The dgSPARSE Library (Deep Graph Sparse Library) is a high performance library for sparse kernel acceleration on GPUs based on CUDA.
.
├── include:
│ └── dgsparse.h: The header file of the dgSPARSE Library.
├── lib:
│ └── dgsparse.so: The dynamic link file of the dgSPARSE Library.
└── src: Some source codes and references of implementations in the dgSPARSE Library.
└── ge-spmm: GE-SpMM implementation.
Check your environment.
export CUDA_HOME=/usr/local/cuda # your cuda path
export LD_LIBRARY_PATH=$CUDA_HOME/lib64
Then build dgsparse through make exp
.
You could run our kernels in the example folder.
Please refer to dgSPARSE Library Documentation for more details.