[Bioperl-l] Help with testing of parallel Smith-Waterman
code on x86 workstations
Jonathan_Epstein at nih.gov
Tue Jul 8 16:56:35 EDT 2003
While we're on this slightly off-topic thread ... I know that the Apple folks used Altivec to squeeze extra performance out of BLAST. Has anyone ever done the same thing with SSE et al for Intel CPUs? Any source code available?
At 03:24 p.m. 7/8/2003 -0400, Aaron J Mackey wrote:
>You should get in touch with Arpith Jacob <jarpith at acm.org>; he has
>already done this as you've described (with both diagonal and horizontal
>vector implementations), and has benchmarks available. It seems a popular
>thing to do these days ;)
>On Tue, 8 Jul 2003, Nemo Physche wrote:
>> Hello all,
>> I'm writing a parallel version of the Smith-Waterman local alignment
>> using SIMD technology (MMX/SSE/SSE2) on Intel x86 platforms. The
>> implementations are based on two published papers in "Bioinformatics."
>> vary according to a wide variety of factors, but is considerably faster than
>> the normal integer implementation of the algorithm.
>> I need help testing the code on a wide variety of platforms including Intel
>> 486/Pentium without MMX, Pentium MMX, Pentium 2/3 and Pentium 4 for a wide
>> variety of clock speeds. The code is written using MMX/SSE/SSE2 instructions
>> and as such does not use AMD specific 3DNow instructions, however the code
>> should run unchanged on AMD processors. Anyone with AMD machines are also
>> The programs are written in C and assembly and should be compilable in Linux
>> Windows (Djgpp) with GCC 3.2.3. The code will be probably be released under
>> the MIT or Apache license.
>> Please contact me if you are interested in helping.
>> Bioperl-l mailing list
>> Bioperl-l at portal.open-bio.org
> Aaron J Mackey
> Pearson Laboratory
> University of Virginia
> (434) 924-2821
> amackey at virginia.edu
>Bioperl-l mailing list
>Bioperl-l at portal.open-bio.org
More information about the Bioperl-l