GitLab

v2.7.0 - HCubature in QuaLiKiz

Important changes:

• HCubature routines used as new default for QuaLiKiz 1D/2D integrations
• Rescaled contour integrals with Ne, changes results slightly (See !63)
• Moved to CeCILL-C license !72, now free to use in integrated frameworks (officially)

Other:

• Fix spurious hanging ifort/12 cases (See !71)
• QLK input and output derived types added
• Much, much more internal refactoring

Upgrade instructions:

git fetch
git checkout v2.7.0
git submodule update --init
make clean; make

If you use old run folders/parameters.json:

• Add absacc1: 0, absacc2: 0, integration_routine: 1 to your parameters.json

NB. This tag was moved from 8297e265 to 5deccef9

This release marks the last one of QuaLiKiz that is build-able with the PGI compiler. Later releases need features that are unavailable of buggy in PGI as of version 19.4. Only minor changes compared to last release:

• Improvements to CI at JET and on gitlab.com
• Updates to TUBS
• Updates to Pythontools

1. Various TUBS bugfixes
2. More graceful warnings regarding non-applicable inputs
3. Bugfix for rot_flag=2 when dimx scan is not a radial scan
4. Bugfix for scaling contour and Newton parameters: central contour width, contour minimum and minimum allowed contour maximum in imaginary axis, and scale used for Newton solver finite differences. They should scale with temperature, but didn't. Was only really noticeable for T < 500eV. Did not affect standard H-mode integrated modelling but perhaps important for L-mode edge

1. Added guards against division by small numbers in integrands. Reproducibility (albeit binary reproducibility with serial-only) now achieved in CI for all test cases.

2. Bugfix for PGI compiler (used on Heimdall for JINTRAC): -O2 or -O3 optimisation caused memory corruption. Workaround with -O1. Impacted calculation of simultaneous ITG+TEM

Includes fix for spurious Ti/Te multiplication of ion heat fluxes.

Most affects cases with fast ions, where the (wrong) ion heat flux is then on order of main ion heat flux.

Together with updates in tag v2.5.3, electrostatic fast ions are now correctly treated in QuaLiKiz

This release treats fast ions more consistently, see !39.

As such, QuaLiKiz can now be used in integrated modelling when running with fast ions. Runs using QuaLiKiz versions before this release should be redone!

This version is able to be compiled for all three compilers on heimdall as of JETTO-SANCO dd835cd4011717836ac0faf5774d7c249822e13a. The following changes have been made since v2.5.1:

• Bugfix in the definition of pinch angle in Krook operator
• Improvement of contour limits for negative temperature and density gradients
• Added compilation options for Heimdall
• Fixes for QLK build inside JETTO
• Input filtering on negative densities and temperatures

This release contains a large refactor of the integrants in all NAG integrals. A lot of integrands are combined into a single one, in which the specific integrant can be chosen with a flag. Specifically:

• 4bac2daf - Generalize trapints.f90
• e1fffe01 - Generalize passints.f90
• 0fbbfe8c - Generalize FLRterms.f90
• 96607b9e - Generalize mod_fluidsol.f90

As of now, MKL is contained in the QuaLiKiz docker 0c3a7ff6, and one small bug was found and fixed in Swap buggy nocol fk/fki lines.

This version is the first release after the full move to GitLab. This enabled the creation of automated regression checks (Continuous Integration, or CI). This in turn allowed for some bugs to be found, greatly increasing the binary-level reproducibility of QuaLiKiz. It should be stressed that these fixes have little physics impact! This release includes numerous improvements of the QuaLiKiz build in JETTO. Other notable changes

Reproducibility:

• Added simple MPI scheme with no OpenMP (new option in QLK interface)
• b898f736 Resolved DAVINT crashes on ifort12 by O2-->O1 optimisation
• b75db360 Bugfix ETG davint integrations
• 985065f7 Redefine arrays of ETG davint integration as [0 ETGscale]

Refactors:

• 206495da Ripped out separated passing FLR calculations. They have been superceded by inclusion directly within passints
• a15595dc Ripped old _cub integrands from calltrapints
• 365ac753 Ripped old _cub integrands from callpassints
• d54ace58 Ripped out FFkerotold from trapints
• c153a00c Refactored asymmetry with unified ecoefs calculation
• 3e295e0c Reduced verbosity if verbose = false

Physics fixes:

• ce851ff6 Bugfix for exotic parameter regimes for multiple solutions per contour
• cfafa1db Bugfix Nathanrule (ETG when gamETG/gamITG>sqrt(mi/me) ). Before was not doing anything
• 37440f07 Bugfix for kperp2 in saturation rule when including rotation
• 9999a77f Bugfix for contour limits in highly driven cases (e.g. L-mode edge)

JETTO-submodule-version QuaLiKiz

Minor aesthetic fixes compared to v2.4.4

This is the first version merged into JETTO development line as a submodule

Single QuaLiKiz version reference

This release merges together the three version of QuaLiKiz, QuaLiKiz-JETTO, QuaLiKiz-CEA and master.

JETTO 2018 reference version

This release is part of the effort to converge to a single canonical QuaLiKiz repository. This version gives identical results for the standard JETTO-QuaLiKiz regression test as the production version of QuaLiKiz inside JETTO v170817, v280818 and v121218 .

This version is included in JETTO as a git sub-module in JETTO feature branch fcasson_qlk_submodule2, commit 0c1d996f427ff0d98b28066ba9d99cc901038b88. (Branched off of JETTO develop commit ed21e4ac0247936103f06ccb3a9a3566e57bbe65 - See also JETTO merge request !119). For more information on the QuaLiKiz-JETTO merge see #29.

Low density trace ions fix

Fixed rounding errors for very low ni/ne (<1e-5) as seen below for trace D in the GA-Standard case. For impurities, it’s not as bad due to the Z^2 term, but can still impact e.g. W-transport lower than 1e-6-1e-7 level.

Detailed change log:

• Bugfix MPI file I/O; Copies of the same file were being written to the same location on disk
• Added write_primi flag to suppress output of primitives
• Extended disk I/O interface to also write 3D arrays
• Fixed name collision with JETTO in top-level Makefile
• Bugfix narrow poloidal eigenfunctions in asymmetry
• Redefined ITG regime kthetarhos<ETGk=2 to kthetarhos<=ETGk=2
• Various other tiny fixes

JETTO reference version

Recap of 2.4:

• Parallelization of standalone I/O and general saturation rules
• Modification of variable names for standalone I/O
• Python I/O tools for standalone mode
• Possibility to output separated ITG, TEM, and ETG fluxes

Specifically for 2.4.2:

• Bugfixes and completion of standalone outputs, following full regression testing

Open Source QuaLiKiz

This is the first open source version of QuaLiKiz!

JETTO/CEA reference version

This version is a carbon copy of the QLK-2.3.0-rc.1 tag from QuaLiKiz repository on the CEA SVN server. It is in sync with the QuaLiKiz version on the JETTO repository. In the near future, this will be merged with the master branch here, with both the Python + MATLAB I/O tools as submodules. The QuaLiKiz version JETTO will be a mirror of here, and thus we will converge to a single QuaLiKiz repo