[gmx-users] Benchs on gcc and pgi

Jones de Andrade johannesrs at gmail.com
Thu May 18 23:47:17 CEST 2006


Hi Erik!

> Normal simulations aren't using any linear algebra stuff at all. BLAS 
> & LAPACK are only useful when you do matrix stuff, for instance
> normal mode analysis.

Ouch, obvious mistake. Sorry, apologize the error from someone programming too much analysis tolls, rather than the simulation programs for a "while". ;)

Anyway, is there any kind of system test for this, for instance some system big enough for the Normal Mode Analysis take some time before ends, and make it possible to bench how it performs?

> GPUs aren't useful for general simulations yet, since
> neighborsearching isn't working. We have good hopes for the future 
> though, and if/when it's working we'll of course include it.

Not wanting to press on this subject, that from previous discussions in this list seems to be a somewhat "sensible" one: You mean that the only and last thing that still need/should be "converted" to make Gromacs run on GPUs (or somehow slacing the job between the CPU and the GPU, as the tests from Folding at Home shows to be the better option to the actual point) is "just" (not "just" as a simples job, but as "the only and last one" job) the neighborsearching? The "impossible" first, and "hard to get any speed improvement" later situation has already grown to this point? Those would be terrific news!

Anyway, thanks a lot.

Jones



More information about the gromacs.org_gmx-users mailing list