VNEC

Computing Environment

Hardware

Intel(R) Xeon(R) Gold 6240 CPU @ 2.60GHz Kunpeng 920 CPU @ 2.60GHz

Software

cmake (3.22.1-1ubuntu1.22.04.1) googletest

Build

mkdir build
cd build
cmake ..
make -j

Dataset Preparation

cd data
# fetch the .tar.gz packages and decompress them to .mtx files.
make all
# convert .mtx files to .sg and .wsg files.
make convert
# create graph data links for each test group
make link

.sg and .wsg format graph can be loaded at a much higher speed than .mtx format graph, which needs more pre-processing.

Note that, for some network reasons, some sparse matrix data cannot be downloaded. In this case, you need to download the data manually from http://sparse.tamu.edu/ and put the mat.tar.gz in the VNEC/data/graph/mat folder. And then re-execute the command

make convert

Run test demos in fp64

cd build
./bin/OpenSpB_test_operation --gtest_filter=mxv_test_VNEC_FP64.amazon*

Note that the test cases (sparse matrices) are written in the google test specification at the end of src/operation/test/test_spmv_vnec.cpp.

Compile and run

./run_vnec.sh