diff --git a/polaris/make.llvm.sh b/polaris/make.llvm.sh index 9ba0f88c61e50ed8bdac27c22f9e67b968ac60b1..6692083cb22c0caf6d3b5763c5f5dcb97a219d80 100755 --- a/polaris/make.llvm.sh +++ b/polaris/make.llvm.sh @@ -3,7 +3,14 @@ module restore module use /soft/modulefiles module load llvm -module load mpiwrappers/cray-mpich-llvm module load cudatoolkit-standalone +module load mpiwrappers/cray-mpich-llvm + +#dbg="-g" +dbg="" + +opt="-O3" + +flags="${dbg} -fopenmp ${opt} -fopenmp-offload-mandatory -fopenmp-targets=nvptx64 --offload-arch=sm_80" -mpicxx -cxx=clang++ -fopenmp -fopenmp-targets=nvptx64-nvidia-cuda --cuda-gpu-arch=sm_80 flops.cpp -o flops.llvm.exe +mpicxx -cxx=clang++ ${flags} flops.cpp -o flops.llvm.exe diff --git a/polaris/make.nvhpc.sh b/polaris/make.nvhpc.sh index dd0e1dd8b244b1447ad9f1287d6bbc140f02aea2..9710f6284718b430e2b4b15b35bee00f5b752cfb 100755 --- a/polaris/make.nvhpc.sh +++ b/polaris/make.nvhpc.sh @@ -2,4 +2,12 @@ module restore -CC -mp=gpu flops.cpp -o flops.nvhpc.exe +#dbg="-g" +dbg="" + +opt="-O3" + +#flags="${dbg} ${opt} -mp=gpu -gpu=cc80,cuda11.0" +flags="${dbg} ${opt} -mp=gpu -gpu=cc80" + +CC ${flags} flops.cpp -o flops.nvhpc.exe diff --git a/polaris/run.llvm.sh b/polaris/run.llvm.sh index 3361e329e61316577f586dc8e0ae386dcbcc9d8d..bec107ac4ceaee7785e727ef36d8d959f7445e71 100755 --- a/polaris/run.llvm.sh +++ b/polaris/run.llvm.sh @@ -7,8 +7,8 @@ module load mpiwrappers/cray-mpich-llvm module load cudatoolkit-standalone EXE=./flops.llvm.exe -OUTER_ITER=10 -INNER_ITER=50 +OUTER_ITER=5 +INNER_ITER=200 export OMP_NUM_THREADS=1