[Mlir-commits] [mlir] [mlir][linalg] Implement patterns for reducing rank of named linalg contraction ops (PR #95710)

[Oleksandr Alex Zinenko]

================
@@ -1692,6 +1692,13 @@ void populateTransposeMatmulPatterns(RewritePatternSet &patterns,
 void populateBlockPackMatmulPatterns(RewritePatternSet &patterns,
                                      const ControlBlockPackMatmulFn &controlFn);
 
+/// Adds patterns that that reduce the rank of named contraction ops that have
+/// unit dimensions in the operand(s) by converting to a senquence of `collapse_shape`,
+/// `<corresponding linalg named op>`, `expand_shape` (if on tensors).  For example a
+/// `linalg.batch_matmul` with unit batch size will convert to `linalg.matmul`
+/// and a `linalg.matvec` with with unit spatial dim in lhs will convert to a `linalg.dot`.
+void populateContractionOpRankReducingPatterns(RewritePatternSet &patterns);
----------------
ftynse wrote:

A callback is more expensive both in runtime and in API complexity. If static rank is sufficient, let's stick with that until there is a real need in more fine-grained control. 

https://github.com/llvm/llvm-project/pull/95710