Review comment - nit

nod-ai · Nov 29, 2024 · 327fc7d · 327fc7d
1 parent 5257be1
commit 327fc7d
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 28 deletions.
diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIEFuseConsumerIntoLoop.cpp b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIEFuseConsumerIntoLoop.cpp
@@ -79,13 +79,14 @@ void AMDAIEFuseConsumerIntoLoopPass::runOnOperation() {
     do {
       Value::user_range users = computeOp->getResult(0).getUsers();
       if (!llvm::hasSingleElement(users)) {
-        computeOp->emitOpError("Expected only one user of the compute op");
+        LLVM_DEBUG(llvm::dbgs()
+                   << "Expected only one user of the compute op\n");
         return signalPassFailure();
       }
 
-      Operation *terminatorStoreOp = *(users.begin());
+      Operation *candidateSliceOp = *(users.begin());
       if (!(isa<tensor::InsertSliceOp, tensor::ParallelInsertSliceOp>(
-              terminatorStoreOp))) {
+              candidateSliceOp))) {
         computeOp = computeOp->getParentOfType<LoopLikeOpInterface>();
         LLVM_DEBUG(llvm::dbgs()
                    << "Going to reattempt fusion because didn't find "
@@ -94,9 +95,9 @@ void AMDAIEFuseConsumerIntoLoopPass::runOnOperation() {
         continue;
       }
       std::optional<scf::SCFFuseConsumerOfSliceResult> fusedConsumer =
-          scf::tileAndFuseConsumerOfSlice(rewriter, terminatorStoreOp);
+          scf::tileAndFuseConsumerOfSlice(rewriter, candidateSliceOp);
       if (!fusedConsumer) {
-        terminatorStoreOp->emitOpError(
+        candidateSliceOp->emitOpError(
             "Failed to fuse any consumer op into the producer");
         return signalPassFailure();
       }

diff --git a/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/fuse_consumer_into_loop.mlir b/compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/fuse_consumer_into_loop.mlir
@@ -182,19 +182,6 @@ module {
 
 // -----
 
-func.func @no_user_of_producer(%arg0: tensor<64xf32>) {
-  %c0 = arith.constant 0 : index
-  %c1 = arith.constant 1 : index
-  %c8 = arith.constant 8 : index
-  scf.for %arg1 = %c0 to %c8 step %c1 {
-    // expected-error @+1 {{Expected only one user of the compute op}}
-    %1 = linalg.elemwise_binary {fun = #linalg.binary_fn<add>} ins(%arg0, %arg0 : tensor<64xf32>, tensor<64xf32>) outs(%arg0 : tensor<64xf32>) -> tensor<64xf32>
-  }
-  return
-}
-
-// -----
-
 func.func @no_consumer_fusion(%arg0: tensor<64xf32>) -> tensor<64xf32> {
   %c0 = arith.constant 0 : index
   %c1 = arith.constant 1 : index
@@ -388,16 +375,6 @@ module {
 
 // -----
 
-func.func @no_user_of_producer(%arg0: tensor<64xf32>) {
-  scf.forall (%arg1, %arg2) in (1,2) {
-    // expected-error @+1 {{Expected only one user of the compute op}}
-    %1 = linalg.elemwise_binary {fun = #linalg.binary_fn<add>} ins(%arg0, %arg0 : tensor<64xf32>, tensor<64xf32>) outs(%arg0 : tensor<64xf32>) -> tensor<64xf32>
-  }
-  return
-}
-
-// -----
-
 func.func @no_consumer_fusion(%arg0: tensor<64xf32>) -> tensor<64xf32> {
   %0 = scf.forall (%arg1, %arg2) in (1,2) shared_outs(%out = %arg0) -> tensor<64xf32> {
     %1 = linalg.elemwise_binary {fun = #linalg.binary_fn<add>} ins(%arg0, %arg0 : tensor<64xf32>, tensor<64xf32>) outs(%arg0 : tensor<64xf32>) -> tensor<64xf32>