* 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 softmax_flash_base_impl.h
* \brief
*/
#if !defined(__ASCENDC_INCLUDE_INTERNAL_HEADERS__)
#pragma message( \
"impl/adv_api/detail/activation/softmax/softmax_flash_base_impl.h is an internal header file and must not be used directly. Functions or variables defined in this file may be removed in the future. Please use \"#include \"adv_api/activation/softmaxflash.h\"\" and use public functions or variables defined in interface headers files.")
#define __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#define __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_SOFTMAX_FLASH_BASE_IMPL_H__
#endif
#ifndef IMPL_ACTIVATION_SOFTMAX_SOFTMAX_FLASH_BASE_IMPL_H
#define IMPL_ACTIVATION_SOFTMAX_SOFTMAX_FLASH_BASE_IMPL_H
#if defined(__NPU_ARCH__) && \
(__NPU_ARCH__ == 3510 || __NPU_ARCH__ == 5102 || __NPU_ARCH__ == 3003 || __NPU_ARCH__ == 3113)
#include "regbase/3510/softmax_flash_impl.h"
#else
#include "softmax_common/softmax_common_flash.h"
#endif
#ifdef ASCENDC_CPU_DEBUG
#include "../../api_check/kernel_check/activation/softmax/softmax_flash/softmax_flash_check.h"
#endif
#include "../../api_check/kernel_api_check.h"
namespace AscendC {
template <typename T, bool isReuseSource = false, bool isBasicBlock = false>
__aicore__ inline void SoftmaxFlashImpl(
const LocalTensor<T>& dstTensor, const LocalTensor<T>& sumTensor, const LocalTensor<T>& maxTensor,
const LocalTensor<T>& srcTensor, const LocalTensor<T>& expMaxTensor, const LocalTensor<T>& inSumTensor,
const LocalTensor<T>& inMaxTensor, const SoftMaxTiling& tiling, bool isUpdate,
const SoftMaxShapeInfo& softmaxShapeInfo)
{
CHECK_FUNC_HIGHLEVEL_API(
SoftmaxFlash, (T, isReuseSource, isBasicBlock),
(dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor, inMaxTensor, tiling, isUpdate,
softmaxShapeInfo));
SoftmaxFlashCommonImpl<T, T, isReuseSource, isBasicBlock>(
dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor, inMaxTensor, tiling, isUpdate,
softmaxShapeInfo);
}
template <typename T, bool isReuseSource = false, bool isBasicBlock = false>
__aicore__ inline void SoftmaxFlashImpl(
const LocalTensor<half>& dstTensor, const LocalTensor<float>& sumTensor, const LocalTensor<float>& maxTensor,
const LocalTensor<half>& srcTensor, const LocalTensor<half>& expMaxTensor, const LocalTensor<float>& inSumTensor,
const LocalTensor<float>& inMaxTensor, const SoftMaxTiling& tiling, bool isUpdate,
const SoftMaxShapeInfo& softmaxShapeInfo)
{
CHECK_FUNC_HIGHLEVEL_API(
SoftmaxFlash, (T, isReuseSource, isBasicBlock),
(dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor, inMaxTensor, tiling, isUpdate,
softmaxShapeInfo));
SoftmaxFlashCommonImpl<half, float, isReuseSource, isBasicBlock>(
dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor, inMaxTensor, tiling, isUpdate,
softmaxShapeInfo);
}
template <typename T, bool isReuseSource = false, bool isBasicBlock = false>
__aicore__ inline void SoftmaxFlashImpl(
const LocalTensor<T>& dstTensor, const LocalTensor<T>& sumTensor, const LocalTensor<T>& maxTensor,
const LocalTensor<T>& srcTensor, const LocalTensor<T>& expMaxTensor, const LocalTensor<T>& inSumTensor,
const LocalTensor<T>& inMaxTensor, const LocalTensor<uint8_t>& sharedTmpBuffer, const SoftMaxTiling& tiling,
bool isUpdate, const SoftMaxShapeInfo& softmaxShapeInfo)
{
CHECK_FUNC_HIGHLEVEL_API(
SoftmaxFlash, (T, isReuseSource, isBasicBlock),
(dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor, inMaxTensor, sharedTmpBuffer, tiling,
isUpdate, softmaxShapeInfo));
SoftmaxFlashTmpBufCommonImpl<T, T, isReuseSource, isBasicBlock>(
dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor, inMaxTensor, sharedTmpBuffer, tiling,
isUpdate, softmaxShapeInfo);
}
template <typename T, bool isReuseSource = false, bool isBasicBlock = false>
__aicore__ inline void SoftmaxFlashImpl(
const LocalTensor<half>& dstTensor, const LocalTensor<float>& sumTensor, const LocalTensor<float>& maxTensor,
const LocalTensor<half>& srcTensor, const LocalTensor<half>& expMaxTensor, const LocalTensor<float>& inSumTensor,
const LocalTensor<float>& inMaxTensor, const LocalTensor<uint8_t>& sharedTmpBuffer, const SoftMaxTiling& tiling,
bool isUpdate, const SoftMaxShapeInfo& softmaxShapeInfo)
{
CHECK_FUNC_HIGHLEVEL_API(
SoftmaxFlash, (T, isReuseSource, isBasicBlock),
(dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor, inMaxTensor, sharedTmpBuffer, tiling,
isUpdate, softmaxShapeInfo));
SoftmaxFlashTmpBufCommonImpl<half, float, isReuseSource, isBasicBlock>(
dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor, inMaxTensor, sharedTmpBuffer, tiling,
isUpdate, softmaxShapeInfo);
}
}
#endif
#if defined(__UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_SOFTMAX_FLASH_BASE_IMPL_H__)
#undef __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#undef __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_SOFTMAX_FLASH_BASE_IMPL_H__
#endif