如何将 cuda-memcheck 应用于具有来自标准 I/O 的管道输入的应用程序

白农

我想将 cuda-memcheck 用于具有标准 I/O 的应用程序。

应用程序 dut 读取标准输入并写入标准输出。

cat input.txt | cuda-memcheck ./dut -dutoptions > output.txt

在这种情况下，dut 应用程序似乎已启动，但 cuda-memcheck 无法正常工作。

我怎样才能做到这一点？

罗伯特·克罗维拉

一种可能的方法：

将要执行的命令放入脚本中。在该脚本上运行 cuda-memcheck。

$ cat input.txt
4

$ cat t2030.cu
#include <iostream>

__global__ void k(int *d){
  *d = 5;
}

int main(){

  int *d = NULL;
#ifndef FAULT
  cudaMalloc(&d, sizeof(d[0]));
#endif
  int a;
  std::cin  >> a;
  std::cout << "value read: " << a << std::endl;
  k<<<1,1>>>(d);
  cudaDeviceSynchronize();
}

$ nvcc -o t2030 t2030.cu
$ cat test
cat input.txt | ./t2030 > output.txt
$ chmod +x test
$ cuda-memcheck ./test
========= CUDA-MEMCHECK
========= ERROR SUMMARY: 0 errors
$ cat output.txt
value read: 4
$ nvcc -DFAULT -o t2030 t2030.cu
$ cuda-memcheck ./test
========= CUDA-MEMCHECK
========= Invalid __global__ write of size 4
=========     at 0x00000050 in k(int*)
=========     by thread (0,0,0) in block (0,0,0)
=========     Address 0x00000000 is out of bounds
=========     Device Frame:k(int*) (k(int*) : 0x50)
=========     Saved host backtrace up to driver entry point at kernel launch time
=========     Host Frame:/lib64/libcuda.so.1 [0x20d6ea]
=========     Host Frame:./t2030 [0x864b]
=========     Host Frame:./t2030 [0x5ec88]
=========     Host Frame:./t2030 [0x3ea0]
=========     Host Frame:./t2030 [0x3d81]
=========     Host Frame:./t2030 [0x3da9]
=========     Host Frame:./t2030 [0x3c2a]
=========     Host Frame:/lib64/libc.so.6 (__libc_start_main + 0xf5) [0x21b15]
=========     Host Frame:./t2030 [0x3a91]
=========
========= Program hit cudaErrorLaunchFailure (error 719) due to "unspecified launch failure" on CUDA API call to cudaDeviceSynchronize.
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:/lib64/libcuda.so.1 [0x3bd253]
=========     Host Frame:./t2030 [0x3a047]
=========     Host Frame:./t2030 [0x3c2f]
=========     Host Frame:/lib64/libc.so.6 (__libc_start_main + 0xf5) [0x21b15]
=========     Host Frame:./t2030 [0x3a91]
=========
========= ERROR SUMMARY: 2 errors
$

请注意，对于 volta 或更新的设备，cuda-memcheck应替换为compute-sanitizer.

本文收集自互联网，转载请注明来源。

如有侵权，请联系 [email protected] 删除。