Re: State of the overlay tree branch?

[Eli Zaretskii <eliz@gnu.org>]

> From: Sebastian Sturm <s.sturm@arkona-technologies.de>
> Date: Wed, 21 Mar 2018 01:36:38 +0100
> 
> this is the profiler report I get for the slow case (BTW, is there a way 
> to have the profiler resolve functions within line-number-at-pos?

Yes: load simple.el manually before running the benchmark.

> with perf, the ("self") time taken by buf_charpos_to_bytepos increases 
> from ~60% (fast case) to >98%. This is the diff generated by perf diff 
> <fast.perf> <slow.perf>:
> 
> # Event 'cycles'
> #
> # Baseline    Delta  Shared Object         Symbol 
> 
> # ........  .......  .................... 
> ..........................................
> #
>      57.77%  +40.30%  emacs-27.0.50         [.] buf_charpos_to_bytepos

This seems to confirm Stefan's guess that converting character
positions to byte positions takes most of the time, which might make
sense with a lot of overlays (because each overlay uses 2 markers).

Does your code call line-number-at-pos at random positions in the
buffer, or are the positions close to one another?  If the latter, you
might be better off calling count-lines directly, starting at the line
where you previously calculated the line number, instead of calling
line-number-at-pos, which always begins at point-min.