MNN/source/backend/opencl/execution/buffer/DepthwiseConvSubgroupBufExe...

51 lines
1.7 KiB
C++

//
// DepthwiseConvSubgroupBufExecution.hpp
// MNN
//
// Created by MNN on 2023/07/01.
// Copyright © 2018, Alibaba Group Holding Limited
//
#ifndef MNN_OPENCL_BUFFER_CLOSED
#ifdef MNN_SUPPORT_INTEL_SUBGROUP
#ifndef DepthwiseConvSubgroupBufExecution_hpp
#define DepthwiseConvSubgroupBufExecution_hpp
#include "ConvBufExecution.hpp"
namespace MNN {
namespace OpenCL {
class DepthwiseConvSubgroupBufExecution : public ConvBufCommonExecution {
public:
DepthwiseConvSubgroupBufExecution(const std::vector<Tensor *> &inputs, const MNN::Op *op, Backend *backend);
virtual ~DepthwiseConvSubgroupBufExecution();
virtual ErrorCode onResize(const std::vector<Tensor *> &inputs, const std::vector<Tensor *> &outputs) override;
virtual ErrorCode onExecute(const std::vector<Tensor *> &inputs, const std::vector<Tensor *> &outputs) override;
private:
void transformWeight(const Tensor *weightDest, const Tensor *source);
const Convolution2DCommon *mConv2dCommonParams;
const Convolution2D *mCon2dParams;
std::vector<int> mStrides{1, 1};
std::vector<int> mPaddings{0, 0};
std::vector<int> mDilations{1, 1};
std::shared_ptr<Tensor> mFilter;
std::shared_ptr<Tensor> mSource;
cl::Kernel mTranseKernel;
cl::Kernel mKernel;
uint32_t mMaxWorkGroupSize;
std::vector<uint32_t> mGlobalWorkSize{1, 1, 1};
std::vector<uint32_t> mLocalWorkSize{1, 1, 1, 1};
std::vector<uint32_t> mTranseGlobalWorkSize{1, 1, 1};
std::vector<uint32_t> mTranseLocalWorkSize{1, 1, 1, 1};
bool mNeedTranse = false;
std::set<std::string> mBuildOptions;
};
} // namespace OpenCL
} // namespace MNN
#endif /* DepthwiseConvSubgroupBufExecution_hpp */
#endif /* MNN_SUPPORT_INTEL_SUBGROUP */
#endif /* MNN_OPENCL_BUFFER_CLOSED */