On Tue, Jan 23, 2001 at 08:58:35PM -0800, Andrew Lenharth wrote: > On Tue, 23 Jan 2001, Adrian Bunk wrote: > > Even if there isn't a generalspeedup, there's software that can be > > compiled with support for 3Dnow! or MMX and there already packages like > > libc6-i586 that are compiled optimized. I don't know if it's worth > > compiling the whole distribution optimized, but at least for some packages > > it is worth it. > > It seems that package pools could solve this. A pool could exist for > different sub-archs containing optimized binaries. As long as entry into > the pool was limited/strictly controlled, the pool could remain fairly > small. There would have to be some constraint, such as "a demonstraited > speed-up of at least 15% in processing data or 15% decrease in cpu > usage on the same load from the non-optimized version". (15% completely > arbitrary). > > This makes it fairly easy for the end-user to use optimized binaries, just > have the optimized line first in source.list. It also let mirrors using > pool-aware mirroring scripts not mirror the optimized binaries if they > don't want to. And since entrance into the pool is very restricted, the > disk space requirement is not huge (hands waving wildly :). > > Andrew Lenharth You dont really need it though. I doubt adding in the MMX/3dnow/etc. stuff will add much extra to a package, and VERY simple parsing of /proc/cpuinfo can tell weather(sp) you should enable it or not. This has been done in things like the utah-glx program. It would probably require some mod's from the authors(or packagers) to make it do that instead of the #ifdef MMX type stuff, but i dont think it would be that difficult. > > > -- > To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org > with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org > > -- Erik Bernhardson journey@jps.net -- It is better to remain silent and be considered a fool, than to speak and remove all doubt. -- Abraham Lincoln
Attachment:
pgpdIff56xNLE.pgp
Description: PGP signature