/**
* 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 rint.asc
 * \brief
 */

#include "acl/acl.h"
#include "data_utils.h"
#include "kernel_operator.h"

template <typename T, int32_t calCount, int32_t dataSize, int32_t sharedTmpBufferSize>
class KernelRint {
public:
    __aicore__ inline KernelRint() {}
    __aicore__ inline void Init(GM_ADDR srcGm, GM_ADDR dstGm, AscendC::TPipe* pipeIn)
    {
        pipe = pipeIn;
        srcGlobal.SetGlobalBuffer(reinterpret_cast<__gm__ T*>(srcGm));
        dstGlobal.SetGlobalBuffer(reinterpret_cast<__gm__ T*>(dstGm));
        pipe->InitBuffer(inQueue, 1, dataSize * sizeof(T));
        pipe->InitBuffer(outQueue, 1, dataSize * sizeof(T));
        if constexpr (sharedTmpBufferSize > 0) {
            pipe->InitBuffer(bufQueue, sharedTmpBufferSize * sizeof(T));
        }
    }
    __aicore__ inline void Process()
    {
        AscendC::AscendCUtils::SetOverflow(1);
        CopyIn();
        Compute();
        CopyOut();
        AscendC::AscendCUtils::SetOverflow(0);
    }

    __aicore__ inline void CopyIn()
    {
        AscendC::LocalTensor<T> srcLocal = inQueue.AllocTensor<T>();
        AscendC::DataCopy(srcLocal, srcGlobal, dataSize);
        inQueue.EnQue(srcLocal);
    }
    __aicore__ inline void Compute()
    {
        AscendC::LocalTensor<T> dstLocal = outQueue.AllocTensor<T>();
        AscendC::LocalTensor<T> srcLocal = inQueue.DeQue<T>();
        AscendC::Duplicate(dstLocal, (T)0, dataSize);
        if constexpr (sharedTmpBufferSize > 0) {
            AscendC::LocalTensor<uint8_t> tmpBuf = bufQueue.Get<uint8_t>();
            AscendC::Rint(dstLocal, srcLocal, tmpBuf, calCount);
        } else {
            AscendC::Rint(dstLocal, srcLocal, calCount);
        }
        outQueue.EnQue(dstLocal);
        inQueue.FreeTensor(srcLocal);
    }
    __aicore__ inline void CopyOut()
    {
        AscendC::LocalTensor<T> dstLocal = outQueue.DeQue<T>();
        AscendC::DataCopy(dstGlobal, dstLocal, dataSize);
        outQueue.FreeTensor(dstLocal);
    }

private:
    AscendC::TPipe* pipe;
    AscendC::TQue<AscendC::QuePosition::VECIN, 1> inQueue;
    AscendC::TQue<AscendC::QuePosition::VECOUT, 1> outQueue;
    AscendC::TBuf<AscendC::TPosition::VECCALC> bufQueue;
    AscendC::GlobalTensor<T> srcGlobal;
    AscendC::GlobalTensor<T> dstGlobal;
};

__global__ __aicore__ void rint_custom(GM_ADDR srcGm, GM_ADDR dstGm)
{
    KERNEL_TASK_TYPE_DEFAULT(KERNEL_TYPE_AIV_ONLY);
    AscendC::TPipe pipe;
    constexpr uint32_t dataSize = 1024;
    constexpr uint32_t calCount = 1024;
    constexpr uint32_t sharedTmpBufSize = 1024;
    KernelRint<float, calCount, (dataSize + 31) / 32 * 32, sharedTmpBufSize> op;
    op.Init(srcGm, dstGm, &pipe);
    op.Process();
}

static bool CompareResult(const void* outputData, uint32_t outSize)
{
    void* goldenData;
    aclrtMallocHost((void**)(&goldenData), outSize);
    size_t goldenSize = outSize;
    bool ret = ReadFile("./output/golden.bin", goldenSize, goldenData, goldenSize);
    if (ret) {
        printf("ReadFile golden.bin success!\n");
    } else {
        printf("test failed!\n");
        return false;
    }
    constexpr float EPS = 1e-4;
    int64_t wrongNum = 0;

    for (size_t i = 0; i < outSize / sizeof(float); i++) {
        float a = (reinterpret_cast<const float*>(outputData))[i];
        float b = (reinterpret_cast<const float*>(goldenData))[i];
        float ae = std::abs(a - b);
        float re = ae / std::abs(b);
        if (ae > EPS && re > EPS) {
            printf("CompareResult golden.bin failed output is %lf, golden is %lf\n", a, b);
            wrongNum++;
        }
    }
    aclrtFreeHost(goldenData);
    if (wrongNum != 0) {
        return false;
    } else {
        printf("CompareResult golden.bin success!\n");
        return true;
    }
}

int32_t main(int32_t argc, char* argv[])
{
    size_t param1FileSize = 1024 * sizeof(int32_t);
    size_t param2FileSize = 1024 * sizeof(int32_t);
    uint32_t numBlocks = 1;

    aclInit(nullptr);
    aclrtContext context;
    int32_t deviceId = 0;
    aclrtSetDevice(deviceId);
    aclrtCreateContext(&context, deviceId);
    aclrtStream stream = nullptr;
    aclrtCreateStream(&stream);

    uint8_t* param1Host;
    uint8_t* param1Device;
    aclrtMallocHost((void**)(&param1Host), param1FileSize);
    aclrtMalloc((void**)&param1Device, param1FileSize, ACL_MEM_MALLOC_HUGE_FIRST);
    ReadFile("./input/input_src.bin", param1FileSize, param1Host, param1FileSize);
    aclrtMemcpy(param1Device, param1FileSize, param1Host, param1FileSize, ACL_MEMCPY_HOST_TO_DEVICE);

    uint8_t* param2Host;
    uint8_t* param2Device;
    aclrtMallocHost((void**)(&param2Host), param2FileSize);
    aclrtMalloc((void**)&param2Device, param2FileSize, ACL_MEM_MALLOC_HUGE_FIRST);

    rint_custom<<<numBlocks, nullptr, stream>>>(param1Device, param2Device);
    aclrtSynchronizeStream(stream);

    aclrtFree(param1Device);
    aclrtFreeHost(param1Host);

    aclrtMemcpy(param2Host, param2FileSize, param2Device, param2FileSize, ACL_MEMCPY_DEVICE_TO_HOST);
    WriteFile("./output/output.bin", param2Host, param2FileSize);

    bool goldenResult = true;
    goldenResult = CompareResult(param2Host, param2FileSize);
    if (goldenResult) {
        printf("test pass!\n");
    } else {
        printf("test failed!\n");
    }

    aclrtFree(param2Device);
    aclrtFreeHost(param2Host);

    aclrtDestroyStream(stream);
    aclrtDestroyContext(context);
    aclrtResetDevice(deviceId);
    aclFinalize();

    return 0;
}