Repository: llnl/sundials · Tag: v7.6.0 · Commit: ddf5dab · Released by: gardner48
SUNDIALS now has official Python interfaces! With this release, we are shipping a beta version of the sundials4py Python module (created with nanobind and litgen). sundials4py provides explicit interfaces to most features of SUNDIALS. See the Python section of the user guide for more information.
Added functions to CVODE(S) and IDA(S) to set the maximum number of inequality constraint failures in a step attempt (CVodeSetMaxNumConstraintFails and IDASetMaxNumConstraintFails) and to retrieve the total number of failed step attempts due to an inequality constraint violation (CVodeGetNumConstraintFails and IDAGetNumConstraintFails). As a result, constraint failures are no longer included in the number of step failures due to a solver failure (i.e., the values returned by CVodeGetNumStepSolveFails and IDAGetNumStepSolveFails). The functions CVodeGetNumConstraintCorrections and IDAGetNumConstraintCorrections were also added to retrieve the number of steps where the corrector was modified to satisfy an inequality constraint without failing the step.
The functions CVodeGetUserDataB and IDAGetUserDataB were added to CVODES and IDAS, respectively.
Fixed a bug in the CVODE(S) inequality constraint handling where the predicted state was used to compute the step size reduction factor which could lead to an insufficient reduction in the step size or, when the prediction violates the constraints, an infinitely large step size in the next step attempt (Issue #702).
On the initial time step with a user-supplied initial step size, ARKODE and CVODE(S) will now return ARK_TOO_CLOSE or CV_TOO_CLOSE, respectively, when the requested output time is the same as, or within numerical roundoff of, the initial time (Issue #722). Before a TOO_CLOSE error would only be returned when internally estimating the initial step size. In IDA(S), added a IDA_TOO_CLOSE return value for when the initial and output time are too close. Previously, IDA(S) would return IDA_ILL_INPUT.
Fixed a bug in ARKODE, CVODE(S), and IDA(S) where the linear solver counters were not reset on reinitialization until the next call to advance the system. As such, non-zero linear solver statistics could be returned if retrieving or printing linear solver counters between reinitialization and the next call to advance the system.
In CVODES and IDA, added missing return flag names to CVodeGetReturnFlagName and IDAGetReturnFlagName, respectively.
The SPRKStep module now accounts for zero coefficients in the SPRK tables, eliminating extraneous function evaluations.
A bug was fixed in KINSOL where the information logging function would always be called even when informational logging was disabled (Issue #801).
A bug preventing a user supplied SUNStepper_ResetCheckpointIndex function from being called was fixed.
The interface to Ginkgo batched linear solvers has been updated to fix build errors when using 64-bit index types (Issue #797). Note, only the batched dense matrix in Ginkgo is currently compatible with 64-bit indexing (as of Ginkgo 1.10).
The Kokkos N_Vector now properly handles unmanaged views. Previously, if a Kokkos N_Vector was created from an unmanaged view, the view would become a managed view and the data would be freed unexpectedly.
Fixed a CMake bug which resulted in static targets depending on shared targets when building both types of libraries in the same build (Issue #692).
Some installed Fortran example makefiles were not linking to sundials_fcore_mod and sundials_core libraries as they should be. This is now fixed.
The N_Vector_S typedef to N_Vector* is deprecated and will be removed in the next major release.
The CSC_MAT and CSR_MAT macros defined in sunmatrix_sparse.h will be removed in the next major release. Use SUN_CSC_MAT and SUN_CSR_MAT instead.
SUNDIALSFileOpen and SUNDIALSFileClose will be removed in the next major release. Use SUNFileOpen and SUNFileClose instead.
The Convert methods on the sundials::kokkos:Vector, sundials::kokkos::DenseMatrix, sundials::ginkgo::Matrix, sundials::ginkgo::BatchMatrix, sundials::kokkos::DenseLinearSolver, sundials::ginkgo::LinearSolver, and sundials::ginkgo::BatchLinearSolver classes have been deprecated and will be removed in the next major release. The method get, should be used instead.
This release has 15 assets:
Visit the release page to download them.
—
You are receiving this because you are watching this repository.
View it on GitHub or unsubscribe from all notifications for this repository.![]()
To unsubscribe from the SUNDIALS-USERS list: write to: mailto:SUNDIALS-USERS-...@LISTSERV.LLNL.GOV