[Mlir-commits] [mlir] [mlir][vector] Remove MatrixMultiplyOp and FlatTransposeOp from Vector dialect (PR #144307)

[Adam Siemieniuk]

================
@@ -303,6 +303,28 @@ void populateVectorRankReducingFMAPattern(RewritePatternSet &patterns);
 void populateVectorToFromElementsToShuffleTreePatterns(
     RewritePatternSet &patterns, PatternBenefit benefit = 1);
 
+/// Populate the pattern set with the following patterns:
+///
+/// [ContractionOpToMatmulOpLowering]
+/// Lowers `vector.transpose` to `llvm.intr.matrix.flat_transpose`.
+///
+/// Given the high benefit, this will be prioriotised over other
+/// transpose-lowering patterns. As such, the convert-vector-to-llvm pass will
+/// only run this registration conditionally.
----------------
adam-smnk wrote:

I think this and the below descriptions have switched places 😉

https://github.com/llvm/llvm-project/pull/144307