[mlir][Linalg] Add bounded recursion declaration to FMAOp -> LLVM conversion.

FMAOp -> LLVM conversion is done progressively by peeling off 1 dimension from FMAOp at each pattern iteration. Add the recursively bounded property declaration to the pattern so that the rewriter can apply it multiple times.

Without this, FMAOps with 3+D do not lower to LLVM.

Differential Revision: https://reviews.llvm.org/D113886

1 files changed, 6 insertions, 0 deletions

diff --git a/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp b/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
index 966d3f3b8fce..65816a2d0580 100644
--- a/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
+++ b/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
@@ -752,6 +752,12 @@ class VectorFMAOpNDRewritePattern : public OpRewritePattern<FMAOp> {
 public:
   using OpRewritePattern<FMAOp>::OpRewritePattern;
 
+  void initialize() {
+    // This pattern recursively unpacks one dimension at a time. The recursion
+    // bounded as the rank is strictly decreasing.
+    setHasBoundedRewriteRecursion();
+  }
+
   LogicalResult matchAndRewrite(FMAOp op,
                                 PatternRewriter &rewriter) const override {
     auto vType = op.getVectorType();