Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | Allow specifying extra cflags/ldflags.HEADmaster | Anton Khirnov | 8 days |
| | |||
* | mg2d: add MG2DContext.adaptive_step | Anton Khirnov | 8 days |
| | | | | Unused, for API compatibility with the adaptive step branch. | ||
* | x86: add a misc utility header | Anton Khirnov | 8 days |
| | |||
* | Makefile: simplify nasm rules | Anton Khirnov | 8 days |
| | | | | There is no need to run nasm twice. | ||
* | x86inc.asm: update to current master 04f14f43 | Anton Khirnov | 8 days |
| | | | | | | | Requires changing residual calc functions to AVX2. Also, supply the private prefix via nasm -D option rather than modifying x86inc.asm. | ||
* | residual_calc: accept all diff coefficients in a single array | Anton Khirnov | 9 days |
| | | | | | | | Plus an offset parameter that signals the distance between different coefficients. This allows to avoid passing so many pointers around, which reduces register pressure and simplifies writing SIMD. Seems also to be a little faster. | ||
* | residual_calc: rename stride to u_stride | Anton Khirnov | 9 days |
| | | | | | Make it explicit that it only applies to u, as other arrays are not indexed beyond curent line. | ||
* | residual_calc: typedef computation kernels | Anton Khirnov | 9 days |
| | |||
* | mg2d: use correct stride | Anton Khirnov | 9 days |
| | |||
* | Switch from meson to configure+make. | Anton Khirnov | 9 days |
| | | | | Meson is an extra dependency and generally sucks. | ||
* | README: fix description taken verbatim from libteukolskydata | Anton Khirnov | 10 days |
| | |||
* | Add README. | Anton Khirnov | 2021-10-10 |
| | |||
* | meson.build: add a trailing slash to asm include path | Anton Khirnov | 2020-07-13 |
| | | | | Required by earlier nasm versions. | ||
* | Track forgotten timer files. | Anton Khirnov | 2020-07-13 |
| | |||
* | mg2d: avoid leaks in mg_dh_init() | Anton Khirnov | 2020-07-12 |
| | |||
* | mg2d: add logging to some error paths | Anton Khirnov | 2020-07-11 |
| | |||
* | Switch to external ndarray library. | Anton Khirnov | 2020-06-25 |
| | |||
* | bicgstab: drop unused label | Anton Khirnov | 2020-01-26 |
| | |||
* | common: print the source location in mg2di_assert() | Anton Khirnov | 2020-01-26 |
| | |||
* | mg2d: create a per-level logging context | Anton Khirnov | 2020-01-26 |
| | |||
* | mg2d: export the residual norm to the caller | Anton Khirnov | 2020-01-20 |
| | |||
* | mg2d: add a designated error code for reaching maxiter | Anton Khirnov | 2020-01-20 |
| | | | | | Allows to reliably distinguish it from other error cases. Also specify that the values of u are defined for this error code. | ||
* | transfer: slightly relax the domain overlap check | Anton Khirnov | 2020-01-20 |
| | | | | | Allow the destination domain to be outside the source by a small margin (1e-12). | ||
* | common: do not interpret printf sequences in mg2di_assert() | Anton Khirnov | 2019-07-10 |
| | |||
* | mg2d: calculate the cfl from continuous diff coeffs | Anton Khirnov | 2019-07-06 |
| | |||
* | mg2d: reduce restriction order to 1 | Anton Khirnov | 2019-06-29 |
| | | | | This seems to work more reliably. | ||
* | egs: reduce the max number of BiCGSTAB iterations to 4 | Anton Khirnov | 2019-06-29 |
| | | | | | Since we typically do many multigrid iterations per step, being more aggressive about a full inverse gives better performance. | ||
* | egs: sync the residual maximum across components. | Anton Khirnov | 2019-06-29 |
| | |||
* | transfer: implement and use 1D interpolation | Anton Khirnov | 2019-06-29 |
| | | | | Stop abusing "2D of y size 1" for this. | ||
* | mg2d: make init_guess() MPI-aware | Anton Khirnov | 2019-06-29 |
| | | | | API and ABI break. | ||
* | implement several relax steps per sync | Anton Khirnov | 2019-06-29 |
| | |||
* | egs: make array allocation more readable | Anton Khirnov | 2019-06-29 |
| | |||
* | egs: allocate mpi sync parameters only after we know the fd stencil | Anton Khirnov | 2019-06-18 |
| | | | | | Also, forbid it to change after first init. That is not ever used and allowing it unnecessarily complicates the logic. | ||
* | egs: skip boundary corners for inter-component boundaries | Anton Khirnov | 2019-06-18 |
| | |||
* | egs: simplify reflection boundary handling | Anton Khirnov | 2019-06-14 |
| | |||
* | rescalc: improve reflection boundary conditions | Anton Khirnov | 2019-06-14 |
| | | | | | Make parameter names more clear/consistent, document them, implement missing 1U boundary. | ||
* | egs: constify a function argument | Anton Khirnov | 2019-06-14 |
| | |||
* | mg2d: add API for specifying singular diff coeffs at the boundaries | Anton Khirnov | 2019-06-13 |
| | | | | API and ABI break. | ||
* | mg2d: relax condition on divergence | Anton Khirnov | 2019-06-13 |
| | |||
* | ndarray: fix handling SLICE_NULL | Anton Khirnov | 2019-06-13 |
| | |||
* | mg2d: implement multicomponent solves for coarser levels | Anton Khirnov | 2019-06-13 |
| | |||
* | mg2d: do not pass NULL as a dummy MPI datatype | Anton Khirnov | 2019-06-04 |
| | | | | Use MPI_DATATYPE_NULL. | ||
* | mg2d: use the all-comp global residual norm for divergence testing | Anton Khirnov | 2019-06-04 |
| | | | | | Residual may grow dramatically in certain components while globally decreasing. | ||
* | mg2d: export the local component extents in public API | Anton Khirnov | 2019-06-02 |
| | |||
* | egs: handle falloff boundaries properly with multiple components | Anton Khirnov | 2019-06-02 |
| | |||
* | mg2d: add support for MPI-based multi-component solves | Anton Khirnov | 2019-05-23 |
| | | | | | For the moment, only the finest component is distributed, any coarser levels are gathered to rank 0. That should change in the future. | ||
* | egs: add support for MPI-based multi-component solves | Anton Khirnov | 2019-05-23 |
| | |||
* | mg2d_test: scale the tolerance with step size | Anton Khirnov | 2019-05-23 |
| | |||
* | transfer: allow optional extrapolation | Anton Khirnov | 2019-05-23 |
| | |||
* | Make the ghost points explicit in prolongation | Anton Khirnov | 2019-05-21 |
| |