[Mlir-commits] [mlir] [mlir][linalg] Remove `computeStaticLoopSizes` (PR #124778)

[llvmlistbot at llvm.org]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-mlir

Author: None (josel-amd)

<details>
<summary>Changes</summary>

`computeStaticLoopSizes()` seems to do the same as `getStaticLoopRanges()`, just in a slightly different way.

This PR is more of a question to the linalg experts. Is there any use for `computeStaticLoopSizes()` or can it be safely removed? :) 

---
Full diff: https://github.com/llvm/llvm-project/pull/124778.diff


3 Files Affected:

- (modified) mlir/include/mlir/Dialect/Linalg/IR/LinalgInterfaces.td (-5) 
- (modified) mlir/lib/Dialect/Linalg/IR/LinalgInterfaces.cpp (-13) 
- (modified) mlir/lib/Dialect/Linalg/Transforms/ConstantFold.cpp (+1-1) 


``````````diff
diff --git a/mlir/include/mlir/Dialect/Linalg/IR/LinalgInterfaces.td b/mlir/include/mlir/Dialect/Linalg/IR/LinalgInterfaces.td
index 244db23925ab3c..4fb2d420093673 100644
--- a/mlir/include/mlir/Dialect/Linalg/IR/LinalgInterfaces.td
+++ b/mlir/include/mlir/Dialect/Linalg/IR/LinalgInterfaces.td
@@ -860,11 +860,6 @@ def LinalgStructuredInterface
     /// `createFlatListOfOperandDims`.
     SmallVector<Range, 4> createLoopRanges(OpBuilder &b, Location loc);
 
-    /// Compute the static loop sizes necessary to vectorize the computation.
-    /// This is done by applying `getShapesToLoopsMap` to
-    /// `createFlatListOfOperandStaticDims`.
-    SmallVector<int64_t, 4> computeStaticLoopSizes();
-
     /// Returns the value that expresses the shape of the output in terms of
     /// shape of the input operands where possible
     LogicalResult reifyResultShapes(OpBuilder &b,
diff --git a/mlir/lib/Dialect/Linalg/IR/LinalgInterfaces.cpp b/mlir/lib/Dialect/Linalg/IR/LinalgInterfaces.cpp
index caf9cdb3a3eb4f..52f2129c77ee3f 100644
--- a/mlir/lib/Dialect/Linalg/IR/LinalgInterfaces.cpp
+++ b/mlir/lib/Dialect/Linalg/IR/LinalgInterfaces.cpp
@@ -1094,19 +1094,6 @@ SmallVector<Range, 4> LinalgOp::createLoopRanges(OpBuilder &b, Location loc) {
   return res;
 }
 
-SmallVector<int64_t, 4> LinalgOp::computeStaticLoopSizes() {
-  AffineMap map = getLoopsToShapesMap();
-  unsigned numDims = map.getNumDims(), numRes = map.getNumResults();
-  SmallVector<int64_t, 4> allShapeSizes = createFlatListOfOperandStaticDims();
-  SmallVector<int64_t, 4> res(numDims, 0);
-  for (unsigned idx = 0; idx < numRes; ++idx) {
-    auto result = map.getResult(idx);
-    if (auto d = dyn_cast<AffineDimExpr>(result))
-      res[d.getPosition()] = allShapeSizes[idx];
-  }
-  return res;
-}
-
 /// Visitor to check if any of the given set of positions from AffineDimExprs
 /// are used within an AffineExpr.
 struct HasAffineDimExprVisitor
diff --git a/mlir/lib/Dialect/Linalg/Transforms/ConstantFold.cpp b/mlir/lib/Dialect/Linalg/Transforms/ConstantFold.cpp
index 2e6079e1402e1d..b53180b5cf7c34 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/ConstantFold.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/ConstantFold.cpp
@@ -130,7 +130,7 @@ class FoldConstantBase : public OpInterfaceRewritePattern<LinalgOp> {
         return failure();
     }
 
-    SmallVector<int64_t, 4> loopBounds = linalgOp.computeStaticLoopSizes();
+    SmallVector<int64_t, 4> loopBounds = linalgOp.getStaticLoopRanges();
     int64_t numElements = outputType.getNumElements();
 
     // Use APInt/APFloat instead of Attribute here for constructing the output.

``````````

</details>


https://github.com/llvm/llvm-project/pull/124778