portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-18 07:47:06 +01:00
Code Releases Activity

Crusher updates

Browse Source
This commit is contained in:
Peter Boyle
2022-05-25 12:36:09 -04:00
parent 239afb18fb
commit 136d843ce7
8 changed files with 62 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
systems/Crusher/dwf8.slurm
View File

@ -12,37 +12,30 @@
#SBATCH --gpu-bind=map_gpu:0,1,2,3,7,6,5,4
DIR=.
module list
source setup.sh
export MPICH_OFI_NIC_POLICY=GPU
export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
export MPICH_GPU_SUPPORT_ENABLED=1
#export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
#export MPICH_SMP_SINGLE_COPY_MODE=CMA
export MPICH_SMP_SINGLE_COPY_MODE=NONE
#export MPICH_SMP_SINGLE_COPY_MODE=NONE
export OMP_NUM_THREADS=1
echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
PARAMS=" --accelerator-threads 16 --grid 64.64.64.256 --mpi 2.2.2.8 --comms-overlap --shm 2048 --shm-mpi 0"
for vol in 64.64.64.256 64.64.64.128 32.32.32.256 32.32.32.128
do
PARAMS=" --accelerator-threads 8 --grid $vol --mpi 2.2.2.8 --comms-overlap --shm 2048 --shm-mpi 1"
echo $PARAMS
#srun --gpus-per-task 1 -N8 -n64 ./benchmarks/Benchmark_dwf_fp32 $PARAMS > dwf.64.64.64.256.8node
srun --gpus-per-task 1 -N8 -n64 ./benchmarks/Benchmark_dwf_fp32 $PARAMS > dwf.${vol}.8node.shm-mpi1
done
PARAMS=" --accelerator-threads 16 --grid 64.64.64.32 --mpi 4.4.4.1 --comms-overlap --shm 2048 --shm-mpi 1"
PARAMS=" --accelerator-threads 8 --grid 64.64.64.32 --mpi 2.2.2.8 --comms-overlap --shm 2048 --shm-mpi 1"
echo $PARAMS
srun --gpus-per-task 1 -N8 -n64 ./benchmarks/Benchmark_dwf_fp32 $PARAMS > dwf.64.64.64.32.8node
srun --gpus-per-task 1 -N8 -n64 ./benchmarks/Benchmark_ITT $PARAMS > itt.8node
PARAMS=" --accelerator-threads 16 --grid 64.64.64.32 --mpi 4.4.4.1 --comms-overlap --shm 2048 --shm-mpi 0"
PARAMS=" --accelerator-threads 8 --grid 64.64.64.32 --mpi 2.2.2.8 --comms-overlap --shm 2048 --shm-mpi 0"
echo $PARAMS
#srun --gpus-per-task 1 -N8 -n64 ./benchmarks/Benchmark_dwf_fp32 $PARAMS > dwf.64.64.64.32.8node.shm0
PARAMS=" --accelerator-threads 16 --grid 64.64.64.32 --mpi 2.2.2.8 --comms-overlap --shm 2048 --shm-mpi 1"
echo $PARAMS
#srun --gpus-per-task 1 -N8 -n64 ./benchmarks/Benchmark_ITT $PARAMS > itt.8node
PARAMS=" --accelerator-threads 16 --grid 64.64.64.32 --mpi 2.2.2.8 --comms-overlap --shm 2048 --shm-mpi 0"
echo $PARAMS
#srun --gpus-per-task 1 -N8 -n64 ./benchmarks/Benchmark_ITT $PARAMS > itt.8node_shm0
srun --gpus-per-task 1 -N8 -n64 ./benchmarks/Benchmark_ITT $PARAMS > itt.8node_shm0