[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
bug#17806: 24.4.50; Branch emacs-24 fails to beuild on Debian
From: |
Eli Zaretskii |
Subject: |
bug#17806: 24.4.50; Branch emacs-24 fails to beuild on Debian |
Date: |
Tue, 24 Jun 2014 19:39:15 +0300 |
> Date: Tue, 24 Jun 2014 09:27:50 -0700
> From: Paul Eggert <eggert@cs.ucla.edu>
> Cc: 17806@debbugs.gnu.org
>
> Dmitry Antipov wrote:
> > RedHat's compiler engineer claims that LTO has made substantial progress
> > in gcc-4.9: http://vmakarov.fedorapeople.org/spec/2014/2014.html
>
> Those comments say that GCC 4.9.0 compiles much faster than GCC 4.8.3
> when LTO is used. And that's understandable: in 4.9.0, -flto defaults
> to -fno-fat-lto-objects. Unfortunately, this is the change that broke
> the Emacs build. Emacs has worked around the problem by compiling with
> -ffat-lto-objects, but this negates the compile-time performance
> advantage of 4.9.0.
>
> 4.9.0 has not made significant improvements in run-time performance over
> 4.8.3 when LTO is used, and for Emacs LTO has only minor (and sometimes
> even negative) run-time performance improvements compared to non-LTO.
I wonder if we should say something about this in PROBLEMS. After
all, LTO is the latest buzz-hot feature, so people will probably try
it with Emacs.