* 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 kernel_operator_vec_scatter_intf_impl.h
* \brief
*/
#ifndef ASCENDC_MODULE_OPERATOR_VEC_SCATTER_INTERFACE_IMPL_H
#define ASCENDC_MODULE_OPERATOR_VEC_SCATTER_INTERFACE_IMPL_H
#include "kernel_tensor.h"
#include "kernel_check.h"
#if __NPU_ARCH__ == 1001
#include "dav_c100/kernel_operator_vec_scatter_impl.h"
#elif __NPU_ARCH__ == 2002
#include "dav_m200/kernel_operator_vec_scatter_impl.h"
#elif __NPU_ARCH__ == 2201
#include "dav_c220/kernel_operator_vec_scatter_impl.h"
#elif __NPU_ARCH__ == 3002
#include "dav_m300/kernel_operator_vec_scatter_impl.h"
#elif __NPU_ARCH__ == 3102
#include "dav_m310/kernel_operator_vec_scatter_impl.h"
#elif __NPU_ARCH__ == 3101
#include "dav_c310/kernel_operator_vec_scatter_impl.h"
#elif (__NPU_ARCH__ == 5102)
#include "dav_m510/kernel_operator_vec_scatter_impl.h"
#elif (__NPU_ARCH__ == 3003)
#include "dav_l300/kernel_operator_vec_scatter_impl.h"
#elif (__NPU_ARCH__ == 3113)
#include "dav_l311/kernel_operator_vec_scatter_impl.h"
#endif
#pragma begin_pipe(V)
namespace AscendC {
template <typename T>
__aicore__ inline void ScatterCheck()
{
using PrimType = PrimT<T>;
#if (__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102)
ASCENDC_ASSERT((SupportType<PrimType, uint8_t, int8_t, half, bfloat16_t, uint16_t, int16_t, float, uint32_t,
int32_t, uint64_t, int64_t>()), {KERNEL_LOG(KERNEL_ERROR,
"Failed to check dtype in Scatter, current api support dtype combination is src and dst both: "
"uint8 / int8 / half / bfloat16_t / uint16_t / int16_t / float / uint32_t / int32_t, uint64_t, int64_t");});
#elif ((__NPU_ARCH__ == 3002) || (__NPU_ARCH__ == 3102) || (__NPU_ARCH__ == 3101))
ASCENDC_ASSERT((SupportType<PrimType, uint8_t, int8_t, half, bfloat16_t, uint16_t, int16_t, float, uint32_t,
int32_t>()), {KERNEL_LOG(KERNEL_ERROR,
"Failed to check dtype in Scatter, current api support dtype combination is src and "
"dst both: uint8 / int8 / half / bfloat16_t / uint16_t / int16_t / float / uint32_t / int32_t");});
#else
ASCENDC_ASSERT((SupportType<PrimType, half, uint16_t, int16_t, float, uint32_t, int32_t>()),
{KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Scatter, current api support dtype combination is src and "
"dst both: half / uint16_t / int16_t / float / uint32_t / int32_t");});
#endif
}
* @ingroup scatter Level 0
* @brief scatter element from dst according to dstOffset
* @param [out] dst output LocalTensor
* @param [in] src input LocalTensor
* @param [in] dstOffset input LocalTensor
* @param [in] dstBaseAddr base address of dst
* @param [in] mask valid element count
* @param [in] repeatTime repeat times
* @param [in] dstRepStride dst repeat stride
*/
template <typename T>
__aicore__ inline void Scatter(const LocalTensor<T>& dst, const LocalTensor<T>& src,
const LocalTensor<uint32_t>& dstOffset, const uint32_t dstBaseAddr, const uint64_t mask,
const uint8_t repeatTime, const uint8_t srcRepStride)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFunScatter(dst, src, dstOffset, dstBaseAddr, mask, repeatTime, srcRepStride, "Scatter")) {
ASCENDC_REPORT_CHECK_ERROR("Scatter", KernelFuncType::MASK_COUNT_MODE);
}
#endif
using PrimType = PrimT<T>;
ScatterCheck<T>();
const uint32_t dstLength = dst.GetSize();
ScatterImpl((__ubuf__ PrimType *)dst.GetPhyAddr(), (__ubuf__ PrimType *)src.GetPhyAddr(),
(__ubuf__ uint32_t *)dstOffset.GetPhyAddr(), dstLength, dstBaseAddr, mask, repeatTime, srcRepStride);
}
* @ingroup scatter Level 0
* @brief scatter element from dst according to dstOffset
* @param [out] dst output LocalTensor
* @param [in] src input LocalTensor
* @param [in] dstOffset input LocalTensor
* @param [in] dstBaseAddr base address of dst
* @param [in] mask valid element count(bit mode)
* @param [in] repeatTime repeat times
* @param [in] dstRepStride dst repeat stride
*/
template <typename T>
__aicore__ inline void Scatter(const LocalTensor<T>& dst, const LocalTensor<T>& src,
const LocalTensor<uint32_t>& dstOffset, const uint32_t dstBaseAddr, const uint64_t mask[],
const uint8_t repeatTime, const uint8_t srcRepStride)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFunScatter(dst, src, dstOffset, dstBaseAddr, mask, repeatTime, srcRepStride, "Scatter")) {
ASCENDC_REPORT_CHECK_ERROR("Scatter", KernelFuncType::MASK_BIT_MODE);
}
#endif
using PrimType = PrimT<T>;
ScatterCheck<T>();
const uint32_t dstLength = dst.GetSize();
ScatterImpl((__ubuf__ PrimType *)dst.GetPhyAddr(), (__ubuf__ PrimType *)src.GetPhyAddr(),
(__ubuf__ uint32_t *)dstOffset.GetPhyAddr(), dstLength, dstBaseAddr, mask, repeatTime, srcRepStride);
}
* @ingroup scatter Level 2
* @brief scatter element from dst according to dstOffset
* @param [out] dst output LocalTensor
* @param [in] src input LocalTensor
* @param [in] dstOffset input LocalTensor
* @param [in] dstBaseAddr base address of dst
* @param [in] count element count
*/
template <typename T>
__aicore__ inline void Scatter(const LocalTensor<T>& dst, const LocalTensor<T>& src,
const LocalTensor<uint32_t>& dstOffset, const uint32_t dstBaseAddr, const uint32_t count)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFunScatter(dst, src, dstOffset, dstBaseAddr, count, "Scatter")) {
ASCENDC_REPORT_CHECK_ERROR("Scatter", KernelFuncType::NONE_MODE);
}
#endif
using PrimType = PrimT<T>;
ScatterCheck<T>();
uint32_t vectorRegWidth = 256;
#if (__NPU_ARCH__ == 3002) || (__NPU_ARCH__ == 3102) || (__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102)
ScatterImpl((__ubuf__ PrimType *)dst.GetPhyAddr(), (__ubuf__ PrimType *)src.GetPhyAddr(),
(__ubuf__ uint32_t *)dstOffset.GetPhyAddr(), dstBaseAddr, count);
#else
uint32_t elementCountSingleRepeat;
if constexpr (sizeof(PrimType) == sizeof(uint16_t)) {
elementCountSingleRepeat = 128;
} else {
elementCountSingleRepeat = 64;
}
#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3003) || \
(__NPU_ARCH__ == 3113))
vectorRegWidth = VECTOR_REG_WIDTH;
#endif
#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3003) || \
(__NPU_ARCH__ == 3113))
elementCountSingleRepeat = vectorRegWidth / sizeof(T);
uint32_t repeatStride = vectorRegWidth / ONE_BLK_SIZE;
#endif
const uint32_t elementCountTail = count % elementCountSingleRepeat;
const uint8_t repeatTime = count / elementCountSingleRepeat;
#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3003) || \
(__NPU_ARCH__ == 3113))
if (repeatTime > 0) {
Scatter(dst, src, dstOffset, dstBaseAddr, (uint64_t)elementCountSingleRepeat, repeatTime,
repeatStride);
}
if (elementCountTail > 0) {
const uint32_t offset = count - elementCountTail;
Scatter(dst, src[offset], dstOffset[offset], dstBaseAddr, (uint64_t)elementCountTail, 1,
repeatStride);
}
#else
if (repeatTime > 0) {
Scatter(dst, src, dstOffset, dstBaseAddr, static_cast<uint64_t>(elementCountSingleRepeat), repeatTime,
DEFAULT_REPEAT_STRIDE);
}
if (elementCountTail > 0) {
const uint32_t offset = count - elementCountTail;
Scatter(dst, src[offset], dstOffset[offset], dstBaseAddr, static_cast<uint64_t>(elementCountTail), 1,
DEFAULT_REPEAT_STRIDE);
}
#endif
#endif
}
}
#pragma end_pipe
#endif
