* 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 softmaxflash.h
* \brief SoftmaxFlash api of AscendC, this api will be deprecated, please use SoftmaxFlashV2 api instead
*/
#ifndef LIB_SOFTMAX_SOFTMAXFLASH_H
#define LIB_SOFTMAX_SOFTMAXFLASH_H
#include "kernel_tensor.h"
#include "kernel_tiling/kernel_tiling.h"
#if defined(__NPU_ARCH__) && (__NPU_ARCH__ == 2201 || __NPU_ARCH__ == 2002 || __NPU_ARCH__ == 3101 || __NPU_ARCH__ == 3002 || __NPU_ARCH__ == 3003 || __NPU_ARCH__ == 3113 || __NPU_ARCH__ == 5102)
#include "../../../impl/adv_api/detail/activation/softmax/softmax_common.h"
#include "../../../impl/adv_api/detail/activation/softmax/softmax_flash_base_impl.h"
#pragma begin_pipe(V)
namespace AscendC {
* \ingroup SoftmaxFlash
* \brief softmax calculation in FlashAttention process: you can refet to the paper:
* <Flashattention: Fast and memory-efficient exact attention with io-awareness>
* if isUpdate = false:
* max = rowmax(x), y = exp(x-max), sum = rowsum(y), y = y / sum
* if isUpdate = true:
* max = max(rowmax(x), inmax), y = exp(x-max), sum = rowsum(y)
* expmax = exp(inmax - max), sum = expmax * insum + sum
* expmax = expmax * insum / sum, y = y / sum
* \note support data type: half and float
* \param [out] dstTensor: output y
* \param [out] sumTensor: output sum
* \param [out] maxTensor: output max
* \param [out] expMaxTensor: output expmax, only valid when isUpdate is true
* \param [in] srcTensor: input x
* \param [in] inSumTensor: input last insum
* \param [in] inMaxTensor: input last inmax
* \param [in] softmaxShapeInfo: input src shape
* \param [in] tiling: input softmaxtiling
* \param [in] isUpdate: whether to update the sum and expmax
* \param [in] isReuseSource: reserved param
* \param [in] isBasicBlock: if src shape[m,k] satisfy the condition(m%8 == 0 && k%64 == 0),
* you can set true to improve performance
*/
template <typename T, bool isReuseSource = false, bool isBasicBlock = false>
[[deprecated(__FILE__ " is deprecated, please use softmaxflashv2.h instead!")]] __aicore__ inline void SoftmaxFlash(
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 = false,
const SoftMaxShapeInfo& softmaxShapeInfo = {})
{
if ASCEND_IS_AIC {
return;
}
TRACE_START(TraceId::SoftmaxFlash);
SoftmaxFlashImpl<T, isReuseSource, isBasicBlock>(dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor,
inMaxTensor, tiling, isUpdate, softmaxShapeInfo);
TRACE_STOP(TraceId::SoftmaxFlash);
}
* \ingroup SoftmaxFlash
* \brief softmax calculation in FlashAttention process: you can refet to the paper:
* <Flashattention: Fast and memory-efficient exact attention with io-awareness>
* if isUpdate = false:
* max = rowmax(x), y = exp(x-max), sum = rowsum(y), y = y / sum
* if isUpdate = true:
* max = max(rowmax(x), inmax), y = exp(x-max), sum = rowsum(y)
* expmax = exp(inmax - max), sum = expmax * insum + sum
* expmax = expmax * insum / sum, y = y / sum
* \param [out] dstTensor: output y with dtype of half
* \param [out] sumTensor: output sum with dtype of float
* \param [out] maxTensor: output max with dtype of float
* \param [out] expMaxTensor: output expmax with dtype of half, only valid when isUpdate is true
* \param [in] srcTensor: input x with dtype of half
* \param [in] inSumTensor: input last insum with dtype of float
* \param [in] inMaxTensor: input last inmax with dtype of float
* \param [in] softmaxShapeInfo: input src shape
* \param [in] tiling: input softmaxtiling
* \param [in] isUpdate: whether to update the sum and expmax
* \param [in] isReuseSource: reserved param
* \param [in] isBasicBlock: if src shape[m,k] satisfy the condition(m%8 == 0 && k%64 == 0),
* you can set true to improve performance
*/
template <typename T, bool isReuseSource = false, bool isBasicBlock = false>
[[deprecated(__FILE__ " is deprecated, please use softmaxflashv2.h instead!")]]
__aicore__ inline void SoftmaxFlash(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 = false, const SoftMaxShapeInfo& softmaxShapeInfo = {})
{
if ASCEND_IS_AIC {
return;
}
TRACE_START(TraceId::SoftmaxFlash);
SoftmaxFlashImpl<T, isReuseSource, isBasicBlock>(dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor,
inMaxTensor, tiling, isUpdate, softmaxShapeInfo);
TRACE_STOP(TraceId::SoftmaxFlash);
}
* \ingroup SoftmaxFlash
* \brief softmax calculation in FlashAttention process: you can refet to the paper:
* <Flashattention: Fast and memory-efficient exact attention with io-awareness>
* if isUpdate = false:
* max = rowmax(x), y = exp(x-max), sum = rowsum(y), y = y / sum
* if isUpdate = true:
* max = max(rowmax(x), inmax), y = exp(x-max), sum = rowsum(y)
* expmax = exp(inmax - max), sum = expmax * insum + sum
* expmax = expmax * insum / sum, y = y / sum
* \note support data type: half and float
* \param [out] dstTensor: output y
* \param [out] sumTensor: output sum
* \param [out] maxTensor: output max
* \param [out] expMaxTensor: output expmax, only valid when isUpdate is true
* \param [in] srcTensor: input x
* \param [in] inSumTensor: input last insum
* \param [in] inMaxTensor: input last inmax
* \param [in] sharedTmpBuffer: input local temporary Tensor, you can get the range by tilingfunc of
* GetSoftMaxFlashMinTmpSize/GetSoftMaxFlashMaxTmpSize
* \param [in] softmaxShapeInfo: input src shape
* \param [in] tiling: input softmaxtiling
* \param [in] isUpdate: whether to update the sum and expmax
* \param [in] isReuseSource: reserved param
* \param [in] isBasicBlock: if src shape[m,k] satisfy the condition(m%8 == 0 && k%64 == 0),
* you can set true to improve performance
*/
template <typename T, bool isReuseSource = false, bool isBasicBlock = false>
[[deprecated(__FILE__ " is deprecated, please use softmaxflashv2.h instead!")]]
__aicore__ inline void SoftmaxFlash(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 = false, const SoftMaxShapeInfo& softmaxShapeInfo = {})
{
if ASCEND_IS_AIC {
return;
}
TRACE_START(TraceId::SoftmaxFlash);
SoftmaxFlashImpl<T, isReuseSource, isBasicBlock>(dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor,
inMaxTensor, sharedTmpBuffer, tiling, isUpdate, softmaxShapeInfo);
TRACE_STOP(TraceId::SoftmaxFlash);
}
* \ingroup SoftmaxFlash
* \brief softmax calculation in FlashAttention process: you can refet to the paper:
* <Flashattention: Fast and memory-efficient exact attention with io-awareness>
* if isUpdate = false:
* max = rowmax(x), y = exp(x-max), sum = rowsum(y), y = y / sum
* if isUpdate = true:
* max = max(rowmax(x), inmax), y = exp(x-max), sum = rowsum(y)
* expmax = exp(inmax - max), sum = expmax * insum + sum
* expmax = expmax * insum / sum, y = y / sum
* \param [out] dstTensor: output y with dtype of half
* \param [out] sumTensor: output sum with dtype of float
* \param [out] maxTensor: output max with dtype of float
* \param [out] expMaxTensor: output expmax with dtype of half, only valid when isUpdate is true
* \param [in] srcTensor: input x with dtype of half
* \param [in] inSumTensor: input last insum with dtype of float
* \param [in] inMaxTensor: input last inmax with dtype of float
* \param [in] sharedTmpBuffer: input local temporary Tensor, you can get the range by tilingfunc of
* GetSoftMaxFlashMinTmpSize/GetSoftMaxFlashMaxTmpSize
* \param [in] softmaxShapeInfo: input src shape
* \param [in] tiling: input softmaxtiling
* \param [in] isUpdate: whether to update the sum and expmax
* \param [in] isReuseSource: reserved param
* \param [in] isBasicBlock: if src shape[m,k] satisfy the condition(m%8 == 0 && k%64 == 0),
* you can set true to improve performance
*/
template <typename T, bool isReuseSource = false, bool isBasicBlock = false>
[[deprecated(__FILE__ " is deprecated, please use softmaxflashv2.h instead!")]]
__aicore__ inline void SoftmaxFlash(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 = false,
const SoftMaxShapeInfo& softmaxShapeInfo = {})
{
if ASCEND_IS_AIC {
return;
}
TRACE_START(TraceId::SoftmaxFlash);
SoftmaxFlashImpl<T, isReuseSource, isBasicBlock>(dstTensor, sumTensor, maxTensor, srcTensor, expMaxTensor, inSumTensor,
inMaxTensor, sharedTmpBuffer, tiling, isUpdate, softmaxShapeInfo);
TRACE_STOP(TraceId::SoftmaxFlash);
}
}
#pragma end_pipe
#endif
#endif
