[Mlir-commits] [mlir] [MLIR] Folding unpack and pack sequence in data layout propagation from padded domain (PR #138332)

Zhuoran Yin llvmlistbot at llvm.org
Tue May 6 09:48:04 PDT 2025


================
@@ -298,20 +298,56 @@ getOrCreatePackedViewOfOperand(OpBuilder &b, Location loc, PackInfo packInfo,
   return std::make_tuple(packedOperand, indexingMap);
 }
 
+static bool isGenericOutsNotUsed(linalg::GenericOp genericOp) {
+  int numDpsOuts = genericOp.getNumDpsInits();
+  for (int i = 0; i < numDpsOuts; ++i) {
+    Block *block = genericOp.getBody();
+    int numBlockArgs = block->getNumArguments();
+    int matchingInitArgIndex = numBlockArgs - numDpsOuts + i;
+    return block->getArgument(matchingInitArgIndex).use_empty();
+  }
+  return true;
+}
+
 /// Pack a genericOp and return it.
 static GenericOp packGenericOp(RewriterBase &rewriter, GenericOp genericOp,
                                Value dest, AffineMap packedOutIndexingMap,
                                const PackInfo &packInfo) {
   Location loc = genericOp.getLoc();
   SmallVector<Value> inputOperands;
+  SmallVector<Value> inputOperandsFromUnpackedSource;
   SmallVector<AffineMap> indexingMaps;
+
+  // Note: canUnpackPackFold needs to also guarantee the generic body
+  // doesn't have gather semantics. Since such scenarios has been
+  // rejected by both BubbleUpPackOpThroughGenericOp and
+  // PushDownUnPackOpThroughGenericOp, we can safely assume
+  // canUnpackPackFold is as long as init is not used.
+  bool canUnpackPackFold = isGenericOutsNotUsed(genericOp);
----------------
jerryyin wrote:

> We need either an assertion or the actual check

Sounds good, will just go with an actual check as I currently do.

> my assumption is that you only want to support elementwise operations when all the outs are not used

Yes, exactly. I think I misunderstood about what `isElementwise()` and probably misunderstood it as `isUnaryOp()`, therefore giving a non-relevant counter example in my last response. Then upon second review, I realized that this is dependent on element mappable traits which all arith op carries. I'll make sure to add this `isElementwise()` in the condition check. Thanks for raising the concern.

https://github.com/llvm/llvm-project/pull/138332


More information about the Mlir-commits mailing list