Mixed precision
Integrate a few patches in the mixed precision reduction/restore kernels required for the branch potrf MP.
Merge request reports
Activity
changed milestone to %Chameleon 1.3.0
assigned to @faverge
added 5 commits
- a2468f0d - lapacke_dlamch: Add a chameleon wrapper to encapsulate and refactor the test...
- a3995108 - Factorize calls to lapacke_[sd]lamch by replacing them by calls to CHAMELEON[sd]lamch
- e73ffe5a - testing/parameter: Make sure we don't use mixed precision with global and...
- 5b45db8c - Preventing the conversion in half precision for diagonal tiles
- 18d20a78 - Adapting gered and gerst codelets cuda parts
Toggle commit listadded 5 commits
- 2e71ddb1 - lapacke_dlamch: Add a chameleon wrapper to encapsulate and refactor the test...
- b6a26ed7 - Factorize calls to lapacke_[sd]lamch by replacing them by calls to CHAMELEON[sd]lamch
- 3288250e - testing/parameter: Make sure we don't use mixed precision with global and...
- cf37cbb0 - Preventing the conversion in half precision for diagonal tiles
- 8cd96bbc - Adapting gered and gerst codelets cuda parts
Toggle commit list- Resolved by Mathieu Faverge
- Resolved by Mathieu Faverge
- Resolved by Mathieu Faverge
- Resolved by Mathieu Faverge
- Resolved by Mathieu Faverge
- Resolved by Mathieu Faverge
- Resolved by Mathieu Faverge
- Resolved by Mathieu Faverge
Please register or sign in to reply