2017F SNU Multicore Computing Project
Place CIFAR10 dataset at data/cifar-10.bin
The project uses cmake build system. Running cmake
will generate GNU Makefile.
$ cmake .
After generating Makefile you can choose which one to build
$ make omp # for single-cpu implementation
$ make opencl # for single-gpu implementation
$ make opencl2 # for multiple gpu implementation
$ make mpi # for mpi with multiple gpu implementation
$ make snucl # for SNUCL implementation
$ make all # to make all of above
and running
$ thorq --add ./omp <input.bmp> <output.bmp>
$ thorq --add --mode single --device gpu/7970 ./opencl <input.bmp> <output.bmp>
$ thorq --add --mode single --device gpu/7970 ./opencl2 <input.bmp> <output.bmp>
$ thorq --add --mode mpi --node 4 --device gpu/7970 ./mpi <input.bmp> <output.bmp>
$ thorq --add --mode snucl --node 4 --device gpu/7970 ./snucl <input.bmp> <output.bmp>
or you can use my thorq wrapper
$ python3 thorq.py --add ./omp <input.bmp> <output.bmp>
$ python3 thorq.py --add --mode single --device gpu/7970 ./opencl <input.bmp> <output.bmp>
$ python3 thorq.py --add --mode single --device gpu/7970 ./opencl2 <input.bmp> <output.bmp>
$ python3 thorq.py --add --mode mpi --node 4 --device gpu/7970 ./mpi <input.bmp> <output.bmp>
$ python3 thorq.py --add --mode snucl --node 4 --device gpu/7970 ./snucl <input.bmp> <output.bmp>