[Mlir-commits] [mlir] [mlir][x86] Fix - Replace `load` with `transfer_read` to support on tensor type. (PR #194543)

[Arun Thangamani]

================
@@ -98,7 +98,22 @@ static void packNonUnitDimOperandToVNNI(mlir::PatternRewriter &rewriter,
             });
       });
 
-  auto vec1 = vector::LoadOp::create(rewriter, loc, flatTy, srcBuff, indexVals);
+  int64_t srcRank;
+  if (auto memrefTy = dyn_cast<MemRefType>(srcBuff.getType())) {
----------------
arun-thmn wrote:

This piece is of code is removed as part of getting `rank` via `sharedType`.

https://github.com/llvm/llvm-project/pull/194543