Re: gcc4: missing atomic builtins?
Corinna Vinschen wrote: > The last i586 was built in 1999 and had a 300 MHz clock rate. The i686 > has been introduced in 1995 and in 1999 the P3 had a minimum clock rate > of 450 MHz. > > I don't think we really need a default compiler and a default installation > sticking to the old i586 design any longer. And, btw, is there any > Linux distro still using a i586 compiler? > > If there are really still i586 users out there, hey, it's all available > in sources. Right. I'll set i686 as the default arch. If anyone really does want to run Cygwin on an i586, they will be SOL, as it's entirely possible they won't be able to run the DLL or anything because everything in the distro will use instructions that their CPU doesn't handle; they'll have to stick with 1.5, or cross-boostrap the entire distro themselves starting from scratch. To Andy's previous post: I think the documentation suggests we want to use -mtune=generic for the best results across the broadest range of CPUs. At the moment, generic and i686 are basically the same, but since -march sets the *minimum* level CPU whereas mtune is aimed at the *typical* CPU (since it will only pessimise, not break anything if incorrct) I think it makes sense to anchor -march and let -mtune float. cheers, DaveK -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: gcc4: missing atomic builtins?
Hello,
* On Mon, Mar 30, 2009 at 10:43:36AM +0200 Corinna Vinschen wrote:
> And, btw, is there any
> Linux distro still using a i586 compiler?
I am happiln running Debian Etch ("oldstable") on a Am486DX4-100 with a
2.6.18 kernel. I am not completely sure if Lenny ("stable") still
supports it, though.
So: Yes, there a distros that still support such old hardware.
Regards,
Spiro.
--
Spiro R. Trikaliotis http://opencbm.sf.net/
http://www.trikaliotis.net/ http://www.viceteam.org/
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
Re: gcc4: missing atomic builtins?
On Mar 29 09:55, Andy Koppe wrote: > > Well, I guess the question is, what's the minimum level of CPU we want to > > support, and I can configure the next version --with-arch that level, and as > > long as it's >=486 we'll be ok. I think the question is mainly "686, or is > > there any reason to have 586 as the minimum supported CPU"? > > Windows 2000 and XP only require 586. Can't imagine many people using > them on a Pentium machine though, and the 686 CMOV instruction would > be nice to have. Also, should -with-tune be set to something recent? The last i586 was built in 1999 and had a 300 MHz clock rate. The i686 has been introduced in 1995 and in 1999 the P3 had a minimum clock rate of 450 MHz. I don't think we really need a default compiler and a default installation sticking to the old i586 design any longer. And, btw, is there any Linux distro still using a i586 compiler? If there are really still i586 users out there, hey, it's all available in sources. Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Project Co-Leader cygwin AT cygwin DOT com Red Hat -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: gcc4: missing atomic builtins?
> Well, I guess the question is, what's the minimum level of CPU we want to > support, and I can configure the next version --with-arch that level, and as > long as it's >=486 we'll be ok. I think the question is mainly "686, or is > there any reason to have 586 as the minimum supported CPU"? Windows 2000 and XP only require 586. Can't imagine many people using them on a Pentium machine though, and the 686 CMOV instruction would be nice to have. Also, should -with-tune be set to something recent? Andy -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Re: gcc4: missing atomic builtins?
Yaakov (Cygwin/X) wrote: > Building djvulibre-3.5.21 with gcc4, while linking libdjvulibre: > > .libs/atomic.o:atomic.cpp:(.text+0x3b): undefined reference to > `___sync_bool_compare_and_swap_4' > .libs/atomic.o:atomic.cpp:(.text+0x65): undefined reference to > `___sync_add_and_fetch_4' > > Some googling shows that the culprit *may be* gcc4: > > http://sourceware.org/ml/crossgcc/2008-06/msg00022.html > > Any ideas? Well, I guess the question is, what's the minimum level of CPU we want to support, and I can configure the next version --with-arch that level, and as long as it's >=486 we'll be ok. I think the question is mainly "686, or is there any reason to have 586 as the minimum supported CPU"? (Not that < [56]86 users couldn't add the -march= flag to their builds, but this will determine what minimum level of cpu all the default code is generated for, and as most packages will be built with the default, we run the risk of having things in the distro that they can't install and run successfully from binaries.) cheers, DaveK -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
gcc4: missing atomic builtins?
-BEGIN PGP SIGNED MESSAGE- Hash: SHA256 Dave, Building djvulibre-3.5.21 with gcc4, while linking libdjvulibre: .libs/atomic.o:atomic.cpp:(.text+0x3b): undefined reference to `___sync_bool_compare_and_swap_4' .libs/atomic.o:atomic.cpp:(.text+0x65): undefined reference to `___sync_add_and_fetch_4' Some googling shows that the culprit *may be* gcc4: http://sourceware.org/ml/crossgcc/2008-06/msg00022.html Any ideas? Yaakov -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.9 (Cygwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iEYEAREIAAYFAknNMlwACgkQpiWmPGlmQSMzYwCg4EVtxHSvT98JxVKq4LLdXio6 VhMAmgJxWtOTYCeLGay2ZG2lN7B/0yJP =RJil -END PGP SIGNATURE- -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
