Use out-of-place trmm when available #978
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
msimberg
force-pushed
the
rocblas-trmm3
branch
4 times, most recently
from
0695032
to
1e46a36
Compare
msimberg
commented
Sep 8, 2023
Neglect 4.X and require >= 5. |
|
Please run the tests with the CI image to be sure that everything works correctly. |
|
cscs-ci run |
1 similar comment
|
cscs-ci run |
msimberg
force-pushed
the
rocblas-trmm3
branch
from
b12b618
to
5fbaa0d
Compare
|
cscs-ci run |
|
The tests pass on hohgant. I've yet to verify that performance is unchanged. |
msimberg
force-pushed
the
rocblas-trmm3
branch
from
5fbaa0d
to
fe46988
Compare
msimberg
added
Type:Bug
|
There seems to be no measurable impact from this change. |
|
cscs-ci run |
rasolca
approved these changes
Nov 29, 2023
|
cscs-ci run |
|
@msimberg This PR is based on a very old master which makes the CI fails. |
|
cscs-ci run |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I'll try to summarize the various stages of trmm evolution since it's a bit complicated to follow in the code:
rocblas_Xtrmmrocblas_Xtrmm_outofplaceROCBLAS_V3definition that should enable the three-parameter version, but which only seems to have an effect in documentation...Given that
trmm_outofplacewas deprecated in 5.6.Xtrmm_outofplaceI would probably for the moment actually not put an upper bound on when to switch back from
trmm_outofplacetotrmm. It's a guessing game I suppose until they actually release 5.7.0 or 6.0.0 but it seems safer to assume thattrmm_outofplaceis going to actually be removed in the next major release rather than one minor release after it was deprecated.I have not tested this with rocblas 4.X (I'm not even sure if it works without these changes, if I remember correctly CMake support was badly broken before 5.something) but in theory it should work. I don't how much we care about that?
I also have not run large-scale runs with this change, so I don't know if it affects performance (smaller tests say no).
I've also put a version cap on when the workaround for certain
Ops not being supported by rocblas is applied.