This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: gcc compile-time performance
- From: dewar at gnat dot com (Robert Dewar)
- To: davem at redhat dot com, degger at fhm dot edu
- Cc: gcc at gcc dot gnu dot org, scott at coyotegulch dot com
- Date: Sun, 19 May 2002 07:12:18 -0400 (EDT)
- Subject: Re: gcc compile-time performance
> I'm still convinced that the more demanding CPUs nowaday are a huge
> factor in the slowness of GCC, be it directly or indirectly. For
> instance some transformations wouldn't make sense on a simple minded CPU
> like a 8051 because (nearly) all instructions take the same time and
> branches are no issue there.
Well sure if you go back to processors where branches are not an issue
(or even loads and stores) then you are in a different world, but we have
been in the new world of processors where thisw is not the case for a long
time now, and certainly this does not account for the 2.8 to 3.1 slowdown
of a factor of 2 or more. The EPIC stuff is new, but
a) there is no reason that this should slow things down
b) not much of the specialized epic stuff is in yet