asdConvolve_complex64

该样例旨在提供快速上手、开发和调试算子的最小化实现，其核心目标是使用最精简的代码展示算子的核心功能，而非提供生产级的安全保障。

不推荐用户直接将样例作为业务代码，若用户将示例代码应用在自身的真实业务场景中且发生了安全问题，则需用户自行承担。

#include <iostream>
#include "asdsip.h"
#include "filter_api.h"
#include "acl/acl.h"
#include "acl/acl_base.h"
#include "acl_meta.h"
#include <complex>
#include <vector>
#include <time.h>

using namespace AsdSip;

#define ASD_STATUS_CHECK(err)                                                \
    do {                                                                     \
        AsdSip::AspbStatus err_ = (err);                                     \
        if (err_ != AsdSip::ErrorType::ACL_SUCCESS) {                                      \
            std::cout << "Execute failed." << std::endl; \
            exit(-1);                                                        \
        }                                                                    \
    } while (0)

#define CHECK_RET(cond, return_expr) \
    do {                             \
        if (!(cond)) {               \
            return_expr;             \
        }                            \
    } while (0)

#define LOG_PRINT(message, ...)         \
    do {                                \
        printf(message, ##__VA_ARGS__); \
    } while (0)

int64_t GetShapeSize(const std::vector<int64_t> &shape)
{
    int64_t shapeSize = 1;
    for (auto i : shape) {
        shapeSize *= i;
    }
    return shapeSize;
}

int Init(int32_t deviceId, aclrtStream *stream)
{
    // 固定写法，acl初始化
    auto ret = aclInit(nullptr);
    CHECK_RET(ret == ::ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
    ret = aclrtSetDevice(deviceId);
    CHECK_RET(ret == ::ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return ret);
    ret = aclrtCreateStream(stream);
    CHECK_RET(ret == ::ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return ret);
    return 0;
}

template <typename T>
int CreateAclTensor(const std::vector<T> &hostData, const std::vector<int64_t> &shape, void **deviceAddr,
    aclDataType dataType, aclTensor **tensor)
{
    auto size = GetShapeSize(shape) * sizeof(T);
    // 调用aclrtMalloc申请device侧内存
    auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
    CHECK_RET(ret == ::ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return ret);
    // 调用aclrtMemcpy将host侧数据复制到device侧内存上
    ret = aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
    CHECK_RET(ret == ::ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return ret);

    // 计算连续tensor的strides
    std::vector<int64_t> strides(shape.size(), 1);
    for (int64_t i = shape.size() - 2; i >= 0; i--) {
        strides[i] = shape[i + 1] * strides[i + 1];
    }

    // 调用aclCreateTensor接口创建aclTensor
    *tensor = aclCreateTensor(shape.data(),
        shape.size(),
        dataType,
        strides.data(),
        0,
        aclFormat::ACL_FORMAT_ND,
        shape.data(),
        shape.size(),
        *deviceAddr);
    return 0;
}

template <typename T>
void printTensor(std::vector<T> tensorData, int64_t tensorSize)
{
    for (int64_t i = 0; i < tensorSize; i++) {
        std::cout << tensorData[i] << " ";
    }
    std::cout << std::endl;
}


int main(int argc, char **argv)
{

    int deviceId = 0;

    aclrtStream stream;
    aclmdlRI modelRI;

    auto ret = Init(deviceId, &stream);
    CHECK_RET(ret == ::ACL_SUCCESS, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);

    int64_t signalLen = 128;
    int64_t kernelLen = 8;
    int64_t batchCount = 2;

    std::vector<std::complex<float>> tensorSignalData;
    tensorSignalData.reserve(signalLen * batchCount);

    std::vector<float> tensorKernelData;
    tensorKernelData.reserve(kernelLen);

    for (int64_t i = 0; i < signalLen * batchCount; i++) {
        tensorSignalData[i] = {1.0, 1.0};
    }

    for (int64_t i = 0; i < kernelLen; i++) {
        tensorKernelData[i] = 1.0 + i;
    }

    std::vector<std::complex<float>> tensorOutData;
    tensorOutData.reserve(signalLen * batchCount);

    for (int64_t i = 0; i < signalLen * batchCount; i++) {
        tensorOutData[i] = {-1.0, -1.0} ;
    }

    std::vector<int64_t> signalShape = {batchCount, signalLen};
    std::vector<int64_t> kernelShape = {kernelLen};
    std::vector<int64_t> resultShape = {batchCount, signalLen};

    aclTensor *signal = nullptr;
    aclTensor *kernel = nullptr;
    aclTensor *output = nullptr;
    void *signalDeviceAddr = nullptr;
    void *kernelDeviceAddr = nullptr;
    void *outputDeviceAddr = nullptr;

    ret = CreateAclTensor<std::complex<float>>(
        tensorSignalData, signalShape, &signalDeviceAddr, aclDataType::ACL_COMPLEX64, &signal);
    CHECK_RET(ret == ::ACL_SUCCESS, return ret);

    ret = CreateAclTensor<float>(
        tensorKernelData, kernelShape, &kernelDeviceAddr, aclDataType::ACL_FLOAT, &kernel);
    CHECK_RET(ret == ::ACL_SUCCESS, return ret);

    ret = CreateAclTensor<std::complex<float>>(
        tensorOutData, resultShape, &outputDeviceAddr, aclDataType::ACL_COMPLEX64, &output);
    CHECK_RET(ret == ::ACL_SUCCESS, return ret);

    size_t lwork = 0;
    AsdSip::asdConvolveGetWorkspaceSize(signalLen, kernelLen, lwork);
    void *buffer = nullptr;

    std::cout << "lwork = " << lwork << std::endl;
    if (lwork > 0) {
        ret = aclrtMalloc(&buffer, static_cast<int64_t>(lwork), ACL_MEM_MALLOC_HUGE_FIRST);
        CHECK_RET(ret == ::ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret);
    }

    auto convolve_func = [signal, kernel, output, stream, buffer]() {
        int IterNum = 1000;
        for (int i = 0; i < IterNum; i++) {
            ASD_STATUS_CHECK(AsdSip::asdConvolve(signal, kernel, output, asdConvolveMode_t::ASD_CONVOLVE_SAME, stream, buffer));
        }
    };

    aclmdlRICaptureBegin(stream, ACL_MODEL_RI_CAPTURE_MODE_GLOBAL);
    convolve_func();
    aclmdlRICaptureEnd(stream, &modelRI);

    struct timespec start, end;
    clock_gettime(CLOCK_MONOTONIC, &start);

    aclmdlRIExecuteAsync(modelRI, stream);
    aclrtSynchronizeStream(stream);
    clock_gettime(CLOCK_MONOTONIC, &end);
    uint64_t elapsed_time = (end.tv_sec - start.tv_sec) * 1000000000 + (end.tv_nsec - start.tv_nsec);
    printf("ACL graph Execution time: %llu ms\n", elapsed_time / 1000 / 1000);

    clock_gettime(CLOCK_MONOTONIC, &start);
    convolve_func();
    aclrtSynchronizeStream(stream);
    clock_gettime(CLOCK_MONOTONIC, &end);
    elapsed_time = (end.tv_sec - start.tv_sec) * 1000000000 + (end.tv_nsec - start.tv_nsec);
    printf("normal Execution time: %llu ms\n", elapsed_time / 1000 / 1000);

    aclmdlRIDestroy(modelRI);
    aclDestroyTensor(signal);
    aclDestroyTensor(kernel);
    aclDestroyTensor(output);
    aclrtFree(signalDeviceAddr);
    aclrtFree(kernelDeviceAddr);
    aclrtFree(outputDeviceAddr);

    aclrtDestroyStream(stream);
    aclrtResetDevice(deviceId);
    aclFinalize();
    return 0;
}