[Mlir-commits] [mlir] [mlir][vector] Add ElementwiseToOuterproduct (PR #93664)

[Andrzej Warzyński]

================
@@ -1795,6 +1795,81 @@ struct BreakDownVectorReduction final : OpRewritePattern<vector::ReductionOp> {
   unsigned maxNumElementsToExtract = 0;
 };
 
+/// Pattern aiming to fold a series of ops mulf(tr(broadcast(A)), broadcast(B))
+/// into vector.outerproduct(A, B) such as :
+/// ```mlir
+///  %lhsBcast = vector.broadcast %lhs : vector<4xi32> to vector<4x4xi32>
+///  %lhsT = vector.transpose %lhsBcast, [1, 0] : vector<4x4xi32> to
+///  vector<4x4xi32> %rhsBcast = vector.broadcast %rhs : vector<4xi32> to
+///  vector<4x4xi32> %mul = arith.muli %lhsT, %rhsBcast : vector<4x4xi32>
+///```
+/// Becomes :
+///```mlir
+///  %res = vector.outerproduct %lhs, %rhs : vector<4xi32>, vector<4xi32>
+///```
+/// Edge Cases where broadcast ops are not 1D to 2D as follow are not handled.
+/// %ex1 = vector.broadcast %lhsCast : vector<1x4xf32> to vector<4x4xf32>
+/// %ex2 = vector.broadcast %lhsCast : f32 to vector<4x4xf32>
+/// %ex3 = vector.broadcast %lhsCast : vector<1x1xf32> to vector<4x4xf32>
+
----------------
banach-space wrote:

> I think removing the ``` is easier to read. 

I agree, though it looks like Hugo is merely trying to be consistent with other patterns in this file. Shall we reformat all MLIR examples in the file? I'm happy to volunteer. 

https://github.com/llvm/llvm-project/pull/93664