For those that are interested, the timings below compare an 'openssl.exe speed'
run between a normal build and a PGO build of libeay32.dll. Machine is a
Windows Server 2008 x64 (dual quad-core 1.6GHz L5310 Xeons), build was done via
Visual Studio 2008 x64. (Mind you, it wasn't straight forward; the `perl
Configure VC-WIN64A && ms\do_win64a && nmake -f ms\ntdll.mk` approach does not
lend itself well to being augmented with an interim PGO/PGU cycle).
Normal build:
OpenSSL 0.9.8g 19 Oct 2007
built on: Thu Feb 28 07:36:23 2008
options:bn(64,64) md2(int) rc4(ptr,int) des(idx,cisc,4,long) aes(partial)
idea(int) blowfish(idx)
compiler: cl /MD /Ox /W3 /Gs0 /GF /Gy /nologo -DWIN32_LEAN_AND_MEAN -DL_ENDIAN
-DDSO_WIN32 -DOPENSSL_SYSNAME_WIN32 -DOPENSSL_SYSNAM
E_WINNT -DUNICODE -D_UNICODE -D_CRT_SECURE_NO_DEPRECATE
-D_CRT_NONSTDC_NO_DEPRECATE -DOPENSSL_USE_APPLINK -I. /Fdout32dll -DOPENSSL_
NO_CAMELLIA -DOPENSSL_NO_SEED -DOPENSSL_NO_RC5 -DOPENSSL_NO_MDC2
-DOPENSSL_NO_TLSEXT -DOPENSSL_NO_KRB5 -DOPENSSL_NO_DYNAMIC_ENGINE
available timing options: TIMEB HZ=1000
timing function used: ftime
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
md2 875.77k 1796.56k 2436.60k 2675.31k 2755.84k
mdc2 0.00 0.00 0.00 0.00 0.00
md4 18029.35k 59417.30k 156723.18k 263896.44k 332222.10k
md5 14864.85k 47665.93k 119987.24k 191630.11k 233178.82k
hmac(md5) 15873.99k 48995.30k 121695.28k 192041.39k 233138.32k
sha1 15757.69k 48281.50k 113801.70k 172427.71k 203391.01k
rmd160 11715.11k 31724.71k 64739.40k 87609.48k 97883.41k
rc4 215534.64k 234810.58k 237402.24k 246108.49k 250668.10k
des cbc 28718.27k 30039.78k 30554.03k 30640.52k 30595.82k
des ede3 11309.21k 11658.13k 11733.94k 11753.62k 11733.67k
idea cbc 31037.31k 32688.19k 33143.45k 32886.83k 33245.25k
seed cbc 0.00 0.00 0.00 0.00 0.00
rc2 cbc 14291.86k 14642.36k 14711.69k 14732.36k 14722.67k
rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00
blowfish cbc 49446.55k 53773.13k 55151.93k 55434.38k 55581.30k
cast cbc 47798.34k 51087.75k 51829.52k 52983.47k 53371.13k
aes-128 cbc 75743.64k 81166.99k 82100.40k 82423.07k 82402.83k
aes-192 cbc 65382.76k 70522.14k 71453.22k 71697.50k 71943.46k
aes-256 cbc 59747.92k 63072.24k 63646.49k 63815.96k 63828.10k
camellia-128 cbc 0.00 0.00 0.00 0.00 0.00
camellia-192 cbc 0.00 0.00 0.00 0.00 0.00
camellia-256 cbc 0.00 0.00 0.00 0.00 0.00
sha256 12607.81k 30391.44k 54975.72k 69835.96k 75735.09k
sha512 9071.95k 36271.14k 64255.90k 95176.38k 111448.75k
aes-128 ige 79081.86k 87086.51k 88510.77k 87792.86k 88885.91k
aes-192 ige 70984.62k 75217.29k 77083.46k 76555.86k 77083.46k
aes-256 ige 62345.66k 66391.83k 67636.43k 67432.54k 67855.27k
sign verify sign/s verify/s
rsa 512 bits 0.000478s 0.000032s 2093.7 31697.7
rsa 1024 bits 0.001691s 0.000077s 591.4 12926.0
rsa 2048 bits 0.008902s 0.000229s 112.3 4373.8
rsa 4096 bits 0.054600s 0.000762s 18.3 1312.5
sign verify sign/s verify/s
dsa 512 bits 0.000330s 0.000339s 3034.4 2947.7
dsa 1024 bits 0.000774s 0.000869s 1292.4 1151.1
dsa 2048 bits 0.002274s 0.002739s 439.7 365.1
PGO-enabled build (the instrumentation was captured via an openssl.exe speed
run, then everything was rebuilt with the new profiling information):
OpenSSL 0.9.8g 19 Oct 2007
built on: Thu Feb 28 19:23:56 2008
options:bn(64,64) md2(int) rc4(ptr,int) des(idx,cisc,4,long) aes(partial)
idea(int) blowfish(idx)
compiler: cl /MD /Ox /favor:INTEL64 /GA /GL /W3 /Gs0 /GF /Gy /nologo
-DWIN32_LEAN_AND_MEAN -DL_ENDIAN -DDSO_WIN32 -DOPENSSL_SYSNAME
_WIN32 -DOPENSSL_SYSNAME_WINNT -DUNICODE -D_UNICODE -D_CRT_SECURE_NO_DEPRECATE
-D_CRT_NONSTDC_NO_DEPRECATE -DOPENSSL_USE_APPLINK -I.
/Fdout32dll -DOPENSSL_NO_CAMELLIA -DOPENSSL_NO_SEED -DOPENSSL_NO_RC5
-DOPENSSL_NO_MDC2 -DOPENSSL_NO_TLSEXT -DOPENSSL_NO_KRB5 -DOPEN
SSL_NO_DYNAMIC_ENGINE
available timing options: TIMEB HZ=1000
timing function used: ftime
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
md2 884.43k 1812.43k 2454.02k 2694.54k 2773.61k
mdc2 0.00 0.00 0.00 0.00 0.00
md4 19307.46k 62482.07k 162648.73k 268865.64k 333460.19k
md5 16051.68k 50344.23k 123079.07k 193341.58k 233178.82k
hmac(md5) 16751.25k 51000.39k 123794.25k 194631.28k 233788.06k
sha1 16778.05k 50490.06k 117057.15k 175585.72k 204350.99k
rmd160 12193.19k 32864.28k 65725.35k 88336.01k 97990.60k
rc4 210451.78k 234318.66k 236899.41k 246651.22k 250705.56k
des cbc 28948.69k 29831.47k 30082.87k 30210.17k 30294.72k
des ede3 11638.72k 11850.82k 11910.12k 11910.07k 11928.90k
idea cbc 32198.86k 33039.02k 33504.18k 33662.15k 33662.15k
seed cbc 0.00 0.00 0.00 0.00 0.00
rc2 cbc 14197.53k 14484.34k 14542.73k 14552.82k 14572.41k
rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00
blowfish cbc 50021.51k 54041.60k 55016.28k 55288.24k 55434.38k
cast cbc 48447.06k 52078.89k 53371.13k 53644.18k 53773.13k
aes-128 cbc 77385.68k 81462.57k 82423.07k 82402.83k 83055.52k
aes-192 cbc 67216.41k 70745.17k 71943.46k 71697.50k 71928.04k
aes-256 cbc 60090.32k 63072.24k 63634.42k 63453.92k 63634.42k
camellia-128 cbc 0.00 0.00 0.00 0.00 0.00
camellia-192 cbc 0.00 0.00 0.00 0.00 0.00
camellia-256 cbc 0.00 0.00 0.00 0.00 0.00
sha256 12712.42k 30552.64k 55650.44k 70179.20k 76000.98k
sha512 9450.36k 37685.73k 65577.63k 96887.12k 113063.54k
aes-128 ige 81166.99k 87086.51k 88138.78k 88534.12k 89240.51k
aes-192 ige 70984.62k 75217.29k 76538.39k 76818.75k 77101.18k
aes-256 ige 63072.24k 66182.31k 66801.58k 67636.43k 67650.06k
sign verify sign/s verify/s
rsa 512 bits 0.000408s 0.000032s 2452.6 31412.8
rsa 1024 bits 0.001786s 0.000086s 559.8 11671.4
rsa 2048 bits 0.010049s 0.000267s 99.5 3748.3
rsa 4096 bits 0.064000s 0.000895s 15.6 1117.3
sign verify sign/s verify/s
dsa 512 bits 0.000330s 0.000350s 3027.9 2860.1
dsa 1024 bits 0.000869s 0.000976s 1150.8 1024.2
dsa 2048 bits 0.002667s 0.003168s 374.9 315.7
I find the results a bit unexciting (compared to say, Python, where PGO
provides a good 35% speedup or so); some runs are faster, but there are a few
that are actually slower.
Regards,
Trent.
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List [email protected]
Automated List Manager [EMAIL PROTECTED]
