[Mlir-commits] [mlir] [mlir][linalg] Add scalable vectorisation for depthwise convolutions (PR #81625)

[Cullen Rhodes]

================
@@ -3171,8 +3253,9 @@ struct Conv1DGenerator
     if (!lhs || !rhs)
       return nullptr;
 
-    if (isa<FloatType>(resTy.getElementType()))
+    if (isa<FloatType>(resTy.getElementType())) {
       return rewriter.create<vector::FMAOp>(loc, lhs, rhs, res);
+    }
----------------
c-rhodes wrote:

unrelated change

https://github.com/llvm/llvm-project/pull/81625