You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
86 lines
2.3 KiB
86 lines
2.3 KiB
/* 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
|
|
#include "paddle/platform/device_context.h"
|
|
|
|
namespace paddle {
|
|
namespace platform {
|
|
|
|
template <typename DeviceContext>
|
|
struct ForRange {
|
|
ForRange(const DeviceContext& dev_ctx, size_t limit);
|
|
|
|
template <typename Function>
|
|
void operator()(Function func) const;
|
|
};
|
|
|
|
template <>
|
|
struct ForRange<CPUDeviceContext> {
|
|
ForRange(const CPUDeviceContext& dev_ctx, size_t limit) : limit_(limit) {}
|
|
|
|
template <typename Function>
|
|
void operator()(Function func) const {
|
|
for (size_t i = 0; i < limit_; ++i) {
|
|
func(i);
|
|
}
|
|
}
|
|
|
|
size_t limit_;
|
|
};
|
|
|
|
#ifdef __NVCC__
|
|
template <typename Function>
|
|
__global__ static void ForRangeElemwiseOpGridIsOne(Function func) {
|
|
size_t idx = static_cast<size_t>(threadIdx.x);
|
|
func(idx);
|
|
}
|
|
|
|
template <typename Function>
|
|
__global__ static void ForRangeElemwiseOp(Function func, int limit) {
|
|
size_t idx = static_cast<size_t>(blockIdx.x * blockDim.x + threadIdx.x);
|
|
if (idx < limit) {
|
|
func(idx);
|
|
}
|
|
}
|
|
|
|
template <>
|
|
struct ForRange<CUDADeviceContext> {
|
|
ForRange(const CUDADeviceContext& dev_ctx, size_t limit)
|
|
: dev_ctx_(dev_ctx), limit_(static_cast<int>(limit)) {}
|
|
|
|
template <typename Function>
|
|
inline void operator()(Function func) const {
|
|
constexpr int num_threads = 1024;
|
|
int block_size = limit_ <= num_threads ? limit_ : num_threads;
|
|
int grid_size = (limit_ + num_threads - 1) / num_threads;
|
|
|
|
if (grid_size == 1) {
|
|
ForRangeElemwiseOpGridIsOne<<<1, block_size, 0, dev_ctx_.stream()>>>(
|
|
func);
|
|
} else {
|
|
ForRangeElemwiseOp<<<grid_size, block_size, 0, dev_ctx_.stream()>>>(
|
|
func, limit_);
|
|
}
|
|
}
|
|
|
|
const CUDADeviceContext& dev_ctx_;
|
|
int limit_;
|
|
};
|
|
|
|
#endif
|
|
|
|
} // namespace platform
|
|
} // namespace paddle
|