From 7bceeab6f7f58693eb4f760abb32420dee7420cf Mon Sep 17 00:00:00 2001 From: Emilien Bauer Date: Wed, 4 Oct 2023 16:33:21 +0100 Subject: [PATCH 1/6] Get those slurm batch files out of Archer2. --- fast/slurm/job-devito-xdsl-1.slurm | 61 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl-128.slurm | 62 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl-16.slurm | 62 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl-2.slurm | 62 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl-32.slurm | 62 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl-4.slurm | 61 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl-64.slurm | 62 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl-8.slurm | 61 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl-wave.slurm | 61 +++++++++++++++++++++++++ fast/slurm/job-devito-xdsl.slurm | 59 ++++++++++++++++++++++++ fast/slurm/wave-job-nn1.slurm | 65 +++++++++++++++++++++++++++ fast/slurm/wave-job-nn128.slurm | 62 +++++++++++++++++++++++++ fast/slurm/wave-job-nn16.slurm | 62 +++++++++++++++++++++++++ fast/slurm/wave-job-nn2.slurm | 62 +++++++++++++++++++++++++ fast/slurm/wave-job-nn32.slurm | 62 +++++++++++++++++++++++++ fast/slurm/wave-job-nn4.slurm | 61 +++++++++++++++++++++++++ fast/slurm/wave-job-nn64.slurm | 61 +++++++++++++++++++++++++ fast/slurm/wave-job-nn8.slurm | 61 +++++++++++++++++++++++++ 18 files changed, 1109 insertions(+) create mode 100644 fast/slurm/job-devito-xdsl-1.slurm create mode 100644 fast/slurm/job-devito-xdsl-128.slurm create mode 100644 fast/slurm/job-devito-xdsl-16.slurm create mode 100644 fast/slurm/job-devito-xdsl-2.slurm create mode 100644 fast/slurm/job-devito-xdsl-32.slurm create mode 100644 fast/slurm/job-devito-xdsl-4.slurm create mode 100644 fast/slurm/job-devito-xdsl-64.slurm create mode 100644 fast/slurm/job-devito-xdsl-8.slurm create mode 100644 fast/slurm/job-devito-xdsl-wave.slurm create mode 100644 fast/slurm/job-devito-xdsl.slurm create mode 100644 fast/slurm/wave-job-nn1.slurm create mode 100644 fast/slurm/wave-job-nn128.slurm create mode 100644 fast/slurm/wave-job-nn16.slurm create mode 100644 fast/slurm/wave-job-nn2.slurm create mode 100644 fast/slurm/wave-job-nn32.slurm create mode 100644 fast/slurm/wave-job-nn4.slurm create mode 100644 fast/slurm/wave-job-nn64.slurm create mode 100644 fast/slurm/wave-job-nn8.slurm diff --git a/fast/slurm/job-devito-xdsl-1.slurm b/fast/slurm/job-devito-xdsl-1.slurm new file mode 100644 index 0000000000..2da2be2c90 --- /dev/null +++ b/fast/slurm/job-devito-xdsl-1.slurm @@ -0,0 +1,61 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:15:00 +#SBATCH --nodes=1 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-1-full.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/job-devito-xdsl-128.slurm b/fast/slurm/job-devito-xdsl-128.slurm new file mode 100644 index 0000000000..d71205364a --- /dev/null +++ b/fast/slurm/job-devito-xdsl-128.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:15:00 +#SBATCH --nodes=128 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-128.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/job-devito-xdsl-16.slurm b/fast/slurm/job-devito-xdsl-16.slurm new file mode 100644 index 0000000000..624128d4d5 --- /dev/null +++ b/fast/slurm/job-devito-xdsl-16.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=16 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-16.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/job-devito-xdsl-2.slurm b/fast/slurm/job-devito-xdsl-2.slurm new file mode 100644 index 0000000000..c9bbfbe621 --- /dev/null +++ b/fast/slurm/job-devito-xdsl-2.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=2 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-2.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/job-devito-xdsl-32.slurm b/fast/slurm/job-devito-xdsl-32.slurm new file mode 100644 index 0000000000..2a30c8d150 --- /dev/null +++ b/fast/slurm/job-devito-xdsl-32.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=32 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-32-full.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/job-devito-xdsl-4.slurm b/fast/slurm/job-devito-xdsl-4.slurm new file mode 100644 index 0000000000..dc6bd71574 --- /dev/null +++ b/fast/slurm/job-devito-xdsl-4.slurm @@ -0,0 +1,61 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=4 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-4-full.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/job-devito-xdsl-64.slurm b/fast/slurm/job-devito-xdsl-64.slurm new file mode 100644 index 0000000000..fa872a7261 --- /dev/null +++ b/fast/slurm/job-devito-xdsl-64.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:12:00 +#SBATCH --nodes=64 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-64.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/job-devito-xdsl-8.slurm b/fast/slurm/job-devito-xdsl-8.slurm new file mode 100644 index 0000000000..ea435be7e1 --- /dev/null +++ b/fast/slurm/job-devito-xdsl-8.slurm @@ -0,0 +1,61 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=8 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-8.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 diffusion_3D_wBCs.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/job-devito-xdsl-wave.slurm b/fast/slurm/job-devito-xdsl-wave.slurm new file mode 100644 index 0000000000..98ac935e15 --- /dev/null +++ b/fast/slurm/job-devito-xdsl-wave.slurm @@ -0,0 +1,61 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=1 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-1-full.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +# srun --distribution=block:block --hint=nomultithread python3 run_benchmark.py 2d5pt -nt 100 --xdsl --devito --openmp --mpi -d 2000 2000 --repeat 1 +srun --distribution=block:block --hint=nomultithread python3 nd_nwave_devito_nodamp.py -d 300 300 300 --nt 100 + diff --git a/fast/slurm/job-devito-xdsl.slurm b/fast/slurm/job-devito-xdsl.slurm new file mode 100644 index 0000000000..fecbdc0f6a --- /dev/null +++ b/fast/slurm/job-devito-xdsl.slurm @@ -0,0 +1,59 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=1 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-1-full.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +srun --distribution=block:block --hint=nomultithread python3 run_benchmark.py 2d5pt -nt 100 --xdsl --devito --openmp --mpi -d 2000 2000 --repeat 1 diff --git a/fast/slurm/wave-job-nn1.slurm b/fast/slurm/wave-job-nn1.slurm new file mode 100644 index 0000000000..09345ef9ba --- /dev/null +++ b/fast/slurm/wave-job-nn1.slurm @@ -0,0 +1,65 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:20:00 +#SBATCH --nodes=1 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-1-full.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export TMPDIR=/work/d011/d011/shared/temp/ + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + + +# DEVITO_MPI=0 python setup_wave3d.py -d 1024 1024 1024 --nt 512 -so 4 +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/wave-job-nn128.slurm b/fast/slurm/wave-job-nn128.slurm new file mode 100644 index 0000000000..2fd2e19d21 --- /dev/null +++ b/fast/slurm/wave-job-nn128.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=128 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-128.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=diag2 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/wave-job-nn16.slurm b/fast/slurm/wave-job-nn16.slurm new file mode 100644 index 0000000000..f5eb65befa --- /dev/null +++ b/fast/slurm/wave-job-nn16.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=16 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-16.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/wave-job-nn2.slurm b/fast/slurm/wave-job-nn2.slurm new file mode 100644 index 0000000000..69f5345ad4 --- /dev/null +++ b/fast/slurm/wave-job-nn2.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:20:00 +#SBATCH --nodes=2 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-2.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/wave-job-nn32.slurm b/fast/slurm/wave-job-nn32.slurm new file mode 100644 index 0000000000..302f4b1608 --- /dev/null +++ b/fast/slurm/wave-job-nn32.slurm @@ -0,0 +1,62 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:15:00 +#SBATCH --nodes=32 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-32.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/wave-job-nn4.slurm b/fast/slurm/wave-job-nn4.slurm new file mode 100644 index 0000000000..6a3d2b8e56 --- /dev/null +++ b/fast/slurm/wave-job-nn4.slurm @@ -0,0 +1,61 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=4 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-4-full.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/wave-job-nn64.slurm b/fast/slurm/wave-job-nn64.slurm new file mode 100644 index 0000000000..27a14f9dd7 --- /dev/null +++ b/fast/slurm/wave-job-nn64.slurm @@ -0,0 +1,61 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:10:00 +#SBATCH --nodes=64 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-64.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive + +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 diff --git a/fast/slurm/wave-job-nn8.slurm b/fast/slurm/wave-job-nn8.slurm new file mode 100644 index 0000000000..69008296b9 --- /dev/null +++ b/fast/slurm/wave-job-nn8.slurm @@ -0,0 +1,61 @@ +#!/bin/bash + +# Slurm job options (job-name, compute nodes, job time) +#SBATCH --job-name=Devito_MPI_Job +#SBATCH --time=00:08:00 +#SBATCH --nodes=8 +#SBATCH --ntasks-per-node=8 +#SBATCH --cpus-per-task=16 +#SBATCH --switches=1@360 # Each group has 128 nodes + +# Replace [budget code] below with your project code (e.g. t01) +#SBATCH --account=d011 +#SBATCH --partition=standard +#SBATCH --qos=standard +#SBATCH -o ./jobs-output/output-8.%j.out # STDOUT + +# Propagate the cpus-per-task setting from script to srun commands +# By default, Slurm does not propagate this setting from the sbatch +# options to srun commands in the job script. If this is not done, +# process/thread pinning may be incorrect leading to poor performance +export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK + +export SHARED=/work/d011/d011/shared +module use $SHARED/modules +module load sc-23 +module load cray-mpich + + +cd $SHARED/software/devito/fast + +# Set the number of threads to 16 and specify placement +# There are 16 OpenMP threads per MPI process +# We want one thread per physical core +export OMP_NUM_THREADS=16 +export OMP_PLACES=cores + +# Devito-specific env variables +export DEVITO_ARCH=cray +export DEVITO_LANGUAGE=openmp +export DEVITO_LOGGING=DEBUG +export DEVITO_MPI=1 +export DEVITO_AUTOTUNING=aggressive +# export DEVITO_PROFILING=advanced2 + +# Archer specific +# export MPICH_OFI_STARTUP_CONNECT=1 +# export MPICH_OFI_RMA_STARTUP_CONNECT=1 +export FI_OFI_RXM_SAR_LIMIT=524288 +export FI_OFI_RXM_BUFFER_SIZE=131072 +export MPICH_SMP_SINGLE_COPY_SIZE=16384 +export CRAY_OMP_CHECK_AFFINITY=TRUE +export SLURM_CPU_FREQ_REQ=2250000 + +# Launch the parallel job +# Using nodes x ntasks-per-node MPI processes +# 8 MPI processes per node +# 16 OpenMP threads per MPI process +# Additional srun options to pin one thread per physical core + +DEVITO_MPI=diag2 srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --devito 1 +srun --distribution=block:block --hint=nomultithread python3 wave3d_b.py -d 1024 1024 1024 --nt 512 -so 4 --xdsl 1 From b34968ab8446010ea1c973137d26a693c1016dd3 Mon Sep 17 00:00:00 2001 From: Emilien Bauer Date: Mon, 9 Oct 2023 12:46:09 +0100 Subject: [PATCH 2/6] Clarity renames. --- fast/slurm/{job-devito-xdsl-1.slurm => diffusion-1.slurm} | 2 +- fast/slurm/{job-devito-xdsl-128.slurm => diffusion-128.slurm} | 2 +- fast/slurm/{job-devito-xdsl-16.slurm => diffusion-16.slurm} | 2 +- fast/slurm/{job-devito-xdsl-2.slurm => diffusion-2.slurm} | 2 +- fast/slurm/{job-devito-xdsl-32.slurm => diffusion-32.slurm} | 2 +- fast/slurm/{job-devito-xdsl-4.slurm => diffusion-4.slurm} | 2 +- fast/slurm/{job-devito-xdsl-64.slurm => diffusion-64.slurm} | 2 +- fast/slurm/{job-devito-xdsl-8.slurm => diffusion-8.slurm} | 2 +- fast/slurm/{wave-job-nn1.slurm => wave-1.slurm} | 2 +- fast/slurm/{wave-job-nn128.slurm => wave-128.slurm} | 2 +- fast/slurm/{wave-job-nn16.slurm => wave-16.slurm} | 2 +- fast/slurm/{wave-job-nn2.slurm => wave-2.slurm} | 2 +- fast/slurm/{wave-job-nn32.slurm => wave-32.slurm} | 2 +- fast/slurm/{wave-job-nn4.slurm => wave-4.slurm} | 2 +- fast/slurm/{wave-job-nn64.slurm => wave-64.slurm} | 2 +- fast/slurm/{wave-job-nn8.slurm => wave-8.slurm} | 2 +- 16 files changed, 16 insertions(+), 16 deletions(-) rename fast/slurm/{job-devito-xdsl-1.slurm => diffusion-1.slurm} (97%) rename fast/slurm/{job-devito-xdsl-128.slurm => diffusion-128.slurm} (97%) rename fast/slurm/{job-devito-xdsl-16.slurm => diffusion-16.slurm} (97%) rename fast/slurm/{job-devito-xdsl-2.slurm => diffusion-2.slurm} (97%) rename fast/slurm/{job-devito-xdsl-32.slurm => diffusion-32.slurm} (97%) rename fast/slurm/{job-devito-xdsl-4.slurm => diffusion-4.slurm} (97%) rename fast/slurm/{job-devito-xdsl-64.slurm => diffusion-64.slurm} (97%) rename fast/slurm/{job-devito-xdsl-8.slurm => diffusion-8.slurm} (97%) rename fast/slurm/{wave-job-nn1.slurm => wave-1.slurm} (97%) rename fast/slurm/{wave-job-nn128.slurm => wave-128.slurm} (97%) rename fast/slurm/{wave-job-nn16.slurm => wave-16.slurm} (97%) rename fast/slurm/{wave-job-nn2.slurm => wave-2.slurm} (97%) rename fast/slurm/{wave-job-nn32.slurm => wave-32.slurm} (97%) rename fast/slurm/{wave-job-nn4.slurm => wave-4.slurm} (97%) rename fast/slurm/{wave-job-nn64.slurm => wave-64.slurm} (97%) rename fast/slurm/{wave-job-nn8.slurm => wave-8.slurm} (97%) diff --git a/fast/slurm/job-devito-xdsl-1.slurm b/fast/slurm/diffusion-1.slurm similarity index 97% rename from fast/slurm/job-devito-xdsl-1.slurm rename to fast/slurm/diffusion-1.slurm index 2da2be2c90..0d38fdbd15 100644 --- a/fast/slurm/job-devito-xdsl-1.slurm +++ b/fast/slurm/diffusion-1.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-1-full.%j.out # STDOUT +#SBATCH -o ./jobs-output/diffusion-1.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/job-devito-xdsl-128.slurm b/fast/slurm/diffusion-128.slurm similarity index 97% rename from fast/slurm/job-devito-xdsl-128.slurm rename to fast/slurm/diffusion-128.slurm index d71205364a..d78da834dd 100644 --- a/fast/slurm/job-devito-xdsl-128.slurm +++ b/fast/slurm/diffusion-128.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-128.%j.out # STDOUT +#SBATCH -o ./jobs-output/diffusion-128.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/job-devito-xdsl-16.slurm b/fast/slurm/diffusion-16.slurm similarity index 97% rename from fast/slurm/job-devito-xdsl-16.slurm rename to fast/slurm/diffusion-16.slurm index 624128d4d5..540e30aea2 100644 --- a/fast/slurm/job-devito-xdsl-16.slurm +++ b/fast/slurm/diffusion-16.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-16.%j.out # STDOUT +#SBATCH -o ./jobs-output/diffusion-16.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/job-devito-xdsl-2.slurm b/fast/slurm/diffusion-2.slurm similarity index 97% rename from fast/slurm/job-devito-xdsl-2.slurm rename to fast/slurm/diffusion-2.slurm index c9bbfbe621..8d24017ff7 100644 --- a/fast/slurm/job-devito-xdsl-2.slurm +++ b/fast/slurm/diffusion-2.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-2.%j.out # STDOUT +#SBATCH -o ./jobs-output/diffusion-2.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/job-devito-xdsl-32.slurm b/fast/slurm/diffusion-32.slurm similarity index 97% rename from fast/slurm/job-devito-xdsl-32.slurm rename to fast/slurm/diffusion-32.slurm index 2a30c8d150..d24ec70256 100644 --- a/fast/slurm/job-devito-xdsl-32.slurm +++ b/fast/slurm/diffusion-32.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-32-full.%j.out # STDOUT +#SBATCH -o ./jobs-output/diffusion-32.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/job-devito-xdsl-4.slurm b/fast/slurm/diffusion-4.slurm similarity index 97% rename from fast/slurm/job-devito-xdsl-4.slurm rename to fast/slurm/diffusion-4.slurm index dc6bd71574..0aaa718456 100644 --- a/fast/slurm/job-devito-xdsl-4.slurm +++ b/fast/slurm/diffusion-4.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-4-full.%j.out # STDOUT +#SBATCH -o ./jobs-output/diffusion-4.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/job-devito-xdsl-64.slurm b/fast/slurm/diffusion-64.slurm similarity index 97% rename from fast/slurm/job-devito-xdsl-64.slurm rename to fast/slurm/diffusion-64.slurm index fa872a7261..8ecea840b3 100644 --- a/fast/slurm/job-devito-xdsl-64.slurm +++ b/fast/slurm/diffusion-64.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-64.%j.out # STDOUT +#SBATCH -o ./jobs-output/diffusion-64.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/job-devito-xdsl-8.slurm b/fast/slurm/diffusion-8.slurm similarity index 97% rename from fast/slurm/job-devito-xdsl-8.slurm rename to fast/slurm/diffusion-8.slurm index ea435be7e1..3369a0e765 100644 --- a/fast/slurm/job-devito-xdsl-8.slurm +++ b/fast/slurm/diffusion-8.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-8.%j.out # STDOUT +#SBATCH -o ./jobs-output/diffusion-8.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/wave-job-nn1.slurm b/fast/slurm/wave-1.slurm similarity index 97% rename from fast/slurm/wave-job-nn1.slurm rename to fast/slurm/wave-1.slurm index 09345ef9ba..f212a6f2e3 100644 --- a/fast/slurm/wave-job-nn1.slurm +++ b/fast/slurm/wave-1.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-1-full.%j.out # STDOUT +#SBATCH -o ./jobs-output/wave-1.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/wave-job-nn128.slurm b/fast/slurm/wave-128.slurm similarity index 97% rename from fast/slurm/wave-job-nn128.slurm rename to fast/slurm/wave-128.slurm index 2fd2e19d21..d2abb6d154 100644 --- a/fast/slurm/wave-job-nn128.slurm +++ b/fast/slurm/wave-128.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-128.%j.out # STDOUT +#SBATCH -o ./jobs-output/wave-128.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/wave-job-nn16.slurm b/fast/slurm/wave-16.slurm similarity index 97% rename from fast/slurm/wave-job-nn16.slurm rename to fast/slurm/wave-16.slurm index f5eb65befa..c6c547a135 100644 --- a/fast/slurm/wave-job-nn16.slurm +++ b/fast/slurm/wave-16.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-16.%j.out # STDOUT +#SBATCH -o ./jobs-output/wave-16.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/wave-job-nn2.slurm b/fast/slurm/wave-2.slurm similarity index 97% rename from fast/slurm/wave-job-nn2.slurm rename to fast/slurm/wave-2.slurm index 69f5345ad4..70fec5de36 100644 --- a/fast/slurm/wave-job-nn2.slurm +++ b/fast/slurm/wave-2.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-2.%j.out # STDOUT +#SBATCH -o ./jobs-output/wave-2.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/wave-job-nn32.slurm b/fast/slurm/wave-32.slurm similarity index 97% rename from fast/slurm/wave-job-nn32.slurm rename to fast/slurm/wave-32.slurm index 302f4b1608..f234fffc31 100644 --- a/fast/slurm/wave-job-nn32.slurm +++ b/fast/slurm/wave-32.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-32.%j.out # STDOUT +#SBATCH -o ./jobs-output/wave-32.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/wave-job-nn4.slurm b/fast/slurm/wave-4.slurm similarity index 97% rename from fast/slurm/wave-job-nn4.slurm rename to fast/slurm/wave-4.slurm index 6a3d2b8e56..ef7d5f0d6b 100644 --- a/fast/slurm/wave-job-nn4.slurm +++ b/fast/slurm/wave-4.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-4-full.%j.out # STDOUT +#SBATCH -o ./jobs-output/wave-4.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/wave-job-nn64.slurm b/fast/slurm/wave-64.slurm similarity index 97% rename from fast/slurm/wave-job-nn64.slurm rename to fast/slurm/wave-64.slurm index 27a14f9dd7..de9cc146c6 100644 --- a/fast/slurm/wave-job-nn64.slurm +++ b/fast/slurm/wave-64.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-64.%j.out # STDOUT +#SBATCH -o ./jobs-output/wave-64.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch diff --git a/fast/slurm/wave-job-nn8.slurm b/fast/slurm/wave-8.slurm similarity index 97% rename from fast/slurm/wave-job-nn8.slurm rename to fast/slurm/wave-8.slurm index 69008296b9..7b8e130d76 100644 --- a/fast/slurm/wave-job-nn8.slurm +++ b/fast/slurm/wave-8.slurm @@ -12,7 +12,7 @@ #SBATCH --account=d011 #SBATCH --partition=standard #SBATCH --qos=standard -#SBATCH -o ./jobs-output/output-8.%j.out # STDOUT +#SBATCH -o ./jobs-output/wave-8.%j.out # STDOUT # Propagate the cpus-per-task setting from script to srun commands # By default, Slurm does not propagate this setting from the sbatch From 9e475e9ea071581f779c3987ca6ad18b34c49424 Mon Sep 17 00:00:00 2001 From: Emilien Bauer Date: Mon, 9 Oct 2023 13:41:26 +0100 Subject: [PATCH 3/6] slurm -> slurm-jobs --- fast/{slurm => slurm-jobs}/diffusion-1.slurm | 0 fast/{slurm => slurm-jobs}/diffusion-128.slurm | 0 fast/{slurm => slurm-jobs}/diffusion-16.slurm | 0 fast/{slurm => slurm-jobs}/diffusion-2.slurm | 0 fast/{slurm => slurm-jobs}/diffusion-32.slurm | 0 fast/{slurm => slurm-jobs}/diffusion-4.slurm | 0 fast/{slurm => slurm-jobs}/diffusion-64.slurm | 0 fast/{slurm => slurm-jobs}/diffusion-8.slurm | 0 fast/{slurm => slurm-jobs}/job-devito-xdsl-wave.slurm | 0 fast/{slurm => slurm-jobs}/job-devito-xdsl.slurm | 0 fast/{slurm => slurm-jobs}/wave-1.slurm | 0 fast/{slurm => slurm-jobs}/wave-128.slurm | 0 fast/{slurm => slurm-jobs}/wave-16.slurm | 0 fast/{slurm => slurm-jobs}/wave-2.slurm | 0 fast/{slurm => slurm-jobs}/wave-32.slurm | 0 fast/{slurm => slurm-jobs}/wave-4.slurm | 0 fast/{slurm => slurm-jobs}/wave-64.slurm | 0 fast/{slurm => slurm-jobs}/wave-8.slurm | 0 18 files changed, 0 insertions(+), 0 deletions(-) rename fast/{slurm => slurm-jobs}/diffusion-1.slurm (100%) rename fast/{slurm => slurm-jobs}/diffusion-128.slurm (100%) rename fast/{slurm => slurm-jobs}/diffusion-16.slurm (100%) rename fast/{slurm => slurm-jobs}/diffusion-2.slurm (100%) rename fast/{slurm => slurm-jobs}/diffusion-32.slurm (100%) rename fast/{slurm => slurm-jobs}/diffusion-4.slurm (100%) rename fast/{slurm => slurm-jobs}/diffusion-64.slurm (100%) rename fast/{slurm => slurm-jobs}/diffusion-8.slurm (100%) rename fast/{slurm => slurm-jobs}/job-devito-xdsl-wave.slurm (100%) rename fast/{slurm => slurm-jobs}/job-devito-xdsl.slurm (100%) rename fast/{slurm => slurm-jobs}/wave-1.slurm (100%) rename fast/{slurm => slurm-jobs}/wave-128.slurm (100%) rename fast/{slurm => slurm-jobs}/wave-16.slurm (100%) rename fast/{slurm => slurm-jobs}/wave-2.slurm (100%) rename fast/{slurm => slurm-jobs}/wave-32.slurm (100%) rename fast/{slurm => slurm-jobs}/wave-4.slurm (100%) rename fast/{slurm => slurm-jobs}/wave-64.slurm (100%) rename fast/{slurm => slurm-jobs}/wave-8.slurm (100%) diff --git a/fast/slurm/diffusion-1.slurm b/fast/slurm-jobs/diffusion-1.slurm similarity index 100% rename from fast/slurm/diffusion-1.slurm rename to fast/slurm-jobs/diffusion-1.slurm diff --git a/fast/slurm/diffusion-128.slurm b/fast/slurm-jobs/diffusion-128.slurm similarity index 100% rename from fast/slurm/diffusion-128.slurm rename to fast/slurm-jobs/diffusion-128.slurm diff --git a/fast/slurm/diffusion-16.slurm b/fast/slurm-jobs/diffusion-16.slurm similarity index 100% rename from fast/slurm/diffusion-16.slurm rename to fast/slurm-jobs/diffusion-16.slurm diff --git a/fast/slurm/diffusion-2.slurm b/fast/slurm-jobs/diffusion-2.slurm similarity index 100% rename from fast/slurm/diffusion-2.slurm rename to fast/slurm-jobs/diffusion-2.slurm diff --git a/fast/slurm/diffusion-32.slurm b/fast/slurm-jobs/diffusion-32.slurm similarity index 100% rename from fast/slurm/diffusion-32.slurm rename to fast/slurm-jobs/diffusion-32.slurm diff --git a/fast/slurm/diffusion-4.slurm b/fast/slurm-jobs/diffusion-4.slurm similarity index 100% rename from fast/slurm/diffusion-4.slurm rename to fast/slurm-jobs/diffusion-4.slurm diff --git a/fast/slurm/diffusion-64.slurm b/fast/slurm-jobs/diffusion-64.slurm similarity index 100% rename from fast/slurm/diffusion-64.slurm rename to fast/slurm-jobs/diffusion-64.slurm diff --git a/fast/slurm/diffusion-8.slurm b/fast/slurm-jobs/diffusion-8.slurm similarity index 100% rename from fast/slurm/diffusion-8.slurm rename to fast/slurm-jobs/diffusion-8.slurm diff --git a/fast/slurm/job-devito-xdsl-wave.slurm b/fast/slurm-jobs/job-devito-xdsl-wave.slurm similarity index 100% rename from fast/slurm/job-devito-xdsl-wave.slurm rename to fast/slurm-jobs/job-devito-xdsl-wave.slurm diff --git a/fast/slurm/job-devito-xdsl.slurm b/fast/slurm-jobs/job-devito-xdsl.slurm similarity index 100% rename from fast/slurm/job-devito-xdsl.slurm rename to fast/slurm-jobs/job-devito-xdsl.slurm diff --git a/fast/slurm/wave-1.slurm b/fast/slurm-jobs/wave-1.slurm similarity index 100% rename from fast/slurm/wave-1.slurm rename to fast/slurm-jobs/wave-1.slurm diff --git a/fast/slurm/wave-128.slurm b/fast/slurm-jobs/wave-128.slurm similarity index 100% rename from fast/slurm/wave-128.slurm rename to fast/slurm-jobs/wave-128.slurm diff --git a/fast/slurm/wave-16.slurm b/fast/slurm-jobs/wave-16.slurm similarity index 100% rename from fast/slurm/wave-16.slurm rename to fast/slurm-jobs/wave-16.slurm diff --git a/fast/slurm/wave-2.slurm b/fast/slurm-jobs/wave-2.slurm similarity index 100% rename from fast/slurm/wave-2.slurm rename to fast/slurm-jobs/wave-2.slurm diff --git a/fast/slurm/wave-32.slurm b/fast/slurm-jobs/wave-32.slurm similarity index 100% rename from fast/slurm/wave-32.slurm rename to fast/slurm-jobs/wave-32.slurm diff --git a/fast/slurm/wave-4.slurm b/fast/slurm-jobs/wave-4.slurm similarity index 100% rename from fast/slurm/wave-4.slurm rename to fast/slurm-jobs/wave-4.slurm diff --git a/fast/slurm/wave-64.slurm b/fast/slurm-jobs/wave-64.slurm similarity index 100% rename from fast/slurm/wave-64.slurm rename to fast/slurm-jobs/wave-64.slurm diff --git a/fast/slurm/wave-8.slurm b/fast/slurm-jobs/wave-8.slurm similarity index 100% rename from fast/slurm/wave-8.slurm rename to fast/slurm-jobs/wave-8.slurm From 4693ccd52c1e8155348705a036f58e6b8b86df0f Mon Sep 17 00:00:00 2001 From: Emilien Bauer Date: Wed, 11 Oct 2023 10:22:17 +0100 Subject: [PATCH 4/6] Switch to BENCH logging. --- fast/- | Bin 0 -> 16112 bytes fast/async.so | Bin 0 -> 16304 bytes fast/logging.patch | 55 +++++++++++++++++++++ fast/slurm-jobs/diffusion-1.slurm | 2 +- fast/slurm-jobs/diffusion-128.slurm | 2 +- fast/slurm-jobs/diffusion-16.slurm | 2 +- fast/slurm-jobs/diffusion-2.slurm | 2 +- fast/slurm-jobs/diffusion-32.slurm | 2 +- fast/slurm-jobs/diffusion-4.slurm | 2 +- fast/slurm-jobs/diffusion-64.slurm | 2 +- fast/slurm-jobs/diffusion-8.slurm | 2 +- fast/slurm-jobs/job-devito-xdsl-wave.slurm | 2 +- fast/slurm-jobs/job-devito-xdsl.slurm | 2 +- fast/slurm-jobs/wave-1.slurm | 2 +- fast/slurm-jobs/wave-128.slurm | 2 +- fast/slurm-jobs/wave-16.slurm | 2 +- fast/slurm-jobs/wave-2.slurm | 2 +- fast/slurm-jobs/wave-32.slurm | 2 +- fast/slurm-jobs/wave-4.slurm | 2 +- fast/slurm-jobs/wave-64.slurm | 2 +- fast/slurm-jobs/wave-8.slurm | 2 +- 21 files changed, 73 insertions(+), 18 deletions(-) create mode 100755 fast/- create mode 100755 fast/async.so create mode 100644 fast/logging.patch diff --git a/fast/- b/fast/- new file mode 100755 index 0000000000000000000000000000000000000000..a34fa05f31c8725b7b6ae6d1190a8fc0bc1674a6 GIT binary patch literal 16112 zcmeHOeQaA-6+gCL@|?2!8;*Cx1Hs+<<`T?laPLuQnoYJ+;*0PLOH+GY>jvefm zE~zwTD6BPGp#lvi_@^MjstLpf)7UB)Eemah_D4Y-6p4?lu+BzMvJXl@UCcT6-ebQ$ z+tUyqLf{+aoqK-gf=}|)MjcfZ@1(DKq#je2lbvYKFXiB%Mzh43+U@F5bov%4KfUTwddMyL zClqBnV#zM2>~hME>8wf?Q`KEq|FRXOIE|a7UnhOHgn=pN@J_U&_TFyc^leb>RqNjp z)i0(hY#{g@_)olOw~;(bWx~52S-1IvKfQk#s};)(q6;fX$cw)M{tLt^>4z|YO8EU% z^uJRDzp9F#-=jX~3aiCJ{q`5RQYDVnO7T2g1*c1>lKxk#=+k-khB=p3!w?-7GMQSJ=#-w1OEvZg8}Pj2&7_~Bjd@0nYJP+%QV64-!x!G zW2xB4SlWuEHVv$aClj$vk!|sqlvS3P!&4D+`&c3pAKM*+k%>ujXDpS7#ladIkENt{ z7)N}gm5N2i*Wkg531A=@os7rA$w<`3u8moJlZjz#ESV5!ID*{h2c=7LBdrD#aklH- z6-!tfr%@}FOi&D@CBU9o+Dav-RV1-Q6e1H7@o5fFc#DmP;_>9Lj372XJTWb*GFYcq ziLzT|-xFL?9M_P!J&h_WSV?8%{DVT|rqRjB`QA${-y<_-T|?aSR3}L;7@|3nz^|Gv4P#XA8@IjBw$G7x38D8N?d#qa@4E z7O>p#L*i))mOm-@!^G2+FMmYv-zJ`>V)-Wo|2Xk91(jPgBYKdcp4|o~By)px{%)(^M+oCHNi0(-bQ27yLHj zY3h{s3I4;x(-brhfwp;b>KliPw}=Aw8wR`*ThUH4+!#rDZQe;*1}JcDQ&x;u1NXfU*C zBcAwnBRhD)$iCj6ozGE!Uu#>?douJr%116UUnulvk8j5kXzTnWLIC~QUtyQdqC9sk zijC~^M)tU|f7w%nBn9%4vH#9FBYVooo(=E6Yc8BU70!MX0jiE&F!mh1U}V=J{e$

CH^_=-bB|vv6qMWF2*a0h zD}?kf;q2Mm!zc}BU&%csDP6dwe&#z+5ukXD7q4neaIMrbP|H9q1GNm)GEmDvEd#X- z)G|=Z!2ewaO0VMhh1aDictCz1a63JKQ*py{uv=EeF~cu4)Ju<5b?*ZDKEQ*=MtU<7 zNRQ(CH7l|WYfPudB`>I%9!*52AduJ=}B9e-Xi@v|zyx4}&CDjecdBFZ=P>PQU-2P@w;(rvl{SBSO8*u1( z#9KGX^D&5(g+7~q^^#Cq;FaH_pj^9E)?2B*(1u^5!3kWf(BA2Vc9hX4F`h*Q(5|$- zk(;}=$h+uGAo8B8e4sUY#e2}8%LlHTaFwPD{jAW6*AlK`;FCH4E_4Wr5AK5EW4)kh z63VjR2B=%kOpX3~GjLvqdFZ(Y?__%+qdywf=k{i>MVA{!a3q|Vo74Sgq34~@pl}Tq z3mo8XLGRc7^Pw3N$KsWOuXDeGBlE_UZueS%kGBuKjPl@2r&rVdB5<#JZzl5qEN^t* z2m2oI-}9=JgQne$FP42nx{^1zzld_U=cUEM-c<+RE!u;&1jok;$b6!Gi0t7+Cp2l& zug4?XsYgEtJpScpNIugEy7!CVuRAiNq27Ty|8loSw)ch}oQIR0*yRy`jBiNOF>de; zd0;V8W91n=GY_XbLG%aX(>!_|>V*O2LDz-luj&5tFw=q8)|GV_f4E-L^?LuQbGoKC zH6ndd$UC7?JMQ%&-=wW=AnMk1L8ZI{yqaF8A>OR*6?zSt-qMWx_uVRWUgON#lqNd0eYlzdZ@q+ai0IJ8^q!1zUe znwurB*WZWz-aJ(AUYvRJE6p8b???Nl#_tNep+V}kw3O_Frzk&*;o`Tn4z_Qqm-?^j z2Sxk3m;7!SYHB8Y%aGn6`Sbb_k>8dUh3h{T?dZA=tFa$%()q#m-8fz-E~4(Br$grN zm*)!lYymQz&?fJNziorPxE35d18q9$Ke2vRhl?alMV`j?mh* zy=<0DnSHbBl#+|K=~u~Ve+n_|Q<;5K+54sc(r=gIxgT@LV-Cyg_bGdy(r3bUOb?)d za_Y?Cqso50E#Z~o4=VezUuaI4@3 z4HR-rop%3ldy72pdpG){^LP4p`1*k8%Y_0Z{?OS%VIR_GUMUowLQ3T_Nx-#x1GuJq zu0_qd_W%Y>;u$P4R7CsMYtNdreFIz#jeK`u%mE(LU>Ozk!XR4%t&&7g6^NQb67A*7ojJe|PJW z?$)kQ>&j4Tuw?h3+kJwdE+IxGRP}{g?ah;yTKO-?0Drf~-|6vXI-n}d?{x6=Z6KB6 zj8@$s=!GDa->Hw)u97V9_kxQQ&x7G=#q)QAG^VMr{Lk}+qyW^7g|1a9e!76L@cn|n zSLVmeuc-L=+c>^YtJ?_VR!e)nUx_!<2=Vnszj{=zlmwnp4KeOhJjZ!R>G2Z-w} z9#iyZik?;UyrT8$W5X*P-;J(av*x}2C4<}WW7tXm(%S;J2fA)qIw{ztAG-~?J8o34 zGxAG$%DV=yI;Rt29J}!su*?3jBkIz#9pPIY__Ye>>jqg-Xdd8#DHuA=|tD)?hn@TaTb$E)DKLfjF$h=`sBE5&)f ziXRUnubM4%7X!Qht_I!TS)e6wlTw`a2Q7uDn@zQ0U`eRlgT!l+Nb`UT*L; z#4F_ohErDBn#39qvVk)SuG+2mAJgx7yuN zwqe4)vq97=?bNVs!6@G?Avb@}mJ7!&6LKfj)lQJoI;h-(A)0dT0vWJx)3{_yi0D_v zohhmlw4=kkWRptJzR99&M~DoL_juShvEYg?FK_d5cX0uHBrzEnjihnSI69p`HA$@$ z1W2+gmP*rx7CU323>Ro4WT0vjaSL$i8&6>XtAUXuR%lHa0%D*9Qpsq`BI9Gj*nH9=&lsHWvn`zlTn--}$IGPry8TBHnkSyexe#9NvT#Mw zB@US9VXS@wR_G^?&3Jyu^a9>Dsn9%&s%+2mwp)~p?Rmb*l;@YICW>;$EYOAc?j z=gVYIb7rS~3^B?H&(DiDZFTeJ=Kxd@MNQ|Kg{@e#+xwLrQz0yBmwprYw_&H#Ud`Y9 zigLW{f_bJigzfF~ypZX)IdRGXx6A(M-Iv{-=SNKWIf?b1{m&@-Vl#fFsi>jmcM2`0 zm-&5Iq4wFH=T%Id(>rJTUvSt@-7GybJt^7Zk@LpDL9Ec+nV)0$`H%OVI`x@n`VF*g zv)S{!d6q6rEUYKF63_TED4?;&?elz-_vg-1a%^(8r3H*?DVK*?0p*pBIs9riqrp6Zf@tjBJgJe?oHIQ~pfYG$g~Q>^y#A~J+2 z^`eOF)o>`{Y;UJL_KHeac|tjyQ-RU?z!mt<@$>zco})Q-tQO0`J7o1qhjVV173ngs QWPhhxsQoo7D(bZVJCdDqe*gdg literal 0 HcmV?d00001 diff --git a/fast/async.so b/fast/async.so new file mode 100755 index 0000000000000000000000000000000000000000..1d8594a3c66aabdb88b4cbd8ed58c5ca2575a1b7 GIT binary patch literal 16304 zcmeHOeQ;FQbwB$d?do$^G!6=2^z52Qw!^aGLnuxOyMW-ySz(MNTLig}m9+Z6t6g>X z2_xFn+mHZThn99~n`V+3r%gM=(@EU?kvOz*z$RFex*gJ?4u80bOl!P5lVTZSLR+lu zIrrYP`?T8iIO8_`XRk)@+~4n_$2$Ek$`DK(J%UKmHma4n3>rHc*+Sz}vEVo$xmZX6t$M8Y)qxwE* z;_&TO^_A;io9Y)!)ogHJMN9CXu&B4ukGWbeale88AHo0ce>=Ok{i&V(_y6VKdwPF(SJNRG9{sC-IDL!1r{;4wjKa}BH%ao5aX!u{8zJNrj_I$kz z?f%%8b zL?k@16%S@a5IynXsc0k=4-Z@5j)=K^DmG+}$76yGg%OQ?rcjLZ>Wf9$uj@c0WaG&LMH5>qh>fG-*MwRet=n*}YYVi>L|b3B?{pdTXC0Xst7Tecc&eQSLi z^7;0S_$%7fpFN*rhNQ#7lC2&79LEOFJ5$TXYXEKjQgP4x$THW0J9dJ`ykW-Y2-qonl1>p zg{QfalQ$r9WY2RqG0g!w0uV2*Zvp{m0hu9Q%(IGbv+$(j?_$y1S%1zM4%FZN(3qq`1M8Lx@YBo>4$&r)}O38y9Pl2;u+JPyPzLF=YClVOsSPVOla}F9>{^FfGxtX9WH>VOlz6pBMN!!n9<{J}dBF z5T>P4_JqJ+CQM7DY)ariB}~hp?4-bdNSKyH*?j_ko-i%_vbzL6M3|O9*?_1VHQ*3aH_=(h9vkLS&gU;&#{ z0q)%Fh^J*+!JmGw)TVz#HB8;xryt(5jIw(AO|xEqa?>{vWcFYyGsAc|U-egrZI5Ez z`TEH}`#369d-~87e}C@}=oi$FPt9fS$6j93J@@p?(Xs31HIMhU*DyZUJbPZc=IK9+ zW5Y{DegPc-*#DZRsq1FupUBrU&~=lIgg*F4@Zn%zu=gRXs98_<&FJat-Rb$vv%q7AFDgGG^ad-Na5llWVzdQY}7_uX%&zyZPm($ZP>*-hY;~RR3@t}jAenUV0;GCYm zsHd-ljz2sXN?#16e+LQ5j$=#C-qh1OQU2WhVgRc-P4^GqqwWS@(CGL*8jM)#ehBUk zKC&P7+cK|FRfo^{_4Lm&pT&GAeZ41rwLAB!=ZlYF>_X{W=K3uu{~INSsOv)JL2`Ps zYX(Dsk&Im08ypBe8vN~`@#xEi{`BR8v8$V!jDg5yT1C%(_ye%nZF>5AW(VT#^!d{g zzoMsqs;9G}d}VSqbC)prA)AEKIFPa#pD5_<%y+O+dirgFL+Q6e=^LT+<;(?SGk=GN z?({j~i2Vqquj=WunRk&DhF3E8VOACUGvW9~<|jh?%TW4CW`)SUmFbm|CW=Bovo9hG zke^fXvj_&bk}46XM4%FZN(3qqs6?O=fl34_5vWAq|2+avz4q};ov-QB+>pZSG=7!E z8Y!QU-wUkccNKPc5e`{QQvC8ky|ma=^_xK73pnstOK( zSZw$p_+kfqQzNDiU%Mry1sacy#KC7q4w}Bjtt4|YD5FY_g%gqCGRmQN3}0KuOn)hT zA`+#t-`9rp;b?R;5uO|qO`S}{&A2%|84-wz2+Nm^AGZ4?I{v>MkPq77_X(OY8BT;J zMBhJXTr5LqL3TGH2Uy?iN%eWbTi=XH{yzk7eZwaC8XS5K@pepNJ_aze#A_*Ez2tK? zx%2PuoNHvh5pA7IEc^;5%;3WW=k0E2MjgEicP!6=bE)Oc+1wp@Y@;`Mh#epEg0t2w z-eU%AUa;MRPig#E&PHecT*9Xecqa{jPdJ3c8+#$~23yD^DP?L}1yvmkD)Rp}0ykVR z58GDYOm-Afu9G3x+|d-)XtTp8j)a-HIhS??w%rajig!Y*-~k>H@^db2K6uW+vA9&= zD*IbFGVkuR+jju`LG!?ys1KZLaXX#A6ujGhG?jV+I%@5Y!?7o{z3)gpIGvB+i(&78 zOX+Lu&!8UmzR@cDZB_8yygxYGaC~fp)aRQA$RB1}pw21FbvR@_)mY96ht_eK^ygZ@ z<$eaVsxt#lv^!Cyb=V!UzF)Wk^KhXByF3bz@(ws%*lzF+IG`1!-g()Tnukj*Ao_#t zb2?mAXcq$12W&SxuDP@uaBd}@TT833{h?~7%T=viyy|kg>S|H`tb)f2yG zz*Qsh4c8eF--ZUoyM7|-ak*Tu8T;`b{XST~9mfl`i=;i^SSjQ8`qi9kHV3H|@W{FF z-#l;>p9KdlgU5yT&+nRb!7b7=UL?`#!z3J(IQAE(fTG0YW>J&9HPM6oQvREjVRMl~ z5ftCF;fc2>^(&Ge>=UGy5`2ZB`JK{6DJG>T95c%QE=BRTAM|Y5n))SJUqlZ3w8(#0`D?QL!nI3b?#CQrjNu~xW6IyF>{+lM%M+-em=?$Ir1IZo33#dY z2bBM!D>Mg8l|vB!8#R>HM1G{puQRPm%%4)2CAXWJMSk2)WtJD>jT9o5#eTnXeaju6 zqHjB?y)*xw%hB(8>CId&h4R#uT<#>w49XW!y5Gv>E}?AB< ze>c|MNkp*Gd&IuAp*o2bASC_o;eQ+I#W#Yi@0a>E)O|(i#kYd2?>X%HQuXnwKXU%D z!~P3tOZCs-e?9i^GD@mzOJnnvMr})D+m=Ruu(31P7%2E1vD;^e@(VQ@p{Y06Xl*`~ zQb`{m0{opFe^;l573@_R#?wWZzU`xwU%{#gfnLy2>Qw{tV-f;?-?v<0{;`%R%-{LZ zxT3`Rcjt3)L8ys@&P__Xn(%mj=u~>Xtvs#T$KR|qOSk+rA+uQ$$_-M=p3f(`%28S> z1-KQ#{61ANxAT;;<2wYd57*_Q{eESS%W{ELbug!}UFrW*VKu#}Qh$AdzgI7=u|xg7 z+m(D+$q^+FD0xK5rl16D(F5|Z8t=#VM~3jZ!MLe~*Nv>*xS@SyW9QnDbwk5L9fRTZ8$an^-`N=% zTI+A`SQlRJ4-c)|cz1EA7GV0Fo$yZaaz&gscKXiR`q3lW(!C?`lSTX;is$nbp-8mO zu)#;+SU_SPS6c<&1bkjeEx>-vQ2Bh`Gk-wXd-LZh^Yp!ZsrElphW~~rrvdyLW-zs> z6YxCdZO{N-joAEW;bM(Ys{9*hW5dWD$4T-^L@zMX%5D^KZ(4?z#_j? z#{R2i_-~ZqUn;}@xD0=}41ZJbi}wfEaUw0%{wCy0_4iJ}*TFue&z%-)Cn{_xVJ6Kf{9cZ0C& zj?e>Jf+6F9?b{#f>NR?UTS8q}AoZxQ!@{~%;{EoOEZUM`-9@or|44xMLg4-fVN`09 zh-LT3;w>9;YYJ^ZDcbl##YuCu38nDJ)>2jR29yPRMHcKv5gjhMSw{7awv`wQ_RZvb zM>{SDYA7^uT% z$zgI(xyh&r_>>-hTOVe9qjAj8<7@DV5#&q6hr?zVe33CVzQ=~K5Ju8f4s_|kd8{Ek zF+POV$4x338z;&cOeTR=0uy)!@rnN-C9QLB;93r+1DQCZ3Ums#@q7)?Z?GOTF>0$QuGKd(<&@?|-%SI&&LAf~l1 zV_yH}rLr$>zoOwm%+L~_{dv93lGod;=lF5|_bPvC2PIy=vrMTDlRvHbi~S?Wk%?NC z7p~*#dd~MTXd;rj9bpLjFlY7GlpjlRPoB44R6IOZ*HtX}K97qi?tcmu#M$KucH@L4)BF|nOQ z1lsGzaO_47KF_ZQDn=PY{me zR+%5W7F(@d0!cZ06N`E`PX+!i=VQg IlEwc21MXAuQvd(} literal 0 HcmV?d00001 diff --git a/fast/logging.patch b/fast/logging.patch new file mode 100644 index 0000000000..72252a13d7 --- /dev/null +++ b/fast/logging.patch @@ -0,0 +1,55 @@ +diff --git a/devito/logger.py b/devito/logger.py +index 2a7eed95a..b104efd34 100644 +--- a/devito/logger.py ++++ b/devito/logger.py +@@ -16,12 +16,15 @@ logger.addHandler(stream_handler) + # Add extra logging levels (note: INFO has value=20, WARNING has value=30) + DEBUG = logging.DEBUG + PERF = 19 ++BENCH = logging.DEBUG + INFO = logging.INFO + WARNING = logging.WARNING + ERROR = logging.ERROR + CRITICAL = logging.CRITICAL + + logging.addLevelName(PERF, "PERF") ++logging.addLevelName(BENCH, "BENCH") ++ + + logger_registry = { + 'DEBUG': DEBUG, +@@ -29,7 +32,8 @@ logger_registry = { + 'INFO': INFO, + 'WARNING': WARNING, + 'ERROR': ERROR, +- 'CRITICAL': CRITICAL ++ 'CRITICAL': CRITICAL, ++ 'BENCH': BENCH + } + + NOCOLOR = '%s' +@@ -133,6 +137,8 @@ def warning(msg, *args, **kwargs): + def error(msg, *args, **kwargs): + log(msg, ERROR, *args, **kwargs) + ++def bench(msg, *args, **kwargs): ++ log(msg, BENCH, *args, **kwargs) + + def debug(msg, *args, **kwargs): + log(msg, DEBUG, *args, **kwargs) +diff --git a/devito/operator/profiling.py b/devito/operator/profiling.py +index f7633a740..2d576df5e 100644 +--- a/devito/operator/profiling.py ++++ b/devito/operator/profiling.py +@@ -460,7 +460,7 @@ class PerformanceSummary(OrderedDict): + + def create_profile(name): + """Create a new Profiler.""" +- if configuration['log-level'] in ['DEBUG', 'PERF'] and \ ++ if configuration['log-level'] in ['DEBUG', 'PERF', 'BENCH'] and \ + configuration['profiling'] == 'basic': + # Enforce performance profiling in DEBUG mode + level = 'advanced' +* Unmerged path fast/slurm/diffusion-16.slurm +* Unmerged path fast/slurm/diffusion-2.slurm +* Unmerged path fast/slurm/diffusion-32.slurm diff --git a/fast/slurm-jobs/diffusion-1.slurm b/fast/slurm-jobs/diffusion-1.slurm index 0d38fdbd15..ee5dc005be 100644 --- a/fast/slurm-jobs/diffusion-1.slurm +++ b/fast/slurm-jobs/diffusion-1.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive # export DEVITO_PROFILING=advanced2 diff --git a/fast/slurm-jobs/diffusion-128.slurm b/fast/slurm-jobs/diffusion-128.slurm index d78da834dd..e6c1a8db90 100644 --- a/fast/slurm-jobs/diffusion-128.slurm +++ b/fast/slurm-jobs/diffusion-128.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/diffusion-16.slurm b/fast/slurm-jobs/diffusion-16.slurm index 540e30aea2..e39179fa6c 100644 --- a/fast/slurm-jobs/diffusion-16.slurm +++ b/fast/slurm-jobs/diffusion-16.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/diffusion-2.slurm b/fast/slurm-jobs/diffusion-2.slurm index 8d24017ff7..cbcbe6517d 100644 --- a/fast/slurm-jobs/diffusion-2.slurm +++ b/fast/slurm-jobs/diffusion-2.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/diffusion-32.slurm b/fast/slurm-jobs/diffusion-32.slurm index d24ec70256..698248a474 100644 --- a/fast/slurm-jobs/diffusion-32.slurm +++ b/fast/slurm-jobs/diffusion-32.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/diffusion-4.slurm b/fast/slurm-jobs/diffusion-4.slurm index 0aaa718456..5f279c5dc2 100644 --- a/fast/slurm-jobs/diffusion-4.slurm +++ b/fast/slurm-jobs/diffusion-4.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive # export DEVITO_PROFILING=advanced2 diff --git a/fast/slurm-jobs/diffusion-64.slurm b/fast/slurm-jobs/diffusion-64.slurm index 8ecea840b3..dfa9f259b9 100644 --- a/fast/slurm-jobs/diffusion-64.slurm +++ b/fast/slurm-jobs/diffusion-64.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/diffusion-8.slurm b/fast/slurm-jobs/diffusion-8.slurm index 3369a0e765..57a2385ca7 100644 --- a/fast/slurm-jobs/diffusion-8.slurm +++ b/fast/slurm-jobs/diffusion-8.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive # export DEVITO_PROFILING=advanced2 diff --git a/fast/slurm-jobs/job-devito-xdsl-wave.slurm b/fast/slurm-jobs/job-devito-xdsl-wave.slurm index 98ac935e15..c622ecc6e2 100644 --- a/fast/slurm-jobs/job-devito-xdsl-wave.slurm +++ b/fast/slurm-jobs/job-devito-xdsl-wave.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 # export DEVITO_PROFILING=advanced2 diff --git a/fast/slurm-jobs/job-devito-xdsl.slurm b/fast/slurm-jobs/job-devito-xdsl.slurm index fecbdc0f6a..001bcff580 100644 --- a/fast/slurm-jobs/job-devito-xdsl.slurm +++ b/fast/slurm-jobs/job-devito-xdsl.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 # export DEVITO_PROFILING=advanced2 diff --git a/fast/slurm-jobs/wave-1.slurm b/fast/slurm-jobs/wave-1.slurm index f212a6f2e3..5e6d73da32 100644 --- a/fast/slurm-jobs/wave-1.slurm +++ b/fast/slurm-jobs/wave-1.slurm @@ -39,7 +39,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive # export DEVITO_PROFILING=advanced2 diff --git a/fast/slurm-jobs/wave-128.slurm b/fast/slurm-jobs/wave-128.slurm index d2abb6d154..553756b7a9 100644 --- a/fast/slurm-jobs/wave-128.slurm +++ b/fast/slurm-jobs/wave-128.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=diag2 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/wave-16.slurm b/fast/slurm-jobs/wave-16.slurm index c6c547a135..2e6eb9997b 100644 --- a/fast/slurm-jobs/wave-16.slurm +++ b/fast/slurm-jobs/wave-16.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/wave-2.slurm b/fast/slurm-jobs/wave-2.slurm index 70fec5de36..a6f270dcf9 100644 --- a/fast/slurm-jobs/wave-2.slurm +++ b/fast/slurm-jobs/wave-2.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/wave-32.slurm b/fast/slurm-jobs/wave-32.slurm index f234fffc31..0d93f7ab2b 100644 --- a/fast/slurm-jobs/wave-32.slurm +++ b/fast/slurm-jobs/wave-32.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/wave-4.slurm b/fast/slurm-jobs/wave-4.slurm index ef7d5f0d6b..1f1f529902 100644 --- a/fast/slurm-jobs/wave-4.slurm +++ b/fast/slurm-jobs/wave-4.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive # export DEVITO_PROFILING=advanced2 diff --git a/fast/slurm-jobs/wave-64.slurm b/fast/slurm-jobs/wave-64.slurm index de9cc146c6..06cdb863db 100644 --- a/fast/slurm-jobs/wave-64.slurm +++ b/fast/slurm-jobs/wave-64.slurm @@ -36,7 +36,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive diff --git a/fast/slurm-jobs/wave-8.slurm b/fast/slurm-jobs/wave-8.slurm index 7b8e130d76..73454898d4 100644 --- a/fast/slurm-jobs/wave-8.slurm +++ b/fast/slurm-jobs/wave-8.slurm @@ -37,7 +37,7 @@ export OMP_PLACES=cores # Devito-specific env variables export DEVITO_ARCH=cray export DEVITO_LANGUAGE=openmp -export DEVITO_LOGGING=DEBUG +export DEVITO_LOGGING=BENCH export DEVITO_MPI=1 export DEVITO_AUTOTUNING=aggressive # export DEVITO_PROFILING=advanced2 From 2581d5111de3475931d7a9e8456d00c82ba9b279 Mon Sep 17 00:00:00 2001 From: Emilien Bauer Date: Thu, 12 Oct 2023 14:12:38 +0100 Subject: [PATCH 5/6] Remove unwanted files. --- fast/- | Bin 16112 -> 0 bytes fast/async.so | Bin 16304 -> 0 bytes fast/logging.patch | 55 --------------------------------------------- 3 files changed, 55 deletions(-) delete mode 100755 fast/- delete mode 100755 fast/async.so delete mode 100644 fast/logging.patch diff --git a/fast/- b/fast/- deleted file mode 100755 index a34fa05f31c8725b7b6ae6d1190a8fc0bc1674a6..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 16112 zcmeHOeQaA-6+gCL@|?2!8;*Cx1Hs+<<`T?laPLuQnoYJ+;*0PLOH+GY>jvefm zE~zwTD6BPGp#lvi_@^MjstLpf)7UB)Eemah_D4Y-6p4?lu+BzMvJXl@UCcT6-ebQ$ z+tUyqLf{+aoqK-gf=}|)MjcfZ@1(DKq#je2lbvYKFXiB%Mzh43+U@F5bov%4KfUTwddMyL zClqBnV#zM2>~hME>8wf?Q`KEq|FRXOIE|a7UnhOHgn=pN@J_U&_TFyc^leb>RqNjp z)i0(hY#{g@_)olOw~;(bWx~52S-1IvKfQk#s};)(q6;fX$cw)M{tLt^>4z|YO8EU% z^uJRDzp9F#-=jX~3aiCJ{q`5RQYDVnO7T2g1*c1>lKxk#=+k-khB=p3!w?-7GMQSJ=#-w1OEvZg8}Pj2&7_~Bjd@0nYJP+%QV64-!x!G zW2xB4SlWuEHVv$aClj$vk!|sqlvS3P!&4D+`&c3pAKM*+k%>ujXDpS7#ladIkENt{ z7)N}gm5N2i*Wkg531A=@os7rA$w<`3u8moJlZjz#ESV5!ID*{h2c=7LBdrD#aklH- z6-!tfr%@}FOi&D@CBU9o+Dav-RV1-Q6e1H7@o5fFc#DmP;_>9Lj372XJTWb*GFYcq ziLzT|-xFL?9M_P!J&h_WSV?8%{DVT|rqRjB`QA${-y<_-T|?aSR3}L;7@|3nz^|Gv4P#XA8@IjBw$G7x38D8N?d#qa@4E z7O>p#L*i))mOm-@!^G2+FMmYv-zJ`>V)-Wo|2Xk91(jPgBYKdcp4|o~By)px{%)(^M+oCHNi0(-bQ27yLHj zY3h{s3I4;x(-brhfwp;b>KliPw}=Aw8wR`*ThUH4+!#rDZQe;*1}JcDQ&x;u1NXfU*C zBcAwnBRhD)$iCj6ozGE!Uu#>?douJr%116UUnulvk8j5kXzTnWLIC~QUtyQdqC9sk zijC~^M)tU|f7w%nBn9%4vH#9FBYVooo(=E6Yc8BU70!MX0jiE&F!mh1U}V=J{e$

CH^_=-bB|vv6qMWF2*a0h zD}?kf;q2Mm!zc}BU&%csDP6dwe&#z+5ukXD7q4neaIMrbP|H9q1GNm)GEmDvEd#X- z)G|=Z!2ewaO0VMhh1aDictCz1a63JKQ*py{uv=EeF~cu4)Ju<5b?*ZDKEQ*=MtU<7 zNRQ(CH7l|WYfPudB`>I%9!*52AduJ=}B9e-Xi@v|zyx4}&CDjecdBFZ=P>PQU-2P@w;(rvl{SBSO8*u1( z#9KGX^D&5(g+7~q^^#Cq;FaH_pj^9E)?2B*(1u^5!3kWf(BA2Vc9hX4F`h*Q(5|$- zk(;}=$h+uGAo8B8e4sUY#e2}8%LlHTaFwPD{jAW6*AlK`;FCH4E_4Wr5AK5EW4)kh z63VjR2B=%kOpX3~GjLvqdFZ(Y?__%+qdywf=k{i>MVA{!a3q|Vo74Sgq34~@pl}Tq z3mo8XLGRc7^Pw3N$KsWOuXDeGBlE_UZueS%kGBuKjPl@2r&rVdB5<#JZzl5qEN^t* z2m2oI-}9=JgQne$FP42nx{^1zzld_U=cUEM-c<+RE!u;&1jok;$b6!Gi0t7+Cp2l& zug4?XsYgEtJpScpNIugEy7!CVuRAiNq27Ty|8loSw)ch}oQIR0*yRy`jBiNOF>de; zd0;V8W91n=GY_XbLG%aX(>!_|>V*O2LDz-luj&5tFw=q8)|GV_f4E-L^?LuQbGoKC zH6ndd$UC7?JMQ%&-=wW=AnMk1L8ZI{yqaF8A>OR*6?zSt-qMWx_uVRWUgON#lqNd0eYlzdZ@q+ai0IJ8^q!1zUe znwurB*WZWz-aJ(AUYvRJE6p8b???Nl#_tNep+V}kw3O_Frzk&*;o`Tn4z_Qqm-?^j z2Sxk3m;7!SYHB8Y%aGn6`Sbb_k>8dUh3h{T?dZA=tFa$%()q#m-8fz-E~4(Br$grN zm*)!lYymQz&?fJNziorPxE35d18q9$Ke2vRhl?alMV`j?mh* zy=<0DnSHbBl#+|K=~u~Ve+n_|Q<;5K+54sc(r=gIxgT@LV-Cyg_bGdy(r3bUOb?)d za_Y?Cqso50E#Z~o4=VezUuaI4@3 z4HR-rop%3ldy72pdpG){^LP4p`1*k8%Y_0Z{?OS%VIR_GUMUowLQ3T_Nx-#x1GuJq zu0_qd_W%Y>;u$P4R7CsMYtNdreFIz#jeK`u%mE(LU>Ozk!XR4%t&&7g6^NQb67A*7ojJe|PJW z?$)kQ>&j4Tuw?h3+kJwdE+IxGRP}{g?ah;yTKO-?0Drf~-|6vXI-n}d?{x6=Z6KB6 zj8@$s=!GDa->Hw)u97V9_kxQQ&x7G=#q)QAG^VMr{Lk}+qyW^7g|1a9e!76L@cn|n zSLVmeuc-L=+c>^YtJ?_VR!e)nUx_!<2=Vnszj{=zlmwnp4KeOhJjZ!R>G2Z-w} z9#iyZik?;UyrT8$W5X*P-;J(av*x}2C4<}WW7tXm(%S;J2fA)qIw{ztAG-~?J8o34 zGxAG$%DV=yI;Rt29J}!su*?3jBkIz#9pPIY__Ye>>jqg-Xdd8#DHuA=|tD)?hn@TaTb$E)DKLfjF$h=`sBE5&)f ziXRUnubM4%7X!Qht_I!TS)e6wlTw`a2Q7uDn@zQ0U`eRlgT!l+Nb`UT*L; z#4F_ohErDBn#39qvVk)SuG+2mAJgx7yuN zwqe4)vq97=?bNVs!6@G?Avb@}mJ7!&6LKfj)lQJoI;h-(A)0dT0vWJx)3{_yi0D_v zohhmlw4=kkWRptJzR99&M~DoL_juShvEYg?FK_d5cX0uHBrzEnjihnSI69p`HA$@$ z1W2+gmP*rx7CU323>Ro4WT0vjaSL$i8&6>XtAUXuR%lHa0%D*9Qpsq`BI9Gj*nH9=&lsHWvn`zlTn--}$IGPry8TBHnkSyexe#9NvT#Mw zB@US9VXS@wR_G^?&3Jyu^a9>Dsn9%&s%+2mwp)~p?Rmb*l;@YICW>;$EYOAc?j z=gVYIb7rS~3^B?H&(DiDZFTeJ=Kxd@MNQ|Kg{@e#+xwLrQz0yBmwprYw_&H#Ud`Y9 zigLW{f_bJigzfF~ypZX)IdRGXx6A(M-Iv{-=SNKWIf?b1{m&@-Vl#fFsi>jmcM2`0 zm-&5Iq4wFH=T%Id(>rJTUvSt@-7GybJt^7Zk@LpDL9Ec+nV)0$`H%OVI`x@n`VF*g zv)S{!d6q6rEUYKF63_TED4?;&?elz-_vg-1a%^(8r3H*?DVK*?0p*pBIs9riqrp6Zf@tjBJgJe?oHIQ~pfYG$g~Q>^y#A~J+2 z^`eOF)o>`{Y;UJL_KHeac|tjyQ-RU?z!mt<@$>zco})Q-tQO0`J7o1qhjVV173ngs QWPhhxsQoo7D(bZVJCdDqe*gdg diff --git a/fast/async.so b/fast/async.so deleted file mode 100755 index 1d8594a3c66aabdb88b4cbd8ed58c5ca2575a1b7..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 16304 zcmeHOeQ;FQbwB$d?do$^G!6=2^z52Qw!^aGLnuxOyMW-ySz(MNTLig}m9+Z6t6g>X z2_xFn+mHZThn99~n`V+3r%gM=(@EU?kvOz*z$RFex*gJ?4u80bOl!P5lVTZSLR+lu zIrrYP`?T8iIO8_`XRk)@+~4n_$2$Ek$`DK(J%UKmHma4n3>rHc*+Sz}vEVo$xmZX6t$M8Y)qxwE* z;_&TO^_A;io9Y)!)ogHJMN9CXu&B4ukGWbeale88AHo0ce>=Ok{i&V(_y6VKdwPF(SJNRG9{sC-IDL!1r{;4wjKa}BH%ao5aX!u{8zJNrj_I$kz z?f%%8b zL?k@16%S@a5IynXsc0k=4-Z@5j)=K^DmG+}$76yGg%OQ?rcjLZ>Wf9$uj@c0WaG&LMH5>qh>fG-*MwRet=n*}YYVi>L|b3B?{pdTXC0Xst7Tecc&eQSLi z^7;0S_$%7fpFN*rhNQ#7lC2&79LEOFJ5$TXYXEKjQgP4x$THW0J9dJ`ykW-Y2-qonl1>p zg{QfalQ$r9WY2RqG0g!w0uV2*Zvp{m0hu9Q%(IGbv+$(j?_$y1S%1zM4%FZN(3qq`1M8Lx@YBo>4$&r)}O38y9Pl2;u+JPyPzLF=YClVOsSPVOla}F9>{^FfGxtX9WH>VOlz6pBMN!!n9<{J}dBF z5T>P4_JqJ+CQM7DY)ariB}~hp?4-bdNSKyH*?j_ko-i%_vbzL6M3|O9*?_1VHQ*3aH_=(h9vkLS&gU;&#{ z0q)%Fh^J*+!JmGw)TVz#HB8;xryt(5jIw(AO|xEqa?>{vWcFYyGsAc|U-egrZI5Ez z`TEH}`#369d-~87e}C@}=oi$FPt9fS$6j93J@@p?(Xs31HIMhU*DyZUJbPZc=IK9+ zW5Y{DegPc-*#DZRsq1FupUBrU&~=lIgg*F4@Zn%zu=gRXs98_<&FJat-Rb$vv%q7AFDgGG^ad-Na5llWVzdQY}7_uX%&zyZPm($ZP>*-hY;~RR3@t}jAenUV0;GCYm zsHd-ljz2sXN?#16e+LQ5j$=#C-qh1OQU2WhVgRc-P4^GqqwWS@(CGL*8jM)#ehBUk zKC&P7+cK|FRfo^{_4Lm&pT&GAeZ41rwLAB!=ZlYF>_X{W=K3uu{~INSsOv)JL2`Ps zYX(Dsk&Im08ypBe8vN~`@#xEi{`BR8v8$V!jDg5yT1C%(_ye%nZF>5AW(VT#^!d{g zzoMsqs;9G}d}VSqbC)prA)AEKIFPa#pD5_<%y+O+dirgFL+Q6e=^LT+<;(?SGk=GN z?({j~i2Vqquj=WunRk&DhF3E8VOACUGvW9~<|jh?%TW4CW`)SUmFbm|CW=Bovo9hG zke^fXvj_&bk}46XM4%FZN(3qqs6?O=fl34_5vWAq|2+avz4q};ov-QB+>pZSG=7!E z8Y!QU-wUkccNKPc5e`{QQvC8ky|ma=^_xK73pnstOK( zSZw$p_+kfqQzNDiU%Mry1sacy#KC7q4w}Bjtt4|YD5FY_g%gqCGRmQN3}0KuOn)hT zA`+#t-`9rp;b?R;5uO|qO`S}{&A2%|84-wz2+Nm^AGZ4?I{v>MkPq77_X(OY8BT;J zMBhJXTr5LqL3TGH2Uy?iN%eWbTi=XH{yzk7eZwaC8XS5K@pepNJ_aze#A_*Ez2tK? zx%2PuoNHvh5pA7IEc^;5%;3WW=k0E2MjgEicP!6=bE)Oc+1wp@Y@;`Mh#epEg0t2w z-eU%AUa;MRPig#E&PHecT*9Xecqa{jPdJ3c8+#$~23yD^DP?L}1yvmkD)Rp}0ykVR z58GDYOm-Afu9G3x+|d-)XtTp8j)a-HIhS??w%rajig!Y*-~k>H@^db2K6uW+vA9&= zD*IbFGVkuR+jju`LG!?ys1KZLaXX#A6ujGhG?jV+I%@5Y!?7o{z3)gpIGvB+i(&78 zOX+Lu&!8UmzR@cDZB_8yygxYGaC~fp)aRQA$RB1}pw21FbvR@_)mY96ht_eK^ygZ@ z<$eaVsxt#lv^!Cyb=V!UzF)Wk^KhXByF3bz@(ws%*lzF+IG`1!-g()Tnukj*Ao_#t zb2?mAXcq$12W&SxuDP@uaBd}@TT833{h?~7%T=viyy|kg>S|H`tb)f2yG zz*Qsh4c8eF--ZUoyM7|-ak*Tu8T;`b{XST~9mfl`i=;i^SSjQ8`qi9kHV3H|@W{FF z-#l;>p9KdlgU5yT&+nRb!7b7=UL?`#!z3J(IQAE(fTG0YW>J&9HPM6oQvREjVRMl~ z5ftCF;fc2>^(&Ge>=UGy5`2ZB`JK{6DJG>T95c%QE=BRTAM|Y5n))SJUqlZ3w8(#0`D?QL!nI3b?#CQrjNu~xW6IyF>{+lM%M+-em=?$Ir1IZo33#dY z2bBM!D>Mg8l|vB!8#R>HM1G{puQRPm%%4)2CAXWJMSk2)WtJD>jT9o5#eTnXeaju6 zqHjB?y)*xw%hB(8>CId&h4R#uT<#>w49XW!y5Gv>E}?AB< ze>c|MNkp*Gd&IuAp*o2bASC_o;eQ+I#W#Yi@0a>E)O|(i#kYd2?>X%HQuXnwKXU%D z!~P3tOZCs-e?9i^GD@mzOJnnvMr})D+m=Ruu(31P7%2E1vD;^e@(VQ@p{Y06Xl*`~ zQb`{m0{opFe^;l573@_R#?wWZzU`xwU%{#gfnLy2>Qw{tV-f;?-?v<0{;`%R%-{LZ zxT3`Rcjt3)L8ys@&P__Xn(%mj=u~>Xtvs#T$KR|qOSk+rA+uQ$$_-M=p3f(`%28S> z1-KQ#{61ANxAT;;<2wYd57*_Q{eESS%W{ELbug!}UFrW*VKu#}Qh$AdzgI7=u|xg7 z+m(D+$q^+FD0xK5rl16D(F5|Z8t=#VM~3jZ!MLe~*Nv>*xS@SyW9QnDbwk5L9fRTZ8$an^-`N=% zTI+A`SQlRJ4-c)|cz1EA7GV0Fo$yZaaz&gscKXiR`q3lW(!C?`lSTX;is$nbp-8mO zu)#;+SU_SPS6c<&1bkjeEx>-vQ2Bh`Gk-wXd-LZh^Yp!ZsrElphW~~rrvdyLW-zs> z6YxCdZO{N-joAEW;bM(Ys{9*hW5dWD$4T-^L@zMX%5D^KZ(4?z#_j? z#{R2i_-~ZqUn;}@xD0=}41ZJbi}wfEaUw0%{wCy0_4iJ}*TFue&z%-)Cn{_xVJ6Kf{9cZ0C& zj?e>Jf+6F9?b{#f>NR?UTS8q}AoZxQ!@{~%;{EoOEZUM`-9@or|44xMLg4-fVN`09 zh-LT3;w>9;YYJ^ZDcbl##YuCu38nDJ)>2jR29yPRMHcKv5gjhMSw{7awv`wQ_RZvb zM>{SDYA7^uT% z$zgI(xyh&r_>>-hTOVe9qjAj8<7@DV5#&q6hr?zVe33CVzQ=~K5Ju8f4s_|kd8{Ek zF+POV$4x338z;&cOeTR=0uy)!@rnN-C9QLB;93r+1DQCZ3Ums#@q7)?Z?GOTF>0$QuGKd(<&@?|-%SI&&LAf~l1 zV_yH}rLr$>zoOwm%+L~_{dv93lGod;=lF5|_bPvC2PIy=vrMTDlRvHbi~S?Wk%?NC z7p~*#dd~MTXd;rj9bpLjFlY7GlpjlRPoB44R6IOZ*HtX}K97qi?tcmu#M$KucH@L4)BF|nOQ z1lsGzaO_47KF_ZQDn=PY{me zR+%5W7F(@d0!cZ06N`E`PX+!i=VQg IlEwc21MXAuQvd(} diff --git a/fast/logging.patch b/fast/logging.patch deleted file mode 100644 index 72252a13d7..0000000000 --- a/fast/logging.patch +++ /dev/null @@ -1,55 +0,0 @@ -diff --git a/devito/logger.py b/devito/logger.py -index 2a7eed95a..b104efd34 100644 ---- a/devito/logger.py -+++ b/devito/logger.py -@@ -16,12 +16,15 @@ logger.addHandler(stream_handler) - # Add extra logging levels (note: INFO has value=20, WARNING has value=30) - DEBUG = logging.DEBUG - PERF = 19 -+BENCH = logging.DEBUG - INFO = logging.INFO - WARNING = logging.WARNING - ERROR = logging.ERROR - CRITICAL = logging.CRITICAL - - logging.addLevelName(PERF, "PERF") -+logging.addLevelName(BENCH, "BENCH") -+ - - logger_registry = { - 'DEBUG': DEBUG, -@@ -29,7 +32,8 @@ logger_registry = { - 'INFO': INFO, - 'WARNING': WARNING, - 'ERROR': ERROR, -- 'CRITICAL': CRITICAL -+ 'CRITICAL': CRITICAL, -+ 'BENCH': BENCH - } - - NOCOLOR = '%s' -@@ -133,6 +137,8 @@ def warning(msg, *args, **kwargs): - def error(msg, *args, **kwargs): - log(msg, ERROR, *args, **kwargs) - -+def bench(msg, *args, **kwargs): -+ log(msg, BENCH, *args, **kwargs) - - def debug(msg, *args, **kwargs): - log(msg, DEBUG, *args, **kwargs) -diff --git a/devito/operator/profiling.py b/devito/operator/profiling.py -index f7633a740..2d576df5e 100644 ---- a/devito/operator/profiling.py -+++ b/devito/operator/profiling.py -@@ -460,7 +460,7 @@ class PerformanceSummary(OrderedDict): - - def create_profile(name): - """Create a new Profiler.""" -- if configuration['log-level'] in ['DEBUG', 'PERF'] and \ -+ if configuration['log-level'] in ['DEBUG', 'PERF', 'BENCH'] and \ - configuration['profiling'] == 'basic': - # Enforce performance profiling in DEBUG mode - level = 'advanced' -* Unmerged path fast/slurm/diffusion-16.slurm -* Unmerged path fast/slurm/diffusion-2.slurm -* Unmerged path fast/slurm/diffusion-32.slurm From b6b04190794d2587fda652d0f177def77bc21dd6 Mon Sep 17 00:00:00 2001 From: Emilien Bauer Date: Thu, 12 Oct 2023 14:29:21 +0100 Subject: [PATCH 6/6] Remove topology hardcoded hack. --- devito/mpi/distributed.py | 3 --- 1 file changed, 3 deletions(-) diff --git a/devito/mpi/distributed.py b/devito/mpi/distributed.py index 50577659cb..387695b7b1 100644 --- a/devito/mpi/distributed.py +++ b/devito/mpi/distributed.py @@ -196,9 +196,6 @@ def __init__(self, shape, dimensions, input_comm=None, topology=None): # mpi4py takes care of that when the object gets out of scope self._input_comm = (input_comm or MPI.COMM_WORLD).Clone() - if len(shape) == 3: - topology = ('*', '*', 1) - if topology is None: # `MPI.Compute_dims` sets the dimension sizes to be as close to each other # as possible, using an appropriate divisibility algorithm. Thus, in 3D: