Merge OpenAI Triton commit 6f5baf6
#2990
Merged
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
This PR refactors the instruction scheduling enums. Now, it is implemented in the MLIR.
This PR implements a specialized codegen for `tt.gather` when it satisfies the conditions of being "warp local": it is possible to compute the output tensor without data movement across warps. `isWarpLocal` is a new function that checks this condition, and places additional restrictions to simplify codegen / separate concerns from `ttg.convert_layout`. This enables `tt.gather` to generate better code when the layout is suitable. In a subsequent PR, a special pattern will be added to generate optimized layouts for `tt.gather` when possible/profitable to enable the lowering.
### Commits in this PR 1. [Pipeliner] Multi-buffer TMA descriptors 2. Add tests for pipelined descriptor creation 3. Be more conservative about number of TMA buffers to allocate 4. Update golden samples 5. Use correct modulus for tma updates
@lezcano pointed out in another PR that the order is confusing because typically we list the lane ID, warp ID, and blockID in this order.
The AMD runner persists changes to the file system between jobs, so the caches need to be manually cleaned up. Closes #5384
This relands triton-lang/triton#5392 to enable new arch target since backend support has been added--it doesn't depend on the reverted LLVM upgrade in triton-lang/triton#5341; basic necessary enablement is already included in the current llvm version we're using.
Enable the TRITON_KERNEL_OVERRIDE feature to work on AMD assembly and binary. Currently, for the backends, it only works on Nvidia `ptx` and `cubin`. --------- Co-authored-by: Yuanwei Fang <[email protected]>
pbchekin
approved these changes
Dec 11, 2024
whitneywhtsang
force-pushed
the
whitneywhtsang/merge
branch
from
0b33ad4 to
e302ae6
Compare
whitneywhtsang
changed the title
f2574796f5baf6
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 change the Triton base from 4d2e9e5 to 6f5baf6 (Dec 10).
Pass rate: 99.82%->99.81% (#2991)
Please do not squash and merge this PR.