I have the following question. I’m currently learning the Verdi 2.1.3 scripts, I’m progressing quite slowly. We have an operating system for air quality forecast, using WRF and CMAQ models and we visualize the results with PAVE. I know that PAVE they don’t support it anymore.
We compiled all our system on a new machine with x86_64 GNU/Linux and INTEL ifort version 18.104.22.168, but PAVE doesn’t work there. Has anyone managed to launch and work with PAVE on newer Linux and Intel.
The default build now is for gfortran/gcc medium-memory model (I/O API
BIN=Linux2_x86_64gfort_medium; see https://cjcoats.github.io/ioapi/AVAIL.html#build), so that it supports arrays larger than 2 GB provided you have enough RAM and appropriate limits (limit stacksize unlimited; limit memoryuse unlimited). Computational parts of PAVE were extensively re-written to avoid “backwards” loop-nest/subscript-order mis-matches that significantly inhibited performance for large grids.
See https://cjcoats.github.io/pave/PaveManual.html for the manual. The Introduction section describes many of the changes to the new version.
Note that because of shared-library mis-matches and availability of support libraries, you frequently may need to compile some of your own support-libraries (etc.) and then compile PAVE itself.