[gmx-users] gromacs on itanium2
lindahl at cbr.su.se
Tue Mar 3 22:02:56 CET 2009
There have been some reports on newer ia64 processors being quite fast
with the Fortran kernels instead (even faster than asm!), so I would
This has to do with the brain-dead architecture on ia64. The asm
kernels were written for original itanium2 timings, but with the
register rotation loops and lack of dynamic instruction scheduling you
would get lots of pipeline stalls/bubbles if Intel ever changed their
instruction latencies, and that might have happend now.
Considering the limited adoption of ia64 I don't think we're very
likely to keep updating those kernels - the first thing on the agenda
right now will rather be improved performance of x86 SSE kernels for
On Mar 3, 2009, at 8:14 PM, David van der Spoel wrote:
> Marius Retegan wrote:
>> Since I was unable to get a working version for Gromacs 4.0.4 on a
>> Itanium 2 machine with the ia64 nonbonded kernel, I was wondering
>> would be the lost in speed if I would disable the assembly loops?
> About a factor of two. What's the problem with compiling?
>> gmx-users mailing list gmx-users at gromacs.org
>> Please search the archive at http://www.gromacs.org/search before
>> Please don't post (un)subscribe requests to the list. Use the www
>> interface or send it to gmx-users-request at gromacs.org.
>> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> David van der Spoel, Ph.D., Professor of Biology
> Molec. Biophys. group, Dept. of Cell & Molec. Biol., Uppsala
> Box 596, 75124 Uppsala, Sweden. Phone: +46184714205. Fax: +4618511755.
> spoel at xray.bmc.uu.se spoel at gromacs.org http://folding.bmc.uu.se
> gmx-users mailing list gmx-users at gromacs.org
> Please search the archive at http://www.gromacs.org/search before
> Please don't post (un)subscribe requests to the list. Use the www
> interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
More information about the gromacs.org_gmx-users