[LLVMCPU] Lower explicit workgroup-local allocs through dispatch ABI#24279
Open
[LLVMCPU] Lower explicit workgroup-local allocs through dispatch ABI#24279
Conversation
Add #iree_codegen.workgroup_local as a codegen memory-space attribute for authored LLVMCPU dispatch allocations that should use HAL workgroup local memory instead of the thread stack. This deliberately matches the GPU-side semantic split: workgroup/shared memory is represented as memref.alloc in a workgroup memory space, while memref.alloca remains private stack scratch. LLVMCPU uses an IREE-specific memory-space attribute instead of #gpu.address_space<workgroup>, but keeps the same alloc/dealloc ownership model. The assignment pass is intentionally narrow: it only handles entry-block memref.alloc operations in HAL executable exports, computes byte ranges with the target data layout and ABI alignment, rejects dynamic or unsupported layouts, and refuses to overwrite existing local-memory assignments or predeclared export requirements. ConvertToLLVM consumes the assigned range by building memref descriptors from the HAL workgroup local-memory pointer. Matching memref.dealloc operations are erased because the storage is owned by the dispatch frame.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Add #iree_codegen.workgroup_local as a codegen memory-space attribute for authored LLVMCPU dispatch allocations that should use HAL workgroup local memory instead of the thread stack.
This deliberately matches the GPU-side semantic split: workgroup/shared memory is represented as memref.alloc in a workgroup memory space, while memref.alloca remains private stack scratch. LLVMCPU uses an IREE-specific memory-space attribute instead of #gpu.address_space, but keeps the same alloc/dealloc ownership model.
The assignment pass is intentionally narrow: it only handles entry-block memref.alloc operations in HAL executable exports, computes byte ranges with the target data layout and ABI alignment, rejects dynamic or unsupported layouts, and refuses to overwrite existing local-memory assignments or predeclared export requirements.
ConvertToLLVM consumes the assigned range by building memref descriptors from the HAL workgroup local-memory pointer. Matching memref.dealloc operations are erased because the storage is owned by the dispatch frame.