[ET-VK] Introduce BufferMetadata GLSL struct to abstract tensor layout
#13595
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13595
Note: Links to docs will display an error until the docs builds have been completed. This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This was referenced Aug 22, 2025
meta-cla
bot
added
the
CLA Signed
This PR needs a
|
|
@SS-JIA has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
This was referenced Aug 22, 2025
|
This pull request was exported from Phabricator. Differential Revision: D80800082 |
|
This pull request was exported from Phabricator. Differential Revision: D80800082 |
manuelcandales
approved these changes
Aug 22, 2025
|
This pull request was exported from Phabricator. Differential Revision: D80800082 |
3 similar comments
|
This pull request was exported from Phabricator. Differential Revision: D80800082 |
|
This pull request was exported from Phabricator. Differential Revision: D80800082 |
|
This pull request was exported from Phabricator. Differential Revision: D80800082 |
SS-JIA
added a commit
that referenced
this pull request
Aug 23, 2025
#13598) Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #13597 * #13596 * #13595 * #13594 * #13593 * #13600 * #13599 * __->__ #13598 Title says it all. For reduce ops, their signature are not all alike so some extra legwork needs to be done to identify specific arguments that need to be checked. Also included a small update to partitioner logging to improve debuggability. Differential Revision: [D80741737](https://our.internmc.facebook.com/intern/diff/D80741737/) --------- Co-authored-by: ssjia <ssjia@devvm5117.ash0.facebook.com>
SS-JIA
added a commit
that referenced
this pull request
Aug 23, 2025
Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #13597 * #13596 * #13595 * #13594 * #13593 * #13600 * __->__ #13599 * #13598 Title says it all! Differential Revision: [D80741734](https://our.internmc.facebook.com/intern/diff/D80741734/) --------- Co-authored-by: ssjia <ssjia@devvm5117.ash0.facebook.com>
SS-JIA
added a commit
that referenced
this pull request
Aug 23, 2025
Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #13597 * #13596 * #13595 * #13594 * #13593 * __->__ #13600 * #13599 * #13598 As title. Use XNNPACK's FuseBatchNorm pass since it can fuse into linear layers as well. Differential Revision: [D80741735](https://our.internmc.facebook.com/intern/diff/D80741735/) Co-authored-by: ssjia <ssjia@devvm5117.ash0.facebook.com>
SS-JIA
added a commit
that referenced
this pull request
Aug 23, 2025
Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #13597 * #13596 * #13595 * #13594 * __->__ #13593 * #13600 * #13599 * #13598 Differential Revision: [D80800081](https://our.internmc.facebook.com/intern/diff/D80800081) Co-authored-by: ssjia <ssjia@devvm5117.ash0.facebook.com>
SS-JIA
added a commit
that referenced
this pull request
Aug 23, 2025
…code (#13594) Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #13597 * #13596 * #13595 * __->__ #13594 * #13593 * #13600 * #13599 * #13598 Differential Revision: [D80800085](https://our.internmc.facebook.com/intern/diff/D80800085) Co-authored-by: ssjia <ssjia@devvm5117.ash0.facebook.com>
Pull Request resolved: #13595 As title; introduce an consolidated metadata UBO for buffer storage that can be used to abstract tensor indexing operations for buffer-backed tensors. This new metadata UBO is capable of representing tensors of up to 8 dimensions. This upper limit is hardcoded, but can be increased later on without needing to update callsites since everything is abstracted by the BufferMetadata struct. Update the following ops to use this new metadata UBO: * staging shaders (nchw_to_buffer and buffer_to_nchw) * binary op @imported-using-ghimport Differential Revision: [D80800082](https://our.internmc.facebook.com/intern/diff/D80800082/) ghstack-source-id: 305143836
SS-JIA
force-pushed
the
gh/SS-JIA/297/head
branch
from
06bb6ff to
5d2cd16
Compare
SS-JIA
added a commit
that referenced
this pull request
Aug 23, 2025
Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #13597 * __->__ #13596 * #13595 * #13594 * #13593 * #13600 * #13599 * #13598 Differential Revision: [D80800083](https://our.internmc.facebook.com/intern/diff/D80800083) Co-authored-by: ssjia <ssjia@devvm5117.ash0.facebook.com>
SS-JIA
added a commit
that referenced
this pull request
Aug 23, 2025
…13597) Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #13597 * #13596 * #13595 * #13594 * #13593 * #13600 * #13599 * #13598 Differential Revision: [D80800084](https://our.internmc.facebook.com/intern/diff/D80800084) Co-authored-by: ssjia <ssjia@devvm5117.ash0.facebook.com>
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.
Stack from ghstack (oldest at bottom):
BufferMetadataGLSL struct to abstract tensor layout #13595Differential Revision: D80800082