CP2K should be kept in good shape during its development.
Bugs that are fixed should stay fixed, features that are working should remain in this state.
In order to do so, an automated tester downloads, builds, and tests CP2K after each new commit
which changes the CP2K source code.
Correctness runs are parallel runs using a given configuration with full
error checking on a PC architecture. Even though we strive to reach the
above goal, the rapid development of CP2K does imply that not all
features are working or passing the high standard of Fortran conformance
we require during this testing. Failing tests are an indication of
possible problems, but as yet still quite common.
These reference results are not aimed to be used for porting/validating
the code on different platforms. In this case you may easily
expect to have several tests failing with a "WRONG RESULT" status.
Anyway they can still be used to check if errors between the regtest
and the reference are larger than a reasonable value,
clearly identifying a problem with the compiler or the libraries.
Usually an error of about 1.0E-8 or less is acceptable if (total) energies
are compared. Larger errors can be tolerated for other quantities like
NMR shifts (1.0E-5).
Latest CP2K source code
The most recent version of CP2K can be downloaded from