add unaligned and 8 MiB tests to p2p bw

perf-regression/bebop/margo-regression.sbatch

@@ -25,16 +25,32 @@ export I_MPI_FABRICS=shm:ofa
 
 echo "## Margo OFI/PSM2 (round trip, bdw):"
 mpirun ./margo-p2p-latency -i 100000 -n psm2://enp6s0f0:5000
-echo "## Margo OFI/PSM2 (bw, bdw):"
+echo "## Margo OFI/PSM2 (bw, 1MiB, bdw):"
 mpirun ./margo-p2p-bw -x 1048576 -n psm2://enp6s0f0:5000 -c 1 -D 20
-echo "## Margo OFI/PSM2 (bw, 8x concurrency, bdw):"
+echo "## Margo OFI/PSM2 (bw, 1MiB, 8x concurrency, bdw):"
 mpirun ./margo-p2p-bw -x 1048576 -n psm2://enp6s0f0:5000 -c 8 -D 20
+echo "## Margo OFI/PSM2 (bw, 8MiB, bdw):"
+mpirun ./margo-p2p-bw -x 8388608 -n psm2://enp6s0f0:5000 -c 1 -D 20
+echo "## Margo OFI/PSM2 (bw, 8MiB, 8x concurrency, bdw):"
+mpirun ./margo-p2p-bw -x 8388608 -n psm2://enp6s0f0:5000 -c 8 -D 20
+echo "## Margo OFI/PSM2 (bw, 1MB unaligned, bdw):"
+mpirun ./margo-p2p-bw -x 1000000 -n psm2://enp6s0f0:5000 -c 1 -D 20
+echo "## Margo OFI/PSM2 (bw, 1MB unaligned, 8x concurrency, bdw):"
+mpirun ./margo-p2p-bw -x 1000000 -n psm2://enp6s0f0:5000 -c 8 -D 20
 
 echo "## Margo OFI/PSM2 (round trip, bdw, Hg busy spin):"
 mpirun ./margo-p2p-latency -i 100000 -n psm2://enp6s0f0:5000 -t 0,0
-echo "## Margo OFI/PSM2 (bw, bdw, Hg busy spin):"
+echo "## Margo OFI/PSM2 (bw, 1MiB, bdw, Hg busy spin):"
 mpirun ./margo-p2p-bw -x 1048576 -n psm2://enp6s0f0:5000 -c 1 -D 20 -t 0,0
-echo "## Margo OFI/PSM2 (bw, 8x concurrency, bdw, Hg busy spin):"
+echo "## Margo OFI/PSM2 (bw, 1MiB, 8x concurrency, bdw, Hg busy spin):"
 mpirun ./margo-p2p-bw -x 1048576 -n psm2://enp6s0f0:5000 -c 8 -D 20 -t 0,0
+echo "## Margo OFI/PSM2 (bw, 8MiB, bdw, Hg busy spin):"
+mpirun ./margo-p2p-bw -x 8388608 -n psm2://enp6s0f0:5000 -c 1 -D 20 -t 0,0
+echo "## Margo OFI/PSM2 (bw, 8MiB, 8x concurrency, bdw, Hg busy spin):"
+mpirun ./margo-p2p-bw -x 8388608 -n psm2://enp6s0f0:5000 -c 8 -D 20 -t 0,0
+echo "## Margo OFI/PSM2 (bw, 1MB unaligned, bdw, Hg busy spin):"
+mpirun ./margo-p2p-bw -x 1000000 -n psm2://enp6s0f0:5000 -c 1 -D 20 -t 0,0
+echo "## Margo OFI/PSM2 (bw, 1MB unaligned, 8x concurrency, bdw, Hg busy spin):"
+mpirun ./margo-p2p-bw -x 1000000 -n psm2://enp6s0f0:5000 -c 8 -D 20 -t 0,0
 
  

perf-regression/bebop/run-regression.sh

@@ -87,6 +87,7 @@ cp $PREFIX/bin/margo-p2p-bw $JOBDIR
 cp $PREFIX/libexec/osu-micro-benchmarks/mpi/pt2pt/osu_latency $JOBDIR
 # cp $PREFIX/bin/mercury-runner $JOBDIR
 cd $JOBDIR
+export SANDBOX
 sbatch --wait --export=ALL ./margo-regression.sbatch
 
 echo "=== JOB DONE, COLLECTING AND SENDING RESULTS ==="