Whn using two square matrices, A and B, each of size NxN, with A being the Cholesky factorization of a positive definite matrix with a zeroed lower triangle, and B being an identity matrix, the program segfaults. I checked the obvious, and it still fails. It looks like anything beyond tile size either segfaults or gives nondeterministic results. The code is an adapted version from the available examples in the official documentation.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Example using cusparse and cusolverSpDcsrlsvchol | 7 | 57 | October 8, 2024 | |
| Cusparse cholesky & structural zeros - preconditioned conjugate gradient | 3 | 1050 | March 15, 2021 | |
| Abnormally slow performance | 3 | 839 | February 4, 2019 | |
| cuSolverSp_LowlevelCholesky fails with CUSOLVER_STATUS_INTERNAL_ERROR | 0 | 586 | March 8, 2020 | |
| Cannot manage to use cuSOLVER to solve a linear system | 3 | 685 | May 24, 2018 | |
| cuSOLVER solving a overdetermined linear equation system | 8 | 1261 | September 19, 2021 | |
| cusolverDnDpotrf does not support float | 0 | 526 | August 30, 2017 | |
| CUDA Cholesky factorization gives NaN when the input matrix is positive definite matrix | 3 | 1021 | March 10, 2022 | |
| Cholesky factorizing of A*D^2*A’ ... | 2 | 650 | May 15, 2013 | |
| Incomplete-LU and Cholesky Preconditioned | 4 | 5337 | November 8, 2012 |