Question

我正在试验Thrust示例monte-carlo.ru： https://github.com/thrust/thrust/blob/master/examples/monte_carlo.cu。
这个问题出现在代码的和平中：

    float estimate = thrust::transform_reduce(thrust::counting_iterator<int>(0),
            thrust::counting_iterator<int>(M),
            estimate_pi(),
            0.0f,
            thrust::plus<float>());

当我将transform_reduce方法的输入序列长度增加到大于M = 87000时，我得到一个错误：
＆＃34;同步：launch_closure_by_value：未知错误＆＃34;

错误之前，屏幕变黑了几秒钟，然后在系统托盘中我看到一条消息＆＃34;视频驱动程序NVidia停止响应并成功恢复了＃34; （我的翻译）然后我重新启动计算机，因为它的行为不稳定。

当我尝试使用cuda-memcheck时，情况发生了变化：我已经得到了相同的错误，长度为M = 30000，尽管在没有cuda-memcheck的情况下运行.exe时程序会成功结束此长度。

以下是cuda-memcheck输出中的几行：

========= Program hit cudaErrorUnknown (error 30) due to "unknown error" on CUDA API call to cudaThreadSynchronize. 
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:C:\Windows\system32\nvcuda.dll (cuProfilerStop + 0xc2d92) [0xe06b2]
=========     Host Frame:C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\bin\cudart64_65.dll (cudaThreadSynchronize + 0xf5) [0x19585]
=========     Host Frame:C:\test\Monte_carlo.exe (thrust::system::cuda::detail::synchronize + 0x47) [0x11117]
...  
========= Program hit cudaErrorUnknown (error 30) due to "unknown error" on CUDA API call to cudaFree. 
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:C:\Windows\system32\nvcuda.dll (cuProfilerStop + 0xc2d92) [0xe06b2]
=========     Host Frame:C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\bin\cudart64_65.dll (cudaFree + 0xfd) [0x1d28d]
=========     Host Frame:C:\test\Monte_carlo.exe (thrust::system::cuda::detail::free > + 0x50) [0x5fa0]

下面还有该程序的完整代码。我在原来只添加了2个更改：try-catch around transform_reduce并从控制台输入M. 我怎样才能理解这个错误的原因？

#include <thrust/random.h>
#include <thrust/iterator/counting_iterator.h>
#include <thrust/functional.h>
#include <thrust/transform_reduce.h>

#include <iostream>
#include <iomanip>
#include <cmath>

// we could vary M & N to find the perf sweet spot

__host__ __device__
unsigned int hash(unsigned int a)
{
    a = (a+0x7ed55d16) + (a<<12);
    a = (a^0xc761c23c) ^ (a>>19);
    a = (a+0x165667b1) + (a<<5);
    a = (a+0xd3a2646c) ^ (a<<9);
    a = (a+0xfd7046c5) + (a<<3);
    a = (a^0xb55a4f09) ^ (a>>16);
    return a;
}

struct estimate_pi : public thrust::unary_function<unsigned int,float>
{
  __host__ __device__
  float operator()(unsigned int thread_id)
  {
    float sum = 0;
    unsigned int N = 10000; // samples per thread

    unsigned int seed = hash(thread_id);

    // seed a random number generator
    thrust::default_random_engine rng(seed);

    // create a mapping from random numbers to [0,1)
    thrust::uniform_real_distribution<float> u01(0,1);

    // take N samples in a quarter circle
    for(unsigned int i = 0; i < N; ++i)
    {
      // draw a sample from the unit square
      float x = u01(rng);
      float y = u01(rng);

      // measure distance from the origin
      float dist = sqrtf(x*x + y*y);

      // add 1.0f if (u0,u1) is inside the quarter circle
      if(dist <= 1.0f)
        sum += 1.0f;
    }

    // multiply by 4 to get the area of the whole circle
    sum *= 4.0f;

    // divide by N
    return sum / N;
  }
};

int main(void)
{
  // use 30K independent seeds
  int M;

    std::cout << "M: ";
    std::cin >> M;
    try
    {
        float estimate = thrust::transform_reduce(thrust::counting_iterator<int>(0),
                                thrust::counting_iterator<int>(M),
                                estimate_pi(),
                                0.0f,
                                thrust::plus<float>());
        estimate /= M;

        std::cout << "M = " << std::setw(6) << M << "  " << std::endl;
        std::cout << std::setprecision(6);
        std::cout << "pi is approximately " << estimate << std::endl;
    }

    catch (thrust::system_error &e)   
    {
        // output an error message and exit
        std::cerr << "Error: " << e.what() << std::endl;
        exit(-1);
    }
    return 0;
}

推力：同步：launch_closure_by_value：未知错误

0 个答案: