111qqz的小窝

老年咸鱼冲锋！

Date Time

18/02/9 14:55

Author

Category

c++ computer vision deep-learning

Tags

Leave a Reply

cuda error checking 学习笔记

由于发现cuda c++ 的 debug方式和c++ 差别很大,因此打算再开一篇,专门记录一些和error checking 以及debug有关的内容.

Error checks in CUDA code can help catch CUDA errors at their source. There are 2 sources of errors in CUDA source code:

  1. Errors from CUDA API calls. For example, a call to  cudaMalloc() might fail.
  2. Errors from CUDA kernel calls. For example, there might be invalid memory access inside a kernel.

All CUDA API calls return a  cudaError value, so these calls are easy to check:

C++
1
2
3
4
5
6
7
8
9
 
 
 
if ( cudaSuccess != cudaMalloc( &fooPtr, fooSize ) )
    printf( "Error!\n" );
 
 
 
 

CUDA kernel invocations do not return any value. Error from a CUDA kernel call can be checked after its execution by calling  cudaGetLastError():
C++
1
2
3
4
5
6
7
8
9
10
 
 
 
fooKernel<<< x, y >>>(); // Kernel call
if ( cudaSuccess != cudaGetLastError() )
    printf( "Error!\n" );
 
 
 
 

 

以及下面是一个check error的宏…

C++
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
 
 
 
#define CHECK_ERROR( err ) \
  if( err != cudaSuccess ) { \
    std::cerr << "ERROR: " << cudaGetErrorString( err ) << std::endl; \
    exit( -1 ); \
  }
 
#define CHECK_LAST_ERROR \
  { cudaError_t err = cudaGetLastError(); \
    if( err != cudaSuccess ) { \
      std::cerr << cudaGetErrorString( err ) << std::endl; \
      exit( -1 ); \
    } \
  }
 
 
 

 

说点什么

您将是第一位评论人！

提醒
wpDiscuz