This is the mail archive of the
mailing list for the GCC project.
Re: Compilation-time suggestion #2
- From: Segher Boessenkool <segher at kernel dot crashing dot org>
- To: Roger Sayle <roger at eyesopen dot com>
- Cc: gcc at gcc dot gnu dot org, Scott Robert Ladd <coyote at coyotegulch dot com>
- Date: Tue, 20 Jan 2004 20:46:01 +0100
- Subject: Re: Compilation-time suggestion #2
- References: <Pine.LNX.email@example.com>
On 19-jan-04, at 18:11, Roger Sayle wrote:
Does "-O2 -funroll-loops -ffast-math -fomit-frame-pointer" produce
a debuggable compiler that is any faster than the current default
-ffast-math shouldn't help any, as GCC doesn't use floating-point
math at all.
-fomit-frame-pointer supposedly makes non-debuggable binaries at
least on some targets (although -O2 is pretty much non-interactively-
debuggable already), but it should help quite a bit on those poor
register-starved architectures out there.
-funroll-loops doesn't help on my primary targets (powerpc*), but I
think it might help on some others (not by all that much, though).
The general idea of providing better compiler options for compiling
the compiler itself sounds fine though, of course :-)