1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
|
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> 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/.
# ====================================================================
#
# SHA256 block transform for x86. September 2007.
#
# Performance in clock cycles per processed byte (less is better):
#
# Pentium PIII P4 AMD K8 Core2
# gcc 46 36 41 27 26
# icc 57 33 38 25 23
# x86 asm 40 30 35 20 20
# x86_64 asm(*) - - 21 15.8 16.5
#
# (*) x86_64 assembler performance is presented for reference
# purposes.
#
# Performance improvement over compiler generated code varies from
# 10% to 40% [see above]. Not very impressive on some µ-archs, but
# it's 5 times smaller and optimizies amount of writes.
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
&asm_init($ARGV[0],"sha512-586.pl",$ARGV[$#ARGV] eq "386");
$A="eax";
$E="edx";
$T="ebx";
$Aoff=&DWP(0,"esp");
$Boff=&DWP(4,"esp");
$Coff=&DWP(8,"esp");
$Doff=&DWP(12,"esp");
$Eoff=&DWP(16,"esp");
$Foff=&DWP(20,"esp");
$Goff=&DWP(24,"esp");
$Hoff=&DWP(28,"esp");
$Xoff=&DWP(32,"esp");
$K256="ebp";
sub BODY_00_15() {
my $in_16_63=shift;
&mov ("ecx",$E);
&add ($T,&DWP(4*(8+15+16-9),"esp")) if ($in_16_63); # T += X[-7]
&ror ("ecx",6);
&mov ("edi",$E);
&ror ("edi",11);
&mov ("esi",$Foff);
&xor ("ecx","edi");
&ror ("edi",25-11);
&mov (&DWP(4*(8+15),"esp"),$T) if ($in_16_63); # save X[0]
&xor ("ecx","edi"); # Sigma1(e)
&mov ("edi",$Goff);
&add ($T,"ecx"); # T += Sigma1(e)
&mov ($Eoff,$E); # modulo-scheduled
&xor ("esi","edi");
&mov ("ecx",$A);
&and ("esi",$E);
&mov ($E,$Doff); # e becomes d, which is e in next iteration
&xor ("esi","edi"); # Ch(e,f,g)
&mov ("edi",$A);
&add ($T,"esi"); # T += Ch(e,f,g)
&ror ("ecx",2);
&add ($T,$Hoff); # T += h
&ror ("edi",13);
&mov ("esi",$Boff);
&xor ("ecx","edi");
&ror ("edi",22-13);
&add ($E,$T); # d += T
&xor ("ecx","edi"); # Sigma0(a)
&mov ("edi",$Coff);
&add ($T,"ecx"); # T += Sigma0(a)
&mov ($Aoff,$A); # modulo-scheduled
&mov ("ecx",$A);
&sub ("esp",4);
&or ($A,"esi"); # a becomes h, which is a in next iteration
&and ("ecx","esi");
&and ($A,"edi");
&mov ("esi",&DWP(0,$K256));
&or ($A,"ecx"); # h=Maj(a,b,c)
&add ($K256,4);
&add ($A,$T); # h += T
&mov ($T,&DWP(4*(8+15+16-1),"esp")) if ($in_16_63); # preload T
&add ($E,"esi"); # d += K256[i]
&add ($A,"esi"); # h += K256[i]
}
&function_begin("sha256_block_data_order");
&mov ("esi",wparam(0)); # ctx
&mov ("edi",wparam(1)); # inp
&mov ("eax",wparam(2)); # num
&mov ("ebx","esp"); # saved sp
&call (&label("pic_point")); # make it PIC!
&set_label("pic_point");
&blindpop($K256);
&lea ($K256,&DWP(&label("K256")."-".&label("pic_point"),$K256));
&sub ("esp",16);
&and ("esp",-64);
&shl ("eax",6);
&add ("eax","edi");
&mov (&DWP(0,"esp"),"esi"); # ctx
&mov (&DWP(4,"esp"),"edi"); # inp
&mov (&DWP(8,"esp"),"eax"); # inp+num*128
&mov (&DWP(12,"esp"),"ebx"); # saved sp
&set_label("loop",16);
# copy input block to stack reversing byte and dword order
for($i=0;$i<4;$i++) {
&mov ("eax",&DWP($i*16+0,"edi"));
&mov ("ebx",&DWP($i*16+4,"edi"));
&mov ("ecx",&DWP($i*16+8,"edi"));
&mov ("edx",&DWP($i*16+12,"edi"));
&bswap ("eax");
&bswap ("ebx");
&bswap ("ecx");
&bswap ("edx");
&push ("eax");
&push ("ebx");
&push ("ecx");
&push ("edx");
}
&add ("edi",64);
&sub ("esp",4*8); # place for A,B,C,D,E,F,G,H
&mov (&DWP(4*(8+16)+4,"esp"),"edi");
# copy ctx->h[0-7] to A,B,C,D,E,F,G,H on stack
&mov ($A,&DWP(0,"esi"));
&mov ("ebx",&DWP(4,"esi"));
&mov ("ecx",&DWP(8,"esi"));
&mov ("edi",&DWP(12,"esi"));
# &mov ($Aoff,$A);
&mov ($Boff,"ebx");
&mov ($Coff,"ecx");
&mov ($Doff,"edi");
&mov ($E,&DWP(16,"esi"));
&mov ("ebx",&DWP(20,"esi"));
&mov ("ecx",&DWP(24,"esi"));
&mov ("edi",&DWP(28,"esi"));
# &mov ($Eoff,$E);
&mov ($Foff,"ebx");
&mov ($Goff,"ecx");
&mov ($Hoff,"edi");
&set_label("00_15",16);
&mov ($T,&DWP(4*(8+15),"esp"));
&BODY_00_15();
&cmp ("esi",0xc19bf174);
&jne (&label("00_15"));
&mov ($T,&DWP(4*(8+15+16-1),"esp")); # preloaded in BODY_00_15(1)
&set_label("16_63",16);
&mov ("esi",$T);
&mov ("ecx",&DWP(4*(8+15+16-14),"esp"));
&shr ($T,3);
&ror ("esi",7);
&xor ($T,"esi");
&ror ("esi",18-7);
&mov ("edi","ecx");
&xor ($T,"esi"); # T = sigma0(X[-15])
&shr ("ecx",10);
&mov ("esi",&DWP(4*(8+15+16),"esp"));
&ror ("edi",17);
&xor ("ecx","edi");
&ror ("edi",19-17);
&add ($T,"esi"); # T += X[-16]
&xor ("edi","ecx") # sigma1(X[-2])
&add ($T,"edi"); # T += sigma1(X[-2])
# &add ($T,&DWP(4*(8+15+16-9),"esp")); # T += X[-7], moved to BODY_00_15(1)
# &mov (&DWP(4*(8+15),"esp"),$T); # save X[0]
&BODY_00_15(1);
&cmp ("esi",0xc67178f2);
&jne (&label("16_63"));
&mov ("esi",&DWP(4*(8+16+64)+0,"esp"));#ctx
# &mov ($A,$Aoff);
&mov ("ebx",$Boff);
&mov ("ecx",$Coff);
&mov ("edi",$Doff);
&add ($A,&DWP(0,"esi"));
&add ("ebx",&DWP(4,"esi"));
&add ("ecx",&DWP(8,"esi"));
&add ("edi",&DWP(12,"esi"));
&mov (&DWP(0,"esi"),$A);
&mov (&DWP(4,"esi"),"ebx");
&mov (&DWP(8,"esi"),"ecx");
&mov (&DWP(12,"esi"),"edi");
# &mov ($E,$Eoff);
&mov ("eax",$Foff);
&mov ("ebx",$Goff);
&mov ("ecx",$Hoff);
&mov ("edi",&DWP(4*(8+16+64)+4,"esp"));#inp
&add ($E,&DWP(16,"esi"));
&add ("eax",&DWP(20,"esi"));
&add ("ebx",&DWP(24,"esi"));
&add ("ecx",&DWP(28,"esi"));
&mov (&DWP(16,"esi"),$E);
&mov (&DWP(20,"esi"),"eax");
&mov (&DWP(24,"esi"),"ebx");
&mov (&DWP(28,"esi"),"ecx");
&add ("esp",4*(8+16+64)); # destroy frame
&sub ($K256,4*64); # rewind K
&cmp ("edi",&DWP(8,"esp")); # are we done yet?
&jb (&label("loop"));
&mov ("esp",&DWP(12,"esp")); # restore sp
&function_end_A();
&set_label("K256",64); # Yes! I keep it in the code segment!
&data_word(0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5);
&data_word(0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5);
&data_word(0xd807aa98,0x12835b01,0x243185be,0x550c7dc3);
&data_word(0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174);
&data_word(0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc);
&data_word(0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da);
&data_word(0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7);
&data_word(0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967);
&data_word(0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13);
&data_word(0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85);
&data_word(0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3);
&data_word(0xd192e819,0xd6990624,0xf40e3585,0x106aa070);
&data_word(0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5);
&data_word(0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3);
&data_word(0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208);
&data_word(0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2);
&function_end_B("sha256_block_data_order");
&asciz("SHA256 block transform for x86, CRYPTOGAMS by <appro\@openssl.org>");
&asm_finish();
|
