v2.3.4

What's Changed

• Add option of task-local checkpointing. by @p-costa in #110
• Updated cuDecomp submodule. by @p-costa in #117

Full Changelog: v2.3.2...v2.3.4

v2.3.2

What's Changed

• Added neater format specifiers. by @p-costa in #101
• Simplify grid binary file I/O. by @p-costa in #102
• Fixed bug when changed the writting of the grid binary file. by @p-costa in #103
• Fix duct output. by @p-costa in #104
• Fix count in MPI_Allreduce statement. by @p-costa in #105
• Add noise to tbl case by default. by @p-costa in #106

Full Changelog: v2.3.1...v2.3.2

v2.3.1

What's Changed

• Fixed subroutine call in the scalar transport equation. by @p-costa in #92
• Improved support for ifx compiler. by @p-costa in #93
• Bump 2decomp submodule. by @p-costa in #94
• Added some extra descriptions. by @p-costa in #95
• Fix issue with namelist files. by @p-costa in #98

Full Changelog: v2.3.0...v2.3.1

v2.3.0

Summary

This release features a few changes, some that break backward compatibility, more specifically:

• We have changed the input file format, which is now based on Fortran namelists, see docs/INFO_INPUT.md for more details.

• We also introduced a gtype input parameter that sets the choice of grid mapping function (rather than being tied to the choice of the initial condition).

• Added more I/O flexibility (including an HDF5 backend).

What's Changed

• added more initial conditions to initgrid. by @p-costa in #76
• bump cudecomp submodule by @p-costa in #78
• fixed initialization of matlab field array by @p-costa in #79
• simplified OMP directives by @p-costa in #81
• fixed normalization of iop initial condition by @p-costa in #82
• minor change for consistency by @p-costa in #84
• Allow for a non-default grid binary file. by @p-costa in #88
• implemented and validated prescription of zero net acceleration by @p-costa in #67
• hdf5 I/O backend by @p-costa in #87
• Bump cudecomp submodule. by @p-costa in #89
• Fixed OOB issue in out1d when averaging along x or y. by @p-costa in #90
• More logical setup of the grid clustering by @p-costa in #85
• Namelist for input files by @p-costa in #86

Full Changelog: v2.2.2...v2.3.0

v2.2.2

Summary

This release features minor extensions that are not exercised at this moment, and a few minor bug fixes.

What's Changed

• added option for computing mean scalar fluxes by @p-costa in #62
• fix computation of retau for pdc case by @p-costa in #64
• Properly rescale initial velocity for iop case. by @p-costa in #65
• Extend integration of wall shear stresses to the different options. by @p-costa in #66
• temporary fix for field data I/O (for visualization, not checkpointing) on GPUs. by @p-costa in #68

Full Changelog: v2.2.1...v2.2.2

v2.2.1

Summary

This release fixes some two minor bugs, notably that the default dns.in file under src/ was not updated following the change in the input file format introduced in #54, which has now been fixed in #60.

What's Changed

• update description of dims by @p-costa in #58
• fixes potential OOB error when only one field is written by @p-costa in #59
• missed updating src/dns.in after changing input file by @p-costa in #60

Full Changelog: v2.2.0...v2.2.1

v2.2.0

Summary

This release features a change in the input file dns.in, which was simplified to avoid a common source of confusion. Instead of prescribing uref, lref, and rey (reference velocity and length scales, and Reynolds number) to calculate the fluid viscosity as visc = uref*lref/rey, we directly prescribe the inverse of the viscosity, visci (visc = visci**(-1)), so all inputs are dimensional (see the updated docs/INFO_INPUT.md file). Note that visci has the same value as the flow Reynolds number for all files under examples, as uref and lref were always equal to 1. This change is backwards-incompatible - former input files should be updated from v2.2.0 onward!

In addition, this release features a minor bugfix, extensions and polishing as detailed below.

What's Changed

• updated GPU implementation of a single scalar by @p-costa in #48
• Add symlink by @p-costa in #49
• polishing python readers by @p-costa in #50
• explicitly import functions from the cudafor module, not implicitly… by @p-costa in #51
• added support for Intel ifx by @p-costa in #53
• fixes bug in the GPU signal processing routines by @p-costa in #56
• changed input files to take visci, the inverse viscosity by @p-costa in #54
• explicitly add optional switch for asynchronous halo exchange on CPUs by @p-costa in #57

Full Changelog: v2.1.1...v2.2.0

v2.1.1

Summary

This release features very minor changes and catches up with changes in the external libraries.

What's Changed

• Minor changes in the eigenvalues by @p-costa in #44
• Bump 2decomp submodule by @p-costa in #45
• Leaner sanity check by @p-costa in #46
• Revert "ssh better for submodules than https" by @p-costa in #47

Full Changelog: v2.1.0...v2.1.1

v2.1.0

Summary

In this release, option SINGLE_PRECISION_POISSON has been removed. While solving the Poisson in lower precision equation yields excellent results for many benchmarks, most cases that benefit from this actually perform well when the whole calculation is performed in lower precision (see #42). Since this mode introduces significant complexity, it will be removed from the main branch for now in favor of a more readable code, a decision that can be reconsidered in the future. This option can still be explored in v2.0.1, and should be valuable for very high Reynolds numbers or other setups with extremely fine grids.

Changes

• Revert mixed precision mode by @p-costa in #42

Full Changelog: v2.0.1...v2.1.0

v2.0.1

Summary

This release features minor changes, to tag the current status before a more relevant release.

Changes

• Bump 2decomp by @p-costa in #31, #34, #35, #36
• added subroutine chk_poisson by @p-costa in #32
• acc_init should be called after setting the device by @p-costa in #33
• Minor changes by @p-costa in #41

Full Changelog: v2.0.0...v2.0.1