[gmx-developers] Nonbonded kernel without forces

Andre Assunção S. T. Ribeiro aastr at iq.ufrj.br
Wed Apr 8 19:57:06 CEST 2009


Yes, I was actually calling the c no-forces kernels and comparing with
assembly kernels for forces. Sorry for that. I expanded the kernellist
arrays in nb_kernel_**.c to include nf terms just like
src/gmxlib/nonbonded/nb_kernel/nb_kernel.c and it became indeed faster.
My question now is that if this is the only way to call a generic nf
assembly kernel, because if so, once my mc integrator is done, it would
not be very practical for users to edit this manually.

cheers

Am Dienstag, den 07.04.2009, 23:08 +0200 schrieb
hessb at mpip-mainz.mpg.de:
> Hi,
> 
> They are indeed faster.
> I put them to use for the test particle insertion.
> If do_force is called without GMX_FORCE_FORCES
> all innerloops will not calculate forces.
> 
> Berk
> 
> > No, they non-force kernels should be at least 50% faster. It is the
> > same code, but with lots of lines and memory operations ripped out.
> >
> > Cheers,
> >
> > Erik
> >
> > On Apr 7, 2009, at 7:57 PM, Andre Assunção S. T. Ribeiro wrote:
> >
> >> Hello all,
> >>
> >> I've been trying to improve performance of the mc integrator i'm
> >> working
> >> on. I tried calling the no-forces kernels for ia32_sse and it seems a
> >> bit slower than the regular kernels with forces. Is this expected? Are
> >> the no-forces kernels as optimized as the ones that do calculate
> >> forces?
> >> I still need to check if the same happens with the c and fortran
> >> codes.
> >>
> >> Thanks
> >>
> >> _______________________________________________
> >> gmx-developers mailing list
> >> gmx-developers at gromacs.org
> >> http://www.gromacs.org/mailman/listinfo/gmx-developers
> >> Please don't post (un)subscribe requests to the list. Use the
> >> www interface or send it to gmx-developers-request at gromacs.org.
> >>
> >
> > _______________________________________________
> > gmx-developers mailing list
> > gmx-developers at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-developers
> > Please don't post (un)subscribe requests to the list. Use the
> > www interface or send it to gmx-developers-request at gromacs.org.
> >
> 
> 
> _______________________________________________
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-developers-request at gromacs.org.




More information about the gromacs.org_gmx-developers mailing list