* Copyright (c) 2025 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.
*/
* \file selectwithbytesmask.h
* \brief
*/
#ifndef LIB_SELECT_SELECT_WITH_BYTES_MASK_H
#define LIB_SELECT_SELECT_WITH_BYTES_MASK_H
#if defined(__NPU_ARCH__) && (__NPU_ARCH__ == 2201 || __NPU_ARCH__ == 2002 || __NPU_ARCH__ == 3101 || \
__NPU_ARCH__ == 5102 || __NPU_ARCH__ == 3003 || __NPU_ARCH__ == 3113)
#include "kernel_tensor.h"
#include "kernel_utils.h"
#if defined(__NPU_ARCH__) && (__NPU_ARCH__ == 3101 || __NPU_ARCH__ == 5102 || \
__NPU_ARCH__ == 3003 || __NPU_ARCH__ == 3113)
#include "../../../impl/adv_api/detail/select/selectwithbytesmask/selectwithbytesmask_c310_impl.h"
#else
#include "../../../impl/adv_api/detail/select/selectwithbytesmask/selectwithbytesmask_impl.h"
#endif
namespace AscendC {
#pragma begin_pipe(V)
* \ingroup Select
* \brief Selects values from two sources(tensor/scalar) and put into dst tensor elementwisely according to the mask
* Tensor Value.
* True: mask tensor position value == 1, select src1 scalar.
* False: mask tensor position value == 0, select src0 Tensor correspoding value.
* \note support data type: T half/float, U uint8_t/int8_t/uint16_t/int16_t/uint32_t/int32_t/bool
* Tensor shape can be considered as [first_axis, last_axis],
* and it requries that mask_first_axis = source_first_axis, mask_last_axis >= source_first_axis
* mask last axis should be 32B aligned and element size should be time of 16
* src last axis should be 32B aligned
* it doesn't allow src/sharedTmpBuffer tensor address overlap.
* \tparam isReuseMask: whether to reuse the mask space for process
* \param [out] dst: output LocalTensor
* \param [in] src0: input LocalTensor
* \param [in] src1: input scalar
* \param [in] mask: mask LocalTensor
* \param [in] sharedTmpBuffer: extra tmp buffer used as intermediate values among calculation process.
* \param [in] info: shape information of input/mask tensors
*/
template <typename T, typename U, bool isReuseMask = true>
__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<T>& src0, T src1, const LocalTensor<U>& mask,
const LocalTensor<uint8_t>& sharedTmpBuffer, const SelectWithBytesMaskShapeInfo& info)
{
SelectWithBytesMaskImpl<T, U, isReuseMask, false>(dst, src0, src1, mask, sharedTmpBuffer, info);
}
* \ingroup Select
* \brief Selects values from two sources(tensor/scalar) and put into dst tensor elementwisely according to the mask
* Tensor Value.
* True: mask tensor position value == 1, select src1 Tensor correspoding value.
* False: mask tensor position value == 0, select src0 scalar.
* \note support data type: T half/float, U uint8_t/int8_t/uint16_t/int16_t/uint32_t/int32_t/bool
* Tensor shape can be considered as [first_axis, last_axis],
* and it requries that mask_first_axis = source_first_axis, mask_last_axis >= source_first_axis
* mask last axis should be 32B aligned and element size should be time of 16
* src last axis should be 32B aligned
* it doesn't allow src/sharedTmpBuffer tensor address overlap.
* \tparam isReuseMask: whether to reuse the mask space for process
* \param [out] dst: output LocalTensor
* \param [in] src0: input scalar
* \param [in] src1: input LocalTensor
* \param [in] mask: mask LocalTensor
* \param [in] sharedTmpBuffer: extra tmp buffer used as intermediate values among calculation process.
* \param [in] info: shape information of input/mask tensors
*/
template <typename T, typename U, bool isReuseMask = true>
__aicore__ inline void Select(const LocalTensor<T>& dst, T src0, const LocalTensor<T>& src1, const LocalTensor<U>& mask,
const LocalTensor<uint8_t>& sharedTmpBuffer, const SelectWithBytesMaskShapeInfo& info)
{
SelectWithBytesMaskImpl<T, U, isReuseMask, true>(dst, src1, src0, mask, sharedTmpBuffer, info);
}
#pragma end_pipe
}
#endif
#endif
