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

[Andrzej Warzyński]

================
@@ -3186,20 +3259,21 @@ struct Conv1DGenerator
     for (int64_t kw = 0; kw < kwSize; ++kw) {
       for (int64_t w = 0; w < wSize; w += wSizeStep) {
         lhsVals.push_back(rewriter.create<vector::ExtractStridedSliceOp>(
-            loc, lhs,
+            loc, maybeMaskedLhs->getResult(0),
             /*offsets=*/ArrayRef<int64_t>{0, w * strideW + kw * dilationW, 0},
             inOutSliceSizes, inOutStrides));
       }
     }
     // Extract rhs slice of size {c} @ [kw].
     for (int64_t kw = 0; kw < kwSize; ++kw) {
       rhsVals.push_back(rewriter.create<vector::ExtractOp>(
-          loc, rhs, /*offsets=*/ArrayRef<int64_t>{kw}));
+          loc, maybeMaskedRhs->getResult(0),
+          /*offsets=*/ArrayRef<int64_t>{kw}));
     }
     // Extract res slice: {n, wSizeStep, c} @ [0, w, 0].
     for (int64_t w = 0; w < wSize; w += wSizeStep) {
       resVals.push_back(rewriter.create<vector::ExtractStridedSliceOp>(
-          loc, res,
+          loc, maybeMaskedRes->getResult(0),
----------------
banach-space wrote:

That's a great catch, thanks! This should be captured by one of the pre-conditions and that's what I'll do.

We don't have a mechanism within TD to trigger this case, which means it's hard to test it. And I'm not sure whether extending TD just to test things is the right thing 🤔  I need to think of a use-case other than testing 😅 

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