Thank you all for your comments, but i still do not have the issue resolved.
After your email below, i tried "-multi" and also "time". I also loaded the
CPU and took some measurements.
But to my surprise, i am getting the same performance numbers, for "time
./openssl speed dsa", in all the three following cases.
Case1: With no load on CPU, and without using Cavium
command used to measure: "time ./openssl speed dsa"
Case2: With CPU loaded, and without using Cavium
command used to measure: "time ./openssl speed dsa"
command used to load CPU: "openssl speed -multi 100000"
Case3: With CPU loaded and using Cavium
command used to measure: "time ./openssl speed dsa"
command used to load CPU: "openssl speed -multi 100000"
The output i am getting in all the three cases are very similar to the one
below.
[EMAIL PROTECTED]:/home/eeliazer/cavium/CN1000-linux2.6-openssl-sdk-4.1
/cavium/apps/openssl-0.9.7d/apps# time ./openssl speed dsa
Doing 512 bit sign dsa's for 10s: 2149 512 bit DSA signs in 1.52s
Doing 512 bit verify dsa's for 10s: 116 512 bit DSA verify in 0.10s
Doing 1024 bit sign dsa's for 10s: 46 1024 bit DSA signs in 0.10s
Doing 1024 bit verify dsa's for 10s: 37 1024 bit DSA verify in 0.10s
Doing 2048 bit sign dsa's for 10s: 13 2048 bit DSA signs in 0.09s
Doing 2048 bit verify dsa's for 10s: 10 2048 bit DSA verify in 0.09s
OpenSSL 0.9.7d 17 Mar 2004
built on: Thu Jan 4 17:58:02 PST 2007
options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial)
idea(int) blowfish(idx)
compiler: gcc -DDSO_DLFCN -DHAVE_DLFCN_H -DOPENSSL_NO_KRB5 -fPIC
-DCAVIUM_ENGINE -DL_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -mcpu=pentium
-Wall -DSHA1_ASM -DMD5_ASM -DRMD160_ASM
available timing options: TIMES TIMEB HZ=100 [sysconf value]
timing function used: times
sign verify sign/s verify/s
dsa 512 bits 0.0007s 0.0009s 1413.8 1160.0
dsa 1024 bits 0.0022s 0.0027s 460.0 370.0
dsa 2048 bits 0.0069s 0.0090s 144.4 111.1
real 4m45.529s
user 0m2.028s
sys 0m0.000s
When i use the engine, i expect to see some difference from the baseline
software performance.
I am noticing no change in numbers, which makes me think somehow i am not
properly exercising cavium hardware accelerator.
Also the 'openssl speed' program seems to be a confusing one to measure
performance.
Any suggestions on a good tool to load CPU in a controlled manner, in this
context?
Your suggestions on how i can really ensure that the hardware is being
exercised will be valuable.
Thanks in advance.
cheers,
Elwin.
On 1/5/07, Mathias Brossard <[EMAIL PROTECTED]> wrote:
Elwin Stelzer Eliazer wrote:
> I am trying to compare the performance benefits in using hardware
> accelerator, and am experimenting the CAVIUM SSL card.
>
> I am comparing the results between 'speed' and 'speed -engine cavium',
> and i do not see any difference in numbers.
> I do see the message 'cavium engine set', but really do not see any
> change in the performance numbers.
Have you tried using the -multi flag ? If that works try increasing the
value to see how it scales.
> Is there a way to check if the hardware is really used? Am i missing
> some needed steps?
Launch openssl with 'time openssl speed rsa...' to see how much CPU is
used. That in turn should tell you if the hardware is actually
offloading the crypto.
Sincerely,
--
Mathias Brossard
______________________________________________________________________
OpenSSL Project http://www.openssl.org
User Support Mailing List openssl-users@openssl.org
Automated List Manager [EMAIL PROTECTED]
