conv_cudnn_op.cc 1.8 KB
Newer Older
武毅 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/* 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. */

C
chengduoZH 已提交
15
#include "paddle/operators/conv_op.h"
武毅 已提交
16 17 18 19 20 21 22 23 24 25 26 27 28 29

namespace paddle {
namespace operators {

class CudnnConvOpMaker : public Conv2DOpMaker {
 public:
  CudnnConvOpMaker(framework::OpProto* proto,
                   framework::OpAttrChecker* op_checker)
      : Conv2DOpMaker(proto, op_checker) {
    AddAttr<int>("workspace_size_MB",
                 "workspace size for cudnn, in MB, "
                 "workspace is a section of GPU memory which will be "
                 "allocated/freed each time the operator runs, larger "
                 "workspace size can increase performance but also requires "
K
Kexin Zhao 已提交
30
                 "better hardware. This size should be chosen carefully.")
武毅 已提交
31 32 33 34 35 36 37 38
        .SetDefault(4096);
  }
};

}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;
C
chengduoZH 已提交
39 40 41
REGISTER_OP(conv_cudnn, ops::ConvOp, ops::CudnnConvOpMaker, conv_cudnn_grad,
            ops::ConvOpGrad);

C
chengduoZH 已提交
42
REGISTER_OP_CPU_KERNEL(conv_cudnn,
C
chengduoZH 已提交
43 44
                       ops::GemmConvKernel<paddle::platform::CPUPlace, float>,
                       ops::GemmConvKernel<paddle::platform::CPUPlace, double>);
武毅 已提交
45
REGISTER_OP_CPU_KERNEL(
C
chengduoZH 已提交
46 47
    conv_cudnn_grad, ops::GemmConvGradKernel<paddle::platform::CPUPlace, float>,
    ops::GemmConvGradKernel<paddle::platform::CPUPlace, double>);