* 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_tiling/kernel_tiling.h"
#include "kernel_operator.h"
#include "arch35/kth_value_axis_one_copy.h"
#include "arch35/kth_value_merge_intra_core.h"
#include "arch35/kth_value_merge_sort_one_core.h"
#include "arch35/kth_value_merge_sort_more_core.h"
#include "arch35/kth_value_non_last_small_axis.h"
#include "arch35/kth_value_radix_more_core.h"
#include "arch35/kth_value_radix_one_core.h"
#include "arch35/kth_value_small_axis_insertion.h"
#include "arch35/kth_value_small_axis_two_stage.h"
#include "arch35/kth_value_tiling_data.h"
#include "arch35/kth_value_tiling_key.h"
using namespace AscendC;
template <uint64_t schId>
__aicore__ inline void RunMergeSortRoute(GM_ADDR x, GM_ADDR y1, GM_ADDR y2, KthValueTilingData* tilingData, TPipe* pipe)
{
constexpr uint64_t isSort32SmallAxis = (schId == KTH_VALUE_SCHID_SORT32_SMALL_AXIS);
if constexpr (IsSameType<bfloat16_t, DTYPE_X>::value) {
KthValue::KthValueMergeSortOneCore<DTYPE_X, float, isSort32SmallAxis> op;
op.Init(x, y1, y2, tilingData, pipe);
op.Process();
} else if constexpr (IsSameType<float, DTYPE_X>::value || IsSameType<half, DTYPE_X>::value) {
KthValue::KthValueMergeSortOneCore<DTYPE_X, DTYPE_X, isSort32SmallAxis> op;
op.Init(x, y1, y2, tilingData, pipe);
op.Process();
}
}
template <uint64_t isInt32>
__aicore__ inline void RunRadixMoreCoreRoute(
GM_ADDR x, GM_ADDR y1, GM_ADDR y2, GM_ADDR workspace, KthValueTilingData* tilingData, TPipe* pipe)
{
if constexpr (sizeof(DTYPE_X) == 1) {
if constexpr (isInt32 == 1) {
KthValue::KthValueRadixMoreCore<DTYPE_X, uint32_t, uint8_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
} else {
KthValue::KthValueRadixMoreCore<DTYPE_X, int64_t, uint8_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
}
if constexpr (sizeof(DTYPE_X) == 2) {
if constexpr (isInt32 == 1) {
KthValue::KthValueRadixMoreCore<DTYPE_X, uint32_t, uint16_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
} else {
KthValue::KthValueRadixMoreCore<DTYPE_X, int64_t, uint16_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
}
if constexpr (sizeof(DTYPE_X) == 4) {
if constexpr (isInt32 == 1) {
KthValue::KthValueRadixMoreCore<DTYPE_X, uint32_t, uint32_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
} else {
KthValue::KthValueRadixMoreCore<DTYPE_X, int64_t, uint32_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
}
if constexpr (sizeof(DTYPE_X) == 8) {
if constexpr (isInt32 == 1) {
KthValue::KthValueRadixMoreCore<DTYPE_X, uint32_t, uint64_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
} else {
KthValue::KthValueRadixMoreCore<DTYPE_X, int64_t, uint64_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
}
}
__aicore__ inline void RunSmallAxisInsertionRoute(
GM_ADDR x, GM_ADDR y1, GM_ADDR y2, KthValueTilingData* tilingData, TPipe* pipe)
{
if constexpr (IsSameType<bfloat16_t, DTYPE_X>::value) {
KthValue::KthValueSmallAxisInsertion<DTYPE_X, float> op;
op.Init(x, y1, y2, tilingData, pipe);
op.Process();
} else {
KthValue::KthValueSmallAxisInsertion<DTYPE_X, DTYPE_X> op;
op.Init(x, y1, y2, tilingData, pipe);
op.Process();
}
}
__aicore__ inline void RunMergeMoreCoreRoute(
GM_ADDR x, GM_ADDR y1, GM_ADDR y2, GM_ADDR workspace, KthValueTilingData* tilingData, TPipe* pipe)
{
if constexpr (IsSameType<float, DTYPE_X>::value) {
KthValue::KthValueMergeSortMoreCore<DTYPE_X, DTYPE_X, false, int64_t> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
}
__aicore__ inline void RunMergeIntraCoreRoute(
GM_ADDR x, GM_ADDR y1, GM_ADDR y2, GM_ADDR workspace, KthValueTilingData* tilingData, TPipe* pipe)
{
if constexpr (IsSameType<float, DTYPE_X>::value) {
KthValue::KthValueMergeIntraCore<DTYPE_X, int64_t, false> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
}
template <bool useMergeSort>
__aicore__ inline void RunNonLastSmallAxisRoute(
GM_ADDR x, GM_ADDR y1, GM_ADDR y2, GM_ADDR workspace, KthValueTilingData* tilingData, TPipe* pipe)
{
if constexpr (useMergeSort) {
if constexpr (
IsSameType<DTYPE_X, float>::value || IsSameType<DTYPE_X, half>::value ||
IsSameType<DTYPE_X, bfloat16_t>::value) {
KthValue::KthValueNonLastSmallAxis<DTYPE_X, false, true> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
} else {
KthValue::KthValueNonLastSmallAxis<DTYPE_X, false, false> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
}
__aicore__ inline void RunAxisOneCopyRoute(
GM_ADDR x, GM_ADDR y1, GM_ADDR y2, GM_ADDR workspace, KthValueTilingData* tilingData, TPipe* pipe)
{
KthValue::KthValueAxisOneCopy<DTYPE_X> op;
op.Init(x, y1, y2, workspace, tilingData, pipe);
op.Process();
}
__aicore__ inline void RunRadixOneCoreRoute(
GM_ADDR x, GM_ADDR y1, GM_ADDR y2, KthValueTilingData* tilingData, TPipe* pipe)
{
KthValue::KthValueRadixOneCore<DTYPE_X> op;
op.Init(x, y1, y2, tilingData, pipe);
op.Process();
}
__aicore__ inline void RunSmallAxisTwoStageRoute(
GM_ADDR x, GM_ADDR y1, GM_ADDR y2, KthValueTilingData* tilingData, TPipe* pipe)
{
KthValue::KthValueSmallAxisTwoStage<DTYPE_X> op;
op.Init(x, y1, y2, tilingData, pipe);
op.Process();
}
template <uint64_t schId, uint64_t isInt32>
__global__ __aicore__ void kth_value(GM_ADDR x, GM_ADDR y1, GM_ADDR y2, GM_ADDR workspace, GM_ADDR tiling)
{
KERNEL_TASK_TYPE_DEFAULT(KERNEL_TYPE_MIX_AIV_1_0);
REGISTER_TILING_DEFAULT(KthValueTilingData);
GET_TILING_DATA_WITH_STRUCT(KthValueTilingData, tilingData, tiling);
GM_ADDR usrWorkspace = AscendC::GetUserWorkspace(workspace);
TPipe pipe;
if constexpr (schId == KTH_VALUE_SCHID_AXIS_ONE_COPY) {
RunAxisOneCopyRoute(x, y1, y2, usrWorkspace, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_MERGE_SORT || schId == KTH_VALUE_SCHID_SORT32_SMALL_AXIS) {
RunMergeSortRoute<schId>(x, y1, y2, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_RADIX_ONE_CORE) {
RunRadixOneCoreRoute(x, y1, y2, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_RADIX_MORE_CORE) {
RunRadixMoreCoreRoute<isInt32>(x, y1, y2, usrWorkspace, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_SMALL_AXIS_INSERTION) {
RunSmallAxisInsertionRoute(x, y1, y2, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_SMALL_AXIS_TWO_STAGE) {
RunSmallAxisTwoStageRoute(x, y1, y2, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_MERGE_MORE_CORE) {
RunMergeMoreCoreRoute(x, y1, y2, usrWorkspace, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_MERGE_INTRA_CORE) {
RunMergeIntraCoreRoute(x, y1, y2, usrWorkspace, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_NON_LAST_SMALL_AXIS) {
RunNonLastSmallAxisRoute<true>(x, y1, y2, usrWorkspace, &tilingData, &pipe);
return;
}
if constexpr (schId == KTH_VALUE_SCHID_NON_LAST_SMALL_AXIS_RADIX) {
RunNonLastSmallAxisRoute<false>(x, y1, y2, usrWorkspace, &tilingData, &pipe);
return;
}
}