You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@kaizhangpnl mentioned at our monthly meeting today the importance of printing out diagnostic quantities for helping us understand problems we encounter. We haven't really discussed how to do this yet. I have some ideas about how we might do this in a GPU environment. If anyone else is interested in this topic, please add your thoughts to this issue.
@kaizhangpnl, maybe we should schedule a meeting to discuss this. I can put together some simple examples of how we might support "debugging prints", and you can see whether you think they are sufficient.
The text was updated successfully, but these errors were encountered:
Hi @jeff-cohere, sorry for the delay. Thank you for taking care of this request. It would be nice to know how C++ in MAMXX handles debugging prints. This issue is perhaps even more important for initialization routines, where we often print out information about global and local tracers indices, aerosols properties, input files, and etc. If it's as simple as Fortran write, we probably could just follow the original code. Otherwise, it's better have to some examples available for the modelers to follow. What do you think? I'd be happy to have a further discussion with you if needed. Thanks!
Hey Kai. Maybe we should chat a bit about specifics. With the introduction of GPUs, certain styles of printing debugging output become less convenient (though they may still be possible). If we can agree on a handful of ways to print quantities, I think it would be easy to make such a capability available in C++.
@kaizhangpnl mentioned at our monthly meeting today the importance of printing out diagnostic quantities for helping us understand problems we encounter. We haven't really discussed how to do this yet. I have some ideas about how we might do this in a GPU environment. If anyone else is interested in this topic, please add your thoughts to this issue.
@kaizhangpnl, maybe we should schedule a meeting to discuss this. I can put together some simple examples of how we might support "debugging prints", and you can see whether you think they are sufficient.
The text was updated successfully, but these errors were encountered: