Use caching allocator for runner (#15730)#15730
Open
kimishpatel wants to merge 58 commits intomainfrom
Open
Conversation
When doing prefill for quantized kv cache, with large prefill length, parallelizing this op helps. Differential Revision: [D84962234](https://our.internmc.facebook.com/intern/diff/D84962234/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D84962234/)! [ghstack-poisoned]
Reason this doesnt directly use Vectorize class is because the equivalent APIs dont exist in Vectorize class Differential Revision: [D84962236](https://our.internmc.facebook.com/intern/diff/D84962236/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D84962236/)! [ghstack-poisoned]
As the title Differential Revision: [D84962233](https://our.internmc.facebook.com/intern/diff/D84962233/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D84962233/)! [ghstack-poisoned]
For small models dequantizing portions of v cache causes extra alloc overhead. Probably a better way to handle this is to dequantize entire v cache outside the model There isnt significant perf advantage from this yet but subsequent diffs will use caching allocator where this refactor help. Differential Revision: [D85532077](https://our.internmc.facebook.com/intern/diff/D85532077/) [ghstack-poisoned]
Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
…allocator" Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
…allocator" Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
…allocator" Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
…allocator" Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
Meant to use this for temp allocator for kernels. Specifically for sdpa, it seems that on iOS there is a significant overhead coming from allocations Differential Revision: [D85532079](https://our.internmc.facebook.com/intern/diff/D85532079/) [ghstack-poisoned]
This allows us to leverage temp memory allocator and if that allocator is caching allocator it reduces the allocaiton overhead. Differential Revision: [D85532076](https://our.internmc.facebook.com/intern/diff/D85532076/) [ghstack-poisoned]
Existing constructors dont compose well such that if you want data loader or data files constructor then you cannot get to override memory allocator. Fix that. Differential Revision: [D86120037](https://our.internmc.facebook.com/intern/diff/D86120037/) [ghstack-poisoned]
kimishpatel
requested review from
jackzhxng,
kirklandsign,
larryliu0820 and
mergennachin
as code owners
This was referenced Nov 10, 2025
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15730
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 2 Cancelled Jobs, 2 Unrelated FailuresAs of commit c75df37 with merge base eef7921 ( CANCELLED JOBS - The following jobs were cancelled. Please retry:
BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This was referenced Nov 10, 2025
kimishpatel
added a commit
that referenced
this pull request
Nov 11, 2025
We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/) ghstack-source-id: 322321964 Pull Request resolved: #15730
meta-cla
Bot
added
the
CLA Signed
kimishpatel
added a commit
that referenced
this pull request
Dec 7, 2025
Pull Request resolved: #15730 We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 327688519 @exported-using-ghexport Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/)
…r runner" We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/) [ghstack-poisoned]
kimishpatel
added a commit
that referenced
this pull request
Dec 9, 2025
Pull Request resolved: #15730 We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 @exported-using-ghexport Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/)
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
meta-codesync
Bot
changed the title
meta-codesync Bot
pushed a commit
that referenced
this pull request
Mar 30, 2026
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038
meta-codesync
Bot
force-pushed
the
gh/kimishpatel/213/head
branch
from
3cd0176 to
32ebe0f
Compare
Contributor
|
@kimishpatel has exported this pull request. If you are a Meta employee, you can view the originating Diff in D86120038. |
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038
meta-codesync
Bot
force-pushed
the
gh/kimishpatel/213/head
branch
from
32ebe0f to
6ebb435
Compare
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
kimishpatel
added a commit
that referenced
this pull request
Apr 15, 2026
Pull Request resolved: #15730 We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 367670359 @exported-using-ghexport Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/)
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
kimishpatel
added a commit
that referenced
this pull request
Apr 16, 2026
Pull Request resolved: #15730 We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 368353572 @exported-using-ghexport Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/)
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
kimishpatel
added a commit
that referenced
this pull request
Apr 17, 2026
Pull Request resolved: #15730 We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 368981082 @exported-using-ghexport Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/)
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
kimishpatel
added a commit
that referenced
this pull request
Apr 17, 2026
Pull Request resolved: #15730 We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 368983883 @exported-using-ghexport Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/)
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
kimishpatel
added a commit
that referenced
this pull request
Apr 20, 2026
Pull Request resolved: #15730 We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 369818674 @exported-using-ghexport Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/)
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
Summary: We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 328001114 exported-using-ghexport Reviewed By: navsud, derekdixu Differential Revision: D86120038 [ghstack-poisoned]
kimishpatel
added a commit
that referenced
this pull request
Apr 24, 2026
Pull Request resolved: #15730 We observed that on iOS it improves perf by 6% because SDPA op does temp allocations. No significant difference on android though. ghstack-source-id: 372501554 @exported-using-ghexport Differential Revision: [D86120038](https://our.internmc.facebook.com/intern/diff/D86120038/)
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
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.
Summary:
We observed that on iOS it improves perf by 6% because SDPA op does temp allocations.
No significant difference on android though.
ghstack-source-id: 328001114
exported-using-ghexport
Reviewed By: navsud, derekdixu
Differential Revision: D86120038