[FFmpeg-devel] MacOS X Compile error with swscale
Mon Mar 31 03:46:16 CEST 2008
On Mon, Mar 31, 2008 at 02:31:47AM +0100, M?ns Rullg?rd wrote:
> Diego Biurrun <diego at biurrun.de> writes:
> > On Sat, Mar 29, 2008 at 05:48:05PM +0100, Michael Niedermayer wrote:
> >> On Sat, Mar 29, 2008 at 05:25:05PM +0100, Diego Biurrun wrote:
> >> > It is trying to use MMX2 code, not MMX code. You could add
> >> > --disable-mmx2 to the command line to make compilation go through
> >> > or try the patch I created to fix the situation.
> >> MMX2 implicates MMX.
> >> Having the defines for MMX2 set but not the ones for MMX is a bug in
> >> configure IMO.
> > That one is fixed easily.
> But is it really a bug?
It didnt compile IIRC so id say there is a bug, somewhere at least.
> Sure, any CPU with MMX2 also supports MMX,
> but I can imagine someone wishing to disable only MMX while keeping
> MMX2 code. Never underestimate the obscurity of people's needs.
Yes, but the code is not designed for that. And simply setting MMX2 but
not MMX #defines is a recipe for disaster. Either people will change
configure (which gets us back to the start) or they will start adding
hacks to make the code compile somehow.
Doing the later would hide the diff by spreading it out over time.
I much rather would like to see what actually needs changing to make
this work so we can judge if its worth for that obscure case.
Also iam not even sure what it means to enable MMX2 but not MMX?
There is no pure MMX2 code.
Should that enable all MMX2+MMX code which doesnt have a MMX counterpart
or disable all functions with the string MMX in them but not MMX2?
Later of course will also contain normal MMX code ...
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Asymptotically faster algorithms should always be preferred if you have
asymptotical amounts of data
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: Digital signature
More information about the ffmpeg-devel