[GPU] Add option for no gpu.block_dim in GPUDistributeScfFor #17214
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
|
I will create an issue later about the |
antiagainst
requested changes
Apr 30, 2024
| @@ -38,13 +42,30 @@ struct DistributeLoop final : public OpRewritePattern<scf::ForOp> { | |||
| if (!numDimAttr) | |||
| return failure(); | |||
|
|
|||
| auto funcOp = forOp->getParentOfType<FunctionOpInterface>(); | |||
There was a problem hiding this comment.
These checks are only neccessary and to fail when useBlockDims is false?
| // CHECK: %[[YLB:.+]] = affine.apply affine_map<()[s0, s1, s2] -> (s0 * s1 + s2)>()[%[[ID]], %[[STEP]], %[[LB]]] | ||
| // CHECK: %[[YSTEP:.+]] = affine.apply affine_map<()[s0, s1] -> (s0 * s1)>()[%[[DIM]], %[[STEP]]] | ||
| // CHECK: scf.for %[[IV:.+]] = %[[YLB]] to %[[UB]] step %[[YSTEP]] { | ||
| // CHECK: memref.store %{{.+}}, %{{.+}}[%[[IV]]] : memref<?xf32> | ||
|
|
||
| // NO-BLOCK-DIM-LABEL: func.func @distribute_to_y |
There was a problem hiding this comment.
These tests are duplicated with the above in the sense they don't check additonal stuff? It's fine to just have one of them and remove the others--saves us efforts to update all of them later.
Max191
force-pushed
the
gpu-distribute-no-block-dim
branch
from
9075fed to
3d10c31
Compare
antiagainst
approved these changes
May 1, 2024
LLITCHEV
pushed a commit
to LLITCHEV/iree
that referenced
this pull request
Jul 30, 2024
…g#17214) This PR adds an option to generate `arith.constant` ops with workgroup sizes in place of `gpu.block_dim` ops in `GPUDistributeScfFor`. The `gpu.block_dim` ops are not currently handled properly on ROCM backends, so this option is needed to support `GPUDistributeScfFor` on the LLVMGPU path. Signed-off-by: Lubo Litchev <lubol@google.com>
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 adds an option to generate
arith.constantops with workgroup sizes in place ofgpu.block_dimops inGPUDistributeScfFor. Thegpu.block_dimops are not currently handled properly on ROCM backends, so this option is needed to supportGPUDistributeScfForon the LLVMGPU path.