BIG Table Lookups or interpolation

[tbenthompson]

future research question: how far can we push interpolation and table lookups?

• can we handle the entire Lei problem inside a table?
• there are four values at each interim.
• so naively, we might have something like a 50^12 table. which seems infeasible.

what about inside a tree instead of a table?

• HUGE parts of the tree of events will get pruned off.
• other large parts of that will be very very unlikely.
• so could we handle the 99% of common simulation paths with a table and then deal with the remaining few paths explicitly?
• this is easily answerable by tracking how many times we repeat the same sampling path in a full imprint analysis of lei. so… we should answer that at some point.

are there efficient batch tree lookup algorithms for gpus? almost certainly yes.

this is not exactly the same problem but is a nice tutorial on implementing a tree-based collision detection algo on the GPU: https://developer.nvidia.com/blog/thinking-parallel-part-ii-tree-traversal-gpu/