Skip to content

Releases: onera/CoMMA

Version v1.3.2

24 Jan 10:48
Compare
Choose a tag to compare

No changes w.r.t. v1.3.1 and v1.3 in terms of agglomeration algorithm. However, some important changes have been introduced in the structure and handling of the project:

  • Add dedicated namespace comma
  • Add choice of special types (for indices, standard integers, and reals) during configuration stage
  • Move headers into include/CoMMA and sources into src
  • Avoid using using namespace foo;
  • Add ctidy
  • [CI] Add tests for 32- and 64bit integer support and for pkgconfig support, and stages for format and linting (both for C++ and python files)

Version v1.3.1

06 Jan 10:21
Compare
Choose a tag to compare

No major changes wrt v1.3. This version only improves the building process. For instance, a make install target is available which installs the headers (and python and documentation if requested) in the given prefix. Support for pkg-config has been added.

Version v1.3

19 Dec 13:37
Compare
Choose a tag to compare

Changelog:

  • Fixes and improvements for correction
  • Restructure tests into small units
  • Clean and improve examples

This version has been registered to the Agency for the Protection of Programs (APP), Paris, with IDDN identification number IDDN.FR.001.420013.000.S.X.2023.000.31235

Version v1.2

19 Dec 13:34
Compare
Choose a tag to compare

This version brings two major improvements:

  • Anisotropic lines: the growth direction is checked before including a new cell. Moreover, when agglomerating, user can choose whether to have or not 3-cells agglomerates if lines have odd length
  • (Drastic) Performance improvements: anisotropic lines build and correction stage have been restructured. Moreover, the subgraphs of coarse cells are not build anymore. The improvements are significant: test on a series of 2D RAE meshes showed that, see plots below, we switch from an exponential (cf baseline picture, notice log scaling) to a linear one (cf improved) wrt to number of cells.

Case description for the performance tests: 2D RAE mesh with a quad BL and triangles elsewhere. Anisotropic agglomeration enabled with an anisotropic threshold of 2 and subset of compliant cells those around the airfoil (inside a given radius). Three arguments configurations tested:

RAE.1: Cardinality fixed to 4 (min=goal=max), no corrections
RAE.2: Cardinality between 2 and 6, no corrections
RAE.3: Cardinality fixed to 4, corrections enabled

image

image

Version v1.1

19 Dec 13:32
Compare
Choose a tag to compare

This release brings several salient changes:

  • Improve seeds pool handling and add some user option to control the order
  • Consider user-defined priority weights used to compute the order of the seeds
  • Consider cell centres to better evaluate the aspect-ratio
  • Make isotropic and anisotropic agglomeration numbering deterministic
  • Add iterative algorithm for the choice of the FC to agglomerate
  • Improve anisotropic algorithm (update seeds pool, start from line end with higher boundary faces,...)
  • Add full support for unsigned types and switch index type to unsigned
  • Improve correction step
  • Improve memory management (avoid copies, set parameters to const,...)
  • Revisit neighbourhood handling