Fix Warpspec Matmul to be compatible with OmniFm Shapes #298
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
Summary: Fixes a couple tutorial assumptions, most notably: 1. It only worked with fp8 and fp16. Now it works with all dtypes for OmniFm. 2. The shapes were not compatible due to layout mismatches. Since every shape will have a layout mismatch this adds an explicit tranpose to allow benchmarking a "best case" although this may not be accurate. 3. Some shapes will never be compatible with TMA as the strides are not divisble by 16. I added an explicit check in the code to simplify this issue, but I will be skipping these. Differential Revision: D77950060
njriasan
temporarily deployed
to
docker-s3-upload
GitHub Actions
Inactive
njriasan
temporarily deployed
to
docker-s3-upload
GitHub Actions
Inactive
|
This pull request was exported from Phabricator. Differential Revision: D77950060 |
facebook-github-bot
pushed a commit
that referenced
this pull request
Jul 22, 2025
Summary: Fixes a couple tutorial assumptions, most notably: 1. It only worked with fp8 and fp16. Now it works with all dtypes for OmniFm. 2. The shapes were not compatible due to layout mismatches. Since every shape will have a layout mismatch this adds an explicit tranpose to allow benchmarking a "best case" although this may not be accurate. 3. Some shapes will never be compatible with TMA as the strides are not divisble by 16. I added an explicit check in the code to simplify this issue, but I will be skipping these. Differential Revision: D77950060
facebook-github-bot
pushed a commit
that referenced
this pull request
Jul 22, 2025
Summary: Fixes a couple tutorial assumptions, most notably: 1. It only worked with fp8 and fp16. Now it works with all dtypes for OmniFm. 2. The shapes were not compatible due to layout mismatches. Since every shape will have a layout mismatch this adds an explicit tranpose to allow benchmarking a "best case" although this may not be accurate. 3. Some shapes will never be compatible with TMA as the strides are not divisble by 16. I added an explicit check in the code to simplify this issue, but I will be skipping these. Differential Revision: D77950060
njriasan
added a commit
that referenced
this pull request
Jul 22, 2025
Summary: Fixes a couple tutorial assumptions, most notably: 1. It only worked with fp8 and fp16. Now it works with all dtypes for OmniFm. 2. The shapes were not compatible due to layout mismatches. Since every shape will have a layout mismatch this adds an explicit tranpose to allow benchmarking a "best case" although this may not be accurate. 3. Some shapes will never be compatible with TMA as the strides are not divisble by 16. I added an explicit check in the code to simplify this issue, but I will be skipping these. Reviewed By: PaulZhang12 Differential Revision: D77950060
njriasan
added a commit
that referenced
this pull request
Jul 22, 2025
Summary: Fixes a couple tutorial assumptions, most notably: 1. It only worked with fp8 and fp16. Now it works with all dtypes for OmniFm. 2. The shapes were not compatible due to layout mismatches. Since every shape will have a layout mismatch this adds an explicit tranpose to allow benchmarking a "best case" although this may not be accurate. 3. Some shapes will never be compatible with TMA as the strides are not divisble by 16. I added an explicit check in the code to simplify this issue, but I will be skipping these. Reviewed By: PaulZhang12 Differential Revision: D77950060
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:
Fixes a couple tutorial assumptions, most notably:
Differential Revision: D77950060