Re: The Emacs master is much slower than the emacs-27 branch.

[Ergus]

On Sat, Dec 05, 2020 at 04:29:02PM +0200, Eli Zaretskii wrote:
> > From: "Basil L. Contovounesios" <contovob@tcd.ie>
> > Cc: ghe@sdf.org,  acm@muc.de,  stephen.berman@gmx.net,  emacs-devel@gnu.org
> > Date: Sat, 05 Dec 2020 13:50:51 +0000
> >
> >   $ for t in /tmp/timings*.txt; do awk '{s+=$1}END{print FILENAME,s/NR}' "$t"; 
> > done
> >   /tmp/timings-27.txt 28.3246
> >   /tmp/timings-28.txt 31.7968
>
> Thanks, so this indicates a 12% slowdown in Emacs 28.

> Which isn't too bad, I guess.
IMHO it depends. If we know the source of the slowdown (for example, a
change introduced like using a new library, a new mode enabled, changes
in rendering/display engine or so) and the change worth it, then maybe
12% is absolutely fine.

But if the slowdown is somehow "unjustified" then 12% is not negligible
at all. Because It could be produced by wrong assumptions and get worth
in the future and harder to detect (like for example excessive GC, extra
redisplay, bad optimizations).

The best approach is maybe what Gregory did. Bisect the history to find
the source of the slowdown.

Actually in my work we have a log where we register the changes and the
commits hashes that impacted performance more than 5-7% either to
improve or degrade it.

Best,
Ergus