Add mpi.sh

fast/mpi.sh

@@ -0,0 +1,52 @@
+#!/bin/bash
+
+# Simple script to run MPI+OMP benchmarks locally,
+# for simple sanity checks.
+
+# Use OpenMP and MPI
+export DEVITO_LANGUAGE=openmp
+export DEVITO_MPI=1
+# Use the default compiler here, cray and MPI doesnt work on my machine.
+unset DEVITO_ARCH
+# Enable debug logging.
+export DEVITO_LOGGING=DEBUG
+# Disable (tile size) autotuning for iteration speed.
+# I enable it sometimes; NB enabling it requires that no explicit tile size
+# is specified in the Operator constructor args.
+unset DEVITO_AUTOTUNING
+
+# Bind threads to physical cores
+export OMP_PLACES=cores
+export OMP_PROC_BIND=true
+
+# Doing 4 ranks X 4 threads locally here
+export OMP_NUM_THREADS=4
+export MPI_NUM_RANKS=4
+
+# Just extract the reported runtime from the whole output of the passed command
+get_runtime() {
+    $@ |& grep 'Operator.*ran' | head -n 1 | rev | cut -d ' ' -f2 | rev
+}
+
+# Iterate over benchmarks and cases, print simple CSV data to stdout
+# Copy-pastes nicely in Google Sheets
+echo bench_name,so,threads,Devito,xDSL
+for bench in "wave2d_b.py -d 2048 2048 --nt 512" "wave3d_b.py -d 512 512 512 --nt 512" "diffusion_3D_wBCs.py -d 512 512 512 --nt 512" "diffusion_2D_wBCs.py -d 2048 2048 --nt 512"
+do
+  # Get the benchmark file for printing
+  bench_name=$(echo $bench | cut -d ' ' -f1)
+  # Iterate over measured space orders
+  for so in 2 4 8
+  do
+
+      # To uncomment to check what's going on without capturing the output.
+      # echo OMP_NUM_THREADS=$threads
+       mpirun -np $MPI_NUM_RANKS --bind-to=core python $bench -so $so --devito 1
+
+      # Get the runtimes
+      devito_time=$(get_runtime mpirun -np $MPI_NUM_RANKS --bind-to=core python $bench -so $so --devito 1)
+      xdsl_time=$(get_runtime mpirun -np $MPI_NUM_RANKS --bind-to=core python $bench -so $so --xdsl 1)
+      # print CSV line
+      echo $bench_name,$so,$devito_time,$xdsl_time
+  done
+done