[FFmpeg-devel] r9017 breaks WMA decoding on Intel Macs
Wed May 30 21:34:04 CEST 2007
On Wed, May 30, 2007 at 07:58:34PM +0100, Patrice Bensoussan wrote:
> On 30 May 2007, at 19:03, Trent Piepho wrote:
> > On Wed, 30 May 2007, Michael Niedermayer wrote:
> >>>> compiles with gcc 2.95. So there isn't any doubt in the patch
> >>>> itself.
> >>> Ok, fine with me. Michael, do you think that the patch I posted
> >>> earlier (100% based on Trent's, only fixing minor issues) should be
> >>> applied?
> >> well these arent the only occurances of this syntax in ffmpeg
> >> also i would like to see benchmarks, gcc has the tendency to do
> >> stupid
> >> things if it can and here it can ... (=more freedom with gcc
> >> generally means
> >> worse code, thats just my experience with gcc, its not always
> >> true, also
> >> gcc should be getting better as the version numbers increase ...)
> >> also i dont see how this additional freedom could lead to better code
> >> here, it can just lead to worse code if gcc doesnt realize that
> >> things
> >> can be addressed via the same register
> > When I checked the asm output, it generates the exact same
> > instructions,
> > just without the semi-incorrect asm syntax.
> Hmmm... Michael, the code currently committed is broken for Intel
> based Macs... so shouldn't we apply this patch anyway to actually
> have a correct code, and then do some benchmarks later? Or simply
> revert the original patch until we can decide what is the best thing
> to do?
wait a moment, what is broken?
apple ships an ancient broken assembler which silently missassembles
if apple would silently misscompile i++ we wouldnt start replacing that
by i=i+1 either ...
replacing all the existing code to workaround that bug (which apple
will likely fix eventually) shouldnt be done carelessly, its not that
hard to put a START/STOP_TIMER aoround the asms which are affected and
and as trent already said that the generated code is identical we shouldnt
see any speed change
(note i would be even happier if the benchmarks would also be done with
gcc 2.95 ...)
if we do care about ancient assemblers we should not forget the less ancient
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
In a rich man's house there is no place to spit but his face.
-- Diogenes of Sinope
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: not available
More information about the ffmpeg-devel