[Mlir-commits] [mlir] [mlir] Add bufferization option for parallel region check (PR #94645)
llvmlistbot at llvm.org
llvmlistbot at llvm.org
Thu Jun 6 10:34:42 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-mlir
Author: None (Max191)
<details>
<summary>Changes</summary>
Handling parallel region RaW conflicts should usually be the responsibility of the source program, rather than bufferization analysis. However, to preserve current functionality, checks on parallel regions is put behind a bufferization in this PR, which is on by default. Default functionality will not change, but this PR enables the option to leave parallelism checks out of the bufferization analysis.
---
Full diff: https://github.com/llvm/llvm-project/pull/94645.diff
2 Files Affected:
- (modified) mlir/include/mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h (+5)
- (modified) mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp (+1-1)
``````````diff
diff --git a/mlir/include/mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h b/mlir/include/mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h
index 2d8add82383be..2fda091e412ae 100644
--- a/mlir/include/mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h
+++ b/mlir/include/mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h
@@ -309,6 +309,11 @@ struct BufferizationOptions {
/// bufferized or not.
bool bufferizeFunctionBoundaries = false;
+ // Specifies whether to account for parallel regions in RaW analysis. If true,
+ // then writes inside of parallel regions that write to buffers defined
+ // outside of the parallel region will be given a new buffer.
+ bool checkParallelRegions = true;
+
/// Certain ops have aliasing OpOperand/OpResult invariants (e.g., scf.for).
/// If this flag is set to `false`, those invariants are no longer enforced
/// with buffer copies.
diff --git a/mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp b/mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp
index 2d329a1f3d889..d0b4e0dd4383e 100644
--- a/mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp
+++ b/mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp
@@ -611,7 +611,7 @@ hasReadAfterWriteInterference(const DenseSet<OpOperand *> &usesRead,
// Before going through the main RaW analysis, find cases where a buffer must
// be privatized due to parallelism. If the result of a write is never read,
// privatization is not necessary (and large parts of the IR are likely dead).
- if (!usesRead.empty()) {
+ if (options.checkParallelRegions && !usesRead.empty()) {
for (OpOperand *uConflictingWrite : usesWrite) {
// Find the allocation point or last write (definition) of the buffer.
// Note: In contrast to `findDefinitions`, this also returns results of
``````````
</details>
https://github.com/llvm/llvm-project/pull/94645
More information about the Mlir-commits
mailing list