Install dependencies

System dependencies:

• Python3
• GNU parallel

Install Python dependencies:

python3 -m pip install -r requirements.txt

Perform the following operations in order

Generate VBR Matrices and their Matrix Market equivalent

./scripts/generate_vbr_matrices_w_all_mostly_dense.sh
./scripts/generate_vbr_matrices_w_some_mostly_sparse.sh
python3 gen.py -o vbr_to_mtx
python3 gen.py -o vbr_to_mtx --dense-blocks-only

The corresponding files will be generated in Generated_VBR/, Generated_VBR_Sparse/, Generated_MMarket/ and Generated_MMarket_Sparse/.

Generate VBR matrices from manually partitioned Suitesparse matrices

python3 scripts/convert_real_to_vbr.py

The corresponding files will be generated in manual_vbr/.

SABLE

Run SABLE to generate CPU code to perform SpMV and SpMM over these generated Matrices

python3 gen.py -o vbr_to_spmv
python3 gen.py -o vbr_to_spmv --dense-blocks-only
python3 gen.py -o vbr_to_spmm
python3 gen.py -o vbr_to_spmm --dense-blocks-only

The corresponding files will be generated in Generated_SpMV/, Generated_SpMV_Sparse/, Generated_SpMM/ and Generated_SpMM_Sparse/.

Run SABLE to generate CUDA code to perform SpMV and SpMM over these Matrices

python3 gen.py -o vbr_to_spmv --cuda
python3 gen.py -o vbr_to_spmv --cuda --dense-blocks-only
python3 gen.py -o vbr_to_spmm --cuda
python3 gen.py -o vbr_to_spmv --cuda --dense-blocks-only

The corresponding files will be generated in Generated_SpMV_cuda/, Generated_SpMV_cuda_Sparse, Generated_SpMM_cuda and Generated_SpMM_cuda_Sparse/.

Benchmark SABLE

python3 bench.py
python3 bench_sparse.py
python3 bench_manualvbr.py
python3 bench_cuda.py
python3 bench_sparse_cuda.py

The corresponding files will be generated in results/.