For more detail in Changelog
The latest tag for monolish container has been removed from this version. Please specify the tag name.
Added
- Add CRS/Dense SOR solver and precondition https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/380 https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/376
- Add CRS.transpose() https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/359 #24
Fixed
- Fix memory leak of cusolver direct solver (getrf, sytrf) https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/364
Changed
- Installtion of monolish for GPU with apt does not support https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/403
- Add nodiscard to solve() https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/400
- Change monolish container name
oss_nvidia
->oss-nvidia
,mkl_nvidia
->mkl-nvidia
https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/398 - Change transpose() function API https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/356 #79
Dense Dense::transpose()
// return A^T -->void Dense::transpose()
// A=A^Tvoid Dense::transpose(const Dense B)
// A=B^T, this function is not changedCOO COO::transpose()
// return A^T -->void COO::transpose()
// A=A^Tvoid COO::transpose(COO B) const
// B=A^T -->void COO::transpose(const COO B)
// A=B^T
- OpenMP parallelized transpose() functions https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/356 #79
- Uses CUDA11 cuSPRASE SpMM https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/355 #76
- cuda11.x cusparse generic API for CRS SpMV https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/352 #65 #75
- cuda 10.1 -> cuda11.4 https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/352 #65
- clang 12.0.0 -> clang12.0.1 https://gitlab.ritc.jp/ricos/monolish/-/merge_requests/352 #65