nifey / nbench

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

nbench

Port of Parboil 2.5 and Rodinia 3.1 benchmarks to CUDA Unified memory (cudaMallocManaged).

Download the parboil dataset from http://impact.crhc.illinois.edu/parboil/parboil_download_page.aspx and rodinia dataset from http://www.cs.virginia.edu/~skadron/lava/Rodinia/Packages/rodinia_3.1.tar.bz2. Then use load_data.sh script to load the data at the expected locations

# After downloading the dataset
bash load_data.sh path/to/pb2.5datasets_standard.tgz path/to/rodinia_3.1.tar.bz2

# Then for running each benchmark
cd bfs
make
./run
  • You can change the dataset used for running by changing the input files and parameters in the run file.
  • sad from parboil and hybridsort, kmeans, mummergpu, myocyte, leukocyte from rodinia benchmarks are not ported as they use old texture API extensively

About


Languages

Language:Cuda 50.3%Language:C 34.1%Language:C++ 14.2%Language:Makefile 1.2%Language:Shell 0.1%