Search found 9 matches

by Michal Kvasnicka
Wed Jan 19, 2011 11:13 am
Forum: Parallelization
Topic: Threading with OpenMP
Replies: 24
Views: 9755

Re: Threading with OpenMP

These OpenMP Microbenchmarks (three separate tests) are intended to measure the overheads of synchronisation (first), loop scheduling (second) and array operations (third) in the OpenMP runtime library. So, each benchmark is able to measure overheads of OpenMP communication processes, but each test ...
by Michal Kvasnicka
Tue Jan 18, 2011 11:51 pm
Forum: Parallelization
Topic: Threading with OpenMP
Replies: 24
Views: 9755

Re: Threading with OpenMP

OK ...well, there is a lot of possible explanations. - CPU caching and RAM memory access scenarios are different for different types of CPUs - memory-CPUcore and inter-core + CPU cache (L1, L2, ...) communication - gcc compiler produce still a bit strange binaries, which has not same performance on ...
by Michal Kvasnicka
Tue Jan 18, 2011 10:21 pm
Forum: Parallelization
Topic: Threading with OpenMP
Replies: 24
Views: 9755

Re: Threading with OpenMP

your parallelized loop should look like this: /* Parallelized loop */ #pragma omp parallel reduction(+:res) for (n = 0; n < N; n++) { /* Operations */ if (test == 1) res[omp_get_thread_num()] += erf(DATA[ptr + n]); else if (test == 2) res[omp_get_thread_num()] += DATA[ptr + n]; else if (test == 3) r...
by Michal Kvasnicka
Tue Jan 18, 2011 9:16 pm
Forum: Parallelization
Topic: Threading with OpenMP
Replies: 24
Views: 9755

Re: Threading with OpenMP

I am missing

#include <omp.h>

in your test code...

Michal
by Michal Kvasnicka
Tue Jan 18, 2011 3:53 pm
Forum: Parallelization
Topic: Threading with OpenMP
Replies: 24
Views: 9755

Re: Threading with OpenMP

And finally, "Hybrid OpenMP MPI Benchmark": http://homb.sourceforge.net/ These benchmarks help you to decide what HW and or OS platform is suitable for your multi-core parallel programming, I hope. The completely different task is, of course, how to write effective parallelized OpenMP vs MPI codes, ...
by Michal Kvasnicka
Tue Jan 18, 2011 3:28 pm
Forum: Parallelization
Topic: Threading with OpenMP
Replies: 24
Views: 9755

Re: Threading with OpenMP

I would like to recommend you the following "OpenMP Microbenchmark" to evaluate you PCs.

http://www2.epcc.ed.ac.uk/computing/res ... index.html

The benchmark is available for Fortran and C. Let us know about results which you get.

Michal
by Michal Kvasnicka
Tue Jan 18, 2011 3:15 pm
Forum: Development
Topic: GPGPU
Replies: 4
Views: 1751

Re: GPGPU

ad1) The forthcoming release of the PGI compiler will not require HW-dependent coding on source level. For more info see: http://www.pgroup.com/about/news.htm#42 In general, you are right, the current state of GPU computing is suitable for photon transport more than for neutron transport. On the oth...
by Michal Kvasnicka
Tue Jan 18, 2011 1:38 pm
Forum: Parallelization
Topic: Threading with OpenMP
Replies: 24
Views: 9755

Re: Threading with OpenMP

First: the example of simple OpenMP code could be better starting point than your description of problem Second: Wikipedia OpenMP page is not very good tutorial. Take a look on: https://computing.llnl.gov/tutorials/openMP/ ,for example. The best source of relevant information is the following book (...
by Michal Kvasnicka
Mon Jan 17, 2011 10:46 am
Forum: Development
Topic: GPGPU
Replies: 4
Views: 1751

GPGPU

I would like to suggest you get some information about GPU-CUDA computing possibilities. There is a lot of references about very successful acceleration of the MonteCarlo simulation in general with speedup ratio about 100 and more. I am not C/C++ expert but I am using last few month GPU toolbox Jack...