[Backend] Codegen for ttg.warp_specialize
#5968
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 is in preparation for warp specialization, which will turn the number of warps into a scoped property of regions. This PR just rearranges the API for looking up the number of warps. In the next PR, the `"ttg.num-warps"` attribute will be moved to `tt.func`.
Warp specialization will cause these to become relative to the current warpgroup, so funnel all the code through a set of common APIs.
Mogball
changed the title
ttg.warp_specializettg.warp_specialize
Mogball
requested review from
antiagainst,
zhanglx13,
ptillet and
Jokeren
as code owners
ThomasRaoux
reviewed
Feb 26, 2025
third_party/nvidia/lib/TritonNVIDIAGPUToLLVM/ConvertWarpSpecializeToLLVM.cpp
Show resolved
Hide resolved
Jokeren
reviewed
Feb 26, 2025
| @@ -875,4 +889,26 @@ tt.func @two_different_ws() { | |||
| tt.return | |||
| } | |||
|
|
|||
| // expected-remark @below {{ptr_allocation_datalayout}} | |||
| // expected-remark @below {{size = 8}} | |||
| tt.func @ptr_allocation_datalayout(%arg0: !tt.ptr<i32>) { | |||
There was a problem hiding this comment.
Do you still need this test?
There was a problem hiding this comment.
Yes, but now it's just testing that PointerType can have its size queried (just not through llvm::DataLayout)
Ah yes, let me update that. |
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Outdated suggestions cannot be applied.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 primarily implements
ConvertWarpSpecializeToLLVM, a pass that runs after all other LLVM conversion that rewrites a warp-specialized function, removingttg.warp_specializeops. This pass generates synchronization by putting all other warp groups into a waiting loop, where each warp waits for a state ID to be populated into shared memory. The ID represents the switch case the warp should branch to.This PR also does a few other things:
ttg.total-num-warpsthrough the compiler so that warp specialization works correctly in the frontend