Update c style cast to c++ style cast

src/Accelerators/NNPA/Conversion/ONNXToZHigh/ONNXLegalityCheck.cpp

@@ -315,7 +315,7 @@ bool meetPoolParamRestrictions(Operation *op, int64_t inputShape,
       return onnxToZHighUnsupportedReport(op, message);
     }
     if (paddingType == "SAME_PADDING") {
-      int64_t reqOutputShape = ceil((float)inputShape / strides);
+      int64_t reqOutputShape = ceil(static_cast<float>(inputShape) / strides);
       if (outputShape != reqOutputShape) {
         std::string message =
             "When the strides (" + std::to_string(strides) +
@@ -329,7 +329,7 @@ bool meetPoolParamRestrictions(Operation *op, int64_t inputShape,
       }
     } else { // VALID_PADDING
       int64_t reqOutputShape =
-          ceil((float)(inputShape - kernelShape + 1) / strides);
+          ceil(static_cast<float>(inputShape - kernelShape + 1) / strides);
       if (outputShape != reqOutputShape) {
         std::string message = "When the strides (" + std::to_string(strides) +
                               ") and the padding type is VALID_PADDING, output "
@@ -1164,7 +1164,7 @@ static bool checkConv2DParamRestrictions(Operation *op, int64_t inputDim,
     }
     if (paddingType == "SAME_PADDING") {
       // height_out restriction.
-      int64_t reqOutputShape = ceil((float)inputDim / stride);
+      int64_t reqOutputShape = ceil(static_cast<float>(inputDim) / stride);
       if (outputDim != reqOutputShape) {
         std::string message =
             "When the strides (" + std::to_string(stride) +
@@ -1189,7 +1189,8 @@ static bool checkConv2DParamRestrictions(Operation *op, int64_t inputDim,
         return onnxToZHighUnsupportedReport(op, message);
       }
       // height_out restriction.
-      int64_t reqOutputShape = ceil((float)(inputDim - kernelDim + 1) / stride);
+      int64_t reqOutputShape =
+          ceil(static_cast<float>(inputDim - kernelDim + 1) / stride);
       if (outputDim != reqOutputShape) {
         std::string message =
             "When the strides (" + std::to_string(stride) +

src/Accelerators/NNPA/Conversion/ONNXToZHigh/RewriteONNXForZHigh.cpp

@@ -240,7 +240,7 @@ ArrayAttr getPadsForNNPAConv(PatternRewriter &rewriter, Value ret) {
 // This function is used for padding attribute in Conv.
 DenseElementsAttr insertZerosForNonPaddedDims(
     PatternRewriter &rewriter, ArrayAttr origAttrs, int extensionLength) {
-  int nDims = (int)origAttrs.getValue().size() / 2;
+  int nDims = static_cast<int>(origAttrs.getValue().size()) / 2;
   int nElements = (nDims + extensionLength) * 2;
   SmallVector<int64_t, 4> pads(nElements, 0);
   for (int i = 0; i < nDims; ++i) {

src/Accelerators/NNPA/Conversion/ZHighToZLow/ZHighToZLow.cpp

@@ -199,7 +199,7 @@ Value insertAllocOrEmitZeroConstant(ArrayRef<IndexExpr> dims,
     // Attribute type: tensor<sizeInBytes x i8>
     int64_t sizeInBytes =
         affine::getIntOrFloatMemRefSizeInBytes(resType).value();
-    char *rawData = (char *)malloc(sizeInBytes);
+    char *rawData = static_cast<char *>(malloc(sizeInBytes));
     assert(rawData && "failed to allocate memory for stickified data");
     memset(rawData, 0, sizeInBytes);
     DenseResourceElementsAttr valueAttr = DenseUI8ResourceElementsAttr::get(
@@ -228,7 +228,7 @@ Value insertShapeMemRefI64(
   MultiDialectBuilder<KrnlBuilder, MathBuilder, MemRefBuilder> create(
       rewriter, loc);
   MemRefType shapeMemRefType = MemRefType::get(
-      {(int64_t)originalDims.size()}, rewriter.getIntegerType(64));
+      {static_cast<int64_t>(originalDims.size())}, rewriter.getIntegerType(64));
   Value shapeMemRef = create.mem.alignedAlloc(shapeMemRefType);
   for (uint64_t i = 0; i < originalDims.size(); ++i) {
     Value dim =
@@ -395,7 +395,7 @@ ZMemRefType convertZTensorToMemRefType(Type type) {
                "wrong concatenated dimension size");
         int64_t s = shape[rank - 1] / 4;
         // ((s + 64 - 1) / 64) * 64;
-        int64_t s_pad = ceil((double)s / 64) * 64;
+        int64_t s_pad = ceil(static_cast<double>(s) / 64) * 64;
         int64_t pad_size = s_pad - s;
         AffineExpr constExprS = getAffineConstantExpr(s, b.getContext());
         if (rank == 2) {
@@ -431,7 +431,8 @@ ZMemRefType convertZTensorToMemRefType(Type type) {
                "in affine_map generation.");
         assert((hidden_size % 3) == 0 && "wrong concatenated dimension size.");
         int64_t s = hidden_size / 3;
-        int64_t s_pad = ceil((float)s / 64) * 64; // ((s + 64 - 1) / 64) * 64;
+        int64_t s_pad =
+            ceil(static_cast<float>(s) / 64) * 64; // ((s + 64 - 1) / 64) * 64;
         int64_t pad_size = s_pad - s;
         AffineExpr constExprS = getAffineConstantExpr(s, b.getContext());
         if (rank == 2) {
@@ -723,7 +724,7 @@ struct ZHighToZLowStickifiedConstantOpLowering : public ConversionPattern {
     // Validate the stickified tensor.
     int64_t memRefSizeInBytes = getMemRefEltSizeInBytes(normalizedType);
     memRefSizeInBytes *= normalizedType.getNumElements();
-    assert((data.size() == (uint64_t)memRefSizeInBytes) &&
+    assert((data.size() == static_cast<uint64_t>(memRefSizeInBytes)) &&
            "The stickified tensor's buffer size and MemRef's size mismatched");
 
     // Create a KrnlGlobalOp.
@@ -1565,7 +1566,7 @@ struct ZHighToZLowStickifiedConstantOfShapeOpLowering
 
     // Convert the scalar value to dlfloat16.
     // Use uint16_t as container.
-    float valueF32 = (float)value.getValueAsDouble();
+    float valueF32 = static_cast<float>(value.getValueAsDouble());
     uint16_t valueDLF16;
     fp32_to_dlf16(&valueF32, &valueDLF16, 1);
 
@@ -1709,10 +1710,10 @@ struct ZHighToZLowDataConversionLowering
 
     if (enableParallel) {
       int64_t parId;
-      int64_t tripCount =
-          flattenedOutputDims[0].isLiteral()
-              ? std::ceil(flattenedOutputDims[0].getLiteral() / (float)archVL)
-              : -1;
+      int64_t tripCount = flattenedOutputDims[0].isLiteral()
+                              ? std::ceil(flattenedOutputDims[0].getLiteral() /
+                                          static_cast<float>(archVL))
+                              : -1;
       if (findSuitableParallelDimension(lbs, flattenedOutputDims, 0, 1, parId,
               /*min iter for going parallel*/ 1024)) {
         create.krnl.parallel(blockedLoopDef[0]);