This package based on program CSparse from SuiteSparse 5.4.0
go test -v -cpuprofile cpu.prof -memprofile mem.prof -coverprofile=coverage.out -run=BenchmarkLU -bench=BenchmarkLU -benchmemgoos: linux
goarch: amd64
pkg: github.com/Konstantin8105/sparse
BenchmarkLU/Sparse:__30:__300-4 37153 30785 ns/op 51328 B/op 28 allocs/op
BenchmarkLU/Dense_:__30:__300-4 41449 29516 ns/op 9236 B/op 11 allocs/op
BenchmarkLU/Sparse:_100:_1070-4 9572 108839 ns/op 201328 B/op 28 allocs/op
BenchmarkLU/Dense_:_100:_1070-4 4323 233880 ns/op 85005 B/op 11 allocs/op
BenchmarkLU/Sparse:_300:_3270-4 3404 324523 ns/op 566384 B/op 28 allocs/op
BenchmarkLU/Dense_:_300:_3270-4 476 2494250 ns/op 739173 B/op 20 allocs/op
BenchmarkLU/Sparse:1000:10970-4 1029 1065713 ns/op 1794812 B/op 28 allocs/op
BenchmarkLU/Dense_:1000:10970-4 18 63434980 ns/op 8061283 B/op 47 allocs/op
BenchmarkLU/Sparse:3000:32970-4 378 3206861 ns/op 5382921 B/op 28 allocs/op
BenchmarkLU/Dense_:3000:32970-4 2 802618764 ns/op 72295580 B/op 111 allocs/op
Using sparse algorithm is effective in test case for square matrixes with size more 50.
rm -rf /tmp/CXSparse
cp -R ./CXSparse/ /tmp/
sed -i 's/CS_ENTRY/double/g' /tmp/CXSparse/Source/*.c
sed -i 's/CS_INT/csi/g' /tmp/CXSparse/Source/*.c
meld /tmp/CXSparse/ ./CSparse/- Check new version of
CSparseis exist on page. - Download new version.
- Compare file-by-file with file in folder
CSparseof that package. - Inject changes into Go files of package.
Note: CSparse haven`t any updates at the last few years, so that package is actual at the future.
This package transpiled CSparse from C to Go by c4go.
go test -v -cpuprofile=cpu.prof -memprofile=mem.prof -run=Benchmark -bench=Benchmark -benchmem
go tool pprof cpu.prof
go tool pprof mem.prof
go test -coverprofile=coverage.out -run=TestLU
go tool cover -html=coverage.out
- Variables
css.lnzandcss.unzis not float typedouble, better to use integer type likeint.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent