Skip to content

[ET-VK] Using push constants for buffer to image prepack nodes. #11252

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

Conversation

trivedivivek
Copy link
Contributor

@trivedivivek trivedivivek commented May 30, 2025

Stack from ghstack (oldest at bottom):

This diff enables the use of push constants for buffer to image prepack nodes in the Vulkan runtime graph. Push constants are a more efficient way to pass small amounts of data to shaders, compared to using uniform buffers.

  • The nchw_to_*.yaml files have been updated to include the USE_PUSH_CONST flag which is True by default, and enables the use of push constants for all nchw_to_* operations.
  • New variants of the nchw_to_* operation have been added with suffix _no_pc, which do not use push constants. These variants are used for compatibility with testing and utility functions.
  • The Convolution.cpp and Staging.cpp files have been updated to pass empty parameter buffers and instead use push constants.

Differential Revision: D70102398

This diff enables the use of push constants for buffer to image prepack nodes in the Vulkan runtime graph. Push constants are a more efficient way to pass small amounts of data to shaders, compared to using uniform buffers.
*   The `nchw_to_*.yaml` files have been updated to include the `USE_PUSH_CONST` flag which is `True` by default, and enables the use of push constants for all `nchw_to_*` operations.
*   New variants of the `nchw_to_*` operation have been added with suffix `_no_pc`, which do not use push constants. These variants are used for compatibility with testing and utility functions.
*   The `Convolution.cpp` and `Staging.cpp` files have been updated to pass empty parameter buffers and instead use push constants.

Differential Revision: [D70102398](https://our.internmc.facebook.com/intern/diff/D70102398/)

[ghstack-poisoned]
Copy link

pytorch-bot bot commented May 30, 2025

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/11252

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit ec94749 with merge base 1a11267 (image):
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

trivedivivek added a commit that referenced this pull request May 30, 2025
This diff enables the use of push constants for buffer to image prepack nodes in the Vulkan runtime graph. Push constants are a more efficient way to pass small amounts of data to shaders, compared to using uniform buffers.
*   The `nchw_to_*.yaml` files have been updated to include the `USE_PUSH_CONST` flag which is `True` by default, and enables the use of push constants for all `nchw_to_*` operations.
*   New variants of the `nchw_to_*` operation have been added with suffix `_no_pc`, which do not use push constants. These variants are used for compatibility with testing and utility functions.
*   The `Convolution.cpp` and `Staging.cpp` files have been updated to pass empty parameter buffers and instead use push constants.

Differential Revision: [D70102398](https://our.internmc.facebook.com/intern/diff/D70102398/)

ghstack-source-id: 287186740
Pull Request resolved: #11252
@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 May 30, 2025
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D70102398

@trivedivivek trivedivivek added the release notes: vulkan Changes to the Vulkan backend delegate label May 30, 2025
…odes."

This diff enables the use of push constants for buffer to image prepack nodes in the Vulkan runtime graph. Push constants are a more efficient way to pass small amounts of data to shaders, compared to using uniform buffers.
*   The `nchw_to_*.yaml` files have been updated to include the `USE_PUSH_CONST` flag which is `True` by default, and enables the use of push constants for all `nchw_to_*` operations.
*   New variants of the `nchw_to_*` operation have been added with suffix `_no_pc`, which do not use push constants. These variants are used for compatibility with testing and utility functions.
*   The `Convolution.cpp` and `Staging.cpp` files have been updated to pass empty parameter buffers and instead use push constants.

Differential Revision: [D70102398](https://our.internmc.facebook.com/intern/diff/D70102398/)

[ghstack-poisoned]
trivedivivek added a commit that referenced this pull request May 30, 2025
Pull Request resolved: #11252

This diff enables the use of push constants for buffer to image prepack nodes in the Vulkan runtime graph. Push constants are a more efficient way to pass small amounts of data to shaders, compared to using uniform buffers.
*   The `nchw_to_*.yaml` files have been updated to include the `USE_PUSH_CONST` flag which is `True` by default, and enables the use of push constants for all `nchw_to_*` operations.
*   New variants of the `nchw_to_*` operation have been added with suffix `_no_pc`, which do not use push constants. These variants are used for compatibility with testing and utility functions.
*   The `Convolution.cpp` and `Staging.cpp` files have been updated to pass empty parameter buffers and instead use push constants.
ghstack-source-id: 287222797
@exported-using-ghexport

Differential Revision: [D70102398](https://our.internmc.facebook.com/intern/diff/D70102398/)
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D70102398

…odes."

This diff enables the use of push constants for buffer to image prepack nodes in the Vulkan runtime graph. Push constants are a more efficient way to pass small amounts of data to shaders, compared to using uniform buffers.
*   The `nchw_to_*.yaml` files have been updated to include the `USE_PUSH_CONST` flag which is `True` by default, and enables the use of push constants for all `nchw_to_*` operations.
*   New variants of the `nchw_to_*` operation have been added with suffix `_no_pc`, which do not use push constants. These variants are used for compatibility with testing and utility functions.
*   The `Convolution.cpp` and `Staging.cpp` files have been updated to pass empty parameter buffers and instead use push constants.

Differential Revision: [D70102398](https://our.internmc.facebook.com/intern/diff/D70102398/)

[ghstack-poisoned]
trivedivivek added a commit that referenced this pull request May 30, 2025
Pull Request resolved: #11252

This diff enables the use of push constants for buffer to image prepack nodes in the Vulkan runtime graph. Push constants are a more efficient way to pass small amounts of data to shaders, compared to using uniform buffers.
*   The `nchw_to_*.yaml` files have been updated to include the `USE_PUSH_CONST` flag which is `True` by default, and enables the use of push constants for all `nchw_to_*` operations.
*   New variants of the `nchw_to_*` operation have been added with suffix `_no_pc`, which do not use push constants. These variants are used for compatibility with testing and utility functions.
*   The `Convolution.cpp` and `Staging.cpp` files have been updated to pass empty parameter buffers and instead use push constants.
ghstack-source-id: 287225744
@exported-using-ghexport

Differential Revision: [D70102398](https://our.internmc.facebook.com/intern/diff/D70102398/)
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D70102398

…odes."

This diff enables the use of push constants for buffer to image prepack nodes in the Vulkan runtime graph. Push constants are a more efficient way to pass small amounts of data to shaders, compared to using uniform buffers.
*   The `nchw_to_*.yaml` files have been updated to include the `USE_PUSH_CONST` flag which is `True` by default, and enables the use of push constants for all `nchw_to_*` operations.
*   New variants of the `nchw_to_*` operation have been added with suffix `_no_pc`, which do not use push constants. These variants are used for compatibility with testing and utility functions.
*   The `Convolution.cpp` and `Staging.cpp` files have been updated to pass empty parameter buffers and instead use push constants.

Differential Revision: [D70102398](https://our.internmc.facebook.com/intern/diff/D70102398/)

[ghstack-poisoned]
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D70102398

@facebook-github-bot facebook-github-bot merged commit 03b1cc8 into gh/trivedivivek/108/base Jun 3, 2025
98 checks passed
@facebook-github-bot facebook-github-bot deleted the gh/trivedivivek/108/head branch June 3, 2025 04:07
trivedivivek added a commit that referenced this pull request Jun 3, 2025
This PR was created by the merge bot to help merge the original PR into
the main branch.
ghstack PR number: #11252 by
@trivedivivek
^ Please use this as the source of truth for the PR details, comments,
and reviews
ghstack PR base:
https://github.com/pytorch/executorch/tree/gh/trivedivivek/108/base
ghstack PR head:
https://github.com/pytorch/executorch/tree/gh/trivedivivek/108/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/main
Merge bot PR head:
https://github.com/pytorch/executorch/tree/gh/trivedivivek/108/orig
@diff-train-skip-merge

Co-authored-by: Vivek Trivedi <[email protected]>
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 release notes: vulkan Changes to the Vulkan backend delegate
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants