[Mplayer-users] What to do before version 1.0
ak at cave.hop.stu.neva.ru
Sat Jun 16 20:49:42 CEST 2001
On Sat, 16 Jun 2001, Arpi wrote:
> double work - nothing sense.
> and if you've read my benchmark about i586 vs i686 thing here, then
> you know: switching some mmx/sse/3dnow optimization at runtime is not all!
> compilation of the C code (-march,-mcpu) also matters A LOT!
> and you can't do it with runtime pointers.
I know. But... users want binaries. :) And the above steps would greatly
reduce the number of binaries you'll have to make: they would only differ
in -march -mcpu compilation options. If some user, after downloading a
binary, discovers that the performance is not enough, he can always get
the source (rpm), turn on compile-time optimization (#ifdef instead of
function pointers) and get a custom binary for himself.
By the way, do you want to see mplayer in distributions? If you do, you'll
have to provide a way to make distributable binaries. Then it will be up
to the distributors, if they include i386.rpm for everyone, or separate
package for every processor, or something in between.
Mplayer-users mailing list
Mplayer-users at lists.sourceforge.net
More information about the MPlayer-users