Re: GCC 3.2.2 import -- questions
On 11-Feb-2003 Craig Rodrigues wrote: > On Mon, Feb 10, 2003 at 08:44:33PM -0500, Wesley Morgan wrote: >> that are supposedly fixed with 3.2.2... My question is, should I consider >> rebuilding my ports with this new compiler because of stability and/or >> speed improvements? Or is this point release not worth the effort. > > Speed improvements? No. gcc 3.2.2 is definitely slower than gcc 2.95. > There is a lot of arguing on the gcc mailing list right now about this, > but no concrete action to improve the situation yet. Umm, his question was to see if there were bugfixes in the Pentium 4 specific optimizations in the 3.2.1 to 3.2.2 GCC upgrade. Not related to 2.95.x at all. -- John Baldwin <[EMAIL PROTECTED]> <>< http://www.FreeBSD.org/~jhb/ "Power Users Use the Power to Serve!" - http://www.FreeBSD.org/ To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Tue, Feb 11, 2003 at 09:03:28PM -0500, Wesley Morgan wrote: > On Wed, 12 Feb 2003, leafy wrote: > > > > > > > Anders > > Yes I noticed it this morning too. > > The funny thing is that. If you use a non-P4 optmized GCC to compile lcms with P4 >opt, then it passes the test. But with a P4 opted GCC, it borks. Looks like P4 opted >GCC itself is bogus. > > That's odd. Does the FreeBSD build skill the stage2 compiler "rebuild"? I > thought the gcc build process tested itself against itself. FreeBSD doesn't use the gcc build process.. Kris msg52242/pgp0.pgp Description: PGP signature
Re: GCC 3.2.2 import -- questions
On Tue, Feb 11, 2003 at 09:03:28PM -0500, Wesley Morgan wrote: > On Wed, 12 Feb 2003, leafy wrote: > > > > > > > Anders > > Yes I noticed it this morning too. > > The funny thing is that. If you use a non-P4 optmized GCC to compile lcms with P4 >opt, then it passes the test. But with a P4 opted GCC, it borks. Looks like P4 opted >GCC itself is bogus. > > That's odd. Does the FreeBSD build skill the stage2 compiler "rebuild"? I > thought the gcc build process tested itself against itself. > >From personal observations, I would not use -march=p4 with gcc 3.2.x on my FreeBSD system. You're just asking for trouble. -- Steve To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Tue, Feb 11, 2003 at 09:03:28PM -0500, Wesley Morgan wrote: > > The funny thing is that. If you use a non-P4 optmized GCC to compile lcms with P4 >opt, then it passes the test. But with a P4 opted GCC, it borks. Looks like P4 opted >GCC itself is bogus. > > That's odd. Does the FreeBSD build skill the stage2 compiler "rebuild"? I > thought the gcc build process tested itself against itself. > Perhaps it missed some floating point test? which could be utilized by lcms. Jiawei Ye -- "Without the userland, the kernel is useless." --inspired by The Tao of Programming To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Wed, 12 Feb 2003, leafy wrote: > > > > Anders > Yes I noticed it this morning too. > The funny thing is that. If you use a non-P4 optmized GCC to compile lcms with P4 >opt, then it passes the test. But with a P4 opted GCC, it borks. Looks like P4 opted >GCC itself is bogus. That's odd. Does the FreeBSD build skill the stage2 compiler "rebuild"? I thought the gcc build process tested itself against itself. -- Hi! I'm a .signature virus! Copy me into your ~/.signature to help me spread! To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Tue, Feb 11, 2003 at 11:11:39PM +0100, Anders Andersson wrote: > Testing curves join ...failed! > *** Error code 1 > > Stop in /usr/ports/graphics/lcms/work/lcms-1.09/testbed. > *** Error code 1 > > So, the lcms port still fails with CPUTYPE=p4 and there seems to be other > issues still with CPUTYPE=p4 / SSE optimizations. > > Anders Yes I noticed it this morning too. The funny thing is that. If you use a non-P4 optmized GCC to compile lcms with P4 opt, then it passes the test. But with a P4 opted GCC, it borks. Looks like P4 opted GCC itself is bogus. Jiawei Ye -- "Without the userland, the kernel is useless." --inspired by The Tao of Programming To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Tue, Feb 11, 2003 at 10:14:58AM +0800, leafy wrote: > lcms post-build tests now finishes correctly with pentium4 optimizations. > And I have world with the p4 optimization with no ill-effact so far. No, it still fails. This is on a new world built with CPUTYPE?=p4 and then: 'portupgrade -f lcms' [snip] ranlib liblcms.a cd /usr/ports/graphics/lcms/work/lcms-1.09/src/../testbed && /usr/bin/env CFLAGS="-O -pipe -march=pentium4 -I../include" make -E CFLAGS test cc -O -pipe -march=pentium4 -I../include -c testcms.c cc -O -pipe -march=pentium4 -I../include testcms.o ../src/liblcms.a -o testcms - lm ./testcms little cms testbed. Ver 1.09 [build Feb 11 2003 23:06:14] Testing fixed point: 2.8848960205 = 2.8848 0.437499269828536 = 0.4374 Testing fixed scaling...pass. Testing curves join ...failed! *** Error code 1 Stop in /usr/ports/graphics/lcms/work/lcms-1.09/testbed. *** Error code 1 So, the lcms port still fails with CPUTYPE=p4 and there seems to be other issues still with CPUTYPE=p4 / SSE optimizations. Anders -- Anders Andersson To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
Hmmm, fails to build for me: FreeBSD asus 5.0-RELEASE-p1 FreeBSD 5.0-RELEASE-p1 #3: Mon Feb 10 10:39:34 CET 2003 root@asus:/usr/obj/usr/src/sys/ASUS i386 gmake[3]: Entering directory `/usr/ports/lang/gcc32/work/build/gcc' for d in libgcc; do \ if [ -d $d ]; then true; else /bin/sh .././..//gcc-3.2.2/gcc/mkinstalldirs $d; fi; \ done if [ -f stmp-dirs ]; then true; else touch stmp-dirs; fi gmake[3]: Leaving directory `/usr/ports/lang/gcc32/work/build/gcc' gmake[2]: Leaving directory `/usr/ports/lang/gcc32/work/build/gcc' gmake[1]: *** [configure-target-libstdc++-v3] Illegal instruction (core dumped) gmake[1]: Leaving directory `/usr/ports/lang/gcc32/work/build' gmake: *** [bootstrap] Error 2 *** Error code 2 Stop in /usr/ports/lang/gcc32. ** Command failed [exit code 1]: /usr/bin/script -qa /tmp/portupgrade97111.0 make ** Fix the problem and try again. ** The following packages were not installed or upgraded (*:skipped / !:failed) ! lang/gcc32(missing header) Any ideas? /Paul Wesley Morgan wrote: The import of gcc 3.2.2 brings a question to mind... Many people have mentioned problems with SSE / SSE2 instructions, optimizer problems etc that are supposedly fixed with 3.2.2... My question is, should I consider rebuilding my ports with this new compiler because of stability and/or speed improvements? Or is this point release not worth the effort. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Mon, 10 Feb 2003 22:48:31 -0500 Rahul Siddharthan <[EMAIL PROTECTED]> wrote: > To the OP -- any speed improvement from gcc 3.2.1 to 3.2.2 would > probably be marginal. If some particular port really bothers you with > its slow performance, try recompiling (though it's unlikely to help), > otherwise don't bother. If you really have a CPU intensive application you want to run on a P4 I suggest to use icc instead of gcc, as gcc does some things on a P4 which Intel has on the "don't do that on a P4" list. Bye, Alexander. -- Press every key to continue. http://www.Leidinger.net Alexander @ Leidinger.net GPG fingerprint = C518 BC70 E67F 143F BE91 3365 79E2 9C60 B006 3FE7 To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
Craig Rodrigues wrote: > There is a long thread on the GCC mailing list right now complaining > about compile-time speed regressions from 2.95.x, with many complaints > coming from Apple: I don't think the original poster was talking about compile-time speed. The running speed of applications is vastly improved under gcc 3.2.x, sometimes by 30% over gcc 2.95.x, in my experience. To the OP -- any speed improvement from gcc 3.2.1 to 3.2.2 would probably be marginal. If some particular port really bothers you with its slow performance, try recompiling (though it's unlikely to help), otherwise don't bother. - Rahul To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
At 9:43 PM -0500 2/10/03, Craig Rodrigues wrote: There is a long thread on the GCC mailing list right now complaining about compile-time speed regressions from 2.95.x, with many complaints coming from Apple: http://gcc.gnu.org/ml/gcc/2003-02/msg00558.html Whether these complaints lead to actual improvements is yet to be seen I do not follow that mailing list. Looking at the thread you pointed at, I see comments from Apple, openbsd developers, and someone in the linux world. Could someone who is regularly on that mailing list add a comment about the freebsd project's experience with switching from 2.95.x to 3.x? I'm the type of person who decided I had to buy a new machine after gcc 3.x went in, because I couldn't stand the slowdown of the new compiler. To me, the cost of that was $1500 and a fair amount of my spare time to shuffle machines around. Sounds like a good reason to complain, but I wouldn't want to jump into the gcc mailing list if someone from FreeBSD is already covering compile-time performance. -- Garance Alistair Drosehn= [EMAIL PROTECTED] Senior Systems Programmer or [EMAIL PROTECTED] Rensselaer Polytechnic Instituteor [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Mon, Feb 10, 2003 at 09:50:06PM -0500, Scott Dodson wrote: > Excellent, > > Which optimization strings are you using in make.conf if you don't mind? > > -- > Scott Plain cflags and cxxflags taken from /usr/share/examples/etc/make.conf just modify the CPUTYPE as p4 Cheers, Jiawei Ye -- "Without the userland, the kernel is useless." --inspired by The Tao of Programming To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Mon, 10 Feb 2003, Craig Rodrigues wrote: > Many people are upgrading from 4.7.x to -current for the first > time these days, so I thought I would mention that for reference. > > GCC 3.2.2 was an incremental bugfix over GCC 3.2.1, and there are no > earth-shattering performance improvements. I have not done > such benchmarking myself, so have no empirical evidence to support this, > but I am basing this on the traffic I have been watching on the > GCC mailing list, and by reading the release notes > at http://gcc.gnu.org/gcc-3.2/changes.html . Well what I am really interested in is whether or not higher levels of optimization are more reliable now than before. Previously we have been warned against using many of the CPU specific optimizations, especially for the pentium 4, and the release notes offer little to support any conclusions... So without digging through mountains of GCC mailing list archives... Are these optimizations SAFER now? -- Hi! I'm a .signature virus! Copy me into your ~/.signature to help me spread! To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Mon, Feb 10, 2003 at 08:06:19PM -0600, Juli Mallett wrote: > I would assume the OP meant relative to the previous version of GCC in > tree. Current hasn't been 2.95.x for some time. Many people are upgrading from 4.7.x to -current for the first time these days, so I thought I would mention that for reference. GCC 3.2.2 was an incremental bugfix over GCC 3.2.1, and there are no earth-shattering performance improvements. I have not done such benchmarking myself, so have no empirical evidence to support this, but I am basing this on the traffic I have been watching on the GCC mailing list, and by reading the release notes at http://gcc.gnu.org/gcc-3.2/changes.html . There is a long thread on the GCC mailing list right now complaining about compile-time speed regressions from 2.95.x, with many complaints coming from Apple: http://gcc.gnu.org/ml/gcc/2003-02/msg00558.html Whether these complaints lead to actual improvements is yet to be seen -- Craig Rodrigues http://home.attbi.com/~rodrigc [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Mon, Feb 10, 2003 at 08:44:33PM -0500, Wesley Morgan wrote: > The import of gcc 3.2.2 brings a question to mind... Many people have > mentioned problems with SSE / SSE2 instructions, optimizer problems etc > that are supposedly fixed with 3.2.2... My question is, should I consider > rebuilding my ports with this new compiler because of stability and/or > speed improvements? Or is this point release not worth the effort. lcms post-build tests now finishes correctly with pentium4 optimizations. And I have world with the p4 optimization with no ill-effact so far. Jiawei Ye -- "Without the userland, the kernel is useless." --inspired by The Tao of Programming To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
* De: Craig Rodrigues <[EMAIL PROTECTED]> [ Data: 2003-02-10 ] [ Subjecte: Re: GCC 3.2.2 import -- questions ] > On Mon, Feb 10, 2003 at 08:44:33PM -0500, Wesley Morgan wrote: > > that are supposedly fixed with 3.2.2... My question is, should I consider > > rebuilding my ports with this new compiler because of stability and/or > > speed improvements? Or is this point release not worth the effort. > > Speed improvements? No. gcc 3.2.2 is definitely slower than gcc 2.95. > There is a lot of arguing on the gcc mailing list right now about this, > but no concrete action to improve the situation yet. I would assume the OP meant relative to the previous version of GCC in tree. Current hasn't been 2.95.x for some time. -- Juli Mallett <[EMAIL PROTECTED]> AIM: BSDFlata -- IRC: juli on EFnet OpenDarwin, Mono, FreeBSD Developer ircd-hybrid Developer, EFnet addict FreeBSD on MIPS-Anything on FreeBSD To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: GCC 3.2.2 import -- questions
On Mon, Feb 10, 2003 at 08:44:33PM -0500, Wesley Morgan wrote: > that are supposedly fixed with 3.2.2... My question is, should I consider > rebuilding my ports with this new compiler because of stability and/or > speed improvements? Or is this point release not worth the effort. Speed improvements? No. gcc 3.2.2 is definitely slower than gcc 2.95. There is a lot of arguing on the gcc mailing list right now about this, but no concrete action to improve the situation yet. Stability improvements? For a list of bug fixes see: http://gcc.gnu.org/gcc-3.2/changes.html and decide for yourself. -- Craig Rodrigues http://home.attbi.com/~rodrigc [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
GCC 3.2.2 import -- questions
The import of gcc 3.2.2 brings a question to mind... Many people have mentioned problems with SSE / SSE2 instructions, optimizer problems etc that are supposedly fixed with 3.2.2... My question is, should I consider rebuilding my ports with this new compiler because of stability and/or speed improvements? Or is this point release not worth the effort. -- Hi! I'm a .signature virus! Copy me into your ~/.signature to help me spread! To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message