[ET-VK] double, short, and uint16 dtype runtime support #11670
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
Pull Request resolved: #11365 # Context This diff generally aims provide improvements to the existing framework for defining dtype GLSL shader variants, along with setting up support that would be necessary for future shader implementations that wish to support int64 and double dtypes. In order to allow doubles as input/output dtypes for dequantization and quantization, this diff will create the dtype runtime support on the Vulkan backend in Executorch by establishing the relationship between different tensor types and different GLSL types. # Changes The main changes are included in `gen_vulkan_spv.py` which maps the relationship between different dtypes and their GLSL types. For instance, we add aliases for every common dtype which includes `uint8`, `int8`, `uint16`, `int16`, `uint32`, `int32`, `uint64`, `int64`, and `double`. We maintain support for `int`, `uint`, and `bool` alises such that we can avoid making the change overly complex while supporting the most common recognizable alias (int). Furthermore, this diff also modifies the vulkan api to incorporate new types, namely `uint32_t`, `double`, the int16 and int64 variants. We then make sure that the `ShaderNameUtils` (which is commonly used by most operators for creating their variant names), utilizes the new aliasing. Beyond that we also throw an exception to disallow YAML files to include just "int", and to be more specific, like with "int32". We then modify dozens of files to switch to the new alias of int32. Furthermore, we also include double in certain shaders that are used as intermediaries for image to buffer to nchw converisons. ghstack-source-id: 290376491 @exported-using-ghexport Differential Revision: [D75959063](https://our.internmc.facebook.com/intern/diff/D75959063/)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/11670
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit 9a82e98 with merge base 56392aa ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
ahmtox
approved these changes
Jun 14, 2025
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.
This PR was created by the merge bot to help merge the original PR into the main branch.
ghstack PR number: #11365 by @ahmtox
^ 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/ahmtox/7/base
ghstack PR head: https://github.com/pytorch/executorch/tree/gh/ahmtox/7/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/gh/ahmtox/17/orig
Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/ahmtox/7/orig
@diff-train-skip-merge