Single core performance, Collide benchmark
Performance in millions of particle-timesteps / second
Nparticles | SandyBridge | Haswell | Broadwell | KNL | |
1000 | 24.79 (CPU,mpi=1) | 27.74 (CPU,mpi=1) | 25.11 (CPU,mpi=1) | 6.501 (CPU/KNL,mpi=1) | |
2000 | 26.75 (CPU,mpi=1) | 28.13 (CPU,mpi=1) | 26.41 (CPU,mpi=1) | 6.61 (CPU/KNL,mpi=1) | |
4000 | 20.92 (CPU,mpi=1) | 26.74 (CPU,mpi=1) | 25.13 (CPU,mpi=1) | 6.564 (CPU/KNL,mpi=1) | |
8000 | 18.96 (Kokkos/OMP,mpi=1) | 25.58 (CPU,mpi=1) | 24.4 (CPU,mpi=1) | 6.022 (CPU/KNL,mpi=1) | |
16000 | 18.41 (CPU,mpi=1) | 23.86 (CPU,mpi=1) | 22.74 (CPU,mpi=1) | 5.08 (CPU/KNL,mpi=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 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 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 | srun -n 1 -C haswell --ntasks-per-node 1 --cpu_bind=cores -c 2 ./spa_mutrino_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=mutrino.pkg=cpu.kind=core.size=1K.node=1.mpi=1.hyper=1 |
2000 | srun -n 1 -C haswell --ntasks-per-node 1 --cpu_bind=cores -c 2 ./spa_mutrino_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=mutrino.pkg=cpu.kind=core.size=2K.node=1.mpi=1.hyper=1 |
4000 | srun -n 1 -C haswell --ntasks-per-node 1 --cpu_bind=cores -c 2 ./spa_mutrino_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=mutrino.pkg=cpu.kind=core.size=4K.node=1.mpi=1.hyper=1 |
8000 | srun -n 1 -C haswell --ntasks-per-node 1 --cpu_bind=cores -c 2 ./spa_mutrino_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=mutrino.pkg=cpu.kind=core.size=8K.node=1.mpi=1.hyper=1 |
16000 | srun -n 1 -C haswell --ntasks-per-node 1 --cpu_bind=cores -c 2 ./spa_mutrino_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=mutrino.pkg=cpu.kind=core.size=16K.node=1.mpi=1.hyper=1 |
1000 | mpiexec -np 1 -npernode 1 --oversubscribe --bind-to core ./spa_serrano_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=serrano.pkg=cpu.kind=core.size=1K.node=1.mpi=1.hyper=1 |
2000 | mpiexec -np 1 -npernode 1 --oversubscribe --bind-to core ./spa_serrano_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=serrano.pkg=cpu.kind=core.size=2K.node=1.mpi=1.hyper=1 |
4000 | mpiexec -np 1 -npernode 1 --oversubscribe --bind-to core ./spa_serrano_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=serrano.pkg=cpu.kind=core.size=4K.node=1.mpi=1.hyper=1 |
8000 | mpiexec -np 1 -npernode 1 --oversubscribe --bind-to core ./spa_serrano_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=serrano.pkg=cpu.kind=core.size=8K.node=1.mpi=1.hyper=1 |
16000 | mpiexec -np 1 -npernode 1 --oversubscribe --bind-to core ./spa_serrano_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=serrano.pkg=cpu.kind=core.size=16K.node=1.mpi=1.hyper=1 |
1000 | srun -n 1 -C knl --ntasks-per-node 1 --cpu_bind=cores -c 4 ./spa_mutrino_knl -v x 4 -v y 5 -v z 5 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=core.size=1K.node=1.mpi=1.hyper=1 |
2000 | srun -n 1 -C knl --ntasks-per-node 1 --cpu_bind=cores -c 4 ./spa_mutrino_knl -v x 8 -v y 5 -v z 5 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=core.size=2K.node=1.mpi=1.hyper=1 |
4000 | srun -n 1 -C knl --ntasks-per-node 1 --cpu_bind=cores -c 4 ./spa_mutrino_knl -v x 8 -v y 5 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=core.size=4K.node=1.mpi=1.hyper=1 |
8000 | srun -n 1 -C knl --ntasks-per-node 1 --cpu_bind=cores -c 4 ./spa_mutrino_knl -v x 8 -v y 10 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=core.size=8K.node=1.mpi=1.hyper=1 |
16000 | srun -n 1 -C knl --ntasks-per-node 1 --cpu_bind=cores -c 4 ./spa_mutrino_knl -v x 16 -v y 10 -v z 10 -v t 100 -in in.collide.steps -log log.sparta.date=23Dec17.model=collide.machine=mutrino.pkg=cpu_knl.kind=core.size=16K.node=1.mpi=1.hyper=1 |