mirror of
https://github.com/paboyle/Grid.git
synced 2024-11-14 09:45:36 +00:00
25 lines
1.1 KiB
Bash
25 lines
1.1 KiB
Bash
#!/bin/bash
|
|
#SBATCH -A mp13
|
|
#SBATCH -C gpu
|
|
#SBATCH -q regular
|
|
#SBATCH -t 0:20:00
|
|
#SBATCH -n 16
|
|
#SBATCH --ntasks-per-node=4
|
|
#SBATCH -c 32
|
|
#SBATCH --exclusive
|
|
#SBATCH --gpus-per-task=1
|
|
#SBATCH --gpu-bind=map_gpu:0,1,2,3
|
|
|
|
export SLURM_CPU_BIND="cores"
|
|
export MPICH_RDMA_ENABLED_CUDA=1
|
|
export MPICH_GPU_SUPPORT_ENABLED=1
|
|
srun ./benchmarks/Benchmark_comms_host_device --mpi 2.2.2.2 --accelerator-threads 8 > comms.4node
|
|
|
|
OPT="--comms-overlap --comms-concurrent --shm-mpi 0"
|
|
srun ./benchmarks/Benchmark_dwf_fp32 --mpi 2.2.2.2 --grid 64.64.64.64 --accelerator-threads 8 --shm 2048 $OPT > dwf.64.64.64.64.4node.opt0
|
|
srun ./benchmarks/Benchmark_dwf_fp32 --mpi 2.2.2.2 --grid 48.48.48.48 --accelerator-threads 8 --shm 2048 $OPT > dwf.48.48.48.48.4node.opt0
|
|
|
|
OPT="--comms-overlap --comms-concurrent --shm-mpi 1"
|
|
srun ./benchmarks/Benchmark_dwf_fp32 --mpi 2.2.2.2 --grid 64.64.64.64 --accelerator-threads 8 --shm 2048 $OPT > dwf.64.64.64.64.4node.opt1
|
|
srun ./benchmarks/Benchmark_dwf_fp32 --mpi 2.2.2.2 --grid 48.48.48.48 --accelerator-threads 8 --shm 2048 $OPT > dwf.48.48.48.48.4node.opt1
|