It's almost January, so time for the annual diet. Let's include the
dicts backing our metadata instances. This cuts 46% of the memory use of
the .smpl and .cplx members themselves (not accounting for sizes of keys
and values), corresponding to about 10% of the total memory consumption.
For an "oe bake world", that translates to around 63 MB of memory with
the metalayers I have. It doesn't add any noticable time to the
initialization phase.
I admit it's a little hacky, but I think the code-to-comment ratio and
the gain should make up for that. It's a shame Python doesn't expose an
official knob to do this kind of thing.
It's almost January, so time for the annual diet. Let's include the dicts backing our metadata instances. This cuts 46% of the memory use of the .smpl and .cplx members themselves (not accounting for sizes of keys and values), corresponding to about 10% of the total memory consumption. For an "oe bake world", that translates to around 63 MB of memory with the metalayers I have. It doesn't add any noticable time to the initialization phase.
I admit it's a little hacky, but I think the code-to-comment ratio and the gain should make up for that. It's a shame Python doesn't expose an official knob to do this kind of thing.