[Fink-users] CPU-specific tuning in Fink Source Distributions
Hi, When using the Fink source distribution to install packages, are cpu-specific optimizations triggered? Will packages build by Fink for i386 run on any other i386-compatible mac? I'm using MacOS 10.6 on a Core2 Duo and I setup Fink to build programs for i386. Can I safely copy my system to a Core i5 laptop? Best wishes, Renaud. -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users
Re: [Fink-users] CPU-specific tuning in Fink Source Distributions
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 8/18/10 11:50 AM, Renaud Detry wrote: Hi, When using the Fink source distribution to install packages, are cpu-specific optimizations triggered? No. Will packages build by Fink for i386 run on any other i386-compatible mac? Yes. I'm using MacOS 10.6 on a Core2 Duo and I setup Fink to build programs for i386. Can I safely copy my system to a Core i5 laptop? Best wishes, Renaud. As far as we know, yes--there have been no reports of any incompatibilities, at any rate. - -- Alexander Hansen Fink User Liaison -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.10 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkxsDKQACgkQB8UpO3rKjQ8f2ACffug2JBihgFboQMUfeM4bgvNH +gEAn2s0CcKOhvWR/IE6Dy8DUYo8/MLJ =qsAz -END PGP SIGNATURE- -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users
Re: [Fink-users] CPU-specific tuning in Fink Source Distributions
On Aug 18, 2010, at 9:39 AM, Alexander Hansen wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 8/18/10 11:50 AM, Renaud Detry wrote: Hi, When using the Fink source distribution to install packages, are cpu-specific optimizations triggered? No. There is one exception to this that I'm aware of: the gmp package does cpu-specific optimization at compile time. But the cpu-specific optimizations tend to be upward-compatible, so if you compile this package on your least-capable machine, it should run on the others. -- Dave -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users
Re: [Fink-users] CPU-specific tuning in Fink Source Distributions
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 8/18/10 12:42 PM, David R. Morrison wrote: On Aug 18, 2010, at 9:39 AM, Alexander Hansen wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 8/18/10 11:50 AM, Renaud Detry wrote: Hi, When using the Fink source distribution to install packages, are cpu-specific optimizations triggered? No. There is one exception to this that I'm aware of: the gmp package does cpu-specific optimization at compile time. But the cpu-specific optimizations tend to be upward-compatible, so if you compile this package on your least-capable machine, it should run on the others. -- Dave Ah right, forgot about that. libmpfrN and fftw3 also seem to do something along those lines too--at least they do bad things on my G4 when I use a binary from a G5. :-) - -- Alexander Hansen Fink User Liaison -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.10 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkxsDi0ACgkQB8UpO3rKjQ/ZVQCgn1nUDwzGKDEmadzuWV/Dp4pC kDAAniIV73CmyAX65BqpjDCLsdF9HOfO =JQhh -END PGP SIGNATURE- -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users
Re: [Fink-users] CPU-specific tuning in Fink Source Distributions
On 18 Aug 2010, at 18:42, David R. Morrison wrote: There is one exception to this that I'm aware of: the gmp package does cpu-specific optimization at compile time. But the cpu- specific optimizations tend to be upward-compatible, so if you compile this package on your least-capable machine, it should run on the others. atlas and its deps are another exception, the whole purpose of the pkg being to tune things to the specific CPU at hand. And I've probably noticed a couple of other exceptions, from glancing when they were building _ not sure this was more than mere tuning... Anyway, there should be very few. JF Mertens -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users
Re: [Fink-users] CPU-specific tuning in Fink Source Distributions
Thanks for the replies! Do (Mach-O?) binaries encode which extensions to the x86 set they make use of? Can one expect a warning when loading an incompatible binary, or will it run with an undefined behavior? (I know this is a bit out-of-topic, but I think it might interest other readers of this thread.) Best wishes, Renaud. On 18 Aug 2010, at 18:59, Jean-François Mertens wrote: On 18 Aug 2010, at 18:42, David R. Morrison wrote: There is one exception to this that I'm aware of: the gmp package does cpu-specific optimization at compile time. But the cpu- specific optimizations tend to be upward-compatible, so if you compile this package on your least-capable machine, it should run on the others. atlas and its deps are another exception, the whole purpose of the pkg being to tune things to the specific CPU at hand. And I've probably noticed a couple of other exceptions, from glancing when they were building _ not sure this was more than mere tuning... Anyway, there should be very few. JF Mertens -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users
Re: [Fink-users] CPU-specific tuning in Fink Source Distributions
On Aug 18, 2010, at 11:45 AM, Renaud Detry wrote: Thanks for the replies! Do (Mach-O?) binaries encode which extensions to the x86 set they make use of? Yes Can one expect a warning when loading an incompatible binary, Yes or will it run with an undefined behavior? Here's a thread about that: http://forum.soft32.com/mac/Bad-CPU-type-ftopict47031.html -- Dave -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users
Re: [Fink-users] CPU-specific tuning in Fink Source Distributions
Jean-Fran?ois Mertens j...@core.ucl.ac.be said: On 18 Aug 2010, at 18:42, David R. Morrison wrote: There is one exception to this that I'm aware of: the gmp package does cpu-specific optimization at compile time. But the cpu- specific optimizations tend to be upward-compatible, so if you compile this package on your least-capable machine, it should run on the others. atlas and its deps are another exception, the whole purpose of the pkg being to tune things to the specific CPU at hand. And I've probably noticed a couple of other exceptions, from glancing when they were building _ not sure this was more than mere tuning... Anyway, there should be very few. Debian is trying to hash out how to handle this exact issue in their distro...they apparently had had libatlas-CPU1 libatlas-CPU2 etc and even that didn't give maximal optimization for each system. They're currently toying with either minimal-optimization for everyone, or a stub library that triggers an actual local compile from sources with autodetection of appropriate optimization (or tells users to do so via normal package-manager tools). Feel free to visit Subject: Atlas proposal on debian-devel and debian-science-maintainers (both hosted at lists.alioth.debian.org). dan -- Daniel Macks dma...@netspace.org -- This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev ___ Fink-users mailing list Fink-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-users