用CUDA编程

Question

我是使用CUDA的新手，正在尝试编写代码，但是仍然遇到麻烦。

CPU代码正常。但不是GNU代码。

我在编译代码时收到的错误是：

{
    "jsonrpc": "2.0",
    "id": "123",
    "result": {
        "columns": [
            "Unit Number",
            "Tank Number",
            "Volume",
            "Volume Percent",
            "Description",
            "Capacity",
            "Status",
            "Last Updated"
        ],
        "values": [
            [
                null,
                null,
                "0",
                "0",
                "Tank 1",
                "50000",
                "N/A",
                "1970-01-01 09:30:00"
            ],
            [
                "3376",
                "1",
                "18490",
                "68.4815",
                "SmartFill 3376 Tank 1",
                "27000",
                "Offline",
                "2018-06-06 14:28:20"
            ]
        ]
    }
}

和警告：先前的语法错误后，解析将在此处重新开始

ejemplo.cu(59): error: argument of type "double" is incompatible with parameter of type "const void *"

ejemplo.cu(61): error: argument of type "double" is incompatible with parameter of type "const void *"

ejemplo.cu(66): error: identifier "__global__void" is undefined

ejemplo.cu(67): error: expected a ";"

Answer 1

您可能想在下划线后面写上 __ global__ void 。对于不兼容错误，您需要在cudamemcpy中传递指针。 尝试：

cudaMemcpy(SUMAd, &SUMA, sizeof(double), cudaMemcpyHostToDevice);

cudamemcpy参考：http://horacio9573.no-ip.org/cuda/group__CUDART__MEMORY_g48efa06b81cc031b2aa6fdc2e9930741.html