* 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 swiglu.h
* \brief
*/
#if !defined(__ASCENDC_INCLUDE_INTERNAL_HEADERS__)
#define __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#define __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_SWIGLU_H__
#endif
#ifndef LIB_MATH_SWIGLU_H
#define LIB_MATH_SWIGLU_H
#include "kernel_tensor.h"
#if defined(__NPU_ARCH__) && (__NPU_ARCH__ == 3510 || __NPU_ARCH__ == 5102 || \
__NPU_ARCH__ == 3003 || __NPU_ARCH__ == 3113)
#include "../../../impl/adv_api/detail/activation/swiglu/swiglu_3510_impl.h"
#else
#include "../../../impl/adv_api/detail/activation/swiglu/swiglu_common_impl.h"
#endif
namespace AscendC {
#pragma begin_pipe(V)
* \note support data type: half and float
* Function�?
swish(x) = x / (1 + e^(-βx))
x1 = 1 + e^(-βx)
* \param [out] dstTensor, output LocalTensor
* \param [in] srcTensor0, input0 LocalTensor
* \param [in] srcTensor1, input1 LocalTensor
* \param [in] scalarValue, input scalar
*/
template <typename T, bool isReuseSource = false>
__aicore__ inline void SwiGLU(LocalTensor<T>& dstTensor, LocalTensor<T>& srcTensor0, LocalTensor<T>& srcTensor1,
const float& scalarValue)
{
if ASCEND_IS_AIC {
return;
}
SwiGLUImpl<T, isReuseSource>(dstTensor, srcTensor0, srcTensor1, scalarValue);
}
* \note support data type: half and float
*
* \param [out] dstTensor, output LocalTensor
* \param [in] srcTensor0, input0 LocalTensor
* \param [in] srcTensor1, input1 LocalTensor
* \param [in] scalarValue, input scalar
* \param [in] sharedTmpBuffer, input local temporary Tensor
*/
template <typename T, bool isReuseSource = false>
__aicore__ inline void SwiGLU(LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor0,
const LocalTensor<T>& srcTensor1, const float& scalarValue, const LocalTensor<uint8_t>& sharedTmpBuffer)
{
SwiGLUImpl<T, isReuseSource>(dstTensor, srcTensor0, srcTensor1, scalarValue, sharedTmpBuffer);
}
* \note support data type: half and float
*
* \param [out] dstTensor, output LocalTensor
* \param [in] srcTensor0, input0 LocalTensor
* \param [in] srcTensor1, input1 LocalTensor
* \param [in] scalarValue, input scalar
* \param [in] calCount, amount of data to be calculated
*/
template <typename T, bool isReuseSource = false>
__aicore__ inline void SwiGLU(LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor0,
const LocalTensor<T>& srcTensor1, const float& scalarValue, const uint32_t calCount)
{
if ASCEND_IS_AIC {
return;
}
SwiGLUImpl<T, isReuseSource>(dstTensor, srcTensor0, srcTensor1, scalarValue, calCount);
}
* \note support data type: half and float
*
* \param [out] dstTensor, output LocalTensor
* \param [in] srcTensor0, input0 LocalTensor
* \param [in] srcTensor1, input1 LocalTensor
* \param [in] scalarValue, input scalar
* \param [in] sharedTmpBuffer, input local temporary Tensor
* \param [in] calCount, amount of data to be calculated
*/
template <typename T, bool isReuseSource = false>
__aicore__ inline void SwiGLU(LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor0,
const LocalTensor<T>& srcTensor1, const float& scalarValue,
const LocalTensor<uint8_t>& sharedTmpBuffer, const uint32_t calCount)
{
SwiGLUImpl<T, isReuseSource>(dstTensor, srcTensor0, srcTensor1, scalarValue, sharedTmpBuffer, calCount);
}
#pragma end_pipe
}
#endif
#if defined(__UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_SWIGLU_H__)
#undef __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#undef __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_SWIGLU_H__
#endif