Hi Jeseem,
are you sure that the cpu you are using has AESNI enabled?
The toolchain you are using is definitly correct, but the fact that the
engine check was not triggered sounds like your SKU has the instraction
disabled.
You can do a check on the CPU with the below code:
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <cpuid.h>
int Check_CPU_support_AES()
{
#if defined(__INTEL_COMPILER)
int CPUInfo[4] = {-1};
__cpuid(CPUInfo, 1);
return (CPUInfo[2] & 0x2000000);
#else
unsigned int a=1,b,c,d;
__cpuid(1, a,b,c,d);
return (c & 0x2000000);
#endif
}
int main()
{
printf("AESNI support: %X \n", Check_CPU_support_AES());
return 0;
}
Hope this helps
Best Regards
Sergio
On Mon, Oct 4, 2010 at 12:20 AM, Jeseem S <jese...@gmail.com> wrote:
> Hi
> I am trying to run aesni on latest openssl from head on a freebsd machine
> with cpu ( xeon 5620 , which supports AESNI ).
>
> running the command "openssl speed -evp aes256 -engine aesni
> It doesn't look like the aesni engine is being used as I see no performance
> improvement.
>
> I found the check for enabling engine wasn't triggered and so bypassed it
> as below
> --- eng_aesni.c.orig 2010-10-03 23:18:09.000000000 -0700
> +++ eng_aesni.c 2010-10-03 23:03:45.000000000 -0700
> @@ -329,6 +329,7 @@
> IA32CAP OPENSSL_ia32_cpuid(void);
> engage = (OPENSSL_ia32_cpuid() >> 57) & 1;
> }
> + engage=1;
>
> /* Register everything or return with an error */
> if (!ENGINE_set_id(e, aesni_id) ||
>
>
>
> However am hitting following error
> aesni engine added
> To get the most accurate results, try to run this
> program when this computer is idle.
> Doing aes-256-cbc for 3s on 16 size blocks: Illegal instruction
>
> Any ideas.
> the os is freebsd 8.0, gcc is 4.4 and as is 2.19
>
> thanks
> jeseem
>
>
--
preferisco ammazzare il tempo,
preferisco sparare cazzate,
preferisco fare esplodere una moda,
preferisco morire d'amore.
(Caparezza)
