Skip to content

[ET-VK] Using uint16 for quantized linear tiling shader to reduce register pressure and improve performance. #10509

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 11 commits into from
Apr 28, 2025

Conversation

trivedivivek
Copy link
Contributor

@trivedivivek trivedivivek commented Apr 28, 2025

Stack from ghstack (oldest at bottom):

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: D73752090

…rk to improve thread occupancy and performance.

This diff changes tiled 8 bit quantized linear mat mul op to linearly dispatch work which increases thread occupancy and improves performance.

Differential Revision: [D73751979](https://our.internmc.facebook.com/intern/diff/D73751979/)

[ghstack-poisoned]
…ister pressure and improve performance.

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
Copy link

pytorch-bot bot commented Apr 28, 2025

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/10509

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.

trivedivivek added a commit that referenced this pull request Apr 28, 2025
…ister pressure and improve performance.

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

ghstack-source-id: 280586466
Pull Request resolved: #10509
@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Apr 28, 2025
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D73752090

…tiling shader to reduce register pressure and improve performance."

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
… reduce register pressure and improve performance."

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
trivedivivek added a commit that referenced this pull request Apr 28, 2025
…ister pressure and improve performance.

Pull Request resolved: #10509

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.
ghstack-source-id: 280651868
@exported-using-ghexport

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D73752090

…tiling shader to reduce register pressure and improve performance."

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
… reduce register pressure and improve performance."

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
trivedivivek added a commit that referenced this pull request Apr 28, 2025
…ister pressure and improve performance.

Pull Request resolved: #10509

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.
ghstack-source-id: 280715504
@exported-using-ghexport

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D73752090

…tiling shader to reduce register pressure and improve performance."

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
… reduce register pressure and improve performance."

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
trivedivivek added a commit that referenced this pull request Apr 28, 2025
…ister pressure and improve performance.

Pull Request resolved: #10509

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.
ghstack-source-id: 280719995
@exported-using-ghexport

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D73752090

…tiling shader to reduce register pressure and improve performance."

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
… reduce register pressure and improve performance."

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)

[ghstack-poisoned]
trivedivivek added a commit that referenced this pull request Apr 28, 2025
…ister pressure and improve performance.

Pull Request resolved: #10509

This diff reduces int precision for certain variables in 8 bit quantized tiled linear op to reduce register pressure and improve performance.
ghstack-source-id: 280739898
@exported-using-ghexport

Differential Revision: [D73752090](https://our.internmc.facebook.com/intern/diff/D73752090/)
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D73752090

SS-JIA pushed a commit that referenced this pull request Apr 28, 2025
…rk to improve thread occupancy and performance. (#10508)

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at
bottom):
* #10509
* __->__ #10508

This diff changes tiled 8 bit quantized linear mat mul op to linearly
dispatch work which increases thread occupancy and improves performance.

Differential Revision:
[D73751979](https://our.internmc.facebook.com/intern/diff/D73751979/)
@SS-JIA SS-JIA changed the base branch from gh/trivedivivek/79/base to main April 28, 2025 21:57
@SS-JIA SS-JIA merged commit 6ca0800 into main Apr 28, 2025
80 of 81 checks passed
@SS-JIA SS-JIA deleted the gh/trivedivivek/79/head branch April 28, 2025 22:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported topic: not user facing
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants