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

[Cullen Rhodes]

================
@@ -19,10 +19,47 @@ module attributes {transform.with_named_sequence} {
 
 // -----
 
-func.func @depthwise_conv1d_nwc_wc_dyn_ch_dim(%input: memref<3x5x?xf32>, %filter: memref<2x?xf32>, %output: memref<3x2x?xf32>) {
+func.func @depthwise_conv1d_ncw_cw(%input: memref<3x5x4xf32>, %filter: memref<5x1xf32>, %output: memref<3x5x4xf32>) {
----------------
c-rhodes wrote:

please could you add a comment to say the channel dim must be dynamic to vectorize?

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