dfhoughton / evolusim Goto Github PK

TrigPairs are causing a huge amount of garbage collection work and memory thrashing. Pooling them and thus eliminating garbage collection has proven inadequate. We need to keep down the object count itself. We can do this by story geometries as subsequences within a number array and pooling offsets into this array for reuse.

the cell geometry modification makes it obsolete

The charts in the chart tabs now appear in random order.

This will speed up object creation a wee bit, since virtual machines that make a distinction between hash-y objects and object-y objects can create the latter faster, and perhaps access their properties and methods faster as well.

It would be cool when a creature is topicalized to see its stats.

working on the demo javascript is getting annoying

in Animal react method

This should make layout simpler and make it less disruptive to add still more adjustable options.

It would be handy to provide some explanation for every init parameter in the demo.

currently spurious nodes are being created

• stone size
• bounce factor

this will obviate the silly jig to get charting going

• meteor impact (all dead within radius of click)
• disease (random deaths in certain species?)
• overcast (tune sunlight)
• others?

It would be cool to have some counts that didn't rely on the Google charts.

they stop at 2000 clicks

get rid of thingMap and create a TrigPair pool

We can make geometric calculations happen in something more like linear time by dividing the universe into cells such that each cell knows its neighbor cells within the max distance radius. Each cell can also keep a map from radii to sub-lists of the master neighbor list, such that to find the entities influencing a particular animal we iterate over the proper cell indices given its maximum radius of influence and only lazily generate the geometric information as needed. We might even cache all the calculations relating accessible pairs of pixels, but I'd make that a separate issue.

It looks like the visual field is drawn pre-jitter.

The critters seem to prefer to swim northeast or southwest. This is likely due to some bug in the geometry calculation. Should write a unit test for the geometry calculation and then apply it to fix this bias.