* 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.
*/
#include <gtest/gtest.h>
#define private public
#define protected public
#include "kernel_operator.h"
using namespace std;
using namespace AscendC;
namespace AscendC {
template <typename dataType, bool isReuseSrc = false, uint8_t expandLevel = 10>
class KernelExpHighPrecision {
public:
__aicore__ inline KernelExpHighPrecision() {}
__aicore__ inline void Init(GM_ADDR inputXGm, GM_ADDR outputGm, uint32_t totalLength, uint32_t calCount,
uint32_t mode)
{
uint32_t oneBlockNum = 32 / sizeof(dataType);
totalLength = (totalLength + oneBlockNum - 1) / oneBlockNum * oneBlockNum;
this->totalLength = totalLength;
this->calCount = calCount;
this->mode = mode;
this->dataFormat = DataFormat::ND;
inputXGlobal.SetGlobalBuffer(reinterpret_cast<__gm__ dataType*>(inputXGm), totalLength);
outputGlobal.SetGlobalBuffer(reinterpret_cast<__gm__ dataType*>(outputGm), totalLength);
pipe.InitBuffer(inQueueX, 1, sizeof(dataType) * totalLength);
pipe.InitBuffer(outQueue, 1, sizeof(dataType) * totalLength);
}
__aicore__ inline void Process()
{
CopyIn();
Compute();
CopyOut();
}
private:
__aicore__ inline void CopyIn()
{
LocalTensor<dataType> inputXLocal = inQueueX.AllocTensor<dataType>();
DataCopy(inputXLocal, inputXGlobal, totalLength);
inQueueX.EnQue(inputXLocal);
}
__aicore__ inline void Compute()
{
LocalTensor<dataType> inputXLocal = inQueueX.DeQue<dataType>();
LocalTensor<dataType> outputLocal = outQueue.AllocTensor<dataType>();
LocalTensor<uint8_t> stackBuffer;
bool ans = PopStackBuffer<uint8_t, TPosition::LCM>(stackBuffer);
stackBufferSize = stackBuffer.GetSize();
if (mode == 0) {
Exp<dataType, expandLevel, isReuseSrc>(outputLocal, inputXLocal, calCount);
} else {
LocalTensor<uint8_t> stackBuffer;
bool ans = PopStackBuffer<uint8_t, TPosition::LCM>(stackBuffer);
stackBuffer.SetSize(stackBufferSize);
Exp<dataType, expandLevel, isReuseSrc>(outputLocal, inputXLocal, stackBuffer, calCount);
}
outQueue.EnQue<dataType>(outputLocal);
inQueueX.FreeTensor(inputXLocal);
}
__aicore__ inline void CopyOut()
{
LocalTensor<dataType> outputLocal = outQueue.DeQue<dataType>();
DataCopy(outputGlobal, outputLocal, totalLength);
outQueue.FreeTensor(outputLocal);
}
private:
GlobalTensor<dataType> inputXGlobal;
GlobalTensor<dataType> outputGlobal;
TPipe pipe;
TQue<TPosition::VECIN, 1> inQueueX;
TQue<TPosition::VECOUT, 1> outQueue;
uint32_t totalLength;
uint32_t calCount;
uint32_t mode;
uint32_t stackBufferSize = 0;
DataFormat dataFormat;
};
}
template <typename dataType, bool isReuseSrc = false, uint8_t expandLevel = 10>
__global__ __aicore__ void kernel_exphighprecision_operator(GM_ADDR inputXGm, GM_ADDR outputGm, uint32_t totalLength,
uint32_t calCount, uint32_t mode)
{
AscendC::KernelExpHighPrecision<dataType, isReuseSrc, expandLevel> op;
op.Init(inputXGm, outputGm, totalLength, calCount, mode);
op.Process();
}
struct ExpHighPrecisionTestParams {
uint32_t totalLength;
uint32_t calCount;
uint32_t mode;
uint32_t typeSize;
void (*calFunc) (uint8_t*, uint8_t*, uint32_t, uint32_t, uint32_t);
};
class ExpHighPrecisionTestsuite : public testing::Test, public testing::WithParamInterface<ExpHighPrecisionTestParams> {
protected:
void SetUp()
{
AscendC::SetGCoreType(2);
}
void TearDown()
{
AscendC::SetGCoreType(0);
}
};
INSTANTIATE_TEST_CASE_P(TEST_OPEARATION_EXPHIGHPRECISION, ExpHighPrecisionTestsuite,
::testing::Values(
ExpHighPrecisionTestParams {1024, 1024, 0, sizeof(half), kernel_exphighprecision_operator<half, true, 10> },
ExpHighPrecisionTestParams {1024, 1000, 0, sizeof(half), kernel_exphighprecision_operator<half, false, 10>},
ExpHighPrecisionTestParams {1024, 1024, 0, sizeof(float), kernel_exphighprecision_operator<float, true, 10> },
ExpHighPrecisionTestParams {1024, 1000, 0, sizeof(float), kernel_exphighprecision_operator<float, false, 10>},
ExpHighPrecisionTestParams {1024, 1024, 1, sizeof(half), kernel_exphighprecision_operator<half, true, 10> }
));
TEST_P(ExpHighPrecisionTestsuite, ExpHighPrecisionOpTestCase)
{
auto param = GetParam();
uint32_t totalLength = param.totalLength;
uint32_t calCount = param.calCount;
uint32_t mode = param.mode;
uint32_t typeSize = param.typeSize;
uint32_t oneBlockNum = 32 / typeSize;
totalLength = (totalLength + oneBlockNum - 1) / oneBlockNum * oneBlockNum;
uint8_t inputXGm[totalLength * typeSize] { 0x00 };
uint8_t outputGm[totalLength * typeSize] { 0x00 };
param.calFunc(inputXGm, outputGm, totalLength, calCount, mode);
}