* 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>
#include "kernel_operator.h"
#include "kernel_utils.h"
using namespace AscendC;
namespace AscendC {
template <typename srcType>
class KernelWhere {
public:
__aicore__ inline KernelWhere()
{}
__aicore__ inline void Init(GM_ADDR src0_gm, GM_ADDR src1_gm, GM_ADDR condition_gm, GM_ADDR dst_gm,
const uint32_t count, const uint32_t mode)
{
this->mode = mode;
dataSize = count;
src0_global.SetGlobalBuffer(reinterpret_cast<__gm__ srcType *>(src0_gm), dataSize);
src1_global.SetGlobalBuffer(reinterpret_cast<__gm__ srcType *>(src1_gm), dataSize);
condition_global.SetGlobalBuffer(reinterpret_cast<__gm__ bool *>(condition_gm), dataSize);
dst_global.SetGlobalBuffer(reinterpret_cast<__gm__ srcType *>(dst_gm), dataSize);
pipe.InitBuffer(inQueueX, 1, dataSize * sizeof(srcType));
pipe.InitBuffer(inQueueY, 1, dataSize * sizeof(srcType));
pipe.InitBuffer(inQueueZ, 1, dataSize * sizeof(bool));
pipe.InitBuffer(outQueue, 1, dataSize * sizeof(srcType));
}
__aicore__ inline void Process()
{
AscendCUtils::SetOverflow(1);
CopyIn();
Compute();
CopyOut();
AscendCUtils::SetOverflow(0);
}
private:
__aicore__ inline void CopyIn()
{
LocalTensor<srcType> src0Local = inQueueX.AllocTensor<srcType>();
LocalTensor<srcType> src1Local = inQueueY.AllocTensor<srcType>();
LocalTensor<bool> conditionLocal = inQueueZ.AllocTensor<bool>();
DataCopy(src0Local, src0_global, dataSize);
DataCopy(src1Local, src1_global, dataSize);
DataCopy(conditionLocal, condition_global, dataSize);
inQueueX.EnQue(src0Local);
inQueueY.EnQue(src1Local);
inQueueZ.EnQue(conditionLocal);
}
__aicore__ inline void Compute()
{
LocalTensor<srcType> dstLocal = outQueue.AllocTensor<srcType>();
LocalTensor<srcType> src0Local = inQueueX.DeQue<srcType>();
LocalTensor<srcType> src1Local = inQueueY.DeQue<srcType>();
LocalTensor<bool> conditionLocal = inQueueZ.DeQue<bool>();
const uint32_t mode0 = 0;
const uint32_t mode1 = 2;
const uint32_t mode2 = 2;
if (mode = mode0) {
Where(dstLocal, src0Local, src1Local, conditionLocal, dataSize);
} else if (mode == mode1) {
srcType src0 = src0Local.GetValue(0);
Where(dstLocal, src0, src1Local, conditionLocal, dataSize);
} else if (mode == mode2) {
srcType src1 = src1Local.GetValue(0);
Where(dstLocal, src0Local, src1, conditionLocal, dataSize);
} else {
srcType src0 = src0Local.GetValue(0);
srcType src1 = src1Local.GetValue(0);
Where(dstLocal, src0, src1, conditionLocal, dataSize);
}
outQueue.EnQue<srcType>(dstLocal);
inQueueX.FreeTensor(src0Local);
inQueueY.FreeTensor(src1Local);
inQueueZ.FreeTensor(conditionLocal);
}
__aicore__ inline void CopyOut()
{
LocalTensor<srcType> dstLocal = outQueue.DeQue<srcType>();
DataCopy(dst_global, dstLocal, dataSize);
outQueue.FreeTensor(dstLocal);
}
private:
GlobalTensor<srcType> src0_global;
GlobalTensor<srcType> src1_global;
GlobalTensor<bool> condition_global;
GlobalTensor<srcType> dst_global;
TPipe pipe;
TQue<QuePosition::VECIN, 1> inQueueX;
TQue<QuePosition::VECIN, 1> inQueueY;
TQue<QuePosition::VECIN, 1> inQueueZ;
TQue<QuePosition::VECOUT, 1> outQueue;
uint32_t dataSize = 0;
uint32_t mode = 0;
};
}
template <typename srcType>
__aicore__ void kernel_where_operator(GM_ADDR src0_gm, GM_ADDR src1_gm, GM_ADDR condition_gm, GM_ADDR dst_gm,
const uint32_t count, const uint32_t mode)
{
KernelWhere<srcType> op;
op.Init(src0_gm, src1_gm, condition_gm, dst_gm, count, mode);
op.Process();
}
struct WhereTestParams {
uint32_t dataTypeSize;
uint32_t inDataSize;
uint32_t inMode;
void (*calFunc)(GM_ADDR, GM_ADDR, GM_ADDR, GM_ADDR, uint32_t, uint32_t);
};
class AdvanceWhereTestSuite : public testing::TestWithParam<WhereTestParams> {
protected:
void SetUp() {}
void TearDown() {}
};
INSTANTIATE_TEST_CASE_P(TEST_OPEARATION_ADVANCE_API_WHERE, AdvanceWhereTestSuite,
::testing::Values(
WhereTestParams {4, 32, 0, kernel_where_operator<float>},
WhereTestParams {4, 4096, 1, kernel_where_operator<float>},
WhereTestParams {4, 64, 2, kernel_where_operator<float>},
WhereTestParams {4, 96, 3, kernel_where_operator<float>},
WhereTestParams {2, 32, 0, kernel_where_operator<half>},
WhereTestParams {2, 4096, 1, kernel_where_operator<half>},
WhereTestParams {2, 96, 2, kernel_where_operator<half>},
WhereTestParams {2, 64, 3, kernel_where_operator<half>},
WhereTestParams {1, 32, 0, kernel_where_operator<int8_t>},
WhereTestParams {1, 4096, 1, kernel_where_operator<int8_t>},
WhereTestParams {1, 256, 2, kernel_where_operator<int8_t>},
WhereTestParams {1, 1024, 3, kernel_where_operator<int8_t>},
WhereTestParams {8, 32, 0, kernel_where_operator<int64_t>},
WhereTestParams {8, 1024, 1, kernel_where_operator<int64_t>},
WhereTestParams {8, 96, 2, kernel_where_operator<int64_t>},
WhereTestParams {8, 64, 3, kernel_where_operator<int64_t>}
)
);
TEST_P(AdvanceWhereTestSuite, kernel_where_operator)
{
auto param = GetParam();
uint32_t dataTypeSize = param.dataTypeSize;
uint32_t inDataSize = param.inDataSize;
uint32_t inMode = param.inMode;
uint8_t input2Gm[inDataSize * dataTypeSize] = {0};
uint8_t input1Gm[inDataSize * dataTypeSize] = {0};
uint8_t input0Gm[inDataSize * dataTypeSize] = {0};
uint8_t outputGm[inDataSize * dataTypeSize] = {0};
param.calFunc(input0Gm, input1Gm, input2Gm, outputGm, inDataSize, inMode);
for (int32_t i = 0; i < inDataSize; i++) {
EXPECT_EQ(outputGm[i], 0x00);
}
}