* 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 floor.h
* \brief
*/
#ifndef FLOOR_H
#define FLOOR_H
#include "kernel_operator.h"
#include "kernel_tiling/kernel_tiling.h"
#include "floor_tiling_data.h"
#include "floor_tiling_key.h"
namespace MyFloor {
using namespace AscendC;
constexpr int32_t BUFFER_NUM = 2;
template <typename TYPE_X>
class KernelFloor {
public:
__aicore__ inline KernelFloor() {}
__aicore__ inline void Init(GM_ADDR x, GM_ADDR y, const FloorTilingData& tilingData);
__aicore__ inline void Process();
private:
__aicore__ inline void CopyIn(int32_t progress);
__aicore__ inline void Compute(int32_t progress);
__aicore__ inline void CopyOut(int32_t progress);
private:
AscendC::TPipe pipe;
AscendC::TQue<AscendC::TPosition::VECIN, BUFFER_NUM> inQueueX;
AscendC::TQue<AscendC::TPosition::VECOUT, BUFFER_NUM> outQueueY;
AscendC::TBuf<AscendC::TPosition::VECCALC> tmpBuf;
AscendC::GlobalTensor<TYPE_X> xGm;
AscendC::GlobalTensor<TYPE_X> yGm;
uint64_t tileNum;
uint64_t tileDataNum;
uint64_t tailDataNum;
uint64_t processDataNum;
};
template <typename TYPE_X>
__aicore__ inline void MyFloor::KernelFloor<TYPE_X>::Init(GM_ADDR x, GM_ADDR y, const FloorTilingData& tilingData)
{
ASSERT(AscendC::GetBlockNum() != 0 && "block dim can not be zero!");
uint64_t coreId = AscendC::GetBlockIdx();
uint64_t globalBufferIndex = tilingData.bigCoreDataNum * coreId;
this->tileDataNum = tilingData.tileDataNum;
uint64_t coreDataNum;
if (coreId < tilingData.tailBlockNum) {
coreDataNum = tilingData.bigCoreDataNum;
this->tileNum = tilingData.finalBigTileNum;
this->tailDataNum = tilingData.bigTailDataNum;
} else {
coreDataNum = tilingData.smallCoreDataNum;
this->tileNum = tilingData.finalSmallTileNum;
this->tailDataNum = tilingData.smallTailDataNum;
globalBufferIndex -= (tilingData.bigCoreDataNum - tilingData.smallCoreDataNum) * (coreId - tilingData.tailBlockNum);
}
xGm.SetGlobalBuffer((__gm__ TYPE_X*)x + globalBufferIndex, coreDataNum);
yGm.SetGlobalBuffer((__gm__ TYPE_X*)y + globalBufferIndex, coreDataNum);
pipe.InitBuffer(inQueueX, BUFFER_NUM, this->tileDataNum * sizeof(TYPE_X));
pipe.InitBuffer(outQueueY, BUFFER_NUM, this->tileDataNum * sizeof(TYPE_X));
if constexpr (!std::is_same_v<TYPE_X, float32_t>) {
pipe.InitBuffer(tmpBuf, this->tileDataNum * sizeof(float));
}
}
template <typename TYPE_X>
__aicore__ inline void MyFloor::KernelFloor<TYPE_X>::CopyIn(int32_t progress)
{
AscendC::LocalTensor<TYPE_X> xLocal = inQueueX.AllocTensor<TYPE_X>();
AscendC::DataCopy(xLocal, xGm[progress * this->tileDataNum], this->processDataNum);
inQueueX.EnQue(xLocal);
}
template <typename TYPE_X>
__aicore__ inline void MyFloor::KernelFloor<TYPE_X>::Compute(int32_t progress)
{
AscendC::LocalTensor<TYPE_X> xLocal = inQueueX.DeQue<TYPE_X>();
AscendC::LocalTensor<TYPE_X> yLocal = outQueueY.AllocTensor<TYPE_X>();
if constexpr (std::is_same_v<TYPE_X, float>) {
AscendC::Floor(yLocal, xLocal, this->processDataNum);
} else {
AscendC::LocalTensor<float> tmpFloat = tmpBuf.Get<float>();
AscendC::Cast(tmpFloat, xLocal, AscendC::RoundMode::CAST_NONE, this->processDataNum);
AscendC::Floor(tmpFloat, tmpFloat, this->processDataNum);
AscendC::Cast(yLocal, tmpFloat, AscendC::RoundMode::CAST_FLOOR, this->processDataNum);
}
outQueueY.EnQue(yLocal);
inQueueX.FreeTensor(xLocal);
}
template <typename TYPE_X>
__aicore__ inline void MyFloor::KernelFloor<TYPE_X>::CopyOut(int32_t progress)
{
AscendC::LocalTensor<TYPE_X> yLocal = outQueueY.DeQue<TYPE_X>();
if (progress == this->tileNum - 1) {
AscendC::DataCopyExtParams copyParams{1, static_cast<uint32_t>(this->tailDataNum * sizeof(TYPE_X)), 0, 0, 0};
AscendC::DataCopyPad(yGm[progress * this->tileDataNum], yLocal, copyParams);
} else {
AscendC::DataCopy(yGm[progress * this->tileDataNum], yLocal, this->processDataNum);
}
outQueueY.FreeTensor(yLocal);
}
template <typename TYPE_X>
__aicore__ inline void MyFloor::KernelFloor<TYPE_X>::Process()
{
int32_t loopCount = this->tileNum;
this->processDataNum = this->tileDataNum;
for (int32_t i = 0; i < loopCount - 1; i++) {
CopyIn(i);
Compute(i);
CopyOut(i);
}
this->processDataNum = this->tailDataNum;
CopyIn(loopCount - 1);
Compute(loopCount - 1);
CopyOut(loopCount - 1);
}
}
#endif