Releases: modelon-community/Assimulo
Releases · modelon-community/Assimulo
Assimulo-3.6.0
- Added get_sundials_version function (import from assimulo.solvers.sundials).
- Fixed bug that Radau5ODE would not count state events in statistics.
- Removed tests from the Assimulo installation.
- Changed testing framework from
nose
topytest
.
Assimulo-3.5.2
Changes
-
Added support to build with Python 3.12.
-
Loosened restriction on scipy.
-
Fixed issue in setup.cfg with Python version requirement.
Full Changelog: Assimulo-3.5.1...Assimulo-3.5.2
Assimulo-3.5.1
* Fixed build with sundials 7.x
* Added "std=legacy" as default fortran compile flag.
This addresses compilation issues commonly experienced with more modern Fortran standards.
Assimulo-3.5.0
* Changed "numpy.float" to equivalent "numpy.float64" due to DeprecationWarnings in numpy >= 1.20.
* Improved examples with sparse jacobians by omitting the zeros in the jacobians.
* Upgraded to Cython >= 3.
* Removed deprecated build option "sundials_with_superlu", which is checked automatically.
* Added support for Sundials 7.x
Attached wheels are compiled on Ubuntu 22.04 (the linux wheel) and Windows 10 (the windows wheel).
Assimulo 3.4.3
This release contains minor changes to an Assimulo example for improved scipy compliance, in particular this only entails #65 .
Attached wheel for Python 3.8 is compiled on Ubuntu 20.04.
Attached wheel for Python 3.10 is compiled on Ubuntu 22.04, using
Package Version
--------------- -------
Assimulo 3.4.3
contourpy 1.2.0
cycler 0.12.1
Cython 0.29.37
fonttools 4.48.1
kiwisolver 1.4.5
matplotlib 3.8.3
numpy 1.26.4
packaging 23.2
pillow 10.2.0
pip 22.0.2
pyparsing 3.1.1
python-dateutil 2.8.2
scipy 1.12.0
setuptools 59.6.0
six 1.16.0
wheel 0.37.1
Assimulo 3.4.2
Changes since 3.4.2
- Resolved a build issue for SuperLUMT with Sundials 6.x
- Resolved an issue where one of the examples would result in a NameError when using scipy 1.10.1.
Full Changelog: Assimulo-3.4.1...Assimulo-3.4.2
Assimulo-3.4.1
- Restored functionality of CVode with 'external_event_detection' = True, which broke with 3.4.0.
- CVode now allows rtol = 0.
- Added support for CVode (without sensitivities) with non-negative relative tolerance vector.
Requires a SUNDIALS installation supporting this, see README.
Assimulo-3.4
* Removed Radau5ODE Fortran implementation.
This change should be seamless since the C implementation is numerically equivalent.
* Radau5 C with SuperLU now requires compilation with '-D__RADAU5_WITH_SUPERLU'
* Refactored Radau5 C implementation; moved to thirdparty/radau5/.
* Replaced event localization for explicit problems with an equivalent C implementation.
* Added missing counts on event indicator evaluations when setting problem data in solvers.
* Fixed Radau5 bug where simulation would not stop when time limit has been exceeded.
* Radau5(ODE) errors due to Exceptions in callbacks will now raise these Exceptions, rather than Radau5Error.
* KeyboardInterrupt in Radau5ODE solout callback (e.g., event_indicator) will properly terminate simulation.
* Added warning if a solver does not support setting time limits.
* Added so that solver statistics is stored in case of early abort for CVode and Radau5.
* Minor refactoring of Radau5 C implementation:
Now uses actual return value to return error codes. idid is return of solout callback.
* Separate fortran from C flags in setup.py
* Changed "numpy.bool" to equivalent "bool" due to DeprecationWarnings for numpy >= 1.20.
Assimulo-3.3
- Sundials 6.x port
- Radau5ODE and Radau5DAE now correctly terminate from unrecoverable errors in right-hand side and Jacobian calls.
- Added support for sparse LU decompositons (via SuperLU) in Radau5 solver.
This is only available with the C version of the Radau5 solver. - Changed attribute name Radau5ODE.solver to Radau5ODE.implementation.
Removed support for Radau5DAE.implementation = 'c' - Removed unused functionality from Radau5 C implementation: DAEs, 2nd order eq, banded matrix structures.
Assimulo-3.2.9
- Bugfix for C version of Radau5 solver