* 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 select_v2_tiling.cpp
* \brief
*/
#include "log/log.h"
#include "util/math_util.h"
#include "tiling/platform/platform_ascendc.h"
#include "register/op_impl_registry.h"
#include <graph/utils/type_utils.h>
#include "../op_kernel/select_v2_tiling_data.h"
#include "../op_kernel/select_v2_tiling_key.h"
namespace optiling {
#define BLOCK_SIZE 32U
#define UB_DATA_NUM_UINT8 16U
#define UB_DATA_NUM_OTHER 8U
constexpr uint32_t BUFFER_NUM = 2;
constexpr uint32_t WS_SYS_SIZE = 0;
constexpr uint32_t RESERVED_UB_SIZE = 8 * 1024;
struct SelectV2CompileInfo {};
static ge::graphStatus TilingParseForSelectV2([[maybe_unused]] gert::TilingParseContext* context)
{
OP_CHECK_IF(context == nullptr, OP_LOGE(context, "context is nullptr"), return ge::GRAPH_FAILED);
return ge::GRAPH_SUCCESS;
}
static ge::graphStatus GetPlatformInfo(gert::TilingContext* context, uint64_t& ubSize, int64_t& coreNum)
{
OP_CHECK_IF(context == nullptr, OP_LOGE(context, "context is nullptr"), return ge::GRAPH_FAILED);
auto ascendcPlatform = platform_ascendc::PlatformAscendC(context->GetPlatformInfo());
ascendcPlatform.GetCoreMemSize(platform_ascendc::CoreMemType::UB, ubSize);
coreNum = ascendcPlatform.GetCoreNum();
OP_CHECK_IF(coreNum <= 0, OP_LOGE(context, "coreNum is 0"), return ge::GRAPH_FAILED);
OP_CHECK_IF(ubSize <= 0, OP_LOGE(context, "ubSize is 0"), return ge::GRAPH_FAILED);
return ge::GRAPH_SUCCESS;
}
static ge::graphStatus GetWorkspaceSize(gert::TilingContext* context)
{
OP_CHECK_IF(context == nullptr, OP_LOGE(context, "context is nullptr"), return ge::GRAPH_FAILED);
size_t usrSize = 0;
auto ascendcPlatform = platform_ascendc::PlatformAscendC(context->GetPlatformInfo());
uint32_t sysWorkspaceSize = ascendcPlatform.GetLibApiWorkSpaceSize();
size_t* currentWorkspace = context->GetWorkspaceSizes(
1);
currentWorkspace[0] = usrSize + sysWorkspaceSize;
return ge::GRAPH_SUCCESS;
}
static ge::graphStatus GetShapeAttrsInfo(
gert::TilingContext* context, uint64_t ubSize, uint64_t& inputNum, uint64_t& inputBytes, uint64_t& tileBlockNum,
uint64_t& tileDataNum, uint64_t& inputLengthAlgin32)
{
OP_CHECK_IF(
context == nullptr || context->GetInputShape(0) == nullptr, OP_LOGE(context, "context is nullptr"),
return ge::GRAPH_FAILED);
inputNum = context->GetInputShape(0)->GetStorageShape().GetShapeSize();
uint32_t typeLength = 0;
ge::TypeUtils::GetDataTypeLength(context->GetInputDesc(1)->GetDataType(), typeLength);
uint64_t inputLength = inputNum * typeLength;
if (inputNum == 0) {
return ge::GRAPH_FAILED;
}
inputBytes = inputLength / inputNum;
uint64_t ubDataNumber =
(typeLength == sizeof(uint8_t)) ? UB_DATA_NUM_UINT8 : UB_DATA_NUM_OTHER;
tileBlockNum = (ubSize / BLOCK_SIZE) / ubDataNumber;
if (inputBytes == 0) {
return ge::GRAPH_FAILED;
}
tileDataNum = (tileBlockNum * BLOCK_SIZE) / inputBytes;
inputLengthAlgin32 = (((inputLength + BLOCK_SIZE - 1) / BLOCK_SIZE) * BLOCK_SIZE);
return ge::GRAPH_SUCCESS;
}
static ge::graphStatus CalculateCoreBlockNums(
uint64_t inputLengthAlgin32, int64_t coreNum, uint64_t tileBlockNum, uint64_t inputBytes, uint64_t tileDataNum,
uint64_t& smallCoreDataNum, uint64_t& bigCoreDataNum, uint64_t& smallTailDataNum, uint64_t& bigTailDataNum,
uint64_t& finalSmallTileNum, uint64_t& finalBigTileNum, uint64_t& tailBlockNum)
{
if (0 == BLOCK_SIZE || 0 == coreNum || 0 == tileBlockNum || 0 == inputBytes) {
return ge::GRAPH_FAILED;
}
uint64_t everyCoreInputBlockNum = inputLengthAlgin32 / BLOCK_SIZE / coreNum;
tailBlockNum = (inputLengthAlgin32 / BLOCK_SIZE) % coreNum;
smallCoreDataNum = everyCoreInputBlockNum * BLOCK_SIZE / inputBytes;
uint64_t smallTileNum = everyCoreInputBlockNum / tileBlockNum;
finalSmallTileNum = (everyCoreInputBlockNum % tileBlockNum) == 0 ? smallTileNum : smallTileNum + 1;
smallTailDataNum = smallCoreDataNum - (tileDataNum * smallTileNum);
smallTailDataNum = smallTailDataNum == 0 ? tileDataNum : smallTailDataNum;
everyCoreInputBlockNum += 1;
bigCoreDataNum = everyCoreInputBlockNum * BLOCK_SIZE / inputBytes;
uint64_t bigTileNum = everyCoreInputBlockNum / tileBlockNum;
finalBigTileNum = (everyCoreInputBlockNum % tileBlockNum) == 0 ? bigTileNum : bigTileNum + 1;
bigTailDataNum = bigCoreDataNum - tileDataNum * bigTileNum;
bigTailDataNum = bigTailDataNum == 0 ? tileDataNum : bigTailDataNum;
return ge::GRAPH_SUCCESS;
}
static ge::graphStatus SelectV2TilingFunc(gert::TilingContext* context)
{
SelectV2TilingData* tiling = context->GetTilingData<SelectV2TilingData>();
OP_CHECK_NULL_WITH_CONTEXT(context, tiling);
OP_CHECK_IF(
memset_s(tiling, sizeof(SelectV2TilingData), 0, sizeof(SelectV2TilingData)) != EOK,
OP_LOGE(context, "set tiling data error"), return ge::GRAPH_FAILED);
uint64_t ubSize;
int64_t coreNum;
ge::graphStatus ret = GetPlatformInfo(context, ubSize, coreNum);
OP_CHECK_IF(ret != ge::GRAPH_SUCCESS, OP_LOGE(context, "GetPlatformInfo error"), return ge::GRAPH_FAILED);
ubSize = ubSize - RESERVED_UB_SIZE;
uint64_t inputNum, inputBytes, tileBlockNum, tileDataNum, inputLengthAlgin32;
ret = GetShapeAttrsInfo(context, ubSize, inputNum, inputBytes, tileBlockNum, tileDataNum, inputLengthAlgin32);
OP_CHECK_IF(ret != ge::GRAPH_SUCCESS, OP_LOGE(context, "GetShapeAttrsInfo error"), return ge::GRAPH_FAILED);
if (tileDataNum >= inputNum) {
coreNum = 1;
} else {
coreNum = (static_cast<uint64_t>(coreNum) < inputLengthAlgin32 / BLOCK_SIZE) ? coreNum :
inputLengthAlgin32 / BLOCK_SIZE;
}
uint64_t smallCoreDataNum, bigCoreDataNum, smallTailDataNum, bigTailDataNum;
uint64_t finalSmallTileNum, finalBigTileNum, tailBlockNum;
ret = CalculateCoreBlockNums(
inputLengthAlgin32, coreNum, tileBlockNum, inputBytes, tileDataNum, smallCoreDataNum, bigCoreDataNum,
smallTailDataNum, bigTailDataNum, finalSmallTileNum, finalBigTileNum, tailBlockNum);
OP_CHECK_IF(ret != ge::GRAPH_SUCCESS, OP_LOGE(context, "CalculateCoreBlockNums error"), return ge::GRAPH_FAILED);
tiling->smallCoreDataNum = smallCoreDataNum;
tiling->bigCoreDataNum = bigCoreDataNum;
tiling->tileDataNum = tileDataNum;
tiling->smallTailDataNum = smallTailDataNum;
tiling->bigTailDataNum = bigTailDataNum;
tiling->finalSmallTileNum = finalSmallTileNum;
tiling->finalBigTileNum = finalBigTileNum;
tiling->tailBlockNum = tailBlockNum;
OP_CHECK_IF(
GetWorkspaceSize(context) != ge::GRAPH_SUCCESS, OP_LOGE(context, "GetWorkspaceSize error"),
return ge::GRAPH_FAILED);
uint64_t tilingKey = 0;
tilingKey = GET_TPL_TILING_KEY(ELEMENTWISE_TPL_SCH_MODE_0);
context->SetTilingKey(tilingKey);
context->SetBlockDim(coreNum);
return ge::GRAPH_SUCCESS;
}
IMPL_OP_OPTILING(SelectV2).Tiling(SelectV2TilingFunc).TilingParse<SelectV2CompileInfo>(TilingParseForSelectV2);
}