[Mlir-commits] [mlir] [mlir][vector] Refine vectorisation of tensor.extract (PR #109580)

[Nirvedh Meshram]

================
@@ -1012,7 +1019,10 @@ getTensorExtractMemoryAccessPattern(tensor::ExtractOp extractOp,
   bool foundIndexOp = false;
   bool isContiguousLoad = isContiguousLoadIdx(linalgOp, extractOpTrailingIdx,
                                               foundIndexOp, resType);
-  isContiguousLoad &= foundIndexOp;
+  // TODO: Support generating contiguous loads for column vectors - that will
+  // require adding a permutation map to tranfer_read Ops.
+  bool isRowVector = resType.getShape().back() != 1;
+  isContiguousLoad &= (foundIndexOp && isRowVector);
----------------
nirvedhmeshram wrote:

Is this logic actually being exercised by the second test case (the currently unsupported contiguous load) ? IIUC  isLoopInvariantIdx will return false for the second case and this line wont be needed?

https://github.com/llvm/llvm-project/pull/109580