* Copyright (c) 2026 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_TENSOR_API_INCLUDE_COMPILER_INTERNAL_HEADERS)
#define ASCENDC_TENSOR_API_INCLUDE_COMPILER_INTERNAL_HEADERS
#define UNDEF_ASCENDC_TENSOR_API_INCLUDE_COMPILER_INTERNAL_HEADERS_ASCENDC_TENSOR_API_H
#endif
* \file copy_op.h
* \brief
*/
#ifndef INCLUDE_TENSOR_API_ARCH_CUBE_COPY_OP_H
#define INCLUDE_TENSOR_API_ARCH_CUBE_COPY_OP_H
#include "tensor_api/arch/cube/copy_trait.h"
namespace asc {
namespace te {
struct l0c_to_gm_params {
__aicore__ constexpr l0c_to_gm_params() = default;
__aicore__ constexpr l0c_to_gm_params(unit_flag_mode unit_flag) : unit_flag(unit_flag) {}
unit_flag_mode unit_flag = unit_flag_mode::disable;
};
struct l0c_to_ub_params {
__aicore__ constexpr l0c_to_ub_params() = default;
__aicore__ constexpr l0c_to_ub_params(unit_flag_mode unit_flag, uint8_t sub_block_id = 0)
: unit_flag(unit_flag), sub_block_id(sub_block_id)
{}
unit_flag_mode unit_flag = unit_flag_mode::disable;
uint8_t sub_block_id = 0;
};
struct l0c_to_l1_params {
__aicore__ constexpr l0c_to_l1_params() = default;
__aicore__ constexpr l0c_to_l1_params(unit_flag_mode unit_flag) : unit_flag(unit_flag) {}
unit_flag_mode unit_flag = unit_flag_mode::disable;
};
template <typename PadType>
struct img2col_params {
uint16_t m_extension = 0;
uint16_t k_extension = 0;
uint16_t m_start_pos = 0;
uint16_t k_start_pos = 0;
uint8_t pad_list[4] = {0, 0, 0, 0};
uint8_t stride_w = 1;
uint8_t stride_h = 1;
uint8_t filter_w = 1;
uint8_t filter_h = 1;
uint8_t dilation_filter_w = 1;
uint8_t dilation_filter_h = 1;
bool enable_filter_w_extend = false;
bool enable_filter_h_extend = false;
bool enable_transpose = false;
bool enable_f_matrix_ctrl = false;
PadType pad_value = 0;
};
struct copy_gm_to_l1 {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
private:
template <const gm_to_l1_trait& trait, typename DstTensor, typename SrcTensor>
__aicore__ inline static void data_copy_impl(const DstTensor& dst, const SrcTensor& src);
template <
const gm_to_l1_trait& trait, typename DstTensor, typename SrcTensor, typename DstCoord, typename SrcCoord,
typename CopyShape>
__aicore__ inline static void data_copy_impl(
const DstTensor& dst, const SrcTensor& src, const DstCoord& dst_coord, const SrcCoord& src_coord,
const CopyShape& copy_shape);
};
struct copy_l1_to_ub {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
private:
template <const l1_to_ub_trait& trait, typename DstTensor, typename SrcTensor>
__aicore__ inline static void data_copy_impl(const DstTensor& dst, const SrcTensor& src);
template <
const l1_to_ub_trait& trait, typename DstTensor, typename SrcTensor, typename DstCoord, typename SrcCoord,
typename CopyShape>
__aicore__ inline static void data_copy_impl(
const DstTensor& dst, const SrcTensor& src, const DstCoord& dst_coord, const SrcCoord& src_coord,
const CopyShape& copy_shape);
};
struct copy_l1_to_l0a {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
};
struct copy_l1_to_l0b {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
private:
template <const l1_to_l0b_trait& trait, typename DstTensor, typename SrcTensor>
__aicore__ inline static void load_data(const DstTensor& dst, const SrcTensor& src);
template <
const l1_to_l0b_trait& trait, typename DstTensor, typename SrcTensor, typename DstCoord, typename SrcCoord,
typename CopyShape>
__aicore__ inline static void load_data(
const DstTensor& dst, const SrcTensor& src, const DstCoord& dst_coord, const SrcCoord& src_coord,
const CopyShape& copy_shape);
};
struct copy_l1_to_l0scalea {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
private:
template <const l1_to_l0scalea_trait& trait, typename DstTensor, typename SrcTensor>
__aicore__ inline static void load_data(const DstTensor& dst, const SrcTensor& src);
template <
const l1_to_l0scalea_trait& trait, typename DstTensor, typename SrcTensor, typename DstCoord, typename SrcCoord,
typename CopyShape>
__aicore__ inline static void load_data(
const DstTensor& dst, const SrcTensor& src, const DstCoord& dst_coord, const SrcCoord& src_coord,
const CopyShape& copy_shape);
};
struct copy_l1_to_l0scaleb {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
private:
template <const l1_to_l0scaleb_trait& trait, typename DstTensor, typename SrcTensor>
__aicore__ inline static void load_data(const DstTensor& dst, const SrcTensor& src);
template <
const l1_to_l0scaleb_trait& trait, typename DstTensor, typename SrcTensor, typename DstCoord, typename SrcCoord,
typename CopyShape>
__aicore__ inline static void load_data(
const DstTensor& dst, const SrcTensor& src, const DstCoord& dst_coord, const SrcCoord& src_coord,
const CopyShape& copy_shape);
};
struct copy_l1_to_fixbuf {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
private:
template <const l1_to_fixbuf_trait& trait, typename DstTensor, typename SrcTensor>
__aicore__ inline static void data_copy_impl(const DstTensor& dst, const SrcTensor& src);
template <
const l1_to_fixbuf_trait& trait, typename DstTensor, typename SrcTensor, typename DstCoord, typename SrcCoord,
typename CopyShape>
__aicore__ inline static void data_copy_impl(
const DstTensor& dst, const SrcTensor& src, const DstCoord& dst_coord, const SrcCoord& src_coord,
const CopyShape& copy_shape);
};
struct copy_l1_to_biastable {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
private:
template <const l1_to_biastable_trait& trait, typename DstTensor, typename SrcTensor>
__aicore__ inline static void data_copy_impl(const DstTensor& dst, const SrcTensor& src);
template <
const l1_to_biastable_trait& trait, typename DstTensor, typename SrcTensor, typename DstCoord,
typename SrcCoord, typename CopyShape>
__aicore__ inline static void data_copy_impl(
const DstTensor& dst, const SrcTensor& src, const DstCoord& dst_coord, const SrcCoord& src_coord,
const CopyShape& copy_shape);
};
struct copy_l0c_to_gm {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
};
struct copy_l0c_to_ub {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
};
struct copy_l0c_to_l1 {
template <typename Trait, const Trait& trait, typename... Args>
__aicore__ inline static void copy(const Args&... args);
};
}
}
#include "impl/tensor_api/arch/cube/gm_to_l1/copy.h"
#include "impl/tensor_api/arch/cube/l1_to_ub/copy.h"
#include "impl/tensor_api/arch/cube/l1_to_l0a/copy.h"
#include "impl/tensor_api/arch/cube/l1_to_l0b/copy.h"
#include "impl/tensor_api/arch/cube/l1_to_l0scalea/copy.h"
#include "impl/tensor_api/arch/cube/l1_to_l0scaleb/copy.h"
#include "impl/tensor_api/arch/cube/l1_to_fb/copy.h"
#include "impl/tensor_api/arch/cube/l1_to_bt/copy.h"
#include "impl/tensor_api/arch/cube/l0c_to_gm/copy.h"
#include "impl/tensor_api/arch/cube/l0c_to_ub/copy.h"
#include "impl/tensor_api/arch/cube/l0c_to_l1/copy.h"
#endif
#if defined(UNDEF_ASCENDC_TENSOR_API_INCLUDE_COMPILER_INTERNAL_HEADERS_ASCENDC_TENSOR_API_H)
#undef ASCENDC_TENSOR_API_INCLUDE_COMPILER_INTERNAL_HEADERS
#undef UNDEF_ASCENDC_TENSOR_API_INCLUDE_COMPILER_INTERNAL_HEADERS_ASCENDC_TENSOR_API_H
#endif
