[Inductor cutlass backend] Support for StreamK and non-zero workspace sizes #114687
Conversation
This file contains 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
… sizes WIP [ghstack-poisoned]
This was referenced Nov 27, 2023
This was referenced Nov 27, 2023
kadeng
added a commit
that referenced
this pull request
Nov 28, 2023
… sizes WIP ghstack-source-id: f723905c1913492f02786a8c10269c7586fd7372 Pull Request resolved: #114687
…o workspace sizes" WIP [ghstack-poisoned]
…o workspace sizes" WIP [ghstack-poisoned]
This was referenced Dec 5, 2023
…o workspace sizes" WIP [ghstack-poisoned]
…o workspace sizes" WIP [ghstack-poisoned]
…o workspace sizes" WIP [ghstack-poisoned]
kadeng
added a commit
that referenced
this pull request
Dec 7, 2023
… sizes WIP ghstack-source-id: e330a20e8dbb4132dc7bc1c3d830c780157f95bc Pull Request resolved: #114687
…o workspace sizes" WIP [ghstack-poisoned]
kadeng
added a commit
that referenced
this pull request
Dec 7, 2023
… sizes WIP ghstack-source-id: 36620de4afce44499522e1831a4fe2f0476c69d8 Pull Request resolved: #114687
…o workspace sizes" WIP [ghstack-poisoned]
kadeng
added a commit
that referenced
this pull request
Dec 7, 2023
… sizes WIP ghstack-source-id: 833a41137aba9906e4c9c311a0fcd3d13b16b002 Pull Request resolved: #114687
…o workspace sizes" WIP [ghstack-poisoned]
kadeng
added a commit
that referenced
this pull request
Dec 7, 2023
… sizes WIP ghstack-source-id: 1ec1e9d9ee3434fbc24814f0e90ba3ad8db0d646 Pull Request resolved: #114687
…o workspace sizes" The StreamK tile scheduler offers performance improvements for certain otherwise problematic GEMM shapes. In Cutlass, most StreamK ops require temporary working memory, though. This diff adds support for temporary workspace memory being provided to Cutlass ops and enables StreamK-based GEMM Kernels in the Inductor Cutlass backend. [ghstack-poisoned]
…o workspace sizes" The StreamK tile scheduler offers performance improvements for certain otherwise problematic GEMM shapes. In Cutlass, most StreamK ops require temporary working memory, though. This diff adds support for temporary workspace memory being provided to Cutlass ops and enables StreamK-based GEMM Kernels in the Inductor Cutlass backend. [ghstack-poisoned]
…o workspace sizes" The StreamK tile scheduler offers performance improvements for certain otherwise problematic GEMM shapes. In Cutlass, most StreamK ops require temporary working memory, though. This diff adds support for temporary workspace memory being provided to Cutlass ops and enables StreamK-based GEMM Kernels in the Inductor Cutlass backend. [ghstack-poisoned]
…o workspace sizes" The StreamK tile scheduler offers performance improvements for certain otherwise problematic GEMM shapes. In Cutlass, most StreamK ops require temporary working memory, though. This diff adds support for temporary workspace memory being provided to Cutlass ops and enables StreamK-based GEMM Kernels in the Inductor Cutlass backend. [ghstack-poisoned]
…o workspace sizes" The StreamK tile scheduler offers performance improvements for certain otherwise problematic GEMM shapes. In Cutlass, most StreamK ops require temporary working memory, though. This diff adds support for temporary workspace memory being provided to Cutlass ops and enables StreamK-based GEMM Kernels in the Inductor Cutlass backend. [ghstack-poisoned]
…o workspace sizes" The StreamK tile scheduler offers performance improvements for certain otherwise problematic GEMM shapes. In Cutlass, most StreamK ops require temporary working memory, though. This diff adds support for temporary workspace memory being provided to Cutlass ops and enables StreamK-based GEMM Kernels in the Inductor Cutlass backend. [ghstack-poisoned]
This was referenced Dec 12, 2023
Closed
…o workspace sizes" The StreamK tile scheduler offers performance improvements for certain otherwise problematic GEMM shapes. In Cutlass, most StreamK ops require temporary working memory, though. This diff adds support for temporary workspace memory being provided to Cutlass ops and enables StreamK-based GEMM Kernels in the Inductor Cutlass backend. [ghstack-poisoned]
|
Moved to a (draft) feature branch, see #115919 |
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.
Stack from ghstack (oldest at bottom):
The StreamK tile scheduler offers performance improvements for certain otherwise problematic GEMM shapes. In Cutlass, most StreamK ops require temporary working memory, though.
This diff adds support for temporary workspace memory being provided to Cutlass ops and enables StreamK-based GEMM Kernels in the Inductor Cutlass backend.