math.h 3.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
/* Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */

#pragma once

// CUDA and HIP use same api
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)

20
#include "paddle/pten/backends/cuda/cuda_context.h"
C
Chen Weihang 已提交
21
#include "paddle/pten/common/scalar.h"
22 23 24 25 26 27 28 29
#include "paddle/pten/core/dense_tensor.h"

namespace pten {

template <typename T>
void Sign(const CUDAContext& dev_ctx, const DenseTensor& x, DenseTensor* out);

template <typename T>
30 31 32 33 34 35 36 37
void Mean(const CUDAContext& dev_ctx,
          const DenseTensor& x,
          const std::vector<int64_t>& dims,
          bool keep_dim,
          bool reduce_all,
          DataType in_dtype,
          DataType out_dtype,
          DenseTensor* out);
38 39 40 41

template <typename T>
void Scale(const CUDAContext& dev_ctx,
           const DenseTensor& x,
C
Chen Weihang 已提交
42
           const Scalar& scale,
43 44 45 46
           float bias,
           bool bias_after_scale,
           DenseTensor* out);

47
template <typename T>
48 49 50 51 52
void Add(const CUDAContext& dev_ctx,
         const DenseTensor& x,
         const DenseTensor& y,
         int axis,
         DenseTensor* out);
53

54
template <typename T>
55 56 57 58 59
void Subtract(const CUDAContext& dev_ctx,
              const DenseTensor& x,
              const DenseTensor& y,
              int axis,
              DenseTensor* out);
60

61
template <typename T>
62 63 64 65 66
void Divide(const CUDAContext& dev_ctx,
            const DenseTensor& x,
            const DenseTensor& y,
            int axis,
            DenseTensor* out);
67

Y
YuanRisheng 已提交
68
template <typename T>
69 70 71 72 73
void Multiply(const CUDAContext& dev_ctx,
              const DenseTensor& x,
              const DenseTensor& y,
              int axis,
              DenseTensor* out);
74 75 76 77 78 79 80 81 82 83 84

template <typename T>
void Sum(const CUDAContext& dev_ctx,
         const DenseTensor& x,
         const std::vector<int64_t>& dims,
         bool keep_dim,
         bool reduce_all,
         DataType in_dtype,
         DataType out_dtype,
         DenseTensor* out);

85 86
}  // namespace pten

Y
YuanRisheng 已提交
87 88
#define DEFINE_CUDA_ELEMENTWISE_OP(name)                               \
  template <typename T>                                                \
89 90 91 92 93
  void name(const CUDAContext& dev_ctx,                                \
            const DenseTensor& x,                                      \
            const DenseTensor& y,                                      \
            int axis,                                                  \
            DenseTensor* out) {                                        \
Y
YuanRisheng 已提交
94 95 96 97 98 99 100 101 102 103
    std::vector<const DenseTensor*> inputs;                            \
    std::vector<DenseTensor*> outputs;                                 \
    inputs.emplace_back(&x);                                           \
    inputs.emplace_back(&y);                                           \
    outputs.emplace_back(out);                                         \
    out->mutable_data<T>();                                            \
    LaunchElementwiseCudaKernel<ElementwiseType::kBinary, T, T>(       \
        dev_ctx, inputs, &outputs, axis, general::name##Functor<T>()); \
  }

104
#endif