[MPlayer-dev-eng] [PATCH] ppc jumbo patch

[Romain Dolbeau]

Alex Beregszaszi wrote:

>>1) configure patch that adds more flags to the compiler
>>    on some powerpc target (putting them in _mcpu pollute
>>    the output). There shouldn't be any apple-specific
>>    flags, so should work in linux/pcc too.
> 
> I will apply the 970 part immediately, but dunno about the other flags
> and
> their performance increase. Have you done any measurements?

No :-) But they're a subset of the "-fast" option, available
only in Apple's version of gcc. I striped the apple-only
option and kept the others. They shouldn't make much of
a difference on a 74xy, but unaligned loops are slow
on the 970 (well for some values of "slow", anyway).

Feel free to drop them ATM, ot the ask for another version
of the patch w/o them (maybe I'll just add a "--enable-fast"
or sometinh to mplayer ? Maybe other arch may benefit of
aggressive compiler options ?)

> Looks nice, also it would be nice to merge our liba52 with the one
> included
> in libavcodec and maybe backporting to the original tree (altough the
> authors aren't response).

Yep, I asked why two copies ?, but got no anwser. Guess
it's mostly lack of time. One would need to back- (or forward-)
port the changes from the mplayer source tree...

>>3) timer-darwin patch (make the cache works in Darwin/MacOS X)
> 
> Dunno, maybe if someone could point me to a link to Apple Developer
> Tools,
> I could test it. I have downloaded an older version, but after
> installing, it still
> doesn't finds the header files!

Huh ? Normally everything works fine upon install. Unless you
somehow have another non-apple compiler ? Until now, all
OSX10.[x, x >=2] box I've tried compiled mplayer without
problem...

That particular patch fix a bug whereas mplayer simply
freeze and nothing happen anymore. For me it's always in
the cache, bud I've heard others complaining of other
freezes in the same timer function...

-- 
Romain Dolbeau