Open kaneko60 opened 3 years ago
Looks like some good work here! We currently have some QoS implementation embedded into a couple of our Dragonfly models. Utilizing a model net scheduler is an interesting idea.
I will say, merging this into mainline will be a process. We'll want to make sure that if there was simulation determinism prior to these changes that it is maintained. Same thing for parallelism/reverse computation.
Another is we'll want to find a way to distinguish between various QoS implementations in the code base so as not to confuse anyone if they, for instance, tried to configure a dragonfly QoS and accidentally utilize the new QoS model-net-scheduler.
And prior to merging, we'd like some sort of validation to show that this does what it's supposed to and documentation for how to use it. CODES hasn't always been great about documentation but I'm working to try and fix that going forward.
We've got a few other things in the pipeline that we're working on merging as well - including QoS related stuff so this will be an interesting merge process 😅
Thanks again, it's interesting work!
The qos features developed by Lee Savoie are merged.