|
[Sponsors] |
![]() |
![]() |
#1 |
Guest
Posts: n/a
|
Hi all,
Can some one guide me to parallelize a fortran sequential code? Thanks |
|
![]() |
![]() |
![]() |
#2 |
Guest
Posts: n/a
|
Please try to give some details and be specific about what you want. F90/95 have some parallel intrinsics, but you'll probably want to use OpenMP if you're interested in shared-memory parallelism. (Some compilers can even figure some limited cases out automatically.) Annotating for parallelism with OpenMP is normally very easy, profile to see what takes time and annotate those loops. Depending on the memory requirements of the work-load, shared memory parallelism could give you no speedup or nearly optimal speedup for a few processors.
For larger scale or memory-limited workloads, you'll want to use distributed memory which means MPI (though perhaps using a library like PETSc). If want to do this, now would be a very good time to consider a major redesign of your program (though you should be able to reuse core parts like flux computations). |
|
![]() |
![]() |
|
|
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
Discontinuous Galerkin schemes -- fortran source code | diedro | Main CFD Forum | 3 | March 12, 2011 10:20 |
is there any parallel code for the famous Lid Driven Cavity flow? | gholamghar | Main CFD Forum | 0 | August 1, 2010 01:55 |
fortran code problem | ztdep | Main CFD Forum | 5 | September 8, 2009 04:22 |
Fortran 90 faster than C/C++ | B. R. Guirguis | Main CFD Forum | 48 | March 6, 2006 13:49 |
Speeding my Fortran Code | S | Main CFD Forum | 11 | March 31, 2005 14:50 |