Joerg Henrichs
Joerg Henrichs
At this stage, an allocate statement becomes a code block. For the kernel-extraction driver creation, direct support for creating allocate statements would simplify the code (which currently parses a Fortran...
Quoting @arporter : > Following our discussion on Teams, I had a play and it seems the warnings are to do with classes contained within classes. Therefore, contrary to what...
In #1778 a new replace_induction_variable transformation was introduced. We need to find a way to best integrate this into the dependency analysis tools. Some issue to consider: - At this...
The newly integrated message system in the dependency analysis (#1693) stores additional information besides an error/warning code and the message strings. Initially these additional messages are the variable names involved,...
In `domain/lfric/psyir.py` many classes and variables are dynamically created. This causes several problems: - pylint warnings (classes are not found statically, exec is dangerous) - usability issues (you can't search...
At this stage a driver still need to be linked with the infrastructure library, since the kernel includes the kernel meta data, e.g.: ``` use kernel_mod use argument_mod ... type,...
While trying to manually fix up an LFRic driver for a gungho run, I ran into the problem that the kernel used a function that used some constants and configuration...
At the moment calls are just stored as CodeBlocks, which means they can't be further analysed (e.g. if no information about the calls is available, the variable access API still...
At the moment it is possible to read in a 4-d variable using a 2-d array (the NetCDF read code will ignore that there are sizes for dimensions 3 and...
This adds support to create PSyIR expressions for all arguments in an LFRic kernel call. The old way (providing the arguments as strings) is still maintained. This PR declares all...