Version 3.23 (code name "Ethel Bellamy") is the new Long Term Support (LTS) release after version 3.16. It features massive benefits in terms of performance, expecially for large models (for instance the USA classical PSHA model is 3-4 times faster than it was in version 3.16, while the EUR event based model is even faster) and many new features.
The complete set of changes is listed in the changelog:
https://github.com/gem/oq-engine/blob/engine-3.23/debian/changelogSummaries of the most recent changes are reported here:
https://docs.openquake.org/oq-engine/master/manual/release-notes/whats-new-3.17.htmlhttps://docs.openquake.org/oq-engine/master/manual/release-notes/whats-new-3.18.htmlhttps://docs.openquake.org/oq-engine/master/manual/release-notes/whats-new-3.19.htmlhttps://docs.openquake.org/oq-engine/master/manual/release-notes/whats-new-3.20.htmlhttps://docs.openquake.org/oq-engine/master/manual/release-notes/whats-new-3.21.htmlhttps://docs.openquake.org/oq-engine/master/manual/release-notes/whats-new-3.22.htmlhttps://docs.openquake.org/oq-engine/master/manual/release-notes/whats-new-3.23.htmlYou can install on all supported platforms (Linux, Windows, macOS) by
using the universal installer:
https://github.com/gem/oq-engine/blob/engine-3.23/doc/installing/universal.mdWindows user may prefer the native installer:
https://downloads.openquake.org/pkgs/windows/oq-engine/OpenQuake_Engine_3.23.0-1-no-toolkit.exeThe OpenQuake engine manual is available here:
https://docs.openquake.org/oq-engine/3.23/manual/getting-started/index.htmlBefore installing the new release you should uninstall the previous release, if you have one. As always, before upgrading, you should export any output of interests, since exporting an output computed with an older version of the engine may raise an error. Moreover, it is generally not possible to run a risk calculation starting from a hazard computed with an older version.
The results generated by version 3.23 are not expected to be identical to the results generated by version 3.16. In particular in event based calculations we changed the rupture filtering procedure and the random seed generation mechanism, so different ruptures will be sampled. The resulting GMFs will be statistically equivalent, but not identical. Even in classical calculations, improvements in the `ps_grid_spacing` approximation or bug fixes in specific GMPEs can cause slightly different results to be generated.
That's all, please let us know if you have any trouble with the upgrade.
Michele Simionato on behalf of the OpenQuake Team