[Mplayer-users] What to do before version 1.0
Felix von Leitner
leitner at convergence.de
Wed Jun 13 20:07:53 CEST 2001
Thus spake Alex Kanavin (ak at cave.hop.stu.neva.ru):
> > Why?
> Because it's universal and robust.
It is neither universal nor robust.
I have seen more projects not compile because of autoconf bogosity than
because of the actual project source code.
> > The configuration now works fine (as opposed to the typical GNU autoconf
> > monster-bloat script that by the way redefines the meaning of "slow
> > and broken software").
> Slow, yes, but only the first time, then it uses cached test results.
The common case is: "wget, tar xzf, configure, make, make install, rm -rf".
Optimize for the common case, which is the users, not the developers.
> > Your patch will be greatly appreciated.
> > I am curious how you will manage to do this without performance penalty.
> Does using function pointers that were set up at start up result in a
> noticeable peroformance decrease?
Yep. Up to ten times the cycle count on some architectures, plus
prefetching problems, plus cache pollution.
Mplayer-users mailing list
Mplayer-users at lists.sourceforge.net
More information about the MPlayer-users