Releases: kul-optec/superscs
Releases · kul-optec/superscs
SuperSCS version 1.3.2
Features:
- Significantly faster linear algebra
- Optimized Anderson's acceleration
- Introduced and applied coding convention
- Updated benchmarks
- A lot of new documentation
- Using absolute paths in MATLAB
- Profiling framework (gprof)
- Applied SuperSCS on Maros-Meszaros problems
- Updated Makefile
- Achieved high code coverage
- New YAML serializer/parser in C and MATLAB
- Codecov configuration
- Added contributing guide
- Analytics for the website
SuperSCS alpha version
Features
- Implementation of SuperSCS using the same function template as SCS (additional SuperSCS-specific options can be provided in the
Settings
structure - Implementation of restarted limited-memory Broyden directions
- Support for warm start passing an (x,y) solution
- Unit tests (91.6% coverage of scs.c, overall 80% coverage)
- Online API documentation using Doxygen with examples and benchmarks
- Optimizations in memory management (e.g., avoid unnecessary allocations)
- Support of printing to streams using
stdio
- Modification of MEX interface that allows us to call SuperSCS from MATLAB
- Integration with CVX in MATLAB
- A lot more info can be exported to MATLAB regarding the progress of the algorithm (e.g., relative gap at every iteration, primal residual as a function of runtime and more)
- Integration tools: codacy, Travis CI, codecov
- Memory management verification using valgrind
- Makefile with new functionality: run doxygen, generate coverage reports, run unit tests, test using valgrind
- Improvement of linear algebra routines using manual loop unrolling
- Gitter chat