[FFmpeg-cvslog] r11889 - trunk/libavutil/mem.h

[Måns Rullgård]

Reimar D?ffinger <Reimar.Doeffinger at stud.uni-karlsruhe.de> writes:

> Hello,
> On Sun, Feb 10, 2008 at 10:32:37PM +0000, M?ns Rullg?rd wrote:
>> cehoyos <subversion at mplayerhq.hu> writes:
>> > Modified: trunk/libavutil/mem.h
>> > ==============================================================================
>> > --- trunk/libavutil/mem.h	(original)
>> > +++ trunk/libavutil/mem.h	Sat Feb  9 21:47:11 2008
>> > @@ -28,7 +28,11 @@
>> >
>> >  #ifdef __GNUC__
>> >    #define DECLARE_ALIGNED(n,t,v)       t v __attribute__ ((aligned (n)))
>> > +#ifdef __ICC
>> > +  #define DECLARE_ASM_CONST(n,t,v)     const t __attribute__ ((aligned (n))) v
>> > +#else
>> >    #define DECLARE_ASM_CONST(n,t,v)     static const t v attribute_used __attribute__ ((aligned (n)))
>> > +#endif
>> 
>> Why does one of those cases use 'static' and not the other?
>
> Hmm... Have you been sleeping or hiding from the flames under a rock?
> ;-)
> Basically the symbols ICC generates for static variables do not work
> with our asm-MANGLE (at least I think it does, in the cases I really
> looked at it just removed them completely because they seemed unused
> to it for all I can tell).  Making the symbols global forces ICC to
> keep the variables and use the standard symbol naming.

Yes, now that you mention it, I do recall something like that being
discussed.  Would it be worthwhile pestering Intel for an
attribute(used) like gcc has?

-- 
M?ns Rullg?rd
mans at mansr.com