Mail Archives: pgcc/1999/03/03/18:58:33
Marc Lehmann wrote:
>
> On Tue, Mar 02, 1999 at 05:13:25PM -0600, Jeffrey Hundstad wrote:
> > On 2 Mar, Dustin Marquess wrote:
> >
> >
> > > Also, is this old Slackware still a.out, or is it one of the
> > > ELF versions? ELF can be 1-5% slower than a.out..
> >
> > This ELF slowness can be eliminated by creating statically linked
> > executables, correct?
>
> Yes, its related to pic code used in (most) shared libraries.
When compiling with old gcc and pgcc, what does change the libs?
I can confirm the subject: beside that gcc generates smaller code,
it mostly runs faster, too (tested for gzip, bzip2, and some internal
stuff
on a PII and K6). Slightly better numbers I got when using a beast like
CC=pgcc -mamdk6
CFLAGS=-Wall -O6 -fomit-frame-pointer -fno-compare-elim -fopt-reg-stack
\
-frecombine
because these options break the compilition or coredump:
# pgcc fail:
# -fcompare-elim -floop-after-global -fcombine-222
all at the expense of a significant bigger binary.
This really desillusionized me about the capabilities of all this
funky optimizing stuff... Mostly they are not worth the bigger size
they need (and the time figuring out the best switches).
Real optimisation happens in the design phase of a project.
The compiler can help you getting the most of it!
Correct me, if I'm wrong.
Sorry, Marc. Please don't feel offended. As I'm an optimist, I will
look forward and see, how I can help you to get behind it...
Cheers
Hans-Peter
> --
> -----==- |
> ----==-- _ |
> ---==---(_)__ __ ____ __ Marc Lehmann +--
> --==---/ / _ \/ // /\ \/ / pcg AT goof DOT com |e|
> -=====/_/_//_/\_,_/ /_/\_\ XX11-RIPE --+
> The choice of a GNU generation |
> |
- Raw text -