Date: Tue, 2 Jul 2002 23:01:14 -0700 (PDT) From: Matthew Dillon <dillon@apollo.backplane.com> To: "David O'Brien" <dev-null@NUXI.com> Cc: FreeBSD current users <current@FreeBSD.ORG> Subject: Re: -current results (was something funny with soft updates?) Message-ID: <200207030601.g6361E5E010289@apollo.backplane.com> References: <Pine.BSF.4.21.0206281233500.75410-100000@InterJet.elischer.org> <200207020314.g623Eke5038019@apollo.backplane.com> <20020702164756.E70767@dragon.nuxi.com> <200207030106.g6316Rwp008905@apollo.backplane.com> <20020702182303.A94386@dragon.nuxi.com>
next in thread | previous in thread | raw e-mail | index | archive | help
:If you can quantify this, it is something we can pass on to the GCC
:folks. They are rather receptive right now due to wanting GCC 3.1.1 to
:be very high quality. Run-time of the compiler isn't anything that can
:be fixed right now -- but if you show how small (but not 3 line trivial)
:programs run slower that would be of interest.
-Matt
Matthew Dillon
<dillon@backplane.com>
Just play with a couple of interpreted languages, like csh, perl,
etc... run the same program compiled with GCC2 and again with GCC3
doing something cpu-bound, like a loop.
#!bin/csh -f
#
set history = 0
set i = 0
while ($i < 10000)
@ i = $i + 1
end
(csh built with GCC2, -O3)
test3:/home/dillon> time ./x.csh
0.832u 0.848s 0:01.68 99.4% 881+645k 0+0io 0pf+0w
test3:/home/dillon> time ./x.csh
0.926u 0.755s 0:01.68 99.4% 889+654k 0+0io 0pf+0w
(csh built with GCC3, -O3)
test3:/home/dillon> time ./x.csh
0.950u 0.693s 0:01.64 100.0% 900+648k 0+0io 0pf+0w
test3:/home/dillon> time ./x.csh
0.810u 0.833s 0:01.64 100.0% 900+648k 0+0io 0pf+0w
(csh built with GCC3, -O)
test3:/home/dillon> time ./x.csh
0.779u 0.904s 0:01.68 99.4% 760+641k 0+0io 0pf+0w
test3:/home/dillon> time ./x.csh
0.817u 0.864s 0:01.68 99.4% 760+641k 0+0io 0pf+0w
I also ran tests with some of my own code and got similar results.
GCC3-compiled binaries do not appear to improve the performance of
anything, at least not on a Pentium III, and yet we know from experience
the the absolute best commercial compilers for IA32 can get 50-100% more
performance (at least from the few postings to the lists where people
have compared timing results).
I get just about the same performance for GCC2 as I do for GCC3 in
the tests I've run so far. It makes me wonder what the hell GCC3 is
burning all that cpu *on*.
-Matt
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200207030601.g6361E5E010289>