finish helper refactoring

use uniqued constant helpers everywhere
materialize transposed constants directly

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/Conv.cpp

@@ -61,9 +61,9 @@ static Value createPaddedRows(Value tensorValue,
     padBlock->addArgument(rewriter.getIndexType(), loc);
   padOp.getRegion().push_back(padBlock);
   rewriter.setInsertionPointToStart(padBlock);
-  auto zero = arith::ConstantOp::create(
-    rewriter, loc, tensorType.getElementType(), rewriter.getZeroAttr(tensorType.getElementType()));
-  tensor::YieldOp::create(rewriter, loc, zero.getResult());
+  auto zero = getOrCreateConstant(rewriter, padOp.getOperation(), rewriter.getZeroAttr(tensorType.getElementType()),
+                                  tensorType.getElementType());
+  tensor::YieldOp::create(rewriter, loc, zero);
   rewriter.setInsertionPointAfter(padOp);
   return padOp.getResult();
 }
@@ -106,7 +106,7 @@ static Value buildPackedWeight(DenseElementsAttr wDenseAttr,
   }
 
   auto packedAttr = DenseElementsAttr::get(packedWeightType, packedValues);
-  return arith::ConstantOp::create(rewriter, loc, packedWeightType, packedAttr);
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), packedAttr, packedWeightType);
 }
 
 static Value createConvWeightMatrix(Value w,
@@ -158,7 +158,7 @@ static Value buildPackedBias(bool hasBias,
 
   auto packedBiasType = RankedTensorType::get({1, packFactor * numChannelsOut}, outType.getElementType());
   auto packedBiasAttr = DenseElementsAttr::get(packedBiasType, packedValues);
-  return arith::ConstantOp::create(rewriter, loc, packedBiasType, packedBiasAttr).getResult();
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), packedBiasAttr, packedBiasType);
 }
 
 static Value createIm2colRowComputes(Value x,
@@ -214,8 +214,8 @@ static Value createIm2colRowComputes(Value x,
           padBlock->addArgument(rewriter.getIndexType(), loc);
         padOp.getRegion().push_back(padBlock);
         rewriter.setInsertionPointToStart(padBlock);
-        auto zero = arith::ConstantOp::create(rewriter, loc, elemType, rewriter.getFloatAttr(elemType, 0.0));
-        tensor::YieldOp::create(rewriter, loc, zero.getResult());
+        auto zero = getOrCreateConstant(rewriter, padOp.getOperation(), rewriter.getFloatAttr(elemType, 0.0), elemType);
+        tensor::YieldOp::create(rewriter, loc, zero);
         rewriter.setInsertionPointAfter(padOp);
         paddedInput = padOp.getResult();
       }
@@ -223,13 +223,14 @@ static Value createIm2colRowComputes(Value x,
       // Build im2col [numPatches, patchSize] incrementally to keep the IR small
       // until the late PIM unrolling step.
       Value im2colInit = tensor::EmptyOp::create(rewriter, loc, im2colType.getShape(), elemType);
-      auto c0 = arith::ConstantIndexOp::create(rewriter, loc, 0);
-      auto c1 = arith::ConstantIndexOp::create(rewriter, loc, 1);
-      auto cNumPatches = arith::ConstantIndexOp::create(rewriter, loc, numPatches);
-      auto cNumPatchesPerBatch = arith::ConstantIndexOp::create(rewriter, loc, numPatchesPerBatch);
-      auto cOutWidth = arith::ConstantIndexOp::create(rewriter, loc, outWidth);
-      auto cStrideHeight = arith::ConstantIndexOp::create(rewriter, loc, strideHeight);
-      auto cStrideWidth = arith::ConstantIndexOp::create(rewriter, loc, strideWidth);
+      Operation* anchorOp = rewriter.getInsertionBlock()->getParentOp();
+      auto c0 = getOrCreateIndexConstant(rewriter, anchorOp, 0);
+      auto c1 = getOrCreateIndexConstant(rewriter, anchorOp, 1);
+      auto cNumPatches = getOrCreateIndexConstant(rewriter, anchorOp, numPatches);
+      auto cNumPatchesPerBatch = getOrCreateIndexConstant(rewriter, anchorOp, numPatchesPerBatch);
+      auto cOutWidth = getOrCreateIndexConstant(rewriter, anchorOp, outWidth);
+      auto cStrideHeight = getOrCreateIndexConstant(rewriter, anchorOp, strideHeight);
+      auto cStrideWidth = getOrCreateIndexConstant(rewriter, anchorOp, strideWidth);
 
       auto im2colLoop = scf::ForOp::create(rewriter, loc, c0, cNumPatches, c1, ValueRange {im2colInit});
       rewriter.setInsertionPointToStart(im2colLoop.getBody());

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/Elementwise.cpp

@@ -83,7 +83,7 @@ static FailureOr<Value> materializeBroadcastedConstantTensor(Value value,
   }
 
   auto broadcastedAttr = DenseElementsAttr::get(resultType, resultValues);
-  return arith::ConstantOp::create(rewriter, loc, resultType, broadcastedAttr).getResult();
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), broadcastedAttr, resultType);
 }
 
 static FailureOr<Value>
@@ -121,7 +121,7 @@ static FailureOr<Value> materializeReciprocalTensor(Value value,
   }
 
   auto reciprocalAttr = DenseFPElementsAttr::get(resultType, reciprocalValues);
-  return arith::ConstantOp::create(rewriter, loc, resultType, reciprocalAttr).getResult();
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), reciprocalAttr, resultType);
 }
 
 template <typename OnnxOp, typename SpatialOp>

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/Gemm.cpp

@@ -50,38 +50,17 @@ materializeScaledConstantTensor(Value value, float factor, ConversionPatternRewr
     return failure();
 
   auto scaledAttr = DenseFPElementsAttr::get(cast<RankedTensorType>(denseAttr.getType()), scaledValues);
-  return arith::ConstantOp::create(rewriter, loc, denseAttr.getType(), scaledAttr).getResult();
-}
-
-static Value transposeForSpatial(Value value,
-                                 RankedTensorType resultType,
-                                 ArrayRef<int64_t> permutation,
-                                 ConversionPatternRewriter& rewriter,
-                                 Location loc) {
-  return transposeMaybeInCompute(value, resultType, permutation, rewriter, loc);
-}
-
-static Value
-multiplyIndexByConstant(Value value, int64_t multiplier, ConversionPatternRewriter& rewriter, Location loc) {
-  return onnx_mlir::multiplyIndexByConstant(rewriter, value.getDefiningOp(), value, multiplier);
-}
-
-static Value modIndexByConstant(Value value, int64_t divisor, ConversionPatternRewriter& rewriter, Location loc) {
-  return onnx_mlir::modIndexByConstant(rewriter, loc, value, divisor);
-}
-
-static Value createGemmBatchRow(Value lane, int64_t numOutRows, ConversionPatternRewriter& rewriter, Location loc) {
-  return modIndexByConstant(lane, numOutRows, rewriter, loc);
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), scaledAttr, denseAttr.getType());
 }
 
 static Value createGemmBatchKOffset(
   Value lane, int64_t numOutRows, int64_t numKSlices, ConversionPatternRewriter& rewriter, Location loc) {
   if (numKSlices == 1)
-    return getOrCreateHostIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
+    return getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
 
   MLIRContext* context = rewriter.getContext();
   AffineExpr d0 = getAffineDimExpr(0, context);
-  return createAffineApplyOrConstant(
+  return createAffineApplyOrFoldedConstant(
     rewriter, loc, (d0.floorDiv(numOutRows) % numKSlices) * crossbarSize.getValue(), ValueRange {lane});
 }
 
@@ -92,11 +71,11 @@ static Value createGemmBatchHOffset(Value lane,
                                     ConversionPatternRewriter& rewriter,
                                     Location loc) {
   if (numOutHSlices == 1)
-    return getOrCreateHostIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
+    return getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
 
   MLIRContext* context = rewriter.getContext();
   AffineExpr d0 = getAffineDimExpr(0, context);
-  return createAffineApplyOrConstant(
+  return createAffineApplyOrFoldedConstant(
     rewriter, loc, d0.floorDiv(numOutRows * numKSlices) * crossbarSize.getValue(), ValueRange {lane});
 }
 
@@ -115,9 +94,9 @@ createZeroPaddedTensor(Value value, RankedTensorType resultType, ConversionPatte
     padBlock->addArgument(rewriter.getIndexType(), loc);
   padOp.getRegion().push_back(padBlock);
   rewriter.setInsertionPointToStart(padBlock);
-  auto zero = arith::ConstantOp::create(
-    rewriter, loc, sourceType.getElementType(), rewriter.getZeroAttr(sourceType.getElementType()));
-  tensor::YieldOp::create(rewriter, loc, zero.getResult());
+  auto zero = getOrCreateConstant(
+    rewriter, padOp.getOperation(), rewriter.getZeroAttr(sourceType.getElementType()), sourceType.getElementType());
+  tensor::YieldOp::create(rewriter, loc, zero);
   rewriter.setInsertionPointAfter(padOp);
   return padOp.getResult();
 }
@@ -149,7 +128,7 @@ static FailureOr<Value> materializePaddedConstantMatrix(Value value,
       resultValues[row * resultShape[1] + col] = sourceValues[row * sourceShape[1] + col];
 
   auto resultAttr = DenseElementsAttr::get(resultType, resultValues);
-  return arith::ConstantOp::create(rewriter, loc, resultType, resultAttr).getResult();
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), resultAttr, resultType);
 }
 
 static FailureOr<Value> materializePaddedBroadcastedConstantTensor(Value value,
@@ -215,7 +194,7 @@ static FailureOr<Value> materializePaddedBroadcastedConstantTensor(Value value,
   }
 
   auto resultAttr = DenseElementsAttr::get(resultType, resultValues);
-  return arith::ConstantOp::create(rewriter, loc, resultType, resultAttr).getResult();
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), resultAttr, resultType);
 }
 
 static FailureOr<Value> prepareBias(Value c,
@@ -274,7 +253,7 @@ static spatial::SpatComputeBatch createVmmBatch(Value a,
   const int64_t laneCount = partialPiecesType.getDimSize(0);
   auto batchOp = createSpatComputeBatch(
     rewriter, loc, TypeRange {partialPiecesType}, laneCount, ValueRange {b}, ValueRange {a}, [&](detail::SpatComputeBatchBodyArgs args) {
-      Value row = createGemmBatchRow(args.lane, numOutRows, rewriter, loc);
+      Value row = onnx_mlir::modIndexByConstant(rewriter, loc, args.lane, numOutRows);
       Value kOffset = createGemmBatchKOffset(args.lane, numOutRows, numKSlices, rewriter, loc);
       Value hOffset = createGemmBatchHOffset(args.lane, numOutRows, numKSlices, numOutHSlices, rewriter, loc);
 
@@ -312,12 +291,7 @@ static Value createDynamicGemmBatchRow(
 
   MLIRContext* context = rewriter.getContext();
   AffineExpr d0 = getAffineDimExpr(0, context);
-  return createAffineApplyOrConstant(rewriter, loc, d0.floorDiv(numOutCols), ValueRange {lane});
-}
-
-static Value createDynamicGemmBatchColumn(
-  Value lane, int64_t numOutCols, ConversionPatternRewriter& rewriter, Location loc) {
-  return modIndexByConstant(lane, numOutCols, rewriter, loc);
+  return createAffineApplyOrFoldedConstant(rewriter, loc, d0.floorDiv(numOutCols), ValueRange {lane});
 }
 
 static Value
@@ -385,7 +359,7 @@ static Value createScalarTensorConstant(RankedTensorType scalarType,
   auto elementType = scalarType.getElementType();
   auto scalarAttr = rewriter.getFloatAttr(elementType, value);
   auto denseAttr = DenseElementsAttr::get(scalarType, scalarAttr);
-  return arith::ConstantOp::create(rewriter, loc, scalarType, denseAttr).getResult();
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), denseAttr, scalarType);
 }
 
 static Value createBroadcastedBiasScalar(Value bias,
@@ -435,7 +409,7 @@ static spatial::SpatComputeBatch createVvdmulBatch(Value a,
   auto batchOp = createSpatComputeBatch(
     rewriter, loc, TypeRange {scalarPiecesType}, laneCount, ValueRange {}, ValueRange {a, b}, [&](detail::SpatComputeBatchBodyArgs args) {
       Value row = createDynamicGemmBatchRow(args.lane, numOutCols, rewriter, loc);
-      Value column = createDynamicGemmBatchColumn(args.lane, numOutCols, rewriter, loc);
+      Value column = onnx_mlir::modIndexByConstant(rewriter, loc, args.lane, numOutCols);
 
       auto vectorType = RankedTensorType::get({1, reductionSize}, aType.getElementType());
       auto scalarType = RankedTensorType::get({1, 1}, outType.getElementType());
@@ -475,16 +449,16 @@ static spatial::SpatCompute createDynamicGemmOutputCompute(Value scalarPieces,
     Value biasArg = bias ? blockArgs[1] : Value();
     auto scalarType = RankedTensorType::get({1, 1}, outType.getElementType());
     Value outputInit = tensor::EmptyOp::create(rewriter, loc, outType.getShape(), outType.getElementType()).getResult();
-    Value c0 = getOrCreateHostIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
-    Value c1 = getOrCreateHostIndexConstant(rewriter,  rewriter.getInsertionBlock()->getParentOp(), 1);
-    Value cLaneCount = getOrCreateHostIndexConstant(rewriter,  rewriter.getInsertionBlock()->getParentOp(), laneCount);
+    Value c0 = getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
+    Value c1 = getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 1);
+    Value cLaneCount = getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), laneCount);
     auto loop = scf::ForOp::create(rewriter, loc, c0, cLaneCount, c1, ValueRange {outputInit});
     rewriter.setInsertionPointToStart(loop.getBody());
 
     Value lane = loop.getInductionVar();
     Value outputAcc = loop.getRegionIterArgs().front();
     Value row = createDynamicGemmBatchRow(lane, numOutCols, rewriter, loc);
-    Value column = createDynamicGemmBatchColumn(lane, numOutCols, rewriter, loc);
+    Value column = onnx_mlir::modIndexByConstant(rewriter, loc, lane, numOutCols);
     SmallVector<OpFoldResult> scalarOffsets {lane, rewriter.getIndexAttr(0)};
     SmallVector<OpFoldResult> scalarSizes {rewriter.getIndexAttr(1), rewriter.getIndexAttr(1)};
     SmallVector<OpFoldResult> unitStrides {rewriter.getIndexAttr(1), rewriter.getIndexAttr(1)};
@@ -522,7 +496,7 @@ static Value createPartialGroupOffset(Value hSlice,
                                       Location loc) {
   MLIRContext* context = rewriter.getContext();
   AffineExpr d0 = getAffineDimExpr(0, context);
-  return createAffineApplyOrConstant(
+  return createAffineApplyOrFoldedConstant(
     rewriter, loc, d0 * (numKSlices * numOutRows) + kSlice * numOutRows, ValueRange {hSlice});
 }
 
@@ -604,7 +578,9 @@ static spatial::SpatCompute createReductionCompute(Value partialPieces,
     auto buildOutputSlice = [&](Value outputAcc, Value hSlice) -> Value {
       Value reduced =
         reducePartialPiecesForHSlice(partialPiecesArg, hSlice, pieceType, numKSlices, numOutRows, rewriter, loc);
-      Value hOffset = multiplyIndexByConstant(hSlice, crossbarSize.getValue(), rewriter, loc);
+      Value hOffset =
+        onnx_mlir::multiplyIndexByConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), hSlice,
+                                           crossbarSize.getValue());
       if (biasArg) {
         SmallVector<OpFoldResult> biasOffsets {rewriter.getIndexAttr(0), hOffset};
         Value biasSlice =
@@ -620,13 +596,14 @@ static spatial::SpatCompute createReductionCompute(Value partialPieces,
 
     Value paddedOutput = outputInit;
     if (numOutHSlices == 1) {
-      Value hSlice = getOrCreateHostIndexConstant(rewriter,  rewriter.getInsertionBlock()->getParentOp(), 0);
+      Value hSlice = getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
       paddedOutput = buildOutputSlice(outputInit, hSlice);
     }
     else {
-      Value c0 = getOrCreateHostIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
-      Value c1 = getOrCreateHostIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 1);
-      Value cOutHSlices = getOrCreateHostIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), numOutHSlices);
+      Value c0 = getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
+      Value c1 = getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 1);
+      Value cOutHSlices =
+        getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), numOutHSlices);
       auto hLoop = scf::ForOp::create(rewriter, loc, c0, cOutHSlices, c1, ValueRange {outputInit});
       rewriter.setInsertionPointToStart(hLoop.getBody());
 
@@ -763,7 +740,7 @@ LogicalResult GemmToSpatialComputes::matchAndRewrite(ONNXGemmOp gemmOp,
   if (gemmOpAdaptor.getTransB()) {
     auto bShape = bType.getShape();
     auto transposedType = RankedTensorType::get({bShape[1], bShape[0]}, bType.getElementType());
-    b = transposeForSpatial(b, transposedType, {1, 0}, rewriter, loc);
+    b = transposeMaybeInCompute(b, transposedType, {1, 0}, rewriter, loc);
     bType = cast<RankedTensorType>(b.getType());
   }
 

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/ReduceMean.cpp

@@ -76,7 +76,7 @@ static Value computeLaneIndex(Value lane,
                               ConversionPatternRewriter& rewriter,
                               Location loc) {
   if (dimSize == 1)
-    return arith::ConstantIndexOp::create(rewriter, loc, 0);
+    return getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0);
 
   MLIRContext* context = rewriter.getContext();
   AffineExpr d0 = getAffineDimExpr(0, context);
@@ -85,7 +85,7 @@ static Value computeLaneIndex(Value lane,
     expr = expr.floorDiv(stride);
   if (dimSize != 1)
     expr = expr % dimSize;
-  return createAffineApplyOrConstant(rewriter, loc, expr, ValueRange {lane});
+  return createAffineApplyOrFoldedConstant(rewriter, loc, expr, ValueRange {lane});
 }
 
 static FailureOr<Value> buildReduceMeanKeepdimsBatch(Value input,
@@ -236,7 +236,7 @@ static Value squeezeReducedAxes(Value keepdimsValue,
                                 Location loc) {
   if (resultType.getRank() == 0) {
     SmallVector<Value> indices(cast<RankedTensorType>(keepdimsValue.getType()).getRank(),
-                               arith::ConstantIndexOp::create(rewriter, loc, 0));
+                               getOrCreateIndexConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), 0));
     Value element = tensor::ExtractOp::create(rewriter, loc, keepdimsValue, indices);
     return tensor::FromElementsOp::create(rewriter, loc, resultType, ValueRange {element});
   }
@@ -268,7 +268,7 @@ struct ReduceMeanToSpatialCompute : OpConversionPattern<ONNXReduceMeanV13Op> {
       return success();
     }
 
-    auto axes = normalizeAxesChecked(reduceMeanOp.getAxesAttr(), inputType.getRank());
+    auto axes = normalizeAxesChecked(std::optional<ArrayAttr>(reduceMeanOp.getAxesAttr()), inputType.getRank());
     if (failed(axes))
       return failure();
     SmallVector<bool> reducedAxes = buildReducedAxesMask(*axes, inputType.getRank());

src/PIM/Conversion/ONNXToSpatial/Patterns/NN/Pool.cpp

@@ -31,17 +31,18 @@ static Value materializeContiguousTile(ConversionPatternRewriter& rewriter, Loca
 
 static Value
 createPoolFillElement(ConversionPatternRewriter& rewriter, Location loc, Type elementType, bool useMinimumValue) {
+  Operation* anchorOp = rewriter.getInsertionBlock()->getParentOp();
   if (!useMinimumValue)
-    return arith::ConstantOp::create(rewriter, loc, elementType, rewriter.getZeroAttr(elementType));
+    return getOrCreateConstant(rewriter, anchorOp, rewriter.getZeroAttr(elementType), elementType);
 
   if (auto floatType = dyn_cast<FloatType>(elementType)) {
     auto minValue = llvm::APFloat::getInf(floatType.getFloatSemantics(), /*Negative=*/true);
-    return arith::ConstantOp::create(rewriter, loc, elementType, rewriter.getFloatAttr(floatType, minValue));
+    return getOrCreateConstant(rewriter, anchorOp, rewriter.getFloatAttr(floatType, minValue), elementType);
   }
 
   if (auto integerType = dyn_cast<IntegerType>(elementType)) {
     auto minValue = llvm::APInt::getSignedMinValue(integerType.getWidth());
-    return arith::ConstantOp::create(rewriter, loc, elementType, rewriter.getIntegerAttr(integerType, minValue));
+    return getOrCreateConstant(rewriter, anchorOp, rewriter.getIntegerAttr(integerType, minValue), elementType);
   }
 
   llvm_unreachable("unsupported pool element type");
@@ -148,7 +149,7 @@ static FailureOr<Value> createAverageScaleTensor(ConversionPatternRewriter& rewr
   }
 
   auto scaleAttr = DenseElementsAttr::get(scaleType, scaleValues);
-  return arith::ConstantOp::create(rewriter, loc, scaleType, scaleAttr).getResult();
+  return getOrCreateConstant(rewriter, rewriter.getInsertionBlock()->getParentOp(), scaleAttr, scaleType);
 }
 
 template <typename PoolOp>
@@ -265,13 +266,14 @@ struct PoolToSpatialComputeBase : public OpConversionPattern<PoolOp> {
           createPaddedPoolInput(rewriter, loc, poolOp, xArg, xType, padTop, padLeft, padBottom, padRight);
         Value pooledOutputInit = tensor::EmptyOp::create(rewriter, loc, outType.getShape(), outType.getElementType());
 
-        Value c0 = arith::ConstantIndexOp::create(rewriter, loc, 0);
-        Value c1 = arith::ConstantIndexOp::create(rewriter, loc, 1);
-        Value cOutputPatchCount = arith::ConstantIndexOp::create(rewriter, loc, outputPatchCount);
-        Value cOutputPixelsPerBatch = arith::ConstantIndexOp::create(rewriter, loc, outputHeight * outputWidth);
-        Value cOutputWidth = arith::ConstantIndexOp::create(rewriter, loc, outputWidth);
-        Value cStrideHeight = arith::ConstantIndexOp::create(rewriter, loc, strideHeight);
-        Value cStrideWidth = arith::ConstantIndexOp::create(rewriter, loc, strideWidth);
+        Operation* anchorOp = rewriter.getInsertionBlock()->getParentOp();
+        Value c0 = getOrCreateIndexConstant(rewriter, anchorOp, 0);
+        Value c1 = getOrCreateIndexConstant(rewriter, anchorOp, 1);
+        Value cOutputPatchCount = getOrCreateIndexConstant(rewriter, anchorOp, outputPatchCount);
+        Value cOutputPixelsPerBatch = getOrCreateIndexConstant(rewriter, anchorOp, outputHeight * outputWidth);
+        Value cOutputWidth = getOrCreateIndexConstant(rewriter, anchorOp, outputWidth);
+        Value cStrideHeight = getOrCreateIndexConstant(rewriter, anchorOp, strideHeight);
+        Value cStrideWidth = getOrCreateIndexConstant(rewriter, anchorOp, strideWidth);
 
         auto outputLoop = scf::ForOp::create(rewriter, loc, c0, cOutputPatchCount, c1, ValueRange {pooledOutputInit});
         rewriter.setInsertionPointToStart(outputLoop.getBody());
@@ -296,14 +298,14 @@ struct PoolToSpatialComputeBase : public OpConversionPattern<PoolOp> {
           for (int64_t kernelH = 0; kernelH < kernelHeight; ++kernelH) {
             Value paddedInH = windowBaseH;
             if (kernelH * dilationHeight != 0) {
-              Value kernelHOffset = arith::ConstantIndexOp::create(rewriter, loc, kernelH * dilationHeight);
+              Value kernelHOffset = getOrCreateIndexConstant(rewriter, anchorOp, kernelH * dilationHeight);
               paddedInH = arith::AddIOp::create(rewriter, loc, paddedInH, kernelHOffset);
             }
 
             for (int64_t kernelW = 0; kernelW < kernelWidth; ++kernelW) {
               Value paddedInW = windowBaseW;
               if (kernelW * dilationWidth != 0) {
-                Value kernelWOffset = arith::ConstantIndexOp::create(rewriter, loc, kernelW * dilationWidth);
+                Value kernelWOffset = getOrCreateIndexConstant(rewriter, anchorOp, kernelW * dilationWidth);
                 paddedInW = arith::AddIOp::create(rewriter, loc, paddedInW, kernelWOffset);
               }
 

src/PIM/Conversion/ONNXToSpatial/Patterns/NN/Softmax.cpp

@@ -52,9 +52,10 @@ static Value buildLoopSoftmaxNest(Value input,
   if (axis == inputType.getRank() - 1)
     return buildLoopSoftmaxSlice(input, accumulator, inputType, outerIndices, rewriter, loc);
 
-  Value c0 = arith::ConstantIndexOp::create(rewriter, loc, 0);
-  Value c1 = arith::ConstantIndexOp::create(rewriter, loc, 1);
-  Value cUpper = arith::ConstantIndexOp::create(rewriter, loc, inputType.getDimSize(axis));
+  Operation* anchorOp = rewriter.getInsertionBlock()->getParentOp();
+  Value c0 = getOrCreateIndexConstant(rewriter, anchorOp, 0);
+  Value c1 = getOrCreateIndexConstant(rewriter, anchorOp, 1);
+  Value cUpper = getOrCreateIndexConstant(rewriter, anchorOp, inputType.getDimSize(axis));
 
   auto loop = scf::ForOp::create(rewriter, loc, c0, cUpper, c1, ValueRange {accumulator});
   rewriter.setInsertionPointToStart(loop.getBody());

src/PIM/Conversion/ONNXToSpatial/Patterns/Tensor/Resize.cpp

@@ -17,9 +17,10 @@ namespace {
 
 static Value buildNearestAsymmetricIndex(
   Value outputIndex, int64_t inputDim, int64_t outputDim, ConversionPatternRewriter& rewriter, Location loc) {
-  Value cInputDim = arith::ConstantIndexOp::create(rewriter, loc, inputDim);
-  Value cOutputDim = arith::ConstantIndexOp::create(rewriter, loc, outputDim);
-  Value cInputDimLast = arith::ConstantIndexOp::create(rewriter, loc, inputDim - 1);
+  Operation* anchorOp = rewriter.getInsertionBlock()->getParentOp();
+  Value cInputDim = getOrCreateIndexConstant(rewriter, anchorOp, inputDim);
+  Value cOutputDim = getOrCreateIndexConstant(rewriter, anchorOp, outputDim);
+  Value cInputDimLast = getOrCreateIndexConstant(rewriter, anchorOp, inputDim - 1);
   Value scaledIndex = arith::MulIOp::create(rewriter, loc, outputIndex, cInputDim);
   Value inputIndex = arith::DivUIOp::create(rewriter, loc, scaledIndex, cOutputDim);
   return arith::MinUIOp::create(rewriter, loc, inputIndex, cInputDimLast);
@@ -37,12 +38,13 @@ static Value buildNearestResizeLoop(Value input,
   SmallVector<OpFoldResult> unitSizes(resultType.getRank(), rewriter.getIndexAttr(1));
   SmallVector<OpFoldResult> unitStrides(resultType.getRank(), rewriter.getIndexAttr(1));
 
-  Value c0 = arith::ConstantIndexOp::create(rewriter, loc, 0);
-  Value c1 = arith::ConstantIndexOp::create(rewriter, loc, 1);
-  Value cOutputN = arith::ConstantIndexOp::create(rewriter, loc, resultType.getDimSize(0));
-  Value cOutputC = arith::ConstantIndexOp::create(rewriter, loc, resultType.getDimSize(1));
-  Value cOutputH = arith::ConstantIndexOp::create(rewriter, loc, resultType.getDimSize(2));
-  Value cOutputW = arith::ConstantIndexOp::create(rewriter, loc, resultType.getDimSize(3));
+  Operation* anchorOp = rewriter.getInsertionBlock()->getParentOp();
+  Value c0 = getOrCreateIndexConstant(rewriter, anchorOp, 0);
+  Value c1 = getOrCreateIndexConstant(rewriter, anchorOp, 1);
+  Value cOutputN = getOrCreateIndexConstant(rewriter, anchorOp, resultType.getDimSize(0));
+  Value cOutputC = getOrCreateIndexConstant(rewriter, anchorOp, resultType.getDimSize(1));
+  Value cOutputH = getOrCreateIndexConstant(rewriter, anchorOp, resultType.getDimSize(2));
+  Value cOutputW = getOrCreateIndexConstant(rewriter, anchorOp, resultType.getDimSize(3));
 
   Value outputInit = tensor::EmptyOp::create(rewriter, loc, resultType.getShape(), elemType);
 

src/PIM/Conversion/ONNXToSpatial/Patterns/Tensor/Transpose.cpp

@@ -1,9 +1,11 @@
 #include "mlir/Dialect/Linalg/IR/Linalg.h"
+#include "mlir/Dialect/Arith/IR/Arith.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Transforms/DialectConversion.h"
 
 #include "llvm/ADT/SmallVector.h"
 
+#include "src/Accelerators/PIM/Common/IR/ShapeUtils.hpp"
 #include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
 #include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
@@ -30,6 +32,54 @@ static Value createTransposeInit(Value input,
   return tensor::EmptyOp::create(rewriter, loc, sizes, resultType.getElementType()).getResult();
 }
 
+static FailureOr<Value> materializeTransposedConstant(Value input,
+                                                      RankedTensorType resultType,
+                                                      ArrayRef<int64_t> permutation,
+                                                      ConversionPatternRewriter& rewriter,
+                                                      Location loc) {
+  auto denseAttr = getHostConstDenseElementsAttr(input);
+  if (!denseAttr)
+    return failure();
+
+  auto inputType = dyn_cast<RankedTensorType>(denseAttr.getType());
+  if (!inputType || !inputType.hasStaticShape() || !resultType.hasStaticShape()
+      || inputType.getRank() != resultType.getRank()
+      || static_cast<int64_t>(permutation.size()) != inputType.getRank()) {
+    return failure();
+  }
+
+  if (denseAttr.isSplat())
+    return getOrCreateConstant(rewriter,
+                               rewriter.getInsertionBlock()->getParentOp(),
+                               DenseElementsAttr::get(resultType, denseAttr.getSplatValue<Attribute>()),
+                               resultType);
+
+  SmallVector<Attribute> inputValues(denseAttr.getValues<Attribute>());
+  SmallVector<Attribute> resultValues(inputValues.size());
+  SmallVector<int64_t> inputStrides = computeRowMajorStrides(inputType.getShape());
+  SmallVector<int64_t> resultStrides = computeRowMajorStrides(resultType.getShape());
+  SmallVector<int64_t> inputIndices(inputType.getRank(), 0);
+
+  for (auto [linearIndex, value] : llvm::enumerate(inputValues)) {
+    int64_t remaining = static_cast<int64_t>(linearIndex);
+    for (int64_t dim = 0; dim < inputType.getRank(); ++dim) {
+      inputIndices[dim] = inputStrides.empty() ? 0 : remaining / inputStrides[dim];
+      remaining = inputStrides.empty() ? 0 : remaining % inputStrides[dim];
+    }
+
+    int64_t resultLinearIndex = 0;
+    for (int64_t dim = 0; dim < resultType.getRank(); ++dim)
+      resultLinearIndex += inputIndices[permutation[dim]] * resultStrides[dim];
+
+    resultValues[resultLinearIndex] = value;
+  }
+
+  return getOrCreateConstant(rewriter,
+                             rewriter.getInsertionBlock()->getParentOp(),
+                             DenseElementsAttr::get(resultType, resultValues),
+                             resultType);
+}
+
 struct TransposeToLinalgTranspose : OpConversionPattern<ONNXTransposeOp> {
   using OpConversionPattern::OpConversionPattern;
 
@@ -44,6 +94,14 @@ struct TransposeToLinalgTranspose : OpConversionPattern<ONNXTransposeOp> {
     auto permutation = getTransposePermutationChecked(transposeOp.getPermAttr(), inputType.getRank());
     if (failed(permutation))
       return failure();
+    if (isCompileTimeComputable(adaptor.getData())) {
+      auto constantTranspose =
+        materializeTransposedConstant(adaptor.getData(), resultType, *permutation, rewriter, transposeOp.getLoc());
+      if (succeeded(constantTranspose)) {
+        rewriter.replaceOp(transposeOp, *constantTranspose);
+        return success();
+      }
+    }
     Value init = createTransposeInit(adaptor.getData(), resultType, *permutation, rewriter, transposeOp.getLoc());
     Value transposed =
       linalg::TransposeOp::create(rewriter, transposeOp.getLoc(), adaptor.getData(), init, *permutation)