* 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.
*/
#include "kernel_operator.h"
#include "arch35/truncate_div_dag.h"
#include "arch35/truncate_div_struct.h"
#include "atvoss/broadcast/broadcast_sch.h"
using namespace AscendC;
using namespace Ops::Base;
template <uint64_t schMode, bool canUseMul>
__aicore__ inline void truncate_div_half(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR tiling)
{
if constexpr (std::is_same<DTYPE_X2, float>::value) {
if constexpr (canUseMul) {
using OpDag = TruncateDivOp::TruncateDivFloatWithCastScalar<half, float, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, y);
} else {
using OpDag = TruncateDivOp::TruncateDivFloatWithCast<half, float, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
} else if constexpr (std::is_same<DTYPE_X2, half>::value) {
if constexpr (canUseMul) {
using OpDag = TruncateDivOp::TruncateDivFloat16Scalar<half, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, y);
} else {
using OpDag = TruncateDivOp::TruncateDivFloat16<half, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
}
}
template <uint64_t schMode, bool canUseMul>
__aicore__ inline void truncate_div_bfloat16(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR tiling)
{
if constexpr (std::is_same<DTYPE_X2, bfloat16_t>::value) {
if constexpr (canUseMul) {
using OpDag = TruncateDivOp::TruncateDivFloat16Scalar<bfloat16_t, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, y);
} else {
using OpDag = TruncateDivOp::TruncateDivFloat16<bfloat16_t, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
}
}
template <uint64_t schMode, bool canUseMul>
__aicore__ inline void truncate_div_float(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR tiling)
{
if constexpr (std::is_same<DTYPE_X2, float>::value) {
if constexpr (canUseMul) {
using OpDag = TruncateDivOp::TruncateDivFloatScalar<float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, y);
} else {
using OpDag = TruncateDivOp::TruncateDivFloat<float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
} else if constexpr (std::is_same<DTYPE_X2, int32_t>::value) {
using OpDag = TruncateDivOp::TruncateDivFloatToLowBit<float, int32_t, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
} else if constexpr (std::is_same<DTYPE_X2, half>::value) {
if constexpr (canUseMul) {
using OpDag = TruncateDivOp::TruncateDivFloatScalar<float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, y);
} else {
using OpDag = TruncateDivOp::TruncateDivFloatToLowBit<float, half, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
}
}
template <uint64_t schMode, bool canUseMul>
__aicore__ inline void truncate_div_int8(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR tiling)
{
using OpDag = TruncateDivOp::TruncateDivIntS8<int8_t, half>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
template <uint64_t schMode, bool canUseMul>
__aicore__ inline void truncate_div_uint8(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR tiling)
{
using OpDag = TruncateDivOp::TruncateDivIntU8<uint8_t, uint16_t>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
template <uint64_t schMode, bool canUseMul>
__aicore__ inline void truncate_div_int16(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR tiling)
{
if constexpr (std::is_same<DTYPE_X2, int16_t>::value) {
using OpDag = TruncateDivOp::TruncateDivInt<int16_t>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
}
template <uint64_t schMode, bool canUseMul>
__aicore__ inline void truncate_div_int32(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR tiling)
{
if constexpr (std::is_same<DTYPE_X2, int32_t>::value) {
using OpDag = TruncateDivOp::TruncateDivInt<int32_t>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
} else if constexpr (std::is_same<DTYPE_X2, float>::value) {
using OpDag = TruncateDivOp::TruncateDivIntToFloat<int32_t, float, float>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
}
template <uint64_t schMode, bool canUseMul>
__aicore__ inline void truncate_div_int64(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR tiling)
{
using OpDag = TruncateDivOp::TruncateDivInt64<int64_t>::OpDag;
BroadcastSch<schMode, OpDag> sch(tiling);
sch.Process(x1, x2, y);
}
* Supported data type combinations (x1, x2, y):
* | x1 (DTYPE_X1) | x2 (DTYPE_X2) | y (DTYPE_Y) |
* |------------------|-----------------|-----------------|
* | half | float | float |
* | half | half | half |
* | bfloat16_t | bfloat16_t | bfloat16_t |
* | float | float | float |
* | float | int32_t | float |
* | float | half | float |
* | int8_t | int8_t | int8_t |
* | uint8_t | uint8_t | uint8_t |
* | int16_t | int16_t | int16_t |
* | int32_t | int32_t | int32_t |
* | int32_t | float | float |
* | int64_t | int64_t | int64_t |
*/
template <uint64_t schMode, bool canUseMul>
__global__ __aicore__ void truncate_div(GM_ADDR x1, GM_ADDR x2, GM_ADDR y, GM_ADDR workspace, GM_ADDR tiling)
{
if constexpr (std::is_same<DTYPE_X1, half>::value) {
truncate_div_half<schMode, canUseMul>(x1, x2, y, tiling);
} else if constexpr (std::is_same<DTYPE_X1, bfloat16_t>::value) {
truncate_div_bfloat16<schMode, canUseMul>(x1, x2, y, tiling);
} else if constexpr (std::is_same<DTYPE_X1, float>::value) {
truncate_div_float<schMode, canUseMul>(x1, x2, y, tiling);
} else if constexpr (std::is_same<DTYPE_X1, int8_t>::value) {
truncate_div_int8<schMode, canUseMul>(x1, x2, y, tiling);
} else if constexpr (std::is_same<DTYPE_X1, uint8_t>::value) {
truncate_div_uint8<schMode, canUseMul>(x1, x2, y, tiling);
} else if constexpr (std::is_same<DTYPE_X1, int16_t>::value) {
truncate_div_int16<schMode, canUseMul>(x1, x2, y, tiling);
} else if constexpr (std::is_same<DTYPE_X1, int32_t>::value) {
truncate_div_int32<schMode, canUseMul>(x1, x2, y, tiling);
} else if constexpr (std::is_same<DTYPE_X1, int64_t>::value) {
truncate_div_int64<schMode, canUseMul>(x1, x2, y, tiling);
}
}
