[Mlir-commits] [mlir] [mlir][x86] Decouple Accumulator from ADD-Based Vector Contractions (over loops) (PR #204327)

Julian Oppermann llvmlistbot at llvm.org
Thu Jul 23 02:03:02 PDT 2026


================
@@ -373,18 +373,22 @@ struct VectorContractToPackedTypeDotProduct
           return rewriter.notifyMatchFailure(
               contractOp, "The store/write operation of contract operation is "
                           "before the pair contract operation");
-        // Shuffle the accumulators of the contract operations.
-        LogicalResult readShuffle =
-            shuffleAfterReadLikeOp(rewriter, accReadOp0, accReadOp1, contractOp,
-                                   pairContractOp, nonUnitDimValue, accTy);
 
-        if (failed(readShuffle))
-          return rewriter.notifyMatchFailure(
-              contractOp, "Accumulator read is not by transfer_read or load");
+        if (!(dyn_cast<arith::ConstantOp>(accReadOp0))) {
----------------
jopperm wrote:

Same as above.

https://github.com/llvm/llvm-project/pull/204327


More information about the Mlir-commits mailing list