Line data    Source code

       1              : /**
       2              :  * Copyright (c) 2025 Huawei Technologies Co., Ltd.
       3              :  * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
       4              :  * CANN Open Software License Agreement Version 2.0 (the "License").
       5              :  * Please refer to the License for details. You may not use this file except in compliance with the License.
       6              :  * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
       7              :  * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
       8              :  * See LICENSE in the root of the software repository for the full text of the License.
       9              :  */
      10              : 
      11              : #ifndef OPERATOR_KERNEL_POST_PROCESS_DYNAMIC_OUTPUT_H
      12              : #define OPERATOR_KERNEL_POST_PROCESS_DYNAMIC_OUTPUT_H
      13              : 
      14              : #include <vector>
      15              : #include "operator_kernel.h"
      16              : 
      17              : namespace AicpuSchedule {
      18              : class OperatorKernelPostProcessDynamicOutput : public OperatorKernel {
      19              : public:
      20           16 :     OperatorKernelPostProcessDynamicOutput() = default;
      21           22 :     ~OperatorKernelPostProcessDynamicOutput() = default;
      22              :     int32_t Compute(const AicpuTaskInfo& kernelTaskInfo, const RunContext& taskContext) override;
      23              : 
      24              : private:
      25              :     int32_t CopyTensorDesc(const PostprocessDynamicOutputKernelArgs* const param, const RunContext& taskContext) const;
      26              :     int32_t AllocatedOutput(
      27              :         const PostprocessDynamicOutputKernelArgs* const param, Mbuf* const outputMbuf, const uint32_t dynamicFlag,
      28              :         const size_t index, RuntimeTensorDesc** dynamicSrcDescPptr, RuntimeTensorDesc** staticSrcDescPptr) const;
      29              :     int32_t GetRuntimeTensor(
      30              :         const PostprocessDynamicOutputKernelArgs* const param, RuntimeTensorDesc** dynamicSrcDescPtr) const;
      31              :     int32_t PostprocessSetDataLen(
      32              :         const RuntimeTensorDesc* const runtimeTensorDesc, Mbuf* const outputMbuf, const size_t index) const;
      33              :     int32_t GetMbufHeadFromResp(
      34              :         Mbuf** const respMbufPtr, const RunContext& taskContext, void** const customBufPtr,
      35              :         uint32_t* const customBufSizePtr) const;
      36              :     int32_t PostProcessForOutputToAllocate(
      37              :         const PostprocessDynamicOutputKernelArgs* const param, Mbuf** const outputMbufPtr, const size_t index,
      38              :         RuntimeTensorDesc** dynamicSrcDescPptr, void* const customBuf, const uint32_t customBufSize,
      39              :         const RunContext& taskContext) const;
      40              :     int32_t MakeupDynamicOutputMbuf(
      41              :         Mbuf** const outputMbufPtr, const size_t index, const RuntimeTensorDesc* const runtimeTensorDesc,
      42              :         void* const customBuf, const uint32_t customBufSize, const RunContext& taskContext) const;
      43              :     int32_t FreeMbuf(const PostprocessDynamicOutputKernelArgs* const param, const RunContext& taskContext) const;
      44              : 
      45              :     bool IsSupportSdmaCopy() const;
      46              : 
      47              :     bool OptimizedMemCopy(
      48              :         void* const dst_data, const size_t dst_size, const void* const src_data, const size_t src_size) const;
      49              : };
      50              : } // namespace AicpuSchedule
      51              : #endif // OPERATOR_KERNEL_POST_PROCESS_DYNAMIC_OUTPUT_H