Automerge: [mlir][NFC] Avoid using braced initializer lists to call a constructor. (#123714)

In the LLVM style guide, we prefer not using braced initializer lists to
call a constructor. Also, we prefer using an equal before the open curly
brace if we use a braced initializer list when initializing a variable.

See

https://llvm.org/docs/CodingStandards.html#do-not-use-braced-initializer-lists-to-call-a-constructor
for more details.

The style guide does not explain the reason well. There is an article
from abseil, which mentions few benefits. E.g., we can avoid the most
vexing parse, etc. See https://abseil.io/tips/88 for more details.

Signed-off-by: hanhanW <[email protected]>

Author: hanhanW

mlir/lib/Bindings/Python/IRAttributes.cpp

@@ -845,7 +845,7 @@ class PyDenseElementsAttribute
       }
       shapedType = *explicitType;
     } else {
-      SmallVector<int64_t> shape{static_cast<int64_t>(numAttributes)};
+      SmallVector<int64_t> shape = {static_cast<int64_t>(numAttributes)};
       shapedType = mlirRankedTensorTypeGet(
           shape.size(), shape.data(),
           mlirAttributeGetType(pyTryCast<PyAttribute>(attributes[0])),

mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp

@@ -1476,7 +1476,7 @@ class MaterializeResizeBroadcast : public OpRewritePattern<tosa::ResizeOp> {
       reassociationMap.push_back({});
     reassociationMap.back().push_back(builder.getAffineDimExpr(3));
 
-    llvm::SmallVector<int64_t> collapseShape{batch};
+    llvm::SmallVector<int64_t> collapseShape = {batch};
     if (inputH != 1)
       collapseShape.push_back(outputH);
     if (inputW != 1)

mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp

@@ -648,12 +648,12 @@ class FullyConnectedConverter
 
     SmallVector<Value> filteredDims = condenseValues(dynDims);
 
-    SmallVector<int64_t> permutation{1, 0};
+    SmallVector<int64_t> permutation = {1, 0};
     auto permutationAttr = rewriter.getI64TensorAttr(permutation);
     Value permutationValue =
         rewriter.create<arith::ConstantOp>(loc, permutationAttr);
 
-    SmallVector<int64_t> newWeightShape{weightShape[1], weightShape[0]};
+    SmallVector<int64_t> newWeightShape = {weightShape[1], weightShape[0]};
     Type newWeightTy =
         RankedTensorType::get(newWeightShape, weightTy.getElementType());
 

mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp

@@ -182,7 +182,7 @@ struct TransferReadLowering : public OpRewritePattern<vector::TransferReadOp> {
           readOp, "Unsupported data type for tranposition");
 
     // If load is transposed, get the base shape for the tensor descriptor.
-    SmallVector<int64_t> descShape{vecTy.getShape()};
+    SmallVector<int64_t> descShape(vecTy.getShape());
     if (isTransposeLoad)
       std::reverse(descShape.begin(), descShape.end());
     auto descType = xegpu::TensorDescType::get(

mlir/lib/Dialect/ArmNeon/Transforms/LowerContractionToSMMLAPattern.cpp

@@ -126,8 +126,8 @@ class LowerContractionToSMMLAPattern
         loc, op.getResultType(), rewriter.getZeroAttr(op.getResultType()));
 
     SmallVector<int64_t> unrolledSize = *op.getShapeForUnroll();
-    SmallVector<int64_t> smmlaShape{2, 8};
-    SmallVector<int64_t> loopOrder{0, 1};
+    SmallVector<int64_t> smmlaShape = {2, 8};
+    SmallVector<int64_t> loopOrder = {0, 1};
     if (unrolledSize.size() == 3) {
       smmlaShape.insert(smmlaShape.begin(), isVecmat ? 1 : 2);
       loopOrder.push_back(2);

mlir/lib/Dialect/Linalg/TransformOps/GPUHeuristics.cpp

@@ -222,7 +222,7 @@ transform::gpu::CopyMappingInfo::inferNumThreadsImpl(
   // Scale the most minor size to account for the chosen vector size and
   // maximize the number of threads without exceeding the total number of
   // threads.
-  SmallVector<int64_t> scaledSizes{sizes};
+  SmallVector<int64_t> scaledSizes(sizes);
   scaledSizes.back() /= desiredVectorSize;
   if (scaledSizes.back() > totalNumThreads) {
     LDBG("--Too few threads given the required vector size -> FAIL");

mlir/lib/Dialect/Linalg/Transforms/BlockPackMatmul.cpp

@@ -55,7 +55,7 @@ static bool validateFullTilesOnDims(linalg::LinalgOp linalgOp,
 
   // Skip the batch dimension if present.
   // Offset all dimensions accordingly.
-  SmallVector<int64_t, 3> offsetDims{dims};
+  SmallVector<int64_t, 3> offsetDims(dims);
   for (size_t i = 0; i < offsetDims.size(); i++)
     offsetDims[i] += batchDimsOffset;
 
@@ -111,10 +111,10 @@ transposePackedMatmul(RewriterBase &rewriter, linalg::LinalgOp linalgOp,
 
   // Transpose only the dimensions that need that to conform to the provided
   // transpotion settings.
-  SmallVector<int64_t> innerPerm{0, 1};
+  SmallVector<int64_t> innerPerm = {0, 1};
   if (isInnerTransposed != transposeInnerBlocks)
     innerPerm = {1, 0};
-  SmallVector<int64_t> outerPerm{0, 1};
+  SmallVector<int64_t> outerPerm = {0, 1};
   if (isOuterTransposed != transposeOuterBlocks)
     outerPerm = {1, 0};
 

mlir/lib/Dialect/Linalg/Transforms/TransposeConv2D.cpp

@@ -52,7 +52,7 @@ FailureOr<Operation *> transposeConv2DHelper(RewriterBase &rewriter,
                                              FHWCConvOp op) {
   // Construct a permutation of the filter tensor dimensions. For a 2D
   // convolution this will be known statically as [1, 2, 3, 0].
-  SmallVector<int64_t> filterPerm({1, 2, 3, 0});
+  SmallVector<int64_t> filterPerm = {1, 2, 3, 0};
 
   // Create the type for the transposed filter tensor.
   auto filter = op->getOperand(1);

mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp

@@ -86,8 +86,8 @@ extractConvInputSlices(RewriterBase &rewriter, Location loc, Value input,
   if (isSingleChanneled) {
     // Extract input slice of size {wSizeStep} @ [w + kw] for non-channeled
     // convolution.
-    SmallVector<int64_t> sizes{wSizeStep};
-    SmallVector<int64_t> strides{1};
+    SmallVector<int64_t> sizes = {wSizeStep};
+    SmallVector<int64_t> strides = {1};
     for (int64_t kw = 0; kw < kwSize; ++kw) {
       for (int64_t w = 0; w < wSize; w += wSizeStep) {
         result.push_back(rewriter.create<vector::ExtractStridedSliceOp>(
@@ -97,8 +97,8 @@ extractConvInputSlices(RewriterBase &rewriter, Location loc, Value input,
   } else {
     // Extract lhs slice of size {n, wSizeStep, c} @ [0, sw * w + dw * kw, 0]
     // for channeled convolution.
-    SmallVector<int64_t> sizes{nSize, wSizeStep, cSize};
-    SmallVector<int64_t> strides{1, 1, 1};
+    SmallVector<int64_t> sizes = {nSize, wSizeStep, cSize};
+    SmallVector<int64_t> strides = {1, 1, 1};
     for (int64_t kw = 0; kw < kwSize; ++kw) {
       for (int64_t w = 0; w < wSize; w += wSizeStep) {
         result.push_back(rewriter.create<vector::ExtractStridedSliceOp>(
@@ -135,17 +135,17 @@ extractConvResultSlices(RewriterBase &rewriter, Location loc, Value res,
   SmallVector<Value> result;
   if (isSingleChanneled) {
     // Extract res slice: {wSizeStep} @ [w] for non-channeled convolution.
-    SmallVector<int64_t> sizes{wSizeStep};
-    SmallVector<int64_t> strides{1};
+    SmallVector<int64_t> sizes = {wSizeStep};
+    SmallVector<int64_t> strides = {1};
     for (int64_t w = 0; w < wSize; w += wSizeStep) {
       result.push_back(rewriter.create<vector::ExtractStridedSliceOp>(
           loc, res, /*offsets=*/ArrayRef<int64_t>{w}, sizes, strides));
     }
   } else {
     // Extract res slice: {n, wSizeStep, f} @ [0, w, 0] for channeled
     // convolution.
-    SmallVector<int64_t> sizes{nSize, wSizeStep, fSize};
-    SmallVector<int64_t> strides{1, 1, 1};
+    SmallVector<int64_t> sizes = {nSize, wSizeStep, fSize};
+    SmallVector<int64_t> strides = {1, 1, 1};
     for (int64_t w = 0; w < wSize; w += wSizeStep) {
       result.push_back(rewriter.create<vector::ExtractStridedSliceOp>(
           loc, res, /*offsets=*/ArrayRef<int64_t>{0, w, 0}, sizes, strides));
@@ -163,15 +163,15 @@ static Value insertConvResultSlices(RewriterBase &rewriter, Location loc,
   if (isSingleChanneled) {
     // Write back res slice: {wSizeStep} @ [w] for non-channeled convolution.
     // This does not depend on kw.
-    SmallVector<int64_t> strides{1};
+    SmallVector<int64_t> strides = {1};
     for (int64_t w = 0; w < wSize; w += wSizeStep) {
       res = rewriter.create<vector::InsertStridedSliceOp>(
           loc, resVals[w], res, /*offsets=*/ArrayRef<int64_t>{w}, strides);
     }
   } else {
     // Write back res slice: {n, wSizeStep, f} @ [0, w, 0] for channeled
     // convolution. This does not depend on kw.
-    SmallVector<int64_t> strides{1, 1, 1};
+    SmallVector<int64_t> strides = {1, 1, 1};
     for (int64_t w = 0; w < wSize; w += wSizeStep) {
       res = rewriter.create<vector::InsertStridedSliceOp>(
           loc, resVals[w], res, /*offsets=*/ArrayRef<int64_t>{0, w, 0},
@@ -3505,8 +3505,8 @@ struct Conv1DGenerator
     //===------------------------------------------------------------------===//
     // Unroll along kw and read slices of lhs and rhs.
     SmallVector<Value> lhsVals, rhsVals, resVals;
-    auto inOutSliceSizes = SmallVector<int64_t>{nSize, wSizeStep, cSize};
-    auto inOutStrides = SmallVector<int64_t>{1, 1, 1};
+    SmallVector<int64_t> inOutSliceSizes = {nSize, wSizeStep, cSize};
+    SmallVector<int64_t> inOutStrides = {1, 1, 1};
 
     // Extract lhs slice of size {n, wSizeStep, c}
     //   @ [0, sw * w + dw * kw, 0].
@@ -3538,8 +3538,7 @@ struct Conv1DGenerator
 
     // Note - the scalable flags are ignored as flattening combined with
     // scalable vectorization is not supported.
-    auto inOutFlattenSliceSizes =
-        SmallVector<int64_t>{nSize, wSizeStep * cSize};
+    SmallVector<int64_t> inOutFlattenSliceSizes = {nSize, wSizeStep * cSize};
     auto lhsTypeAfterFlattening =
         VectorType::get(inOutFlattenSliceSizes, lhsEltType);
     auto resTypeAfterFlattening =

mlir/lib/Dialect/NVGPU/TransformOps/NVGPUTransformOps.cpp

@@ -740,9 +740,9 @@ static std::tuple<SmallVector<int64_t>, SmallVector<int64_t>,
                   SmallVector<int64_t>>
 makeVectorShapes(ArrayRef<int64_t> lhs, ArrayRef<int64_t> rhs,
                  ArrayRef<int64_t> res) {
-  SmallVector<int64_t> vlhs{lhs};
-  SmallVector<int64_t> vrhs{rhs};
-  SmallVector<int64_t> vres{res};
+  SmallVector<int64_t> vlhs(lhs);
+  SmallVector<int64_t> vrhs(rhs);
+  SmallVector<int64_t> vres(res);
   return std::make_tuple(vlhs, vrhs, vres);
 }