diff options
Diffstat (limited to 'mlir/lib/Conversion/XeGPUToXeVM/XeGPUToXeVM.cpp')
| -rw-r--r-- | mlir/lib/Conversion/XeGPUToXeVM/XeGPUToXeVM.cpp | 5 |
1 files changed, 2 insertions, 3 deletions
diff --git a/mlir/lib/Conversion/XeGPUToXeVM/XeGPUToXeVM.cpp b/mlir/lib/Conversion/XeGPUToXeVM/XeGPUToXeVM.cpp index 9ead1d89069d..71687b1479a7 100644 --- a/mlir/lib/Conversion/XeGPUToXeVM/XeGPUToXeVM.cpp +++ b/mlir/lib/Conversion/XeGPUToXeVM/XeGPUToXeVM.cpp @@ -23,6 +23,7 @@ #include "mlir/Dialect/XeGPU/IR/XeGPU.h" #include "mlir/Pass/Pass.h" #include "mlir/Support/LLVM.h" +#include "llvm/ADT/STLExtras.h" #include "llvm/Support/FormatVariadic.h" #include "mlir/IR/BuiltinTypes.h" @@ -774,9 +775,7 @@ struct ConvertXeGPUToXeVMPass if (rank < 1 || type.getNumElements() == 1) return elemType; // Otherwise, convert the vector to a flat vector type. - int64_t sum = - std::accumulate(type.getShape().begin(), type.getShape().end(), - int64_t{1}, std::multiplies<int64_t>()); + int64_t sum = llvm::product_of(type.getShape()); return VectorType::get(sum, elemType); }); typeConverter.addConversion([&](xegpu::TensorDescType type) -> Type { |