,-1,%g0 + ld [%o2],%o5 + addxcc %o5,%o4,%o5 + addx %g0,0,%g1 + deccc %o3 + bz .L_bn_add_words_return + st %o5,[%o0] + + ld [%o1+4],%o4 + addcc %g1,-1,%g0 + ld [%o2+4],%o5 + addxcc %o5,%o4,%o5 + addx %g0,0,%g1 + deccc %o3 + bz .L_bn_add_words_return + st %o5,[%o0+4] + + ld [%o1+8],%o4 + addcc %g1,-1,%g0 + ld [%o2+8],%o5 + addxcc %o5,%o4,%o5 + st %o5,[%o0+8] + retl + addx %g0,0,%o0 + +.type bn_add_words,#function +.size bn_add_words,(.-bn_add_words) + +.align 32 + +.global bn_sub_words +/* + * BN_ULONG bn_sub_words(rp,ap,bp,n) + * BN_ULONG *rp,*ap,*bp; + * int n; + */ +bn_sub_words: + cmp %o3,0 + bg,a .L_bn_sub_words_proceed + ld [%o1],%o4 + retl + clr %o0 + +.L_bn_sub_words_proceed: + andcc %o3,-4,%g0 + bz .L_bn_sub_words_tail + clr %g1 + ba .L_bn_sub_words_warm_loop + addcc %g0,0,%g0 ! clear carry flag + +.L_bn_sub_words_loop: + ld [%o1],%o4 +.L_bn_sub_words_warm_loop: + ld [%o2],%o5 + ld [%o1+4],%g3 + ld [%o2+4],%g4 + dec 4,%o3 + subxcc %o4,%o5,%o5 + st %o5,[%o0] + + ld [%o1+8],%o4 + ld [%o2+8],%o5 + inc 16,%o1 + subxcc %g3,%g4,%g4 + st %g4,[%o0+4] + + ld [%o1-4],%g3 + ld [%o2+12],%g4 + inc 16,%o2 + subxcc %o4,%o5,%o5 + st %o5,[%o0+8] + + inc 16,%o0 + subxcc %g3,%g4,%g4 + st %g4,[%o0-4] + addx %g0,0,%g1 + andcc %o3,-4,%g0 + bnz,a .L_bn_sub_words_loop + addcc %g1,-1,%g0 + + tst %o3 + nop + bnz,a .L_bn_sub_words_tail + ld [%o1],%o4 +.L_bn_sub_words_return: + retl + mov %g1,%o0 + +.L_bn_sub_words_tail: + addcc %g1,-1,%g0 + ld [%o2],%o5 + subxcc %o4,%o5,%o5 + addx %g0,0,%g1 + deccc %o3 + bz .L_bn_sub_words_return + st %o5,[%o0] + nop + + ld [%o1+4],%o4 + addcc %g1,-1,%g0 + ld [%o2+4],%o5 + subxcc %o4,%o5,%o5 + addx %g0,0,%g1 + deccc %o3 + bz .L_bn_sub_words_return + st %o5,[%o0+4] + + ld [%o1+8],%o4 + addcc %g1,-1,%g0 + ld [%o2+8],%o5 + subxcc %o4,%o5,%o5 + st %o5,[%o0+8] + retl + addx %g0,0,%o0 + +.type bn_sub_words,#function +.size bn_sub_words,(.-bn_sub_words) + +#define FRAME_SIZE -96 + +/* + * Here is register usage map for *all* routines below. + */ +#define t_1 %o0 +#define t_2 %o1 +#define c_1 %o2 +#define c_2 %o3 +#define c_3 %o4 + +#define ap(I) [%i1+4*I] +#define bp(I) [%i2+4*I] +#define rp(I) [%i0+4*I] + +#define a_0 %l0 +#define a_1 %l1 +#define a_2 %l2 +#define a_3 %l3 +#define a_4 %l4 +#define a_5 %l5 +#define a_6 %l6 +#define a_7 %l7 + +#define b_0 %i3 +#define b_1 %i4 +#define b_2 %i5 +#define b_3 %o5 +#define b_4 %g1 +#define b_5 %g2 +#define b_6 %g3 +#define b_7 %g4 + +.align 32 +.global bn_mul_comba8 +/* + * void bn_mul_comba8(r,a,b) + * BN_ULONG *r,*a,*b; + */ +bn_mul_comba8: + save %sp,FRAME_SIZE,%sp + ld ap(0),a_0 + ld bp(0),b_0 + umul a_0,b_0,c_1 !=!mul_add_c(a[0],b[0],c1,c2,c3); + ld bp(1),b_1 + rd %y,c_2 + st c_1,rp(0) !r[0]=c1; + + umul a_0,b_1,t_1 !=!mul_add_c(a[0],b[1],c2,c3,c1); + ld ap(1),a_1 + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc %g0,t_2,c_3 != + addx %g0,%g0,c_1 + ld ap(2),a_2 + umul a_1,b_0,t_1 !mul_add_c(a[1],b[0],c2,c3,c1); + addcc c_2,t_1,c_2 != + rd %y,t_2 + addxcc c_3,t_2,c_3 + st c_2,rp(1) !r[1]=c2; + addx c_1,%g0,c_1 != + + umul a_2,b_0,t_1 !mul_add_c(a[2],b[0],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx %g0,%g0,c_2 + ld bp(2),b_2 + umul a_1,b_1,t_1 !mul_add_c(a[1],b[1],c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + ld bp(3),b_3 + addx c_2,%g0,c_2 != + umul a_0,b_2,t_1 !mul_add_c(a[0],b[2],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx c_2,%g0,c_2 + st c_3,rp(2) !r[2]=c3; + + umul a_0,b_3,t_1 !mul_add_c(a[0],b[3],c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx %g0,%g0,c_3 + umul a_1,b_2,t_1 !=!mul_add_c(a[1],b[2],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + ld ap(3),a_3 + umul a_2,b_1,t_1 !mul_add_c(a[2],b[1],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 != + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + ld ap(4),a_4 + umul a_3,b_0,t_1 !mul_add_c(a[3],b[0],c1,c2,c3);!= + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + st c_1,rp(3) !r[3]=c1; + + umul a_4,b_0,t_1 !mul_add_c(a[4],b[0],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx %g0,%g0,c_1 + umul a_3,b_1,t_1 !mul_add_c(a[3],b[1],c2,c3,c1); + addcc c_2,t_1,c_2 != + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + umul a_2,b_2,t_1 !=!mul_add_c(a[2],b[2],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 != + ld bp(4),b_4 + umul a_1,b_3,t_1 !mul_add_c(a[1],b[3],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + ld bp(5),b_5 + umul a_0,b_4,t_1 !=!mul_add_c(a[0],b[4],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 != + st c_2,rp(4) !r[4]=c2; + + umul a_0,b_5,t_1 !mul_add_c(a[0],b[5],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 + umul a_1,b_4,t_1 !mul_add_c(a[1],b[4],c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + umul a_2,b_3,t_1 !=!mul_add_c(a[2],b[3],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 != + umul a_3,b_2,t_1 !mul_add_c(a[3],b[2],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx c_2,%g0,c_2 + ld ap(5),a_5 + umul a_4,b_1,t_1 !mul_add_c(a[4],b[1],c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + ld ap(6),a_6 + addx c_2,%g0,c_2 != + umul a_5,b_0,t_1 !mul_add_c(a[5],b[0],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx c_2,%g0,c_2 + st c_3,rp(5) !r[5]=c3; + + umul a_6,b_0,t_1 !mul_add_c(a[6],b[0],c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx %g0,%g0,c_3 + umul a_5,b_1,t_1 !=!mul_add_c(a[5],b[1],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + umul a_4,b_2,t_1 !mul_add_c(a[4],b[2],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + umul a_3,b_3,t_1 !mul_add_c(a[3],b[3],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 != + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + umul a_2,b_4,t_1 !mul_add_c(a[2],b[4],c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + ld bp(6),b_6 + addx c_3,%g0,c_3 != + umul a_1,b_5,t_1 !mul_add_c(a[1],b[5],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + ld bp(7),b_7 + umul a_0,b_6,t_1 !mul_add_c(a[0],b[6],c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + st c_1,rp(6) !r[6]=c1; + addx c_3,%g0,c_3 != + + umul a_0,b_7,t_1 !mul_add_c(a[0],b[7],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 != + addx %g0,%g0,c_1 + umul a_1,b_6,t_1 !mul_add_c(a[1],b[6],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + umul a_2,b_5,t_1 !mul_add_c(a[2],b[5],c2,c3,c1); + addcc c_2,t_1,c_2 != + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + umul a_3,b_4,t_1 !=!mul_add_c(a[3],b[4],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 != + umul a_4,b_3,t_1 !mul_add_c(a[4],b[3],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_5,b_2,t_1 !mul_add_c(a[5],b[2],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + ld ap(7),a_7 + umul a_6,b_1,t_1 !=!mul_add_c(a[6],b[1],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 != + umul a_7,b_0,t_1 !mul_add_c(a[7],b[0],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + st c_2,rp(7) !r[7]=c2; + + umul a_7,b_1,t_1 !mul_add_c(a[7],b[1],c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 + umul a_6,b_2,t_1 !=!mul_add_c(a[6],b[2],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 != + umul a_5,b_3,t_1 !mul_add_c(a[5],b[3],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx c_2,%g0,c_2 + umul a_4,b_4,t_1 !mul_add_c(a[4],b[4],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + umul a_3,b_5,t_1 !mul_add_c(a[3],b[5],c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + umul a_2,b_6,t_1 !=!mul_add_c(a[2],b[6],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 != + umul a_1,b_7,t_1 !mul_add_c(a[1],b[7],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 ! + addx c_2,%g0,c_2 + st c_3,rp(8) !r[8]=c3; + + umul a_2,b_7,t_1 !mul_add_c(a[2],b[7],c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx %g0,%g0,c_3 + umul a_3,b_6,t_1 !=!mul_add_c(a[3],b[6],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + umul a_4,b_5,t_1 !mul_add_c(a[4],b[5],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + umul a_5,b_4,t_1 !mul_add_c(a[5],b[4],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 != + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + umul a_6,b_3,t_1 !mul_add_c(a[6],b[3],c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + umul a_7,b_2,t_1 !=!mul_add_c(a[7],b[2],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + st c_1,rp(9) !r[9]=c1; + + umul a_7,b_3,t_1 !mul_add_c(a[7],b[3],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx %g0,%g0,c_1 + umul a_6,b_4,t_1 !mul_add_c(a[6],b[4],c2,c3,c1); + addcc c_2,t_1,c_2 != + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + umul a_5,b_5,t_1 !=!mul_add_c(a[5],b[5],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 != + umul a_4,b_6,t_1 !mul_add_c(a[4],b[6],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_3,b_7,t_1 !mul_add_c(a[3],b[7],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + st c_2,rp(10) !r[10]=c2; + + umul a_4,b_7,t_1 !=!mul_add_c(a[4],b[7],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 != + umul a_5,b_6,t_1 !mul_add_c(a[5],b[6],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx c_2,%g0,c_2 + umul a_6,b_5,t_1 !mul_add_c(a[6],b[5],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + umul a_7,b_4,t_1 !mul_add_c(a[7],b[4],c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + st c_3,rp(11) !r[11]=c3; + addx c_2,%g0,c_2 != + + umul a_7,b_5,t_1 !mul_add_c(a[7],b[5],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx %g0,%g0,c_3 + umul a_6,b_6,t_1 !mul_add_c(a[6],b[6],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 != + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + umul a_5,b_7,t_1 !mul_add_c(a[5],b[7],c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + st c_1,rp(12) !r[12]=c1; + addx c_3,%g0,c_3 != + + umul a_6,b_7,t_1 !mul_add_c(a[6],b[7],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 != + addx %g0,%g0,c_1 + umul a_7,b_6,t_1 !mul_add_c(a[7],b[6],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + st c_2,rp(13) !r[13]=c2; + + umul a_7,b_7,t_1 !=!mul_add_c(a[7],b[7],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + nop != + st c_3,rp(14) !r[14]=c3; + st c_1,rp(15) !r[15]=c1; + + ret + restore %g0,%g0,%o0 + +.type bn_mul_comba8,#function +.size bn_mul_comba8,(.-bn_mul_comba8) + +.align 32 + +.global bn_mul_comba4 +/* + * void bn_mul_comba4(r,a,b) + * BN_ULONG *r,*a,*b; + */ +bn_mul_comba4: + save %sp,FRAME_SIZE,%sp + ld ap(0),a_0 + ld bp(0),b_0 + umul a_0,b_0,c_1 !=!mul_add_c(a[0],b[0],c1,c2,c3); + ld bp(1),b_1 + rd %y,c_2 + st c_1,rp(0) !r[0]=c1; + + umul a_0,b_1,t_1 !=!mul_add_c(a[0],b[1],c2,c3,c1); + ld ap(1),a_1 + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc %g0,t_2,c_3 + addx %g0,%g0,c_1 + ld ap(2),a_2 + umul a_1,b_0,t_1 !=!mul_add_c(a[1],b[0],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 != + st c_2,rp(1) !r[1]=c2; + + umul a_2,b_0,t_1 !mul_add_c(a[2],b[0],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 + ld bp(2),b_2 + umul a_1,b_1,t_1 !=!mul_add_c(a[1],b[1],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 != + ld bp(3),b_3 + umul a_0,b_2,t_1 !mul_add_c(a[0],b[2],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + st c_3,rp(2) !r[2]=c3; + + umul a_0,b_3,t_1 !=!mul_add_c(a[0],b[3],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx %g0,%g0,c_3 != + umul a_1,b_2,t_1 !mul_add_c(a[1],b[2],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + ld ap(3),a_3 + umul a_2,b_1,t_1 !mul_add_c(a[2],b[1],c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + umul a_3,b_0,t_1 !=!mul_add_c(a[3],b[0],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + st c_1,rp(3) !r[3]=c1; + + umul a_3,b_1,t_1 !mul_add_c(a[3],b[1],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx %g0,%g0,c_1 + umul a_2,b_2,t_1 !mul_add_c(a[2],b[2],c2,c3,c1); + addcc c_2,t_1,c_2 != + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + umul a_1,b_3,t_1 !=!mul_add_c(a[1],b[3],c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 != + st c_2,rp(4) !r[4]=c2; + + umul a_2,b_3,t_1 !mul_add_c(a[2],b[3],c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 + umul a_3,b_2,t_1 !mul_add_c(a[3],b[2],c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + st c_3,rp(5) !r[5]=c3; + addx c_2,%g0,c_2 != + + umul a_3,b_3,t_1 !mul_add_c(a[3],b[3],c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + st c_1,rp(6) !r[6]=c1; + st c_2,rp(7) !r[7]=c2; + + ret + restore %g0,%g0,%o0 + +.type bn_mul_comba4,#function +.size bn_mul_comba4,(.-bn_mul_comba4) + +.align 32 + +.global bn_sqr_comba8 +bn_sqr_comba8: + save %sp,FRAME_SIZE,%sp + ld ap(0),a_0 + ld ap(1),a_1 + umul a_0,a_0,c_1 !=!sqr_add_c(a,0,c1,c2,c3); + rd %y,c_2 + st c_1,rp(0) !r[0]=c1; + + ld ap(2),a_2 + umul a_0,a_1,t_1 !=!sqr_add_c2(a,1,0,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc %g0,t_2,c_3 + addx %g0,%g0,c_1 != + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 + st c_2,rp(1) !r[1]=c2; + addx c_1,%g0,c_1 != + + umul a_2,a_0,t_1 !sqr_add_c2(a,2,0,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx %g0,%g0,c_2 + addcc c_3,t_1,c_3 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 != + ld ap(3),a_3 + umul a_1,a_1,t_1 !sqr_add_c(a,1,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + st c_3,rp(2) !r[2]=c3; + + umul a_0,a_3,t_1 !=!sqr_add_c2(a,3,0,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx %g0,%g0,c_3 != + addcc c_1,t_1,c_1 + addxcc c_2,t_2,c_2 + ld ap(4),a_4 + addx c_3,%g0,c_3 != + umul a_1,a_2,t_1 !sqr_add_c2(a,2,1,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + addcc c_1,t_1,c_1 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + st c_1,rp(3) !r[3]=c1; + + umul a_4,a_0,t_1 !sqr_add_c2(a,4,0,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx %g0,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_3,a_1,t_1 !sqr_add_c2(a,3,1,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + ld ap(5),a_5 + umul a_2,a_2,t_1 !sqr_add_c(a,2,c2,c3,c1); + addcc c_2,t_1,c_2 != + rd %y,t_2 + addxcc c_3,t_2,c_3 + st c_2,rp(4) !r[4]=c2; + addx c_1,%g0,c_1 != + + umul a_0,a_5,t_1 !sqr_add_c2(a,5,0,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx %g0,%g0,c_2 + addcc c_3,t_1,c_3 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 != + umul a_1,a_4,t_1 !sqr_add_c2(a,4,1,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + addx c_2,%g0,c_2 + addcc c_3,t_1,c_3 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 != + ld ap(6),a_6 + umul a_2,a_3,t_1 !sqr_add_c2(a,3,2,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + addcc c_3,t_1,c_3 + addxcc c_1,t_2,c_1 != + addx c_2,%g0,c_2 + st c_3,rp(5) !r[5]=c3; + + umul a_6,a_0,t_1 !sqr_add_c2(a,6,0,c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx %g0,%g0,c_3 + addcc c_1,t_1,c_1 != + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + umul a_5,a_1,t_1 !sqr_add_c2(a,5,1,c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + addcc c_1,t_1,c_1 != + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + umul a_4,a_2,t_1 !sqr_add_c2(a,4,2,c1,c2,c3); + addcc c_1,t_1,c_1 != + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + addcc c_1,t_1,c_1 != + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 + ld ap(7),a_7 + umul a_3,a_3,t_1 !=!sqr_add_c(a,3,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + st c_1,rp(6) !r[6]=c1; + + umul a_0,a_7,t_1 !sqr_add_c2(a,7,0,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx %g0,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_1,a_6,t_1 !sqr_add_c2(a,6,1,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_2,a_5,t_1 !sqr_add_c2(a,5,2,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_3,a_4,t_1 !sqr_add_c2(a,4,3,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + st c_2,rp(7) !r[7]=c2; + + umul a_7,a_1,t_1 !sqr_add_c2(a,7,1,c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 + addcc c_3,t_1,c_3 != + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + umul a_6,a_2,t_1 !sqr_add_c2(a,6,2,c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + addcc c_3,t_1,c_3 != + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + umul a_5,a_3,t_1 !sqr_add_c2(a,5,3,c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + addcc c_3,t_1,c_3 != + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + umul a_4,a_4,t_1 !sqr_add_c(a,4,c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + st c_3,rp(8) !r[8]=c3; + addx c_2,%g0,c_2 != + + umul a_2,a_7,t_1 !sqr_add_c2(a,7,2,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx %g0,%g0,c_3 + addcc c_1,t_1,c_1 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + umul a_3,a_6,t_1 !sqr_add_c2(a,6,3,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + addcc c_1,t_1,c_1 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + umul a_4,a_5,t_1 !sqr_add_c2(a,5,4,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + addcc c_1,t_1,c_1 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + st c_1,rp(9) !r[9]=c1; + + umul a_7,a_3,t_1 !sqr_add_c2(a,7,3,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx %g0,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_6,a_4,t_1 !sqr_add_c2(a,6,4,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_5,a_5,t_1 !sqr_add_c(a,5,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + st c_2,rp(10) !r[10]=c2; + + umul a_4,a_7,t_1 !=!sqr_add_c2(a,7,4,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 != + addcc c_3,t_1,c_3 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 + umul a_5,a_6,t_1 !=!sqr_add_c2(a,6,5,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx c_2,%g0,c_2 != + addcc c_3,t_1,c_3 + addxcc c_1,t_2,c_1 + st c_3,rp(11) !r[11]=c3; + addx c_2,%g0,c_2 != + + umul a_7,a_5,t_1 !sqr_add_c2(a,7,5,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx %g0,%g0,c_3 + addcc c_1,t_1,c_1 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + umul a_6,a_6,t_1 !sqr_add_c(a,6,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + st c_1,rp(12) !r[12]=c1; + + umul a_6,a_7,t_1 !sqr_add_c2(a,7,6,c2,c3,c1); + addcc c_2,t_1,c_2 != + rd %y,t_2 + addxcc c_3,t_2,c_3 + addx %g0,%g0,c_1 + addcc c_2,t_1,c_2 != + addxcc c_3,t_2,c_3 + st c_2,rp(13) !r[13]=c2; + addx c_1,%g0,c_1 != + + umul a_7,a_7,t_1 !sqr_add_c(a,7,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 != + st c_3,rp(14) !r[14]=c3; + st c_1,rp(15) !r[15]=c1; + + ret + restore %g0,%g0,%o0 + +.type bn_sqr_comba8,#function +.size bn_sqr_comba8,(.-bn_sqr_comba8) + +.align 32 + +.global bn_sqr_comba4 +/* + * void bn_sqr_comba4(r,a) + * BN_ULONG *r,*a; + */ +bn_sqr_comba4: + save %sp,FRAME_SIZE,%sp + ld ap(0),a_0 + umul a_0,a_0,c_1 !sqr_add_c(a,0,c1,c2,c3); + ld ap(1),a_1 != + rd %y,c_2 + st c_1,rp(0) !r[0]=c1; + + ld ap(2),a_2 + umul a_0,a_1,t_1 !=!sqr_add_c2(a,1,0,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 + addxcc %g0,t_2,c_3 + addx %g0,%g0,c_1 != + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 != + st c_2,rp(1) !r[1]=c2; + + umul a_2,a_0,t_1 !sqr_add_c2(a,2,0,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 != + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 + addcc c_3,t_1,c_3 + addxcc c_1,t_2,c_1 != + addx c_2,%g0,c_2 + ld ap(3),a_3 + umul a_1,a_1,t_1 !sqr_add_c(a,1,c3,c1,c2); + addcc c_3,t_1,c_3 != + rd %y,t_2 + addxcc c_1,t_2,c_1 + st c_3,rp(2) !r[2]=c3; + addx c_2,%g0,c_2 != + + umul a_0,a_3,t_1 !sqr_add_c2(a,3,0,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx %g0,%g0,c_3 + addcc c_1,t_1,c_1 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + umul a_1,a_2,t_1 !sqr_add_c2(a,2,1,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + addx c_3,%g0,c_3 + addcc c_1,t_1,c_1 + addxcc c_2,t_2,c_2 + addx c_3,%g0,c_3 != + st c_1,rp(3) !r[3]=c1; + + umul a_3,a_1,t_1 !sqr_add_c2(a,3,1,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx %g0,%g0,c_1 + addcc c_2,t_1,c_2 + addxcc c_3,t_2,c_3 != + addx c_1,%g0,c_1 + umul a_2,a_2,t_1 !sqr_add_c(a,2,c2,c3,c1); + addcc c_2,t_1,c_2 + rd %y,t_2 != + addxcc c_3,t_2,c_3 + addx c_1,%g0,c_1 + st c_2,rp(4) !r[4]=c2; + + umul a_2,a_3,t_1 !=!sqr_add_c2(a,3,2,c3,c1,c2); + addcc c_3,t_1,c_3 + rd %y,t_2 + addxcc c_1,t_2,c_1 + addx %g0,%g0,c_2 != + addcc c_3,t_1,c_3 + addxcc c_1,t_2,c_1 + st c_3,rp(5) !r[5]=c3; + addx c_2,%g0,c_2 != + + umul a_3,a_3,t_1 !sqr_add_c(a,3,c1,c2,c3); + addcc c_1,t_1,c_1 + rd %y,t_2 + addxcc c_2,t_2,c_2 != + st c_1,rp(6) !r[6]=c1; + st c_2,rp(7) !r[7]=c2; + + ret + restore %g0,%g0,%o0 + +.type bn_sqr_comba4,#function +.size bn_sqr_comba4,(.-bn_sqr_comba4) + +.align 32 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv8plus.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv8plus.S new file mode 100644 index 00000000..63de1860 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv8plus.S @@ -0,0 +1,1558 @@ +.ident "sparcv8plus.s, Version 1.4" +.ident "SPARC v9 ISA artwork by Andy Polyakov " + +/* + * ==================================================================== + * Written by Andy Polyakov for the OpenSSL + * project. + * + * Rights for redistribution and usage in source and binary forms are + * granted according to the OpenSSL license. Warranty of any kind is + * disclaimed. + * ==================================================================== + */ + +/* + * This is my modest contributon to OpenSSL project (see + * http://www.openssl.org/ for more information about it) and is + * a drop-in UltraSPARC ISA replacement for crypto/bn/bn_asm.c + * module. For updates see http://fy.chalmers.se/~appro/hpe/. + * + * Questions-n-answers. + * + * Q. How to compile? + * A. With SC4.x/SC5.x: + * + * cc -xarch=v8plus -c bn_asm.sparc.v8plus.S -o bn_asm.o + * + * and with gcc: + * + * gcc -mcpu=ultrasparc -c bn_asm.sparc.v8plus.S -o bn_asm.o + * + * or if above fails (it does if you have gas installed): + * + * gcc -E bn_asm.sparc.v8plus.S | as -xarch=v8plus /dev/fd/0 -o bn_asm.o + * + * Quick-n-dirty way to fuse the module into the library. + * Provided that the library is already configured and built + * (in 0.9.2 case with no-asm option): + * + * # cd crypto/bn + * # cp /some/place/bn_asm.sparc.v8plus.S . + * # cc -xarch=v8plus -c bn_asm.sparc.v8plus.S -o bn_asm.o + * # make + * # cd ../.. + * # make; make test + * + * Quick-n-dirty way to get rid of it: + * + * # cd crypto/bn + * # touch bn_asm.c + * # make + * # cd ../.. + * # make; make test + * + * Q. V8plus achitecture? What kind of beast is that? + * A. Well, it's rather a programming model than an architecture... + * It's actually v9-compliant, i.e. *any* UltraSPARC, CPU under + * special conditions, namely when kernel doesn't preserve upper + * 32 bits of otherwise 64-bit registers during a context switch. + * + * Q. Why just UltraSPARC? What about SuperSPARC? + * A. Original release did target UltraSPARC only. Now SuperSPARC + * version is provided along. Both version share bn_*comba[48] + * implementations (see comment later in code for explanation). + * But what's so special about this UltraSPARC implementation? + * Why didn't I let compiler do the job? Trouble is that most of + * available compilers (well, SC5.0 is the only exception) don't + * attempt to take advantage of UltraSPARC's 64-bitness under + * 32-bit kernels even though it's perfectly possible (see next + * question). + * + * Q. 64-bit registers under 32-bit kernels? Didn't you just say it + * doesn't work? + * A. You can't adress *all* registers as 64-bit wide:-( The catch is + * that you actually may rely upon %o0-%o5 and %g1-%g4 being fully + * preserved if you're in a leaf function, i.e. such never calling + * any other functions. All functions in this module are leaf and + * 10 registers is a handful. And as a matter of fact none-"comba" + * routines don't require even that much and I could even afford to + * not allocate own stack frame for 'em:-) + * + * Q. What about 64-bit kernels? + * A. What about 'em? Just kidding:-) Pure 64-bit version is currently + * under evaluation and development... + * + * Q. What about shared libraries? + * A. What about 'em? Kidding again:-) Code does *not* contain any + * code position dependencies and it's safe to include it into + * shared library as is. + * + * Q. How much faster does it go? + * A. Do you have a good benchmark? In either case below is what I + * experience with crypto/bn/expspeed.c test program: + * + * v8plus module on U10/300MHz against bn_asm.c compiled with: + * + * cc-5.0 -xarch=v8plus -xO5 -xdepend +7-12% + * cc-4.2 -xarch=v8plus -xO5 -xdepend +25-35% + * egcs-1.1.2 -mcpu=ultrasparc -O3 +35-45% + * + * v8 module on SS10/60MHz against bn_asm.c compiled with: + * + * cc-5.0 -xarch=v8 -xO5 -xdepend +7-10% + * cc-4.2 -xarch=v8 -xO5 -xdepend +10% + * egcs-1.1.2 -mv8 -O3 +35-45% + * + * As you can see it's damn hard to beat the new Sun C compiler + * and it's in first place GNU C users who will appreciate this + * assembler implementation:-) + */ + +/* + * Revision history. + * + * 1.0 - initial release; + * 1.1 - new loop unrolling model(*); + * - some more fine tuning; + * 1.2 - made gas friendly; + * - updates to documentation concerning v9; + * - new performance comparison matrix; + * 1.3 - fixed problem with /usr/ccs/lib/cpp; + * 1.4 - native V9 bn_*_comba[48] implementation (15% more efficient) + * resulting in slight overall performance kick; + * - some retunes; + * - support for GNU as added; + * + * (*) Originally unrolled loop looked like this: + * for (;;) { + * op(p+0); if (--n==0) break; + * op(p+1); if (--n==0) break; + * op(p+2); if (--n==0) break; + * op(p+3); if (--n==0) break; + * p+=4; + * } + * I unroll according to following: + * while (n&~3) { + * op(p+0); op(p+1); op(p+2); op(p+3); + * p+=4; n=-4; + * } + * if (n) { + * op(p+0); if (--n==0) return; + * op(p+2); if (--n==0) return; + * op(p+3); return; + * } + */ + +#if defined(__SUNPRO_C) && defined(__sparcv9) + /* They've said -xarch=v9 at command line */ + .register %g2,#scratch + .register %g3,#scratch +# define FRAME_SIZE -192 +#elif defined(__GNUC__) && defined(__arch64__) + /* They've said -m64 at command line */ + .register %g2,#scratch + .register %g3,#scratch +# define FRAME_SIZE -192 +#else +# define FRAME_SIZE -96 +#endif +/* + * GNU assembler can't stand stuw:-( + */ +#define stuw st + +.section ".text",#alloc,#execinstr +.file "bn_asm.sparc.v8plus.S" + +.align 32 + +.global bn_mul_add_words +/* + * BN_ULONG bn_mul_add_words(rp,ap,num,w) + * BN_ULONG *rp,*ap; + * int num; + * BN_ULONG w; + */ +bn_mul_add_words: + sra %o2,%g0,%o2 ! signx %o2 + brgz,a %o2,.L_bn_mul_add_words_proceed + lduw [%o1],%g2 + retl + clr %o0 + nop + nop + nop + +.L_bn_mul_add_words_proceed: + srl %o3,%g0,%o3 ! clruw %o3 + andcc %o2,-4,%g0 + bz,pn %icc,.L_bn_mul_add_words_tail + clr %o5 + +.L_bn_mul_add_words_loop: ! wow! 32 aligned! + lduw [%o0],%g1 + lduw [%o1+4],%g3 + mulx %o3,%g2,%g2 + add %g1,%o5,%o4 + nop + add %o4,%g2,%o4 + stuw %o4,[%o0] + srlx %o4,32,%o5 + + lduw [%o0+4],%g1 + lduw [%o1+8],%g2 + mulx %o3,%g3,%g3 + add %g1,%o5,%o4 + dec 4,%o2 + add %o4,%g3,%o4 + stuw %o4,[%o0+4] + srlx %o4,32,%o5 + + lduw [%o0+8],%g1 + lduw [%o1+12],%g3 + mulx %o3,%g2,%g2 + add %g1,%o5,%o4 + inc 16,%o1 + add %o4,%g2,%o4 + stuw %o4,[%o0+8] + srlx %o4,32,%o5 + + lduw [%o0+12],%g1 + mulx %o3,%g3,%g3 + add %g1,%o5,%o4 + inc 16,%o0 + add %o4,%g3,%o4 + andcc %o2,-4,%g0 + stuw %o4,[%o0-4] + srlx %o4,32,%o5 + bnz,a,pt %icc,.L_bn_mul_add_words_loop + lduw [%o1],%g2 + + brnz,a,pn %o2,.L_bn_mul_add_words_tail + lduw [%o1],%g2 +.L_bn_mul_add_words_return: + retl + mov %o5,%o0 + +.L_bn_mul_add_words_tail: + lduw [%o0],%g1 + mulx %o3,%g2,%g2 + add %g1,%o5,%o4 + dec %o2 + add %o4,%g2,%o4 + srlx %o4,32,%o5 + brz,pt %o2,.L_bn_mul_add_words_return + stuw %o4,[%o0] + + lduw [%o1+4],%g2 + lduw [%o0+4],%g1 + mulx %o3,%g2,%g2 + add %g1,%o5,%o4 + dec %o2 + add %o4,%g2,%o4 + srlx %o4,32,%o5 + brz,pt %o2,.L_bn_mul_add_words_return + stuw %o4,[%o0+4] + + lduw [%o1+8],%g2 + lduw [%o0+8],%g1 + mulx %o3,%g2,%g2 + add %g1,%o5,%o4 + add %o4,%g2,%o4 + stuw %o4,[%o0+8] + retl + srlx %o4,32,%o0 + +.type bn_mul_add_words,#function +.size bn_mul_add_words,(.-bn_mul_add_words) + +.align 32 + +.global bn_mul_words +/* + * BN_ULONG bn_mul_words(rp,ap,num,w) + * BN_ULONG *rp,*ap; + * int num; + * BN_ULONG w; + */ +bn_mul_words: + sra %o2,%g0,%o2 ! signx %o2 + brgz,a %o2,.L_bn_mul_words_proceeed + lduw [%o1],%g2 + retl + clr %o0 + nop + nop + nop + +.L_bn_mul_words_proceeed: + srl %o3,%g0,%o3 ! clruw %o3 + andcc %o2,-4,%g0 + bz,pn %icc,.L_bn_mul_words_tail + clr %o5 + +.L_bn_mul_words_loop: ! wow! 32 aligned! + lduw [%o1+4],%g3 + mulx %o3,%g2,%g2 + add %g2,%o5,%o4 + nop + stuw %o4,[%o0] + srlx %o4,32,%o5 + + lduw [%o1+8],%g2 + mulx %o3,%g3,%g3 + add %g3,%o5,%o4 + dec 4,%o2 + stuw %o4,[%o0+4] + srlx %o4,32,%o5 + + lduw [%o1+12],%g3 + mulx %o3,%g2,%g2 + add %g2,%o5,%o4 + inc 16,%o1 + stuw %o4,[%o0+8] + srlx %o4,32,%o5 + + mulx %o3,%g3,%g3 + add %g3,%o5,%o4 + inc 16,%o0 + stuw %o4,[%o0-4] + srlx %o4,32,%o5 + andcc %o2,-4,%g0 + bnz,a,pt %icc,.L_bn_mul_words_loop + lduw [%o1],%g2 + nop + nop + + brnz,a,pn %o2,.L_bn_mul_words_tail + lduw [%o1],%g2 +.L_bn_mul_words_return: + retl + mov %o5,%o0 + +.L_bn_mul_words_tail: + mulx %o3,%g2,%g2 + add %g2,%o5,%o4 + dec %o2 + srlx %o4,32,%o5 + brz,pt %o2,.L_bn_mul_words_return + stuw %o4,[%o0] + + lduw [%o1+4],%g2 + mulx %o3,%g2,%g2 + add %g2,%o5,%o4 + dec %o2 + srlx %o4,32,%o5 + brz,pt %o2,.L_bn_mul_words_return + stuw %o4,[%o0+4] + + lduw [%o1+8],%g2 + mulx %o3,%g2,%g2 + add %g2,%o5,%o4 + stuw %o4,[%o0+8] + retl + srlx %o4,32,%o0 + +.type bn_mul_words,#function +.size bn_mul_words,(.-bn_mul_words) + +.align 32 +.global bn_sqr_words +/* + * void bn_sqr_words(r,a,n) + * BN_ULONG *r,*a; + * int n; + */ +bn_sqr_words: + sra %o2,%g0,%o2 ! signx %o2 + brgz,a %o2,.L_bn_sqr_words_proceeed + lduw [%o1],%g2 + retl + clr %o0 + nop + nop + nop + +.L_bn_sqr_words_proceeed: + andcc %o2,-4,%g0 + nop + bz,pn %icc,.L_bn_sqr_words_tail + nop + +.L_bn_sqr_words_loop: ! wow! 32 aligned! + lduw [%o1+4],%g3 + mulx %g2,%g2,%o4 + stuw %o4,[%o0] + srlx %o4,32,%o5 + stuw %o5,[%o0+4] + nop + + lduw [%o1+8],%g2 + mulx %g3,%g3,%o4 + dec 4,%o2 + stuw %o4,[%o0+8] + srlx %o4,32,%o5 + stuw %o5,[%o0+12] + + lduw [%o1+12],%g3 + mulx %g2,%g2,%o4 + srlx %o4,32,%o5 + stuw %o4,[%o0+16] + inc 16,%o1 + stuw %o5,[%o0+20] + + mulx %g3,%g3,%o4 + inc 32,%o0 + stuw %o4,[%o0-8] + srlx %o4,32,%o5 + andcc %o2,-4,%g2 + stuw %o5,[%o0-4] + bnz,a,pt %icc,.L_bn_sqr_words_loop + lduw [%o1],%g2 + nop + + brnz,a,pn %o2,.L_bn_sqr_words_tail + lduw [%o1],%g2 +.L_bn_sqr_words_return: + retl + clr %o0 + +.L_bn_sqr_words_tail: + mulx %g2,%g2,%o4 + dec %o2 + stuw %o4,[%o0] + srlx %o4,32,%o5 + brz,pt %o2,.L_bn_sqr_words_return + stuw %o5,[%o0+4] + + lduw [%o1+4],%g2 + mulx %g2,%g2,%o4 + dec %o2 + stuw %o4,[%o0+8] + srlx %o4,32,%o5 + brz,pt %o2,.L_bn_sqr_words_return + stuw %o5,[%o0+12] + + lduw [%o1+8],%g2 + mulx %g2,%g2,%o4 + srlx %o4,32,%o5 + stuw %o4,[%o0+16] + stuw %o5,[%o0+20] + retl + clr %o0 + +.type bn_sqr_words,#function +.size bn_sqr_words,(.-bn_sqr_words) + +.align 32 +.global bn_div_words +/* + * BN_ULONG bn_div_words(h,l,d) + * BN_ULONG h,l,d; + */ +bn_div_words: + sllx %o0,32,%o0 + or %o0,%o1,%o0 + udivx %o0,%o2,%o0 + retl + srl %o0,%g0,%o0 ! clruw %o0 + +.type bn_div_words,#function +.size bn_div_words,(.-bn_div_words) + +.align 32 + +.global bn_add_words +/* + * BN_ULONG bn_add_words(rp,ap,bp,n) + * BN_ULONG *rp,*ap,*bp; + * int n; + */ +bn_add_words: + sra %o3,%g0,%o3 ! signx %o3 + brgz,a %o3,.L_bn_add_words_proceed + lduw [%o1],%o4 + retl + clr %o0 + +.L_bn_add_words_proceed: + andcc %o3,-4,%g0 + bz,pn %icc,.L_bn_add_words_tail + addcc %g0,0,%g0 ! clear carry flag + +.L_bn_add_words_loop: ! wow! 32 aligned! + dec 4,%o3 + lduw [%o2],%o5 + lduw [%o1+4],%g1 + lduw [%o2+4],%g2 + lduw [%o1+8],%g3 + lduw [%o2+8],%g4 + addccc %o5,%o4,%o5 + stuw %o5,[%o0] + + lduw [%o1+12],%o4 + lduw [%o2+12],%o5 + inc 16,%o1 + addccc %g1,%g2,%g1 + stuw %g1,[%o0+4] + + inc 16,%o2 + addccc %g3,%g4,%g3 + stuw %g3,[%o0+8] + + inc 16,%o0 + addccc %o5,%o4,%o5 + stuw %o5,[%o0-4] + and %o3,-4,%g1 + brnz,a,pt %g1,.L_bn_add_words_loop + lduw [%o1],%o4 + + brnz,a,pn %o3,.L_bn_add_words_tail + lduw [%o1],%o4 +.L_bn_add_words_return: + clr %o0 + retl + movcs %icc,1,%o0 + nop + +.L_bn_add_words_tail: + lduw [%o2],%o5 + dec %o3 + addccc %o5,%o4,%o5 + brz,pt %o3,.L_bn_add_words_return + stuw %o5,[%o0] + + lduw [%o1+4],%o4 + lduw [%o2+4],%o5 + dec %o3 + addccc %o5,%o4,%o5 + brz,pt %o3,.L_bn_add_words_return + stuw %o5,[%o0+4] + + lduw [%o1+8],%o4 + lduw [%o2+8],%o5 + addccc %o5,%o4,%o5 + stuw %o5,[%o0+8] + clr %o0 + retl + movcs %icc,1,%o0 + +.type bn_add_words,#function +.size bn_add_words,(.-bn_add_words) + +.global bn_sub_words +/* + * BN_ULONG bn_sub_words(rp,ap,bp,n) + * BN_ULONG *rp,*ap,*bp; + * int n; + */ +bn_sub_words: + sra %o3,%g0,%o3 ! signx %o3 + brgz,a %o3,.L_bn_sub_words_proceed + lduw [%o1],%o4 + retl + clr %o0 + +.L_bn_sub_words_proceed: + andcc %o3,-4,%g0 + bz,pn %icc,.L_bn_sub_words_tail + addcc %g0,0,%g0 ! clear carry flag + +.L_bn_sub_words_loop: ! wow! 32 aligned! + dec 4,%o3 + lduw [%o2],%o5 + lduw [%o1+4],%g1 + lduw [%o2+4],%g2 + lduw [%o1+8],%g3 + lduw [%o2+8],%g4 + subccc %o4,%o5,%o5 + stuw %o5,[%o0] + + lduw [%o1+12],%o4 + lduw [%o2+12],%o5 + inc 16,%o1 + subccc %g1,%g2,%g2 + stuw %g2,[%o0+4] + + inc 16,%o2 + subccc %g3,%g4,%g4 + stuw %g4,[%o0+8] + + inc 16,%o0 + subccc %o4,%o5,%o5 + stuw %o5,[%o0-4] + and %o3,-4,%g1 + brnz,a,pt %g1,.L_bn_sub_words_loop + lduw [%o1],%o4 + + brnz,a,pn %o3,.L_bn_sub_words_tail + lduw [%o1],%o4 +.L_bn_sub_words_return: + clr %o0 + retl + movcs %icc,1,%o0 + nop + +.L_bn_sub_words_tail: ! wow! 32 aligned! + lduw [%o2],%o5 + dec %o3 + subccc %o4,%o5,%o5 + brz,pt %o3,.L_bn_sub_words_return + stuw %o5,[%o0] + + lduw [%o1+4],%o4 + lduw [%o2+4],%o5 + dec %o3 + subccc %o4,%o5,%o5 + brz,pt %o3,.L_bn_sub_words_return + stuw %o5,[%o0+4] + + lduw [%o1+8],%o4 + lduw [%o2+8],%o5 + subccc %o4,%o5,%o5 + stuw %o5,[%o0+8] + clr %o0 + retl + movcs %icc,1,%o0 + +.type bn_sub_words,#function +.size bn_sub_words,(.-bn_sub_words) + +/* + * Code below depends on the fact that upper parts of the %l0-%l7 + * and %i0-%i7 are zeroed by kernel after context switch. In + * previous versions this comment stated that "the trouble is that + * it's not feasible to implement the mumbo-jumbo in less V9 + * instructions:-(" which apparently isn't true thanks to + * 'bcs,a %xcc,.+8; inc %rd' pair. But the performance improvement + * results not from the shorter code, but from elimination of + * multicycle none-pairable 'rd %y,%rd' instructions. + * + * Andy. + */ + +/* + * Here is register usage map for *all* routines below. + */ +#define t_1 %o0 +#define t_2 %o1 +#define c_12 %o2 +#define c_3 %o3 + +#define ap(I) [%i1+4*I] +#define bp(I) [%i2+4*I] +#define rp(I) [%i0+4*I] + +#define a_0 %l0 +#define a_1 %l1 +#define a_2 %l2 +#define a_3 %l3 +#define a_4 %l4 +#define a_5 %l5 +#define a_6 %l6 +#define a_7 %l7 + +#define b_0 %i3 +#define b_1 %i4 +#define b_2 %i5 +#define b_3 %o4 +#define b_4 %o5 +#define b_5 %o7 +#define b_6 %g1 +#define b_7 %g4 + +.align 32 +.global bn_mul_comba8 +/* + * void bn_mul_comba8(r,a,b) + * BN_ULONG *r,*a,*b; + */ +bn_mul_comba8: + save %sp,FRAME_SIZE,%sp + mov 1,t_2 + lduw ap(0),a_0 + sllx t_2,32,t_2 + lduw bp(0),b_0 != + lduw bp(1),b_1 + mulx a_0,b_0,t_1 !mul_add_c(a[0],b[0],c1,c2,c3); + srlx t_1,32,c_12 + stuw t_1,rp(0) !=!r[0]=c1; + + lduw ap(1),a_1 + mulx a_0,b_1,t_1 !mul_add_c(a[0],b[1],c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(2),a_2 + mulx a_1,b_0,t_1 !=!mul_add_c(a[1],b[0],c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 != + stuw t_1,rp(1) !r[1]=c2; + or c_12,c_3,c_12 + + mulx a_2,b_0,t_1 !mul_add_c(a[2],b[0],c3,c1,c2); + addcc c_12,t_1,c_12 != + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw bp(2),b_2 != + mulx a_1,b_1,t_1 !mul_add_c(a[1],b[1],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + lduw bp(3),b_3 + mulx a_0,b_2,t_1 !mul_add_c(a[0],b[2],c3,c1,c2); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(2) !r[2]=c3; + or c_12,c_3,c_12 != + + mulx a_0,b_3,t_1 !mul_add_c(a[0],b[3],c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_1,b_2,t_1 !=!mul_add_c(a[1],b[2],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + lduw ap(3),a_3 + mulx a_2,b_1,t_1 !mul_add_c(a[2],b[1],c1,c2,c3); + addcc c_12,t_1,c_12 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(4),a_4 + mulx a_3,b_0,t_1 !=!mul_add_c(a[3],b[0],c1,c2,c3);!= + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 != + stuw t_1,rp(3) !r[3]=c1; + or c_12,c_3,c_12 + + mulx a_4,b_0,t_1 !mul_add_c(a[4],b[0],c2,c3,c1); + addcc c_12,t_1,c_12 != + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_3,b_1,t_1 !=!mul_add_c(a[3],b[1],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_2,b_2,t_1 !=!mul_add_c(a[2],b[2],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw bp(4),b_4 != + mulx a_1,b_3,t_1 !mul_add_c(a[1],b[3],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + lduw bp(5),b_5 + mulx a_0,b_4,t_1 !mul_add_c(a[0],b[4],c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(4) !r[4]=c2; + or c_12,c_3,c_12 != + + mulx a_0,b_5,t_1 !mul_add_c(a[0],b[5],c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_1,b_4,t_1 !mul_add_c(a[1],b[4],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_2,b_3,t_1 !mul_add_c(a[2],b[3],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_3,b_2,t_1 !mul_add_c(a[3],b[2],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + lduw ap(5),a_5 + mulx a_4,b_1,t_1 !mul_add_c(a[4],b[1],c3,c1,c2); + addcc c_12,t_1,c_12 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(6),a_6 + mulx a_5,b_0,t_1 !=!mul_add_c(a[5],b[0],c3,c1,c2); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 != + stuw t_1,rp(5) !r[5]=c3; + or c_12,c_3,c_12 + + mulx a_6,b_0,t_1 !mul_add_c(a[6],b[0],c1,c2,c3); + addcc c_12,t_1,c_12 != + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_5,b_1,t_1 !=!mul_add_c(a[5],b[1],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_4,b_2,t_1 !=!mul_add_c(a[4],b[2],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_3,b_3,t_1 !=!mul_add_c(a[3],b[3],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_2,b_4,t_1 !=!mul_add_c(a[2],b[4],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw bp(6),b_6 != + mulx a_1,b_5,t_1 !mul_add_c(a[1],b[5],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + lduw bp(7),b_7 + mulx a_0,b_6,t_1 !mul_add_c(a[0],b[6],c1,c2,c3); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(6) !r[6]=c1; + or c_12,c_3,c_12 != + + mulx a_0,b_7,t_1 !mul_add_c(a[0],b[7],c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_1,b_6,t_1 !mul_add_c(a[1],b[6],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_2,b_5,t_1 !mul_add_c(a[2],b[5],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_3,b_4,t_1 !mul_add_c(a[3],b[4],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_4,b_3,t_1 !mul_add_c(a[4],b[3],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_5,b_2,t_1 !mul_add_c(a[5],b[2],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + lduw ap(7),a_7 + mulx a_6,b_1,t_1 !=!mul_add_c(a[6],b[1],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_7,b_0,t_1 !=!mul_add_c(a[7],b[0],c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 != + stuw t_1,rp(7) !r[7]=c2; + or c_12,c_3,c_12 + + mulx a_7,b_1,t_1 !=!mul_add_c(a[7],b[1],c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + mulx a_6,b_2,t_1 !mul_add_c(a[6],b[2],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + mulx a_5,b_3,t_1 !mul_add_c(a[5],b[3],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + mulx a_4,b_4,t_1 !mul_add_c(a[4],b[4],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + mulx a_3,b_5,t_1 !mul_add_c(a[3],b[5],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + mulx a_2,b_6,t_1 !mul_add_c(a[2],b[6],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + mulx a_1,b_7,t_1 !mul_add_c(a[1],b[7],c3,c1,c2); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + srlx t_1,32,c_12 + stuw t_1,rp(8) !r[8]=c3; + or c_12,c_3,c_12 + + mulx a_2,b_7,t_1 !=!mul_add_c(a[2],b[7],c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + mulx a_3,b_6,t_1 !mul_add_c(a[3],b[6],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_4,b_5,t_1 !mul_add_c(a[4],b[5],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_5,b_4,t_1 !mul_add_c(a[5],b[4],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_6,b_3,t_1 !mul_add_c(a[6],b[3],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_7,b_2,t_1 !mul_add_c(a[7],b[2],c1,c2,c3); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(9) !r[9]=c1; + or c_12,c_3,c_12 != + + mulx a_7,b_3,t_1 !mul_add_c(a[7],b[3],c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_6,b_4,t_1 !mul_add_c(a[6],b[4],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_5,b_5,t_1 !mul_add_c(a[5],b[5],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_4,b_6,t_1 !mul_add_c(a[4],b[6],c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_3,b_7,t_1 !mul_add_c(a[3],b[7],c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(10) !r[10]=c2; + or c_12,c_3,c_12 != + + mulx a_4,b_7,t_1 !mul_add_c(a[4],b[7],c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_5,b_6,t_1 !mul_add_c(a[5],b[6],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_6,b_5,t_1 !mul_add_c(a[6],b[5],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_7,b_4,t_1 !mul_add_c(a[7],b[4],c3,c1,c2); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(11) !r[11]=c3; + or c_12,c_3,c_12 != + + mulx a_7,b_5,t_1 !mul_add_c(a[7],b[5],c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_6,b_6,t_1 !mul_add_c(a[6],b[6],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_5,b_7,t_1 !mul_add_c(a[5],b[7],c1,c2,c3); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(12) !r[12]=c1; + or c_12,c_3,c_12 != + + mulx a_6,b_7,t_1 !mul_add_c(a[6],b[7],c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_7,b_6,t_1 !mul_add_c(a[7],b[6],c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + st t_1,rp(13) !r[13]=c2; + or c_12,c_3,c_12 != + + mulx a_7,b_7,t_1 !mul_add_c(a[7],b[7],c3,c1,c2); + addcc c_12,t_1,t_1 + srlx t_1,32,c_12 != + stuw t_1,rp(14) !r[14]=c3; + stuw c_12,rp(15) !r[15]=c1; + + ret + restore %g0,%g0,%o0 != + +.type bn_mul_comba8,#function +.size bn_mul_comba8,(.-bn_mul_comba8) + +.align 32 + +.global bn_mul_comba4 +/* + * void bn_mul_comba4(r,a,b) + * BN_ULONG *r,*a,*b; + */ +bn_mul_comba4: + save %sp,FRAME_SIZE,%sp + lduw ap(0),a_0 + mov 1,t_2 + lduw bp(0),b_0 + sllx t_2,32,t_2 != + lduw bp(1),b_1 + mulx a_0,b_0,t_1 !mul_add_c(a[0],b[0],c1,c2,c3); + srlx t_1,32,c_12 + stuw t_1,rp(0) !=!r[0]=c1; + + lduw ap(1),a_1 + mulx a_0,b_1,t_1 !mul_add_c(a[0],b[1],c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(2),a_2 + mulx a_1,b_0,t_1 !=!mul_add_c(a[1],b[0],c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 != + stuw t_1,rp(1) !r[1]=c2; + or c_12,c_3,c_12 + + mulx a_2,b_0,t_1 !mul_add_c(a[2],b[0],c3,c1,c2); + addcc c_12,t_1,c_12 != + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw bp(2),b_2 != + mulx a_1,b_1,t_1 !mul_add_c(a[1],b[1],c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 != + lduw bp(3),b_3 + mulx a_0,b_2,t_1 !mul_add_c(a[0],b[2],c3,c1,c2); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(2) !r[2]=c3; + or c_12,c_3,c_12 != + + mulx a_0,b_3,t_1 !mul_add_c(a[0],b[3],c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + mulx a_1,b_2,t_1 !mul_add_c(a[1],b[2],c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 != + add c_3,t_2,c_3 + lduw ap(3),a_3 + mulx a_2,b_1,t_1 !mul_add_c(a[2],b[1],c1,c2,c3); + addcc c_12,t_1,c_12 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_3,b_0,t_1 !mul_add_c(a[3],b[0],c1,c2,c3);!= + addcc c_12,t_1,t_1 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(3) !=!r[3]=c1; + or c_12,c_3,c_12 + + mulx a_3,b_1,t_1 !mul_add_c(a[3],b[1],c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_2,b_2,t_1 !mul_add_c(a[2],b[2],c2,c3,c1); + addcc c_12,t_1,c_12 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_1,b_3,t_1 !mul_add_c(a[1],b[3],c2,c3,c1); + addcc c_12,t_1,t_1 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(4) !=!r[4]=c2; + or c_12,c_3,c_12 + + mulx a_2,b_3,t_1 !mul_add_c(a[2],b[3],c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_3,b_2,t_1 !mul_add_c(a[3],b[2],c3,c1,c2); + addcc c_12,t_1,t_1 != + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(5) !=!r[5]=c3; + or c_12,c_3,c_12 + + mulx a_3,b_3,t_1 !mul_add_c(a[3],b[3],c1,c2,c3); + addcc c_12,t_1,t_1 + srlx t_1,32,c_12 != + stuw t_1,rp(6) !r[6]=c1; + stuw c_12,rp(7) !r[7]=c2; + + ret + restore %g0,%g0,%o0 + +.type bn_mul_comba4,#function +.size bn_mul_comba4,(.-bn_mul_comba4) + +.align 32 + +.global bn_sqr_comba8 +bn_sqr_comba8: + save %sp,FRAME_SIZE,%sp + mov 1,t_2 + lduw ap(0),a_0 + sllx t_2,32,t_2 + lduw ap(1),a_1 + mulx a_0,a_0,t_1 !sqr_add_c(a,0,c1,c2,c3); + srlx t_1,32,c_12 + stuw t_1,rp(0) !r[0]=c1; + + lduw ap(2),a_2 + mulx a_0,a_1,t_1 !=!sqr_add_c2(a,1,0,c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(1) !r[1]=c2; + or c_12,c_3,c_12 + + mulx a_2,a_0,t_1 !sqr_add_c2(a,2,0,c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(3),a_3 + mulx a_1,a_1,t_1 !sqr_add_c(a,1,c3,c1,c2); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(2) !r[2]=c3; + or c_12,c_3,c_12 + + mulx a_0,a_3,t_1 !sqr_add_c2(a,3,0,c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(4),a_4 + mulx a_1,a_2,t_1 !sqr_add_c2(a,2,1,c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + st t_1,rp(3) !r[3]=c1; + or c_12,c_3,c_12 + + mulx a_4,a_0,t_1 !sqr_add_c2(a,4,0,c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_3,a_1,t_1 !sqr_add_c2(a,3,1,c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(5),a_5 + mulx a_2,a_2,t_1 !sqr_add_c(a,2,c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(4) !r[4]=c2; + or c_12,c_3,c_12 + + mulx a_0,a_5,t_1 !sqr_add_c2(a,5,0,c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_1,a_4,t_1 !sqr_add_c2(a,4,1,c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(6),a_6 + mulx a_2,a_3,t_1 !sqr_add_c2(a,3,2,c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(5) !r[5]=c3; + or c_12,c_3,c_12 + + mulx a_6,a_0,t_1 !sqr_add_c2(a,6,0,c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_5,a_1,t_1 !sqr_add_c2(a,5,1,c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_4,a_2,t_1 !sqr_add_c2(a,4,2,c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(7),a_7 + mulx a_3,a_3,t_1 !=!sqr_add_c(a,3,c1,c2,c3); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(6) !r[6]=c1; + or c_12,c_3,c_12 + + mulx a_0,a_7,t_1 !sqr_add_c2(a,7,0,c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_1,a_6,t_1 !sqr_add_c2(a,6,1,c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_2,a_5,t_1 !sqr_add_c2(a,5,2,c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_3,a_4,t_1 !sqr_add_c2(a,4,3,c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(7) !r[7]=c2; + or c_12,c_3,c_12 + + mulx a_7,a_1,t_1 !sqr_add_c2(a,7,1,c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_6,a_2,t_1 !sqr_add_c2(a,6,2,c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_5,a_3,t_1 !sqr_add_c2(a,5,3,c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_4,a_4,t_1 !sqr_add_c(a,4,c3,c1,c2); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(8) !r[8]=c3; + or c_12,c_3,c_12 + + mulx a_2,a_7,t_1 !sqr_add_c2(a,7,2,c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_3,a_6,t_1 !sqr_add_c2(a,6,3,c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_4,a_5,t_1 !sqr_add_c2(a,5,4,c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(9) !r[9]=c1; + or c_12,c_3,c_12 + + mulx a_7,a_3,t_1 !sqr_add_c2(a,7,3,c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_6,a_4,t_1 !sqr_add_c2(a,6,4,c2,c3,c1); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_5,a_5,t_1 !sqr_add_c(a,5,c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(10) !r[10]=c2; + or c_12,c_3,c_12 + + mulx a_4,a_7,t_1 !sqr_add_c2(a,7,4,c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_5,a_6,t_1 !sqr_add_c2(a,6,5,c3,c1,c2); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(11) !r[11]=c3; + or c_12,c_3,c_12 + + mulx a_7,a_5,t_1 !sqr_add_c2(a,7,5,c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_6,a_6,t_1 !sqr_add_c(a,6,c1,c2,c3); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(12) !r[12]=c1; + or c_12,c_3,c_12 + + mulx a_6,a_7,t_1 !sqr_add_c2(a,7,6,c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(13) !r[13]=c2; + or c_12,c_3,c_12 + + mulx a_7,a_7,t_1 !sqr_add_c(a,7,c3,c1,c2); + addcc c_12,t_1,t_1 + srlx t_1,32,c_12 + stuw t_1,rp(14) !r[14]=c3; + stuw c_12,rp(15) !r[15]=c1; + + ret + restore %g0,%g0,%o0 + +.type bn_sqr_comba8,#function +.size bn_sqr_comba8,(.-bn_sqr_comba8) + +.align 32 + +.global bn_sqr_comba4 +/* + * void bn_sqr_comba4(r,a) + * BN_ULONG *r,*a; + */ +bn_sqr_comba4: + save %sp,FRAME_SIZE,%sp + mov 1,t_2 + lduw ap(0),a_0 + sllx t_2,32,t_2 + lduw ap(1),a_1 + mulx a_0,a_0,t_1 !sqr_add_c(a,0,c1,c2,c3); + srlx t_1,32,c_12 + stuw t_1,rp(0) !r[0]=c1; + + lduw ap(2),a_2 + mulx a_0,a_1,t_1 !sqr_add_c2(a,1,0,c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(1) !r[1]=c2; + or c_12,c_3,c_12 + + mulx a_2,a_0,t_1 !sqr_add_c2(a,2,0,c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + lduw ap(3),a_3 + mulx a_1,a_1,t_1 !sqr_add_c(a,1,c3,c1,c2); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(2) !r[2]=c3; + or c_12,c_3,c_12 + + mulx a_0,a_3,t_1 !sqr_add_c2(a,3,0,c1,c2,c3); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_1,a_2,t_1 !sqr_add_c2(a,2,1,c1,c2,c3); + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(3) !r[3]=c1; + or c_12,c_3,c_12 + + mulx a_3,a_1,t_1 !sqr_add_c2(a,3,1,c2,c3,c1); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,c_12 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + mulx a_2,a_2,t_1 !sqr_add_c(a,2,c2,c3,c1); + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(4) !r[4]=c2; + or c_12,c_3,c_12 + + mulx a_2,a_3,t_1 !sqr_add_c2(a,3,2,c3,c1,c2); + addcc c_12,t_1,c_12 + clr c_3 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + addcc c_12,t_1,t_1 + bcs,a %xcc,.+8 + add c_3,t_2,c_3 + srlx t_1,32,c_12 + stuw t_1,rp(5) !r[5]=c3; + or c_12,c_3,c_12 + + mulx a_3,a_3,t_1 !sqr_add_c(a,3,c1,c2,c3); + addcc c_12,t_1,t_1 + srlx t_1,32,c_12 + stuw t_1,rp(6) !r[6]=c1; + stuw c_12,rp(7) !r[7]=c2; + + ret + restore %g0,%g0,%o0 + +.type bn_sqr_comba4,#function +.size bn_sqr_comba4,(.-bn_sqr_comba4) + +.align 32 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv9-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv9-mont.pl new file mode 100644 index 00000000..b8fb1e8a --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv9-mont.pl @@ -0,0 +1,606 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# December 2005 +# +# Pure SPARCv9/8+ and IALU-only bn_mul_mont implementation. The reasons +# for undertaken effort are multiple. First of all, UltraSPARC is not +# the whole SPARCv9 universe and other VIS-free implementations deserve +# optimized code as much. Secondly, newly introduced UltraSPARC T1, +# a.k.a. Niagara, has shared FPU and concurrent FPU-intensive pathes, +# such as sparcv9a-mont, will simply sink it. Yes, T1 is equipped with +# several integrated RSA/DSA accelerator circuits accessible through +# kernel driver [only(*)], but having decent user-land software +# implementation is important too. Finally, reasons like desire to +# experiment with dedicated squaring procedure. Yes, this module +# implements one, because it was easiest to draft it in SPARCv9 +# instructions... + +# (*) Engine accessing the driver in question is on my TODO list. +# For reference, acceleator is estimated to give 6 to 10 times +# improvement on single-threaded RSA sign. It should be noted +# that 6-10x improvement coefficient does not actually mean +# something extraordinary in terms of absolute [single-threaded] +# performance, as SPARCv9 instruction set is by all means least +# suitable for high performance crypto among other 64 bit +# platforms. 6-10x factor simply places T1 in same performance +# domain as say AMD64 and IA-64. Improvement of RSA verify don't +# appear impressive at all, but it's the sign operation which is +# far more critical/interesting. + +# You might notice that inner loops are modulo-scheduled:-) This has +# essentially negligible impact on UltraSPARC performance, it's +# Fujitsu SPARC64 V users who should notice and hopefully appreciate +# the advantage... Currently this module surpasses sparcv9a-mont.pl +# by ~20% on UltraSPARC-III and later cores, but recall that sparcv9a +# module still have hidden potential [see TODO list there], which is +# estimated to be larger than 20%... + +# int bn_mul_mont( +$rp="%i0"; # BN_ULONG *rp, +$ap="%i1"; # const BN_ULONG *ap, +$bp="%i2"; # const BN_ULONG *bp, +$np="%i3"; # const BN_ULONG *np, +$n0="%i4"; # const BN_ULONG *n0, +$num="%i5"; # int num); + +$bits=32; +for (@ARGV) { $bits=64 if (/\-m64/ || /\-xarch\=v9/); } +if ($bits==64) { $bias=2047; $frame=192; } +else { $bias=0; $frame=128; } + +$car0="%o0"; +$car1="%o1"; +$car2="%o2"; # 1 bit +$acc0="%o3"; +$acc1="%o4"; +$mask="%g1"; # 32 bits, what a waste... +$tmp0="%g4"; +$tmp1="%g5"; + +$i="%l0"; +$j="%l1"; +$mul0="%l2"; +$mul1="%l3"; +$tp="%l4"; +$apj="%l5"; +$npj="%l6"; +$tpj="%l7"; + +$fname="bn_mul_mont_int"; + +$code=<<___; +.section ".text",#alloc,#execinstr + +.global $fname +.align 32 +$fname: + cmp %o5,4 ! 128 bits minimum + bge,pt %icc,.Lenter + sethi %hi(0xffffffff),$mask + retl + clr %o0 +.align 32 +.Lenter: + save %sp,-$frame,%sp + sll $num,2,$num ! num*=4 + or $mask,%lo(0xffffffff),$mask + ld [$n0],$n0 + cmp $ap,$bp + and $num,$mask,$num + ld [$bp],$mul0 ! bp[0] + nop + + add %sp,$bias,%o7 ! real top of stack + ld [$ap],$car0 ! ap[0] ! redundant in squaring context + sub %o7,$num,%o7 + ld [$ap+4],$apj ! ap[1] + and %o7,-1024,%o7 + ld [$np],$car1 ! np[0] + sub %o7,$bias,%sp ! alloca + ld [$np+4],$npj ! np[1] + be,pt `$bits==32?"%icc":"%xcc"`,.Lbn_sqr_mont + mov 12,$j + + mulx $car0,$mul0,$car0 ! ap[0]*bp[0] + mulx $apj,$mul0,$tmp0 !prologue! ap[1]*bp[0] + and $car0,$mask,$acc0 + add %sp,$bias+$frame,$tp + ld [$ap+8],$apj !prologue! + + mulx $n0,$acc0,$mul1 ! "t[0]"*n0 + and $mul1,$mask,$mul1 + + mulx $car1,$mul1,$car1 ! np[0]*"t[0]"*n0 + mulx $npj,$mul1,$acc1 !prologue! np[1]*"t[0]"*n0 + srlx $car0,32,$car0 + add $acc0,$car1,$car1 + ld [$np+8],$npj !prologue! + srlx $car1,32,$car1 + mov $tmp0,$acc0 !prologue! + +.L1st: + mulx $apj,$mul0,$tmp0 + mulx $npj,$mul1,$tmp1 + add $acc0,$car0,$car0 + ld [$ap+$j],$apj ! ap[j] + and $car0,$mask,$acc0 + add $acc1,$car1,$car1 + ld [$np+$j],$npj ! np[j] + srlx $car0,32,$car0 + add $acc0,$car1,$car1 + add $j,4,$j ! j++ + mov $tmp0,$acc0 + st $car1,[$tp] + cmp $j,$num + mov $tmp1,$acc1 + srlx $car1,32,$car1 + bl %icc,.L1st + add $tp,4,$tp ! tp++ +!.L1st + + mulx $apj,$mul0,$tmp0 !epilogue! + mulx $npj,$mul1,$tmp1 + add $acc0,$car0,$car0 + and $car0,$mask,$acc0 + add $acc1,$car1,$car1 + srlx $car0,32,$car0 + add $acc0,$car1,$car1 + st $car1,[$tp] + srlx $car1,32,$car1 + + add $tmp0,$car0,$car0 + and $car0,$mask,$acc0 + add $tmp1,$car1,$car1 + srlx $car0,32,$car0 + add $acc0,$car1,$car1 + st $car1,[$tp+4] + srlx $car1,32,$car1 + + add $car0,$car1,$car1 + st $car1,[$tp+8] + srlx $car1,32,$car2 + + mov 4,$i ! i++ + ld [$bp+4],$mul0 ! bp[1] +.Louter: + add %sp,$bias+$frame,$tp + ld [$ap],$car0 ! ap[0] + ld [$ap+4],$apj ! ap[1] + ld [$np],$car1 ! np[0] + ld [$np+4],$npj ! np[1] + ld [$tp],$tmp1 ! tp[0] + ld [$tp+4],$tpj ! tp[1] + mov 12,$j + + mulx $car0,$mul0,$car0 + mulx $apj,$mul0,$tmp0 !prologue! + add $tmp1,$car0,$car0 + ld [$ap+8],$apj !prologue! + and $car0,$mask,$acc0 + + mulx $n0,$acc0,$mul1 + and $mul1,$mask,$mul1 + + mulx $car1,$mul1,$car1 + mulx $npj,$mul1,$acc1 !prologue! + srlx $car0,32,$car0 + add $acc0,$car1,$car1 + ld [$np+8],$npj !prologue! + srlx $car1,32,$car1 + mov $tmp0,$acc0 !prologue! + +.Linner: + mulx $apj,$mul0,$tmp0 + mulx $npj,$mul1,$tmp1 + add $tpj,$car0,$car0 + ld [$ap+$j],$apj ! ap[j] + add $acc0,$car0,$car0 + add $acc1,$car1,$car1 + ld [$np+$j],$npj ! np[j] + and $car0,$mask,$acc0 + ld [$tp+8],$tpj ! tp[j] + srlx $car0,32,$car0 + add $acc0,$car1,$car1 + add $j,4,$j ! j++ + mov $tmp0,$acc0 + st $car1,[$tp] ! tp[j-1] + srlx $car1,32,$car1 + mov $tmp1,$acc1 + cmp $j,$num + bl %icc,.Linner + add $tp,4,$tp ! tp++ +!.Linner + + mulx $apj,$mul0,$tmp0 !epilogue! + mulx $npj,$mul1,$tmp1 + add $tpj,$car0,$car0 + add $acc0,$car0,$car0 + ld [$tp+8],$tpj ! tp[j] + and $car0,$mask,$acc0 + add $acc1,$car1,$car1 + srlx $car0,32,$car0 + add $acc0,$car1,$car1 + st $car1,[$tp] ! tp[j-1] + srlx $car1,32,$car1 + + add $tpj,$car0,$car0 + add $tmp0,$car0,$car0 + and $car0,$mask,$acc0 + add $tmp1,$car1,$car1 + add $acc0,$car1,$car1 + st $car1,[$tp+4] ! tp[j-1] + srlx $car0,32,$car0 + add $i,4,$i ! i++ + srlx $car1,32,$car1 + + add $car0,$car1,$car1 + cmp $i,$num + add $car2,$car1,$car1 + st $car1,[$tp+8] + + srlx $car1,32,$car2 + bl,a %icc,.Louter + ld [$bp+$i],$mul0 ! bp[i] +!.Louter + + add $tp,12,$tp + +.Ltail: + add $np,$num,$np + add $rp,$num,$rp + mov $tp,$ap + sub %g0,$num,%o7 ! k=-num + ba .Lsub + subcc %g0,%g0,%g0 ! clear %icc.c +.align 16 +.Lsub: + ld [$tp+%o7],%o0 + ld [$np+%o7],%o1 + subccc %o0,%o1,%o1 ! tp[j]-np[j] + add $rp,%o7,$i + add %o7,4,%o7 + brnz %o7,.Lsub + st %o1,[$i] + subc $car2,0,$car2 ! handle upmost overflow bit + and $tp,$car2,$ap + andn $rp,$car2,$np + or $ap,$np,$ap + sub %g0,$num,%o7 + +.Lcopy: + ld [$ap+%o7],%o0 ! copy or in-place refresh + st %g0,[$tp+%o7] ! zap tp + st %o0,[$rp+%o7] + add %o7,4,%o7 + brnz %o7,.Lcopy + nop + mov 1,%i0 + ret + restore +___ + +######## +######## .Lbn_sqr_mont gives up to 20% *overall* improvement over +######## code without following dedicated squaring procedure. +######## +$sbit="%i2"; # re-use $bp! + +$code.=<<___; +.align 32 +.Lbn_sqr_mont: + mulx $mul0,$mul0,$car0 ! ap[0]*ap[0] + mulx $apj,$mul0,$tmp0 !prologue! + and $car0,$mask,$acc0 + add %sp,$bias+$frame,$tp + ld [$ap+8],$apj !prologue! + + mulx $n0,$acc0,$mul1 ! "t[0]"*n0 + srlx $car0,32,$car0 + and $mul1,$mask,$mul1 + + mulx $car1,$mul1,$car1 ! np[0]*"t[0]"*n0 + mulx $npj,$mul1,$acc1 !prologue! + and $car0,1,$sbit + ld [$np+8],$npj !prologue! + srlx $car0,1,$car0 + add $acc0,$car1,$car1 + srlx $car1,32,$car1 + mov $tmp0,$acc0 !prologue! + +.Lsqr_1st: + mulx $apj,$mul0,$tmp0 + mulx $npj,$mul1,$tmp1 + add $acc0,$car0,$car0 ! ap[j]*a0+c0 + add $acc1,$car1,$car1 + ld [$ap+$j],$apj ! ap[j] + and $car0,$mask,$acc0 + ld [$np+$j],$npj ! np[j] + srlx $car0,32,$car0 + add $acc0,$acc0,$acc0 + or $sbit,$acc0,$acc0 + mov $tmp1,$acc1 + srlx $acc0,32,$sbit + add $j,4,$j ! j++ + and $acc0,$mask,$acc0 + cmp $j,$num + add $acc0,$car1,$car1 + st $car1,[$tp] + mov $tmp0,$acc0 + srlx $car1,32,$car1 + bl %icc,.Lsqr_1st + add $tp,4,$tp ! tp++ +!.Lsqr_1st + + mulx $apj,$mul0,$tmp0 ! epilogue + mulx $npj,$mul1,$tmp1 + add $acc0,$car0,$car0 ! ap[j]*a0+c0 + add $acc1,$car1,$car1 + and $car0,$mask,$acc0 + srlx $car0,32,$car0 + add $acc0,$acc0,$acc0 + or $sbit,$acc0,$acc0 + srlx $acc0,32,$sbit + and $acc0,$mask,$acc0 + add $acc0,$car1,$car1 + st $car1,[$tp] + srlx $car1,32,$car1 + + add $tmp0,$car0,$car0 ! ap[j]*a0+c0 + add $tmp1,$car1,$car1 + and $car0,$mask,$acc0 + srlx $car0,32,$car0 + add $acc0,$acc0,$acc0 + or $sbit,$acc0,$acc0 + srlx $acc0,32,$sbit + and $acc0,$mask,$acc0 + add $acc0,$car1,$car1 + st $car1,[$tp+4] + srlx $car1,32,$car1 + + add $car0,$car0,$car0 + or $sbit,$car0,$car0 + add $car0,$car1,$car1 + st $car1,[$tp+8] + srlx $car1,32,$car2 + + ld [%sp+$bias+$frame],$tmp0 ! tp[0] + ld [%sp+$bias+$frame+4],$tmp1 ! tp[1] + ld [%sp+$bias+$frame+8],$tpj ! tp[2] + ld [$ap+4],$mul0 ! ap[1] + ld [$ap+8],$apj ! ap[2] + ld [$np],$car1 ! np[0] + ld [$np+4],$npj ! np[1] + mulx $n0,$tmp0,$mul1 + + mulx $mul0,$mul0,$car0 + and $mul1,$mask,$mul1 + + mulx $car1,$mul1,$car1 + mulx $npj,$mul1,$acc1 + add $tmp0,$car1,$car1 + and $car0,$mask,$acc0 + ld [$np+8],$npj ! np[2] + srlx $car1,32,$car1 + add $tmp1,$car1,$car1 + srlx $car0,32,$car0 + add $acc0,$car1,$car1 + and $car0,1,$sbit + add $acc1,$car1,$car1 + srlx $car0,1,$car0 + mov 12,$j + st $car1,[%sp+$bias+$frame] ! tp[0]= + srlx $car1,32,$car1 + add %sp,$bias+$frame+4,$tp + +.Lsqr_2nd: + mulx $apj,$mul0,$acc0 + mulx $npj,$mul1,$acc1 + add $acc0,$car0,$car0 + add $tpj,$car1,$car1 + ld [$ap+$j],$apj ! ap[j] + and $car0,$mask,$acc0 + ld [$np+$j],$npj ! np[j] + srlx $car0,32,$car0 + add $acc1,$car1,$car1 + ld [$tp+8],$tpj ! tp[j] + add $acc0,$acc0,$acc0 + add $j,4,$j ! j++ + or $sbit,$acc0,$acc0 + srlx $acc0,32,$sbit + and $acc0,$mask,$acc0 + cmp $j,$num + add $acc0,$car1,$car1 + st $car1,[$tp] ! tp[j-1] + srlx $car1,32,$car1 + bl %icc,.Lsqr_2nd + add $tp,4,$tp ! tp++ +!.Lsqr_2nd + + mulx $apj,$mul0,$acc0 + mulx $npj,$mul1,$acc1 + add $acc0,$car0,$car0 + add $tpj,$car1,$car1 + and $car0,$mask,$acc0 + srlx $car0,32,$car0 + add $acc1,$car1,$car1 + add $acc0,$acc0,$acc0 + or $sbit,$acc0,$acc0 + srlx $acc0,32,$sbit + and $acc0,$mask,$acc0 + add $acc0,$car1,$car1 + st $car1,[$tp] ! tp[j-1] + srlx $car1,32,$car1 + + add $car0,$car0,$car0 + or $sbit,$car0,$car0 + add $car0,$car1,$car1 + add $car2,$car1,$car1 + st $car1,[$tp+4] + srlx $car1,32,$car2 + + ld [%sp+$bias+$frame],$tmp1 ! tp[0] + ld [%sp+$bias+$frame+4],$tpj ! tp[1] + ld [$ap+8],$mul0 ! ap[2] + ld [$np],$car1 ! np[0] + ld [$np+4],$npj ! np[1] + mulx $n0,$tmp1,$mul1 + and $mul1,$mask,$mul1 + mov 8,$i + + mulx $mul0,$mul0,$car0 + mulx $car1,$mul1,$car1 + and $car0,$mask,$acc0 + add $tmp1,$car1,$car1 + srlx $car0,32,$car0 + add %sp,$bias+$frame,$tp + srlx $car1,32,$car1 + and $car0,1,$sbit + srlx $car0,1,$car0 + mov 4,$j + +.Lsqr_outer: +.Lsqr_inner1: + mulx $npj,$mul1,$acc1 + add $tpj,$car1,$car1 + add $j,4,$j + ld [$tp+8],$tpj + cmp $j,$i + add $acc1,$car1,$car1 + ld [$np+$j],$npj + st $car1,[$tp] + srlx $car1,32,$car1 + bl %icc,.Lsqr_inner1 + add $tp,4,$tp +!.Lsqr_inner1 + + add $j,4,$j + ld [$ap+$j],$apj ! ap[j] + mulx $npj,$mul1,$acc1 + add $tpj,$car1,$car1 + ld [$np+$j],$npj ! np[j] + add $acc0,$car1,$car1 + ld [$tp+8],$tpj ! tp[j] + add $acc1,$car1,$car1 + st $car1,[$tp] + srlx $car1,32,$car1 + + add $j,4,$j + cmp $j,$num + be,pn %icc,.Lsqr_no_inner2 + add $tp,4,$tp + +.Lsqr_inner2: + mulx $apj,$mul0,$acc0 + mulx $npj,$mul1,$acc1 + add $tpj,$car1,$car1 + add $acc0,$car0,$car0 + ld [$ap+$j],$apj ! ap[j] + and $car0,$mask,$acc0 + ld [$np+$j],$npj ! np[j] + srlx $car0,32,$car0 + add $acc0,$acc0,$acc0 + ld [$tp+8],$tpj ! tp[j] + or $sbit,$acc0,$acc0 + add $j,4,$j ! j++ + srlx $acc0,32,$sbit + and $acc0,$mask,$acc0 + cmp $j,$num + add $acc0,$car1,$car1 + add $acc1,$car1,$car1 + st $car1,[$tp] ! tp[j-1] + srlx $car1,32,$car1 + bl %icc,.Lsqr_inner2 + add $tp,4,$tp ! tp++ + +.Lsqr_no_inner2: + mulx $apj,$mul0,$acc0 + mulx $npj,$mul1,$acc1 + add $tpj,$car1,$car1 + add $acc0,$car0,$car0 + and $car0,$mask,$acc0 + srlx $car0,32,$car0 + add $acc0,$acc0,$acc0 + or $sbit,$acc0,$acc0 + srlx $acc0,32,$sbit + and $acc0,$mask,$acc0 + add $acc0,$car1,$car1 + add $acc1,$car1,$car1 + st $car1,[$tp] ! tp[j-1] + srlx $car1,32,$car1 + + add $car0,$car0,$car0 + or $sbit,$car0,$car0 + add $car0,$car1,$car1 + add $car2,$car1,$car1 + st $car1,[$tp+4] + srlx $car1,32,$car2 + + add $i,4,$i ! i++ + ld [%sp+$bias+$frame],$tmp1 ! tp[0] + ld [%sp+$bias+$frame+4],$tpj ! tp[1] + ld [$ap+$i],$mul0 ! ap[j] + ld [$np],$car1 ! np[0] + ld [$np+4],$npj ! np[1] + mulx $n0,$tmp1,$mul1 + and $mul1,$mask,$mul1 + add $i,4,$tmp0 + + mulx $mul0,$mul0,$car0 + mulx $car1,$mul1,$car1 + and $car0,$mask,$acc0 + add $tmp1,$car1,$car1 + srlx $car0,32,$car0 + add %sp,$bias+$frame,$tp + srlx $car1,32,$car1 + and $car0,1,$sbit + srlx $car0,1,$car0 + + cmp $tmp0,$num ! i" +.align 32 +___ +$code =~ s/\`([^\`]*)\`/eval($1)/gem; +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv9a-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv9a-mont.pl new file mode 100755 index 00000000..a14205f2 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/sparcv9a-mont.pl @@ -0,0 +1,882 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# October 2005 +# +# "Teaser" Montgomery multiplication module for UltraSPARC. Why FPU? +# Because unlike integer multiplier, which simply stalls whole CPU, +# FPU is fully pipelined and can effectively emit 48 bit partial +# product every cycle. Why not blended SPARC v9? One can argue that +# making this module dependent on UltraSPARC VIS extension limits its +# binary compatibility. Well yes, it does exclude SPARC64 prior-V(!) +# implementations from compatibility matrix. But the rest, whole Sun +# UltraSPARC family and brand new Fujitsu's SPARC64 V, all support +# VIS extension instructions used in this module. This is considered +# good enough to not care about HAL SPARC64 users [if any] who have +# integer-only pure SPARCv9 module to "fall down" to. + +# USI&II cores currently exhibit uniform 2x improvement [over pre- +# bn_mul_mont codebase] for all key lengths and benchmarks. On USIII +# performance improves few percents for shorter keys and worsens few +# percents for longer keys. This is because USIII integer multiplier +# is >3x faster than USI&II one, which is harder to match [but see +# TODO list below]. It should also be noted that SPARC64 V features +# out-of-order execution, which *might* mean that integer multiplier +# is pipelined, which in turn *might* be impossible to match... On +# additional note, SPARC64 V implements FP Multiply-Add instruction, +# which is perfectly usable in this context... In other words, as far +# as Fujitsu SPARC64 V goes, talk to the author:-) + +# The implementation implies following "non-natural" limitations on +# input arguments: +# - num may not be less than 4; +# - num has to be even; +# Failure to meet either condition has no fatal effects, simply +# doesn't give any performance gain. + +# TODO: +# - modulo-schedule inner loop for better performance (on in-order +# execution core such as UltraSPARC this shall result in further +# noticeable(!) improvement); +# - dedicated squaring procedure[?]; + +###################################################################### +# November 2006 +# +# Modulo-scheduled inner loops allow to interleave floating point and +# integer instructions and minimize Read-After-Write penalties. This +# results in *further* 20-50% perfromance improvement [depending on +# key length, more for longer keys] on USI&II cores and 30-80% - on +# USIII&IV. + +$fname="bn_mul_mont_fpu"; +$bits=32; +for (@ARGV) { $bits=64 if (/\-m64/ || /\-xarch\=v9/); } + +if ($bits==64) { + $bias=2047; + $frame=192; +} else { + $bias=0; + $frame=128; # 96 rounded up to largest known cache-line +} +$locals=64; + +# In order to provide for 32-/64-bit ABI duality, I keep integers wider +# than 32 bit in %g1-%g4 and %o0-%o5. %l0-%l7 and %i0-%i5 are used +# exclusively for pointers, indexes and other small values... +# int bn_mul_mont( +$rp="%i0"; # BN_ULONG *rp, +$ap="%i1"; # const BN_ULONG *ap, +$bp="%i2"; # const BN_ULONG *bp, +$np="%i3"; # const BN_ULONG *np, +$n0="%i4"; # const BN_ULONG *n0, +$num="%i5"; # int num); + +$tp="%l0"; # t[num] +$ap_l="%l1"; # a[num],n[num] are smashed to 32-bit words and saved +$ap_h="%l2"; # to these four vectors as double-precision FP values. +$np_l="%l3"; # This way a bunch of fxtods are eliminated in second +$np_h="%l4"; # loop and L1-cache aliasing is minimized... +$i="%l5"; +$j="%l6"; +$mask="%l7"; # 16-bit mask, 0xffff + +$n0="%g4"; # reassigned(!) to "64-bit" register +$carry="%i4"; # %i4 reused(!) for a carry bit + +# FP register naming chart +# +# ..HILO +# dcba +# -------- +# LOa +# LOb +# LOc +# LOd +# HIa +# HIb +# HIc +# HId +# ..a +# ..b +$ba="%f0"; $bb="%f2"; $bc="%f4"; $bd="%f6"; +$na="%f8"; $nb="%f10"; $nc="%f12"; $nd="%f14"; +$alo="%f16"; $alo_="%f17"; $ahi="%f18"; $ahi_="%f19"; +$nlo="%f20"; $nlo_="%f21"; $nhi="%f22"; $nhi_="%f23"; + +$dota="%f24"; $dotb="%f26"; + +$aloa="%f32"; $alob="%f34"; $aloc="%f36"; $alod="%f38"; +$ahia="%f40"; $ahib="%f42"; $ahic="%f44"; $ahid="%f46"; +$nloa="%f48"; $nlob="%f50"; $nloc="%f52"; $nlod="%f54"; +$nhia="%f56"; $nhib="%f58"; $nhic="%f60"; $nhid="%f62"; + +$ASI_FL16_P=0xD2; # magic ASI value to engage 16-bit FP load + +$code=<<___; +.section ".text",#alloc,#execinstr + +.global $fname +.align 32 +$fname: + save %sp,-$frame-$locals,%sp + + cmp $num,4 + bl,a,pn %icc,.Lret + clr %i0 + andcc $num,1,%g0 ! $num has to be even... + bnz,a,pn %icc,.Lret + clr %i0 ! signal "unsupported input value" + + srl $num,1,$num + sethi %hi(0xffff),$mask + ld [%i4+0],$n0 ! $n0 reassigned, remember? + or $mask,%lo(0xffff),$mask + ld [%i4+4],%o0 + sllx %o0,32,%o0 + or %o0,$n0,$n0 ! $n0=n0[1].n0[0] + + sll $num,3,$num ! num*=8 + + add %sp,$bias,%o0 ! real top of stack + sll $num,2,%o1 + add %o1,$num,%o1 ! %o1=num*5 + sub %o0,%o1,%o0 + and %o0,-2048,%o0 ! optimize TLB utilization + sub %o0,$bias,%sp ! alloca(5*num*8) + + rd %asi,%o7 ! save %asi + add %sp,$bias+$frame+$locals,$tp + add $tp,$num,$ap_l + add $ap_l,$num,$ap_l ! [an]p_[lh] point at the vectors' ends ! + add $ap_l,$num,$ap_h + add $ap_h,$num,$np_l + add $np_l,$num,$np_h + + wr %g0,$ASI_FL16_P,%asi ! setup %asi for 16-bit FP loads + + add $rp,$num,$rp ! readjust input pointers to point + add $ap,$num,$ap ! at the ends too... + add $bp,$num,$bp + add $np,$num,$np + + stx %o7,[%sp+$bias+$frame+48] ! save %asi + + sub %g0,$num,$i ! i=-num + sub %g0,$num,$j ! j=-num + + add $ap,$j,%o3 + add $bp,$i,%o4 + + ld [%o3+4],%g1 ! bp[0] + ld [%o3+0],%o0 + ld [%o4+4],%g5 ! ap[0] + sllx %g1,32,%g1 + ld [%o4+0],%o1 + sllx %g5,32,%g5 + or %g1,%o0,%o0 + or %g5,%o1,%o1 + + add $np,$j,%o5 + + mulx %o1,%o0,%o0 ! ap[0]*bp[0] + mulx $n0,%o0,%o0 ! ap[0]*bp[0]*n0 + stx %o0,[%sp+$bias+$frame+0] + + ld [%o3+0],$alo_ ! load a[j] as pair of 32-bit words + fzeros $alo + ld [%o3+4],$ahi_ + fzeros $ahi + ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words + fzeros $nlo + ld [%o5+4],$nhi_ + fzeros $nhi + + ! transfer b[i] to FPU as 4x16-bit values + ldda [%o4+2]%asi,$ba + fxtod $alo,$alo + ldda [%o4+0]%asi,$bb + fxtod $ahi,$ahi + ldda [%o4+6]%asi,$bc + fxtod $nlo,$nlo + ldda [%o4+4]%asi,$bd + fxtod $nhi,$nhi + + ! transfer ap[0]*b[0]*n0 to FPU as 4x16-bit values + ldda [%sp+$bias+$frame+6]%asi,$na + fxtod $ba,$ba + ldda [%sp+$bias+$frame+4]%asi,$nb + fxtod $bb,$bb + ldda [%sp+$bias+$frame+2]%asi,$nc + fxtod $bc,$bc + ldda [%sp+$bias+$frame+0]%asi,$nd + fxtod $bd,$bd + + std $alo,[$ap_l+$j] ! save smashed ap[j] in double format + fxtod $na,$na + std $ahi,[$ap_h+$j] + fxtod $nb,$nb + std $nlo,[$np_l+$j] ! save smashed np[j] in double format + fxtod $nc,$nc + std $nhi,[$np_h+$j] + fxtod $nd,$nd + + fmuld $alo,$ba,$aloa + fmuld $nlo,$na,$nloa + fmuld $alo,$bb,$alob + fmuld $nlo,$nb,$nlob + fmuld $alo,$bc,$aloc + faddd $aloa,$nloa,$nloa + fmuld $nlo,$nc,$nloc + fmuld $alo,$bd,$alod + faddd $alob,$nlob,$nlob + fmuld $nlo,$nd,$nlod + fmuld $ahi,$ba,$ahia + faddd $aloc,$nloc,$nloc + fmuld $nhi,$na,$nhia + fmuld $ahi,$bb,$ahib + faddd $alod,$nlod,$nlod + fmuld $nhi,$nb,$nhib + fmuld $ahi,$bc,$ahic + faddd $ahia,$nhia,$nhia + fmuld $nhi,$nc,$nhic + fmuld $ahi,$bd,$ahid + faddd $ahib,$nhib,$nhib + fmuld $nhi,$nd,$nhid + + faddd $ahic,$nhic,$dota ! $nhic + faddd $ahid,$nhid,$dotb ! $nhid + + faddd $nloc,$nhia,$nloc + faddd $nlod,$nhib,$nlod + + fdtox $nloa,$nloa + fdtox $nlob,$nlob + fdtox $nloc,$nloc + fdtox $nlod,$nlod + + std $nloa,[%sp+$bias+$frame+0] + add $j,8,$j + std $nlob,[%sp+$bias+$frame+8] + add $ap,$j,%o4 + std $nloc,[%sp+$bias+$frame+16] + add $np,$j,%o5 + std $nlod,[%sp+$bias+$frame+24] + + ld [%o4+0],$alo_ ! load a[j] as pair of 32-bit words + fzeros $alo + ld [%o4+4],$ahi_ + fzeros $ahi + ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words + fzeros $nlo + ld [%o5+4],$nhi_ + fzeros $nhi + + fxtod $alo,$alo + fxtod $ahi,$ahi + fxtod $nlo,$nlo + fxtod $nhi,$nhi + + ldx [%sp+$bias+$frame+0],%o0 + fmuld $alo,$ba,$aloa + ldx [%sp+$bias+$frame+8],%o1 + fmuld $nlo,$na,$nloa + ldx [%sp+$bias+$frame+16],%o2 + fmuld $alo,$bb,$alob + ldx [%sp+$bias+$frame+24],%o3 + fmuld $nlo,$nb,$nlob + + srlx %o0,16,%o7 + std $alo,[$ap_l+$j] ! save smashed ap[j] in double format + fmuld $alo,$bc,$aloc + add %o7,%o1,%o1 + std $ahi,[$ap_h+$j] + faddd $aloa,$nloa,$nloa + fmuld $nlo,$nc,$nloc + srlx %o1,16,%o7 + std $nlo,[$np_l+$j] ! save smashed np[j] in double format + fmuld $alo,$bd,$alod + add %o7,%o2,%o2 + std $nhi,[$np_h+$j] + faddd $alob,$nlob,$nlob + fmuld $nlo,$nd,$nlod + srlx %o2,16,%o7 + fmuld $ahi,$ba,$ahia + add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15] + faddd $aloc,$nloc,$nloc + fmuld $nhi,$na,$nhia + !and %o0,$mask,%o0 + !and %o1,$mask,%o1 + !and %o2,$mask,%o2 + !sllx %o1,16,%o1 + !sllx %o2,32,%o2 + !sllx %o3,48,%o7 + !or %o1,%o0,%o0 + !or %o2,%o0,%o0 + !or %o7,%o0,%o0 ! 64-bit result + srlx %o3,16,%g1 ! 34-bit carry + fmuld $ahi,$bb,$ahib + + faddd $alod,$nlod,$nlod + fmuld $nhi,$nb,$nhib + fmuld $ahi,$bc,$ahic + faddd $ahia,$nhia,$nhia + fmuld $nhi,$nc,$nhic + fmuld $ahi,$bd,$ahid + faddd $ahib,$nhib,$nhib + fmuld $nhi,$nd,$nhid + + faddd $dota,$nloa,$nloa + faddd $dotb,$nlob,$nlob + faddd $ahic,$nhic,$dota ! $nhic + faddd $ahid,$nhid,$dotb ! $nhid + + faddd $nloc,$nhia,$nloc + faddd $nlod,$nhib,$nlod + + fdtox $nloa,$nloa + fdtox $nlob,$nlob + fdtox $nloc,$nloc + fdtox $nlod,$nlod + + std $nloa,[%sp+$bias+$frame+0] + std $nlob,[%sp+$bias+$frame+8] + addcc $j,8,$j + std $nloc,[%sp+$bias+$frame+16] + bz,pn %icc,.L1stskip + std $nlod,[%sp+$bias+$frame+24] + +.align 32 ! incidentally already aligned ! +.L1st: + add $ap,$j,%o4 + add $np,$j,%o5 + ld [%o4+0],$alo_ ! load a[j] as pair of 32-bit words + fzeros $alo + ld [%o4+4],$ahi_ + fzeros $ahi + ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words + fzeros $nlo + ld [%o5+4],$nhi_ + fzeros $nhi + + fxtod $alo,$alo + fxtod $ahi,$ahi + fxtod $nlo,$nlo + fxtod $nhi,$nhi + + ldx [%sp+$bias+$frame+0],%o0 + fmuld $alo,$ba,$aloa + ldx [%sp+$bias+$frame+8],%o1 + fmuld $nlo,$na,$nloa + ldx [%sp+$bias+$frame+16],%o2 + fmuld $alo,$bb,$alob + ldx [%sp+$bias+$frame+24],%o3 + fmuld $nlo,$nb,$nlob + + srlx %o0,16,%o7 + std $alo,[$ap_l+$j] ! save smashed ap[j] in double format + fmuld $alo,$bc,$aloc + add %o7,%o1,%o1 + std $ahi,[$ap_h+$j] + faddd $aloa,$nloa,$nloa + fmuld $nlo,$nc,$nloc + srlx %o1,16,%o7 + std $nlo,[$np_l+$j] ! save smashed np[j] in double format + fmuld $alo,$bd,$alod + add %o7,%o2,%o2 + std $nhi,[$np_h+$j] + faddd $alob,$nlob,$nlob + fmuld $nlo,$nd,$nlod + srlx %o2,16,%o7 + fmuld $ahi,$ba,$ahia + add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15] + and %o0,$mask,%o0 + faddd $aloc,$nloc,$nloc + fmuld $nhi,$na,$nhia + and %o1,$mask,%o1 + and %o2,$mask,%o2 + fmuld $ahi,$bb,$ahib + sllx %o1,16,%o1 + faddd $alod,$nlod,$nlod + fmuld $nhi,$nb,$nhib + sllx %o2,32,%o2 + fmuld $ahi,$bc,$ahic + sllx %o3,48,%o7 + or %o1,%o0,%o0 + faddd $ahia,$nhia,$nhia + fmuld $nhi,$nc,$nhic + or %o2,%o0,%o0 + fmuld $ahi,$bd,$ahid + or %o7,%o0,%o0 ! 64-bit result + faddd $ahib,$nhib,$nhib + fmuld $nhi,$nd,$nhid + addcc %g1,%o0,%o0 + faddd $dota,$nloa,$nloa + srlx %o3,16,%g1 ! 34-bit carry + faddd $dotb,$nlob,$nlob + bcs,a %xcc,.+8 + add %g1,1,%g1 + + stx %o0,[$tp] ! tp[j-1]= + + faddd $ahic,$nhic,$dota ! $nhic + faddd $ahid,$nhid,$dotb ! $nhid + + faddd $nloc,$nhia,$nloc + faddd $nlod,$nhib,$nlod + + fdtox $nloa,$nloa + fdtox $nlob,$nlob + fdtox $nloc,$nloc + fdtox $nlod,$nlod + + std $nloa,[%sp+$bias+$frame+0] + std $nlob,[%sp+$bias+$frame+8] + std $nloc,[%sp+$bias+$frame+16] + std $nlod,[%sp+$bias+$frame+24] + + addcc $j,8,$j + bnz,pt %icc,.L1st + add $tp,8,$tp + +.L1stskip: + fdtox $dota,$dota + fdtox $dotb,$dotb + + ldx [%sp+$bias+$frame+0],%o0 + ldx [%sp+$bias+$frame+8],%o1 + ldx [%sp+$bias+$frame+16],%o2 + ldx [%sp+$bias+$frame+24],%o3 + + srlx %o0,16,%o7 + std $dota,[%sp+$bias+$frame+32] + add %o7,%o1,%o1 + std $dotb,[%sp+$bias+$frame+40] + srlx %o1,16,%o7 + add %o7,%o2,%o2 + srlx %o2,16,%o7 + add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15] + and %o0,$mask,%o0 + and %o1,$mask,%o1 + and %o2,$mask,%o2 + sllx %o1,16,%o1 + sllx %o2,32,%o2 + sllx %o3,48,%o7 + or %o1,%o0,%o0 + or %o2,%o0,%o0 + or %o7,%o0,%o0 ! 64-bit result + ldx [%sp+$bias+$frame+32],%o4 + addcc %g1,%o0,%o0 + ldx [%sp+$bias+$frame+40],%o5 + srlx %o3,16,%g1 ! 34-bit carry + bcs,a %xcc,.+8 + add %g1,1,%g1 + + stx %o0,[$tp] ! tp[j-1]= + add $tp,8,$tp + + srlx %o4,16,%o7 + add %o7,%o5,%o5 + and %o4,$mask,%o4 + sllx %o5,16,%o7 + or %o7,%o4,%o4 + addcc %g1,%o4,%o4 + srlx %o5,48,%g1 + bcs,a %xcc,.+8 + add %g1,1,%g1 + + mov %g1,$carry + stx %o4,[$tp] ! tp[num-1]= + + ba .Louter + add $i,8,$i +.align 32 +.Louter: + sub %g0,$num,$j ! j=-num + add %sp,$bias+$frame+$locals,$tp + + add $ap,$j,%o3 + add $bp,$i,%o4 + + ld [%o3+4],%g1 ! bp[i] + ld [%o3+0],%o0 + ld [%o4+4],%g5 ! ap[0] + sllx %g1,32,%g1 + ld [%o4+0],%o1 + sllx %g5,32,%g5 + or %g1,%o0,%o0 + or %g5,%o1,%o1 + + ldx [$tp],%o2 ! tp[0] + mulx %o1,%o0,%o0 + addcc %o2,%o0,%o0 + mulx $n0,%o0,%o0 ! (ap[0]*bp[i]+t[0])*n0 + stx %o0,[%sp+$bias+$frame+0] + + ! transfer b[i] to FPU as 4x16-bit values + ldda [%o4+2]%asi,$ba + ldda [%o4+0]%asi,$bb + ldda [%o4+6]%asi,$bc + ldda [%o4+4]%asi,$bd + + ! transfer (ap[0]*b[i]+t[0])*n0 to FPU as 4x16-bit values + ldda [%sp+$bias+$frame+6]%asi,$na + fxtod $ba,$ba + ldda [%sp+$bias+$frame+4]%asi,$nb + fxtod $bb,$bb + ldda [%sp+$bias+$frame+2]%asi,$nc + fxtod $bc,$bc + ldda [%sp+$bias+$frame+0]%asi,$nd + fxtod $bd,$bd + ldd [$ap_l+$j],$alo ! load a[j] in double format + fxtod $na,$na + ldd [$ap_h+$j],$ahi + fxtod $nb,$nb + ldd [$np_l+$j],$nlo ! load n[j] in double format + fxtod $nc,$nc + ldd [$np_h+$j],$nhi + fxtod $nd,$nd + + fmuld $alo,$ba,$aloa + fmuld $nlo,$na,$nloa + fmuld $alo,$bb,$alob + fmuld $nlo,$nb,$nlob + fmuld $alo,$bc,$aloc + faddd $aloa,$nloa,$nloa + fmuld $nlo,$nc,$nloc + fmuld $alo,$bd,$alod + faddd $alob,$nlob,$nlob + fmuld $nlo,$nd,$nlod + fmuld $ahi,$ba,$ahia + faddd $aloc,$nloc,$nloc + fmuld $nhi,$na,$nhia + fmuld $ahi,$bb,$ahib + faddd $alod,$nlod,$nlod + fmuld $nhi,$nb,$nhib + fmuld $ahi,$bc,$ahic + faddd $ahia,$nhia,$nhia + fmuld $nhi,$nc,$nhic + fmuld $ahi,$bd,$ahid + faddd $ahib,$nhib,$nhib + fmuld $nhi,$nd,$nhid + + faddd $ahic,$nhic,$dota ! $nhic + faddd $ahid,$nhid,$dotb ! $nhid + + faddd $nloc,$nhia,$nloc + faddd $nlod,$nhib,$nlod + + fdtox $nloa,$nloa + fdtox $nlob,$nlob + fdtox $nloc,$nloc + fdtox $nlod,$nlod + + std $nloa,[%sp+$bias+$frame+0] + std $nlob,[%sp+$bias+$frame+8] + std $nloc,[%sp+$bias+$frame+16] + add $j,8,$j + std $nlod,[%sp+$bias+$frame+24] + + ldd [$ap_l+$j],$alo ! load a[j] in double format + ldd [$ap_h+$j],$ahi + ldd [$np_l+$j],$nlo ! load n[j] in double format + ldd [$np_h+$j],$nhi + + fmuld $alo,$ba,$aloa + fmuld $nlo,$na,$nloa + fmuld $alo,$bb,$alob + fmuld $nlo,$nb,$nlob + fmuld $alo,$bc,$aloc + ldx [%sp+$bias+$frame+0],%o0 + faddd $aloa,$nloa,$nloa + fmuld $nlo,$nc,$nloc + ldx [%sp+$bias+$frame+8],%o1 + fmuld $alo,$bd,$alod + ldx [%sp+$bias+$frame+16],%o2 + faddd $alob,$nlob,$nlob + fmuld $nlo,$nd,$nlod + ldx [%sp+$bias+$frame+24],%o3 + fmuld $ahi,$ba,$ahia + + srlx %o0,16,%o7 + faddd $aloc,$nloc,$nloc + fmuld $nhi,$na,$nhia + add %o7,%o1,%o1 + fmuld $ahi,$bb,$ahib + srlx %o1,16,%o7 + faddd $alod,$nlod,$nlod + fmuld $nhi,$nb,$nhib + add %o7,%o2,%o2 + fmuld $ahi,$bc,$ahic + srlx %o2,16,%o7 + faddd $ahia,$nhia,$nhia + fmuld $nhi,$nc,$nhic + add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15] + ! why? + and %o0,$mask,%o0 + fmuld $ahi,$bd,$ahid + and %o1,$mask,%o1 + and %o2,$mask,%o2 + faddd $ahib,$nhib,$nhib + fmuld $nhi,$nd,$nhid + sllx %o1,16,%o1 + faddd $dota,$nloa,$nloa + sllx %o2,32,%o2 + faddd $dotb,$nlob,$nlob + sllx %o3,48,%o7 + or %o1,%o0,%o0 + faddd $ahic,$nhic,$dota ! $nhic + or %o2,%o0,%o0 + faddd $ahid,$nhid,$dotb ! $nhid + or %o7,%o0,%o0 ! 64-bit result + ldx [$tp],%o7 + faddd $nloc,$nhia,$nloc + addcc %o7,%o0,%o0 + ! end-of-why? + faddd $nlod,$nhib,$nlod + srlx %o3,16,%g1 ! 34-bit carry + fdtox $nloa,$nloa + bcs,a %xcc,.+8 + add %g1,1,%g1 + + fdtox $nlob,$nlob + fdtox $nloc,$nloc + fdtox $nlod,$nlod + + std $nloa,[%sp+$bias+$frame+0] + std $nlob,[%sp+$bias+$frame+8] + addcc $j,8,$j + std $nloc,[%sp+$bias+$frame+16] + bz,pn %icc,.Linnerskip + std $nlod,[%sp+$bias+$frame+24] + + ba .Linner + nop +.align 32 +.Linner: + ldd [$ap_l+$j],$alo ! load a[j] in double format + ldd [$ap_h+$j],$ahi + ldd [$np_l+$j],$nlo ! load n[j] in double format + ldd [$np_h+$j],$nhi + + fmuld $alo,$ba,$aloa + fmuld $nlo,$na,$nloa + fmuld $alo,$bb,$alob + fmuld $nlo,$nb,$nlob + fmuld $alo,$bc,$aloc + ldx [%sp+$bias+$frame+0],%o0 + faddd $aloa,$nloa,$nloa + fmuld $nlo,$nc,$nloc + ldx [%sp+$bias+$frame+8],%o1 + fmuld $alo,$bd,$alod + ldx [%sp+$bias+$frame+16],%o2 + faddd $alob,$nlob,$nlob + fmuld $nlo,$nd,$nlod + ldx [%sp+$bias+$frame+24],%o3 + fmuld $ahi,$ba,$ahia + + srlx %o0,16,%o7 + faddd $aloc,$nloc,$nloc + fmuld $nhi,$na,$nhia + add %o7,%o1,%o1 + fmuld $ahi,$bb,$ahib + srlx %o1,16,%o7 + faddd $alod,$nlod,$nlod + fmuld $nhi,$nb,$nhib + add %o7,%o2,%o2 + fmuld $ahi,$bc,$ahic + srlx %o2,16,%o7 + faddd $ahia,$nhia,$nhia + fmuld $nhi,$nc,$nhic + add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15] + and %o0,$mask,%o0 + fmuld $ahi,$bd,$ahid + and %o1,$mask,%o1 + and %o2,$mask,%o2 + faddd $ahib,$nhib,$nhib + fmuld $nhi,$nd,$nhid + sllx %o1,16,%o1 + faddd $dota,$nloa,$nloa + sllx %o2,32,%o2 + faddd $dotb,$nlob,$nlob + sllx %o3,48,%o7 + or %o1,%o0,%o0 + faddd $ahic,$nhic,$dota ! $nhic + or %o2,%o0,%o0 + faddd $ahid,$nhid,$dotb ! $nhid + or %o7,%o0,%o0 ! 64-bit result + faddd $nloc,$nhia,$nloc + addcc %g1,%o0,%o0 + ldx [$tp+8],%o7 ! tp[j] + faddd $nlod,$nhib,$nlod + srlx %o3,16,%g1 ! 34-bit carry + fdtox $nloa,$nloa + bcs,a %xcc,.+8 + add %g1,1,%g1 + fdtox $nlob,$nlob + addcc %o7,%o0,%o0 + fdtox $nloc,$nloc + bcs,a %xcc,.+8 + add %g1,1,%g1 + + stx %o0,[$tp] ! tp[j-1] + fdtox $nlod,$nlod + + std $nloa,[%sp+$bias+$frame+0] + std $nlob,[%sp+$bias+$frame+8] + std $nloc,[%sp+$bias+$frame+16] + addcc $j,8,$j + std $nlod,[%sp+$bias+$frame+24] + bnz,pt %icc,.Linner + add $tp,8,$tp + +.Linnerskip: + fdtox $dota,$dota + fdtox $dotb,$dotb + + ldx [%sp+$bias+$frame+0],%o0 + ldx [%sp+$bias+$frame+8],%o1 + ldx [%sp+$bias+$frame+16],%o2 + ldx [%sp+$bias+$frame+24],%o3 + + srlx %o0,16,%o7 + std $dota,[%sp+$bias+$frame+32] + add %o7,%o1,%o1 + std $dotb,[%sp+$bias+$frame+40] + srlx %o1,16,%o7 + add %o7,%o2,%o2 + srlx %o2,16,%o7 + add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15] + and %o0,$mask,%o0 + and %o1,$mask,%o1 + and %o2,$mask,%o2 + sllx %o1,16,%o1 + sllx %o2,32,%o2 + sllx %o3,48,%o7 + or %o1,%o0,%o0 + or %o2,%o0,%o0 + ldx [%sp+$bias+$frame+32],%o4 + or %o7,%o0,%o0 ! 64-bit result + ldx [%sp+$bias+$frame+40],%o5 + addcc %g1,%o0,%o0 + ldx [$tp+8],%o7 ! tp[j] + srlx %o3,16,%g1 ! 34-bit carry + bcs,a %xcc,.+8 + add %g1,1,%g1 + + addcc %o7,%o0,%o0 + bcs,a %xcc,.+8 + add %g1,1,%g1 + + stx %o0,[$tp] ! tp[j-1] + add $tp,8,$tp + + srlx %o4,16,%o7 + add %o7,%o5,%o5 + and %o4,$mask,%o4 + sllx %o5,16,%o7 + or %o7,%o4,%o4 + addcc %g1,%o4,%o4 + srlx %o5,48,%g1 + bcs,a %xcc,.+8 + add %g1,1,%g1 + + addcc $carry,%o4,%o4 + stx %o4,[$tp] ! tp[num-1] + mov %g1,$carry + bcs,a %xcc,.+8 + add $carry,1,$carry + + addcc $i,8,$i + bnz %icc,.Louter + nop + + add $tp,8,$tp ! adjust tp to point at the end + orn %g0,%g0,%g4 + sub %g0,$num,%o7 ! n=-num + ba .Lsub + subcc %g0,%g0,%g0 ! clear %icc.c + +.align 32 +.Lsub: + ldx [$tp+%o7],%o0 + add $np,%o7,%g1 + ld [%g1+0],%o2 + ld [%g1+4],%o3 + srlx %o0,32,%o1 + subccc %o0,%o2,%o2 + add $rp,%o7,%g1 + subccc %o1,%o3,%o3 + st %o2,[%g1+0] + add %o7,8,%o7 + brnz,pt %o7,.Lsub + st %o3,[%g1+4] + subc $carry,0,%g4 + sub %g0,$num,%o7 ! n=-num + ba .Lcopy + nop + +.align 32 +.Lcopy: + ldx [$tp+%o7],%o0 + add $rp,%o7,%g1 + ld [%g1+0],%o2 + ld [%g1+4],%o3 + stx %g0,[$tp+%o7] + and %o0,%g4,%o0 + srlx %o0,32,%o1 + andn %o2,%g4,%o2 + andn %o3,%g4,%o3 + or %o2,%o0,%o0 + or %o3,%o1,%o1 + st %o0,[%g1+0] + add %o7,8,%o7 + brnz,pt %o7,.Lcopy + st %o1,[%g1+4] + sub %g0,$num,%o7 ! n=-num + +.Lzap: + stx %g0,[$ap_l+%o7] + stx %g0,[$ap_h+%o7] + stx %g0,[$np_l+%o7] + stx %g0,[$np_h+%o7] + add %o7,8,%o7 + brnz,pt %o7,.Lzap + nop + + ldx [%sp+$bias+$frame+48],%o7 + wr %g0,%o7,%asi ! restore %asi + + mov 1,%i0 +.Lret: + ret + restore +.type $fname,#function +.size $fname,(.-$fname) +.asciz "Montgomery Multipltication for UltraSPARC, CRYPTOGAMS by " +.align 32 +___ + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; + +# Below substitution makes it possible to compile without demanding +# VIS extentions on command line, e.g. -xarch=v9 vs. -xarch=v9a. I +# dare to do this, because VIS capability is detected at run-time now +# and this routine is not called on CPU not capable to execute it. Do +# note that fzeros is not the only VIS dependency! Another dependency +# is implicit and is just _a_ numerical value loaded to %asi register, +# which assembler can't recognize as VIS specific... +$code =~ s/fzeros\s+%f([0-9]+)/ + sprintf(".word\t0x%x\t! fzeros %%f%d",0x81b00c20|($1<<25),$1) + /gem; + +print $code; +# flush +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/via-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/via-mont.pl new file mode 100644 index 00000000..c046a514 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/via-mont.pl @@ -0,0 +1,242 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# Wrapper around 'rep montmul', VIA-specific instruction accessing +# PadLock Montgomery Multiplier. The wrapper is designed as drop-in +# replacement for OpenSSL bn_mul_mont [first implemented in 0.9.9]. +# +# Below are interleaved outputs from 'openssl speed rsa dsa' for 4 +# different software configurations on 1.5GHz VIA Esther processor. +# Lines marked with "software integer" denote performance of hand- +# coded integer-only assembler found in OpenSSL 0.9.7. "Software SSE2" +# refers to hand-coded SSE2 Montgomery multiplication procedure found +# OpenSSL 0.9.9. "Hardware VIA SDK" refers to padlock_pmm routine from +# Padlock SDK 2.0.1 available for download from VIA, which naturally +# utilizes the magic 'repz montmul' instruction. And finally "hardware +# this" refers to *this* implementation which also uses 'repz montmul' +# +# sign verify sign/s verify/s +# rsa 512 bits 0.001720s 0.000140s 581.4 7149.7 software integer +# rsa 512 bits 0.000690s 0.000086s 1450.3 11606.0 software SSE2 +# rsa 512 bits 0.006136s 0.000201s 163.0 4974.5 hardware VIA SDK +# rsa 512 bits 0.000712s 0.000050s 1404.9 19858.5 hardware this +# +# rsa 1024 bits 0.008518s 0.000413s 117.4 2420.8 software integer +# rsa 1024 bits 0.004275s 0.000277s 233.9 3609.7 software SSE2 +# rsa 1024 bits 0.012136s 0.000260s 82.4 3844.5 hardware VIA SDK +# rsa 1024 bits 0.002522s 0.000116s 396.5 8650.9 hardware this +# +# rsa 2048 bits 0.050101s 0.001371s 20.0 729.6 software integer +# rsa 2048 bits 0.030273s 0.001008s 33.0 991.9 software SSE2 +# rsa 2048 bits 0.030833s 0.000976s 32.4 1025.1 hardware VIA SDK +# rsa 2048 bits 0.011879s 0.000342s 84.2 2921.7 hardware this +# +# rsa 4096 bits 0.327097s 0.004859s 3.1 205.8 software integer +# rsa 4096 bits 0.229318s 0.003859s 4.4 259.2 software SSE2 +# rsa 4096 bits 0.233953s 0.003274s 4.3 305.4 hardware VIA SDK +# rsa 4096 bits 0.070493s 0.001166s 14.2 857.6 hardware this +# +# dsa 512 bits 0.001342s 0.001651s 745.2 605.7 software integer +# dsa 512 bits 0.000844s 0.000987s 1185.3 1013.1 software SSE2 +# dsa 512 bits 0.001902s 0.002247s 525.6 444.9 hardware VIA SDK +# dsa 512 bits 0.000458s 0.000524s 2182.2 1909.1 hardware this +# +# dsa 1024 bits 0.003964s 0.004926s 252.3 203.0 software integer +# dsa 1024 bits 0.002686s 0.003166s 372.3 315.8 software SSE2 +# dsa 1024 bits 0.002397s 0.002823s 417.1 354.3 hardware VIA SDK +# dsa 1024 bits 0.000978s 0.001170s 1022.2 855.0 hardware this +# +# dsa 2048 bits 0.013280s 0.016518s 75.3 60.5 software integer +# dsa 2048 bits 0.009911s 0.011522s 100.9 86.8 software SSE2 +# dsa 2048 bits 0.009542s 0.011763s 104.8 85.0 hardware VIA SDK +# dsa 2048 bits 0.002884s 0.003352s 346.8 298.3 hardware this +# +# To give you some other reference point here is output for 2.4GHz P4 +# running hand-coded SSE2 bn_mul_mont found in 0.9.9, i.e. "software +# SSE2" in above terms. +# +# rsa 512 bits 0.000407s 0.000047s 2454.2 21137.0 +# rsa 1024 bits 0.002426s 0.000141s 412.1 7100.0 +# rsa 2048 bits 0.015046s 0.000491s 66.5 2034.9 +# rsa 4096 bits 0.109770s 0.002379s 9.1 420.3 +# dsa 512 bits 0.000438s 0.000525s 2281.1 1904.1 +# dsa 1024 bits 0.001346s 0.001595s 742.7 627.0 +# dsa 2048 bits 0.004745s 0.005582s 210.7 179.1 +# +# Conclusions: +# - VIA SDK leaves a *lot* of room for improvement (which this +# implementation successfully fills:-); +# - 'rep montmul' gives up to >3x performance improvement depending on +# key length; +# - in terms of absolute performance it delivers approximately as much +# as modern out-of-order 32-bit cores [again, for longer keys]. + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "x86asm.pl"; + +&asm_init($ARGV[0],"via-mont.pl"); + +# int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,const BN_ULONG *n0, int num); +$func="bn_mul_mont_padlock"; + +$pad=16*1; # amount of reserved bytes on top of every vector + +# stack layout +$mZeroPrime=&DWP(0,"esp"); # these are specified by VIA +$A=&DWP(4,"esp"); +$B=&DWP(8,"esp"); +$T=&DWP(12,"esp"); +$M=&DWP(16,"esp"); +$scratch=&DWP(20,"esp"); +$rp=&DWP(24,"esp"); # these are mine +$sp=&DWP(28,"esp"); +# &DWP(32,"esp") # 32 byte scratch area +# &DWP(64+(4*$num+$pad)*0,"esp") # padded tp[num] +# &DWP(64+(4*$num+$pad)*1,"esp") # padded copy of ap[num] +# &DWP(64+(4*$num+$pad)*2,"esp") # padded copy of bp[num] +# &DWP(64+(4*$num+$pad)*3,"esp") # padded copy of np[num] +# Note that SDK suggests to unconditionally allocate 2K per vector. This +# has quite an impact on performance. It naturally depends on key length, +# but to give an example 1024 bit private RSA key operations suffer >30% +# penalty. I allocate only as much as actually required... + +&function_begin($func); + &xor ("eax","eax"); + &mov ("ecx",&wparam(5)); # num + # meet VIA's limitations for num [note that the specification + # expresses them in bits, while we work with amount of 32-bit words] + &test ("ecx",3); + &jnz (&label("leave")); # num % 4 != 0 + &cmp ("ecx",8); + &jb (&label("leave")); # num < 8 + &cmp ("ecx",1024); + &ja (&label("leave")); # num > 1024 + + &pushf (); + &cld (); + + &mov ("edi",&wparam(0)); # rp + &mov ("eax",&wparam(1)); # ap + &mov ("ebx",&wparam(2)); # bp + &mov ("edx",&wparam(3)); # np + &mov ("esi",&wparam(4)); # n0 + &mov ("esi",&DWP(0,"esi")); # *n0 + + &lea ("ecx",&DWP($pad,"","ecx",4)); # ecx becomes vector size in bytes + &lea ("ebp",&DWP(64,"","ecx",4)); # allocate 4 vectors + 64 bytes + &neg ("ebp"); + &add ("ebp","esp"); + &and ("ebp",-64); # align to cache-line + &xchg ("ebp","esp"); # alloca + + &mov ($rp,"edi"); # save rp + &mov ($sp,"ebp"); # save esp + + &mov ($mZeroPrime,"esi"); + &lea ("esi",&DWP(64,"esp")); # tp + &mov ($T,"esi"); + &lea ("edi",&DWP(32,"esp")); # scratch area + &mov ($scratch,"edi"); + &mov ("esi","eax"); + + &lea ("ebp",&DWP(-$pad,"ecx")); + &shr ("ebp",2); # restore original num value in ebp + + &xor ("eax","eax"); + + &mov ("ecx","ebp"); + &lea ("ecx",&DWP((32+$pad)/4,"ecx"));# padded tp + scratch + &data_byte(0xf3,0xab); # rep stosl, bzero + + &mov ("ecx","ebp"); + &lea ("edi",&DWP(64+$pad,"esp","ecx",4));# pointer to ap copy + &mov ($A,"edi"); + &data_byte(0xf3,0xa5); # rep movsl, memcpy + &mov ("ecx",$pad/4); + &data_byte(0xf3,0xab); # rep stosl, bzero pad + # edi points at the end of padded ap copy... + + &mov ("ecx","ebp"); + &mov ("esi","ebx"); + &mov ($B,"edi"); + &data_byte(0xf3,0xa5); # rep movsl, memcpy + &mov ("ecx",$pad/4); + &data_byte(0xf3,0xab); # rep stosl, bzero pad + # edi points at the end of padded bp copy... + + &mov ("ecx","ebp"); + &mov ("esi","edx"); + &mov ($M,"edi"); + &data_byte(0xf3,0xa5); # rep movsl, memcpy + &mov ("ecx",$pad/4); + &data_byte(0xf3,0xab); # rep stosl, bzero pad + # edi points at the end of padded np copy... + + # let magic happen... + &mov ("ecx","ebp"); + &mov ("esi","esp"); + &shl ("ecx",5); # convert word counter to bit counter + &align (4); + &data_byte(0xf3,0x0f,0xa6,0xc0);# rep montmul + + &mov ("ecx","ebp"); + &lea ("esi",&DWP(64,"esp")); # tp + # edi still points at the end of padded np copy... + &neg ("ebp"); + &lea ("ebp",&DWP(-$pad,"edi","ebp",4)); # so just "rewind" + &mov ("edi",$rp); # restore rp + &xor ("edx","edx"); # i=0 and clear CF + +&set_label("sub",8); + &mov ("eax",&DWP(0,"esi","edx",4)); + &sbb ("eax",&DWP(0,"ebp","edx",4)); + &mov (&DWP(0,"edi","edx",4),"eax"); # rp[i]=tp[i]-np[i] + &lea ("edx",&DWP(1,"edx")); # i++ + &loop (&label("sub")); # doesn't affect CF! + + &mov ("eax",&DWP(0,"esi","edx",4)); # upmost overflow bit + &sbb ("eax",0); + &and ("esi","eax"); + ¬ ("eax"); + &mov ("ebp","edi"); + &and ("ebp","eax"); + &or ("esi","ebp"); # tp=carry?tp:rp + + &mov ("ecx","edx"); # num + &xor ("edx","edx"); # i=0 + +&set_label("copy",8); + &mov ("eax",&DWP(0,"esi","edx",4)); + &mov (&DWP(64,"esp","edx",4),"ecx"); # zap tp + &mov (&DWP(0,"edi","edx",4),"eax"); + &lea ("edx",&DWP(1,"edx")); # i++ + &loop (&label("copy")); + + &mov ("ebp",$sp); + &xor ("eax","eax"); + + &mov ("ecx",64/4); + &mov ("edi","esp"); # zap frame including scratch area + &data_byte(0xf3,0xab); # rep stosl, bzero + + # zap copies of ap, bp and np + &lea ("edi",&DWP(64+$pad,"esp","edx",4));# pointer to ap + &lea ("ecx",&DWP(3*$pad/4,"edx","edx",2)); + &data_byte(0xf3,0xab); # rep stosl, bzero + + &mov ("esp","ebp"); + &inc ("eax"); # signal "done" + &popf (); +&set_label("leave"); +&function_end($func); + +&asciz("Padlock Montgomery Multiplication, CRYPTOGAMS by "); + +&asm_finish(); diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-gf2m.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-gf2m.S new file mode 100644 index 00000000..9ed29ae0 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-gf2m.S @@ -0,0 +1,347 @@ +.file "crypto/bn/asm/x86-gf2m.s" +.text +.type _mul_1x1_mmx,@function +.align 16 +_mul_1x1_mmx: + subl $36,%esp + movl %eax,%ecx + leal (%eax,%eax,1),%edx + andl $1073741823,%ecx + leal (%edx,%edx,1),%ebp + movl $0,(%esp) + andl $2147483647,%edx + movd %eax,%mm2 + movd %ebx,%mm3 + movl %ecx,4(%esp) + xorl %edx,%ecx + pxor %mm5,%mm5 + pxor %mm4,%mm4 + movl %edx,8(%esp) + xorl %ebp,%edx + movl %ecx,12(%esp) + pcmpgtd %mm2,%mm5 + paddd %mm2,%mm2 + xorl %edx,%ecx + movl %ebp,16(%esp) + xorl %edx,%ebp + pand %mm3,%mm5 + pcmpgtd %mm2,%mm4 + movl %ecx,20(%esp) + xorl %ecx,%ebp + psllq $31,%mm5 + pand %mm3,%mm4 + movl %edx,24(%esp) + movl $7,%esi + movl %ebp,28(%esp) + movl %esi,%ebp + andl %ebx,%esi + shrl $3,%ebx + movl %ebp,%edi + psllq $30,%mm4 + andl %ebx,%edi + shrl $3,%ebx + movd (%esp,%esi,4),%mm0 + movl %ebp,%esi + andl %ebx,%esi + shrl $3,%ebx + movd (%esp,%edi,4),%mm2 + movl %ebp,%edi + psllq $3,%mm2 + andl %ebx,%edi + shrl $3,%ebx + pxor %mm2,%mm0 + movd (%esp,%esi,4),%mm1 + movl %ebp,%esi + psllq $6,%mm1 + andl %ebx,%esi + shrl $3,%ebx + pxor %mm1,%mm0 + movd (%esp,%edi,4),%mm2 + movl %ebp,%edi + psllq $9,%mm2 + andl %ebx,%edi + shrl $3,%ebx + pxor %mm2,%mm0 + movd (%esp,%esi,4),%mm1 + movl %ebp,%esi + psllq $12,%mm1 + andl %ebx,%esi + shrl $3,%ebx + pxor %mm1,%mm0 + movd (%esp,%edi,4),%mm2 + movl %ebp,%edi + psllq $15,%mm2 + andl %ebx,%edi + shrl $3,%ebx + pxor %mm2,%mm0 + movd (%esp,%esi,4),%mm1 + movl %ebp,%esi + psllq $18,%mm1 + andl %ebx,%esi + shrl $3,%ebx + pxor %mm1,%mm0 + movd (%esp,%edi,4),%mm2 + movl %ebp,%edi + psllq $21,%mm2 + andl %ebx,%edi + shrl $3,%ebx + pxor %mm2,%mm0 + movd (%esp,%esi,4),%mm1 + movl %ebp,%esi + psllq $24,%mm1 + andl %ebx,%esi + shrl $3,%ebx + pxor %mm1,%mm0 + movd (%esp,%edi,4),%mm2 + pxor %mm4,%mm0 + psllq $27,%mm2 + pxor %mm2,%mm0 + movd (%esp,%esi,4),%mm1 + pxor %mm5,%mm0 + psllq $30,%mm1 + addl $36,%esp + pxor %mm1,%mm0 + ret +.size _mul_1x1_mmx,.-_mul_1x1_mmx +.type _mul_1x1_ialu,@function +.align 16 +_mul_1x1_ialu: + subl $36,%esp + movl %eax,%ecx + leal (%eax,%eax,1),%edx + leal (,%eax,4),%ebp + andl $1073741823,%ecx + leal (%eax,%eax,1),%edi + sarl $31,%eax + movl $0,(%esp) + andl $2147483647,%edx + movl %ecx,4(%esp) + xorl %edx,%ecx + movl %edx,8(%esp) + xorl %ebp,%edx + movl %ecx,12(%esp) + xorl %edx,%ecx + movl %ebp,16(%esp) + xorl %edx,%ebp + movl %ecx,20(%esp) + xorl %ecx,%ebp + sarl $31,%edi + andl %ebx,%eax + movl %edx,24(%esp) + andl %ebx,%edi + movl %ebp,28(%esp) + movl %eax,%edx + shll $31,%eax + movl %edi,%ecx + shrl $1,%edx + movl $7,%esi + shll $30,%edi + andl %ebx,%esi + shrl $2,%ecx + xorl %edi,%eax + shrl $3,%ebx + movl $7,%edi + andl %ebx,%edi + shrl $3,%ebx + xorl %ecx,%edx + xorl (%esp,%esi,4),%eax + movl $7,%esi + andl %ebx,%esi + shrl $3,%ebx + movl (%esp,%edi,4),%ebp + movl $7,%edi + movl %ebp,%ecx + shll $3,%ebp + andl %ebx,%edi + shrl $29,%ecx + xorl %ebp,%eax + shrl $3,%ebx + xorl %ecx,%edx + movl (%esp,%esi,4),%ecx + movl $7,%esi + movl %ecx,%ebp + shll $6,%ecx + andl %ebx,%esi + shrl $26,%ebp + xorl %ecx,%eax + shrl $3,%ebx + xorl %ebp,%edx + movl (%esp,%edi,4),%ebp + movl $7,%edi + movl %ebp,%ecx + shll $9,%ebp + andl %ebx,%edi + shrl $23,%ecx + xorl %ebp,%eax + shrl $3,%ebx + xorl %ecx,%edx + movl (%esp,%esi,4),%ecx + movl $7,%esi + movl %ecx,%ebp + shll $12,%ecx + andl %ebx,%esi + shrl $20,%ebp + xorl %ecx,%eax + shrl $3,%ebx + xorl %ebp,%edx + movl (%esp,%edi,4),%ebp + movl $7,%edi + movl %ebp,%ecx + shll $15,%ebp + andl %ebx,%edi + shrl $17,%ecx + xorl %ebp,%eax + shrl $3,%ebx + xorl %ecx,%edx + movl (%esp,%esi,4),%ecx + movl $7,%esi + movl %ecx,%ebp + shll $18,%ecx + andl %ebx,%esi + shrl $14,%ebp + xorl %ecx,%eax + shrl $3,%ebx + xorl %ebp,%edx + movl (%esp,%edi,4),%ebp + movl $7,%edi + movl %ebp,%ecx + shll $21,%ebp + andl %ebx,%edi + shrl $11,%ecx + xorl %ebp,%eax + shrl $3,%ebx + xorl %ecx,%edx + movl (%esp,%esi,4),%ecx + movl $7,%esi + movl %ecx,%ebp + shll $24,%ecx + andl %ebx,%esi + shrl $8,%ebp + xorl %ecx,%eax + shrl $3,%ebx + xorl %ebp,%edx + movl (%esp,%edi,4),%ebp + movl %ebp,%ecx + shll $27,%ebp + movl (%esp,%esi,4),%edi + shrl $5,%ecx + movl %edi,%esi + xorl %ebp,%eax + shll $30,%edi + xorl %ecx,%edx + shrl $2,%esi + xorl %edi,%eax + xorl %esi,%edx + addl $36,%esp + ret +.size _mul_1x1_ialu,.-_mul_1x1_ialu +.globl bn_GF2m_mul_2x2 +.type bn_GF2m_mul_2x2,@function +.align 16 +bn_GF2m_mul_2x2: +.L_bn_GF2m_mul_2x2_begin: + call .L000PIC_me_up +.L000PIC_me_up: + popl %edx + leal _GLOBAL_OFFSET_TABLE_+[.-.L000PIC_me_up](%edx),%edx + movl OPENSSL_ia32cap_P@GOT(%edx),%edx + movl (%edx),%eax + movl 4(%edx),%edx + testl $8388608,%eax + jz .L001ialu + testl $16777216,%eax + jz .L002mmx + testl $2,%edx + jz .L002mmx + movups 8(%esp),%xmm0 + shufps $177,%xmm0,%xmm0 +.byte 102,15,58,68,192,1 + movl 4(%esp),%eax + movups %xmm0,(%eax) + ret +.align 16 +.L002mmx: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + movl 24(%esp),%eax + movl 32(%esp),%ebx + call _mul_1x1_mmx + movq %mm0,%mm7 + movl 28(%esp),%eax + movl 36(%esp),%ebx + call _mul_1x1_mmx + movq %mm0,%mm6 + movl 24(%esp),%eax + movl 32(%esp),%ebx + xorl 28(%esp),%eax + xorl 36(%esp),%ebx + call _mul_1x1_mmx + pxor %mm7,%mm0 + movl 20(%esp),%eax + pxor %mm6,%mm0 + movq %mm0,%mm2 + psllq $32,%mm0 + popl %edi + psrlq $32,%mm2 + popl %esi + pxor %mm6,%mm0 + popl %ebx + pxor %mm7,%mm2 + movq %mm0,(%eax) + popl %ebp + movq %mm2,8(%eax) + emms + ret +.align 16 +.L001ialu: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + subl $20,%esp + movl 44(%esp),%eax + movl 52(%esp),%ebx + call _mul_1x1_ialu + movl %eax,8(%esp) + movl %edx,12(%esp) + movl 48(%esp),%eax + movl 56(%esp),%ebx + call _mul_1x1_ialu + movl %eax,(%esp) + movl %edx,4(%esp) + movl 44(%esp),%eax + movl 52(%esp),%ebx + xorl 48(%esp),%eax + xorl 56(%esp),%ebx + call _mul_1x1_ialu + movl 40(%esp),%ebp + movl (%esp),%ebx + movl 4(%esp),%ecx + movl 8(%esp),%edi + movl 12(%esp),%esi + xorl %edx,%eax + xorl %ecx,%edx + xorl %ebx,%eax + movl %ebx,(%ebp) + xorl %edi,%edx + movl %esi,12(%ebp) + xorl %esi,%eax + addl $20,%esp + xorl %esi,%edx + popl %edi + xorl %edx,%eax + popl %esi + movl %edx,8(%ebp) + popl %ebx + movl %eax,4(%ebp) + popl %ebp + ret +.size bn_GF2m_mul_2x2,.-.L_bn_GF2m_mul_2x2_begin +.byte 71,70,40,50,94,109,41,32,77,117,108,116,105,112,108,105 +.byte 99,97,116,105,111,110,32,102,111,114,32,120,56,54,44,32 +.byte 67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97 +.byte 112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103 +.byte 62,0 +.comm OPENSSL_ia32cap_P,8,4 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-gf2m.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-gf2m.pl new file mode 100644 index 00000000..b5795302 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-gf2m.pl @@ -0,0 +1,313 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# May 2011 +# +# The module implements bn_GF2m_mul_2x2 polynomial multiplication used +# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for +# the time being... Except that it has three code paths: pure integer +# code suitable for any x86 CPU, MMX code suitable for PIII and later +# and PCLMULQDQ suitable for Westmere and later. Improvement varies +# from one benchmark and µ-arch to another. Below are interval values +# for 163- and 571-bit ECDH benchmarks relative to compiler-generated +# code: +# +# PIII 16%-30% +# P4 12%-12% +# Opteron 18%-40% +# Core2 19%-44% +# Atom 38%-64% +# Westmere 53%-121%(PCLMULQDQ)/20%-32%(MMX) +# Sandy Bridge 72%-127%(PCLMULQDQ)/27%-23%(MMX) +# +# Note that above improvement coefficients are not coefficients for +# bn_GF2m_mul_2x2 itself. For example 120% ECDH improvement is result +# of bn_GF2m_mul_2x2 being >4x faster. As it gets faster, benchmark +# is more and more dominated by other subroutines, most notably by +# BN_GF2m_mod[_mul]_arr... + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "x86asm.pl"; + +&asm_init($ARGV[0],$0,$x86only = $ARGV[$#ARGV] eq "386"); + +$sse2=0; +for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } + +&external_label("OPENSSL_ia32cap_P") if ($sse2); + +$a="eax"; +$b="ebx"; +($a1,$a2,$a4)=("ecx","edx","ebp"); + +$R="mm0"; +@T=("mm1","mm2"); +($A,$B,$B30,$B31)=("mm2","mm3","mm4","mm5"); +@i=("esi","edi"); + + if (!$x86only) { +&function_begin_B("_mul_1x1_mmx"); + &sub ("esp",32+4); + &mov ($a1,$a); + &lea ($a2,&DWP(0,$a,$a)); + &and ($a1,0x3fffffff); + &lea ($a4,&DWP(0,$a2,$a2)); + &mov (&DWP(0*4,"esp"),0); + &and ($a2,0x7fffffff); + &movd ($A,$a); + &movd ($B,$b); + &mov (&DWP(1*4,"esp"),$a1); # a1 + &xor ($a1,$a2); # a1^a2 + &pxor ($B31,$B31); + &pxor ($B30,$B30); + &mov (&DWP(2*4,"esp"),$a2); # a2 + &xor ($a2,$a4); # a2^a4 + &mov (&DWP(3*4,"esp"),$a1); # a1^a2 + &pcmpgtd($B31,$A); # broadcast 31st bit + &paddd ($A,$A); # $A<<=1 + &xor ($a1,$a2); # a1^a4=a1^a2^a2^a4 + &mov (&DWP(4*4,"esp"),$a4); # a4 + &xor ($a4,$a2); # a2=a4^a2^a4 + &pand ($B31,$B); + &pcmpgtd($B30,$A); # broadcast 30th bit + &mov (&DWP(5*4,"esp"),$a1); # a1^a4 + &xor ($a4,$a1); # a1^a2^a4 + &psllq ($B31,31); + &pand ($B30,$B); + &mov (&DWP(6*4,"esp"),$a2); # a2^a4 + &mov (@i[0],0x7); + &mov (&DWP(7*4,"esp"),$a4); # a1^a2^a4 + &mov ($a4,@i[0]); + &and (@i[0],$b); + &shr ($b,3); + &mov (@i[1],$a4); + &psllq ($B30,30); + &and (@i[1],$b); + &shr ($b,3); + &movd ($R,&DWP(0,"esp",@i[0],4)); + &mov (@i[0],$a4); + &and (@i[0],$b); + &shr ($b,3); + for($n=1;$n<9;$n++) { + &movd (@T[1],&DWP(0,"esp",@i[1],4)); + &mov (@i[1],$a4); + &psllq (@T[1],3*$n); + &and (@i[1],$b); + &shr ($b,3); + &pxor ($R,@T[1]); + + push(@i,shift(@i)); push(@T,shift(@T)); + } + &movd (@T[1],&DWP(0,"esp",@i[1],4)); + &pxor ($R,$B30); + &psllq (@T[1],3*$n++); + &pxor ($R,@T[1]); + + &movd (@T[0],&DWP(0,"esp",@i[0],4)); + &pxor ($R,$B31); + &psllq (@T[0],3*$n); + &add ("esp",32+4); + &pxor ($R,@T[0]); + &ret (); +&function_end_B("_mul_1x1_mmx"); + } + +($lo,$hi)=("eax","edx"); +@T=("ecx","ebp"); + +&function_begin_B("_mul_1x1_ialu"); + &sub ("esp",32+4); + &mov ($a1,$a); + &lea ($a2,&DWP(0,$a,$a)); + &lea ($a4,&DWP(0,"",$a,4)); + &and ($a1,0x3fffffff); + &lea (@i[1],&DWP(0,$lo,$lo)); + &sar ($lo,31); # broadcast 31st bit + &mov (&DWP(0*4,"esp"),0); + &and ($a2,0x7fffffff); + &mov (&DWP(1*4,"esp"),$a1); # a1 + &xor ($a1,$a2); # a1^a2 + &mov (&DWP(2*4,"esp"),$a2); # a2 + &xor ($a2,$a4); # a2^a4 + &mov (&DWP(3*4,"esp"),$a1); # a1^a2 + &xor ($a1,$a2); # a1^a4=a1^a2^a2^a4 + &mov (&DWP(4*4,"esp"),$a4); # a4 + &xor ($a4,$a2); # a2=a4^a2^a4 + &mov (&DWP(5*4,"esp"),$a1); # a1^a4 + &xor ($a4,$a1); # a1^a2^a4 + &sar (@i[1],31); # broardcast 30th bit + &and ($lo,$b); + &mov (&DWP(6*4,"esp"),$a2); # a2^a4 + &and (@i[1],$b); + &mov (&DWP(7*4,"esp"),$a4); # a1^a2^a4 + &mov ($hi,$lo); + &shl ($lo,31); + &mov (@T[0],@i[1]); + &shr ($hi,1); + + &mov (@i[0],0x7); + &shl (@i[1],30); + &and (@i[0],$b); + &shr (@T[0],2); + &xor ($lo,@i[1]); + + &shr ($b,3); + &mov (@i[1],0x7); # 5-byte instruction!? + &and (@i[1],$b); + &shr ($b,3); + &xor ($hi,@T[0]); + &xor ($lo,&DWP(0,"esp",@i[0],4)); + &mov (@i[0],0x7); + &and (@i[0],$b); + &shr ($b,3); + for($n=1;$n<9;$n++) { + &mov (@T[1],&DWP(0,"esp",@i[1],4)); + &mov (@i[1],0x7); + &mov (@T[0],@T[1]); + &shl (@T[1],3*$n); + &and (@i[1],$b); + &shr (@T[0],32-3*$n); + &xor ($lo,@T[1]); + &shr ($b,3); + &xor ($hi,@T[0]); + + push(@i,shift(@i)); push(@T,shift(@T)); + } + &mov (@T[1],&DWP(0,"esp",@i[1],4)); + &mov (@T[0],@T[1]); + &shl (@T[1],3*$n); + &mov (@i[1],&DWP(0,"esp",@i[0],4)); + &shr (@T[0],32-3*$n); $n++; + &mov (@i[0],@i[1]); + &xor ($lo,@T[1]); + &shl (@i[1],3*$n); + &xor ($hi,@T[0]); + &shr (@i[0],32-3*$n); + &xor ($lo,@i[1]); + &xor ($hi,@i[0]); + + &add ("esp",32+4); + &ret (); +&function_end_B("_mul_1x1_ialu"); + +# void bn_GF2m_mul_2x2(BN_ULONG *r, BN_ULONG a1, BN_ULONG a0, BN_ULONG b1, BN_ULONG b0); +&function_begin_B("bn_GF2m_mul_2x2"); +if (!$x86only) { + &picmeup("edx","OPENSSL_ia32cap_P"); + &mov ("eax",&DWP(0,"edx")); + &mov ("edx",&DWP(4,"edx")); + &test ("eax",1<<23); # check MMX bit + &jz (&label("ialu")); +if ($sse2) { + &test ("eax",1<<24); # check FXSR bit + &jz (&label("mmx")); + &test ("edx",1<<1); # check PCLMULQDQ bit + &jz (&label("mmx")); + + &movups ("xmm0",&QWP(8,"esp")); + &shufps ("xmm0","xmm0",0b10110001); + &pclmulqdq ("xmm0","xmm0",1); + &mov ("eax",&DWP(4,"esp")); + &movups (&QWP(0,"eax"),"xmm0"); + &ret (); + +&set_label("mmx",16); +} + &push ("ebp"); + &push ("ebx"); + &push ("esi"); + &push ("edi"); + &mov ($a,&wparam(1)); + &mov ($b,&wparam(3)); + &call ("_mul_1x1_mmx"); # a1·b1 + &movq ("mm7",$R); + + &mov ($a,&wparam(2)); + &mov ($b,&wparam(4)); + &call ("_mul_1x1_mmx"); # a0·b0 + &movq ("mm6",$R); + + &mov ($a,&wparam(1)); + &mov ($b,&wparam(3)); + &xor ($a,&wparam(2)); + &xor ($b,&wparam(4)); + &call ("_mul_1x1_mmx"); # (a0+a1)·(b0+b1) + &pxor ($R,"mm7"); + &mov ($a,&wparam(0)); + &pxor ($R,"mm6"); # (a0+a1)·(b0+b1)-a1·b1-a0·b0 + + &movq ($A,$R); + &psllq ($R,32); + &pop ("edi"); + &psrlq ($A,32); + &pop ("esi"); + &pxor ($R,"mm6"); + &pop ("ebx"); + &pxor ($A,"mm7"); + &movq (&QWP(0,$a),$R); + &pop ("ebp"); + &movq (&QWP(8,$a),$A); + &emms (); + &ret (); +&set_label("ialu",16); +} + &push ("ebp"); + &push ("ebx"); + &push ("esi"); + &push ("edi"); + &stack_push(4+1); + + &mov ($a,&wparam(1)); + &mov ($b,&wparam(3)); + &call ("_mul_1x1_ialu"); # a1·b1 + &mov (&DWP(8,"esp"),$lo); + &mov (&DWP(12,"esp"),$hi); + + &mov ($a,&wparam(2)); + &mov ($b,&wparam(4)); + &call ("_mul_1x1_ialu"); # a0·b0 + &mov (&DWP(0,"esp"),$lo); + &mov (&DWP(4,"esp"),$hi); + + &mov ($a,&wparam(1)); + &mov ($b,&wparam(3)); + &xor ($a,&wparam(2)); + &xor ($b,&wparam(4)); + &call ("_mul_1x1_ialu"); # (a0+a1)·(b0+b1) + + &mov ("ebp",&wparam(0)); + @r=("ebx","ecx","edi","esi"); + &mov (@r[0],&DWP(0,"esp")); + &mov (@r[1],&DWP(4,"esp")); + &mov (@r[2],&DWP(8,"esp")); + &mov (@r[3],&DWP(12,"esp")); + + &xor ($lo,$hi); + &xor ($hi,@r[1]); + &xor ($lo,@r[0]); + &mov (&DWP(0,"ebp"),@r[0]); + &xor ($hi,@r[2]); + &mov (&DWP(12,"ebp"),@r[3]); + &xor ($lo,@r[3]); + &stack_pop(4+1); + &xor ($hi,@r[3]); + &pop ("edi"); + &xor ($lo,$hi); + &pop ("esi"); + &mov (&DWP(8,"ebp"),$hi); + &pop ("ebx"); + &mov (&DWP(4,"ebp"),$lo); + &pop ("ebp"); + &ret (); +&function_end_B("bn_GF2m_mul_2x2"); + +&asciz ("GF(2^m) Multiplication for x86, CRYPTOGAMS by "); + +&asm_finish(); diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-mont.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-mont.S new file mode 100644 index 00000000..c701e9e3 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-mont.S @@ -0,0 +1,460 @@ +.file "crypto/bn/asm/x86-mont.s" +.text +.globl bn_mul_mont +.type bn_mul_mont,@function +.align 16 +bn_mul_mont: +.L_bn_mul_mont_begin: + pushl %ebp + pushl %ebx + pushl %esi + pushl %edi + xorl %eax,%eax + movl 40(%esp),%edi + cmpl $4,%edi + jl .L000just_leave + leal 20(%esp),%esi + leal 24(%esp),%edx + movl %esp,%ebp + addl $2,%edi + negl %edi + leal -32(%esp,%edi,4),%esp + negl %edi + movl %esp,%eax + subl %edx,%eax + andl $2047,%eax + subl %eax,%esp + xorl %esp,%edx + andl $2048,%edx + xorl $2048,%edx + subl %edx,%esp + andl $-64,%esp + movl (%esi),%eax + movl 4(%esi),%ebx + movl 8(%esi),%ecx + movl 12(%esi),%edx + movl 16(%esi),%esi + movl (%esi),%esi + movl %eax,4(%esp) + movl %ebx,8(%esp) + movl %ecx,12(%esp) + movl %edx,16(%esp) + movl %esi,20(%esp) + leal -3(%edi),%ebx + movl %ebp,24(%esp) + call .L001PIC_me_up +.L001PIC_me_up: + popl %eax + leal _GLOBAL_OFFSET_TABLE_+[.-.L001PIC_me_up](%eax),%eax + movl OPENSSL_ia32cap_P@GOT(%eax),%eax + btl $26,(%eax) + jnc .L002non_sse2 + movl $-1,%eax + movd %eax,%mm7 + movl 8(%esp),%esi + movl 12(%esp),%edi + movl 16(%esp),%ebp + xorl %edx,%edx + xorl %ecx,%ecx + movd (%edi),%mm4 + movd (%esi),%mm5 + movd (%ebp),%mm3 + pmuludq %mm4,%mm5 + movq %mm5,%mm2 + movq %mm5,%mm0 + pand %mm7,%mm0 + pmuludq 20(%esp),%mm5 + pmuludq %mm5,%mm3 + paddq %mm0,%mm3 + movd 4(%ebp),%mm1 + movd 4(%esi),%mm0 + psrlq $32,%mm2 + psrlq $32,%mm3 + incl %ecx +.align 16 +.L0031st: + pmuludq %mm4,%mm0 + pmuludq %mm5,%mm1 + paddq %mm0,%mm2 + paddq %mm1,%mm3 + movq %mm2,%mm0 + pand %mm7,%mm0 + movd 4(%ebp,%ecx,4),%mm1 + paddq %mm0,%mm3 + movd 4(%esi,%ecx,4),%mm0 + psrlq $32,%mm2 + movd %mm3,28(%esp,%ecx,4) + psrlq $32,%mm3 + leal 1(%ecx),%ecx + cmpl %ebx,%ecx + jl .L0031st + pmuludq %mm4,%mm0 + pmuludq %mm5,%mm1 + paddq %mm0,%mm2 + paddq %mm1,%mm3 + movq %mm2,%mm0 + pand %mm7,%mm0 + paddq %mm0,%mm3 + movd %mm3,28(%esp,%ecx,4) + psrlq $32,%mm2 + psrlq $32,%mm3 + paddq %mm2,%mm3 + movq %mm3,32(%esp,%ebx,4) + incl %edx +.L004outer: + xorl %ecx,%ecx + movd (%edi,%edx,4),%mm4 + movd (%esi),%mm5 + movd 32(%esp),%mm6 + movd (%ebp),%mm3 + pmuludq %mm4,%mm5 + paddq %mm6,%mm5 + movq %mm5,%mm0 + movq %mm5,%mm2 + pand %mm7,%mm0 + pmuludq 20(%esp),%mm5 + pmuludq %mm5,%mm3 + paddq %mm0,%mm3 + movd 36(%esp),%mm6 + movd 4(%ebp),%mm1 + movd 4(%esi),%mm0 + psrlq $32,%mm2 + psrlq $32,%mm3 + paddq %mm6,%mm2 + incl %ecx + decl %ebx +.L005inner: + pmuludq %mm4,%mm0 + pmuludq %mm5,%mm1 + paddq %mm0,%mm2 + paddq %mm1,%mm3 + movq %mm2,%mm0 + movd 36(%esp,%ecx,4),%mm6 + pand %mm7,%mm0 + movd 4(%ebp,%ecx,4),%mm1 + paddq %mm0,%mm3 + movd 4(%esi,%ecx,4),%mm0 + psrlq $32,%mm2 + movd %mm3,28(%esp,%ecx,4) + psrlq $32,%mm3 + paddq %mm6,%mm2 + decl %ebx + leal 1(%ecx),%ecx + jnz .L005inner + movl %ecx,%ebx + pmuludq %mm4,%mm0 + pmuludq %mm5,%mm1 + paddq %mm0,%mm2 + paddq %mm1,%mm3 + movq %mm2,%mm0 + pand %mm7,%mm0 + paddq %mm0,%mm3 + movd %mm3,28(%esp,%ecx,4) + psrlq $32,%mm2 + psrlq $32,%mm3 + movd 36(%esp,%ebx,4),%mm6 + paddq %mm2,%mm3 + paddq %mm6,%mm3 + movq %mm3,32(%esp,%ebx,4) + leal 1(%edx),%edx + cmpl %ebx,%edx + jle .L004outer + emms + jmp .L006common_tail +.align 16 +.L002non_sse2: + movl 8(%esp),%esi + leal 1(%ebx),%ebp + movl 12(%esp),%edi + xorl %ecx,%ecx + movl %esi,%edx + andl $1,%ebp + subl %edi,%edx + leal 4(%edi,%ebx,4),%eax + orl %edx,%ebp + movl (%edi),%edi + jz .L007bn_sqr_mont + movl %eax,28(%esp) + movl (%esi),%eax + xorl %edx,%edx +.align 16 +.L008mull: + movl %edx,%ebp + mull %edi + addl %eax,%ebp + leal 1(%ecx),%ecx + adcl $0,%edx + movl (%esi,%ecx,4),%eax + cmpl %ebx,%ecx + movl %ebp,28(%esp,%ecx,4) + jl .L008mull + movl %edx,%ebp + mull %edi + movl 20(%esp),%edi + addl %ebp,%eax + movl 16(%esp),%esi + adcl $0,%edx + imull 32(%esp),%edi + movl %eax,32(%esp,%ebx,4) + xorl %ecx,%ecx + movl %edx,36(%esp,%ebx,4) + movl %ecx,40(%esp,%ebx,4) + movl (%esi),%eax + mull %edi + addl 32(%esp),%eax + movl 4(%esi),%eax + adcl $0,%edx + incl %ecx + jmp .L0092ndmadd +.align 16 +.L0101stmadd: + movl %edx,%ebp + mull %edi + addl 32(%esp,%ecx,4),%ebp + leal 1(%ecx),%ecx + adcl $0,%edx + addl %eax,%ebp + movl (%esi,%ecx,4),%eax + adcl $0,%edx + cmpl %ebx,%ecx + movl %ebp,28(%esp,%ecx,4) + jl .L0101stmadd + movl %edx,%ebp + mull %edi + addl 32(%esp,%ebx,4),%eax + movl 20(%esp),%edi + adcl $0,%edx + movl 16(%esp),%esi + addl %eax,%ebp + adcl $0,%edx + imull 32(%esp),%edi + xorl %ecx,%ecx + addl 36(%esp,%ebx,4),%edx + movl %ebp,32(%esp,%ebx,4) + adcl $0,%ecx + movl (%esi),%eax + movl %edx,36(%esp,%ebx,4) + movl %ecx,40(%esp,%ebx,4) + mull %edi + addl 32(%esp),%eax + movl 4(%esi),%eax + adcl $0,%edx + movl $1,%ecx +.align 16 +.L0092ndmadd: + movl %edx,%ebp + mull %edi + addl 32(%esp,%ecx,4),%ebp + leal 1(%ecx),%ecx + adcl $0,%edx + addl %eax,%ebp + movl (%esi,%ecx,4),%eax + adcl $0,%edx + cmpl %ebx,%ecx + movl %ebp,24(%esp,%ecx,4) + jl .L0092ndmadd + movl %edx,%ebp + mull %edi + addl 32(%esp,%ebx,4),%ebp + adcl $0,%edx + addl %eax,%ebp + adcl $0,%edx + movl %ebp,28(%esp,%ebx,4) + xorl %eax,%eax + movl 12(%esp),%ecx + addl 36(%esp,%ebx,4),%edx + adcl 40(%esp,%ebx,4),%eax + leal 4(%ecx),%ecx + movl %edx,32(%esp,%ebx,4) + cmpl 28(%esp),%ecx + movl %eax,36(%esp,%ebx,4) + je .L006common_tail + movl (%ecx),%edi + movl 8(%esp),%esi + movl %ecx,12(%esp) + xorl %ecx,%ecx + xorl %edx,%edx + movl (%esi),%eax + jmp .L0101stmadd +.align 16 +.L007bn_sqr_mont: + movl %ebx,(%esp) + movl %ecx,12(%esp) + movl %edi,%eax + mull %edi + movl %eax,32(%esp) + movl %edx,%ebx + shrl $1,%edx + andl $1,%ebx + incl %ecx +.align 16 +.L011sqr: + movl (%esi,%ecx,4),%eax + movl %edx,%ebp + mull %edi + addl %ebp,%eax + leal 1(%ecx),%ecx + adcl $0,%edx + leal (%ebx,%eax,2),%ebp + shrl $31,%eax + cmpl (%esp),%ecx + movl %eax,%ebx + movl %ebp,28(%esp,%ecx,4) + jl .L011sqr + movl (%esi,%ecx,4),%eax + movl %edx,%ebp + mull %edi + addl %ebp,%eax + movl 20(%esp),%edi + adcl $0,%edx + movl 16(%esp),%esi + leal (%ebx,%eax,2),%ebp + imull 32(%esp),%edi + shrl $31,%eax + movl %ebp,32(%esp,%ecx,4) + leal (%eax,%edx,2),%ebp + movl (%esi),%eax + shrl $31,%edx + movl %ebp,36(%esp,%ecx,4) + movl %edx,40(%esp,%ecx,4) + mull %edi + addl 32(%esp),%eax + movl %ecx,%ebx + adcl $0,%edx + movl 4(%esi),%eax + movl $1,%ecx +.align 16 +.L0123rdmadd: + movl %edx,%ebp + mull %edi + addl 32(%esp,%ecx,4),%ebp + adcl $0,%edx + addl %eax,%ebp + movl 4(%esi,%ecx,4),%eax + adcl $0,%edx + movl %ebp,28(%esp,%ecx,4) + movl %edx,%ebp + mull %edi + addl 36(%esp,%ecx,4),%ebp + leal 2(%ecx),%ecx + adcl $0,%edx + addl %eax,%ebp + movl (%esi,%ecx,4),%eax + adcl $0,%edx + cmpl %ebx,%ecx + movl %ebp,24(%esp,%ecx,4) + jl .L0123rdmadd + movl %edx,%ebp + mull %edi + addl 32(%esp,%ebx,4),%ebp + adcl $0,%edx + addl %eax,%ebp + adcl $0,%edx + movl %ebp,28(%esp,%ebx,4) + movl 12(%esp),%ecx + xorl %eax,%eax + movl 8(%esp),%esi + addl 36(%esp,%ebx,4),%edx + adcl 40(%esp,%ebx,4),%eax + movl %edx,32(%esp,%ebx,4) + cmpl %ebx,%ecx + movl %eax,36(%esp,%ebx,4) + je .L006common_tail + movl 4(%esi,%ecx,4),%edi + leal 1(%ecx),%ecx + movl %edi,%eax + movl %ecx,12(%esp) + mull %edi + addl 32(%esp,%ecx,4),%eax + adcl $0,%edx + movl %eax,32(%esp,%ecx,4) + xorl %ebp,%ebp + cmpl %ebx,%ecx + leal 1(%ecx),%ecx + je .L013sqrlast + movl %edx,%ebx + shrl $1,%edx + andl $1,%ebx +.align 16 +.L014sqradd: + movl (%esi,%ecx,4),%eax + movl %edx,%ebp + mull %edi + addl %ebp,%eax + leal (%eax,%eax,1),%ebp + adcl $0,%edx + shrl $31,%eax + addl 32(%esp,%ecx,4),%ebp + leal 1(%ecx),%ecx + adcl $0,%eax + addl %ebx,%ebp + adcl $0,%eax + cmpl (%esp),%ecx + movl %ebp,28(%esp,%ecx,4) + movl %eax,%ebx + jle .L014sqradd + movl %edx,%ebp + addl %edx,%edx + shrl $31,%ebp + addl %ebx,%edx + adcl $0,%ebp +.L013sqrlast: + movl 20(%esp),%edi + movl 16(%esp),%esi + imull 32(%esp),%edi + addl 32(%esp,%ecx,4),%edx + movl (%esi),%eax + adcl $0,%ebp + movl %edx,32(%esp,%ecx,4) + movl %ebp,36(%esp,%ecx,4) + mull %edi + addl 32(%esp),%eax + leal -1(%ecx),%ebx + adcl $0,%edx + movl $1,%ecx + movl 4(%esi),%eax + jmp .L0123rdmadd +.align 16 +.L006common_tail: + movl 16(%esp),%ebp + movl 4(%esp),%edi + leal 32(%esp),%esi + movl (%esi),%eax + movl %ebx,%ecx + xorl %edx,%edx +.align 16 +.L015sub: + sbbl (%ebp,%edx,4),%eax + movl %eax,(%edi,%edx,4) + decl %ecx + movl 4(%esi,%edx,4),%eax + leal 1(%edx),%edx + jge .L015sub + sbbl $0,%eax + andl %eax,%esi + notl %eax + movl %edi,%ebp + andl %eax,%ebp + orl %ebp,%esi +.align 16 +.L016copy: + movl (%esi,%ebx,4),%eax + movl %eax,(%edi,%ebx,4) + movl %ecx,32(%esp,%ebx,4) + decl %ebx + jge .L016copy + movl 24(%esp),%esp + movl $1,%eax +.L000just_leave: + popl %edi + popl %esi + popl %ebx + popl %ebp + ret +.size bn_mul_mont,.-.L_bn_mul_mont_begin +.byte 77,111,110,116,103,111,109,101,114,121,32,77,117,108,116,105 +.byte 112,108,105,99,97,116,105,111,110,32,102,111,114,32,120,56 +.byte 54,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121 +.byte 32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46 +.byte 111,114,103,62,0 +.comm OPENSSL_ia32cap_P,8,4 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-mont.pl new file mode 100755 index 00000000..e8f6b050 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86-mont.pl @@ -0,0 +1,593 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# October 2005 +# +# This is a "teaser" code, as it can be improved in several ways... +# First of all non-SSE2 path should be implemented (yes, for now it +# performs Montgomery multiplication/convolution only on SSE2-capable +# CPUs such as P4, others fall down to original code). Then inner loop +# can be unrolled and modulo-scheduled to improve ILP and possibly +# moved to 128-bit XMM register bank (though it would require input +# rearrangement and/or increase bus bandwidth utilization). Dedicated +# squaring procedure should give further performance improvement... +# Yet, for being draft, the code improves rsa512 *sign* benchmark by +# 110%(!), rsa1024 one - by 70% and rsa4096 - by 20%:-) + +# December 2006 +# +# Modulo-scheduling SSE2 loops results in further 15-20% improvement. +# Integer-only code [being equipped with dedicated squaring procedure] +# gives ~40% on rsa512 sign benchmark... + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); +require "x86asm.pl"; + +&asm_init($ARGV[0],$0); + +$sse2=0; +for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } + +&external_label("OPENSSL_ia32cap_P") if ($sse2); + +&function_begin("bn_mul_mont"); + +$i="edx"; +$j="ecx"; +$ap="esi"; $tp="esi"; # overlapping variables!!! +$rp="edi"; $bp="edi"; # overlapping variables!!! +$np="ebp"; +$num="ebx"; + +$_num=&DWP(4*0,"esp"); # stack top layout +$_rp=&DWP(4*1,"esp"); +$_ap=&DWP(4*2,"esp"); +$_bp=&DWP(4*3,"esp"); +$_np=&DWP(4*4,"esp"); +$_n0=&DWP(4*5,"esp"); $_n0q=&QWP(4*5,"esp"); +$_sp=&DWP(4*6,"esp"); +$_bpend=&DWP(4*7,"esp"); +$frame=32; # size of above frame rounded up to 16n + + &xor ("eax","eax"); + &mov ("edi",&wparam(5)); # int num + &cmp ("edi",4); + &jl (&label("just_leave")); + + &lea ("esi",&wparam(0)); # put aside pointer to argument block + &lea ("edx",&wparam(1)); # load ap + &mov ("ebp","esp"); # saved stack pointer! + &add ("edi",2); # extra two words on top of tp + &neg ("edi"); + &lea ("esp",&DWP(-$frame,"esp","edi",4)); # alloca($frame+4*(num+2)) + &neg ("edi"); + + # minimize cache contention by arraning 2K window between stack + # pointer and ap argument [np is also position sensitive vector, + # but it's assumed to be near ap, as it's allocated at ~same + # time]. + &mov ("eax","esp"); + &sub ("eax","edx"); + &and ("eax",2047); + &sub ("esp","eax"); # this aligns sp and ap modulo 2048 + + &xor ("edx","esp"); + &and ("edx",2048); + &xor ("edx",2048); + &sub ("esp","edx"); # this splits them apart modulo 4096 + + &and ("esp",-64); # align to cache line + + ################################# load argument block... + &mov ("eax",&DWP(0*4,"esi"));# BN_ULONG *rp + &mov ("ebx",&DWP(1*4,"esi"));# const BN_ULONG *ap + &mov ("ecx",&DWP(2*4,"esi"));# const BN_ULONG *bp + &mov ("edx",&DWP(3*4,"esi"));# const BN_ULONG *np + &mov ("esi",&DWP(4*4,"esi"));# const BN_ULONG *n0 + #&mov ("edi",&DWP(5*4,"esi"));# int num + + &mov ("esi",&DWP(0,"esi")); # pull n0[0] + &mov ($_rp,"eax"); # ... save a copy of argument block + &mov ($_ap,"ebx"); + &mov ($_bp,"ecx"); + &mov ($_np,"edx"); + &mov ($_n0,"esi"); + &lea ($num,&DWP(-3,"edi")); # num=num-1 to assist modulo-scheduling + #&mov ($_num,$num); # redundant as $num is not reused + &mov ($_sp,"ebp"); # saved stack pointer! + +if($sse2) { +$acc0="mm0"; # mmx register bank layout +$acc1="mm1"; +$car0="mm2"; +$car1="mm3"; +$mul0="mm4"; +$mul1="mm5"; +$temp="mm6"; +$mask="mm7"; + + &picmeup("eax","OPENSSL_ia32cap_P"); + &bt (&DWP(0,"eax"),26); + &jnc (&label("non_sse2")); + + &mov ("eax",-1); + &movd ($mask,"eax"); # mask 32 lower bits + + &mov ($ap,$_ap); # load input pointers + &mov ($bp,$_bp); + &mov ($np,$_np); + + &xor ($i,$i); # i=0 + &xor ($j,$j); # j=0 + + &movd ($mul0,&DWP(0,$bp)); # bp[0] + &movd ($mul1,&DWP(0,$ap)); # ap[0] + &movd ($car1,&DWP(0,$np)); # np[0] + + &pmuludq($mul1,$mul0); # ap[0]*bp[0] + &movq ($car0,$mul1); + &movq ($acc0,$mul1); # I wish movd worked for + &pand ($acc0,$mask); # inter-register transfers + + &pmuludq($mul1,$_n0q); # *=n0 + + &pmuludq($car1,$mul1); # "t[0]"*np[0]*n0 + &paddq ($car1,$acc0); + + &movd ($acc1,&DWP(4,$np)); # np[1] + &movd ($acc0,&DWP(4,$ap)); # ap[1] + + &psrlq ($car0,32); + &psrlq ($car1,32); + + &inc ($j); # j++ +&set_label("1st",16); + &pmuludq($acc0,$mul0); # ap[j]*bp[0] + &pmuludq($acc1,$mul1); # np[j]*m1 + &paddq ($car0,$acc0); # +=c0 + &paddq ($car1,$acc1); # +=c1 + + &movq ($acc0,$car0); + &pand ($acc0,$mask); + &movd ($acc1,&DWP(4,$np,$j,4)); # np[j+1] + &paddq ($car1,$acc0); # +=ap[j]*bp[0]; + &movd ($acc0,&DWP(4,$ap,$j,4)); # ap[j+1] + &psrlq ($car0,32); + &movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[j-1]= + &psrlq ($car1,32); + + &lea ($j,&DWP(1,$j)); + &cmp ($j,$num); + &jl (&label("1st")); + + &pmuludq($acc0,$mul0); # ap[num-1]*bp[0] + &pmuludq($acc1,$mul1); # np[num-1]*m1 + &paddq ($car0,$acc0); # +=c0 + &paddq ($car1,$acc1); # +=c1 + + &movq ($acc0,$car0); + &pand ($acc0,$mask); + &paddq ($car1,$acc0); # +=ap[num-1]*bp[0]; + &movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[num-2]= + + &psrlq ($car0,32); + &psrlq ($car1,32); + + &paddq ($car1,$car0); + &movq (&QWP($frame,"esp",$num,4),$car1); # tp[num].tp[num-1] + + &inc ($i); # i++ +&set_label("outer"); + &xor ($j,$j); # j=0 + + &movd ($mul0,&DWP(0,$bp,$i,4)); # bp[i] + &movd ($mul1,&DWP(0,$ap)); # ap[0] + &movd ($temp,&DWP($frame,"esp")); # tp[0] + &movd ($car1,&DWP(0,$np)); # np[0] + &pmuludq($mul1,$mul0); # ap[0]*bp[i] + + &paddq ($mul1,$temp); # +=tp[0] + &movq ($acc0,$mul1); + &movq ($car0,$mul1); + &pand ($acc0,$mask); + + &pmuludq($mul1,$_n0q); # *=n0 + + &pmuludq($car1,$mul1); + &paddq ($car1,$acc0); + + &movd ($temp,&DWP($frame+4,"esp")); # tp[1] + &movd ($acc1,&DWP(4,$np)); # np[1] + &movd ($acc0,&DWP(4,$ap)); # ap[1] + + &psrlq ($car0,32); + &psrlq ($car1,32); + &paddq ($car0,$temp); # +=tp[1] + + &inc ($j); # j++ + &dec ($num); +&set_label("inner"); + &pmuludq($acc0,$mul0); # ap[j]*bp[i] + &pmuludq($acc1,$mul1); # np[j]*m1 + &paddq ($car0,$acc0); # +=c0 + &paddq ($car1,$acc1); # +=c1 + + &movq ($acc0,$car0); + &movd ($temp,&DWP($frame+4,"esp",$j,4));# tp[j+1] + &pand ($acc0,$mask); + &movd ($acc1,&DWP(4,$np,$j,4)); # np[j+1] + &paddq ($car1,$acc0); # +=ap[j]*bp[i]+tp[j] + &movd ($acc0,&DWP(4,$ap,$j,4)); # ap[j+1] + &psrlq ($car0,32); + &movd (&DWP($frame-4,"esp",$j,4),$car1);# tp[j-1]= + &psrlq ($car1,32); + &paddq ($car0,$temp); # +=tp[j+1] + + &dec ($num); + &lea ($j,&DWP(1,$j)); # j++ + &jnz (&label("inner")); + + &mov ($num,$j); + &pmuludq($acc0,$mul0); # ap[num-1]*bp[i] + &pmuludq($acc1,$mul1); # np[num-1]*m1 + &paddq ($car0,$acc0); # +=c0 + &paddq ($car1,$acc1); # +=c1 + + &movq ($acc0,$car0); + &pand ($acc0,$mask); + &paddq ($car1,$acc0); # +=ap[num-1]*bp[i]+tp[num-1] + &movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[num-2]= + &psrlq ($car0,32); + &psrlq ($car1,32); + + &movd ($temp,&DWP($frame+4,"esp",$num,4)); # += tp[num] + &paddq ($car1,$car0); + &paddq ($car1,$temp); + &movq (&QWP($frame,"esp",$num,4),$car1); # tp[num].tp[num-1] + + &lea ($i,&DWP(1,$i)); # i++ + &cmp ($i,$num); + &jle (&label("outer")); + + &emms (); # done with mmx bank + &jmp (&label("common_tail")); + +&set_label("non_sse2",16); +} + +if (0) { + &mov ("esp",$_sp); + &xor ("eax","eax"); # signal "not fast enough [yet]" + &jmp (&label("just_leave")); + # While the below code provides competitive performance for + # all key lengthes on modern Intel cores, it's still more + # than 10% slower for 4096-bit key elsewhere:-( "Competitive" + # means compared to the original integer-only assembler. + # 512-bit RSA sign is better by ~40%, but that's about all + # one can say about all CPUs... +} else { +$inp="esi"; # integer path uses these registers differently +$word="edi"; +$carry="ebp"; + + &mov ($inp,$_ap); + &lea ($carry,&DWP(1,$num)); + &mov ($word,$_bp); + &xor ($j,$j); # j=0 + &mov ("edx",$inp); + &and ($carry,1); # see if num is even + &sub ("edx",$word); # see if ap==bp + &lea ("eax",&DWP(4,$word,$num,4)); # &bp[num] + &or ($carry,"edx"); + &mov ($word,&DWP(0,$word)); # bp[0] + &jz (&label("bn_sqr_mont")); + &mov ($_bpend,"eax"); + &mov ("eax",&DWP(0,$inp)); + &xor ("edx","edx"); + +&set_label("mull",16); + &mov ($carry,"edx"); + &mul ($word); # ap[j]*bp[0] + &add ($carry,"eax"); + &lea ($j,&DWP(1,$j)); + &adc ("edx",0); + &mov ("eax",&DWP(0,$inp,$j,4)); # ap[j+1] + &cmp ($j,$num); + &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]= + &jl (&label("mull")); + + &mov ($carry,"edx"); + &mul ($word); # ap[num-1]*bp[0] + &mov ($word,$_n0); + &add ("eax",$carry); + &mov ($inp,$_np); + &adc ("edx",0); + &imul ($word,&DWP($frame,"esp")); # n0*tp[0] + + &mov (&DWP($frame,"esp",$num,4),"eax"); # tp[num-1]= + &xor ($j,$j); + &mov (&DWP($frame+4,"esp",$num,4),"edx"); # tp[num]= + &mov (&DWP($frame+8,"esp",$num,4),$j); # tp[num+1]= + + &mov ("eax",&DWP(0,$inp)); # np[0] + &mul ($word); # np[0]*m + &add ("eax",&DWP($frame,"esp")); # +=tp[0] + &mov ("eax",&DWP(4,$inp)); # np[1] + &adc ("edx",0); + &inc ($j); + + &jmp (&label("2ndmadd")); + +&set_label("1stmadd",16); + &mov ($carry,"edx"); + &mul ($word); # ap[j]*bp[i] + &add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j] + &lea ($j,&DWP(1,$j)); + &adc ("edx",0); + &add ($carry,"eax"); + &mov ("eax",&DWP(0,$inp,$j,4)); # ap[j+1] + &adc ("edx",0); + &cmp ($j,$num); + &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]= + &jl (&label("1stmadd")); + + &mov ($carry,"edx"); + &mul ($word); # ap[num-1]*bp[i] + &add ("eax",&DWP($frame,"esp",$num,4)); # +=tp[num-1] + &mov ($word,$_n0); + &adc ("edx",0); + &mov ($inp,$_np); + &add ($carry,"eax"); + &adc ("edx",0); + &imul ($word,&DWP($frame,"esp")); # n0*tp[0] + + &xor ($j,$j); + &add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num] + &mov (&DWP($frame,"esp",$num,4),$carry); # tp[num-1]= + &adc ($j,0); + &mov ("eax",&DWP(0,$inp)); # np[0] + &mov (&DWP($frame+4,"esp",$num,4),"edx"); # tp[num]= + &mov (&DWP($frame+8,"esp",$num,4),$j); # tp[num+1]= + + &mul ($word); # np[0]*m + &add ("eax",&DWP($frame,"esp")); # +=tp[0] + &mov ("eax",&DWP(4,$inp)); # np[1] + &adc ("edx",0); + &mov ($j,1); + +&set_label("2ndmadd",16); + &mov ($carry,"edx"); + &mul ($word); # np[j]*m + &add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j] + &lea ($j,&DWP(1,$j)); + &adc ("edx",0); + &add ($carry,"eax"); + &mov ("eax",&DWP(0,$inp,$j,4)); # np[j+1] + &adc ("edx",0); + &cmp ($j,$num); + &mov (&DWP($frame-8,"esp",$j,4),$carry); # tp[j-1]= + &jl (&label("2ndmadd")); + + &mov ($carry,"edx"); + &mul ($word); # np[j]*m + &add ($carry,&DWP($frame,"esp",$num,4)); # +=tp[num-1] + &adc ("edx",0); + &add ($carry,"eax"); + &adc ("edx",0); + &mov (&DWP($frame-4,"esp",$num,4),$carry); # tp[num-2]= + + &xor ("eax","eax"); + &mov ($j,$_bp); # &bp[i] + &add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num] + &adc ("eax",&DWP($frame+8,"esp",$num,4)); # +=tp[num+1] + &lea ($j,&DWP(4,$j)); + &mov (&DWP($frame,"esp",$num,4),"edx"); # tp[num-1]= + &cmp ($j,$_bpend); + &mov (&DWP($frame+4,"esp",$num,4),"eax"); # tp[num]= + &je (&label("common_tail")); + + &mov ($word,&DWP(0,$j)); # bp[i+1] + &mov ($inp,$_ap); + &mov ($_bp,$j); # &bp[++i] + &xor ($j,$j); + &xor ("edx","edx"); + &mov ("eax",&DWP(0,$inp)); + &jmp (&label("1stmadd")); + +&set_label("bn_sqr_mont",16); +$sbit=$num; + &mov ($_num,$num); + &mov ($_bp,$j); # i=0 + + &mov ("eax",$word); # ap[0] + &mul ($word); # ap[0]*ap[0] + &mov (&DWP($frame,"esp"),"eax"); # tp[0]= + &mov ($sbit,"edx"); + &shr ("edx",1); + &and ($sbit,1); + &inc ($j); +&set_label("sqr",16); + &mov ("eax",&DWP(0,$inp,$j,4)); # ap[j] + &mov ($carry,"edx"); + &mul ($word); # ap[j]*ap[0] + &add ("eax",$carry); + &lea ($j,&DWP(1,$j)); + &adc ("edx",0); + &lea ($carry,&DWP(0,$sbit,"eax",2)); + &shr ("eax",31); + &cmp ($j,$_num); + &mov ($sbit,"eax"); + &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]= + &jl (&label("sqr")); + + &mov ("eax",&DWP(0,$inp,$j,4)); # ap[num-1] + &mov ($carry,"edx"); + &mul ($word); # ap[num-1]*ap[0] + &add ("eax",$carry); + &mov ($word,$_n0); + &adc ("edx",0); + &mov ($inp,$_np); + &lea ($carry,&DWP(0,$sbit,"eax",2)); + &imul ($word,&DWP($frame,"esp")); # n0*tp[0] + &shr ("eax",31); + &mov (&DWP($frame,"esp",$j,4),$carry); # tp[num-1]= + + &lea ($carry,&DWP(0,"eax","edx",2)); + &mov ("eax",&DWP(0,$inp)); # np[0] + &shr ("edx",31); + &mov (&DWP($frame+4,"esp",$j,4),$carry); # tp[num]= + &mov (&DWP($frame+8,"esp",$j,4),"edx"); # tp[num+1]= + + &mul ($word); # np[0]*m + &add ("eax",&DWP($frame,"esp")); # +=tp[0] + &mov ($num,$j); + &adc ("edx",0); + &mov ("eax",&DWP(4,$inp)); # np[1] + &mov ($j,1); + +&set_label("3rdmadd",16); + &mov ($carry,"edx"); + &mul ($word); # np[j]*m + &add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j] + &adc ("edx",0); + &add ($carry,"eax"); + &mov ("eax",&DWP(4,$inp,$j,4)); # np[j+1] + &adc ("edx",0); + &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j-1]= + + &mov ($carry,"edx"); + &mul ($word); # np[j+1]*m + &add ($carry,&DWP($frame+4,"esp",$j,4)); # +=tp[j+1] + &lea ($j,&DWP(2,$j)); + &adc ("edx",0); + &add ($carry,"eax"); + &mov ("eax",&DWP(0,$inp,$j,4)); # np[j+2] + &adc ("edx",0); + &cmp ($j,$num); + &mov (&DWP($frame-8,"esp",$j,4),$carry); # tp[j]= + &jl (&label("3rdmadd")); + + &mov ($carry,"edx"); + &mul ($word); # np[j]*m + &add ($carry,&DWP($frame,"esp",$num,4)); # +=tp[num-1] + &adc ("edx",0); + &add ($carry,"eax"); + &adc ("edx",0); + &mov (&DWP($frame-4,"esp",$num,4),$carry); # tp[num-2]= + + &mov ($j,$_bp); # i + &xor ("eax","eax"); + &mov ($inp,$_ap); + &add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num] + &adc ("eax",&DWP($frame+8,"esp",$num,4)); # +=tp[num+1] + &mov (&DWP($frame,"esp",$num,4),"edx"); # tp[num-1]= + &cmp ($j,$num); + &mov (&DWP($frame+4,"esp",$num,4),"eax"); # tp[num]= + &je (&label("common_tail")); + + &mov ($word,&DWP(4,$inp,$j,4)); # ap[i] + &lea ($j,&DWP(1,$j)); + &mov ("eax",$word); + &mov ($_bp,$j); # ++i + &mul ($word); # ap[i]*ap[i] + &add ("eax",&DWP($frame,"esp",$j,4)); # +=tp[i] + &adc ("edx",0); + &mov (&DWP($frame,"esp",$j,4),"eax"); # tp[i]= + &xor ($carry,$carry); + &cmp ($j,$num); + &lea ($j,&DWP(1,$j)); + &je (&label("sqrlast")); + + &mov ($sbit,"edx"); # zaps $num + &shr ("edx",1); + &and ($sbit,1); +&set_label("sqradd",16); + &mov ("eax",&DWP(0,$inp,$j,4)); # ap[j] + &mov ($carry,"edx"); + &mul ($word); # ap[j]*ap[i] + &add ("eax",$carry); + &lea ($carry,&DWP(0,"eax","eax")); + &adc ("edx",0); + &shr ("eax",31); + &add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j] + &lea ($j,&DWP(1,$j)); + &adc ("eax",0); + &add ($carry,$sbit); + &adc ("eax",0); + &cmp ($j,$_num); + &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]= + &mov ($sbit,"eax"); + &jle (&label("sqradd")); + + &mov ($carry,"edx"); + &add ("edx","edx"); + &shr ($carry,31); + &add ("edx",$sbit); + &adc ($carry,0); +&set_label("sqrlast"); + &mov ($word,$_n0); + &mov ($inp,$_np); + &imul ($word,&DWP($frame,"esp")); # n0*tp[0] + + &add ("edx",&DWP($frame,"esp",$j,4)); # +=tp[num] + &mov ("eax",&DWP(0,$inp)); # np[0] + &adc ($carry,0); + &mov (&DWP($frame,"esp",$j,4),"edx"); # tp[num]= + &mov (&DWP($frame+4,"esp",$j,4),$carry); # tp[num+1]= + + &mul ($word); # np[0]*m + &add ("eax",&DWP($frame,"esp")); # +=tp[0] + &lea ($num,&DWP(-1,$j)); + &adc ("edx",0); + &mov ($j,1); + &mov ("eax",&DWP(4,$inp)); # np[1] + + &jmp (&label("3rdmadd")); +} + +&set_label("common_tail",16); + &mov ($np,$_np); # load modulus pointer + &mov ($rp,$_rp); # load result pointer + &lea ($tp,&DWP($frame,"esp")); # [$ap and $bp are zapped] + + &mov ("eax",&DWP(0,$tp)); # tp[0] + &mov ($j,$num); # j=num-1 + &xor ($i,$i); # i=0 and clear CF! + +&set_label("sub",16); + &sbb ("eax",&DWP(0,$np,$i,4)); + &mov (&DWP(0,$rp,$i,4),"eax"); # rp[i]=tp[i]-np[i] + &dec ($j); # doesn't affect CF! + &mov ("eax",&DWP(4,$tp,$i,4)); # tp[i+1] + &lea ($i,&DWP(1,$i)); # i++ + &jge (&label("sub")); + + &sbb ("eax",0); # handle upmost overflow bit + &and ($tp,"eax"); + ¬ ("eax"); + &mov ($np,$rp); + &and ($np,"eax"); + &or ($tp,$np); # tp=carry?tp:rp + +&set_label("copy",16); # copy or in-place refresh + &mov ("eax",&DWP(0,$tp,$num,4)); + &mov (&DWP(0,$rp,$num,4),"eax"); # rp[i]=tp[i] + &mov (&DWP($frame,"esp",$num,4),$j); # zap temporary vector + &dec ($num); + &jge (&label("copy")); + + &mov ("esp",$_sp); # pull saved stack pointer + &mov ("eax",1); +&set_label("just_leave"); +&function_end("bn_mul_mont"); + +&asciz("Montgomery Multiplication for x86, CRYPTOGAMS by "); + +&asm_finish(); diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86.pl new file mode 100644 index 00000000..1bc4f1bb --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86.pl @@ -0,0 +1,28 @@ +#!/usr/local/bin/perl + +push(@INC,"perlasm","../../perlasm"); +require "x86asm.pl"; + +require("x86/mul_add.pl"); +require("x86/mul.pl"); +require("x86/sqr.pl"); +require("x86/div.pl"); +require("x86/add.pl"); +require("x86/sub.pl"); +require("x86/comba.pl"); + +&asm_init($ARGV[0],$0); + +&bn_mul_add_words("bn_mul_add_words"); +&bn_mul_words("bn_mul_words"); +&bn_sqr_words("bn_sqr_words"); +&bn_div_words("bn_div_words"); +&bn_add_words("bn_add_words"); +&bn_sub_words("bn_sub_words"); +&bn_mul_comba("bn_mul_comba8",8); +&bn_mul_comba("bn_mul_comba4",4); +&bn_sqr_comba("bn_sqr_comba8",8); +&bn_sqr_comba("bn_sqr_comba4",4); + +&asm_finish(); + diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/add.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/add.pl new file mode 100644 index 00000000..0b5cf583 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/add.pl @@ -0,0 +1,76 @@ +#!/usr/local/bin/perl +# x86 assember + +sub bn_add_words + { + local($name)=@_; + + &function_begin($name,""); + + &comment(""); + $a="esi"; + $b="edi"; + $c="eax"; + $r="ebx"; + $tmp1="ecx"; + $tmp2="edx"; + $num="ebp"; + + &mov($r,&wparam(0)); # get r + &mov($a,&wparam(1)); # get a + &mov($b,&wparam(2)); # get b + &mov($num,&wparam(3)); # get num + &xor($c,$c); # clear carry + &and($num,0xfffffff8); # num / 8 + + &jz(&label("aw_finish")); + + &set_label("aw_loop",0); + for ($i=0; $i<8; $i++) + { + &comment("Round $i"); + + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0)); # *b + &add($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &add($tmp1,$tmp2); + &adc($c,0); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + } + + &comment(""); + &add($a,32); + &add($b,32); + &add($r,32); + &sub($num,8); + &jnz(&label("aw_loop")); + + &set_label("aw_finish",0); + &mov($num,&wparam(3)); # get num + &and($num,7); + &jz(&label("aw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0));# *b + &add($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &add($tmp1,$tmp2); + &adc($c,0); + &dec($num) if ($i != 6); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *a + &jz(&label("aw_end")) if ($i != 6); + } + &set_label("aw_end",0); + +# &mov("eax",$c); # $c is "eax" + + &function_end($name); + } + +1; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/comba.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/comba.pl new file mode 100644 index 00000000..22912536 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/comba.pl @@ -0,0 +1,277 @@ +#!/usr/local/bin/perl +# x86 assember + +sub mul_add_c + { + local($a,$ai,$b,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_; + + # pos == -1 if eax and edx are pre-loaded, 0 to load from next + # words, and 1 if load return value + + &comment("mul a[$ai]*b[$bi]"); + + # "eax" and "edx" will always be pre-loaded. + # &mov("eax",&DWP($ai*4,$a,"",0)) ; + # &mov("edx",&DWP($bi*4,$b,"",0)); + + &mul("edx"); + &add($c0,"eax"); + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # laod next a + &mov("eax",&wparam(0)) if $pos > 0; # load r[] + ### + &adc($c1,"edx"); + &mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 0; # laod next b + &mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 1; # laod next b + ### + &adc($c2,0); + # is pos > 1, it means it is the last loop + &mov(&DWP($i*4,"eax","",0),$c0) if $pos > 0; # save r[]; + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # laod next a + } + +sub sqr_add_c + { + local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_; + + # pos == -1 if eax and edx are pre-loaded, 0 to load from next + # words, and 1 if load return value + + &comment("sqr a[$ai]*a[$bi]"); + + # "eax" and "edx" will always be pre-loaded. + # &mov("eax",&DWP($ai*4,$a,"",0)) ; + # &mov("edx",&DWP($bi*4,$b,"",0)); + + if ($ai == $bi) + { &mul("eax");} + else + { &mul("edx");} + &add($c0,"eax"); + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a + ### + &adc($c1,"edx"); + &mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos == 1) && ($na != $nb); + ### + &adc($c2,0); + # is pos > 1, it means it is the last loop + &mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[]; + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b + } + +sub sqr_add_c2 + { + local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_; + + # pos == -1 if eax and edx are pre-loaded, 0 to load from next + # words, and 1 if load return value + + &comment("sqr a[$ai]*a[$bi]"); + + # "eax" and "edx" will always be pre-loaded. + # &mov("eax",&DWP($ai*4,$a,"",0)) ; + # &mov("edx",&DWP($bi*4,$a,"",0)); + + if ($ai == $bi) + { &mul("eax");} + else + { &mul("edx");} + &add("eax","eax"); + ### + &adc("edx","edx"); + ### + &adc($c2,0); + &add($c0,"eax"); + &adc($c1,"edx"); + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a + &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b + &adc($c2,0); + &mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[]; + &mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos <= 1) && ($na != $nb); + ### + } + +sub bn_mul_comba + { + local($name,$num)=@_; + local($a,$b,$c0,$c1,$c2); + local($i,$as,$ae,$bs,$be,$ai,$bi); + local($tot,$end); + + &function_begin_B($name,""); + + $c0="ebx"; + $c1="ecx"; + $c2="ebp"; + $a="esi"; + $b="edi"; + + $as=0; + $ae=0; + $bs=0; + $be=0; + $tot=$num+$num-1; + + &push("esi"); + &mov($a,&wparam(1)); + &push("edi"); + &mov($b,&wparam(2)); + &push("ebp"); + &push("ebx"); + + &xor($c0,$c0); + &mov("eax",&DWP(0,$a,"",0)); # load the first word + &xor($c1,$c1); + &mov("edx",&DWP(0,$b,"",0)); # load the first second + + for ($i=0; $i<$tot; $i++) + { + $ai=$as; + $bi=$bs; + $end=$be+1; + + &comment("################## Calculate word $i"); + + for ($j=$bs; $j<$end; $j++) + { + &xor($c2,$c2) if ($j == $bs); + if (($j+1) == $end) + { + $v=1; + $v=2 if (($i+1) == $tot); + } + else + { $v=0; } + if (($j+1) != $end) + { + $na=($ai-1); + $nb=($bi+1); + } + else + { + $na=$as+($i < ($num-1)); + $nb=$bs+($i >= ($num-1)); + } +#printf STDERR "[$ai,$bi] -> [$na,$nb]\n"; + &mul_add_c($a,$ai,$b,$bi,$c0,$c1,$c2,$v,$i,$na,$nb); + if ($v) + { + &comment("saved r[$i]"); + # &mov("eax",&wparam(0)); + # &mov(&DWP($i*4,"eax","",0),$c0); + ($c0,$c1,$c2)=($c1,$c2,$c0); + } + $ai--; + $bi++; + } + $as++ if ($i < ($num-1)); + $ae++ if ($i >= ($num-1)); + + $bs++ if ($i >= ($num-1)); + $be++ if ($i < ($num-1)); + } + &comment("save r[$i]"); + # &mov("eax",&wparam(0)); + &mov(&DWP($i*4,"eax","",0),$c0); + + &pop("ebx"); + &pop("ebp"); + &pop("edi"); + &pop("esi"); + &ret(); + &function_end_B($name); + } + +sub bn_sqr_comba + { + local($name,$num)=@_; + local($r,$a,$c0,$c1,$c2)=@_; + local($i,$as,$ae,$bs,$be,$ai,$bi); + local($b,$tot,$end,$half); + + &function_begin_B($name,""); + + $c0="ebx"; + $c1="ecx"; + $c2="ebp"; + $a="esi"; + $r="edi"; + + &push("esi"); + &push("edi"); + &push("ebp"); + &push("ebx"); + &mov($r,&wparam(0)); + &mov($a,&wparam(1)); + &xor($c0,$c0); + &xor($c1,$c1); + &mov("eax",&DWP(0,$a,"",0)); # load the first word + + $as=0; + $ae=0; + $bs=0; + $be=0; + $tot=$num+$num-1; + + for ($i=0; $i<$tot; $i++) + { + $ai=$as; + $bi=$bs; + $end=$be+1; + + &comment("############### Calculate word $i"); + for ($j=$bs; $j<$end; $j++) + { + &xor($c2,$c2) if ($j == $bs); + if (($ai-1) < ($bi+1)) + { + $v=1; + $v=2 if ($i+1) == $tot; + } + else + { $v=0; } + if (!$v) + { + $na=$ai-1; + $nb=$bi+1; + } + else + { + $na=$as+($i < ($num-1)); + $nb=$bs+($i >= ($num-1)); + } + if ($ai == $bi) + { + &sqr_add_c($r,$a,$ai,$bi, + $c0,$c1,$c2,$v,$i,$na,$nb); + } + else + { + &sqr_add_c2($r,$a,$ai,$bi, + $c0,$c1,$c2,$v,$i,$na,$nb); + } + if ($v) + { + &comment("saved r[$i]"); + #&mov(&DWP($i*4,$r,"",0),$c0); + ($c0,$c1,$c2)=($c1,$c2,$c0); + last; + } + $ai--; + $bi++; + } + $as++ if ($i < ($num-1)); + $ae++ if ($i >= ($num-1)); + + $bs++ if ($i >= ($num-1)); + $be++ if ($i < ($num-1)); + } + &mov(&DWP($i*4,$r,"",0),$c0); + &pop("ebx"); + &pop("ebp"); + &pop("edi"); + &pop("esi"); + &ret(); + &function_end_B($name); + } + +1; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/div.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/div.pl new file mode 100644 index 00000000..0e90152c --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/div.pl @@ -0,0 +1,15 @@ +#!/usr/local/bin/perl +# x86 assember + +sub bn_div_words + { + local($name)=@_; + + &function_begin($name,""); + &mov("edx",&wparam(0)); # + &mov("eax",&wparam(1)); # + &mov("ebx",&wparam(2)); # + &div("ebx"); + &function_end($name); + } +1; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/f b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/f new file mode 100644 index 00000000..22e41122 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/f @@ -0,0 +1,3 @@ +#!/usr/local/bin/perl +# x86 assember + diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/mul.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/mul.pl new file mode 100644 index 00000000..674cb9b0 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/mul.pl @@ -0,0 +1,77 @@ +#!/usr/local/bin/perl +# x86 assember + +sub bn_mul_words + { + local($name)=@_; + + &function_begin($name,""); + + &comment(""); + $Low="eax"; + $High="edx"; + $a="ebx"; + $w="ecx"; + $r="edi"; + $c="esi"; + $num="ebp"; + + &xor($c,$c); # clear carry + &mov($r,&wparam(0)); # + &mov($a,&wparam(1)); # + &mov($num,&wparam(2)); # + &mov($w,&wparam(3)); # + + &and($num,0xfffffff8); # num / 8 + &jz(&label("mw_finish")); + + &set_label("mw_loop",0); + for ($i=0; $i<32; $i+=4) + { + &comment("Round $i"); + + &mov("eax",&DWP($i,$a,"",0)); # *a + &mul($w); # *a * w + &add("eax",$c); # L(t)+=c + # XXX + + &adc("edx",0); # H(t)+=carry + &mov(&DWP($i,$r,"",0),"eax"); # *r= L(t); + + &mov($c,"edx"); # c= H(t); + } + + &comment(""); + &add($a,32); + &add($r,32); + &sub($num,8); + &jz(&label("mw_finish")); + &jmp(&label("mw_loop")); + + &set_label("mw_finish",0); + &mov($num,&wparam(2)); # get num + &and($num,7); + &jnz(&label("mw_finish2")); + &jmp(&label("mw_end")); + + &set_label("mw_finish2",1); + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov("eax",&DWP($i*4,$a,"",0));# *a + &mul($w); # *a * w + &add("eax",$c); # L(t)+=c + # XXX + &adc("edx",0); # H(t)+=carry + &mov(&DWP($i*4,$r,"",0),"eax");# *r= L(t); + &mov($c,"edx"); # c= H(t); + &dec($num) if ($i != 7-1); + &jz(&label("mw_end")) if ($i != 7-1); + } + &set_label("mw_end",0); + &mov("eax",$c); + + &function_end($name); + } + +1; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/mul_add.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/mul_add.pl new file mode 100644 index 00000000..61830d3a --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/mul_add.pl @@ -0,0 +1,87 @@ +#!/usr/local/bin/perl +# x86 assember + +sub bn_mul_add_words + { + local($name)=@_; + + &function_begin($name,""); + + &comment(""); + $Low="eax"; + $High="edx"; + $a="ebx"; + $w="ebp"; + $r="edi"; + $c="esi"; + + &xor($c,$c); # clear carry + &mov($r,&wparam(0)); # + + &mov("ecx",&wparam(2)); # + &mov($a,&wparam(1)); # + + &and("ecx",0xfffffff8); # num / 8 + &mov($w,&wparam(3)); # + + &push("ecx"); # Up the stack for a tmp variable + + &jz(&label("maw_finish")); + + &set_label("maw_loop",0); + + &mov(&swtmp(0),"ecx"); # + + for ($i=0; $i<32; $i+=4) + { + &comment("Round $i"); + + &mov("eax",&DWP($i,$a,"",0)); # *a + &mul($w); # *a * w + &add("eax",$c); # L(t)+= *r + &mov($c,&DWP($i,$r,"",0)); # L(t)+= *r + &adc("edx",0); # H(t)+=carry + &add("eax",$c); # L(t)+=c + &adc("edx",0); # H(t)+=carry + &mov(&DWP($i,$r,"",0),"eax"); # *r= L(t); + &mov($c,"edx"); # c= H(t); + } + + &comment(""); + &mov("ecx",&swtmp(0)); # + &add($a,32); + &add($r,32); + &sub("ecx",8); + &jnz(&label("maw_loop")); + + &set_label("maw_finish",0); + &mov("ecx",&wparam(2)); # get num + &and("ecx",7); + &jnz(&label("maw_finish2")); # helps branch prediction + &jmp(&label("maw_end")); + + &set_label("maw_finish2",1); + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov("eax",&DWP($i*4,$a,"",0));# *a + &mul($w); # *a * w + &add("eax",$c); # L(t)+=c + &mov($c,&DWP($i*4,$r,"",0)); # L(t)+= *r + &adc("edx",0); # H(t)+=carry + &add("eax",$c); + &adc("edx",0); # H(t)+=carry + &dec("ecx") if ($i != 7-1); + &mov(&DWP($i*4,$r,"",0),"eax"); # *r= L(t); + &mov($c,"edx"); # c= H(t); + &jz(&label("maw_end")) if ($i != 7-1); + } + &set_label("maw_end",0); + &mov("eax",$c); + + &pop("ecx"); # clear variable from + + &function_end($name); + } + +1; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/sqr.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/sqr.pl new file mode 100644 index 00000000..1f90993c --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/sqr.pl @@ -0,0 +1,60 @@ +#!/usr/local/bin/perl +# x86 assember + +sub bn_sqr_words + { + local($name)=@_; + + &function_begin($name,""); + + &comment(""); + $r="esi"; + $a="edi"; + $num="ebx"; + + &mov($r,&wparam(0)); # + &mov($a,&wparam(1)); # + &mov($num,&wparam(2)); # + + &and($num,0xfffffff8); # num / 8 + &jz(&label("sw_finish")); + + &set_label("sw_loop",0); + for ($i=0; $i<32; $i+=4) + { + &comment("Round $i"); + &mov("eax",&DWP($i,$a,"",0)); # *a + # XXX + &mul("eax"); # *a * *a + &mov(&DWP($i*2,$r,"",0),"eax"); # + &mov(&DWP($i*2+4,$r,"",0),"edx");# + } + + &comment(""); + &add($a,32); + &add($r,64); + &sub($num,8); + &jnz(&label("sw_loop")); + + &set_label("sw_finish",0); + &mov($num,&wparam(2)); # get num + &and($num,7); + &jz(&label("sw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov("eax",&DWP($i*4,$a,"",0)); # *a + # XXX + &mul("eax"); # *a * *a + &mov(&DWP($i*8,$r,"",0),"eax"); # + &dec($num) if ($i != 7-1); + &mov(&DWP($i*8+4,$r,"",0),"edx"); + &jz(&label("sw_end")) if ($i != 7-1); + } + &set_label("sw_end",0); + + &function_end($name); + } + +1; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/sub.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/sub.pl new file mode 100644 index 00000000..837b0e1b --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86/sub.pl @@ -0,0 +1,76 @@ +#!/usr/local/bin/perl +# x86 assember + +sub bn_sub_words + { + local($name)=@_; + + &function_begin($name,""); + + &comment(""); + $a="esi"; + $b="edi"; + $c="eax"; + $r="ebx"; + $tmp1="ecx"; + $tmp2="edx"; + $num="ebp"; + + &mov($r,&wparam(0)); # get r + &mov($a,&wparam(1)); # get a + &mov($b,&wparam(2)); # get b + &mov($num,&wparam(3)); # get num + &xor($c,$c); # clear carry + &and($num,0xfffffff8); # num / 8 + + &jz(&label("aw_finish")); + + &set_label("aw_loop",0); + for ($i=0; $i<8; $i++) + { + &comment("Round $i"); + + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0)); # *b + &sub($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &sub($tmp1,$tmp2); + &adc($c,0); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *r + } + + &comment(""); + &add($a,32); + &add($b,32); + &add($r,32); + &sub($num,8); + &jnz(&label("aw_loop")); + + &set_label("aw_finish",0); + &mov($num,&wparam(3)); # get num + &and($num,7); + &jz(&label("aw_end")); + + for ($i=0; $i<7; $i++) + { + &comment("Tail Round $i"); + &mov($tmp1,&DWP($i*4,$a,"",0)); # *a + &mov($tmp2,&DWP($i*4,$b,"",0));# *b + &sub($tmp1,$c); + &mov($c,0); + &adc($c,$c); + &sub($tmp1,$tmp2); + &adc($c,0); + &dec($num) if ($i != 6); + &mov(&DWP($i*4,$r,"",0),$tmp1); # *a + &jz(&label("aw_end")) if ($i != 6); + } + &set_label("aw_end",0); + +# &mov("eax",$c); # $c is "eax" + + &function_end($name); + } + +1; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-gcc.c b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-gcc.c new file mode 100644 index 00000000..329946e5 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-gcc.c @@ -0,0 +1,606 @@ +#include "../bn_lcl.h" +#if !(defined(__GNUC__) && __GNUC__>=2) +# include "../bn_asm.c" /* kind of dirty hack for Sun Studio */ +#else +/* + * x86_64 BIGNUM accelerator version 0.1, December 2002. + * + * Implemented by Andy Polyakov for the OpenSSL + * project. + * + * Rights for redistribution and usage in source and binary forms are + * granted according to the OpenSSL license. Warranty of any kind is + * disclaimed. + * + * Q. Version 0.1? It doesn't sound like Andy, he used to assign real + * versions, like 1.0... + * A. Well, that's because this code is basically a quick-n-dirty + * proof-of-concept hack. As you can see it's implemented with + * inline assembler, which means that you're bound to GCC and that + * there might be enough room for further improvement. + * + * Q. Why inline assembler? + * A. x86_64 features own ABI which I'm not familiar with. This is + * why I decided to let the compiler take care of subroutine + * prologue/epilogue as well as register allocation. For reference. + * Win64 implements different ABI for AMD64, different from Linux. + * + * Q. How much faster does it get? + * A. 'apps/openssl speed rsa dsa' output with no-asm: + * + * sign verify sign/s verify/s + * rsa 512 bits 0.0006s 0.0001s 1683.8 18456.2 + * rsa 1024 bits 0.0028s 0.0002s 356.0 6407.0 + * rsa 2048 bits 0.0172s 0.0005s 58.0 1957.8 + * rsa 4096 bits 0.1155s 0.0018s 8.7 555.6 + * sign verify sign/s verify/s + * dsa 512 bits 0.0005s 0.0006s 2100.8 1768.3 + * dsa 1024 bits 0.0014s 0.0018s 692.3 559.2 + * dsa 2048 bits 0.0049s 0.0061s 204.7 165.0 + * + * 'apps/openssl speed rsa dsa' output with this module: + * + * sign verify sign/s verify/s + * rsa 512 bits 0.0004s 0.0000s 2767.1 33297.9 + * rsa 1024 bits 0.0012s 0.0001s 867.4 14674.7 + * rsa 2048 bits 0.0061s 0.0002s 164.0 5270.0 + * rsa 4096 bits 0.0384s 0.0006s 26.1 1650.8 + * sign verify sign/s verify/s + * dsa 512 bits 0.0002s 0.0003s 4442.2 3786.3 + * dsa 1024 bits 0.0005s 0.0007s 1835.1 1497.4 + * dsa 2048 bits 0.0016s 0.0020s 620.4 504.6 + * + * For the reference. IA-32 assembler implementation performs + * very much like 64-bit code compiled with no-asm on the same + * machine. + */ + +#ifdef _WIN64 +#define BN_ULONG unsigned long long +#else +#define BN_ULONG unsigned long +#endif + +#undef mul +#undef mul_add +#undef sqr + +/* + * "m"(a), "+m"(r) is the way to favor DirectPath µ-code; + * "g"(0) let the compiler to decide where does it + * want to keep the value of zero; + */ +#define mul_add(r,a,word,carry) do { \ + register BN_ULONG high,low; \ + asm ("mulq %3" \ + : "=a"(low),"=d"(high) \ + : "a"(word),"m"(a) \ + : "cc"); \ + asm ("addq %2,%0; adcq %3,%1" \ + : "+r"(carry),"+d"(high)\ + : "a"(low),"g"(0) \ + : "cc"); \ + asm ("addq %2,%0; adcq %3,%1" \ + : "+m"(r),"+d"(high) \ + : "r"(carry),"g"(0) \ + : "cc"); \ + carry=high; \ + } while (0) + +#define mul(r,a,word,carry) do { \ + register BN_ULONG high,low; \ + asm ("mulq %3" \ + : "=a"(low),"=d"(high) \ + : "a"(word),"g"(a) \ + : "cc"); \ + asm ("addq %2,%0; adcq %3,%1" \ + : "+r"(carry),"+d"(high)\ + : "a"(low),"g"(0) \ + : "cc"); \ + (r)=carry, carry=high; \ + } while (0) + +#define sqr(r0,r1,a) \ + asm ("mulq %2" \ + : "=a"(r0),"=d"(r1) \ + : "a"(a) \ + : "cc"); + +BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w) + { + BN_ULONG c1=0; + + if (num <= 0) return(c1); + + while (num&~3) + { + mul_add(rp[0],ap[0],w,c1); + mul_add(rp[1],ap[1],w,c1); + mul_add(rp[2],ap[2],w,c1); + mul_add(rp[3],ap[3],w,c1); + ap+=4; rp+=4; num-=4; + } + if (num) + { + mul_add(rp[0],ap[0],w,c1); if (--num==0) return c1; + mul_add(rp[1],ap[1],w,c1); if (--num==0) return c1; + mul_add(rp[2],ap[2],w,c1); return c1; + } + + return(c1); + } + +BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w) + { + BN_ULONG c1=0; + + if (num <= 0) return(c1); + + while (num&~3) + { + mul(rp[0],ap[0],w,c1); + mul(rp[1],ap[1],w,c1); + mul(rp[2],ap[2],w,c1); + mul(rp[3],ap[3],w,c1); + ap+=4; rp+=4; num-=4; + } + if (num) + { + mul(rp[0],ap[0],w,c1); if (--num == 0) return c1; + mul(rp[1],ap[1],w,c1); if (--num == 0) return c1; + mul(rp[2],ap[2],w,c1); + } + return(c1); + } + +void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n) + { + if (n <= 0) return; + + while (n&~3) + { + sqr(r[0],r[1],a[0]); + sqr(r[2],r[3],a[1]); + sqr(r[4],r[5],a[2]); + sqr(r[6],r[7],a[3]); + a+=4; r+=8; n-=4; + } + if (n) + { + sqr(r[0],r[1],a[0]); if (--n == 0) return; + sqr(r[2],r[3],a[1]); if (--n == 0) return; + sqr(r[4],r[5],a[2]); + } + } + +BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d) +{ BN_ULONG ret,waste; + + asm ("divq %4" + : "=a"(ret),"=d"(waste) + : "a"(l),"d"(h),"g"(d) + : "cc"); + + return ret; +} + +BN_ULONG bn_add_words (BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,int n) +{ BN_ULONG ret=0,i=0; + + if (n <= 0) return 0; + + asm ( + " subq %2,%2 \n" + ".p2align 4 \n" + "1: movq (%4,%2,8),%0 \n" + " adcq (%5,%2,8),%0 \n" + " movq %0,(%3,%2,8) \n" + " leaq 1(%2),%2 \n" + " loop 1b \n" + " sbbq %0,%0 \n" + : "=&a"(ret),"+c"(n),"=&r"(i) + : "r"(rp),"r"(ap),"r"(bp) + : "cc" + ); + + return ret&1; +} + +#ifndef SIMICS +BN_ULONG bn_sub_words (BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,int n) +{ BN_ULONG ret=0,i=0; + + if (n <= 0) return 0; + + asm ( + " subq %2,%2 \n" + ".p2align 4 \n" + "1: movq (%4,%2,8),%0 \n" + " sbbq (%5,%2,8),%0 \n" + " movq %0,(%3,%2,8) \n" + " leaq 1(%2),%2 \n" + " loop 1b \n" + " sbbq %0,%0 \n" + : "=&a"(ret),"+c"(n),"=&r"(i) + : "r"(rp),"r"(ap),"r"(bp) + : "cc" + ); + + return ret&1; +} +#else +/* Simics 1.4<7 has buggy sbbq:-( */ +#define BN_MASK2 0xffffffffffffffffL +BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) + { + BN_ULONG t1,t2; + int c=0; + + if (n <= 0) return((BN_ULONG)0); + + for (;;) + { + t1=a[0]; t2=b[0]; + r[0]=(t1-t2-c)&BN_MASK2; + if (t1 != t2) c=(t1 < t2); + if (--n <= 0) break; + + t1=a[1]; t2=b[1]; + r[1]=(t1-t2-c)&BN_MASK2; + if (t1 != t2) c=(t1 < t2); + if (--n <= 0) break; + + t1=a[2]; t2=b[2]; + r[2]=(t1-t2-c)&BN_MASK2; + if (t1 != t2) c=(t1 < t2); + if (--n <= 0) break; + + t1=a[3]; t2=b[3]; + r[3]=(t1-t2-c)&BN_MASK2; + if (t1 != t2) c=(t1 < t2); + if (--n <= 0) break; + + a+=4; + b+=4; + r+=4; + } + return(c); + } +#endif + +/* mul_add_c(a,b,c0,c1,c2) -- c+=a*b for three word number c=(c2,c1,c0) */ +/* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */ +/* sqr_add_c(a,i,c0,c1,c2) -- c+=a[i]^2 for three word number c=(c2,c1,c0) */ +/* sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number c=(c2,c1,c0) */ + +#if 0 +/* original macros are kept for reference purposes */ +#define mul_add_c(a,b,c0,c1,c2) { \ + BN_ULONG ta=(a),tb=(b); \ + t1 = ta * tb; \ + t2 = BN_UMULT_HIGH(ta,tb); \ + c0 += t1; t2 += (c0 for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# May 2011 +# +# The module implements bn_GF2m_mul_2x2 polynomial multiplication used +# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for +# the time being... Except that it has two code paths: code suitable +# for any x86_64 CPU and PCLMULQDQ one suitable for Westmere and +# later. Improvement varies from one benchmark and µ-arch to another. +# Vanilla code path is at most 20% faster than compiler-generated code +# [not very impressive], while PCLMULQDQ - whole 85%-160% better on +# 163- and 571-bit ECDH benchmarks on Intel CPUs. Keep in mind that +# these coefficients are not ones for bn_GF2m_mul_2x2 itself, as not +# all CPU time is burnt in it... + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open OUT,"| \"$^X\" $xlate $flavour $output"; +*STDOUT=*OUT; + +($lo,$hi)=("%rax","%rdx"); $a=$lo; +($i0,$i1)=("%rsi","%rdi"); +($t0,$t1)=("%rbx","%rcx"); +($b,$mask)=("%rbp","%r8"); +($a1,$a2,$a4,$a8,$a12,$a48)=map("%r$_",(9..15)); +($R,$Tx)=("%xmm0","%xmm1"); + +$code.=<<___; +.text + +.type _mul_1x1,\@abi-omnipotent +.align 16 +_mul_1x1: + sub \$128+8,%rsp + mov \$-1,$a1 + lea ($a,$a),$i0 + shr \$3,$a1 + lea (,$a,4),$i1 + and $a,$a1 # a1=a&0x1fffffffffffffff + lea (,$a,8),$a8 + sar \$63,$a # broadcast 63rd bit + lea ($a1,$a1),$a2 + sar \$63,$i0 # broadcast 62nd bit + lea (,$a1,4),$a4 + and $b,$a + sar \$63,$i1 # boardcast 61st bit + mov $a,$hi # $a is $lo + shl \$63,$lo + and $b,$i0 + shr \$1,$hi + mov $i0,$t1 + shl \$62,$i0 + and $b,$i1 + shr \$2,$t1 + xor $i0,$lo + mov $i1,$t0 + shl \$61,$i1 + xor $t1,$hi + shr \$3,$t0 + xor $i1,$lo + xor $t0,$hi + + mov $a1,$a12 + movq \$0,0(%rsp) # tab[0]=0 + xor $a2,$a12 # a1^a2 + mov $a1,8(%rsp) # tab[1]=a1 + mov $a4,$a48 + mov $a2,16(%rsp) # tab[2]=a2 + xor $a8,$a48 # a4^a8 + mov $a12,24(%rsp) # tab[3]=a1^a2 + + xor $a4,$a1 + mov $a4,32(%rsp) # tab[4]=a4 + xor $a4,$a2 + mov $a1,40(%rsp) # tab[5]=a1^a4 + xor $a4,$a12 + mov $a2,48(%rsp) # tab[6]=a2^a4 + xor $a48,$a1 # a1^a4^a4^a8=a1^a8 + mov $a12,56(%rsp) # tab[7]=a1^a2^a4 + xor $a48,$a2 # a2^a4^a4^a8=a1^a8 + + mov $a8,64(%rsp) # tab[8]=a8 + xor $a48,$a12 # a1^a2^a4^a4^a8=a1^a2^a8 + mov $a1,72(%rsp) # tab[9]=a1^a8 + xor $a4,$a1 # a1^a8^a4 + mov $a2,80(%rsp) # tab[10]=a2^a8 + xor $a4,$a2 # a2^a8^a4 + mov $a12,88(%rsp) # tab[11]=a1^a2^a8 + + xor $a4,$a12 # a1^a2^a8^a4 + mov $a48,96(%rsp) # tab[12]=a4^a8 + mov $mask,$i0 + mov $a1,104(%rsp) # tab[13]=a1^a4^a8 + and $b,$i0 + mov $a2,112(%rsp) # tab[14]=a2^a4^a8 + shr \$4,$b + mov $a12,120(%rsp) # tab[15]=a1^a2^a4^a8 + mov $mask,$i1 + and $b,$i1 + shr \$4,$b + + movq (%rsp,$i0,8),$R # half of calculations is done in SSE2 + mov $mask,$i0 + and $b,$i0 + shr \$4,$b +___ + for ($n=1;$n<8;$n++) { + $code.=<<___; + mov (%rsp,$i1,8),$t1 + mov $mask,$i1 + mov $t1,$t0 + shl \$`8*$n-4`,$t1 + and $b,$i1 + movq (%rsp,$i0,8),$Tx + shr \$`64-(8*$n-4)`,$t0 + xor $t1,$lo + pslldq \$$n,$Tx + mov $mask,$i0 + shr \$4,$b + xor $t0,$hi + and $b,$i0 + shr \$4,$b + pxor $Tx,$R +___ + } +$code.=<<___; + mov (%rsp,$i1,8),$t1 + mov $t1,$t0 + shl \$`8*$n-4`,$t1 + movq $R,$i0 + shr \$`64-(8*$n-4)`,$t0 + xor $t1,$lo + psrldq \$8,$R + xor $t0,$hi + movq $R,$i1 + xor $i0,$lo + xor $i1,$hi + + add \$128+8,%rsp + ret +.Lend_mul_1x1: +.size _mul_1x1,.-_mul_1x1 +___ + +($rp,$a1,$a0,$b1,$b0) = $win64? ("%rcx","%rdx","%r8", "%r9","%r10") : # Win64 order + ("%rdi","%rsi","%rdx","%rcx","%r8"); # Unix order + +$code.=<<___; +.extern OPENSSL_ia32cap_P +.globl bn_GF2m_mul_2x2 +.type bn_GF2m_mul_2x2,\@abi-omnipotent +.align 16 +bn_GF2m_mul_2x2: + mov OPENSSL_ia32cap_P(%rip),%rax + bt \$33,%rax + jnc .Lvanilla_mul_2x2 + + movq $a1,%xmm0 + movq $b1,%xmm1 + movq $a0,%xmm2 +___ +$code.=<<___ if ($win64); + movq 40(%rsp),%xmm3 +___ +$code.=<<___ if (!$win64); + movq $b0,%xmm3 +___ +$code.=<<___; + movdqa %xmm0,%xmm4 + movdqa %xmm1,%xmm5 + pclmulqdq \$0,%xmm1,%xmm0 # a1·b1 + pxor %xmm2,%xmm4 + pxor %xmm3,%xmm5 + pclmulqdq \$0,%xmm3,%xmm2 # a0·b0 + pclmulqdq \$0,%xmm5,%xmm4 # (a0+a1)·(b0+b1) + xorps %xmm0,%xmm4 + xorps %xmm2,%xmm4 # (a0+a1)·(b0+b1)-a0·b0-a1·b1 + movdqa %xmm4,%xmm5 + pslldq \$8,%xmm4 + psrldq \$8,%xmm5 + pxor %xmm4,%xmm2 + pxor %xmm5,%xmm0 + movdqu %xmm2,0($rp) + movdqu %xmm0,16($rp) + ret + +.align 16 +.Lvanilla_mul_2x2: + lea -8*17(%rsp),%rsp +___ +$code.=<<___ if ($win64); + mov `8*17+40`(%rsp),$b0 + mov %rdi,8*15(%rsp) + mov %rsi,8*16(%rsp) +___ +$code.=<<___; + mov %r14,8*10(%rsp) + mov %r13,8*11(%rsp) + mov %r12,8*12(%rsp) + mov %rbp,8*13(%rsp) + mov %rbx,8*14(%rsp) +.Lbody_mul_2x2: + mov $rp,32(%rsp) # save the arguments + mov $a1,40(%rsp) + mov $a0,48(%rsp) + mov $b1,56(%rsp) + mov $b0,64(%rsp) + + mov \$0xf,$mask + mov $a1,$a + mov $b1,$b + call _mul_1x1 # a1·b1 + mov $lo,16(%rsp) + mov $hi,24(%rsp) + + mov 48(%rsp),$a + mov 64(%rsp),$b + call _mul_1x1 # a0·b0 + mov $lo,0(%rsp) + mov $hi,8(%rsp) + + mov 40(%rsp),$a + mov 56(%rsp),$b + xor 48(%rsp),$a + xor 64(%rsp),$b + call _mul_1x1 # (a0+a1)·(b0+b1) +___ + @r=("%rbx","%rcx","%rdi","%rsi"); +$code.=<<___; + mov 0(%rsp),@r[0] + mov 8(%rsp),@r[1] + mov 16(%rsp),@r[2] + mov 24(%rsp),@r[3] + mov 32(%rsp),%rbp + + xor $hi,$lo + xor @r[1],$hi + xor @r[0],$lo + mov @r[0],0(%rbp) + xor @r[2],$hi + mov @r[3],24(%rbp) + xor @r[3],$lo + xor @r[3],$hi + xor $hi,$lo + mov $hi,16(%rbp) + mov $lo,8(%rbp) + + mov 8*10(%rsp),%r14 + mov 8*11(%rsp),%r13 + mov 8*12(%rsp),%r12 + mov 8*13(%rsp),%rbp + mov 8*14(%rsp),%rbx +___ +$code.=<<___ if ($win64); + mov 8*15(%rsp),%rdi + mov 8*16(%rsp),%rsi +___ +$code.=<<___; + lea 8*17(%rsp),%rsp + ret +.Lend_mul_2x2: +.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2 +.asciz "GF(2^m) Multiplication for x86_64, CRYPTOGAMS by " +.align 16 +___ + +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +if ($win64) { +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind + +.type se_handler,\@abi-omnipotent +.align 16 +se_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 152($context),%rax # pull context->Rsp + mov 248($context),%rbx # pull context->Rip + + lea .Lbody_mul_2x2(%rip),%r10 + cmp %r10,%rbx # context->Rip<"prologue" label + jb .Lin_prologue + + mov 8*10(%rax),%r14 # mimic epilogue + mov 8*11(%rax),%r13 + mov 8*12(%rax),%r12 + mov 8*13(%rax),%rbp + mov 8*14(%rax),%rbx + mov 8*15(%rax),%rdi + mov 8*16(%rax),%rsi + + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + +.Lin_prologue: + lea 8*17(%rax),%rax + mov %rax,152($context) # restore context->Rsp + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$154,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size se_handler,.-se_handler + +.section .pdata +.align 4 + .rva _mul_1x1 + .rva .Lend_mul_1x1 + .rva .LSEH_info_1x1 + + .rva .Lvanilla_mul_2x2 + .rva .Lend_mul_2x2 + .rva .LSEH_info_2x2 +.section .xdata +.align 8 +.LSEH_info_1x1: + .byte 0x01,0x07,0x02,0x00 + .byte 0x07,0x01,0x11,0x00 # sub rsp,128+8 +.LSEH_info_2x2: + .byte 9,0,0,0 + .rva se_handler +___ +} + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont.S new file mode 100644 index 00000000..95e29052 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont.S @@ -0,0 +1,1374 @@ +.text + +.globl bn_mul_mont +.type bn_mul_mont,@function +.align 16 +bn_mul_mont: + testl $3,%r9d + jnz .Lmul_enter + cmpl $8,%r9d + jb .Lmul_enter + cmpq %rsi,%rdx + jne .Lmul4x_enter + jmp .Lsqr4x_enter + +.align 16 +.Lmul_enter: + pushq %rbx + pushq %rbp + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + + movl %r9d,%r9d + leaq 2(%r9),%r10 + movq %rsp,%r11 + negq %r10 + leaq (%rsp,%r10,8),%rsp + andq $-1024,%rsp + + movq %r11,8(%rsp,%r9,8) +.Lmul_body: + movq %rdx,%r12 + movq (%r8),%r8 + movq (%r12),%rbx + movq (%rsi),%rax + + xorq %r14,%r14 + xorq %r15,%r15 + + movq %r8,%rbp + mulq %rbx + movq %rax,%r10 + movq (%rcx),%rax + + imulq %r10,%rbp + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r10 + movq 8(%rsi),%rax + adcq $0,%rdx + movq %rdx,%r13 + + leaq 1(%r15),%r15 + jmp .L1st_enter + +.align 16 +.L1st: + addq %rax,%r13 + movq (%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%r13 + movq %r10,%r11 + adcq $0,%rdx + movq %r13,-16(%rsp,%r15,8) + movq %rdx,%r13 + +.L1st_enter: + mulq %rbx + addq %rax,%r11 + movq (%rcx,%r15,8),%rax + adcq $0,%rdx + leaq 1(%r15),%r15 + movq %rdx,%r10 + + mulq %rbp + cmpq %r9,%r15 + jne .L1st + + addq %rax,%r13 + movq (%rsi),%rax + adcq $0,%rdx + addq %r11,%r13 + adcq $0,%rdx + movq %r13,-16(%rsp,%r15,8) + movq %rdx,%r13 + movq %r10,%r11 + + xorq %rdx,%rdx + addq %r11,%r13 + adcq $0,%rdx + movq %r13,-8(%rsp,%r9,8) + movq %rdx,(%rsp,%r9,8) + + leaq 1(%r14),%r14 + jmp .Louter +.align 16 +.Louter: + movq (%r12,%r14,8),%rbx + xorq %r15,%r15 + movq %r8,%rbp + movq (%rsp),%r10 + mulq %rbx + addq %rax,%r10 + movq (%rcx),%rax + adcq $0,%rdx + + imulq %r10,%rbp + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r10 + movq 8(%rsi),%rax + adcq $0,%rdx + movq 8(%rsp),%r10 + movq %rdx,%r13 + + leaq 1(%r15),%r15 + jmp .Linner_enter + +.align 16 +.Linner: + addq %rax,%r13 + movq (%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + movq (%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %r13,-16(%rsp,%r15,8) + movq %rdx,%r13 + +.Linner_enter: + mulq %rbx + addq %rax,%r11 + movq (%rcx,%r15,8),%rax + adcq $0,%rdx + addq %r11,%r10 + movq %rdx,%r11 + adcq $0,%r11 + leaq 1(%r15),%r15 + + mulq %rbp + cmpq %r9,%r15 + jne .Linner + + addq %rax,%r13 + movq (%rsi),%rax + adcq $0,%rdx + addq %r10,%r13 + movq (%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %r13,-16(%rsp,%r15,8) + movq %rdx,%r13 + + xorq %rdx,%rdx + addq %r11,%r13 + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-8(%rsp,%r9,8) + movq %rdx,(%rsp,%r9,8) + + leaq 1(%r14),%r14 + cmpq %r9,%r14 + jl .Louter + + xorq %r14,%r14 + movq (%rsp),%rax + leaq (%rsp),%rsi + movq %r9,%r15 + jmp .Lsub +.align 16 +.Lsub: sbbq (%rcx,%r14,8),%rax + movq %rax,(%rdi,%r14,8) + movq 8(%rsi,%r14,8),%rax + leaq 1(%r14),%r14 + decq %r15 + jnz .Lsub + + sbbq $0,%rax + xorq %r14,%r14 + andq %rax,%rsi + notq %rax + movq %rdi,%rcx + andq %rax,%rcx + movq %r9,%r15 + orq %rcx,%rsi +.align 16 +.Lcopy: + movq (%rsi,%r14,8),%rax + movq %r14,(%rsp,%r14,8) + movq %rax,(%rdi,%r14,8) + leaq 1(%r14),%r14 + subq $1,%r15 + jnz .Lcopy + + movq 8(%rsp,%r9,8),%rsi + movq $1,%rax + movq (%rsi),%r15 + movq 8(%rsi),%r14 + movq 16(%rsi),%r13 + movq 24(%rsi),%r12 + movq 32(%rsi),%rbp + movq 40(%rsi),%rbx + leaq 48(%rsi),%rsp +.Lmul_epilogue: + .byte 0xf3,0xc3 +.size bn_mul_mont,.-bn_mul_mont +.type bn_mul4x_mont,@function +.align 16 +bn_mul4x_mont: +.Lmul4x_enter: + pushq %rbx + pushq %rbp + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + + movl %r9d,%r9d + leaq 4(%r9),%r10 + movq %rsp,%r11 + negq %r10 + leaq (%rsp,%r10,8),%rsp + andq $-1024,%rsp + + movq %r11,8(%rsp,%r9,8) +.Lmul4x_body: + movq %rdi,16(%rsp,%r9,8) + movq %rdx,%r12 + movq (%r8),%r8 + movq (%r12),%rbx + movq (%rsi),%rax + + xorq %r14,%r14 + xorq %r15,%r15 + + movq %r8,%rbp + mulq %rbx + movq %rax,%r10 + movq (%rcx),%rax + + imulq %r10,%rbp + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r10 + movq 8(%rsi),%rax + adcq $0,%rdx + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq 8(%rcx),%rax + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq 16(%rsi),%rax + adcq $0,%rdx + addq %r11,%rdi + leaq 4(%r15),%r15 + adcq $0,%rdx + movq %rdi,(%rsp) + movq %rdx,%r13 + jmp .L1st4x +.align 16 +.L1st4x: + mulq %rbx + addq %rax,%r10 + movq -16(%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq -8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-24(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq -8(%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq (%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-16(%rsp,%r15,8) + movq %rdx,%r13 + + mulq %rbx + addq %rax,%r10 + movq (%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq 8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-8(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq 8(%rcx,%r15,8),%rax + adcq $0,%rdx + leaq 4(%r15),%r15 + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq -16(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-32(%rsp,%r15,8) + movq %rdx,%r13 + cmpq %r9,%r15 + jl .L1st4x + + mulq %rbx + addq %rax,%r10 + movq -16(%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq -8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-24(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq -8(%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq (%rsi),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-16(%rsp,%r15,8) + movq %rdx,%r13 + + xorq %rdi,%rdi + addq %r10,%r13 + adcq $0,%rdi + movq %r13,-8(%rsp,%r15,8) + movq %rdi,(%rsp,%r15,8) + + leaq 1(%r14),%r14 +.align 4 +.Louter4x: + movq (%r12,%r14,8),%rbx + xorq %r15,%r15 + movq (%rsp),%r10 + movq %r8,%rbp + mulq %rbx + addq %rax,%r10 + movq (%rcx),%rax + adcq $0,%rdx + + imulq %r10,%rbp + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r10 + movq 8(%rsi),%rax + adcq $0,%rdx + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq 8(%rcx),%rax + adcq $0,%rdx + addq 8(%rsp),%r11 + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq 16(%rsi),%rax + adcq $0,%rdx + addq %r11,%rdi + leaq 4(%r15),%r15 + adcq $0,%rdx + movq %rdi,(%rsp) + movq %rdx,%r13 + jmp .Linner4x +.align 16 +.Linner4x: + mulq %rbx + addq %rax,%r10 + movq -16(%rcx,%r15,8),%rax + adcq $0,%rdx + addq -16(%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq -8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-24(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq -8(%rcx,%r15,8),%rax + adcq $0,%rdx + addq -8(%rsp,%r15,8),%r11 + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq (%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-16(%rsp,%r15,8) + movq %rdx,%r13 + + mulq %rbx + addq %rax,%r10 + movq (%rcx,%r15,8),%rax + adcq $0,%rdx + addq (%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq 8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-8(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq 8(%rcx,%r15,8),%rax + adcq $0,%rdx + addq 8(%rsp,%r15,8),%r11 + adcq $0,%rdx + leaq 4(%r15),%r15 + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq -16(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-32(%rsp,%r15,8) + movq %rdx,%r13 + cmpq %r9,%r15 + jl .Linner4x + + mulq %rbx + addq %rax,%r10 + movq -16(%rcx,%r15,8),%rax + adcq $0,%rdx + addq -16(%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq -8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-24(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq -8(%rcx,%r15,8),%rax + adcq $0,%rdx + addq -8(%rsp,%r15,8),%r11 + adcq $0,%rdx + leaq 1(%r14),%r14 + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq (%rsi),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-16(%rsp,%r15,8) + movq %rdx,%r13 + + xorq %rdi,%rdi + addq %r10,%r13 + adcq $0,%rdi + addq (%rsp,%r9,8),%r13 + adcq $0,%rdi + movq %r13,-8(%rsp,%r15,8) + movq %rdi,(%rsp,%r15,8) + + cmpq %r9,%r14 + jl .Louter4x + movq 16(%rsp,%r9,8),%rdi + movq 0(%rsp),%rax + pxor %xmm0,%xmm0 + movq 8(%rsp),%rdx + shrq $2,%r9 + leaq (%rsp),%rsi + xorq %r14,%r14 + + subq 0(%rcx),%rax + movq 16(%rsi),%rbx + movq 24(%rsi),%rbp + sbbq 8(%rcx),%rdx + leaq -1(%r9),%r15 + jmp .Lsub4x +.align 16 +.Lsub4x: + movq %rax,0(%rdi,%r14,8) + movq %rdx,8(%rdi,%r14,8) + sbbq 16(%rcx,%r14,8),%rbx + movq 32(%rsi,%r14,8),%rax + movq 40(%rsi,%r14,8),%rdx + sbbq 24(%rcx,%r14,8),%rbp + movq %rbx,16(%rdi,%r14,8) + movq %rbp,24(%rdi,%r14,8) + sbbq 32(%rcx,%r14,8),%rax + movq 48(%rsi,%r14,8),%rbx + movq 56(%rsi,%r14,8),%rbp + sbbq 40(%rcx,%r14,8),%rdx + leaq 4(%r14),%r14 + decq %r15 + jnz .Lsub4x + + movq %rax,0(%rdi,%r14,8) + movq 32(%rsi,%r14,8),%rax + sbbq 16(%rcx,%r14,8),%rbx + movq %rdx,8(%rdi,%r14,8) + sbbq 24(%rcx,%r14,8),%rbp + movq %rbx,16(%rdi,%r14,8) + + sbbq $0,%rax + movq %rbp,24(%rdi,%r14,8) + xorq %r14,%r14 + andq %rax,%rsi + notq %rax + movq %rdi,%rcx + andq %rax,%rcx + leaq -1(%r9),%r15 + orq %rcx,%rsi + + movdqu (%rsi),%xmm1 + movdqa %xmm0,(%rsp) + movdqu %xmm1,(%rdi) + jmp .Lcopy4x +.align 16 +.Lcopy4x: + movdqu 16(%rsi,%r14,1),%xmm2 + movdqu 32(%rsi,%r14,1),%xmm1 + movdqa %xmm0,16(%rsp,%r14,1) + movdqu %xmm2,16(%rdi,%r14,1) + movdqa %xmm0,32(%rsp,%r14,1) + movdqu %xmm1,32(%rdi,%r14,1) + leaq 32(%r14),%r14 + decq %r15 + jnz .Lcopy4x + + shlq $2,%r9 + movdqu 16(%rsi,%r14,1),%xmm2 + movdqa %xmm0,16(%rsp,%r14,1) + movdqu %xmm2,16(%rdi,%r14,1) + movq 8(%rsp,%r9,8),%rsi + movq $1,%rax + movq (%rsi),%r15 + movq 8(%rsi),%r14 + movq 16(%rsi),%r13 + movq 24(%rsi),%r12 + movq 32(%rsi),%rbp + movq 40(%rsi),%rbx + leaq 48(%rsi),%rsp +.Lmul4x_epilogue: + .byte 0xf3,0xc3 +.size bn_mul4x_mont,.-bn_mul4x_mont +.type bn_sqr4x_mont,@function +.align 16 +bn_sqr4x_mont: +.Lsqr4x_enter: + pushq %rbx + pushq %rbp + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + + shll $3,%r9d + xorq %r10,%r10 + movq %rsp,%r11 + subq %r9,%r10 + movq (%r8),%r8 + leaq -72(%rsp,%r10,2),%rsp + andq $-1024,%rsp + + + + + + + + + + + + movq %rdi,32(%rsp) + movq %rcx,40(%rsp) + movq %r8,48(%rsp) + movq %r11,56(%rsp) +.Lsqr4x_body: + + + + + + + + leaq 32(%r10),%rbp + leaq (%rsi,%r9,1),%rsi + + movq %r9,%rcx + + + movq -32(%rsi,%rbp,1),%r14 + leaq 64(%rsp,%r9,2),%rdi + movq -24(%rsi,%rbp,1),%rax + leaq -32(%rdi,%rbp,1),%rdi + movq -16(%rsi,%rbp,1),%rbx + movq %rax,%r15 + + mulq %r14 + movq %rax,%r10 + movq %rbx,%rax + movq %rdx,%r11 + movq %r10,-24(%rdi,%rbp,1) + + xorq %r10,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + movq %r11,-16(%rdi,%rbp,1) + + leaq -16(%rbp),%rcx + + + movq 8(%rsi,%rcx,1),%rbx + mulq %r15 + movq %rax,%r12 + movq %rbx,%rax + movq %rdx,%r13 + + xorq %r11,%r11 + addq %r12,%r10 + leaq 16(%rcx),%rcx + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %rbx,%rax + adcq %rdx,%r11 + movq %r10,-8(%rdi,%rcx,1) + jmp .Lsqr4x_1st + +.align 16 +.Lsqr4x_1st: + movq (%rsi,%rcx,1),%rbx + xorq %r12,%r12 + mulq %r15 + addq %rax,%r13 + movq %rbx,%rax + adcq %rdx,%r12 + + xorq %r10,%r10 + addq %r13,%r11 + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + movq %r11,(%rdi,%rcx,1) + + + movq 8(%rsi,%rcx,1),%rbx + xorq %r13,%r13 + mulq %r15 + addq %rax,%r12 + movq %rbx,%rax + adcq %rdx,%r13 + + xorq %r11,%r11 + addq %r12,%r10 + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %rbx,%rax + adcq %rdx,%r11 + movq %r10,8(%rdi,%rcx,1) + + movq 16(%rsi,%rcx,1),%rbx + xorq %r12,%r12 + mulq %r15 + addq %rax,%r13 + movq %rbx,%rax + adcq %rdx,%r12 + + xorq %r10,%r10 + addq %r13,%r11 + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + movq %r11,16(%rdi,%rcx,1) + + + movq 24(%rsi,%rcx,1),%rbx + xorq %r13,%r13 + mulq %r15 + addq %rax,%r12 + movq %rbx,%rax + adcq %rdx,%r13 + + xorq %r11,%r11 + addq %r12,%r10 + leaq 32(%rcx),%rcx + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %rbx,%rax + adcq %rdx,%r11 + movq %r10,-8(%rdi,%rcx,1) + + cmpq $0,%rcx + jne .Lsqr4x_1st + + xorq %r12,%r12 + addq %r11,%r13 + adcq $0,%r12 + mulq %r15 + addq %rax,%r13 + adcq %rdx,%r12 + + movq %r13,(%rdi) + leaq 16(%rbp),%rbp + movq %r12,8(%rdi) + jmp .Lsqr4x_outer + +.align 16 +.Lsqr4x_outer: + movq -32(%rsi,%rbp,1),%r14 + leaq 64(%rsp,%r9,2),%rdi + movq -24(%rsi,%rbp,1),%rax + leaq -32(%rdi,%rbp,1),%rdi + movq -16(%rsi,%rbp,1),%rbx + movq %rax,%r15 + + movq -24(%rdi,%rbp,1),%r10 + xorq %r11,%r11 + mulq %r14 + addq %rax,%r10 + movq %rbx,%rax + adcq %rdx,%r11 + movq %r10,-24(%rdi,%rbp,1) + + xorq %r10,%r10 + addq -16(%rdi,%rbp,1),%r11 + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + movq %r11,-16(%rdi,%rbp,1) + + leaq -16(%rbp),%rcx + xorq %r12,%r12 + + + movq 8(%rsi,%rcx,1),%rbx + xorq %r13,%r13 + addq 8(%rdi,%rcx,1),%r12 + adcq $0,%r13 + mulq %r15 + addq %rax,%r12 + movq %rbx,%rax + adcq %rdx,%r13 + + xorq %r11,%r11 + addq %r12,%r10 + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %rbx,%rax + adcq %rdx,%r11 + movq %r10,8(%rdi,%rcx,1) + + leaq 16(%rcx),%rcx + jmp .Lsqr4x_inner + +.align 16 +.Lsqr4x_inner: + movq (%rsi,%rcx,1),%rbx + xorq %r12,%r12 + addq (%rdi,%rcx,1),%r13 + adcq $0,%r12 + mulq %r15 + addq %rax,%r13 + movq %rbx,%rax + adcq %rdx,%r12 + + xorq %r10,%r10 + addq %r13,%r11 + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + movq %r11,(%rdi,%rcx,1) + + movq 8(%rsi,%rcx,1),%rbx + xorq %r13,%r13 + addq 8(%rdi,%rcx,1),%r12 + adcq $0,%r13 + mulq %r15 + addq %rax,%r12 + movq %rbx,%rax + adcq %rdx,%r13 + + xorq %r11,%r11 + addq %r12,%r10 + leaq 16(%rcx),%rcx + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %rbx,%rax + adcq %rdx,%r11 + movq %r10,-8(%rdi,%rcx,1) + + cmpq $0,%rcx + jne .Lsqr4x_inner + + xorq %r12,%r12 + addq %r11,%r13 + adcq $0,%r12 + mulq %r15 + addq %rax,%r13 + adcq %rdx,%r12 + + movq %r13,(%rdi) + movq %r12,8(%rdi) + + addq $16,%rbp + jnz .Lsqr4x_outer + + + movq -32(%rsi),%r14 + leaq 64(%rsp,%r9,2),%rdi + movq -24(%rsi),%rax + leaq -32(%rdi,%rbp,1),%rdi + movq -16(%rsi),%rbx + movq %rax,%r15 + + xorq %r11,%r11 + mulq %r14 + addq %rax,%r10 + movq %rbx,%rax + adcq %rdx,%r11 + movq %r10,-24(%rdi) + + xorq %r10,%r10 + addq %r13,%r11 + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + movq %r11,-16(%rdi) + + movq -8(%rsi),%rbx + mulq %r15 + addq %rax,%r12 + movq %rbx,%rax + adcq $0,%rdx + + xorq %r11,%r11 + addq %r12,%r10 + movq %rdx,%r13 + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %rbx,%rax + adcq %rdx,%r11 + movq %r10,-8(%rdi) + + xorq %r12,%r12 + addq %r11,%r13 + adcq $0,%r12 + mulq %r15 + addq %rax,%r13 + movq -16(%rsi),%rax + adcq %rdx,%r12 + + movq %r13,(%rdi) + movq %r12,8(%rdi) + + mulq %rbx + addq $16,%rbp + xorq %r14,%r14 + subq %r9,%rbp + xorq %r15,%r15 + + addq %r12,%rax + adcq $0,%rdx + movq %rax,8(%rdi) + movq %rdx,16(%rdi) + movq %r15,24(%rdi) + + movq -16(%rsi,%rbp,1),%rax + leaq 64(%rsp,%r9,2),%rdi + xorq %r10,%r10 + movq -24(%rdi,%rbp,2),%r11 + + leaq (%r14,%r10,2),%r12 + shrq $63,%r10 + leaq (%rcx,%r11,2),%r13 + shrq $63,%r11 + orq %r10,%r13 + movq -16(%rdi,%rbp,2),%r10 + movq %r11,%r14 + mulq %rax + negq %r15 + movq -8(%rdi,%rbp,2),%r11 + adcq %rax,%r12 + movq -8(%rsi,%rbp,1),%rax + movq %r12,-32(%rdi,%rbp,2) + adcq %rdx,%r13 + + leaq (%r14,%r10,2),%rbx + movq %r13,-24(%rdi,%rbp,2) + sbbq %r15,%r15 + shrq $63,%r10 + leaq (%rcx,%r11,2),%r8 + shrq $63,%r11 + orq %r10,%r8 + movq 0(%rdi,%rbp,2),%r10 + movq %r11,%r14 + mulq %rax + negq %r15 + movq 8(%rdi,%rbp,2),%r11 + adcq %rax,%rbx + movq 0(%rsi,%rbp,1),%rax + movq %rbx,-16(%rdi,%rbp,2) + adcq %rdx,%r8 + leaq 16(%rbp),%rbp + movq %r8,-40(%rdi,%rbp,2) + sbbq %r15,%r15 + jmp .Lsqr4x_shift_n_add + +.align 16 +.Lsqr4x_shift_n_add: + leaq (%r14,%r10,2),%r12 + shrq $63,%r10 + leaq (%rcx,%r11,2),%r13 + shrq $63,%r11 + orq %r10,%r13 + movq -16(%rdi,%rbp,2),%r10 + movq %r11,%r14 + mulq %rax + negq %r15 + movq -8(%rdi,%rbp,2),%r11 + adcq %rax,%r12 + movq -8(%rsi,%rbp,1),%rax + movq %r12,-32(%rdi,%rbp,2) + adcq %rdx,%r13 + + leaq (%r14,%r10,2),%rbx + movq %r13,-24(%rdi,%rbp,2) + sbbq %r15,%r15 + shrq $63,%r10 + leaq (%rcx,%r11,2),%r8 + shrq $63,%r11 + orq %r10,%r8 + movq 0(%rdi,%rbp,2),%r10 + movq %r11,%r14 + mulq %rax + negq %r15 + movq 8(%rdi,%rbp,2),%r11 + adcq %rax,%rbx + movq 0(%rsi,%rbp,1),%rax + movq %rbx,-16(%rdi,%rbp,2) + adcq %rdx,%r8 + + leaq (%r14,%r10,2),%r12 + movq %r8,-8(%rdi,%rbp,2) + sbbq %r15,%r15 + shrq $63,%r10 + leaq (%rcx,%r11,2),%r13 + shrq $63,%r11 + orq %r10,%r13 + movq 16(%rdi,%rbp,2),%r10 + movq %r11,%r14 + mulq %rax + negq %r15 + movq 24(%rdi,%rbp,2),%r11 + adcq %rax,%r12 + movq 8(%rsi,%rbp,1),%rax + movq %r12,0(%rdi,%rbp,2) + adcq %rdx,%r13 + + leaq (%r14,%r10,2),%rbx + movq %r13,8(%rdi,%rbp,2) + sbbq %r15,%r15 + shrq $63,%r10 + leaq (%rcx,%r11,2),%r8 + shrq $63,%r11 + orq %r10,%r8 + movq 32(%rdi,%rbp,2),%r10 + movq %r11,%r14 + mulq %rax + negq %r15 + movq 40(%rdi,%rbp,2),%r11 + adcq %rax,%rbx + movq 16(%rsi,%rbp,1),%rax + movq %rbx,16(%rdi,%rbp,2) + adcq %rdx,%r8 + movq %r8,24(%rdi,%rbp,2) + sbbq %r15,%r15 + addq $32,%rbp + jnz .Lsqr4x_shift_n_add + + leaq (%r14,%r10,2),%r12 + shrq $63,%r10 + leaq (%rcx,%r11,2),%r13 + shrq $63,%r11 + orq %r10,%r13 + movq -16(%rdi),%r10 + movq %r11,%r14 + mulq %rax + negq %r15 + movq -8(%rdi),%r11 + adcq %rax,%r12 + movq -8(%rsi),%rax + movq %r12,-32(%rdi) + adcq %rdx,%r13 + + leaq (%r14,%r10,2),%rbx + movq %r13,-24(%rdi) + sbbq %r15,%r15 + shrq $63,%r10 + leaq (%rcx,%r11,2),%r8 + shrq $63,%r11 + orq %r10,%r8 + mulq %rax + negq %r15 + adcq %rax,%rbx + adcq %rdx,%r8 + movq %rbx,-16(%rdi) + movq %r8,-8(%rdi) + movq 40(%rsp),%rsi + movq 48(%rsp),%r8 + xorq %rcx,%rcx + movq %r9,0(%rsp) + subq %r9,%rcx + movq 64(%rsp),%r10 + movq %r8,%r14 + leaq 64(%rsp,%r9,2),%rax + leaq 64(%rsp,%r9,1),%rdi + movq %rax,8(%rsp) + leaq (%rsi,%r9,1),%rsi + xorq %rbp,%rbp + + movq 0(%rsi,%rcx,1),%rax + movq 8(%rsi,%rcx,1),%r9 + imulq %r10,%r14 + movq %rax,%rbx + jmp .Lsqr4x_mont_outer + +.align 16 +.Lsqr4x_mont_outer: + xorq %r11,%r11 + mulq %r14 + addq %rax,%r10 + movq %r9,%rax + adcq %rdx,%r11 + movq %r8,%r15 + + xorq %r10,%r10 + addq 8(%rdi,%rcx,1),%r11 + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + + imulq %r11,%r15 + + movq 16(%rsi,%rcx,1),%rbx + xorq %r13,%r13 + addq %r11,%r12 + adcq $0,%r13 + mulq %r15 + addq %rax,%r12 + movq %rbx,%rax + adcq %rdx,%r13 + movq %r12,8(%rdi,%rcx,1) + + xorq %r11,%r11 + addq 16(%rdi,%rcx,1),%r10 + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %r9,%rax + adcq %rdx,%r11 + + movq 24(%rsi,%rcx,1),%r9 + xorq %r12,%r12 + addq %r10,%r13 + adcq $0,%r12 + mulq %r15 + addq %rax,%r13 + movq %r9,%rax + adcq %rdx,%r12 + movq %r13,16(%rdi,%rcx,1) + + xorq %r10,%r10 + addq 24(%rdi,%rcx,1),%r11 + leaq 32(%rcx),%rcx + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + jmp .Lsqr4x_mont_inner + +.align 16 +.Lsqr4x_mont_inner: + movq (%rsi,%rcx,1),%rbx + xorq %r13,%r13 + addq %r11,%r12 + adcq $0,%r13 + mulq %r15 + addq %rax,%r12 + movq %rbx,%rax + adcq %rdx,%r13 + movq %r12,-8(%rdi,%rcx,1) + + xorq %r11,%r11 + addq (%rdi,%rcx,1),%r10 + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %r9,%rax + adcq %rdx,%r11 + + movq 8(%rsi,%rcx,1),%r9 + xorq %r12,%r12 + addq %r10,%r13 + adcq $0,%r12 + mulq %r15 + addq %rax,%r13 + movq %r9,%rax + adcq %rdx,%r12 + movq %r13,(%rdi,%rcx,1) + + xorq %r10,%r10 + addq 8(%rdi,%rcx,1),%r11 + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + + + movq 16(%rsi,%rcx,1),%rbx + xorq %r13,%r13 + addq %r11,%r12 + adcq $0,%r13 + mulq %r15 + addq %rax,%r12 + movq %rbx,%rax + adcq %rdx,%r13 + movq %r12,8(%rdi,%rcx,1) + + xorq %r11,%r11 + addq 16(%rdi,%rcx,1),%r10 + adcq $0,%r11 + mulq %r14 + addq %rax,%r10 + movq %r9,%rax + adcq %rdx,%r11 + + movq 24(%rsi,%rcx,1),%r9 + xorq %r12,%r12 + addq %r10,%r13 + adcq $0,%r12 + mulq %r15 + addq %rax,%r13 + movq %r9,%rax + adcq %rdx,%r12 + movq %r13,16(%rdi,%rcx,1) + + xorq %r10,%r10 + addq 24(%rdi,%rcx,1),%r11 + leaq 32(%rcx),%rcx + adcq $0,%r10 + mulq %r14 + addq %rax,%r11 + movq %rbx,%rax + adcq %rdx,%r10 + cmpq $0,%rcx + jne .Lsqr4x_mont_inner + + subq 0(%rsp),%rcx + movq %r8,%r14 + + xorq %r13,%r13 + addq %r11,%r12 + adcq $0,%r13 + mulq %r15 + addq %rax,%r12 + movq %r9,%rax + adcq %rdx,%r13 + movq %r12,-8(%rdi) + + xorq %r11,%r11 + addq (%rdi),%r10 + adcq $0,%r11 + movq 0(%rsi,%rcx,1),%rbx + addq %rbp,%r10 + adcq $0,%r11 + + imulq 16(%rdi,%rcx,1),%r14 + xorq %r12,%r12 + movq 8(%rsi,%rcx,1),%r9 + addq %r10,%r13 + movq 16(%rdi,%rcx,1),%r10 + adcq $0,%r12 + mulq %r15 + addq %rax,%r13 + movq %rbx,%rax + adcq %rdx,%r12 + movq %r13,(%rdi) + + xorq %rbp,%rbp + addq 8(%rdi),%r12 + adcq %rbp,%rbp + addq %r11,%r12 + leaq 16(%rdi),%rdi + adcq $0,%rbp + movq %r12,-8(%rdi) + cmpq 8(%rsp),%rdi + jb .Lsqr4x_mont_outer + + movq 0(%rsp),%r9 + movq %rbp,(%rdi) + movq 64(%rsp,%r9,1),%rax + leaq 64(%rsp,%r9,1),%rbx + movq 40(%rsp),%rsi + shrq $5,%r9 + movq 8(%rbx),%rdx + xorq %rbp,%rbp + + movq 32(%rsp),%rdi + subq 0(%rsi),%rax + movq 16(%rbx),%r10 + movq 24(%rbx),%r11 + sbbq 8(%rsi),%rdx + leaq -1(%r9),%rcx + jmp .Lsqr4x_sub +.align 16 +.Lsqr4x_sub: + movq %rax,0(%rdi,%rbp,8) + movq %rdx,8(%rdi,%rbp,8) + sbbq 16(%rsi,%rbp,8),%r10 + movq 32(%rbx,%rbp,8),%rax + movq 40(%rbx,%rbp,8),%rdx + sbbq 24(%rsi,%rbp,8),%r11 + movq %r10,16(%rdi,%rbp,8) + movq %r11,24(%rdi,%rbp,8) + sbbq 32(%rsi,%rbp,8),%rax + movq 48(%rbx,%rbp,8),%r10 + movq 56(%rbx,%rbp,8),%r11 + sbbq 40(%rsi,%rbp,8),%rdx + leaq 4(%rbp),%rbp + decq %rcx + jnz .Lsqr4x_sub + + movq %rax,0(%rdi,%rbp,8) + movq 32(%rbx,%rbp,8),%rax + sbbq 16(%rsi,%rbp,8),%r10 + movq %rdx,8(%rdi,%rbp,8) + sbbq 24(%rsi,%rbp,8),%r11 + movq %r10,16(%rdi,%rbp,8) + + sbbq $0,%rax + movq %r11,24(%rdi,%rbp,8) + xorq %rbp,%rbp + andq %rax,%rbx + notq %rax + movq %rdi,%rsi + andq %rax,%rsi + leaq -1(%r9),%rcx + orq %rsi,%rbx + + pxor %xmm0,%xmm0 + leaq 64(%rsp,%r9,8),%rsi + movdqu (%rbx),%xmm1 + leaq (%rsi,%r9,8),%rsi + movdqa %xmm0,64(%rsp) + movdqa %xmm0,(%rsi) + movdqu %xmm1,(%rdi) + jmp .Lsqr4x_copy +.align 16 +.Lsqr4x_copy: + movdqu 16(%rbx,%rbp,1),%xmm2 + movdqu 32(%rbx,%rbp,1),%xmm1 + movdqa %xmm0,80(%rsp,%rbp,1) + movdqa %xmm0,96(%rsp,%rbp,1) + movdqa %xmm0,16(%rsi,%rbp,1) + movdqa %xmm0,32(%rsi,%rbp,1) + movdqu %xmm2,16(%rdi,%rbp,1) + movdqu %xmm1,32(%rdi,%rbp,1) + leaq 32(%rbp),%rbp + decq %rcx + jnz .Lsqr4x_copy + + movdqu 16(%rbx,%rbp,1),%xmm2 + movdqa %xmm0,80(%rsp,%rbp,1) + movdqa %xmm0,16(%rsi,%rbp,1) + movdqu %xmm2,16(%rdi,%rbp,1) + movq 56(%rsp),%rsi + movq $1,%rax + movq 0(%rsi),%r15 + movq 8(%rsi),%r14 + movq 16(%rsi),%r13 + movq 24(%rsi),%r12 + movq 32(%rsi),%rbp + movq 40(%rsi),%rbx + leaq 48(%rsi),%rsp +.Lsqr4x_epilogue: + .byte 0xf3,0xc3 +.size bn_sqr4x_mont,.-bn_sqr4x_mont +.byte 77,111,110,116,103,111,109,101,114,121,32,77,117,108,116,105,112,108,105,99,97,116,105,111,110,32,102,111,114,32,120,56,54,95,54,52,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0 +.align 16 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont.pl new file mode 100755 index 00000000..17fb94c8 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont.pl @@ -0,0 +1,1681 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# October 2005. +# +# Montgomery multiplication routine for x86_64. While it gives modest +# 9% improvement of rsa4096 sign on Opteron, rsa512 sign runs more +# than twice, >2x, as fast. Most common rsa1024 sign is improved by +# respectful 50%. It remains to be seen if loop unrolling and +# dedicated squaring routine can provide further improvement... + +# July 2011. +# +# Add dedicated squaring procedure. Performance improvement varies +# from platform to platform, but in average it's ~5%/15%/25%/33% +# for 512-/1024-/2048-/4096-bit RSA *sign* benchmarks respectively. + +# August 2011. +# +# Unroll and modulo-schedule inner loops in such manner that they +# are "fallen through" for input lengths of 8, which is critical for +# 1024-bit RSA *sign*. Average performance improvement in comparison +# to *initial* version of this module from 2005 is ~0%/30%/40%/45% +# for 512-/1024-/2048-/4096-bit RSA *sign* benchmarks respectively. + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open OUT,"| \"$^X\" $xlate $flavour $output"; +*STDOUT=*OUT; + +# int bn_mul_mont( +$rp="%rdi"; # BN_ULONG *rp, +$ap="%rsi"; # const BN_ULONG *ap, +$bp="%rdx"; # const BN_ULONG *bp, +$np="%rcx"; # const BN_ULONG *np, +$n0="%r8"; # const BN_ULONG *n0, +$num="%r9"; # int num); +$lo0="%r10"; +$hi0="%r11"; +$hi1="%r13"; +$i="%r14"; +$j="%r15"; +$m0="%rbx"; +$m1="%rbp"; + +$code=<<___; +.text + +.globl bn_mul_mont +.type bn_mul_mont,\@function,6 +.align 16 +bn_mul_mont: + test \$3,${num}d + jnz .Lmul_enter + cmp \$8,${num}d + jb .Lmul_enter + cmp $ap,$bp + jne .Lmul4x_enter + jmp .Lsqr4x_enter + +.align 16 +.Lmul_enter: + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + + mov ${num}d,${num}d + lea 2($num),%r10 + mov %rsp,%r11 + neg %r10 + lea (%rsp,%r10,8),%rsp # tp=alloca(8*(num+2)) + and \$-1024,%rsp # minimize TLB usage + + mov %r11,8(%rsp,$num,8) # tp[num+1]=%rsp +.Lmul_body: + mov $bp,%r12 # reassign $bp +___ + $bp="%r12"; +$code.=<<___; + mov ($n0),$n0 # pull n0[0] value + mov ($bp),$m0 # m0=bp[0] + mov ($ap),%rax + + xor $i,$i # i=0 + xor $j,$j # j=0 + + mov $n0,$m1 + mulq $m0 # ap[0]*bp[0] + mov %rax,$lo0 + mov ($np),%rax + + imulq $lo0,$m1 # "tp[0]"*n0 + mov %rdx,$hi0 + + mulq $m1 # np[0]*m1 + add %rax,$lo0 # discarded + mov 8($ap),%rax + adc \$0,%rdx + mov %rdx,$hi1 + + lea 1($j),$j # j++ + jmp .L1st_enter + +.align 16 +.L1st: + add %rax,$hi1 + mov ($ap,$j,8),%rax + adc \$0,%rdx + add $hi0,$hi1 # np[j]*m1+ap[j]*bp[0] + mov $lo0,$hi0 + adc \$0,%rdx + mov $hi1,-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$hi1 + +.L1st_enter: + mulq $m0 # ap[j]*bp[0] + add %rax,$hi0 + mov ($np,$j,8),%rax + adc \$0,%rdx + lea 1($j),$j # j++ + mov %rdx,$lo0 + + mulq $m1 # np[j]*m1 + cmp $num,$j + jne .L1st + + add %rax,$hi1 + mov ($ap),%rax # ap[0] + adc \$0,%rdx + add $hi0,$hi1 # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $hi1,-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$hi1 + mov $lo0,$hi0 + + xor %rdx,%rdx + add $hi0,$hi1 + adc \$0,%rdx + mov $hi1,-8(%rsp,$num,8) + mov %rdx,(%rsp,$num,8) # store upmost overflow bit + + lea 1($i),$i # i++ + jmp .Louter +.align 16 +.Louter: + mov ($bp,$i,8),$m0 # m0=bp[i] + xor $j,$j # j=0 + mov $n0,$m1 + mov (%rsp),$lo0 + mulq $m0 # ap[0]*bp[i] + add %rax,$lo0 # ap[0]*bp[i]+tp[0] + mov ($np),%rax + adc \$0,%rdx + + imulq $lo0,$m1 # tp[0]*n0 + mov %rdx,$hi0 + + mulq $m1 # np[0]*m1 + add %rax,$lo0 # discarded + mov 8($ap),%rax + adc \$0,%rdx + mov 8(%rsp),$lo0 # tp[1] + mov %rdx,$hi1 + + lea 1($j),$j # j++ + jmp .Linner_enter + +.align 16 +.Linner: + add %rax,$hi1 + mov ($ap,$j,8),%rax + adc \$0,%rdx + add $lo0,$hi1 # np[j]*m1+ap[j]*bp[i]+tp[j] + mov (%rsp,$j,8),$lo0 + adc \$0,%rdx + mov $hi1,-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$hi1 + +.Linner_enter: + mulq $m0 # ap[j]*bp[i] + add %rax,$hi0 + mov ($np,$j,8),%rax + adc \$0,%rdx + add $hi0,$lo0 # ap[j]*bp[i]+tp[j] + mov %rdx,$hi0 + adc \$0,$hi0 + lea 1($j),$j # j++ + + mulq $m1 # np[j]*m1 + cmp $num,$j + jne .Linner + + add %rax,$hi1 + mov ($ap),%rax # ap[0] + adc \$0,%rdx + add $lo0,$hi1 # np[j]*m1+ap[j]*bp[i]+tp[j] + mov (%rsp,$j,8),$lo0 + adc \$0,%rdx + mov $hi1,-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$hi1 + + xor %rdx,%rdx + add $hi0,$hi1 + adc \$0,%rdx + add $lo0,$hi1 # pull upmost overflow bit + adc \$0,%rdx + mov $hi1,-8(%rsp,$num,8) + mov %rdx,(%rsp,$num,8) # store upmost overflow bit + + lea 1($i),$i # i++ + cmp $num,$i + jl .Louter + + xor $i,$i # i=0 and clear CF! + mov (%rsp),%rax # tp[0] + lea (%rsp),$ap # borrow ap for tp + mov $num,$j # j=num + jmp .Lsub +.align 16 +.Lsub: sbb ($np,$i,8),%rax + mov %rax,($rp,$i,8) # rp[i]=tp[i]-np[i] + mov 8($ap,$i,8),%rax # tp[i+1] + lea 1($i),$i # i++ + dec $j # doesnn't affect CF! + jnz .Lsub + + sbb \$0,%rax # handle upmost overflow bit + xor $i,$i + and %rax,$ap + not %rax + mov $rp,$np + and %rax,$np + mov $num,$j # j=num + or $np,$ap # ap=borrow?tp:rp +.align 16 +.Lcopy: # copy or in-place refresh + mov ($ap,$i,8),%rax + mov $i,(%rsp,$i,8) # zap temporary vector + mov %rax,($rp,$i,8) # rp[i]=tp[i] + lea 1($i),$i + sub \$1,$j + jnz .Lcopy + + mov 8(%rsp,$num,8),%rsi # restore %rsp + mov \$1,%rax + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lmul_epilogue: + ret +.size bn_mul_mont,.-bn_mul_mont +___ +{{{ +my @A=("%r10","%r11"); +my @N=("%r13","%rdi"); +$code.=<<___; +.type bn_mul4x_mont,\@function,6 +.align 16 +bn_mul4x_mont: +.Lmul4x_enter: + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + + mov ${num}d,${num}d + lea 4($num),%r10 + mov %rsp,%r11 + neg %r10 + lea (%rsp,%r10,8),%rsp # tp=alloca(8*(num+4)) + and \$-1024,%rsp # minimize TLB usage + + mov %r11,8(%rsp,$num,8) # tp[num+1]=%rsp +.Lmul4x_body: + mov $rp,16(%rsp,$num,8) # tp[num+2]=$rp + mov %rdx,%r12 # reassign $bp +___ + $bp="%r12"; +$code.=<<___; + mov ($n0),$n0 # pull n0[0] value + mov ($bp),$m0 # m0=bp[0] + mov ($ap),%rax + + xor $i,$i # i=0 + xor $j,$j # j=0 + + mov $n0,$m1 + mulq $m0 # ap[0]*bp[0] + mov %rax,$A[0] + mov ($np),%rax + + imulq $A[0],$m1 # "tp[0]"*n0 + mov %rdx,$A[1] + + mulq $m1 # np[0]*m1 + add %rax,$A[0] # discarded + mov 8($ap),%rax + adc \$0,%rdx + mov %rdx,$N[1] + + mulq $m0 + add %rax,$A[1] + mov 8($np),%rax + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 + add %rax,$N[1] + mov 16($ap),%rax + adc \$0,%rdx + add $A[1],$N[1] + lea 4($j),$j # j++ + adc \$0,%rdx + mov $N[1],(%rsp) + mov %rdx,$N[0] + jmp .L1st4x +.align 16 +.L1st4x: + mulq $m0 # ap[j]*bp[0] + add %rax,$A[0] + mov -16($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov -8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[0],-24(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[1] + mov -8($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov ($ap,$j,8),%rax + adc \$0,%rdx + add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[1],-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[0] + mov ($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov 8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[0],-8(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[1] + mov 8($np,$j,8),%rax + adc \$0,%rdx + lea 4($j),$j # j++ + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov -16($ap,$j,8),%rax + adc \$0,%rdx + add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[1],-32(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + cmp $num,$j + jl .L1st4x + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[0] + mov -16($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov -8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[0],-24(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[1] + mov -8($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov ($ap),%rax # ap[0] + adc \$0,%rdx + add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[1],-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + + xor $N[1],$N[1] + add $A[0],$N[0] + adc \$0,$N[1] + mov $N[0],-8(%rsp,$j,8) + mov $N[1],(%rsp,$j,8) # store upmost overflow bit + + lea 1($i),$i # i++ +.align 4 +.Louter4x: + mov ($bp,$i,8),$m0 # m0=bp[i] + xor $j,$j # j=0 + mov (%rsp),$A[0] + mov $n0,$m1 + mulq $m0 # ap[0]*bp[i] + add %rax,$A[0] # ap[0]*bp[i]+tp[0] + mov ($np),%rax + adc \$0,%rdx + + imulq $A[0],$m1 # tp[0]*n0 + mov %rdx,$A[1] + + mulq $m1 # np[0]*m1 + add %rax,$A[0] # "$N[0]", discarded + mov 8($ap),%rax + adc \$0,%rdx + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[1] + mov 8($np),%rax + adc \$0,%rdx + add 8(%rsp),$A[1] # +tp[1] + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov 16($ap),%rax + adc \$0,%rdx + add $A[1],$N[1] # np[j]*m1+ap[j]*bp[i]+tp[j] + lea 4($j),$j # j+=2 + adc \$0,%rdx + mov $N[1],(%rsp) # tp[j-1] + mov %rdx,$N[0] + jmp .Linner4x +.align 16 +.Linner4x: + mulq $m0 # ap[j]*bp[i] + add %rax,$A[0] + mov -16($np,$j,8),%rax + adc \$0,%rdx + add -16(%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j] + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov -8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] + adc \$0,%rdx + mov $N[0],-24(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[1] + mov -8($np,$j,8),%rax + adc \$0,%rdx + add -8(%rsp,$j,8),$A[1] + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov ($ap,$j,8),%rax + adc \$0,%rdx + add $A[1],$N[1] + adc \$0,%rdx + mov $N[1],-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[0] + mov ($np,$j,8),%rax + adc \$0,%rdx + add (%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j] + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov 8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] + adc \$0,%rdx + mov $N[0],-8(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[1] + mov 8($np,$j,8),%rax + adc \$0,%rdx + add 8(%rsp,$j,8),$A[1] + adc \$0,%rdx + lea 4($j),$j # j++ + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov -16($ap,$j,8),%rax + adc \$0,%rdx + add $A[1],$N[1] + adc \$0,%rdx + mov $N[1],-32(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + cmp $num,$j + jl .Linner4x + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[0] + mov -16($np,$j,8),%rax + adc \$0,%rdx + add -16(%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j] + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov -8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] + adc \$0,%rdx + mov $N[0],-24(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[1] + mov -8($np,$j,8),%rax + adc \$0,%rdx + add -8(%rsp,$j,8),$A[1] + adc \$0,%rdx + lea 1($i),$i # i++ + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov ($ap),%rax # ap[0] + adc \$0,%rdx + add $A[1],$N[1] + adc \$0,%rdx + mov $N[1],-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + + xor $N[1],$N[1] + add $A[0],$N[0] + adc \$0,$N[1] + add (%rsp,$num,8),$N[0] # pull upmost overflow bit + adc \$0,$N[1] + mov $N[0],-8(%rsp,$j,8) + mov $N[1],(%rsp,$j,8) # store upmost overflow bit + + cmp $num,$i + jl .Louter4x +___ +{ +my @ri=("%rax","%rdx",$m0,$m1); +$code.=<<___; + mov 16(%rsp,$num,8),$rp # restore $rp + mov 0(%rsp),@ri[0] # tp[0] + pxor %xmm0,%xmm0 + mov 8(%rsp),@ri[1] # tp[1] + shr \$2,$num # num/=4 + lea (%rsp),$ap # borrow ap for tp + xor $i,$i # i=0 and clear CF! + + sub 0($np),@ri[0] + mov 16($ap),@ri[2] # tp[2] + mov 24($ap),@ri[3] # tp[3] + sbb 8($np),@ri[1] + lea -1($num),$j # j=num/4-1 + jmp .Lsub4x +.align 16 +.Lsub4x: + mov @ri[0],0($rp,$i,8) # rp[i]=tp[i]-np[i] + mov @ri[1],8($rp,$i,8) # rp[i]=tp[i]-np[i] + sbb 16($np,$i,8),@ri[2] + mov 32($ap,$i,8),@ri[0] # tp[i+1] + mov 40($ap,$i,8),@ri[1] + sbb 24($np,$i,8),@ri[3] + mov @ri[2],16($rp,$i,8) # rp[i]=tp[i]-np[i] + mov @ri[3],24($rp,$i,8) # rp[i]=tp[i]-np[i] + sbb 32($np,$i,8),@ri[0] + mov 48($ap,$i,8),@ri[2] + mov 56($ap,$i,8),@ri[3] + sbb 40($np,$i,8),@ri[1] + lea 4($i),$i # i++ + dec $j # doesnn't affect CF! + jnz .Lsub4x + + mov @ri[0],0($rp,$i,8) # rp[i]=tp[i]-np[i] + mov 32($ap,$i,8),@ri[0] # load overflow bit + sbb 16($np,$i,8),@ri[2] + mov @ri[1],8($rp,$i,8) # rp[i]=tp[i]-np[i] + sbb 24($np,$i,8),@ri[3] + mov @ri[2],16($rp,$i,8) # rp[i]=tp[i]-np[i] + + sbb \$0,@ri[0] # handle upmost overflow bit + mov @ri[3],24($rp,$i,8) # rp[i]=tp[i]-np[i] + xor $i,$i # i=0 + and @ri[0],$ap + not @ri[0] + mov $rp,$np + and @ri[0],$np + lea -1($num),$j + or $np,$ap # ap=borrow?tp:rp + + movdqu ($ap),%xmm1 + movdqa %xmm0,(%rsp) + movdqu %xmm1,($rp) + jmp .Lcopy4x +.align 16 +.Lcopy4x: # copy or in-place refresh + movdqu 16($ap,$i),%xmm2 + movdqu 32($ap,$i),%xmm1 + movdqa %xmm0,16(%rsp,$i) + movdqu %xmm2,16($rp,$i) + movdqa %xmm0,32(%rsp,$i) + movdqu %xmm1,32($rp,$i) + lea 32($i),$i + dec $j + jnz .Lcopy4x + + shl \$2,$num + movdqu 16($ap,$i),%xmm2 + movdqa %xmm0,16(%rsp,$i) + movdqu %xmm2,16($rp,$i) +___ +} +$code.=<<___; + mov 8(%rsp,$num,8),%rsi # restore %rsp + mov \$1,%rax + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lmul4x_epilogue: + ret +.size bn_mul4x_mont,.-bn_mul4x_mont +___ +}}} + {{{ +###################################################################### +# void bn_sqr4x_mont( +my $rptr="%rdi"; # const BN_ULONG *rptr, +my $aptr="%rsi"; # const BN_ULONG *aptr, +my $bptr="%rdx"; # not used +my $nptr="%rcx"; # const BN_ULONG *nptr, +my $n0 ="%r8"; # const BN_ULONG *n0); +my $num ="%r9"; # int num, has to be divisible by 4 and + # not less than 8 + +my ($i,$j,$tptr)=("%rbp","%rcx",$rptr); +my @A0=("%r10","%r11"); +my @A1=("%r12","%r13"); +my ($a0,$a1,$ai)=("%r14","%r15","%rbx"); + +$code.=<<___; +.type bn_sqr4x_mont,\@function,6 +.align 16 +bn_sqr4x_mont: +.Lsqr4x_enter: + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + + shl \$3,${num}d # convert $num to bytes + xor %r10,%r10 + mov %rsp,%r11 # put aside %rsp + sub $num,%r10 # -$num + mov ($n0),$n0 # *n0 + lea -72(%rsp,%r10,2),%rsp # alloca(frame+2*$num) + and \$-1024,%rsp # minimize TLB usage + ############################################################## + # Stack layout + # + # +0 saved $num, used in reduction section + # +8 &t[2*$num], used in reduction section + # +32 saved $rptr + # +40 saved $nptr + # +48 saved *n0 + # +56 saved %rsp + # +64 t[2*$num] + # + mov $rptr,32(%rsp) # save $rptr + mov $nptr,40(%rsp) + mov $n0, 48(%rsp) + mov %r11, 56(%rsp) # save original %rsp +.Lsqr4x_body: + ############################################################## + # Squaring part: + # + # a) multiply-n-add everything but a[i]*a[i]; + # b) shift result of a) by 1 to the left and accumulate + # a[i]*a[i] products; + # + lea 32(%r10),$i # $i=-($num-32) + lea ($aptr,$num),$aptr # end of a[] buffer, ($aptr,$i)=&ap[2] + + mov $num,$j # $j=$num + + # comments apply to $num==8 case + mov -32($aptr,$i),$a0 # a[0] + lea 64(%rsp,$num,2),$tptr # end of tp[] buffer, &tp[2*$num] + mov -24($aptr,$i),%rax # a[1] + lea -32($tptr,$i),$tptr # end of tp[] window, &tp[2*$num-"$i"] + mov -16($aptr,$i),$ai # a[2] + mov %rax,$a1 + + mul $a0 # a[1]*a[0] + mov %rax,$A0[0] # a[1]*a[0] + mov $ai,%rax # a[2] + mov %rdx,$A0[1] + mov $A0[0],-24($tptr,$i) # t[1] + + xor $A0[0],$A0[0] + mul $a0 # a[2]*a[0] + add %rax,$A0[1] + mov $ai,%rax + adc %rdx,$A0[0] + mov $A0[1],-16($tptr,$i) # t[2] + + lea -16($i),$j # j=-16 + + + mov 8($aptr,$j),$ai # a[3] + mul $a1 # a[2]*a[1] + mov %rax,$A1[0] # a[2]*a[1]+t[3] + mov $ai,%rax + mov %rdx,$A1[1] + + xor $A0[1],$A0[1] + add $A1[0],$A0[0] + lea 16($j),$j + adc \$0,$A0[1] + mul $a0 # a[3]*a[0] + add %rax,$A0[0] # a[3]*a[0]+a[2]*a[1]+t[3] + mov $ai,%rax + adc %rdx,$A0[1] + mov $A0[0],-8($tptr,$j) # t[3] + jmp .Lsqr4x_1st + +.align 16 +.Lsqr4x_1st: + mov ($aptr,$j),$ai # a[4] + xor $A1[0],$A1[0] + mul $a1 # a[3]*a[1] + add %rax,$A1[1] # a[3]*a[1]+t[4] + mov $ai,%rax + adc %rdx,$A1[0] + + xor $A0[0],$A0[0] + add $A1[1],$A0[1] + adc \$0,$A0[0] + mul $a0 # a[4]*a[0] + add %rax,$A0[1] # a[4]*a[0]+a[3]*a[1]+t[4] + mov $ai,%rax # a[3] + adc %rdx,$A0[0] + mov $A0[1],($tptr,$j) # t[4] + + + mov 8($aptr,$j),$ai # a[5] + xor $A1[1],$A1[1] + mul $a1 # a[4]*a[3] + add %rax,$A1[0] # a[4]*a[3]+t[5] + mov $ai,%rax + adc %rdx,$A1[1] + + xor $A0[1],$A0[1] + add $A1[0],$A0[0] + adc \$0,$A0[1] + mul $a0 # a[5]*a[2] + add %rax,$A0[0] # a[5]*a[2]+a[4]*a[3]+t[5] + mov $ai,%rax + adc %rdx,$A0[1] + mov $A0[0],8($tptr,$j) # t[5] + + mov 16($aptr,$j),$ai # a[6] + xor $A1[0],$A1[0] + mul $a1 # a[5]*a[3] + add %rax,$A1[1] # a[5]*a[3]+t[6] + mov $ai,%rax + adc %rdx,$A1[0] + + xor $A0[0],$A0[0] + add $A1[1],$A0[1] + adc \$0,$A0[0] + mul $a0 # a[6]*a[2] + add %rax,$A0[1] # a[6]*a[2]+a[5]*a[3]+t[6] + mov $ai,%rax # a[3] + adc %rdx,$A0[0] + mov $A0[1],16($tptr,$j) # t[6] + + + mov 24($aptr,$j),$ai # a[7] + xor $A1[1],$A1[1] + mul $a1 # a[6]*a[5] + add %rax,$A1[0] # a[6]*a[5]+t[7] + mov $ai,%rax + adc %rdx,$A1[1] + + xor $A0[1],$A0[1] + add $A1[0],$A0[0] + lea 32($j),$j + adc \$0,$A0[1] + mul $a0 # a[7]*a[4] + add %rax,$A0[0] # a[7]*a[4]+a[6]*a[5]+t[6] + mov $ai,%rax + adc %rdx,$A0[1] + mov $A0[0],-8($tptr,$j) # t[7] + + cmp \$0,$j + jne .Lsqr4x_1st + + xor $A1[0],$A1[0] + add $A0[1],$A1[1] + adc \$0,$A1[0] + mul $a1 # a[7]*a[5] + add %rax,$A1[1] + adc %rdx,$A1[0] + + mov $A1[1],($tptr) # t[8] + lea 16($i),$i + mov $A1[0],8($tptr) # t[9] + jmp .Lsqr4x_outer + +.align 16 +.Lsqr4x_outer: # comments apply to $num==6 case + mov -32($aptr,$i),$a0 # a[0] + lea 64(%rsp,$num,2),$tptr # end of tp[] buffer, &tp[2*$num] + mov -24($aptr,$i),%rax # a[1] + lea -32($tptr,$i),$tptr # end of tp[] window, &tp[2*$num-"$i"] + mov -16($aptr,$i),$ai # a[2] + mov %rax,$a1 + + mov -24($tptr,$i),$A0[0] # t[1] + xor $A0[1],$A0[1] + mul $a0 # a[1]*a[0] + add %rax,$A0[0] # a[1]*a[0]+t[1] + mov $ai,%rax # a[2] + adc %rdx,$A0[1] + mov $A0[0],-24($tptr,$i) # t[1] + + xor $A0[0],$A0[0] + add -16($tptr,$i),$A0[1] # a[2]*a[0]+t[2] + adc \$0,$A0[0] + mul $a0 # a[2]*a[0] + add %rax,$A0[1] + mov $ai,%rax + adc %rdx,$A0[0] + mov $A0[1],-16($tptr,$i) # t[2] + + lea -16($i),$j # j=-16 + xor $A1[0],$A1[0] + + + mov 8($aptr,$j),$ai # a[3] + xor $A1[1],$A1[1] + add 8($tptr,$j),$A1[0] + adc \$0,$A1[1] + mul $a1 # a[2]*a[1] + add %rax,$A1[0] # a[2]*a[1]+t[3] + mov $ai,%rax + adc %rdx,$A1[1] + + xor $A0[1],$A0[1] + add $A1[0],$A0[0] + adc \$0,$A0[1] + mul $a0 # a[3]*a[0] + add %rax,$A0[0] # a[3]*a[0]+a[2]*a[1]+t[3] + mov $ai,%rax + adc %rdx,$A0[1] + mov $A0[0],8($tptr,$j) # t[3] + + lea 16($j),$j + jmp .Lsqr4x_inner + +.align 16 +.Lsqr4x_inner: + mov ($aptr,$j),$ai # a[4] + xor $A1[0],$A1[0] + add ($tptr,$j),$A1[1] + adc \$0,$A1[0] + mul $a1 # a[3]*a[1] + add %rax,$A1[1] # a[3]*a[1]+t[4] + mov $ai,%rax + adc %rdx,$A1[0] + + xor $A0[0],$A0[0] + add $A1[1],$A0[1] + adc \$0,$A0[0] + mul $a0 # a[4]*a[0] + add %rax,$A0[1] # a[4]*a[0]+a[3]*a[1]+t[4] + mov $ai,%rax # a[3] + adc %rdx,$A0[0] + mov $A0[1],($tptr,$j) # t[4] + + mov 8($aptr,$j),$ai # a[5] + xor $A1[1],$A1[1] + add 8($tptr,$j),$A1[0] + adc \$0,$A1[1] + mul $a1 # a[4]*a[3] + add %rax,$A1[0] # a[4]*a[3]+t[5] + mov $ai,%rax + adc %rdx,$A1[1] + + xor $A0[1],$A0[1] + add $A1[0],$A0[0] + lea 16($j),$j # j++ + adc \$0,$A0[1] + mul $a0 # a[5]*a[2] + add %rax,$A0[0] # a[5]*a[2]+a[4]*a[3]+t[5] + mov $ai,%rax + adc %rdx,$A0[1] + mov $A0[0],-8($tptr,$j) # t[5], "preloaded t[1]" below + + cmp \$0,$j + jne .Lsqr4x_inner + + xor $A1[0],$A1[0] + add $A0[1],$A1[1] + adc \$0,$A1[0] + mul $a1 # a[5]*a[3] + add %rax,$A1[1] + adc %rdx,$A1[0] + + mov $A1[1],($tptr) # t[6], "preloaded t[2]" below + mov $A1[0],8($tptr) # t[7], "preloaded t[3]" below + + add \$16,$i + jnz .Lsqr4x_outer + + # comments apply to $num==4 case + mov -32($aptr),$a0 # a[0] + lea 64(%rsp,$num,2),$tptr # end of tp[] buffer, &tp[2*$num] + mov -24($aptr),%rax # a[1] + lea -32($tptr,$i),$tptr # end of tp[] window, &tp[2*$num-"$i"] + mov -16($aptr),$ai # a[2] + mov %rax,$a1 + + xor $A0[1],$A0[1] + mul $a0 # a[1]*a[0] + add %rax,$A0[0] # a[1]*a[0]+t[1], preloaded t[1] + mov $ai,%rax # a[2] + adc %rdx,$A0[1] + mov $A0[0],-24($tptr) # t[1] + + xor $A0[0],$A0[0] + add $A1[1],$A0[1] # a[2]*a[0]+t[2], preloaded t[2] + adc \$0,$A0[0] + mul $a0 # a[2]*a[0] + add %rax,$A0[1] + mov $ai,%rax + adc %rdx,$A0[0] + mov $A0[1],-16($tptr) # t[2] + + mov -8($aptr),$ai # a[3] + mul $a1 # a[2]*a[1] + add %rax,$A1[0] # a[2]*a[1]+t[3], preloaded t[3] + mov $ai,%rax + adc \$0,%rdx + + xor $A0[1],$A0[1] + add $A1[0],$A0[0] + mov %rdx,$A1[1] + adc \$0,$A0[1] + mul $a0 # a[3]*a[0] + add %rax,$A0[0] # a[3]*a[0]+a[2]*a[1]+t[3] + mov $ai,%rax + adc %rdx,$A0[1] + mov $A0[0],-8($tptr) # t[3] + + xor $A1[0],$A1[0] + add $A0[1],$A1[1] + adc \$0,$A1[0] + mul $a1 # a[3]*a[1] + add %rax,$A1[1] + mov -16($aptr),%rax # a[2] + adc %rdx,$A1[0] + + mov $A1[1],($tptr) # t[4] + mov $A1[0],8($tptr) # t[5] + + mul $ai # a[2]*a[3] +___ +{ +my ($shift,$carry)=($a0,$a1); +my @S=(@A1,$ai,$n0); +$code.=<<___; + add \$16,$i + xor $shift,$shift + sub $num,$i # $i=16-$num + xor $carry,$carry + + add $A1[0],%rax # t[5] + adc \$0,%rdx + mov %rax,8($tptr) # t[5] + mov %rdx,16($tptr) # t[6] + mov $carry,24($tptr) # t[7] + + mov -16($aptr,$i),%rax # a[0] + lea 64(%rsp,$num,2),$tptr + xor $A0[0],$A0[0] # t[0] + mov -24($tptr,$i,2),$A0[1] # t[1] + + lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift + shr \$63,$A0[0] + lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 | + shr \$63,$A0[1] + or $A0[0],$S[1] # | t[2*i]>>63 + mov -16($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch + mov $A0[1],$shift # shift=t[2*i+1]>>63 + mul %rax # a[i]*a[i] + neg $carry # mov $carry,cf + mov -8($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch + adc %rax,$S[0] + mov -8($aptr,$i),%rax # a[i+1] # prefetch + mov $S[0],-32($tptr,$i,2) + adc %rdx,$S[1] + + lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1 | shift + mov $S[1],-24($tptr,$i,2) + sbb $carry,$carry # mov cf,$carry + shr \$63,$A0[0] + lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 | + shr \$63,$A0[1] + or $A0[0],$S[3] # | t[2*i]>>63 + mov 0($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch + mov $A0[1],$shift # shift=t[2*i+1]>>63 + mul %rax # a[i]*a[i] + neg $carry # mov $carry,cf + mov 8($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch + adc %rax,$S[2] + mov 0($aptr,$i),%rax # a[i+1] # prefetch + mov $S[2],-16($tptr,$i,2) + adc %rdx,$S[3] + lea 16($i),$i + mov $S[3],-40($tptr,$i,2) + sbb $carry,$carry # mov cf,$carry + jmp .Lsqr4x_shift_n_add + +.align 16 +.Lsqr4x_shift_n_add: + lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift + shr \$63,$A0[0] + lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 | + shr \$63,$A0[1] + or $A0[0],$S[1] # | t[2*i]>>63 + mov -16($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch + mov $A0[1],$shift # shift=t[2*i+1]>>63 + mul %rax # a[i]*a[i] + neg $carry # mov $carry,cf + mov -8($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch + adc %rax,$S[0] + mov -8($aptr,$i),%rax # a[i+1] # prefetch + mov $S[0],-32($tptr,$i,2) + adc %rdx,$S[1] + + lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1 | shift + mov $S[1],-24($tptr,$i,2) + sbb $carry,$carry # mov cf,$carry + shr \$63,$A0[0] + lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 | + shr \$63,$A0[1] + or $A0[0],$S[3] # | t[2*i]>>63 + mov 0($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch + mov $A0[1],$shift # shift=t[2*i+1]>>63 + mul %rax # a[i]*a[i] + neg $carry # mov $carry,cf + mov 8($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch + adc %rax,$S[2] + mov 0($aptr,$i),%rax # a[i+1] # prefetch + mov $S[2],-16($tptr,$i,2) + adc %rdx,$S[3] + + lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift + mov $S[3],-8($tptr,$i,2) + sbb $carry,$carry # mov cf,$carry + shr \$63,$A0[0] + lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 | + shr \$63,$A0[1] + or $A0[0],$S[1] # | t[2*i]>>63 + mov 16($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch + mov $A0[1],$shift # shift=t[2*i+1]>>63 + mul %rax # a[i]*a[i] + neg $carry # mov $carry,cf + mov 24($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch + adc %rax,$S[0] + mov 8($aptr,$i),%rax # a[i+1] # prefetch + mov $S[0],0($tptr,$i,2) + adc %rdx,$S[1] + + lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1 | shift + mov $S[1],8($tptr,$i,2) + sbb $carry,$carry # mov cf,$carry + shr \$63,$A0[0] + lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 | + shr \$63,$A0[1] + or $A0[0],$S[3] # | t[2*i]>>63 + mov 32($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch + mov $A0[1],$shift # shift=t[2*i+1]>>63 + mul %rax # a[i]*a[i] + neg $carry # mov $carry,cf + mov 40($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch + adc %rax,$S[2] + mov 16($aptr,$i),%rax # a[i+1] # prefetch + mov $S[2],16($tptr,$i,2) + adc %rdx,$S[3] + mov $S[3],24($tptr,$i,2) + sbb $carry,$carry # mov cf,$carry + add \$32,$i + jnz .Lsqr4x_shift_n_add + + lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift + shr \$63,$A0[0] + lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 | + shr \$63,$A0[1] + or $A0[0],$S[1] # | t[2*i]>>63 + mov -16($tptr),$A0[0] # t[2*i+2] # prefetch + mov $A0[1],$shift # shift=t[2*i+1]>>63 + mul %rax # a[i]*a[i] + neg $carry # mov $carry,cf + mov -8($tptr),$A0[1] # t[2*i+2+1] # prefetch + adc %rax,$S[0] + mov -8($aptr),%rax # a[i+1] # prefetch + mov $S[0],-32($tptr) + adc %rdx,$S[1] + + lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1|shift + mov $S[1],-24($tptr) + sbb $carry,$carry # mov cf,$carry + shr \$63,$A0[0] + lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 | + shr \$63,$A0[1] + or $A0[0],$S[3] # | t[2*i]>>63 + mul %rax # a[i]*a[i] + neg $carry # mov $carry,cf + adc %rax,$S[2] + adc %rdx,$S[3] + mov $S[2],-16($tptr) + mov $S[3],-8($tptr) +___ +} +############################################################## +# Montgomery reduction part, "word-by-word" algorithm. +# +{ +my ($topbit,$nptr)=("%rbp",$aptr); +my ($m0,$m1)=($a0,$a1); +my @Ni=("%rbx","%r9"); +$code.=<<___; + mov 40(%rsp),$nptr # restore $nptr + mov 48(%rsp),$n0 # restore *n0 + xor $j,$j + mov $num,0(%rsp) # save $num + sub $num,$j # $j=-$num + mov 64(%rsp),$A0[0] # t[0] # modsched # + mov $n0,$m0 # # modsched # + lea 64(%rsp,$num,2),%rax # end of t[] buffer + lea 64(%rsp,$num),$tptr # end of t[] window + mov %rax,8(%rsp) # save end of t[] buffer + lea ($nptr,$num),$nptr # end of n[] buffer + xor $topbit,$topbit # $topbit=0 + + mov 0($nptr,$j),%rax # n[0] # modsched # + mov 8($nptr,$j),$Ni[1] # n[1] # modsched # + imulq $A0[0],$m0 # m0=t[0]*n0 # modsched # + mov %rax,$Ni[0] # # modsched # + jmp .Lsqr4x_mont_outer + +.align 16 +.Lsqr4x_mont_outer: + xor $A0[1],$A0[1] + mul $m0 # n[0]*m0 + add %rax,$A0[0] # n[0]*m0+t[0] + mov $Ni[1],%rax + adc %rdx,$A0[1] + mov $n0,$m1 + + xor $A0[0],$A0[0] + add 8($tptr,$j),$A0[1] + adc \$0,$A0[0] + mul $m0 # n[1]*m0 + add %rax,$A0[1] # n[1]*m0+t[1] + mov $Ni[0],%rax + adc %rdx,$A0[0] + + imulq $A0[1],$m1 + + mov 16($nptr,$j),$Ni[0] # n[2] + xor $A1[1],$A1[1] + add $A0[1],$A1[0] + adc \$0,$A1[1] + mul $m1 # n[0]*m1 + add %rax,$A1[0] # n[0]*m1+"t[1]" + mov $Ni[0],%rax + adc %rdx,$A1[1] + mov $A1[0],8($tptr,$j) # "t[1]" + + xor $A0[1],$A0[1] + add 16($tptr,$j),$A0[0] + adc \$0,$A0[1] + mul $m0 # n[2]*m0 + add %rax,$A0[0] # n[2]*m0+t[2] + mov $Ni[1],%rax + adc %rdx,$A0[1] + + mov 24($nptr,$j),$Ni[1] # n[3] + xor $A1[0],$A1[0] + add $A0[0],$A1[1] + adc \$0,$A1[0] + mul $m1 # n[1]*m1 + add %rax,$A1[1] # n[1]*m1+"t[2]" + mov $Ni[1],%rax + adc %rdx,$A1[0] + mov $A1[1],16($tptr,$j) # "t[2]" + + xor $A0[0],$A0[0] + add 24($tptr,$j),$A0[1] + lea 32($j),$j + adc \$0,$A0[0] + mul $m0 # n[3]*m0 + add %rax,$A0[1] # n[3]*m0+t[3] + mov $Ni[0],%rax + adc %rdx,$A0[0] + jmp .Lsqr4x_mont_inner + +.align 16 +.Lsqr4x_mont_inner: + mov ($nptr,$j),$Ni[0] # n[4] + xor $A1[1],$A1[1] + add $A0[1],$A1[0] + adc \$0,$A1[1] + mul $m1 # n[2]*m1 + add %rax,$A1[0] # n[2]*m1+"t[3]" + mov $Ni[0],%rax + adc %rdx,$A1[1] + mov $A1[0],-8($tptr,$j) # "t[3]" + + xor $A0[1],$A0[1] + add ($tptr,$j),$A0[0] + adc \$0,$A0[1] + mul $m0 # n[4]*m0 + add %rax,$A0[0] # n[4]*m0+t[4] + mov $Ni[1],%rax + adc %rdx,$A0[1] + + mov 8($nptr,$j),$Ni[1] # n[5] + xor $A1[0],$A1[0] + add $A0[0],$A1[1] + adc \$0,$A1[0] + mul $m1 # n[3]*m1 + add %rax,$A1[1] # n[3]*m1+"t[4]" + mov $Ni[1],%rax + adc %rdx,$A1[0] + mov $A1[1],($tptr,$j) # "t[4]" + + xor $A0[0],$A0[0] + add 8($tptr,$j),$A0[1] + adc \$0,$A0[0] + mul $m0 # n[5]*m0 + add %rax,$A0[1] # n[5]*m0+t[5] + mov $Ni[0],%rax + adc %rdx,$A0[0] + + + mov 16($nptr,$j),$Ni[0] # n[6] + xor $A1[1],$A1[1] + add $A0[1],$A1[0] + adc \$0,$A1[1] + mul $m1 # n[4]*m1 + add %rax,$A1[0] # n[4]*m1+"t[5]" + mov $Ni[0],%rax + adc %rdx,$A1[1] + mov $A1[0],8($tptr,$j) # "t[5]" + + xor $A0[1],$A0[1] + add 16($tptr,$j),$A0[0] + adc \$0,$A0[1] + mul $m0 # n[6]*m0 + add %rax,$A0[0] # n[6]*m0+t[6] + mov $Ni[1],%rax + adc %rdx,$A0[1] + + mov 24($nptr,$j),$Ni[1] # n[7] + xor $A1[0],$A1[0] + add $A0[0],$A1[1] + adc \$0,$A1[0] + mul $m1 # n[5]*m1 + add %rax,$A1[1] # n[5]*m1+"t[6]" + mov $Ni[1],%rax + adc %rdx,$A1[0] + mov $A1[1],16($tptr,$j) # "t[6]" + + xor $A0[0],$A0[0] + add 24($tptr,$j),$A0[1] + lea 32($j),$j + adc \$0,$A0[0] + mul $m0 # n[7]*m0 + add %rax,$A0[1] # n[7]*m0+t[7] + mov $Ni[0],%rax + adc %rdx,$A0[0] + cmp \$0,$j + jne .Lsqr4x_mont_inner + + sub 0(%rsp),$j # $j=-$num # modsched # + mov $n0,$m0 # # modsched # + + xor $A1[1],$A1[1] + add $A0[1],$A1[0] + adc \$0,$A1[1] + mul $m1 # n[6]*m1 + add %rax,$A1[0] # n[6]*m1+"t[7]" + mov $Ni[1],%rax + adc %rdx,$A1[1] + mov $A1[0],-8($tptr) # "t[7]" + + xor $A0[1],$A0[1] + add ($tptr),$A0[0] # +t[8] + adc \$0,$A0[1] + mov 0($nptr,$j),$Ni[0] # n[0] # modsched # + add $topbit,$A0[0] + adc \$0,$A0[1] + + imulq 16($tptr,$j),$m0 # m0=t[0]*n0 # modsched # + xor $A1[0],$A1[0] + mov 8($nptr,$j),$Ni[1] # n[1] # modsched # + add $A0[0],$A1[1] + mov 16($tptr,$j),$A0[0] # t[0] # modsched # + adc \$0,$A1[0] + mul $m1 # n[7]*m1 + add %rax,$A1[1] # n[7]*m1+"t[8]" + mov $Ni[0],%rax # # modsched # + adc %rdx,$A1[0] + mov $A1[1],($tptr) # "t[8]" + + xor $topbit,$topbit + add 8($tptr),$A1[0] # +t[9] + adc $topbit,$topbit + add $A0[1],$A1[0] + lea 16($tptr),$tptr # "t[$num]>>128" + adc \$0,$topbit + mov $A1[0],-8($tptr) # "t[9]" + cmp 8(%rsp),$tptr # are we done? + jb .Lsqr4x_mont_outer + + mov 0(%rsp),$num # restore $num + mov $topbit,($tptr) # save $topbit +___ +} +############################################################## +# Post-condition, 4x unrolled copy from bn_mul_mont +# +{ +my ($tptr,$nptr)=("%rbx",$aptr); +my @ri=("%rax","%rdx","%r10","%r11"); +$code.=<<___; + mov 64(%rsp,$num),@ri[0] # tp[0] + lea 64(%rsp,$num),$tptr # upper half of t[2*$num] holds result + mov 40(%rsp),$nptr # restore $nptr + shr \$5,$num # num/4 + mov 8($tptr),@ri[1] # t[1] + xor $i,$i # i=0 and clear CF! + + mov 32(%rsp),$rptr # restore $rptr + sub 0($nptr),@ri[0] + mov 16($tptr),@ri[2] # t[2] + mov 24($tptr),@ri[3] # t[3] + sbb 8($nptr),@ri[1] + lea -1($num),$j # j=num/4-1 + jmp .Lsqr4x_sub +.align 16 +.Lsqr4x_sub: + mov @ri[0],0($rptr,$i,8) # rp[i]=tp[i]-np[i] + mov @ri[1],8($rptr,$i,8) # rp[i]=tp[i]-np[i] + sbb 16($nptr,$i,8),@ri[2] + mov 32($tptr,$i,8),@ri[0] # tp[i+1] + mov 40($tptr,$i,8),@ri[1] + sbb 24($nptr,$i,8),@ri[3] + mov @ri[2],16($rptr,$i,8) # rp[i]=tp[i]-np[i] + mov @ri[3],24($rptr,$i,8) # rp[i]=tp[i]-np[i] + sbb 32($nptr,$i,8),@ri[0] + mov 48($tptr,$i,8),@ri[2] + mov 56($tptr,$i,8),@ri[3] + sbb 40($nptr,$i,8),@ri[1] + lea 4($i),$i # i++ + dec $j # doesn't affect CF! + jnz .Lsqr4x_sub + + mov @ri[0],0($rptr,$i,8) # rp[i]=tp[i]-np[i] + mov 32($tptr,$i,8),@ri[0] # load overflow bit + sbb 16($nptr,$i,8),@ri[2] + mov @ri[1],8($rptr,$i,8) # rp[i]=tp[i]-np[i] + sbb 24($nptr,$i,8),@ri[3] + mov @ri[2],16($rptr,$i,8) # rp[i]=tp[i]-np[i] + + sbb \$0,@ri[0] # handle upmost overflow bit + mov @ri[3],24($rptr,$i,8) # rp[i]=tp[i]-np[i] + xor $i,$i # i=0 + and @ri[0],$tptr + not @ri[0] + mov $rptr,$nptr + and @ri[0],$nptr + lea -1($num),$j + or $nptr,$tptr # tp=borrow?tp:rp + + pxor %xmm0,%xmm0 + lea 64(%rsp,$num,8),$nptr + movdqu ($tptr),%xmm1 + lea ($nptr,$num,8),$nptr + movdqa %xmm0,64(%rsp) # zap lower half of temporary vector + movdqa %xmm0,($nptr) # zap upper half of temporary vector + movdqu %xmm1,($rptr) + jmp .Lsqr4x_copy +.align 16 +.Lsqr4x_copy: # copy or in-place refresh + movdqu 16($tptr,$i),%xmm2 + movdqu 32($tptr,$i),%xmm1 + movdqa %xmm0,80(%rsp,$i) # zap lower half of temporary vector + movdqa %xmm0,96(%rsp,$i) # zap lower half of temporary vector + movdqa %xmm0,16($nptr,$i) # zap upper half of temporary vector + movdqa %xmm0,32($nptr,$i) # zap upper half of temporary vector + movdqu %xmm2,16($rptr,$i) + movdqu %xmm1,32($rptr,$i) + lea 32($i),$i + dec $j + jnz .Lsqr4x_copy + + movdqu 16($tptr,$i),%xmm2 + movdqa %xmm0,80(%rsp,$i) # zap lower half of temporary vector + movdqa %xmm0,16($nptr,$i) # zap upper half of temporary vector + movdqu %xmm2,16($rptr,$i) +___ +} +$code.=<<___; + mov 56(%rsp),%rsi # restore %rsp + mov \$1,%rax + mov 0(%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lsqr4x_epilogue: + ret +.size bn_sqr4x_mont,.-bn_sqr4x_mont +___ +}}} +$code.=<<___; +.asciz "Montgomery Multiplication for x86_64, CRYPTOGAMS by " +.align 16 +___ + +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +if ($win64) { +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type mul_handler,\@abi-omnipotent +.align 16 +mul_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # end of prologue label + cmp %r10,%rbx # context->RipRsp + + mov 4(%r11),%r10d # HandlerData[1] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lcommon_seh_tail + + mov 192($context),%r10 # pull $num + mov 8(%rax,%r10,8),%rax # pull saved stack pointer + lea 48(%rax),%rax + + mov -8(%rax),%rbx + mov -16(%rax),%rbp + mov -24(%rax),%r12 + mov -32(%rax),%r13 + mov -40(%rax),%r14 + mov -48(%rax),%r15 + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + + jmp .Lcommon_seh_tail +.size mul_handler,.-mul_handler + +.type sqr_handler,\@abi-omnipotent +.align 16 +sqr_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + lea .Lsqr4x_body(%rip),%r10 + cmp %r10,%rbx # context->Rip<.Lsqr_body + jb .Lcommon_seh_tail + + mov 152($context),%rax # pull context->Rsp + + lea .Lsqr4x_epilogue(%rip),%r10 + cmp %r10,%rbx # context->Rip>=.Lsqr_epilogue + jae .Lcommon_seh_tail + + mov 56(%rax),%rax # pull saved stack pointer + lea 48(%rax),%rax + + mov -8(%rax),%rbx + mov -16(%rax),%rbp + mov -24(%rax),%r12 + mov -32(%rax),%r13 + mov -40(%rax),%r14 + mov -48(%rax),%r15 + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + +.Lcommon_seh_tail: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$154,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size sqr_handler,.-sqr_handler + +.section .pdata +.align 4 + .rva .LSEH_begin_bn_mul_mont + .rva .LSEH_end_bn_mul_mont + .rva .LSEH_info_bn_mul_mont + + .rva .LSEH_begin_bn_mul4x_mont + .rva .LSEH_end_bn_mul4x_mont + .rva .LSEH_info_bn_mul4x_mont + + .rva .LSEH_begin_bn_sqr4x_mont + .rva .LSEH_end_bn_sqr4x_mont + .rva .LSEH_info_bn_sqr4x_mont + +.section .xdata +.align 8 +.LSEH_info_bn_mul_mont: + .byte 9,0,0,0 + .rva mul_handler + .rva .Lmul_body,.Lmul_epilogue # HandlerData[] +.LSEH_info_bn_mul4x_mont: + .byte 9,0,0,0 + .rva mul_handler + .rva .Lmul4x_body,.Lmul4x_epilogue # HandlerData[] +.LSEH_info_bn_sqr4x_mont: + .byte 9,0,0,0 + .rva sqr_handler +___ +} + +print $code; +close STDOUT; diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont5.S b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont5.S new file mode 100644 index 00000000..49ec6ac6 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont5.S @@ -0,0 +1,784 @@ +.text + +.globl bn_mul_mont_gather5 +.type bn_mul_mont_gather5,@function +.align 64 +bn_mul_mont_gather5: + testl $3,%r9d + jnz .Lmul_enter + cmpl $8,%r9d + jb .Lmul_enter + jmp .Lmul4x_enter + +.align 16 +.Lmul_enter: + movl %r9d,%r9d + movl 8(%rsp),%r10d + pushq %rbx + pushq %rbp + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + movq %rsp,%rax + leaq 2(%r9),%r11 + negq %r11 + leaq (%rsp,%r11,8),%rsp + andq $-1024,%rsp + + movq %rax,8(%rsp,%r9,8) +.Lmul_body: + movq %rdx,%r12 + movq %r10,%r11 + shrq $3,%r10 + andq $7,%r11 + notq %r10 + leaq .Lmagic_masks(%rip),%rax + andq $3,%r10 + leaq 96(%r12,%r11,8),%r12 + movq 0(%rax,%r10,8),%xmm4 + movq 8(%rax,%r10,8),%xmm5 + movq 16(%rax,%r10,8),%xmm6 + movq 24(%rax,%r10,8),%xmm7 + + movq -96(%r12),%xmm0 + movq -32(%r12),%xmm1 + pand %xmm4,%xmm0 + movq 32(%r12),%xmm2 + pand %xmm5,%xmm1 + movq 96(%r12),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + por %xmm2,%xmm0 + leaq 256(%r12),%r12 + por %xmm3,%xmm0 + +.byte 102,72,15,126,195 + + movq (%r8),%r8 + movq (%rsi),%rax + + xorq %r14,%r14 + xorq %r15,%r15 + + movq -96(%r12),%xmm0 + movq -32(%r12),%xmm1 + pand %xmm4,%xmm0 + movq 32(%r12),%xmm2 + pand %xmm5,%xmm1 + + movq %r8,%rbp + mulq %rbx + movq %rax,%r10 + movq (%rcx),%rax + + movq 96(%r12),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + + imulq %r10,%rbp + movq %rdx,%r11 + + por %xmm2,%xmm0 + leaq 256(%r12),%r12 + por %xmm3,%xmm0 + + mulq %rbp + addq %rax,%r10 + movq 8(%rsi),%rax + adcq $0,%rdx + movq %rdx,%r13 + + leaq 1(%r15),%r15 + jmp .L1st_enter + +.align 16 +.L1st: + addq %rax,%r13 + movq (%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%r13 + movq %r10,%r11 + adcq $0,%rdx + movq %r13,-16(%rsp,%r15,8) + movq %rdx,%r13 + +.L1st_enter: + mulq %rbx + addq %rax,%r11 + movq (%rcx,%r15,8),%rax + adcq $0,%rdx + leaq 1(%r15),%r15 + movq %rdx,%r10 + + mulq %rbp + cmpq %r9,%r15 + jne .L1st + +.byte 102,72,15,126,195 + + addq %rax,%r13 + movq (%rsi),%rax + adcq $0,%rdx + addq %r11,%r13 + adcq $0,%rdx + movq %r13,-16(%rsp,%r15,8) + movq %rdx,%r13 + movq %r10,%r11 + + xorq %rdx,%rdx + addq %r11,%r13 + adcq $0,%rdx + movq %r13,-8(%rsp,%r9,8) + movq %rdx,(%rsp,%r9,8) + + leaq 1(%r14),%r14 + jmp .Louter +.align 16 +.Louter: + xorq %r15,%r15 + movq %r8,%rbp + movq (%rsp),%r10 + + movq -96(%r12),%xmm0 + movq -32(%r12),%xmm1 + pand %xmm4,%xmm0 + movq 32(%r12),%xmm2 + pand %xmm5,%xmm1 + + mulq %rbx + addq %rax,%r10 + movq (%rcx),%rax + adcq $0,%rdx + + movq 96(%r12),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + + imulq %r10,%rbp + movq %rdx,%r11 + + por %xmm2,%xmm0 + leaq 256(%r12),%r12 + por %xmm3,%xmm0 + + mulq %rbp + addq %rax,%r10 + movq 8(%rsi),%rax + adcq $0,%rdx + movq 8(%rsp),%r10 + movq %rdx,%r13 + + leaq 1(%r15),%r15 + jmp .Linner_enter + +.align 16 +.Linner: + addq %rax,%r13 + movq (%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + movq (%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %r13,-16(%rsp,%r15,8) + movq %rdx,%r13 + +.Linner_enter: + mulq %rbx + addq %rax,%r11 + movq (%rcx,%r15,8),%rax + adcq $0,%rdx + addq %r11,%r10 + movq %rdx,%r11 + adcq $0,%r11 + leaq 1(%r15),%r15 + + mulq %rbp + cmpq %r9,%r15 + jne .Linner + +.byte 102,72,15,126,195 + + addq %rax,%r13 + movq (%rsi),%rax + adcq $0,%rdx + addq %r10,%r13 + movq (%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %r13,-16(%rsp,%r15,8) + movq %rdx,%r13 + + xorq %rdx,%rdx + addq %r11,%r13 + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-8(%rsp,%r9,8) + movq %rdx,(%rsp,%r9,8) + + leaq 1(%r14),%r14 + cmpq %r9,%r14 + jl .Louter + + xorq %r14,%r14 + movq (%rsp),%rax + leaq (%rsp),%rsi + movq %r9,%r15 + jmp .Lsub +.align 16 +.Lsub: sbbq (%rcx,%r14,8),%rax + movq %rax,(%rdi,%r14,8) + movq 8(%rsi,%r14,8),%rax + leaq 1(%r14),%r14 + decq %r15 + jnz .Lsub + + sbbq $0,%rax + xorq %r14,%r14 + andq %rax,%rsi + notq %rax + movq %rdi,%rcx + andq %rax,%rcx + movq %r9,%r15 + orq %rcx,%rsi +.align 16 +.Lcopy: + movq (%rsi,%r14,8),%rax + movq %r14,(%rsp,%r14,8) + movq %rax,(%rdi,%r14,8) + leaq 1(%r14),%r14 + subq $1,%r15 + jnz .Lcopy + + movq 8(%rsp,%r9,8),%rsi + movq $1,%rax + movq (%rsi),%r15 + movq 8(%rsi),%r14 + movq 16(%rsi),%r13 + movq 24(%rsi),%r12 + movq 32(%rsi),%rbp + movq 40(%rsi),%rbx + leaq 48(%rsi),%rsp +.Lmul_epilogue: + .byte 0xf3,0xc3 +.size bn_mul_mont_gather5,.-bn_mul_mont_gather5 +.type bn_mul4x_mont_gather5,@function +.align 16 +bn_mul4x_mont_gather5: +.Lmul4x_enter: + movl %r9d,%r9d + movl 8(%rsp),%r10d + pushq %rbx + pushq %rbp + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + movq %rsp,%rax + leaq 4(%r9),%r11 + negq %r11 + leaq (%rsp,%r11,8),%rsp + andq $-1024,%rsp + + movq %rax,8(%rsp,%r9,8) +.Lmul4x_body: + movq %rdi,16(%rsp,%r9,8) + movq %rdx,%r12 + movq %r10,%r11 + shrq $3,%r10 + andq $7,%r11 + notq %r10 + leaq .Lmagic_masks(%rip),%rax + andq $3,%r10 + leaq 96(%r12,%r11,8),%r12 + movq 0(%rax,%r10,8),%xmm4 + movq 8(%rax,%r10,8),%xmm5 + movq 16(%rax,%r10,8),%xmm6 + movq 24(%rax,%r10,8),%xmm7 + + movq -96(%r12),%xmm0 + movq -32(%r12),%xmm1 + pand %xmm4,%xmm0 + movq 32(%r12),%xmm2 + pand %xmm5,%xmm1 + movq 96(%r12),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + por %xmm2,%xmm0 + leaq 256(%r12),%r12 + por %xmm3,%xmm0 + +.byte 102,72,15,126,195 + movq (%r8),%r8 + movq (%rsi),%rax + + xorq %r14,%r14 + xorq %r15,%r15 + + movq -96(%r12),%xmm0 + movq -32(%r12),%xmm1 + pand %xmm4,%xmm0 + movq 32(%r12),%xmm2 + pand %xmm5,%xmm1 + + movq %r8,%rbp + mulq %rbx + movq %rax,%r10 + movq (%rcx),%rax + + movq 96(%r12),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + + imulq %r10,%rbp + movq %rdx,%r11 + + por %xmm2,%xmm0 + leaq 256(%r12),%r12 + por %xmm3,%xmm0 + + mulq %rbp + addq %rax,%r10 + movq 8(%rsi),%rax + adcq $0,%rdx + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq 8(%rcx),%rax + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq 16(%rsi),%rax + adcq $0,%rdx + addq %r11,%rdi + leaq 4(%r15),%r15 + adcq $0,%rdx + movq %rdi,(%rsp) + movq %rdx,%r13 + jmp .L1st4x +.align 16 +.L1st4x: + mulq %rbx + addq %rax,%r10 + movq -16(%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq -8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-24(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq -8(%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq (%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-16(%rsp,%r15,8) + movq %rdx,%r13 + + mulq %rbx + addq %rax,%r10 + movq (%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq 8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-8(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq 8(%rcx,%r15,8),%rax + adcq $0,%rdx + leaq 4(%r15),%r15 + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq -16(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-32(%rsp,%r15,8) + movq %rdx,%r13 + cmpq %r9,%r15 + jl .L1st4x + + mulq %rbx + addq %rax,%r10 + movq -16(%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq -8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %r13,-24(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq -8(%rcx,%r15,8),%rax + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq (%rsi),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %rdi,-16(%rsp,%r15,8) + movq %rdx,%r13 + +.byte 102,72,15,126,195 + + xorq %rdi,%rdi + addq %r10,%r13 + adcq $0,%rdi + movq %r13,-8(%rsp,%r15,8) + movq %rdi,(%rsp,%r15,8) + + leaq 1(%r14),%r14 +.align 4 +.Louter4x: + xorq %r15,%r15 + movq -96(%r12),%xmm0 + movq -32(%r12),%xmm1 + pand %xmm4,%xmm0 + movq 32(%r12),%xmm2 + pand %xmm5,%xmm1 + + movq (%rsp),%r10 + movq %r8,%rbp + mulq %rbx + addq %rax,%r10 + movq (%rcx),%rax + adcq $0,%rdx + + movq 96(%r12),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + + imulq %r10,%rbp + movq %rdx,%r11 + + por %xmm2,%xmm0 + leaq 256(%r12),%r12 + por %xmm3,%xmm0 + + mulq %rbp + addq %rax,%r10 + movq 8(%rsi),%rax + adcq $0,%rdx + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq 8(%rcx),%rax + adcq $0,%rdx + addq 8(%rsp),%r11 + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq 16(%rsi),%rax + adcq $0,%rdx + addq %r11,%rdi + leaq 4(%r15),%r15 + adcq $0,%rdx + movq %rdx,%r13 + jmp .Linner4x +.align 16 +.Linner4x: + mulq %rbx + addq %rax,%r10 + movq -16(%rcx,%r15,8),%rax + adcq $0,%rdx + addq -16(%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq -8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %rdi,-32(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq -8(%rcx,%r15,8),%rax + adcq $0,%rdx + addq -8(%rsp,%r15,8),%r11 + adcq $0,%rdx + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq (%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %r13,-24(%rsp,%r15,8) + movq %rdx,%r13 + + mulq %rbx + addq %rax,%r10 + movq (%rcx,%r15,8),%rax + adcq $0,%rdx + addq (%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq 8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %rdi,-16(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq 8(%rcx,%r15,8),%rax + adcq $0,%rdx + addq 8(%rsp,%r15,8),%r11 + adcq $0,%rdx + leaq 4(%r15),%r15 + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq -16(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %r13,-40(%rsp,%r15,8) + movq %rdx,%r13 + cmpq %r9,%r15 + jl .Linner4x + + mulq %rbx + addq %rax,%r10 + movq -16(%rcx,%r15,8),%rax + adcq $0,%rdx + addq -16(%rsp,%r15,8),%r10 + adcq $0,%rdx + movq %rdx,%r11 + + mulq %rbp + addq %rax,%r13 + movq -8(%rsi,%r15,8),%rax + adcq $0,%rdx + addq %r10,%r13 + adcq $0,%rdx + movq %rdi,-32(%rsp,%r15,8) + movq %rdx,%rdi + + mulq %rbx + addq %rax,%r11 + movq -8(%rcx,%r15,8),%rax + adcq $0,%rdx + addq -8(%rsp,%r15,8),%r11 + adcq $0,%rdx + leaq 1(%r14),%r14 + movq %rdx,%r10 + + mulq %rbp + addq %rax,%rdi + movq (%rsi),%rax + adcq $0,%rdx + addq %r11,%rdi + adcq $0,%rdx + movq %r13,-24(%rsp,%r15,8) + movq %rdx,%r13 + +.byte 102,72,15,126,195 + movq %rdi,-16(%rsp,%r15,8) + + xorq %rdi,%rdi + addq %r10,%r13 + adcq $0,%rdi + addq (%rsp,%r9,8),%r13 + adcq $0,%rdi + movq %r13,-8(%rsp,%r15,8) + movq %rdi,(%rsp,%r15,8) + + cmpq %r9,%r14 + jl .Louter4x + movq 16(%rsp,%r9,8),%rdi + movq 0(%rsp),%rax + pxor %xmm0,%xmm0 + movq 8(%rsp),%rdx + shrq $2,%r9 + leaq (%rsp),%rsi + xorq %r14,%r14 + + subq 0(%rcx),%rax + movq 16(%rsi),%rbx + movq 24(%rsi),%rbp + sbbq 8(%rcx),%rdx + leaq -1(%r9),%r15 + jmp .Lsub4x +.align 16 +.Lsub4x: + movq %rax,0(%rdi,%r14,8) + movq %rdx,8(%rdi,%r14,8) + sbbq 16(%rcx,%r14,8),%rbx + movq 32(%rsi,%r14,8),%rax + movq 40(%rsi,%r14,8),%rdx + sbbq 24(%rcx,%r14,8),%rbp + movq %rbx,16(%rdi,%r14,8) + movq %rbp,24(%rdi,%r14,8) + sbbq 32(%rcx,%r14,8),%rax + movq 48(%rsi,%r14,8),%rbx + movq 56(%rsi,%r14,8),%rbp + sbbq 40(%rcx,%r14,8),%rdx + leaq 4(%r14),%r14 + decq %r15 + jnz .Lsub4x + + movq %rax,0(%rdi,%r14,8) + movq 32(%rsi,%r14,8),%rax + sbbq 16(%rcx,%r14,8),%rbx + movq %rdx,8(%rdi,%r14,8) + sbbq 24(%rcx,%r14,8),%rbp + movq %rbx,16(%rdi,%r14,8) + + sbbq $0,%rax + movq %rbp,24(%rdi,%r14,8) + xorq %r14,%r14 + andq %rax,%rsi + notq %rax + movq %rdi,%rcx + andq %rax,%rcx + leaq -1(%r9),%r15 + orq %rcx,%rsi + + movdqu (%rsi),%xmm1 + movdqa %xmm0,(%rsp) + movdqu %xmm1,(%rdi) + jmp .Lcopy4x +.align 16 +.Lcopy4x: + movdqu 16(%rsi,%r14,1),%xmm2 + movdqu 32(%rsi,%r14,1),%xmm1 + movdqa %xmm0,16(%rsp,%r14,1) + movdqu %xmm2,16(%rdi,%r14,1) + movdqa %xmm0,32(%rsp,%r14,1) + movdqu %xmm1,32(%rdi,%r14,1) + leaq 32(%r14),%r14 + decq %r15 + jnz .Lcopy4x + + shlq $2,%r9 + movdqu 16(%rsi,%r14,1),%xmm2 + movdqa %xmm0,16(%rsp,%r14,1) + movdqu %xmm2,16(%rdi,%r14,1) + movq 8(%rsp,%r9,8),%rsi + movq $1,%rax + movq (%rsi),%r15 + movq 8(%rsi),%r14 + movq 16(%rsi),%r13 + movq 24(%rsi),%r12 + movq 32(%rsi),%rbp + movq 40(%rsi),%rbx + leaq 48(%rsi),%rsp +.Lmul4x_epilogue: + .byte 0xf3,0xc3 +.size bn_mul4x_mont_gather5,.-bn_mul4x_mont_gather5 +.globl bn_scatter5 +.type bn_scatter5,@function +.align 16 +bn_scatter5: + cmpq $0,%rsi + jz .Lscatter_epilogue + leaq (%rdx,%rcx,8),%rdx +.Lscatter: + movq (%rdi),%rax + leaq 8(%rdi),%rdi + movq %rax,(%rdx) + leaq 256(%rdx),%rdx + subq $1,%rsi + jnz .Lscatter +.Lscatter_epilogue: + .byte 0xf3,0xc3 +.size bn_scatter5,.-bn_scatter5 + +.globl bn_gather5 +.type bn_gather5,@function +.align 16 +bn_gather5: + movq %rcx,%r11 + shrq $3,%rcx + andq $7,%r11 + notq %rcx + leaq .Lmagic_masks(%rip),%rax + andq $3,%rcx + leaq 96(%rdx,%r11,8),%rdx + movq 0(%rax,%rcx,8),%xmm4 + movq 8(%rax,%rcx,8),%xmm5 + movq 16(%rax,%rcx,8),%xmm6 + movq 24(%rax,%rcx,8),%xmm7 + jmp .Lgather +.align 16 +.Lgather: + movq -96(%rdx),%xmm0 + movq -32(%rdx),%xmm1 + pand %xmm4,%xmm0 + movq 32(%rdx),%xmm2 + pand %xmm5,%xmm1 + movq 96(%rdx),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + por %xmm2,%xmm0 + leaq 256(%rdx),%rdx + por %xmm3,%xmm0 + + movq %xmm0,(%rdi) + leaq 8(%rdi),%rdi + subq $1,%rsi + jnz .Lgather + .byte 0xf3,0xc3 +.LSEH_end_bn_gather5: +.size bn_gather5,.-bn_gather5 +.align 64 +.Lmagic_masks: +.long 0,0, 0,0, 0,0, -1,-1 +.long 0,0, 0,0, 0,0, 0,0 +.byte 77,111,110,116,103,111,109,101,114,121,32,77,117,108,116,105,112,108,105,99,97,116,105,111,110,32,119,105,116,104,32,115,99,97,116,116,101,114,47,103,97,116,104,101,114,32,102,111,114,32,120,56,54,95,54,52,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0 diff --git a/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont5.pl b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont5.pl new file mode 100755 index 00000000..dae0fe24 --- /dev/null +++ b/ics-openvpn-stripped/main/openssl/crypto/bn/asm/x86_64-mont5.pl @@ -0,0 +1,1071 @@ +#!/usr/bin/env perl + +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# August 2011. +# +# Companion to x86_64-mont.pl that optimizes cache-timing attack +# countermeasures. The subroutines are produced by replacing bp[i] +# references in their x86_64-mont.pl counterparts with cache-neutral +# references to powers table computed in BN_mod_exp_mont_consttime. +# In addition subroutine that scatters elements of the powers table +# is implemented, so that scatter-/gathering can be tuned without +# bn_exp.c modifications. + +$flavour = shift; +$output = shift; +if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } + +$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open OUT,"| \"$^X\" $xlate $flavour $output"; +*STDOUT=*OUT; + +# int bn_mul_mont_gather5( +$rp="%rdi"; # BN_ULONG *rp, +$ap="%rsi"; # const BN_ULONG *ap, +$bp="%rdx"; # const BN_ULONG *bp, +$np="%rcx"; # const BN_ULONG *np, +$n0="%r8"; # const BN_ULONG *n0, +$num="%r9"; # int num, + # int idx); # 0 to 2^5-1, "index" in $bp holding + # pre-computed powers of a', interlaced + # in such manner that b[0] is $bp[idx], + # b[1] is [2^5+idx], etc. +$lo0="%r10"; +$hi0="%r11"; +$hi1="%r13"; +$i="%r14"; +$j="%r15"; +$m0="%rbx"; +$m1="%rbp"; + +$code=<<___; +.text + +.globl bn_mul_mont_gather5 +.type bn_mul_mont_gather5,\@function,6 +.align 64 +bn_mul_mont_gather5: + test \$3,${num}d + jnz .Lmul_enter + cmp \$8,${num}d + jb .Lmul_enter + jmp .Lmul4x_enter + +.align 16 +.Lmul_enter: + mov ${num}d,${num}d + mov `($win64?56:8)`(%rsp),%r10d # load 7th argument + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 +___ +$code.=<<___ if ($win64); + lea -0x28(%rsp),%rsp + movaps %xmm6,(%rsp) + movaps %xmm7,0x10(%rsp) +.Lmul_alloca: +___ +$code.=<<___; + mov %rsp,%rax + lea 2($num),%r11 + neg %r11 + lea (%rsp,%r11,8),%rsp # tp=alloca(8*(num+2)) + and \$-1024,%rsp # minimize TLB usage + + mov %rax,8(%rsp,$num,8) # tp[num+1]=%rsp +.Lmul_body: + mov $bp,%r12 # reassign $bp +___ + $bp="%r12"; + $STRIDE=2**5*8; # 5 is "window size" + $N=$STRIDE/4; # should match cache line size +$code.=<<___; + mov %r10,%r11 + shr \$`log($N/8)/log(2)`,%r10 + and \$`$N/8-1`,%r11 + not %r10 + lea .Lmagic_masks(%rip),%rax + and \$`2**5/($N/8)-1`,%r10 # 5 is "window size" + lea 96($bp,%r11,8),$bp # pointer within 1st cache line + movq 0(%rax,%r10,8),%xmm4 # set of masks denoting which + movq 8(%rax,%r10,8),%xmm5 # cache line contains element + movq 16(%rax,%r10,8),%xmm6 # denoted by 7th argument + movq 24(%rax,%r10,8),%xmm7 + + movq `0*$STRIDE/4-96`($bp),%xmm0 + movq `1*$STRIDE/4-96`($bp),%xmm1 + pand %xmm4,%xmm0 + movq `2*$STRIDE/4-96`($bp),%xmm2 + pand %xmm5,%xmm1 + movq `3*$STRIDE/4-96`($bp),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + por %xmm2,%xmm0 + lea $STRIDE($bp),$bp + por %xmm3,%xmm0 + + movq %xmm0,$m0 # m0=bp[0] + + mov ($n0),$n0 # pull n0[0] value + mov ($ap),%rax + + xor $i,$i # i=0 + xor $j,$j # j=0 + + movq `0*$STRIDE/4-96`($bp),%xmm0 + movq `1*$STRIDE/4-96`($bp),%xmm1 + pand %xmm4,%xmm0 + movq `2*$STRIDE/4-96`($bp),%xmm2 + pand %xmm5,%xmm1 + + mov $n0,$m1 + mulq $m0 # ap[0]*bp[0] + mov %rax,$lo0 + mov ($np),%rax + + movq `3*$STRIDE/4-96`($bp),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + + imulq $lo0,$m1 # "tp[0]"*n0 + mov %rdx,$hi0 + + por %xmm2,%xmm0 + lea $STRIDE($bp),$bp + por %xmm3,%xmm0 + + mulq $m1 # np[0]*m1 + add %rax,$lo0 # discarded + mov 8($ap),%rax + adc \$0,%rdx + mov %rdx,$hi1 + + lea 1($j),$j # j++ + jmp .L1st_enter + +.align 16 +.L1st: + add %rax,$hi1 + mov ($ap,$j,8),%rax + adc \$0,%rdx + add $hi0,$hi1 # np[j]*m1+ap[j]*bp[0] + mov $lo0,$hi0 + adc \$0,%rdx + mov $hi1,-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$hi1 + +.L1st_enter: + mulq $m0 # ap[j]*bp[0] + add %rax,$hi0 + mov ($np,$j,8),%rax + adc \$0,%rdx + lea 1($j),$j # j++ + mov %rdx,$lo0 + + mulq $m1 # np[j]*m1 + cmp $num,$j + jne .L1st + + movq %xmm0,$m0 # bp[1] + + add %rax,$hi1 + mov ($ap),%rax # ap[0] + adc \$0,%rdx + add $hi0,$hi1 # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $hi1,-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$hi1 + mov $lo0,$hi0 + + xor %rdx,%rdx + add $hi0,$hi1 + adc \$0,%rdx + mov $hi1,-8(%rsp,$num,8) + mov %rdx,(%rsp,$num,8) # store upmost overflow bit + + lea 1($i),$i # i++ + jmp .Louter +.align 16 +.Louter: + xor $j,$j # j=0 + mov $n0,$m1 + mov (%rsp),$lo0 + + movq `0*$STRIDE/4-96`($bp),%xmm0 + movq `1*$STRIDE/4-96`($bp),%xmm1 + pand %xmm4,%xmm0 + movq `2*$STRIDE/4-96`($bp),%xmm2 + pand %xmm5,%xmm1 + + mulq $m0 # ap[0]*bp[i] + add %rax,$lo0 # ap[0]*bp[i]+tp[0] + mov ($np),%rax + adc \$0,%rdx + + movq `3*$STRIDE/4-96`($bp),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + + imulq $lo0,$m1 # tp[0]*n0 + mov %rdx,$hi0 + + por %xmm2,%xmm0 + lea $STRIDE($bp),$bp + por %xmm3,%xmm0 + + mulq $m1 # np[0]*m1 + add %rax,$lo0 # discarded + mov 8($ap),%rax + adc \$0,%rdx + mov 8(%rsp),$lo0 # tp[1] + mov %rdx,$hi1 + + lea 1($j),$j # j++ + jmp .Linner_enter + +.align 16 +.Linner: + add %rax,$hi1 + mov ($ap,$j,8),%rax + adc \$0,%rdx + add $lo0,$hi1 # np[j]*m1+ap[j]*bp[i]+tp[j] + mov (%rsp,$j,8),$lo0 + adc \$0,%rdx + mov $hi1,-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$hi1 + +.Linner_enter: + mulq $m0 # ap[j]*bp[i] + add %rax,$hi0 + mov ($np,$j,8),%rax + adc \$0,%rdx + add $hi0,$lo0 # ap[j]*bp[i]+tp[j] + mov %rdx,$hi0 + adc \$0,$hi0 + lea 1($j),$j # j++ + + mulq $m1 # np[j]*m1 + cmp $num,$j + jne .Linner + + movq %xmm0,$m0 # bp[i+1] + + add %rax,$hi1 + mov ($ap),%rax # ap[0] + adc \$0,%rdx + add $lo0,$hi1 # np[j]*m1+ap[j]*bp[i]+tp[j] + mov (%rsp,$j,8),$lo0 + adc \$0,%rdx + mov $hi1,-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$hi1 + + xor %rdx,%rdx + add $hi0,$hi1 + adc \$0,%rdx + add $lo0,$hi1 # pull upmost overflow bit + adc \$0,%rdx + mov $hi1,-8(%rsp,$num,8) + mov %rdx,(%rsp,$num,8) # store upmost overflow bit + + lea 1($i),$i # i++ + cmp $num,$i + jl .Louter + + xor $i,$i # i=0 and clear CF! + mov (%rsp),%rax # tp[0] + lea (%rsp),$ap # borrow ap for tp + mov $num,$j # j=num + jmp .Lsub +.align 16 +.Lsub: sbb ($np,$i,8),%rax + mov %rax,($rp,$i,8) # rp[i]=tp[i]-np[i] + mov 8($ap,$i,8),%rax # tp[i+1] + lea 1($i),$i # i++ + dec $j # doesnn't affect CF! + jnz .Lsub + + sbb \$0,%rax # handle upmost overflow bit + xor $i,$i + and %rax,$ap + not %rax + mov $rp,$np + and %rax,$np + mov $num,$j # j=num + or $np,$ap # ap=borrow?tp:rp +.align 16 +.Lcopy: # copy or in-place refresh + mov ($ap,$i,8),%rax + mov $i,(%rsp,$i,8) # zap temporary vector + mov %rax,($rp,$i,8) # rp[i]=tp[i] + lea 1($i),$i + sub \$1,$j + jnz .Lcopy + + mov 8(%rsp,$num,8),%rsi # restore %rsp + mov \$1,%rax +___ +$code.=<<___ if ($win64); + movaps (%rsi),%xmm6 + movaps 0x10(%rsi),%xmm7 + lea 0x28(%rsi),%rsi +___ +$code.=<<___; + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lmul_epilogue: + ret +.size bn_mul_mont_gather5,.-bn_mul_mont_gather5 +___ +{{{ +my @A=("%r10","%r11"); +my @N=("%r13","%rdi"); +$code.=<<___; +.type bn_mul4x_mont_gather5,\@function,6 +.align 16 +bn_mul4x_mont_gather5: +.Lmul4x_enter: + mov ${num}d,${num}d + mov `($win64?56:8)`(%rsp),%r10d # load 7th argument + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 +___ +$code.=<<___ if ($win64); + lea -0x28(%rsp),%rsp + movaps %xmm6,(%rsp) + movaps %xmm7,0x10(%rsp) +.Lmul4x_alloca: +___ +$code.=<<___; + mov %rsp,%rax + lea 4($num),%r11 + neg %r11 + lea (%rsp,%r11,8),%rsp # tp=alloca(8*(num+4)) + and \$-1024,%rsp # minimize TLB usage + + mov %rax,8(%rsp,$num,8) # tp[num+1]=%rsp +.Lmul4x_body: + mov $rp,16(%rsp,$num,8) # tp[num+2]=$rp + mov %rdx,%r12 # reassign $bp +___ + $bp="%r12"; + $STRIDE=2**5*8; # 5 is "window size" + $N=$STRIDE/4; # should match cache line size +$code.=<<___; + mov %r10,%r11 + shr \$`log($N/8)/log(2)`,%r10 + and \$`$N/8-1`,%r11 + not %r10 + lea .Lmagic_masks(%rip),%rax + and \$`2**5/($N/8)-1`,%r10 # 5 is "window size" + lea 96($bp,%r11,8),$bp # pointer within 1st cache line + movq 0(%rax,%r10,8),%xmm4 # set of masks denoting which + movq 8(%rax,%r10,8),%xmm5 # cache line contains element + movq 16(%rax,%r10,8),%xmm6 # denoted by 7th argument + movq 24(%rax,%r10,8),%xmm7 + + movq `0*$STRIDE/4-96`($bp),%xmm0 + movq `1*$STRIDE/4-96`($bp),%xmm1 + pand %xmm4,%xmm0 + movq `2*$STRIDE/4-96`($bp),%xmm2 + pand %xmm5,%xmm1 + movq `3*$STRIDE/4-96`($bp),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + por %xmm2,%xmm0 + lea $STRIDE($bp),$bp + por %xmm3,%xmm0 + + movq %xmm0,$m0 # m0=bp[0] + mov ($n0),$n0 # pull n0[0] value + mov ($ap),%rax + + xor $i,$i # i=0 + xor $j,$j # j=0 + + movq `0*$STRIDE/4-96`($bp),%xmm0 + movq `1*$STRIDE/4-96`($bp),%xmm1 + pand %xmm4,%xmm0 + movq `2*$STRIDE/4-96`($bp),%xmm2 + pand %xmm5,%xmm1 + + mov $n0,$m1 + mulq $m0 # ap[0]*bp[0] + mov %rax,$A[0] + mov ($np),%rax + + movq `3*$STRIDE/4-96`($bp),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + + imulq $A[0],$m1 # "tp[0]"*n0 + mov %rdx,$A[1] + + por %xmm2,%xmm0 + lea $STRIDE($bp),$bp + por %xmm3,%xmm0 + + mulq $m1 # np[0]*m1 + add %rax,$A[0] # discarded + mov 8($ap),%rax + adc \$0,%rdx + mov %rdx,$N[1] + + mulq $m0 + add %rax,$A[1] + mov 8($np),%rax + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 + add %rax,$N[1] + mov 16($ap),%rax + adc \$0,%rdx + add $A[1],$N[1] + lea 4($j),$j # j++ + adc \$0,%rdx + mov $N[1],(%rsp) + mov %rdx,$N[0] + jmp .L1st4x +.align 16 +.L1st4x: + mulq $m0 # ap[j]*bp[0] + add %rax,$A[0] + mov -16($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov -8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[0],-24(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[1] + mov -8($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov ($ap,$j,8),%rax + adc \$0,%rdx + add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[1],-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[0] + mov ($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov 8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[0],-8(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[1] + mov 8($np,$j,8),%rax + adc \$0,%rdx + lea 4($j),$j # j++ + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov -16($ap,$j,8),%rax + adc \$0,%rdx + add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[1],-32(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + cmp $num,$j + jl .L1st4x + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[0] + mov -16($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov -8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[0],-24(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[0] + add %rax,$A[1] + mov -8($np,$j,8),%rax + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov ($ap),%rax # ap[0] + adc \$0,%rdx + add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0] + adc \$0,%rdx + mov $N[1],-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + + movq %xmm0,$m0 # bp[1] + + xor $N[1],$N[1] + add $A[0],$N[0] + adc \$0,$N[1] + mov $N[0],-8(%rsp,$j,8) + mov $N[1],(%rsp,$j,8) # store upmost overflow bit + + lea 1($i),$i # i++ +.align 4 +.Louter4x: + xor $j,$j # j=0 + movq `0*$STRIDE/4-96`($bp),%xmm0 + movq `1*$STRIDE/4-96`($bp),%xmm1 + pand %xmm4,%xmm0 + movq `2*$STRIDE/4-96`($bp),%xmm2 + pand %xmm5,%xmm1 + + mov (%rsp),$A[0] + mov $n0,$m1 + mulq $m0 # ap[0]*bp[i] + add %rax,$A[0] # ap[0]*bp[i]+tp[0] + mov ($np),%rax + adc \$0,%rdx + + movq `3*$STRIDE/4-96`($bp),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + + imulq $A[0],$m1 # tp[0]*n0 + mov %rdx,$A[1] + + por %xmm2,%xmm0 + lea $STRIDE($bp),$bp + por %xmm3,%xmm0 + + mulq $m1 # np[0]*m1 + add %rax,$A[0] # "$N[0]", discarded + mov 8($ap),%rax + adc \$0,%rdx + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[1] + mov 8($np),%rax + adc \$0,%rdx + add 8(%rsp),$A[1] # +tp[1] + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov 16($ap),%rax + adc \$0,%rdx + add $A[1],$N[1] # np[j]*m1+ap[j]*bp[i]+tp[j] + lea 4($j),$j # j+=2 + adc \$0,%rdx + mov %rdx,$N[0] + jmp .Linner4x +.align 16 +.Linner4x: + mulq $m0 # ap[j]*bp[i] + add %rax,$A[0] + mov -16($np,$j,8),%rax + adc \$0,%rdx + add -16(%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j] + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov -8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] + adc \$0,%rdx + mov $N[1],-32(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[1] + mov -8($np,$j,8),%rax + adc \$0,%rdx + add -8(%rsp,$j,8),$A[1] + adc \$0,%rdx + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov ($ap,$j,8),%rax + adc \$0,%rdx + add $A[1],$N[1] + adc \$0,%rdx + mov $N[0],-24(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[0] + mov ($np,$j,8),%rax + adc \$0,%rdx + add (%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j] + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov 8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] + adc \$0,%rdx + mov $N[1],-16(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[1] + mov 8($np,$j,8),%rax + adc \$0,%rdx + add 8(%rsp,$j,8),$A[1] + adc \$0,%rdx + lea 4($j),$j # j++ + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov -16($ap,$j,8),%rax + adc \$0,%rdx + add $A[1],$N[1] + adc \$0,%rdx + mov $N[0],-40(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + cmp $num,$j + jl .Linner4x + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[0] + mov -16($np,$j,8),%rax + adc \$0,%rdx + add -16(%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j] + adc \$0,%rdx + mov %rdx,$A[1] + + mulq $m1 # np[j]*m1 + add %rax,$N[0] + mov -8($ap,$j,8),%rax + adc \$0,%rdx + add $A[0],$N[0] + adc \$0,%rdx + mov $N[1],-32(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[1] + + mulq $m0 # ap[j]*bp[i] + add %rax,$A[1] + mov -8($np,$j,8),%rax + adc \$0,%rdx + add -8(%rsp,$j,8),$A[1] + adc \$0,%rdx + lea 1($i),$i # i++ + mov %rdx,$A[0] + + mulq $m1 # np[j]*m1 + add %rax,$N[1] + mov ($ap),%rax # ap[0] + adc \$0,%rdx + add $A[1],$N[1] + adc \$0,%rdx + mov $N[0],-24(%rsp,$j,8) # tp[j-1] + mov %rdx,$N[0] + + movq %xmm0,$m0 # bp[i+1] + mov $N[1],-16(%rsp,$j,8) # tp[j-1] + + xor $N[1],$N[1] + add $A[0],$N[0] + adc \$0,$N[1] + add (%rsp,$num,8),$N[0] # pull upmost overflow bit + adc \$0,$N[1] + mov $N[0],-8(%rsp,$j,8) + mov $N[1],(%rsp,$j,8) # store upmost overflow bit + + cmp $num,$i + jl .Louter4x +___ +{ +my @ri=("%rax","%rdx",$m0,$m1); +$code.=<<___; + mov 16(%rsp,$num,8),$rp # restore $rp + mov 0(%rsp),@ri[0] # tp[0] + pxor %xmm0,%xmm0 + mov 8(%rsp),@ri[1] # tp[1] + shr \$2,$num # num/=4 + lea (%rsp),$ap # borrow ap for tp + xor $i,$i # i=0 and clear CF! + + sub 0($np),@ri[0] + mov 16($ap),@ri[2] # tp[2] + mov 24($ap),@ri[3] # tp[3] + sbb 8($np),@ri[1] + lea -1($num),$j # j=num/4-1 + jmp .Lsub4x +.align 16 +.Lsub4x: + mov @ri[0],0($rp,$i,8) # rp[i]=tp[i]-np[i] + mov @ri[1],8($rp,$i,8) # rp[i]=tp[i]-np[i] + sbb 16($np,$i,8),@ri[2] + mov 32($ap,$i,8),@ri[0] # tp[i+1] + mov 40($ap,$i,8),@ri[1] + sbb 24($np,$i,8),@ri[3] + mov @ri[2],16($rp,$i,8) # rp[i]=tp[i]-np[i] + mov @ri[3],24($rp,$i,8) # rp[i]=tp[i]-np[i] + sbb 32($np,$i,8),@ri[0] + mov 48($ap,$i,8),@ri[2] + mov 56($ap,$i,8),@ri[3] + sbb 40($np,$i,8),@ri[1] + lea 4($i),$i # i++ + dec $j # doesnn't affect CF! + jnz .Lsub4x + + mov @ri[0],0($rp,$i,8) # rp[i]=tp[i]-np[i] + mov 32($ap,$i,8),@ri[0] # load overflow bit + sbb 16($np,$i,8),@ri[2] + mov @ri[1],8($rp,$i,8) # rp[i]=tp[i]-np[i] + sbb 24($np,$i,8),@ri[3] + mov @ri[2],16($rp,$i,8) # rp[i]=tp[i]-np[i] + + sbb \$0,@ri[0] # handle upmost overflow bit + mov @ri[3],24($rp,$i,8) # rp[i]=tp[i]-np[i] + xor $i,$i # i=0 + and @ri[0],$ap + not @ri[0] + mov $rp,$np + and @ri[0],$np + lea -1($num),$j + or $np,$ap # ap=borrow?tp:rp + + movdqu ($ap),%xmm1 + movdqa %xmm0,(%rsp) + movdqu %xmm1,($rp) + jmp .Lcopy4x +.align 16 +.Lcopy4x: # copy or in-place refresh + movdqu 16($ap,$i),%xmm2 + movdqu 32($ap,$i),%xmm1 + movdqa %xmm0,16(%rsp,$i) + movdqu %xmm2,16($rp,$i) + movdqa %xmm0,32(%rsp,$i) + movdqu %xmm1,32($rp,$i) + lea 32($i),$i + dec $j + jnz .Lcopy4x + + shl \$2,$num + movdqu 16($ap,$i),%xmm2 + movdqa %xmm0,16(%rsp,$i) + movdqu %xmm2,16($rp,$i) +___ +} +$code.=<<___; + mov 8(%rsp,$num,8),%rsi # restore %rsp + mov \$1,%rax +___ +$code.=<<___ if ($win64); + movaps (%rsi),%xmm6 + movaps 0x10(%rsi),%xmm7 + lea 0x28(%rsi),%rsi +___ +$code.=<<___; + mov (%rsi),%r15 + mov 8(%rsi),%r14 + mov 16(%rsi),%r13 + mov 24(%rsi),%r12 + mov 32(%rsi),%rbp + mov 40(%rsi),%rbx + lea 48(%rsi),%rsp +.Lmul4x_epilogue: + ret +.size bn_mul4x_mont_gather5,.-bn_mul4x_mont_gather5 +___ +}}} + +{ +my ($inp,$num,$tbl,$idx)=$win64?("%rcx","%rdx","%r8", "%r9") : # Win64 order + ("%rdi","%rsi","%rdx","%rcx"); # Unix order +my $out=$inp; +my $STRIDE=2**5*8; +my $N=$STRIDE/4; + +$code.=<<___; +.globl bn_scatter5 +.type bn_scatter5,\@abi-omnipotent +.align 16 +bn_scatter5: + cmp \$0, $num + jz .Lscatter_epilogue + lea ($tbl,$idx,8),$tbl +.Lscatter: + mov ($inp),%rax + lea 8($inp),$inp + mov %rax,($tbl) + lea 32*8($tbl),$tbl + sub \$1,$num + jnz .Lscatter +.Lscatter_epilogue: + ret +.size bn_scatter5,.-bn_scatter5 + +.globl bn_gather5 +.type bn_gather5,\@abi-omnipotent +.align 16 +bn_gather5: +___ +$code.=<<___ if ($win64); +.LSEH_begin_bn_gather5: + # I can't trust assembler to use specific encoding:-( + .byte 0x48,0x83,0xec,0x28 #sub \$0x28,%rsp + .byte 0x0f,0x29,0x34,0x24 #movaps %xmm6,(%rsp) + .byte 0x0f,0x29,0x7c,0x24,0x10 #movdqa %xmm7,0x10(%rsp) +___ +$code.=<<___; + mov $idx,%r11 + shr \$`log($N/8)/log(2)`,$idx + and \$`$N/8-1`,%r11 + not $idx + lea .Lmagic_masks(%rip),%rax + and \$`2**5/($N/8)-1`,$idx # 5 is "window size" + lea 96($tbl,%r11,8),$tbl # pointer within 1st cache line + movq 0(%rax,$idx,8),%xmm4 # set of masks denoting which + movq 8(%rax,$idx,8),%xmm5 # cache line contains element + movq 16(%rax,$idx,8),%xmm6 # denoted by 7th argument + movq 24(%rax,$idx,8),%xmm7 + jmp .Lgather +.align 16 +.Lgather: + movq `0*$STRIDE/4-96`($tbl),%xmm0 + movq `1*$STRIDE/4-96`($tbl),%xmm1 + pand %xmm4,%xmm0 + movq `2*$STRIDE/4-96`($tbl),%xmm2 + pand %xmm5,%xmm1 + movq `3*$STRIDE/4-96`($tbl),%xmm3 + pand %xmm6,%xmm2 + por %xmm1,%xmm0 + pand %xmm7,%xmm3 + por %xmm2,%xmm0 + lea $STRIDE($tbl),$tbl + por %xmm3,%xmm0 + + movq %xmm0,($out) # m0=bp[0] + lea 8($out),$out + sub \$1,$num + jnz .Lgather +___ +$code.=<<___ if ($win64); + movaps (%rsp),%xmm6 + movaps 0x10(%rsp),%xmm7 + lea 0x28(%rsp),%rsp +___ +$code.=<<___; + ret +.LSEH_end_bn_gather5: +.size bn_gather5,.-bn_gather5 +___ +} +$code.=<<___; +.align 64 +.Lmagic_masks: + .long 0,0, 0,0, 0,0, -1,-1 + .long 0,0, 0,0, 0,0, 0,0 +.asciz "Montgomery Multiplication with scatter/gather for x86_64, CRYPTOGAMS by " +___ + +# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, +# CONTEXT *context,DISPATCHER_CONTEXT *disp) +if ($win64) { +$rec="%rcx"; +$frame="%rdx"; +$context="%r8"; +$disp="%r9"; + +$code.=<<___; +.extern __imp_RtlVirtualUnwind +.type mul_handler,\@abi-omnipotent +.align 16 +mul_handler: + push %rsi + push %rdi + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + pushfq + sub \$64,%rsp + + mov 120($context),%rax # pull context->Rax + mov 248($context),%rbx # pull context->Rip + + mov 8($disp),%rsi # disp->ImageBase + mov 56($disp),%r11 # disp->HandlerData + + mov 0(%r11),%r10d # HandlerData[0] + lea (%rsi,%r10),%r10 # end of prologue label + cmp %r10,%rbx # context->RipRipRsp + + mov 8(%r11),%r10d # HandlerData[2] + lea (%rsi,%r10),%r10 # epilogue label + cmp %r10,%rbx # context->Rip>=epilogue label + jae .Lcommon_seh_tail + + mov 192($context),%r10 # pull $num + mov 8(%rax,%r10,8),%rax # pull saved stack pointer + + movaps (%rax),%xmm0 + movaps 16(%rax),%xmm1 + lea `40+48`(%rax),%rax + + mov -8(%rax),%rbx + mov -16(%rax),%rbp + mov -24(%rax),%r12 + mov -32(%rax),%r13 + mov -40(%rax),%r14 + mov -48(%rax),%r15 + mov %rbx,144($context) # restore context->Rbx + mov %rbp,160($context) # restore context->Rbp + mov %r12,216($context) # restore context->R12 + mov %r13,224($context) # restore context->R13 + mov %r14,232($context) # restore context->R14 + mov %r15,240($context) # restore context->R15 + movups %xmm0,512($context) # restore context->Xmm6 + movups %xmm1,528($context) # restore context->Xmm7 + +.Lcommon_seh_tail: + mov 8(%rax),%rdi + mov 16(%rax),%rsi + mov %rax,152($context) # restore context->Rsp + mov %rsi,168($context) # restore context->Rsi + mov %rdi,176($context) # restore context->Rdi + + mov 40($disp),%rdi # disp->ContextRecord + mov $context,%rsi # context + mov \$154,%ecx # sizeof(CONTEXT) + .long 0xa548f3fc # cld; rep movsq + + mov $disp,%rsi + xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER + mov 8(%rsi),%rdx # arg2, disp->ImageBase + mov 0(%rsi),%r8 # arg3, disp->ControlPc + mov 16(%rsi),%r9 # arg4, disp->FunctionEntry + mov 40(%rsi),%r10 # disp->ContextRecord + lea 56(%rsi),%r11 # &disp->HandlerData + lea 24(%rsi),%r12 # &disp->EstablisherFrame + mov %r10,32(%rsp) # arg5 + mov %r11,40(%rsp) # arg6 + mov %r12,48(%rsp) # arg7 + mov %rcx,56(%rsp) # arg8, (NULL) + call *__imp_RtlVirtualUnwind(%rip) + + mov \$1,%eax # ExceptionContinueSearch + add \$64,%rsp + popfq + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + pop %rdi + pop %rsi + ret +.size mul_handler,.-mul_handler + +.section .pdata +.align 4 + .rva .LSEH_begin_bn_mul_mont_gather5 + .rva .LSEH_end_bn_mul_mont_gather5 + .rva .LSEH_info_bn_mul_mont_gather5 + + .rva .LSEH_begin_bn_mul4x_mont_gather5 + .rva .LSEH_end_bn_mul4x_mont_gather5 + .rva .LSEH_info_bn_mul4x_mont_gather5 + + .rva .LSEH_begin_bn_gather5 + .rva .LSEH_end_bn_gather5 + .rva .LSEH_info_bn_gather5 + +.section .xdata +.align 8 +.LSEH_info_bn_mul_mont_gather5: + .byte 9,0,0,0 + .rva mul_handler + .rva .Lmul_alloca,.Lmul_body,.Lmul_epilogue # HandlerData[] +.align 8 +.LSEH_info_bn_mul4x_mont_gather5: + .byte 9,0,0,0 + .rva mul_handler + .rva .Lmul4x_alloca,.Lmul4x_body,.Lmul4x_epilogue # HandlerData[] +.align 8 +.LSEH_info_bn_gather5: + .byte 0x01,0x0d,0x05,0x00 + .byte 0x0d,0x78,0x01,0x00 #movaps 0x10(rsp),xmm7 + .byte 0x08,0x68,0x00,0x00 #movaps (rsp),xmm6 + .byte 0x04,0x42,0x00,0x00 #sub rsp,0x28 +.align 8 +___ +} + +$code =~ s/\`([^\`]*)\`/eval($1)/gem; + +print $code; +close STDOUT; -- cgit v1.2.3