diff options
| author | Tor Aamodt <[email protected]> | 2010-10-01 08:55:28 -0800 |
|---|---|---|
| committer | Tor Aamodt <[email protected]> | 2010-10-01 08:55:28 -0800 |
| commit | 11b308e7363e937966b035b4891db32b4eece3bf (patch) | |
| tree | 50ca4c9ad6f163ac4acb2bf505e64dfebed66947 /benchmarks/CUDA/DG/readme | |
| parent | bb820c116764d7a1b8e071137d32b74e7f34dd2f (diff) | |
integrating recent changes from fermi-test into fermi
(i'll use "fermi" for more disruptive changes to the pipeline model such
as updating the MSHRs and getting rid of the warp tracker, ripping out DWF, etc...)
[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 7805]
Diffstat (limited to 'benchmarks/CUDA/DG/readme')
| -rw-r--r-- | benchmarks/CUDA/DG/readme | 73 |
1 files changed, 0 insertions, 73 deletions
diff --git a/benchmarks/CUDA/DG/readme b/benchmarks/CUDA/DG/readme deleted file mode 100644 index cd2e7e6..0000000 --- a/benchmarks/CUDA/DG/readme +++ /dev/null @@ -1,73 +0,0 @@ -# ------------------------------------------------------------------------ -# see the copyright information for MIDG in license.midg -# MIDG: a MIni Discontinuous Galerkin solver for CPUs or GPUS with MPI parallelism -# Copyright (c) 2008, Tim Warburton -# http://www.nudg.org -# ------------------------------------------------------------------------ - -# The following simulation codes all run in parallel -# using ParMetis for mesh partitioning and MPI for -# parallel computation. - -# First compile the ParMetis library -cd 3rdParty/ParMetis-3.1/ -make - -# you may need to adjust the paths for the MPI compilers in the Makefiles - -# ------------------------------------------------------------------------ - -# To compile the CPU/MPI only codes: - -# CPU+MPI+2d time-domain Maxwells (TM mode) with N=6 -make -f MakefileCPU2d N=6 clean -make -f MakefileCPU2d N=6 - -# CPU+MPI+3d timw-domain Maxwells with N=6 -make -f MakefileCPU3d N=6 clean -make -f MakefileCPU3d N=6 - -# ------------------------------------------------------------------------ - -# To compile the GPU/MPI only codes: - -# NOTE FOR GPU CODES: -# 1. you may wish to change the device assignment made -# near the start of the main function in MaxwellsDriver#d.m -# 2. you will need to have nvidia's cuda framework installed -# and may need to adjust the path that the Makefiles use -# to search for the nvcc compiler, libraries and include -# files - -# GPU+MPI+2d time-domain Maxwells (TM mode) with N=6 -make -f MakefileGPU2d N=6 clean -make -f MakefileGPU2d N=6 - -# GPU+MPI+3d timw-domain Maxwells with N=6 -make -f MakefileGPU3d N=6 clean -make -f MakefileGPU3d N=6 - - -# ------------------------------------------------------------------------ - -# To run the various configurations (in serial) with sample meshes: - -./MaxwellsCPU2d Meshes/block2.neu -./MaxwellsCPU3d Meshes/cubeK268.neu -./MaxwellsGPU2d Meshes/block2.neu -./MaxwellsGPU3d Meshes/cubeK268.neu - -# ------------------------------------------------------------------------ - -# To run the various configurations (in parallel using MPI) with sample meshes: - -mpirun -np 2 ./MaxwellsCPU2d Meshes/block2.neu -mpirun -np 2 ./MaxwellsCPU3d Meshes/cubeK268.neu - -mpirun -np 2 ./MaxwellsGPU2d Meshes/block2.neu -mpirun -np 2 ./MaxwellsGPU3d Meshes/cubeK268.neu - -# When running with multiple GPUs you may wish to adjust the device -# used by each cpu. - -# ------------------------------------------------------------------------ |