* Copyright (c) 2026 Huawei Technologies Co., Ltd.
* This program is free software, you can redistribute it and/or modify it under the terms and conditions of
* CANN Open Software License Agreement Version 2.0 (the "License").
* Please refer to the License for details. You may not use this file except in compliance with the License.
* THIS SOFTWARE IS PROVIDED ON AN " AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
* See LICENSE in the root of the software repository for the full text of the License.
*/
#ifndef OPS_GET_SHAPE_OP_KERNEL_ARCH35_GET_SHAPE_H_
#define OPS_GET_SHAPE_OP_KERNEL_ARCH35_GET_SHAPE_H_
#include "kernel_operator.h"
#include "kernel_operator_list_tensor_intf.h"
#include "get_shape_tiling_data.h"
namespace NsGetShape {
using namespace AscendC;
class KernelGetShape {
public:
__aicore__ inline KernelGetShape() {}
__aicore__ inline void Init(GM_ADDR x, GM_ADDR y, const GetShapeTilingData& tilingData);
__aicore__ inline void Process();
private:
TPipe pipe_;
TQue<QuePosition::VECOUT, GetShapeConst::BUFFER_NUM> outQueue_;
GlobalTensor<int32_t> yGm_;
ListTensorDesc inputList_;
uint32_t inputNum_ = 0;
int64_t totalDimNum_ = 0;
int32_t shapeValues_[GetShapeConst::MAX_TOTAL_DIM] = {0};
};
__aicore__ inline void KernelGetShape::Init(GM_ADDR x, GM_ADDR y, const GetShapeTilingData& tilingData)
{
inputList_ = ListTensorDesc(reinterpret_cast<__gm__ void*>(x));
yGm_.SetGlobalBuffer(reinterpret_cast<__gm__ int32_t*>(y));
pipe_.InitBuffer(outQueue_, GetShapeConst::BUFFER_NUM, GetShapeConst::MAX_TOTAL_DIM * sizeof(int32_t));
inputNum_ = static_cast<uint32_t>(tilingData.inputNum);
if (inputNum_ == 0) {
return;
}
totalDimNum_ = 0;
for (uint32_t i = 0; i < inputNum_; ++i) {
__gm__ int64_t* xGmPtr = inputList_.GetDataPtr<int64_t>(i);
GlobalTensor<int64_t> xGm;
xGm.SetGlobalBuffer(xGmPtr);
int64_t dimNum = xGm.GetValue(3);
if (dimNum < 0 || dimNum > GetShapeConst::MAX_DIM_PER_TENSOR) {
inputNum_ = 0;
totalDimNum_ = 0;
return;
}
if (totalDimNum_ + dimNum > GetShapeConst::MAX_TOTAL_DIM) {
inputNum_ = 0;
totalDimNum_ = 0;
return;
}
for (int64_t d = 0; d < dimNum; ++d) {
shapeValues_[totalDimNum_ + d] = static_cast<int32_t>(xGm.GetValue(4 + d));
}
totalDimNum_ += dimNum;
}
}
__aicore__ inline void KernelGetShape::Process()
{
if (totalDimNum_ <= 0 || totalDimNum_ > GetShapeConst::MAX_TOTAL_DIM) {
return;
}
LocalTensor<int32_t> yUb = outQueue_.AllocTensor<int32_t>();
for (int64_t i = 0; i < totalDimNum_; ++i) {
yUb.SetValue(i, shapeValues_[i]);
}
event_t evt = static_cast<event_t>(GetTPipePtr()->FetchEventID(HardEvent::S_MTE3));
SetFlag<HardEvent::S_MTE3>(evt);
WaitFlag<HardEvent::S_MTE3>(evt);
GetTPipePtr()->ReleaseEventID<HardEvent::S_MTE3>(evt);
outQueue_.EnQue(yUb);
LocalTensor<int32_t> yOut = outQueue_.DeQue<int32_t>();
DataCopyExtParams copyParams = {static_cast<uint16_t>(1), static_cast<uint32_t>(totalDimNum_ * sizeof(int32_t)), 0,
0, 0};
DataCopyPad(yGm_, yOut, copyParams);
outQueue_.FreeTensor(yOut);
}
}
#endif