Improve transpose basis performance for magma backend
#1385
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
|
Here is an example performance comparison for the non-tensor basis application with P = 20, Q = 24 (p = 3 tet) on a V100. This PR should also help improve performance for the Before:
After:
|
|
Hi @sebastiangrimberg, I'm trying to get caught up on the activity with these PRs. I'd vote for going ahead and merging this into #1301 as you are currently proposing, then we can merge #1301 into main, then rebase #1366 with the updated kernels so that I can create/add the updated MI250X and A100 data to that PR. Is that what you had in mind? Or do you prefer to do #1366 before #1301? |
sebastiangrimberg
force-pushed
the
sjg/magma-transpose-opt
branch
from
0544c4a to
3e5ab5d
Compare
sebastiangrimberg
added a commit
to sebastiangrimberg/libCEED
that referenced
this pull request
Oct 27, 2023
Improve transpose basis performance for `magma` backend
sebastiangrimberg
added a commit
that referenced
this pull request
Nov 1, 2023
Improve transpose basis performance for `magma` backend
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Outdated suggestions cannot be applied.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.
In profiling it turns out that the cost of the
magma_dlasetcall to zero outVfor the basis transpose application was quite expensive and the cause of the transpose kernels being 10-20% slower than the non-transpose ones. This PR removes these calls and modifies the relevantmagmabackend kernels to write toVinstead of accumulating into it after it was set to zero.NOTE: This PR currently targets #1301, and can be either merged in there or we can wait until #1301 is approved and merged to
mainbefore rebasing and merging this one.