Release Notes OVERFLOW 2.4

2.4b 4 November 2022

o Moved “makeall_cmake” script to “cmakeall”.
o Can specify FLAG_VERSION in cmakeall for different Intel compiler flags.
o Improved cmakeall and cmake.sys documentation.

Compiler options:
o Added -mcmodel=medium to gfortran flags to avoid link problems.
o Added -fp-model precise to Intel compiler flags to avoid 32-bit
floating-point errors.
o Updated Makelist (for compile testing).

o Added DELTA>1 eigenvalue smoothing for HLLE++ left-hand side
per suggestion of Josh Elliott (Boeing).
o Added dual number utility code and example.
o Allowed ILHS=16/17/18 when using multigrid.
o Improved use of “fold-over” grid subsets for DCF inverse maps
to eliminate orphans.
o Updated test/airfoil_drop_2d Mach=0.2 case to use NQT=104.

Bugs fixed:
o Fixed misplaced parentheses in NQT=103/104 SA diffusion terms.
o Fixed some XML parser problems (for reading Config.xml, Scenario.xml).
o Fixed rotor disk model problem with multigrid and grid splitting.
o Fixed two problems with SPLITM subsets and XINTOUT mixed mode,
causing extra fringe points from Pegasus5 to be shown in both
PLOT3D-style output files and TRIQ files.

2.4a 8 September 2022

Bugs fixed:
o Fixed GLOBAL_LINEAR_SOLVE to work with data surface grids.
o Fixed some $OMP PRIVATE lists, especially for low-Mach preconditioning
and/or C-grids.
o Fixed default GMP Constraint Center to be the body center of mass (CM).
o Improved some srb_axi and airfoil_drop_2d test case input files.
o Removed Baldwin-Barth from flat_plate* test cases.

2.4 20 July 2022

The big change for OVERFLOW Version 2.4 is the addition of the Parallel SSOR
linear solver, essentially a restricted additive Schwarz option, as described
in AIAA Paper # 2021-2747, which can be accessed through a new logical
variable GLOBAL_LINEAR_SOLVE in &GLOBAL. The basic idea is to make the
overset boundaries more implicit by communicating delta Q updates during the
linear solver process, doing a better job of solving the linear problem and
hopefully resulting in better nonlinear convergence, at the cost of more MPI
communication and higher memory usage. More details on how to use this new
option are discussed in doc/readmes/README.gls. The 2.4 release should
maintain backwards compatibility for most users, with slight changes to wall
functions, but for those users who have their own modifications to the code,
be aware that there were fairly significant changes at the OVERFL level to
accommodate inter-processor communication during the linear solve.

Further notes:

1: For startup and for higher speed flows, the new GLOBAL_LINEAR_SOLVE option
can be used in conjunction with ILHS = 6/7/8 to use more dissipative
Steger-Warming flux Jacobians as well as implicit physical and overset
boundary conditions; this should help users who have had difficulty getting
the Improved SSOR (ILHS = 26/27/28) to start. Be aware that starting
simulations using a constant DTPHYS may result in very high CFL numbers near
the wall and might destabilize the simulation.

2: The Improved and Parallel SSOR schemes have been adjusted to allow for
two-point boundary condition linearizations, improving convergence when using
symmetry planes.