* 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 erf.h
* \brief Defines a series of interface used to do elementwise math Erf calculation.
* Formula: Error function or Gauss error function.
* The Erf function does not have an elementary function expression, and there is calculating by
* function approximation.
* The approximate calculation formula is as follows:
* Erf(x) = P(Clip(x)) / Q(Clip(x))
* Clip(x) = Min(-3.92, Max(x, 3.92))
* P(x) = (((((0.053443748819x^2+0.75517016694e1)x^2+0.10162808918e3)x^2+0.13938061484e4)x^2+0.50637915060e4)x^2
+0.29639384698e
* Q(x) = ((((x^2+0.31212858877e2)x^2+0.39856963806e3)x^2+0.30231248150e4)x^2+0.13243365831e5)x^2+0.26267224157e5
*/
#ifndef LIB_MATH_ERF_H
#define LIB_MATH_ERF_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"
#if defined(__NPU_ARCH__) && (__NPU_ARCH__ == 3101 || __NPU_ARCH__ == 5102 || __NPU_ARCH__ == 3003 || \
__NPU_ARCH__ == 3113)
#include "erf_utils.h"
#include "../../../impl/adv_api/detail/math/erf/erf_c310_impl.h"
#else
#include "../../../impl/adv_api/detail/math/erf/erf_common_impl.h"
#endif
namespace AscendC {
#pragma begin_pipe(V)
#if defined(__NPU_ARCH__) && (__NPU_ARCH__ == 3101 || __NPU_ARCH__ == 5102 || __NPU_ARCH__ == 3003 || \
__NPU_ARCH__ == 3113)
* \ingroup Erf
* \brief compute Erf elementwisely
* \tparam T: half/float
* \tparam isReuseSource: whether allows API to modify source data, usually for performance reason
* \tparam defaultErfConfig: the config for Erf, choose different algo
* \param [out] dstTensor: output LocalTensor
* \param [in] srcTensor: input LocalTensor
* \param [in] sharedTmpBuffer: extra temporary shared space used for intermediate values among calculation process,
* whose required space size should refer to corresponding tiling API, which is defined at erf_tiling.h.
* Generally, the more space you allocate, the better performance you will achieve, and the performance
* reaches peak when buffer size is maximum(calculated by tiling function). Moreover, it is not guaranteed
* that the shared space will be cleared after usage, the data could be anything.
* \param [in] calCount: the number of elements to be processed.
* \note src/dst Tensor must be 32B aligned, and it doesn't allow src/dst/sharedTmpBuffer tensor address overlap.
*/
template <typename T, bool isReuseSource = false, const ErfConfig& config = defaultErfConfig>
__aicore__ inline void Erf(const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor,
const LocalTensor<uint8_t>& sharedTmpBuffer, const uint32_t calCount)
{
ErfImpl<T, isReuseSource, config>(dstTensor, srcTensor, sharedTmpBuffer, calCount);
}
* \ingroup Erf
* \brief compute Erf elementwisely for whole source tensor
* \tparam T: half/float
* \tparam isReuseSource: whether allows API to modify source data, usually for performance reason
* \tparam defaultErfConfig: the config for Erf, choose different algo
* \param [out] dstTensor: output LocalTensor
* \param [in] srcTensor: input LocalTensor
* \param [in] sharedTmpBuffer: extra temporary shared space used for intermediate values among calculation process,
* whose required space size should refer to corresponding tiling API, which is defined at erf_tiling.h.
* Generally, the more space you allocate, the better performance you will achieve, and the performance
* reaches peak when buffer size is maximum(calculated by tiling function). Moreover, it is not guaranteed
* that the shared space will be cleared after usage, the data could be anything.
* \note src/dst Tensor must be 32B aligned, and it doesn't allow src/dst/sharedTmpBuffer tensor address overlap.
*/
template <typename T, bool isReuseSource = false, const ErfConfig& config = defaultErfConfig>
__aicore__ inline void Erf(
const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor, const LocalTensor<uint8_t>& sharedTmpBuffer)
{
Erf<T, isReuseSource, config>(dstTensor, srcTensor, sharedTmpBuffer, srcTensor.GetSize());
}
* \ingroup Erf
* \brief compute Erf elementwisely
* \tparam T: half/float
* \tparam isReuseSource: whether allows API to modify source data, usually for performance reason
* \tparam defaultErfConfig: the config for Erf, choose different algo
* \param [out] dstTensor: output LocalTensor
* \param [in] srcTensor: input LocalTensor
* \param [in] calCount: the number of elements to be processed.
* \note src/dst Tensor must be 32B aligned, and it doesn't allow src/dst/sharedTmpBuffer tensor address overlap.
*/
template <typename T, bool isReuseSource = false, const ErfConfig& config = defaultErfConfig>
__aicore__ inline void Erf(const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor, const uint32_t calCount)
{
ErfImpl<T, isReuseSource, config>(dstTensor, srcTensor, calCount);
}
* \ingroup Erf
* \brief compute Erf elementwisely for whole source tensor
* \tparam T: half/float
* \tparam isReuseSource: whether allows API to modify source data, usually for performance reason
* \tparam defaultErfConfig: the config for Erf, choose different algo
* \param [out] dstTensor: output LocalTensor
* \param [in] srcTensor: input LocalTensor
* \note src/dst Tensor must be 32B aligned, and it doesn't allow src/dst/sharedTmpBuffer tensor address overlap.
*/
template <typename T, bool isReuseSource = false, const ErfConfig& config = defaultErfConfig>
__aicore__ inline void Erf(const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor)
{
Erf<T, isReuseSource, config>(dstTensor, srcTensor, srcTensor.GetSize());
}
#else
* \ingroup Erf
* \brief compute Erf elementwisely
* \tparam T: half/float
* \tparam isReuseSource: whether allows API to modify source data, usually for performance reason
* \param [out] dstTensor: output LocalTensor
* \param [in] srcTensor: input LocalTensor
* \param [in] sharedTmpBuffer: extra temporary shared space used for intermediate values among calculation process,
* whose required space size should refer to corresponding tiling API, which is defined at erf_tiling.h.
* Generally, the more space you allocate, the better performance you will achieve, and the performance
* reaches peak when buffer size is maximum(calculated by tiling function). Moreover, it is not guaranteed
* that the shared space will be cleared after usage, the data could be anything.
* \param [in] calCount: the number of elements to be processed.
* \note src/dst Tensor must be 32B aligned, and it doesn't allow src/dst/sharedTmpBuffer tensor address overlap.
*/
template <typename T, bool isReuseSource = false>
__aicore__ inline void Erf(const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor,
const LocalTensor<uint8_t>& sharedTmpBuffer, const uint32_t calCount)
{
ErfImpl<T, isReuseSource>(dstTensor, srcTensor, sharedTmpBuffer, calCount);
}
* \ingroup Erf
* \brief compute Erf elementwisely for whole source tensor
* \tparam T: half/float
* \tparam isReuseSource: whether allows API to modify source data, usually for performance reason
* \param [out] dstTensor: output LocalTensor
* \param [in] srcTensor: input LocalTensor
* \param [in] sharedTmpBuffer: extra temporary shared space used for intermediate values among calculation process,
* whose required space size should refer to corresponding tiling API, which is defined at erf_tiling.h.
* Generally, the more space you allocate, the better performance you will achieve, and the performance
* reaches peak when buffer size is maximum(calculated by tiling function). Moreover, it is not guaranteed
* that the shared space will be cleared after usage, the data could be anything.
* \note src/dst Tensor must be 32B aligned, and it doesn't allow src/dst/sharedTmpBuffer tensor address overlap.
*/
template <typename T, bool isReuseSource = false>
__aicore__ inline void Erf(
const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor, const LocalTensor<uint8_t>& sharedTmpBuffer)
{
Erf<T, isReuseSource>(dstTensor, srcTensor, sharedTmpBuffer, srcTensor.GetSize());
}
* \ingroup Erf
* \brief compute Erf elementwisely
* \tparam T: half/float
* \tparam isReuseSource: whether allows API to modify source data, usually for performance reason
* \param [out] dstTensor: output LocalTensor
* \param [in] srcTensor: input LocalTensor
* \param [in] calCount: the number of elements to be processed.
* \note src/dst Tensor must be 32B aligned, and it doesn't allow src/dst/sharedTmpBuffer tensor address overlap.
*/
template <typename T, bool isReuseSource = false>
__aicore__ inline void Erf(const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor, const uint32_t calCount)
{
ErfImpl<T, isReuseSource>(dstTensor, srcTensor, calCount);
}
* \ingroup Erf
* \brief compute Erf elementwisely for whole source tensor
* \tparam T: half/float
* \tparam isReuseSource: whether allows API to modify source data, usually for performance reason
* \param [out] dstTensor: output LocalTensor
* \param [in] srcTensor: input LocalTensor
* \note src/dst Tensor must be 32B aligned, and it doesn't allow src/dst/sharedTmpBuffer tensor address overlap.
*/
template <typename T, bool isReuseSource = false>
__aicore__ inline void Erf(const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor)
{
Erf<T, isReuseSource>(dstTensor, srcTensor, srcTensor.GetSize());
}
#endif
#pragma end_pipe
}
#endif
#endif
