Re: emacs rendering comparisson between emacs23 and emacs26.3

[Eli Zaretskii]

> From: Richard Stallman <address@hidden>
> Date: Fri, 27 Mar 2020 22:40:47 -0400
> Cc: address@hidden, address@hidden
> 
> To tell the people with the strongest commitment to freedom to use an
> old version of Emacs is to choose failure.  We need to re-optimize the
> Emacs master so it gives good performance on those machines.

That is of course the ideal we should always strive to, but it's
unreasonable to expect newer versions of Emacs to stay as fast as the
oder ones: we do add functionality, and some of that eats up CPU
cycles.  We do try to optimize performance, and occasionally do make
newer versions faster (e.g., Emacs 26 is about 30% faster than Emacs
25 on my machine, when scrolling through a large C file), but those
are exceptions rather than the rule, because significant speedups are
only possible by redesign and reimplementation of core functions,
something we do rarely.

> Suggesting that we disable certain features is ok.

Then my suggestion is to disable font-lock: that makes scrolling
through xdisp.c in Emacs 26.3 8 times faster, and almost as fast as
Emacs 23.4.

> Regarding C++ indentation, I have two suggestions.
> 
> * We could support a faster indentation mode for C code
> and for C++ files that use a limited repertoire of C++ constructs.
> 
> * Maybe we can come up with additional tricks comparable to
> {-in-column-0 to optimize parsing.  Emacs users would put these things
> into source files so as to get better peformance.

I think the way to increase performance of C/C++ fontification,
indentation, etc. is to use syntactic analysis in C (or another
similarly fast compiled language).  Several programs/libraries are
available out there for this purpose, and we should encourage
interested individuals to work on making these used by our major modes
that support programming languages.  The time we could do that with
ad-hoc regexps and the likes of syntax-ppss has gone, IMO.