Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | mg2d: extend stats logging | Anton Khirnov | 2019-04-01 |
| | |||
* | egs_exact: do not construct the matrix more often than necessary | Anton Khirnov | 2019-03-26 |
| | | | | It does not change unless the diff coeffs change. | ||
* | egs_exact: avoid explicit matrix transpoe | Anton Khirnov | 2019-03-26 |
| | |||
* | mg2d: do not leak the diff_coeffs interpolator | Anton Khirnov | 2019-03-26 |
| | |||
* | egs: add more timers | Anton Khirnov | 2019-03-25 |
| | |||
* | mg2d: use appropriate full-weighted restriction for 3rd order FDs | Anton Khirnov | 2019-03-25 |
| | |||
* | mg2d: use the new transfer API for inter-grid transfers | Anton Khirnov | 2019-03-25 |
| | |||
* | mg2d: use ndarray for internal arrays | Anton Khirnov | 2019-03-22 |
| | |||
* | ell_grid_solve: switch to ndarray in its external API | Anton Khirnov | 2019-03-22 |
| | |||
* | mg2d: remove unused variable. | Anton Khirnov | 2019-03-18 |
| | |||
* | mg2d: add bicubic prolongation and generic interpolation | Anton Khirnov | 2019-03-18 |
| | |||
* | mg2d: make the refinement depth/exact solve size configurable | Anton Khirnov | 2019-03-11 |
| | | | | API bump | ||
* | mg2d: allocate separate data arrays for the multigrid layer | Anton Khirnov | 2019-03-11 |
| | | | | | | | Do not use those from the topmost level directly. This introduces an additional copy, but allows us to decouple level allocation from the solver allocation. | ||
* | mg2d: remove unused variable | Anton Khirnov | 2019-03-11 |
| | |||
* | mg2d: refactor levels allocation | Anton Khirnov | 2019-03-04 |
| | | | | This should simplify future improvements. | ||
* | mg2d: be less strict for converging to machine epsilon | Anton Khirnov | 2019-03-04 |
| | |||
* | mg2d: fail if the residual grows too much from the original one | Anton Khirnov | 2019-03-04 |
| | |||
* | mg2d: take into account the factor in front of ∂_xx when computing the CFL ↵ | Anton Khirnov | 2019-03-04 |
| | | | | factor | ||
* | Implement C/r falloff boundary condition. | Anton Khirnov | 2019-03-04 |
| | | | | API bump. | ||
* | mg2d: print a more explicit error message when we reach maxiter | Anton Khirnov | 2019-02-24 |
| | |||
* | Rename fixdiff boundary condition to reflect. | Anton Khirnov | 2019-02-08 |
| | | | | | | | This is what it actually does, we do not support setting arbitrary derivative values. API bump. | ||
* | mg2d: relax the condition on residual convergence | Anton Khirnov | 2019-02-02 |
| | | | | | Apparently in certain cases the residual can briefly grow, even though the process converges after. | ||
* | mg2d: print stats properly for exact solves | Anton Khirnov | 2019-02-02 |
| | |||
* | mg2d: account for effect of the ~u term on the relaxation factor | Anton Khirnov | 2019-01-29 |
| | | | | | The maximum allowed time step in the presence of a -K * u term goes like 2 (dx ** 2) / (4 + K (dx ** 2)) | ||
* | Fix callback type mismatch | Anton Khirnov | 2019-01-29 |
| | |||
* | mg2d: fail when a full multigrid iteration does not reduce the residual | Anton Khirnov | 2019-01-29 |
| | |||
* | Solve the discretized system exactly on the coarsest level. | Anton Khirnov | 2019-01-26 |
| | |||
* | ell_relax -> ell_grid_solve | Anton Khirnov | 2019-01-26 |
| | | | | | Generalize the API to allow for multiple solver types. This is done in preparation for the exact linear system inversion solver. | ||
* | mg2d: factor out the boundary condition-related API | Anton Khirnov | 2019-01-25 |
| | |||
* | Stop duplicating some constants between ell_relax and mg2d. | Anton Khirnov | 2019-01-25 |
| | |||
* | mg2d: split the constant definitions into their own header | Anton Khirnov | 2019-01-25 |
| | | | | This will allow sharing them with ell_relax. | ||
* | mg2d: log the convergence factors after relax/correct | Anton Khirnov | 2019-01-22 |
| | |||
* | mg2d: fix typo | Anton Khirnov | 2019-01-21 |
| | | | | | Do nb_relax_post steps in the post-correct relaxation phase, not nb_relax_pre. | ||
* | mg2d: failer earlier when relaxation does not converge | Anton Khirnov | 2019-01-21 |
| | |||
* | mg2d: check and propagate the error from mg_solve_subgrid() | Anton Khirnov | 2019-01-21 |
| | |||
* | mg2d: factor out the relaxation step call | Anton Khirnov | 2019-01-16 |
| | |||
* | mg2d: add a context variable to control the log level | Anton Khirnov | 2019-01-16 |
| | | | | | | Simpler to use than overriding the log callback. API bump. | ||
* | mg2d: add a function for getting maximum supported fd_stencil | Anton Khirnov | 2019-01-16 |
| | | | | | | | This is needed for the caller to know the number of boundary zones that can be filled. API bump. | ||
* | ell_relax: compute the residual norm in residual_calc() | Anton Khirnov | 2019-01-13 |
| | | | | It is cheap and avoids an extra step in mg2d. | ||
* | mg2d: make the cfl factor runtime-configurable | Anton Khirnov | 2019-01-13 |
| | |||
* | mg2d: ignore padding values in findmax() | Anton Khirnov | 2019-01-09 |
| | |||
* | mg2d: remove a mistakenly committed hunk. | Anton Khirnov | 2018-12-28 |
| | |||
* | mg2d: factor out restriction/prolongation calls | Anton Khirnov | 2018-12-28 |
| | | | | | Also generalize the check for full interpolation vs special-cased restrict/prolong functions. | ||
* | mg2d: print the overhead time in stats | Anton Khirnov | 2018-12-28 |
| | |||
* | mg2d: do not select the first coarser level too close to the finest one | Anton Khirnov | 2018-12-28 |
| | |||
* | mg2d: simplify and speed up prolongation | Anton Khirnov | 2018-12-27 |
| | |||
* | Add threading support through libthreadpool. | Anton Khirnov | 2018-12-27 |
| | |||
* | mg2d: initialize the solver stepsize before it is used | Anton Khirnov | 2018-12-27 |
| | |||
* | Add CPU feature detection. | Anton Khirnov | 2018-12-27 |
| | | | | Will be used for dynamically dispatching future asm functions. | ||
* | mg2d: define log levels | Anton Khirnov | 2018-12-06 |
| |