[AMDAIEControlCodeLowering] Separate DmaCpyNdToHalfDmaCpyNd to a new pass

compiler/plugins/target/AMD-AIE/iree-amd-aie/Test/samples/matmul_pack_peel_objectfifo.mlir

@@ -1,6 +1,6 @@
 // This pipeline is obtained by going into Passes.cpp, and dumping the pass pipeline (at the end of addAMDAIEObjectFifoLoweringPasses) using `passManager.dump()`. This test is included, as it can be useful to have a reference in IR of all the passes that are run.
 
-// RUN: iree-opt --pass-pipeline="builtin.module(fold-memref-alias-ops,iree-amdaie-distribute-l1-allocations,iree-amdaie-convert-to-dma,iree-amdaie-normalize-loop-bounds,iree-amdaie-insert-cores,iree-amdaie-localize-logicalobjectfifo,cse,iree-amdaie-distribute-cores-and-objectfifos,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-split-logical-objectfifos-for-connection-reuse,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-assign-tiles,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-dma-to-circular-dma,func.func(iree-amdaie-create-aie-workgroup),cse,iree-amdaie-dma-cse,iree-amdaie-hoist-logical-objectfifo,iree-amdaie-canonicalize-doubly-strided-op{fold-single-dims=false},iree-amdaie-flatten-logicalobjectfifo,iree-amdaie-assign-logical-objectfifo-depth{l1-buffer-depth=2 l2-buffer-depth=2 l3-buffer-depth=1},iree-amdaie-access-to-acquire-release,iree-amdaie-none-access-to-temporary-buffer,iree-amdaie-assign-connection-types,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-dma-composition{only-zero-stride-on-outer-dim=true},cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-dma-cse,iree-amdaie-assign-npu-dma-bd-ids,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-controlcode-loop-unroll,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-dma-cse,iree-amdaie-canonicalize-doubly-strided-op{fold-single-dims=false},canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-convert-core-forall-to-for,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-assign-channels,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-objfifo-bufferization,iree-amdaie-connection-to-flow,iree-amdaie-assign-packet-ids,iree-amdaie-controlcode-lowering,iree-amdaie-controlcode-to-transaction,iree-amdaie-acquire-release-to-use-lock,iree-amdaie-canonicalize-npu-dma-cpy-nd{nb-dimensions=4},canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-sink-into-core,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-lower-to-aie,iree-amdaie-remove-memoryspace)" --split-input-file %s | FileCheck %s
+// RUN: iree-opt --pass-pipeline="builtin.module(fold-memref-alias-ops,iree-amdaie-distribute-l1-allocations,iree-amdaie-convert-to-dma,iree-amdaie-normalize-loop-bounds,iree-amdaie-insert-cores,iree-amdaie-localize-logicalobjectfifo,cse,iree-amdaie-distribute-cores-and-objectfifos,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-split-logical-objectfifos-for-connection-reuse,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-assign-tiles,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-dma-to-circular-dma,func.func(iree-amdaie-create-aie-workgroup),cse,iree-amdaie-dma-cse,iree-amdaie-hoist-logical-objectfifo,iree-amdaie-canonicalize-doubly-strided-op{fold-single-dims=false},iree-amdaie-flatten-logicalobjectfifo,iree-amdaie-assign-logical-objectfifo-depth{l1-buffer-depth=2 l2-buffer-depth=2 l3-buffer-depth=1},iree-amdaie-access-to-acquire-release,iree-amdaie-none-access-to-temporary-buffer,iree-amdaie-assign-connection-types,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-dma-composition{only-zero-stride-on-outer-dim=true},cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-dma-cse,iree-amdaie-assign-npu-dma-bd-ids,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-controlcode-loop-unroll,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-dma-cse,iree-amdaie-canonicalize-doubly-strided-op{fold-single-dims=false},canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-convert-core-forall-to-for,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-assign-channels,cse,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-objfifo-bufferization,iree-amdaie-connection-to-flow,iree-amdaie-assign-packet-ids,iree-amdaie-npu-dma-to-half-dma-cpy-nd,iree-amdaie-controlcode-lowering,iree-amdaie-controlcode-to-transaction,iree-amdaie-acquire-release-to-use-lock,iree-amdaie-canonicalize-npu-dma-cpy-nd{nb-dimensions=4},canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-sink-into-core,canonicalize{  max-iterations=10 max-num-rewrites=-1 region-simplify=normal test-convergence=false top-down=true},iree-amdaie-lower-to-aie,iree-amdaie-remove-memoryspace)" --split-input-file %s | FileCheck %s
 
 
 

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIEControlCodeLowering.cpp

@@ -18,80 +18,6 @@
 
 namespace mlir::iree_compiler::AMDAIE {
 
-struct DmaCpyNdToHalfDmaCpyNdConverter final
-    : OpConversionPattern<AMDAIE::NpuDmaCpyNdOp> {
-  using OpConversionPattern::OpConversionPattern;
-
-  LogicalResult matchAndRewrite(
-      AMDAIE::NpuDmaCpyNdOp dmaOp, OpAdaptor adaptor,
-      ConversionPatternRewriter &rewriter) const override {
-    LLVM_DEBUG(llvm::dbgs() << "matchAndRewrite[AMDAIE::NpuDmaCpyNdOp]\n");
-    AMDAIE::ConnectionOp connectionOp = dmaOp.getConnectionOp();
-    if (!connectionOp) {
-      return dmaOp.emitOpError()
-             << "should operate on an `amdaie.connection` op";
-    }
-    // Convert source half.
-    Value source =
-        dmaOp.getSource() ? dmaOp.getSource() : connectionOp.getSource();
-    if (connectionOp.getSourceChannels().size() != 1)
-      return connectionOp.emitOpError() << "expected a single source channel";
-    auto sourceChannelOp = dyn_cast<AMDAIE::ChannelOp>(
-        connectionOp.getSourceChannels()[0].getDefiningOp());
-    bool hasAsyncSourceToken =
-        llvm::any_of(dmaOp.getAsyncTokens(), [](Value token) {
-          return isa<AMDAIE::AsyncSourceTokenType>(token.getType());
-        });
-    SmallVector<Type> resultTypes = {
-        rewriter.getType<AMDAIE::AsyncTokenType>()};
-    TypeRange sourceResultTypes =
-        hasAsyncSourceToken ? TypeRange{resultTypes} : TypeRange{};
-    rewriter.setInsertionPoint(dmaOp);
-    auto sourceDma = rewriter.create<AMDAIE::NpuHalfDmaCpyNdOp>(
-        dmaOp.getLoc(), sourceResultTypes, connectionOp, source,
-        dmaOp.getSourceMixedOffsets(), dmaOp.getSourceMixedSizes(),
-        dmaOp.getSourceMixedStrides(), dmaOp.getSourceBdId(), sourceChannelOp);
-
-    // Convert target half.
-    Value target =
-        dmaOp.getTarget() ? dmaOp.getTarget() : connectionOp.getTarget();
-    if (connectionOp.getTargetChannels().size() != 1)
-      return connectionOp.emitOpError() << "expected a single target channel";
-    auto targetChannelOp = dyn_cast<AMDAIE::ChannelOp>(
-        connectionOp.getTargetChannels()[0].getDefiningOp());
-    bool hasAsyncTargetToken =
-        llvm::any_of(dmaOp.getAsyncTokens(), [](Value token) {
-          return isa<AMDAIE::AsyncTargetTokenType>(token.getType());
-        });
-    TypeRange targetResultTypes =
-        hasAsyncTargetToken ? TypeRange{resultTypes} : TypeRange{};
-    auto targetDma = rewriter.create<AMDAIE::NpuHalfDmaCpyNdOp>(
-        dmaOp.getLoc(), targetResultTypes, connectionOp, target,
-        dmaOp.getTargetMixedOffsets(), dmaOp.getTargetMixedSizes(),
-        dmaOp.getTargetMixedStrides(), dmaOp.getTargetBdId(), targetChannelOp);
-    if (dmaOp.getNumResults() == 1) {
-      if (sourceDma.getNumResults() == 1) {
-        rewriter.replaceUsesWithIf(
-            dmaOp.getResult(0), sourceDma.getResult(0), [&](OpOperand &use) {
-              return isa<AMDAIE::AsyncSourceTokenType>(use.get().getType()) &&
-                     isa<AMDAIE::NpuDmaWaitOp>(use.getOwner());
-            });
-      }
-      if (targetDma.getNumResults() == 1) {
-        rewriter.replaceUsesWithIf(
-            dmaOp.getResult(0), targetDma.getResult(0), [&](OpOperand &use) {
-              return isa<AMDAIE::AsyncTargetTokenType>(use.get().getType()) &&
-                     isa<AMDAIE::NpuDmaWaitOp>(use.getOwner());
-            });
-      }
-      if (!dmaOp.getResult(0).use_empty())
-        return dmaOp.emitOpError() << "should not have any uses anymore";
-    }
-    rewriter.eraseOp(dmaOp);
-    return success();
-  }
-};
-
 struct HalfDmaCpyNdToNpuConverter final
     : OpConversionPattern<AMDAIE::NpuHalfDmaCpyNdOp> {
   using OpConversionPattern::OpConversionPattern;
@@ -159,22 +85,18 @@ struct HalfDmaCpyNdToNpuConverter final
         if (stride == 0) {
           repeatCount = size;
         } else {
-          iterationStride =
-              std::max(stride * elemWidthInBits / minStrideBitWidth,
-                       (int64_t)1);
+          iterationStride = std::max(
+              stride * elemWidthInBits / minStrideBitWidth, (int64_t)1);
           iterationSize = size;
-          if (stride == 1)
-            size = (size * elemWidthInBits) / minStrideBitWidth;
+          if (stride == 1) size = (size * elemWidthInBits) / minStrideBitWidth;
           repeatCount = iterationSize;
         }
       } else {
         staticStrides.push_back(
-            std::max(stride * elemWidthInBits / minStrideBitWidth,
-                     (int64_t)1));
+            std::max(stride * elemWidthInBits / minStrideBitWidth, (int64_t)1));
         // Innermost size needs to account for addressing granularity.
         if (iter.index() == (sizes.size() - 1)) {
-          staticSizes.push_back(size * elemWidthInBits /
-                                minStrideBitWidth);
+          staticSizes.push_back(size * elemWidthInBits / minStrideBitWidth);
         } else {
           staticSizes.push_back(size);
         }
@@ -323,9 +245,7 @@ void AMDAIEControlCodeLoweringPass::runOnOperation() {
   RewritePatternSet patterns(context);
   ConversionTarget conversionTarget(*context);
   conversionTarget.addLegalDialect<AMDAIEDialect>();
-  conversionTarget
-      .addIllegalOp<AMDAIE::NpuHalfDmaCpyNdOp, AMDAIE::NpuDmaCpyNdOp>();
-  patterns.insert<DmaCpyNdToHalfDmaCpyNdConverter>(context);
+  conversionTarget.addIllegalOp<AMDAIE::NpuHalfDmaCpyNdOp>();
   patterns.insert<HalfDmaCpyNdToNpuConverter>(context, deviceModel);
   if (failed(applyPartialConversion(parentOp, conversionTarget,
                                     std::move(patterns)))) {

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIENpuDmaToHalfDmaCpyNd.cpp

@@ -0,0 +1,126 @@
+// Copyright 2024 The IREE Authors
+//
+// Licensed under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+#include <algorithm>
+
+#include "iree-amd-aie/IR/AMDAIEOps.h"
+#include "iree-amd-aie/Transforms/AMDAIEDmaUtils.h"
+#include "iree-amd-aie/Transforms/AMDAIEUtils.h"
+#include "iree-amd-aie/Transforms/Passes.h"
+#include "iree-amd-aie/Transforms/Transforms.h"
+#include "mlir/Transforms/DialectConversion.h"
+#include "mlir/Transforms/GreedyPatternRewriteDriver.h"
+
+#define DEBUG_TYPE "iree-amdaie-npu-dma-to-half-dma-cpy-nd"
+
+namespace mlir::iree_compiler::AMDAIE {
+
+struct NpuDmaToHalfDmaCpyNdConverter final
+    : OpConversionPattern<AMDAIE::NpuDmaCpyNdOp> {
+  using OpConversionPattern::OpConversionPattern;
+
+  LogicalResult matchAndRewrite(
+      AMDAIE::NpuDmaCpyNdOp dmaOp, OpAdaptor adaptor,
+      ConversionPatternRewriter &rewriter) const override {
+    LLVM_DEBUG(llvm::dbgs() << "matchAndRewrite[AMDAIE::NpuDmaCpyNdOp]\n");
+    AMDAIE::ConnectionOp connectionOp = dmaOp.getConnectionOp();
+    if (!connectionOp) {
+      return dmaOp.emitOpError()
+             << "should operate on an `amdaie.connection` op";
+    }
+    // Convert source half.
+    Value source =
+        dmaOp.getSource() ? dmaOp.getSource() : connectionOp.getSource();
+    if (connectionOp.getSourceChannels().size() != 1)
+      return connectionOp.emitOpError() << "expected a single source channel";
+    auto sourceChannelOp = dyn_cast<AMDAIE::ChannelOp>(
+        connectionOp.getSourceChannels()[0].getDefiningOp());
+    bool hasAsyncSourceToken =
+        llvm::any_of(dmaOp.getAsyncTokens(), [](Value token) {
+          return isa<AMDAIE::AsyncSourceTokenType>(token.getType());
+        });
+    SmallVector<Type> resultTypes = {
+        rewriter.getType<AMDAIE::AsyncTokenType>()};
+    TypeRange sourceResultTypes =
+        hasAsyncSourceToken ? TypeRange{resultTypes} : TypeRange{};
+    rewriter.setInsertionPoint(dmaOp);
+    auto sourceDma = rewriter.create<AMDAIE::NpuHalfDmaCpyNdOp>(
+        dmaOp.getLoc(), sourceResultTypes, connectionOp, source,
+        dmaOp.getSourceMixedOffsets(), dmaOp.getSourceMixedSizes(),
+        dmaOp.getSourceMixedStrides(), dmaOp.getSourceBdId(), sourceChannelOp);
+
+    // Convert target half.
+    Value target =
+        dmaOp.getTarget() ? dmaOp.getTarget() : connectionOp.getTarget();
+    if (connectionOp.getTargetChannels().size() != 1)
+      return connectionOp.emitOpError() << "expected a single target channel";
+    auto targetChannelOp = dyn_cast<AMDAIE::ChannelOp>(
+        connectionOp.getTargetChannels()[0].getDefiningOp());
+    bool hasAsyncTargetToken =
+        llvm::any_of(dmaOp.getAsyncTokens(), [](Value token) {
+          return isa<AMDAIE::AsyncTargetTokenType>(token.getType());
+        });
+    TypeRange targetResultTypes =
+        hasAsyncTargetToken ? TypeRange{resultTypes} : TypeRange{};
+    auto targetDma = rewriter.create<AMDAIE::NpuHalfDmaCpyNdOp>(
+        dmaOp.getLoc(), targetResultTypes, connectionOp, target,
+        dmaOp.getTargetMixedOffsets(), dmaOp.getTargetMixedSizes(),
+        dmaOp.getTargetMixedStrides(), dmaOp.getTargetBdId(), targetChannelOp);
+    if (dmaOp.getNumResults() == 1) {
+      if (sourceDma.getNumResults() == 1) {
+        rewriter.replaceUsesWithIf(
+            dmaOp.getResult(0), sourceDma.getResult(0), [&](OpOperand &use) {
+              return isa<AMDAIE::AsyncSourceTokenType>(use.get().getType()) &&
+                     isa<AMDAIE::NpuDmaWaitOp>(use.getOwner());
+            });
+      }
+      if (targetDma.getNumResults() == 1) {
+        rewriter.replaceUsesWithIf(
+            dmaOp.getResult(0), targetDma.getResult(0), [&](OpOperand &use) {
+              return isa<AMDAIE::AsyncTargetTokenType>(use.get().getType()) &&
+                     isa<AMDAIE::NpuDmaWaitOp>(use.getOwner());
+            });
+      }
+      if (!dmaOp.getResult(0).use_empty())
+        return dmaOp.emitOpError() << "should not have any uses anymore";
+    }
+    rewriter.eraseOp(dmaOp);
+    return success();
+  }
+};
+
+namespace {
+class AMDAIENpuDmaToHalfDmaCpyNdPass
+    : public impl::AMDAIENpuDmaToHalfDmaCpyNdBase<
+          AMDAIENpuDmaToHalfDmaCpyNdPass> {
+ public:
+  void getDependentDialects(DialectRegistry &registry) const override {
+    registry.insert<AMDAIEDialect>();
+  }
+  void runOnOperation() override;
+};
+
+void AMDAIENpuDmaToHalfDmaCpyNdPass::runOnOperation() {
+  Operation *parentOp = getOperation();
+  MLIRContext *context = &getContext();
+  RewritePatternSet patterns(context);
+  ConversionTarget conversionTarget(*context);
+  conversionTarget.addLegalDialect<AMDAIEDialect>();
+  conversionTarget.addIllegalOp<AMDAIE::NpuDmaCpyNdOp>();
+  patterns.insert<NpuDmaToHalfDmaCpyNdConverter>(context);
+  if (failed(applyPartialConversion(parentOp, conversionTarget,
+                                    std::move(patterns)))) {
+    return signalPassFailure();
+  }
+}
+
+}  // namespace
+
+std::unique_ptr<Pass> createAMDAIENpuDmaToHalfDmaCpyNdPass() {
+  return std::make_unique<AMDAIENpuDmaToHalfDmaCpyNdPass>();
+}
+
+}  // namespace mlir::iree_compiler::AMDAIE

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/CMakeLists.txt

@@ -95,6 +95,7 @@ iree_cc_library(
     "AMDAIEMapForallToCores.cpp"
     "AMDAIENoneAccessToTemporaryBuffer.cpp"
     "AMDAIENormalizeLoopBounds.cpp"
+    "AMDAIENpuDmaToHalfDmaCpyNd.cpp"
     "AMDAIEObjFifoBufferization.cpp"
     "AMDAIEPackAndTranspose.cpp"
     "AMDAIEPad.cpp"

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/PassDetail.h

@@ -73,6 +73,7 @@ namespace mlir::iree_compiler::AMDAIE {
 #define GEN_PASS_DEF_AMDAIEMAPFORALLTOCORES
 #define GEN_PASS_DEF_AMDAIENONEACCESSTOTEMPORARYBUFFER
 #define GEN_PASS_DEF_AMDAIENORMALIZELOOPBOUNDS
+#define GEN_PASS_DEF_AMDAIENPUDMATOHALFDMACPYND
 #define GEN_PASS_DEF_AMDAIEOBJFIFOBUFFERIZATION
 #define GEN_PASS_DEF_AMDAIEPACKANDTRANSPOSE
 #define GEN_PASS_DEF_AMDAIECONVERTTODMA

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.cpp

@@ -664,6 +664,7 @@ void addAMDAIEObjectFifoLoweringPasses(
   passManager.addPass(createAMDAIEConnectionToFlowPass());
   passManager.addPass(createAMDAIEAssignPacketIdsPass());
 
+  passManager.addPass(createAMDAIENpuDmaToHalfDmaCpyNdPass());
   passManager.addPass(createAMDAIEControlCodeLoweringPass());
   passManager.addPass(createAMDAIEControlCodeToTransactionPass());
 

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.h

@@ -121,7 +121,8 @@ std::unique_ptr<Pass> createAMDAIEControlCodeForallToForPass();
 /// Pass to unroll the loops within the control code regions.
 std::unique_ptr<Pass> createAMDAIEControlCodeLoopUnrollPass();
 
-/// Pass to convert control code DMA operations into NPU writes and syncs.
+/// Pass to convert control code HalfDmaCpyNd into NPU WriteBd, AddressPatch,
+/// PushToQueue operations.
 std::unique_ptr<Pass> createAMDAIEControlCodeLoweringPass();
 
 /// Pass to convert control code into a transaction binary.
@@ -253,6 +254,9 @@ std::unique_ptr<Pass> createAMDAIENoneAccessToTemporaryBufferPass();
 /// Normalize the loop bounds of `scf.for` and `scf.forall`.
 std::unique_ptr<Pass> createAMDAIENormalizeLoopBoundsPass();
 
+/// Pass to convert control code DMA operations into HalfDmaCpyNd.
+std::unique_ptr<Pass> createAMDAIENpuDmaToHalfDmaCpyNdPass();
+
 /// Create a pass to bufferize logical objectFifos.
 std::unique_ptr<Pass> createAMDAIEObjFifoBufferizationPass();
 

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/Passes.td

@@ -532,6 +532,12 @@ def AMDAIENormalizeLoopBounds :
   let constructor = "mlir::iree_compiler::AMDAIE::createAMDAIENormalizeLoopBoundsPass()";
 }
 
+def AMDAIENpuDmaToHalfDmaCpyNd :
+    Pass<"iree-amdaie-npu-dma-to-half-dma-cpy-nd", ""> {
+  let summary = "Lower DmaCpNd ops to HalfDmaCpyNd ops";
+  let constructor = "mlir::iree_compiler::AMDAIE::createAMDAIENpuDmaToHalfDmaCpyNdPass()";
+}
+
 def AMDAIEObjFifoBufferization :
     Pass<"iree-amdaie-objfifo-bufferization", ""> {
   let summary = "Bufferize logical objectFifos.";

compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/CMakeLists.txt

@@ -63,6 +63,7 @@ iree_lit_test_suite(
     "map_forall_to_cores.mlir"
     "none_access_to_temporary_buffer.mlir"
     "normalize_loop_bounds.mlir"
+    "npu_dma_to_half_dma_cpy_nd.mlir"
     "obj_fifo_bufferization.mlir"
     "pack_and_transpose_level1.mlir"
     "pack_and_transpose_level2.mlir"