Prepare benchmark
ACCEL-SIM provides a repo that collects gpu apps, to get it
git clone https://github.com/accel-sim/gpu-app-collection.git
There are many benchmark in it, we only need ispass-2009 at this LAB, to build it
# in Docker, <CUDA_DIR>=/usr/local/cuda-11.0
export CUDA_INSTALL_PATH=<CUDA_DIR>
export PATH=$CUDA_INSTALL_PATH/bin:$PATH
source ./src/setup_environment
make ispass-2009 -i -j -C ./src
To test the build
ls bin/<cuda-version>/release
If you find ispass-2009-BFS
and ispass-2009-NN
in the output , then the app is built successfully.
We provide data at here. You need to place <data_dirs>
under <gpu-app-collection>
. To test the data
# at <gpu-app-collection>
ls data_dirs/cuda/ispass-2009
If you find ispass-2009-BFS
and ispass-2009-NN
in the output , then the data is prepared successfully.