I was just thinking whether it made sense for us to be cashing all the things, as we do now. In order to answer that I'd want to know for each thing orca is keeping track of:
How many times did orca generate this thing?
How many times did orca use a cached copy of this thing?
How much time was spent in generating this thing?
How much ram was used to hang on to the cached copy of this thing?
Things I could diagnose with this data from after a full model run include:
Things that are never used. Consider removing from the code.
Things that are never retrieved from the cache. Consider not caching that thing.
Things that are fast to compute. Consider not caching that thing.
I was just thinking whether it made sense for us to be cashing all the things, as we do now. In order to answer that I'd want to know for each thing orca is keeping track of:
Things I could diagnose with this data from after a full model run include: