Swift
|- Dataset
|- Dataset_samples
|- FigurePlot
|- script
|- SOTA_script
|- src
|- test
clean.sh
compile.sh
Makefile
Readme
SOTA_method_compile.sh
cd Dataset_samples
./Data_get.sh
1 Modify the CUDA-related sections in the Makefile (lines 13 and 17).
2 ./compile.sh
git clone --recursive https://github.com/CRAFT-THU/RoDe.git
2 ./SOTA_method_compile.sh
1 cd script
2 ./Function_test.sh
1 cd script
2 ./Run_CMA_compare.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
3 ./Run_FP32.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
4 ./Run_FP64.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
5 ./Run_idealCMA.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
6 ./Run_irr_opt.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
7 ./Run_irr_regular_ratio.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
8 ./Run_irregular_K.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
9 ./Run_LoadBalance.sh
10 ./Run_MTX_info.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
11 ./Run_preprocess.sh /Direct/to/Swift/Dataset_samples test_mtx_list.txt
12 ./Run_Swift_ncu.sh /Direct/to/Swift/Dataset_samples /Direct/to/Swift/FigurePlot/ncu test_mtx_list.txt
13 ./Run_SOTA_method.sh /Direct/to/Dataset_samples test_mtx_list.txt
14 ./Run_SOTA_preprocess.sh /Direct/to/Dataset_samples test_mtx_list.txt
15 ./Run_SOTA_ncu.sh /Direct/to/Swift/Dataset_samples /Direct/to/Swift/FigurePlot/SOTA_ncu test_mtx_list.txt
1 Down load the dataset of sparse matrix (Section H)
2 Change the dataset path in the script execution command (/Direct/to/Dataset_samples -> /Direct/to/Data).
3 Replace the test_mtx_list.txt to mtxlist.txt for 1-8,10-14 in Section E
./Function_test.sh (This script is used to test if the compilation is successful)
./Run_CMA_compare.sh (This script is used to compare the Swift with or without coalesced memory access)
./Run_FP32.sh (This script is used to test the Swift with FP32)
./Run_FP64.sh (This script is used to test the Swift with FP64)
./Run_idealCMA (This script is used to test the ideal situation of matrix multiplication with or without coalesced memory access)
./Run_irr_opt (This script is used to compare the Swift with or without optimization of the regular part)
./Run_irr_regular_ratio.sh (This script is used to categorize the ratio of regular and irregular parts of the sparse matrix after the pre-processing of Swift)
./Run_irregular_K.sh (This script is used to compare the performances of Swift with cuSPARSE under various irregular K)
./Run_LoadBalance.sh (This script is used to compare the performance of SpMM with and without load balance)
./Run_preprocess.sh (This script is used to test the preprocess time of Swift)
./Run_SOTA_method.sh (This script is used to test the SOTA methods)
./Run_SOTA_preprocess.sh (This script is used to test the preprocess of SOTA methods)
./Run_SOTA_ncu.sh (This script is used to test the memory bandwidth of SOTA via the Nsight compute)
./Run_Swift_ncu.sh (This script is used to test the memory bandwidth of Swift via the Nsight compute)
cd Dataset
./Data_get.sh