[Codegen] Replace LLVMGPUVectorize with LLVMGPUTileAndFuse for supported cases #18474
+411
−323
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 adds direct support for bufferizing iree_gpu.barrier_region. Now we can directly handle this operation during bufferization rather than requiring it to be decomposed before lowering. For now this simply bufferizes to two barriers at the beginning and the end of the region. In the future we could opt to either keep the region, allowing for some additional analysis, or drop the barriers in certain cases. Those options are left as TODO and this is kept simple for now.
The way that barriers are currently inserted for forall fusion is fragile and trying to model "WaR" conflicts on tensors (kind of). We instead want to put the barrier around the body of the whole scf.forall. See this comment: iree-org#18490 (comment)
…ted cases The plan is to deprecate the LLVMGPUVectorize pipeline as it is no longer maintained. This patch is the first step towards deprecating it. The pipeline is still used by default for convolutions (if LLVMGPUVectorDistribute does not pick it up). 1) (this) Replace default use cases of LLVMGPUVectorize with LLVMGPUTileAndFuse. 2) Replace convolution focused uses of LLVMGPUVectorize. 3) Separate out the lowering config logic for pre-bufferized dispatches (that targets LLVMGPUBaseDistribute). 4) Delete the vectorize pipeline.
qedawkins
force-pushed
the
deprecate_llvmgpu_vectorize
branch
from
931f7da
to
056fcf2
Compare
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Outdated suggestions cannot be applied.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.
The plan is to deprecate the LLVMGPUVectorize pipeline as it is no
longer maintained. This patch is the first step towards deprecating it.
The pipeline is still used by default for convolutions (if
LLVMGPUVectorDistribute does not pick it up).
LLVMGPUTileAndFuse.
(that targets LLVMGPUBaseDistribute).