search

compdyn / partmc

Particle-resolved stochastic atmospheric aerosol model
http://lagrange.mechse.illinois.edu/partmc/
GNU General Public License v2.0
27 stars 15 forks source link

Develop 114 adapt gpu #127

Closed cguzman95 closed 4 years ago

cguzman95 commented 4 years ago

Adaptation of calc_deriv and calc_jac are done. Some reactions are not computed on GPU and calculated on CPU after. Unit tests working fine with GPU flag ON. Missing test it with some mechanisms that use the non-implemented reactions (HL, SIMPOL, aqueous and condensed_arrhenius) and other implemented reactions to ensure all the contributions are added correctly.

NOTE: There are a few changes in .f files like camp_core and rxn_data. These changes are only to ensure update_data works by setting updating only the first cell. I encourage to merge also this changes as provisional until we correctly update all cells.

mattldawson commented 4 years ago

one of the recent changes to the update data objects is to allow them to update any cell - i think this is already working in chem_mod (i’ll check tomorrow)

cguzman95 commented 4 years ago

Nice - I will check it also (and fix the compilation errors caused by forgetting to test DEBUG flag)

cguzman95 commented 4 years ago

There is a bug with new unit tests GPU - I'm fixing that in #126

mattldawson commented 4 years ago

that's cool - do you still want me to merge this?

cguzman95 commented 4 years ago

Yes, please. And remember to check my big comment on #92 when you can

mattldawson commented 4 years ago

ok, I may not get to that for a few days - I need to focus on the MONARCH implementation, but I'll start going through the list as I have time

cguzman95 commented 4 years ago

Ok, no worries - I will fix the GPU issues on #126 alongside correct testing, and go with documentation and other things, so I have entertainment xD

mattldawson commented 4 years ago

@cguzman95 - do you want to keep this branch open after the merge or should I delete it?