> So, has anybody done any benchmarks on common platforms of OpenSSL versus
> the BSAFE crypto stuff? Specifically the RSA operations
Almost two month ago I've challenged this list (RSAref users to be more
accurate) to submit results from 'apps/openssl speed rsa dsa' but
haven't got any replies whatsoever:-(
> and rc4/rc5. Just
> HOW much does the assembly help things with OpenSSL?
>
> I'd especially be interested in the timings against something other than
> Solaris/sparc or Windows/x86.
Hmmm... You wonder how much does assembly stuff helps, but don't want to
hear about x86... But most assembly stuff *is* x86. In rc4/rc5 case the
*only* assembly stuff is x86... In RSA case it's currently only x86,
SPARC and MIPS III/IV that are backed up by assembler crafted bignums.
So it's not much left to tell about and I think I can say few words
about MIPS:-) The catch is that MIPS III/IV compliant CPU offers 128-bit
multiplication (I mean 64-bit number times 64-bit number produces
128-bit result) while there're no C compilers available providing means
to address the result(*). Therefore you have to replace *one* 64-bit
multiplication with *four* 32-bit ones plus some additions to calculate
same result. Speed-up factor should therefore be over 4 in *ideal* case
and is something like 3-3.5 in practice. So that if BSAFE doesn't offer
MIPS assembler module...
(*) MIPSpro C has an intrinsic to address 64 most significant bits of
the 128-bit result, but the last time I checked the compiler still
generated two multiplications.
Andy.
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
