Commit df9680e5 authored by Antonio Ragagnin's avatar Antonio Ragagnin 💬
Browse files

Edit leonardo_booster.sh

parent 59ddb204

leonardo_booster.sh

+4 −4
Original line number Original line Diff line number Diff line
@@ -18,8 +18,8 @@ OMP_GPU=100 #number of particles per GPU thread 
#

#

# test tree build on GPU and CPU

# test tree build on GPU and CPU

#

#

OMP_NUM_THREADS=4 CC='gcc' HW_FLAGS='-fopenmp -O3' srun -n 1 python -m hotwheels_core.wrap src/test_target.cpp

OMP_NUM_THREADS=4 CC='gcc' EC_FLAGS='-fopenmp -O3' srun -n 1 python -m hotwheels_core.wrap src/test_target.cpp

OMP_NUM_THREADS=4 CC='nvc++' HW_FLAGS='-mp=gpu -gpu=rdc,managed -gpu=cc80  -Minfo=mp,accel -Minline=10000 -O3' srun -n 1 python -m hotwheels_core.wrap src/test_target.cpp

OMP_NUM_THREADS=4 CC='nvc++' EC_FLAGS='-mp=gpu -gpu=rdc,managed -gpu=cc80  -Minfo=mp,accel -Minline=10000 -O3' srun -n 1 python -m hotwheels_core.wrap src/test_target.cpp





for ppl in "${PPL[@]}" ; do # loop over particles per leaf

for ppl in "${PPL[@]}" ; do # loop over particles per leaf

    for n in "${N[@]}" ; do # loop over n of particles

    for n in "${N[@]}" ; do # loop over n of particles
@@ -27,14 +27,14 @@ for ppl in "${PPL[@]}" ; do # loop over particles per leaf 
            for threads in "${THREADS[@]}" ; do #loop over number of OpenMPThreads

            for threads in "${THREADS[@]}" ; do #loop over number of OpenMPThreads

                if [[ "$mode" == "cpu"  ]]; then

                if [[ "$mode" == "cpu"  ]]; then

                    export CC=gcc

                    export CC=gcc

                    export HW_FLAGS='-O3 -fopenmp'

                    export EC_FLAGS='-O3 -fopenmp'

                    export OMP_NUM_THREADS=$threads

                    export OMP_NUM_THREADS=$threads

                else

                else

                    if [[ $threads != 8 ]]; then

                    if [[ $threads != 8 ]]; then

                       continue

                       continue

                    fi

                    fi

                    export CC=nvc++

                    export CC=nvc++

                    export HW_FLAGS='-mp=gpu -gpu=rdc,managed -gpu=cc80  -Minfo=mp,accel -Minline=1000 -O3'

                    export EC_FLAGS='-mp=gpu -gpu=rdc,managed -gpu=cc80  -Minfo=mp,accel -Minline=1000 -O3'

                    # gpu always uses 32 cpu cores but actually doesnt 

                    # gpu always uses 32 cpu cores but actually doesnt 

                    export OMP_NUM_THREADS=32 change much

                    export OMP_NUM_THREADS=32 change much

                    threads=$(python -c "print(int($N//100))") #for the text hilight: "

                    threads=$(python -c "print(int($N//100))") #for the text hilight: "