public inbox for gcc@gcc.gnu.org
 help / color / mirror / Atom feed
* A nice improvement in GCC-3.3
@ 2003-03-25 21:41 Art Haas
  2003-03-27  3:32 ` Richard Henderson
  0 siblings, 1 reply; 2+ messages in thread
From: Art Haas @ 2003-03-25 21:41 UTC (permalink / raw)
  To: gcc

Hi.

I'd like to report a large _speedup_ in the compiler when compiling the
latest parrot versions. Parrot is planned to be the run-time core of
Perl 6, and the file 'core_ops_cg.c' in parrot has always been trouble
to compile on my old, slow machine (i586). Memory requirements were
usually around 180-190 megs, and this file would often take between 10
and 15 minutes to compile. My latest build of GCC-3.3 taken from CVS
about two days ago though now compiles this file in under 5 minutes and
the memory needs topped out around 90 M. Much, much better. This file
makes heavy use of computed goto statements - perhaps this file can be a
test case for testing the compilers handling of such code?

For comparison, I just recompiled the file with GCC-3.2.3 (Debian), and
top reported memory use well over 100 M for much of the compile.
Apparently the parrot developers are also making this file a bit easier
to compile as well, but the improvements in GCC-3.3 in compiling this
file are so large I want to let the developers know and give an example
of a major improvement in the compiler.

Here's a link to parrot/perl6 for those interested ...

http://dev.perl.org/

My thanks to everyone working on GCC.

Art Haas
-- 
They that can give up essential liberty to obtain a little temporary safety
deserve neither liberty nor safety.
 -- Benjamin Franklin, Historical Review of Pennsylvania, 1759

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: A nice improvement in GCC-3.3
  2003-03-25 21:41 A nice improvement in GCC-3.3 Art Haas
@ 2003-03-27  3:32 ` Richard Henderson
  0 siblings, 0 replies; 2+ messages in thread
From: Richard Henderson @ 2003-03-27  3:32 UTC (permalink / raw)
  To: Art Haas; +Cc: gcc

On Tue, Mar 25, 2003 at 03:23:33PM -0600, Art Haas wrote:
> This file makes heavy use of computed goto statements - perhaps this
> file can be a test case for testing the compilers handling of such code?

We already have some.  Indeed, this fix was associated with PR 2001.
And before you ask, this can't be backported to 3.2 easily.  It relies
on infrastructure created during the 3.3 series.


r~

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2003-03-27  1:57 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-03-25 21:41 A nice improvement in GCC-3.3 Art Haas
2003-03-27  3:32 ` Richard Henderson

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).