diff options
| author | Parménides GV <parmegv@sdf.org> | 2014-04-07 20:43:34 +0200 |
|---|---|---|
| committer | Parménides GV <parmegv@sdf.org> | 2014-04-08 11:43:27 +0200 |
| commit | c206a91d320995f37f8abb33188bfd384249da3d (patch) | |
| tree | 10a7d8a9dd7f24437ac4851b8d01edbd5dd3ee3b /openssl/crypto/des/times/usparc.cc | |
| parent | 910b0e1746ab3f63e63808b198ad51fec5b635e5 (diff) | |
Next step: compile jni sources correctly.
Diffstat (limited to 'openssl/crypto/des/times/usparc.cc')
| -rw-r--r-- | openssl/crypto/des/times/usparc.cc | 31 |
1 files changed, 0 insertions, 31 deletions
diff --git a/openssl/crypto/des/times/usparc.cc b/openssl/crypto/des/times/usparc.cc deleted file mode 100644 index 0864285e..00000000 --- a/openssl/crypto/des/times/usparc.cc +++ /dev/null @@ -1,31 +0,0 @@ -solaris 2.5.1 usparc 167mhz?? - SC4.0 cc -fast -Xa -xO5 - -For the ultra sparc, SunC 4.0 cc -fast -Xa -xO5, running 'des_opts' -gives a speed of 475,000 des/s while 'speed' gives 417,000 des/s. -I believe the difference is tied up in optimisation that the compiler -is able to perform when the code is 'inlined'. For 'speed', the DES -routines are being linked from a library. I'll record the higher -speed since if performance is everything, you can always inline -'des_enc.c'. - -[ 16-Jan-06 - I've been playing with the - '-xtarget=ultra -xarch=v8plus -Xa -xO5 -Xa' - and while it makes the des_opts numbers much slower, it makes the - actual 'speed' numbers look better which is a realistic version of - using the libraries. ] - -options des ecb/s -16 r1 p 475516.90 100.0% -16 r2 p 439388.10 92.4% -16 c i 427001.40 89.8% -16 c p 419516.50 88.2% - 4 r2 p 409491.70 86.1% - 4 r1 p 404266.90 85.0% - 4 c p 398121.00 83.7% - 4 c i 370588.40 77.9% - 4 r1 i 362742.20 76.3% -16 r2 i 331275.50 69.7% -16 r1 i 324730.60 68.3% - 4 r2 i 63535.10 13.4% <-- very very weird, must be cache problems. --DDES_UNROLL -DDES_RISC1 -DDES_PTR - |