bug#52753: 29.0.50; Printing long list-like structures fails

[Mattias Engdegård]

30 jan. 2022 kl. 10.16 skrev Ihor Radchenko <yantar92@gmail.com>:

> the speedup will only be relevant when
> explicitly allocated C arrays are used to store B-tree segments: all the
> tree data must be physically located within continuous segment of RAM
> address space.

Lisp vectors are contiguous in memory, and your keys are small integers and 
thus unboxed.

> People do say
> that splay trees can be good, but never have real world examples.

The trouble with splay trees seems to be that they require a very skewed access 
distribution to be worthwhile -- every read of something that isn't the latest 
accessed element becomes a tree modification. But that implies that the access 
pattern exhibits spatial and/or temporal locality, which makes caching an 
attractive alternative (explicit or those in your computer).