portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-20 00:36:55 +01:00
Code Releases Activity

Aurora MPI standalone benchmake and options that work well

Browse Source
This commit is contained in:
Peter Boyle
2024-02-06 16:28:40 +00:00
parent 2a0d75bac2
commit 5bfa88be85
9 changed files with 426 additions and 118 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
MPI_benchmark/bench2.pbs Normal file
View File

@ -0,0 +1,22 @@
#!/bin/bash
#PBS -q EarlyAppAccess
#PBS -l select=2
#PBS -l walltime=01:00:00
#PBS -A LatticeQCD_aesp_CNDA
export TZ='/usr/share/zoneinfo/US/Central'
export OMP_PROC_BIND=spread
export OMP_NUM_THREADS=3
unset OMP_PLACES
cd $PBS_O_WORKDIR
NNODES=`wc -l < $PBS_NODEFILE`
NRANKS=12 # Number of MPI ranks per node
NDEPTH=4 # Number of hardware threads per rank, spacing between MPI ranks on a node
NTHREADS=$OMP_NUM_THREADS # Number of OMP threads per rank, given to OMP_NUM_THREADS
NTOTRANKS=$(( NNODES * NRANKS ))
CMD="mpiexec -np 2 -ppn 1 -envall ./gpu_tile_compact.sh ./halo_mpi --mpi 2.1.1.1"
$CMD