[Bioclusters] Cluster specs

Joe Landman bioclusters@bioinformatics.org
Thu, 04 Nov 2004 13:46:53 -0500

John Van Workum wrote:

>>Have a good hard look at Opterons.  Better performers by far on bio
>>codes than Athlons, and on chem codes.  The 64 bit OS and builds have
>>been shown to give (a free and) significant performance advantage over
>>the pure 32 bit binaries.
>But to take full advantage of a 64-bit OS, shouldn't 32 bit code be rewritten
>for 64-bit OS? 

Under gcc, use -m64

You can rewrite if you wish.  The compilers are smart enough to do a 
good job on it.

>Is the performance boost you are seeing mostly due to the
>higher performance of the HW (faster FSB),

FSB issues are not relevant for many codes in this space.  What matters 
is the memory latency for pointer chasers like BLAST, memory access 
patterns in general to exploit large caches, and the use of SIMD and 
variants for codes such as HMMer.

> better compiler, both? How did you
>test and compare the two binaries?

One SuSE 9.0 AMD64 system (dual 246's, 2 GB ram).  Gcc 3.3.2.  Used -m32 
for 32 bit binaries (checked to make sure they ran on Pentium/Athlons), 
used -m64 for 64 bit binaries.  Same input decks.  Same data set.

I will check with the folks who own the paper that was written and see 
if we can distribute it. 


>Bioclusters maillist  -  Bioclusters@bioinformatics.org

Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman@scalableinformatics.com
web  : http://www.scalableinformatics.com
phone: +1 734 612 4615