#!/bin/sh
#SBATCH --account=gm2dwf
#SBATCH --nodes=16
#SBATCH --ntasks=64
#SBATCH --ntasks-per-node=4
#SBATCH --cpus-per-task=12
#SBATCH --time=0:30:00
#SBATCH --partition=booster
#SBATCH --gres=gpu:4

export OMP_NUM_THREADS=4
export OMPI_MCA_btl=^uct,openib
export UCX_TLS=gdr_copy,rc,rc_x,sm,cuda_copy,cuda_ipc
export UCX_RNDV_SCHEME=put_zcopy
export UCX_RNDV_THRESH=16384
export UCX_IB_GPU_DIRECT_RDMA=yes
export UCX_MEMTYPE_CACHE=n
OPT="--comms-overlap --comms-concurrent"


srun -N 16  -n $SLURM_NTASKS \
	./benchmarks/Benchmark_dwf_fp32 \
	$OPT \
	--mpi 2.2.2.8 \
	--accelerator-threads 8 \
	--grid 64.64.64.256 \
	--shm 2048 > dwf.16node.perf