Segmentation fault when using cusolverSpScsrlsvchol in CUDA for sparse linear problems.

huangapple
go评论109阅读模式
英文:

Segmentation fault when using cusolverSpScsrlsvchol in CUDA for sparse linear problems

问题

我正在尝试将一个线性问题移植到CUDA以加速求解时间。我已成功使用cusolverDn来处理GPU上的密集问题。然而，当我尝试使用cusolverSpScsrlsvchol处理稀疏问题时，我一直收到分段错误。

为了调试这个问题，我使用了CUDA计算消毒剂，并收到了以下输出：

  1. $ /c/Programme/NVIDIA\ GPU\ Computing\ Toolkit/CUDA/v11.7/bin/compute-sanitizer.bat --tool memcheck bin/FEMaster_gpu.exe
  2. ========= COMPUTE-SANITIZER
  4. ========= Error: process didn't terminate successfully
  5. ========= Target application returned an error
  6. ========= ERROR SUMMARY: 0 errors
  7. Segmentation fault

我将问题缩小到以下最小代码段：

  1. cusolverSpHandle_t handle_cusolver_sp;
  2. cusparseHandle_t   handle_cusparse;
  4. // loading handles
  5. cusolverSpCreate(&handle_cusolver_sp);
  6. cusparseCreate  (&handle_cusparse);
  8. // get properties
  9. cudaSetDevice(0);
  11. // create csr arrays on cpu
  12. float host_csr_values[4]{1,1,1,1};
  13. int   host_csr_col_id[4]{0,1,2,3};
  14. int   host_csr_row_pt[5]{0,1,2,3,4};
  15. float host_rhs       [4]{0,3,7,1};
  16. int   host_singular  [1]{0};
  18. // allocate arrays on the gpu
  19. float* dev_csr_values;
  20. int  * dev_csr_col_id;
  21. int  * dev_csr_row_pt;
  22. float* dev_rhs;
  23. int  * dev_singular;
  25. runtime_assert_cuda(cudaMalloc((void**) &dev_csr_values,4 * sizeof(float)));
  26. runtime_assert_cuda(cudaMalloc((void**) &dev_csr_col_id,4 * sizeof(int  )));
  27. runtime_assert_cuda(cudaMalloc((void**) &dev_csr_row_pt,5 * sizeof(int  )));
  28. runtime_assert_cuda(cudaMalloc((void**) &dev_rhs       ,4 * sizeof(float)));
  29. runtime_assert_cuda(cudaMalloc((void**) &dev_singular  ,1 * sizeof(int  )));
  31. // move data to gpu
  32. runtime_assert_cuda(cudaMemcpy(dev_csr_values, host_csr_values, 4 * sizeof(float), cudaMemcpyHostToDevice));
  33. runtime_assert_cuda(cudaMemcpy(dev_csr_col_id, host_csr_col_id, 4 * sizeof(int  ), cudaMemcpyHostToDevice));
  34. runtime_assert_cuda(cudaMemcpy(dev_csr_row_pt, host_csr_row_pt, 5 * sizeof(int  ), cudaMemcpyHostToDevice));
  35. runtime_assert_cuda(cudaMemcpy(dev_rhs       , host_rhs       , 4 * sizeof(float), cudaMemcpyHostToDevice));
  37. // create matrix descriptor
  38. cusparseMatDescr_t descr;
  39. runtime_assert_cuda(cusparseCreateMatDescr(&descr));
  40. runtime_assert_cuda(cusparseSetMatType     (descr, CUSPARSE_MATRIX_TYPE_GENERAL));
  41. runtime_assert_cuda(cusparseSetMatIndexBase(descr, CUSPARSE_INDEX_BASE_ZERO    ));
  43. runtime_assert_cuda(cusolverSpScsrlsvchol(handle_cusolver_sp,
  44.                                           4,
  45.                                           4,
  46.                                           descr,
  47.                                           dev_csr_values,
  48.                                           dev_csr_row_pt,
  49.                                           dev_csr_col_id,
  50.                                           dev_rhs,
  51.                                           0,    // tolerance
  52.                                           0,    // reorder
  53.                                           dev_rhs,
  54.                                           dev_singular));

我放入稀疏矩阵的值是对角矩阵的值。

为简单起见，我删除了内存释放、输出检索和其他类似调用。代码看起来很简单，但在调用cusolverSpScsrlsvchol时导致分段错误。我已经被这个问题困扰了一天多，但无法弄清楚为什么它不起作用。非常感谢任何帮助！

英文:

I'm trying to port a linear problem to CUDA in order to speed up solving times. I have successfully used cusolverDn to handle dense problems on the GPU. However, when I attempted to apply it to sparse problems using cusolverSpScsrlsvchol, I keep getting a segmentation fault.

To debug the issue, I used the CUDA compute sanitizers and received the following output:

  1. $ /c/Programme/NVIDIA\ GPU\ Computing\ Toolkit/CUDA/v11.7/bin/compute-sanitizer.bat --tool memcheck bin/FEMaster_gpu.exe
  2. ========= COMPUTE-SANITIZER
  3. ========= Error: process didn't terminate successfully
  4. ========= Target application returned an error
  5. ========= ERROR SUMMARY: 0 errors
  6. Segmentation fault

I narrowed down the problem to the following minimal code snippet:

  1. cusolverSpHandle_t handle_cusolver_sp;
  2.     cusparseHandle_t   handle_cusparse;
  4.     // loading handles
  5.     cusolverSpCreate(&handle_cusolver_sp);
  6.     cusparseCreate  (&handle_cusparse);
  8.     // get properties
  9.     cudaSetDevice(0);
  11.     // create csr arrays on cpu
  12.     float host_csr_values[4]{1,1,1,1};
  13.     int   host_csr_col_id[4]{0,1,2,3};
  14.     int   host_csr_row_pt[5]{0,1,2,3,4};
  15.     float host_rhs       [4]{0,3,7,1};
  16.     int   host_singular  [1]{0};
  18.     // allocate arrays on the gpu
  19.     float* dev_csr_values;
  20.     int  * dev_csr_col_id;
  21.     int  * dev_csr_row_pt;
  22.     float* dev_rhs;
  23.     int  * dev_singular;
  25.     runtime_assert_cuda(cudaMalloc((void**) &dev_csr_values,4 * sizeof(float)));
  26.     runtime_assert_cuda(cudaMalloc((void**) &dev_csr_col_id,4 * sizeof(int  )));
  27.     runtime_assert_cuda(cudaMalloc((void**) &dev_csr_row_pt,5 * sizeof(int  )));
  28.     runtime_assert_cuda(cudaMalloc((void**) &dev_rhs       ,4 * sizeof(float)));
  29.     runtime_assert_cuda(cudaMalloc((void**) &dev_singular  ,1 * sizeof(int  )));
  31.     // move data to gpu
  32.     runtime_assert_cuda(cudaMemcpy(dev_csr_values, host_csr_values, 4 * sizeof(float), cudaMemcpyHostToDevice));
  33.     runtime_assert_cuda(cudaMemcpy(dev_csr_col_id, host_csr_col_id, 4 * sizeof(int  ), cudaMemcpyHostToDevice));
  34.     runtime_assert_cuda(cudaMemcpy(dev_csr_row_pt, host_csr_row_pt, 5 * sizeof(int  ), cudaMemcpyHostToDevice));
  35.     runtime_assert_cuda(cudaMemcpy(dev_rhs       , host_rhs       , 4 * sizeof(float), cudaMemcpyHostToDevice));
  37.     // create matrix descriptor
  38.     cusparseMatDescr_t descr;
  39.     runtime_assert_cuda(cusparseCreateMatDescr(&descr));
  40.     runtime_assert_cuda(cusparseSetMatType     (descr, CUSPARSE_MATRIX_TYPE_GENERAL));
  41.     runtime_assert_cuda(cusparseSetMatIndexBase(descr, CUSPARSE_INDEX_BASE_ZERO    ));
  43.     runtime_assert_cuda(cusolverSpScsrlsvchol(handle_cusolver_sp,
  44.                                               4,
  45.                                               4,
  46.                                               descr,
  47.                                               dev_csr_values,
  48.                                               dev_csr_row_pt,
  49.                                               dev_csr_col_id,
  50.                                               dev_rhs,
  51.                                               0,    // tolerance
  52.                                               0,    // reorder
  53.                                               dev_rhs,
  54.                                               dev_singular));

The values I put in there for the sparse matrix is the one for a diagonal matrix.

I removed the memory deallocation, output retrieval, and other similar calls for simplicity. The code seems straightforward, but it results in a segmentation fault. The issue occurs specifically during the call to cusolverSpScsrlsvchol.

I've been stuck on this problem for over a day and I can't figure out why it's not working. Any help would be greatly appreciated!

答案1

得分: 2

API中指出，奇异性参数应该在主机内存空间中，而不是设备内存中。

英文:

The API states that the singularity parameter is supposed to be in host memory space, not device.

huangapple
  • 本文由 发表于 2023年6月6日 17:31:59
  • 转载请务必保留本文链接：https://go.coder-hub.com/76413263.html
  • c++
  • cuda
  • cusolver
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定