Charm++ parallel programming system

[Ben Porebski <ben.porebski AT monash.edu>]

Hi Eduardo,

I have recently been successful in compiling charm and namd for the xeon phi.

My most successful tests have been with charm 6.5.0/namd 2.9 compiled as net-x86_64-smp-icc.

I had to make a few modifications to a makefile and ensure that everything is compiled with the flag -mmic.

Everything will compile and run, however performance of a single phi card is roughly equivalent to 2x xeon E5462's.

I suspect that the phi's are capable of more performance for molecular dynamics, but I'll have to wait until the developers implement optimisations.

The -mmic flag compiles native code for the phi card, although the cards support host delegation using openMP.

I again suspect that openMP will be required for maximum performance, rather than running code natively on the phi.

If you have any further questions, or need help compiling for the phi, feel free to email me.

Cheers, Ben

On Fri, Mar 29, 2013 at 6:02 AM, Eduardo <erocha.ssa AT gmail.com> wrote:

Hi,

Is there already a version of AMPI/Charm for Intel Xeon Phi ? I mean,
a version that would treat the board just as a regular x86 node.

Thanks in advance,
Eduardo
_______________________________________________
charm mailing list
charm AT cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/charm