This is the mail archive of the
mailing list for the GCC project.
Re: [www] Document C++ -O0 improvements
- From: "Giovanni Bajo" <giovannibajo at libero dot it>
- To: "Bonzini Paolo" <paolo dot bonzini at lu dot unisi dot ch>,"Gerald Pfeifer" <gerald at pfeifer dot com>
- Cc: <gcc-patches at gcc dot gnu dot org>,"Bernardo Innocenti" <bernie at develer dot com>,"Paolo Bonzini" <bonzini at gnu dot org>
- Date: Fri, 29 Oct 2004 15:45:19 +0200
- Subject: Re: [www] Document C++ -O0 improvements
- References: <46267C3A67BB0845BA6115D53D4668C803C2C396@campus9.usilu.net>
Bonzini Paolo wrote:
>> ...though, given the quote above, isn't the statement a bit too
>> strong ("any previous version")?
> any 3.x version?
No, it is faster than 2.95 as well. I would argue that, in the worse cases, it
is about the same speed or just a tad slower, but for many others it is much
For compilations which are template-bound like PR8361, the gain is smaller, at
least for the average users. To me, for instance, this already means a lot
because we go almost the same speed *and* do two-phase template lookup instead
of just one-phase lookup (just to name one thing G++ 4.0 and 2.95 did not).
Moreover, to tell you everything, much of the point of this patch is to tell
users that we *are* doing progress even compared to 2.95. What I often hear
from the "field" is that most people do not know why they should upgrade from
2.95, and the only things they *all* know is that newer versions of GCC are
much slower. This is why, for instance, I believe that gcc-4.0/changes.html is
still totally incomplete: it does not say -- in user terms -- why we bumped the
big number. The paragraph about tree-ssa is just a list of techinical names
which mean nothing to non compiler developers. I am actually planning to
rewrite that adding many examples of code snippets that we now optimize better,
with side-by-side assembly comparison with 2.95 or 3.4.
The patch is now in, thanks Gerald!