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

[Diego Caballero]

================
@@ -1609,6 +1610,20 @@ static LogicalResult reductionPreconditions(LinalgOp op) {
 }
 
 static LogicalResult vectorizeDynamicLinalgOpPrecondition(linalg::LinalgOp op) {
+  if (auto conv = dyn_cast<linalg::DepthwiseConv1DNwcWcOp>(op.getOperation())) {
----------------
dcaballe wrote:

could we move all this conv specific logic to a specific `*Precondition` function?

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