strided_memcpy.h 1.7 KB
Newer Older
Y
Yu Yang 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.

   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
16
#include "paddle/operators/detail/strided_memcpy.h"
Y
Yu Yang 已提交
17 18 19 20

namespace paddle {
namespace operators {

21
// Strided memory copy from src to dst.
Y
Yu Yang 已提交
22
//
23 24 25 26
// The src and dst should be both on dev_ctx.GetPlace(), otherwise, there will
// be a segment fault.
//
// The stride of an array (also referred to as increment, pitch or step size) is
Y
Yu Yang 已提交
27 28 29 30 31 32 33 34 35
// the number of locations in memory between beginnings of successive array
// elements
//
// For example, for tensor like [1, 3, 300, 300]. If there is no padding, the
// stride is [270000, 90000, 300, 1].
//
// NOTE: When use GPU, the memcpy is async. To sync memcpy, please invoke
// `dev_ctx.Wait()`.
template <typename T>
36 37 38 39
inline void StridedMemcpy(const platform::DeviceContext& dev_ctx, const T* src,
                          const framework::DDim& src_stride,
                          const framework::DDim& dst_dim,
                          const framework::DDim& dst_stride, T* dst) {
Y
Yu Yang 已提交
40
  using namespace detail;
41
  StridedCopyDimVisitor<T> func(dev_ctx, src, src_stride, dst_stride, dst);
Y
Yu Yang 已提交
42 43 44 45
  boost::apply_visitor(func, dst_dim);
}
}  // namespace operators
}  // namespace paddle