#!/bin/bash #PBS -l select=512 #PBS -q EarlyAppAccess #PBS -A LatticeQCD_aesp_CNDA #PBS -l walltime=6:00:00 #PBS -N reproBigJob #PBS -k doe #export OMP_PROC_BIND=spread #unset OMP_PLACES #module load oneapi/eng-compiler/2023.05.15.003 #module load mpich/51.2/icc-all-deterministic-pmix-gpu # 56 cores / 6 threads ~9 export OMP_NUM_THREADS=6 export MPIR_CVAR_CH4_OFI_ENABLE_GPU_PIPELINE=1 export MPIR_CVAR_CH4_OFI_GPU_PIPELINE_D2H_ENGINE_TYPE=0 export MPIR_CVAR_CH4_OFI_GPU_PIPELINE_H2D_ENGINE_TYPE=0 export MPIR_CVAR_CH4_OFI_GPU_PIPELINE_BUFFER_SZ=10485760 export MPIR_CVAR_CH4_OFI_GPU_PIPELINE_THRESHOLD=131072 export MPIR_CVAR_CH4_OFI_GPU_PIPELINE_NUM_BUFFERS_PER_CHUNK=16 export MPIR_CVAR_CH4_OFI_GPU_PIPELINE_MAX_NUM_BUFFERS=16 #export MPIR_CVAR_GPU_USE_IMMEDIATE_COMMAND_LIST=1 #export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 export SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE=1 export SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE_FOR_D2D_COPY=1 export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file" export GRID_PRINT_ENTIRE_LOG=0 export GRID_CHECKSUM_RECV_BUF=0 export GRID_CHECKSUM_SEND_BUF=0 export MPICH_OFI_NIC_POLICY=GPU #export MPIR_CVAR_ALLREDUCE_DEVICE_COLLECTIVE=0 #export MPIR_CVAR_REDUCE_DEVICE_COLLECTIVE=0 #export MPIR_CVAR_ALLREDUCE_INTRA_ALGORITHM=recursive_doubling #unset MPIR_CVAR_CH4_COLL_SELECTION_TUNING_JSON_FILE #unset MPIR_CVAR_COLL_SELECTION_TUNING_JSON_FILE #unset MPIR_CVAR_CH4_POSIX_COLL_SELECTION_TUNING_JSON_FILE cd $PBS_O_WORKDIR cp $PBS_NODEFILE nodefile DIR=reproBigJob.$PBS_JOBID mkdir -p $DIR cd $DIR cp $PBS_NODEFILE nodefile BINARY=../Test_dwf_mixedcg_prec echo > pingjob < command-line env > environment $CMD grep Oops Grid.stderr.* > failures.$PBS_JOBID rm core.*