portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

Script update

Browse Source
This commit is contained in:
Peter Boyle 2023-03-27 17:29:43 -07:00
parent 8feedb4f6f
commit 0efa107cb6
3 changed files with 12 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
systems/PVC/benchmarks/run-1tile.sh
View File

@ -4,7 +4,7 @@
#SBATCH -p QZ1J-ICX-PVC #SBATCH -p QZ1J-ICX-PVC
##SBATCH -p QZ1J-SPR-PVC-2C ##SBATCH -p QZ1J-SPR-PVC-2C
source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh #source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
export NT=8 export NT=8

14
systems/PVC/benchmarks/run-2tile-mpi.sh
View File

@ -4,7 +4,7 @@
#SBATCH -p QZ1J-ICX-PVC #SBATCH -p QZ1J-ICX-PVC
source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh #source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
export NT=16 export NT=16
@ -19,16 +19,14 @@ export SYCL_DEVICE_FILTER=gpu,level_zero
export I_MPI_OFFLOAD_CELL=tile export I_MPI_OFFLOAD_CELL=tile
export EnableImplicitScaling=0 export EnableImplicitScaling=0
export EnableWalkerPartition=0 export EnableWalkerPartition=0
export SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS=1 #export SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS=1
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 #export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
export SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE=0 export SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE=0
for i in 0 for i in 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
do do
mpiexec -launcher ssh -n 2 -host localhost ./wrap4gpu.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT --shm-mpi 1 --device-mem 32768 mpiexec -launcher ssh -n 2 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT --shm-mpi 0 --device-mem 32768 > 1.1.1.2.log$i
mpiexec -launcher ssh -n 2 -host localhost ./wrap4gpu.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --shm-mpi 1 --device-mem 32768 mpiexec -launcher ssh -n 2 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --shm-mpi 0 --device-mem 32768 > 2.1.1.1.log$i
done done
#mpiexec -launcher ssh -n 2 -host localhost ./wrap4gpu.sh ./Benchmark_halo --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT --shm-mpi 1 > halo.2tile.1x2.log
#mpiexec -launcher ssh -n 2 -host localhost ./wrap4gpu.sh ./Benchmark_halo --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --shm-mpi 1 > halo.2tile.2x1.log

10
systems/PVC/benchmarks/wrap.sh
View File

@ -5,10 +5,10 @@ export ZE_AFFINITY_MASK=0.$MPI_LOCALRANKID
echo Ranke $MPI_LOCALRANKID ZE_AFFINITY_MASK is $ZE_AFFINITY_MASK echo Ranke $MPI_LOCALRANKID ZE_AFFINITY_MASK is $ZE_AFFINITY_MASK
if [ $MPI_LOCALRANKID = "0" ] #if [ $MPI_LOCALRANKID = "0" ]
then #then
# ~psteinbr/build_pti/ze_tracer -h $@ # ~psteinbr/build_pti/ze_tracer -h $@
onetrace --chrome-device-timeline $@ # onetrace --chrome-device-timeline $@
else #else
$@ $@
fi #fi