Weak scaling performance, Collide benchmark, KNL, 1M particles/node
Performance in millions of particle-timesteps / second / node

Nodes CPU/KNL (mpi,hyper) Kokkos/KNL (mpi,thread,hyper) Kokkos/serial/KNL (mpi,hyper)
1 326.8 (128,2) 320.1 (64,4,4) 320.6 (128,2)
2 288.1 (128,2) 285.8 (64,4,4) 281.6 (128,2)
4 275.3 (128,2) 272.1 (64,4,4) 268.2 (128,2)
8 262.5 (128,2) 261.2 (32,8,4) 254.9 (128,2)
16 255 (128,2) 255.2 (32,8,4) 194 (64,1)
32 235.7 (128,2) 245.8 (32,8,4) 235.8 (128,2)
64 234.9 (128,2) 236.5 (32,8,4) 209.6 (128,2)

Run commands and logfile links for column CPU/KNL

1 srun -n 128 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_knl -v x 64 -v y 40 -v z 40 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=weak.size=1M.node=1.mpi=128.hyper=2
2 srun -n 256 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_knl -v x 64 -v y 40 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=weak.size=1M.node=2.mpi=128.hyper=2
4 srun -n 512 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_knl -v x 64 -v y 80 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=weak.size=1M.node=4.mpi=128.hyper=2
8 srun -n 1024 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_knl -v x 128 -v y 80 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=weak.size=1M.node=8.mpi=128.hyper=2
16 srun -n 2048 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_knl -v x 128 -v y 80 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=weak.size=1M.node=16.mpi=128.hyper=2
32 srun -n 4096 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_knl -v x 128 -v y 160 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=weak.size=1M.node=32.mpi=128.hyper=2
64 srun -n 8192 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_knl -v x 256 -v y 160 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=weak.size=1M.node=64.mpi=128.hyper=2

Run commands and logfile links for column Kokkos/KNL

1 setenv OMP_NUM_THREADS 4; srun -n 64 -C knl --ntasks-per-node 64 --cpu_bind=cores -c 4 ./spa_mutrino_kokkos_knl -sf kk -k on t 4 -pk kokkos reduction parallel/reduce comm classic -v x 64 -v y 40 -v z 40 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_knl.kind=weak.size=1M.node=1.mpi=64.thread=4.hyper=4
2 setenv OMP_NUM_THREADS 4; srun -n 128 -C knl --ntasks-per-node 64 --cpu_bind=cores -c 4 ./spa_mutrino_kokkos_knl -sf kk -k on t 4 -pk kokkos reduction parallel/reduce comm classic -v x 64 -v y 40 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_knl.kind=weak.size=1M.node=2.mpi=64.thread=4.hyper=4
4 setenv OMP_NUM_THREADS 4; srun -n 256 -C knl --ntasks-per-node 64 --cpu_bind=cores -c 4 ./spa_mutrino_kokkos_knl -sf kk -k on t 4 -pk kokkos reduction parallel/reduce comm classic -v x 64 -v y 80 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_knl.kind=weak.size=1M.node=4.mpi=64.thread=4.hyper=4
8 setenv OMP_NUM_THREADS 8; srun -n 256 -C knl --ntasks-per-node 32 --cpu_bind=cores -c 8 ./spa_mutrino_kokkos_knl -sf kk -k on t 8 -pk kokkos reduction parallel/reduce comm classic -v x 128 -v y 80 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_knl.kind=weak.size=1M.node=8.mpi=32.thread=8.hyper=4
16 setenv OMP_NUM_THREADS 8; srun -n 512 -C knl --ntasks-per-node 32 --cpu_bind=cores -c 8 ./spa_mutrino_kokkos_knl -sf kk -k on t 8 -pk kokkos reduction parallel/reduce comm classic -v x 128 -v y 80 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_knl.kind=weak.size=1M.node=16.mpi=32.thread=8.hyper=4
32 setenv OMP_NUM_THREADS 8; srun -n 1024 -C knl --ntasks-per-node 32 --cpu_bind=cores -c 8 ./spa_mutrino_kokkos_knl -sf kk -k on t 8 -pk kokkos reduction parallel/reduce comm classic -v x 128 -v y 160 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_knl.kind=weak.size=1M.node=32.mpi=32.thread=8.hyper=4
64 setenv OMP_NUM_THREADS 8; srun -n 2048 -C knl --ntasks-per-node 32 --cpu_bind=cores -c 8 ./spa_mutrino_kokkos_knl -sf kk -k on t 8 -pk kokkos reduction parallel/reduce comm classic -v x 256 -v y 160 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_knl.kind=weak.size=1M.node=64.mpi=32.thread=8.hyper=4

Run commands and logfile links for column Kokkos/serial/KNL

1 srun -n 128 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_kokkos_serial_knl -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 64 -v y 40 -v z 40 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_serial_knl.kind=weak.size=1M.node=1.mpi=128.hyper=2
2 srun -n 256 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_kokkos_serial_knl -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 64 -v y 40 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_serial_knl.kind=weak.size=1M.node=2.mpi=128.hyper=2
4 srun -n 512 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_kokkos_serial_knl -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 64 -v y 80 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_serial_knl.kind=weak.size=1M.node=4.mpi=128.hyper=2
8 srun -n 1024 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_kokkos_serial_knl -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 128 -v y 80 -v z 80 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_serial_knl.kind=weak.size=1M.node=8.mpi=128.hyper=2
16 srun -n 1024 -C knl --ntasks-per-node 64 --cpu_bind=rank -c 4 ./spa_mutrino_kokkos_serial_knl -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 128 -v y 80 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_serial_knl.kind=weak.size=1M.node=16.mpi=64.hyper=1
32 srun -n 4096 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_kokkos_serial_knl -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 128 -v y 160 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_serial_knl.kind=weak.size=1M.node=32.mpi=128.hyper=2
64 srun -n 8192 -C knl --ntasks-per-node 128 --cpu_bind=threads -c 2 ./spa_mutrino_kokkos_serial_knl -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 256 -v y 160 -v z 160 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=kokkos_serial_knl.kind=weak.size=1M.node=64.mpi=128.hyper=2