pytorch/c10/util/qint32.h
Igor Sugak f1e89fbe53 [pytorch] add missing host-device attribute to fix clang build (#37358)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/37358

Test Plan:
```lang=bash
buck build mode/opt -c fbcode.cuda_use_clang=true //vision/fair/detectron2/tools:benchmark
```

Reviewed By: ngimel

Differential Revision: D21262235

fbshipit-source-id: 00633352d87da0881b2cc90759265fa0d0bd96be
2020-04-27 18:24:20 -07:00

19 lines
319 B
C++

#pragma once
#include <cstdint>
#include <c10/macros/Macros.h>
namespace c10 {
/**
* qint32 is for signed 32 bit quantized Tensors
*/
struct alignas(4) qint32 {
using underlying = int32_t;
int32_t val_;
qint32() = default;
C10_HOST_DEVICE explicit qint32(int32_t val) : val_(val) {}
};
} // namespace c10