Single core performance, Collide benchmark, SandyBridge
Performance in millions of particle-timesteps / second
Nparticles | CPU (mpi) | Kokkos/OMP (mpi) | Kokkos/serial (mpi) | |
1000 | 24.79 (1) | 13.67 (1) | 19.76 (1) | |
2000 | 26.75 (1) | 9.111 (1) | 22.18 (1) | |
4000 | 20.92 (1) | 13.05 (1) | 19.24 (1) | |
8000 | 18.48 (1) | 18.96 (1) | 18.27 (1) | |
16000 | 18.41 (1) | 8.998 (1) | 16.78 (1) |
1000 | mpirun -n 1 -N 1 --bind-to core spa_chama_cpu -v x 4 -v y 5 -v z 5 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=cpu.kind=core.size=1K.node=1.mpi=1 |
2000 | mpirun -n 1 -N 1 --bind-to core spa_chama_cpu -v x 8 -v y 5 -v z 5 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=cpu.kind=core.size=2K.node=1.mpi=1 |
4000 | mpirun -n 1 -N 1 --bind-to core spa_chama_cpu -v x 8 -v y 5 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=cpu.kind=core.size=4K.node=1.mpi=1 |
8000 | mpirun -n 1 -N 1 --bind-to core spa_chama_cpu -v x 8 -v y 10 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=cpu.kind=core.size=8K.node=1.mpi=1 |
16000 | mpirun -n 1 -N 1 --bind-to core spa_chama_cpu -v x 16 -v y 10 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=cpu.kind=core.size=16K.node=1.mpi=1 |
1000 | mpirun -n 1 -N 1 --bind-to socket spa_chama_kokkos_omp -sf kk -k on t 1 -pk kokkos reduction parallel/reduce comm classic -v x 4 -v y 5 -v z 5 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_omp.kind=core.size=1K.node=1.mpi=1.thread=1 |
2000 | mpirun -n 1 -N 1 --bind-to socket spa_chama_kokkos_omp -sf kk -k on t 1 -pk kokkos reduction parallel/reduce comm classic -v x 8 -v y 5 -v z 5 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_omp.kind=core.size=2K.node=1.mpi=1.thread=1 |
4000 | mpirun -n 1 -N 1 --bind-to socket spa_chama_kokkos_omp -sf kk -k on t 1 -pk kokkos reduction parallel/reduce comm classic -v x 8 -v y 5 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_omp.kind=core.size=4K.node=1.mpi=1.thread=1 |
8000 | mpirun -n 1 -N 1 --bind-to socket spa_chama_kokkos_omp -sf kk -k on t 1 -pk kokkos reduction parallel/reduce comm classic -v x 8 -v y 10 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_omp.kind=core.size=8K.node=1.mpi=1.thread=1 |
16000 | mpirun -n 1 -N 1 --bind-to socket spa_chama_kokkos_omp -sf kk -k on t 1 -pk kokkos reduction parallel/reduce comm classic -v x 16 -v y 10 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_omp.kind=core.size=16K.node=1.mpi=1.thread=1 |
1000 | mpirun -n 1 -N 1 --bind-to core spa_chama_kokkos_serial -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 4 -v y 5 -v z 5 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_serial.kind=core.size=1K.node=1.mpi=1 |
2000 | mpirun -n 1 -N 1 --bind-to core spa_chama_kokkos_serial -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 8 -v y 5 -v z 5 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_serial.kind=core.size=2K.node=1.mpi=1 |
4000 | mpirun -n 1 -N 1 --bind-to core spa_chama_kokkos_serial -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 8 -v y 5 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_serial.kind=core.size=4K.node=1.mpi=1 |
8000 | mpirun -n 1 -N 1 --bind-to core spa_chama_kokkos_serial -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 8 -v y 10 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_serial.kind=core.size=8K.node=1.mpi=1 |
16000 | mpirun -n 1 -N 1 --bind-to core spa_chama_kokkos_serial -sf kk -k on -pk kokkos reduction parallel/reduce comm classic -v x 16 -v y 10 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=chama.pkg=kokkos_serial.kind=core.size=16K.node=1.mpi=1 |