Raffaele Solcà
Raffaele Solcà
Reported by @RMeli in #603, but moved here as it is reproducible: I get the following failure on hohgant, when running the RANK_6 tests: ``` 57: Test command: /scratch/e1000/rmeli/git/DLA-Future/build2/test/unit/multiplication/test_multiplication_hermitian 57:...
Allow matrices to be used for vectors. - [ ] refactor distribution - [ ] refactor matrix
To simplify memory management and reduce memory usage, tiles should be allocated just before the first use. Matrix constructors should be revised to allow tile by tile allocation (beside given...
As suggested by @albestro: https://github.com/eth-cscs/DLA-Future/pull/574#issuecomment-1162264161
The Cholesky decomposition doens't performs well on a MI50. Using 10240 matrices the double precision performance is just 150 GFlop/s, and increasing the matrix size to 20480 the performance even...
Preconditions rank-k updates written in the proposal document are wrong as A must NOT be a square matrix: https://eel.is/c++draft/linalg#algs.blas3.rankk-3.2 https://eel.is/c++draft/linalg#algs.blas3.rankk-4.1 Similarly, preconditions of rank-2k are wrong. https://eel.is/c++draft/linalg#algs.blas3.rank2k-4.1 and https://eel.is/c++draft/linalg#algs.blas3.rank2k-4.1 should...
I noticed that the doc action on master can fail if a PR is merged while the action is running. It should not be a big problem as a new...
Modified rocm release pipeline: - clang -> gcc12 (cray-mpich%clang can't be installed) - added test stage that run on beverin's MI250x Open tasks for other PRs: - stdexec pipeline -...