* 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.
*/
#if !defined(__ASCENDC_INCLUDE_INTERNAL_HEADERS__)
#pragma message( \
"impl/adv_api/detail/math/frac/frac_common_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/math/frac.h\"\" and use public functions or variables defined in interface headers files.")
#define __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#define __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_MATH_FRAC_FRAC_COMMON_IMPL_H__
#endif
#ifndef IMPL_MATH_FRAC_FRAC_COMMON_IMPL_H
#define IMPL_MATH_FRAC_FRAC_COMMON_IMPL_H
#include "kernel_tensor.h"
#include "kernel_basic_intf.h"
#include "../../common/check.h"
#ifdef ASCENDC_CPU_DEBUG
#include "../../api_check/kernel_check/math/frac/frac_check.h"
#endif
#include "../../api_check/kernel_api_check.h"
#if defined(__NPU_ARCH__) && __NPU_ARCH__ == 2002
#include "frac_v200_impl.h"
#elif defined(__NPU_ARCH__) && __NPU_ARCH__ == 2201
#include "frac_v220_impl.h"
#endif
namespace AscendC {
template <typename T>
__aicore__ inline void FracCompute(
const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor, const LocalTensor<float>& tmpTensor,
const uint32_t splitSize)
{
TruncCastForFrac(dstTensor, srcTensor, dstTensor);
Sub<T, false>(
dstTensor, srcTensor, dstTensor, MASK_PLACEHOLDER, 1,
{1, 1, 1, DEFAULT_REPEAT_STRIDE, DEFAULT_REPEAT_STRIDE, DEFAULT_REPEAT_STRIDE});
PipeBarrier<PIPE_V>();
}
template <>
__aicore__ inline void FracCompute(
const LocalTensor<half>& dstTensor, const LocalTensor<half>& srcTensor, const LocalTensor<float>& tmpTensor,
const uint32_t splitSize)
{
LocalTensor<float> srcTmpTensor = tmpTensor;
LocalTensor<float> dstTmpTensor = tmpTensor[splitSize];
Cast<float, half, false>(
srcTmpTensor, srcTensor, RoundMode::CAST_NONE, MASK_PLACEHOLDER, 1,
{1, 1, DEFAULT_REPEAT_STRIDE, DEFAULT_REPEAT_STRIDE / 2});
PipeBarrier<PIPE_V>();
TruncCastForFrac(dstTmpTensor, srcTmpTensor, dstTmpTensor);
Sub<float, false>(
dstTmpTensor, srcTmpTensor, dstTmpTensor, MASK_PLACEHOLDER, 1,
{1, 1, 1, DEFAULT_REPEAT_STRIDE, DEFAULT_REPEAT_STRIDE, DEFAULT_REPEAT_STRIDE});
PipeBarrier<PIPE_V>();
Cast<half, float, false>(
dstTensor, dstTmpTensor, RoundMode::CAST_NONE, MASK_PLACEHOLDER, 1,
{1, 1, DEFAULT_REPEAT_STRIDE / 2, DEFAULT_REPEAT_STRIDE});
PipeBarrier<PIPE_V>();
}
template <typename T, bool isReuseSource = false>
__aicore__ inline void FracImpl(
const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor, const LocalTensor<uint8_t>& sharedTmpBuffer,
const uint32_t calCount)
{
if ASCEND_IS_AIC {
return;
}
CHECK_FUNC_HIGHLEVEL_API(Frac, (T, isReuseSource), (dstTensor, srcTensor, sharedTmpBuffer, calCount));
uint32_t bufferSize = sharedTmpBuffer.GetSize();
uint32_t tmpBufferSize = bufferSize / sizeof(float);
CheckTmpBufferSize(tmpBufferSize, 0, bufferSize);
LocalTensor<float> tmpBuffer = sharedTmpBuffer.ReinterpretCast<float>();
uint32_t stackSize = 0;
uint32_t round = 1;
uint32_t tail = 0;
constexpr uint8_t FRAC_HALF_CALC_PROCEDURE = 2;
if constexpr (sizeof(T) == sizeof(half)) {
stackSize = tmpBufferSize / FRAC_HALF_CALC_PROCEDURE / ONE_BLK_SIZE * ONE_BLK_SIZE;
CheckTmpBufferSize(stackSize, 0, bufferSize);
round = calCount / stackSize;
tail = calCount % stackSize;
SetMaskCount();
SetVectorMask<half, MaskMode::COUNTER>(0, stackSize);
} else {
SetMaskCount();
SetVectorMask<half, MaskMode::COUNTER>(0, calCount);
}
uint32_t offset = 0;
for (uint32_t i = 0; i < round; i++) {
FracCompute(dstTensor[offset], srcTensor[offset], tmpBuffer, stackSize);
offset = offset + stackSize;
}
if (tail != 0) {
SetVectorMask<half, MaskMode::COUNTER>(0, tail);
FracCompute(dstTensor[offset], srcTensor[offset], tmpBuffer, stackSize);
}
SetMaskNorm();
ResetMask();
}
template <typename T, bool isReuseSource = false>
__aicore__ inline void FracImpl(
const LocalTensor<T>& dstTensor, const LocalTensor<T>& srcTensor, const uint32_t calCount)
{
if ASCEND_IS_AIC {
return;
}
LocalTensor<uint8_t> sharedTmpBuffer;
bool ans = PopStackBuffer<uint8_t, TPosition::LCM>(sharedTmpBuffer);
ASCENDC_ASSERT((ans), { KERNEL_LOG(KERNEL_ERROR, "PopStackBuffer Error!"); });
FracImpl<T, isReuseSource>(dstTensor, srcTensor, sharedTmpBuffer, calCount);
}
}
#endif
#if defined(__UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_MATH_FRAC_FRAC_COMMON_IMPL_H__)
#undef __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#undef __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_MATH_FRAC_FRAC_COMMON_IMPL_H__
#endif