[ROCm] Improvements for vectorized elementwise kernels #143269
Closed
Conversation
This file contains hidden or 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
thread_work_size of size 8 and 16 enabled for dtype size 2 and 1 respectively vec_size of 8 and 16 for dtype size 2 and 1 implemented; but not enabled.
DropOut maxed at vec_size of 8
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/143269
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (3 Unrelated Failures)As of commit b779f2b with merge base 95b41d2 ( FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
UNSTABLE - The following job failed but was likely due to flakiness present on trunk and has been marked as unstable:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
8000
jeffdaily
reviewed
Dec 16, 2024
jeffdaily
reviewed
Dec 16, 2024
jeffdaily
approved these changes
Dec 16, 2024
|
@Mellonta has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
@Mellonta , |
Our internal builds include all tests in OSS CI. Could you please fix them? |
|
@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here |
|
Rebase failed due to Command Raised by https://github.com/pytorch/pytorch/actions/runs/12540729651 |
|
@pytorchbot rebase |
|
@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here |
|
Rebase failed due to Command Raised by https://github.com/pytorch/pytorch/actions/runs/12754592663 |
|
@Mellonta |
|
@pytorchbot rebase |
|
@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here |
|
Rebase failed due to Command Raised by https://github.com/pytorch/pytorch/actions/runs/12760808972 |
|
@Mellonta |
|
@Mellonta has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
pruthvistony
added a commit
to ROCm/pytorch
that referenced
this pull request
Jan 31, 2025
… (#1874) * Make io_size calculation as minimum of size of input and output size, rather than the summation of all sizes * for e.g, for torch.add() on half dtypes (bfloat16/float16), calc_io_size() returns 6 causing elems_per_thread to be 4 * But elems_per_thread = 8 works better on half datypes for AMD gpus * Enable *_load_dwordx4 ISA for 16-bit and 8-bit dtypes on AMD gpus by using vector size of 8 and 16 respectively Co-author: @akadutta Pull Request resolved: pytorch#143269 Approved by: https://github.com/jeffdaily, https://github.com/pruthvistony Co-authored-by: Pruthvi Madugundu <pruthvigithub@gmail.com>
jerrymannil
added a commit
to ROCm/pytorch
that referenced
this pull request
Feb 7, 2025
* Make io_size calculation as minimum of size of input and output size, rather than the summation of all sizes * for e.g, for torch.add() on half dtypes (bfloat16/float16), calc_io_size() returns 6 causing elems_per_thread to be 4 * But elems_per_thread = 8 works better on half datypes for AMD gpus * Enable *_load_dwordx4 ISA for 16-bit and 8-bit dtypes on AMD gpus by using vector size of 8 and 16 respectively Co-author: @akadutta Pull Request resolved: pytorch#143269 Approved by: https://github.com/jeffdaily, https://github.com/pruthvistony Co-authored-by: Pruthvi Madugundu <pruthvigithub@gmail.com>
pruthvistony
added a commit
to ROCm/pytorch
that referenced
this pull request
Feb 21, 2025
… (#1924) * Make io_size calculation as minimum of size of input and output size, rather than the summation of all sizes * for e.g, for torch.add() on half dtypes (bfloat16/float16), calc_io_size() returns 6 causing elems_per_thread to be 4 * But elems_per_thread = 8 works better on half datypes for AMD gpus * Enable *_load_dwordx4 ISA for 16-bit and 8-bit dtypes on AMD gpus by using vector size of 8 and 16 respectively Co-author: @akadutta Pull Request resolved: pytorch#143269 Approved by: https://github.com/jeffdaily, https://github.com/pruthvistony Co-authored-by: Pruthvi Madugundu <pruthvigithub@gmail.com>
dnikolaev-amd
pushed a commit
to ROCm/pytorch
that referenced
this pull request
Apr 17, 2025
… (#1874) * Make io_size calculation as minimum of size of input and output size, rather than the summation of all sizes * for e.g, for torch.add() on half dtypes (bfloat16/float16), calc_io_size() returns 6 causing elems_per_thread to be 4 * But elems_per_thread = 8 works better on half datypes for AMD gpus * Enable *_load_dwordx4 ISA for 16-bit and 8-bit dtypes on AMD gpus by using vector size of 8 and 16 respectively Co-author: @akadutta Pull Request resolved: pytorch#143269 Approved by: https://github.com/jeffdaily, https://github.com/pruthvistony Co-authored-by: Pruthvi Madugundu <pruthvigithub@gmail.com> (cherry picked from commit 4686828)
dnikolaev-amd
pushed a commit
to ROCm/pytorch
that referenced
this pull request
Apr 24, 2025
… (#1874) * Make io_size calculation as minimum of size of input and output size, rather than the summation of all sizes * for e.g, for torch.add() on half dtypes (bfloat16/float16), calc_io_size() returns 6 causing elems_per_thread to be 4 * But elems_per_thread = 8 works better on half datypes for AMD gpus * Enable *_load_dwordx4 ISA for 16-bit and 8-bit dtypes on AMD gpus by using vector size of 8 and 16 respectively Co-author: @akadutta Pull Request resolved: pytorch#143269 Approved by: https://github.com/jeffdaily, https://github.com/pruthvistony Co-authored-by: Pruthvi Madugundu <pruthvigithub@gmail.com> (cherry picked from commit 4686828)
dnikolaev-amd
pushed a commit
to ROCm/pytorch
that referenced
this pull request
Apr 24, 2025
… (#1874) * Make io_size calculation as minimum of size of input and output size, rather than the summation of all sizes * for e.g, for torch.add() on half dtypes (bfloat16/float16), calc_io_size() returns 6 causing elems_per_thread to be 4 * But elems_per_thread = 8 works better on half datypes for AMD gpus * Enable *_load_dwordx4 ISA for 16-bit and 8-bit dtypes on AMD gpus by using vector size of 8 and 16 respectively Co-author: @akadutta Pull Request resolved: pytorch#143269 Approved by: https://github.com/jeffdaily, https://github.com/pruthvistony Co-authored-by: Pruthvi Madugundu <pruthvigithub@gmail.com> (cherry picked from commit 4686828)
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.
Co-author: @akadutta
cc @jeffdaily @sunway513 @jithunnair-amd @pruthvistony @ROCmSupport @dllehr-amd @jataylo @hongxiayang @naromero77amd