nvml sample(written in c), nvml API reference is here: https://docs.nvidia.com/deploy/nvml-api/index.html

nvml_sample.c

#include <stdio.h>
#include <nvml.h>

int main(int argc, char **argv)
{
    nvmlReturn_t result;

    result = nvmlInit();
    if (NVML_SUCCESS != result)
    {
        printf("Failed to initialize NVML: %s\n", nvmlErrorString(result));
        printf("Press ENTER to continue...\n");
        getchar();
        return 1;
    }

    unsigned int device_count;
    result = nvmlDeviceGetCount(&device_count);
    if (NVML_SUCCESS != result)
    {
        printf("Failed to query device count: %s\n", nvmlErrorString(result));
        return 1;
    }

    for (int i = 0; i < device_count; i++)
    {
        nvmlDevice_t device;
        result = nvmlDeviceGetHandleByIndex(i, &device);
        if (NVML_SUCCESS != result)
        {
            printf("Failed to get device handle: %s\n", nvmlErrorString(result));
            return 1;
        }

        nvmlUtilization_t util;
        result = nvmlDeviceGetUtilizationRates(device, &util);
        if (NVML_SUCCESS != result)
        {
            printf("Failed to get utilization rates: %s\n", nvmlErrorString(result));
            return 1;
        }

        printf("gpu: %d\n", util.gpu);
        printf("memory: %d\n", util.gpu);

        nvmlMemory_t memory;
        result = nvmlDeviceGetMemoryInfo(device, &memory);
        if (NVML_SUCCESS != result)
        {
            printf("Failed to get device memory info: %s\n", nvmlErrorString(result));
            return 1;
        }

        printf("free: %lld\n", memory.free);
        printf("used: %lld\n", memory.used);
        printf("total: %lld\n", memory.total);
    }

    nvmlShutdown();
}

build sample

$ gcc -std=c99 -Wall nvml_test.c -I /usr/local/include -I /usr/local/cuda-9.1/include -o nvml_test -lnvidia-ml -L /usr/lib64/nvidia