|
From: | Gregory Heytings |
Subject: | Re: The Emacs master is much slower than the emacs-27 branch. |
Date: | Sat, 05 Dec 2020 12:55:19 +0000 |
User-agent: | Alpine 2.22 (NEB 394 2020-01-19) |
I don't know what happens on your computer (what is your platform/compiler?) but on mine (Debian GNU/Linux with the latest Linux kernel and the latest GCC) the effect on -O0 is almost the same as on -Og for this particular benchmark. I just tried again on the emacs-27 branch:-Og: ~67s -O0: ~63s -O1: ~26s -O2: ~18s -O3: ~17sSo when I want to compare Emacs 27 and Emacs 28 benchmarks, -Og builds are off limits. Now can I use O0 builds or do I have to use O1-O3 builds exclusively?
The emacs-27 branch since commit 365dad197bac5deec9244fd9c189d23c46c99b31 (Sep 4 2019) is about three times slower (at least on this benchmark, with GCC 10.2, on a 64-bit GNU/Linux) than the master branch since commit b9b32074cecb0e0d20cf527eadacbb0b7f5cced9 (Apr 7 2020) with -Og and -O0. With -O[1-3] it is not.
I'd guess that Eli will merge b9b32074cecb0e0d20cf527eadacbb0b7f5cced9 into emacs-27 very soon after the Emacs 27.2 release, and this slowdown will disappear.
[Prev in Thread] | Current Thread | [Next in Thread] |