[FFmpeg-devel] [PATCH] Make HAVE_FAST_UNALIGNED allow unaligned memory accesses

[Måns Rullgård]

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

> On Tue, Jul 15, 2008 at 08:23:12PM +0100, M?ns Rullg?rd wrote:
>> > diff --git a/libavutil/intreadwrite.h b/libavutil/intreadwrite.h
>> > index cbf5163..12dec54 100644
>> > --- a/libavutil/intreadwrite.h
>> > +++ b/libavutil/intreadwrite.h
>> > @@ -23,6 +23,8 @@
>> >  #include "config.h"
>> >  #include "bswap.h"
>> >
>> > +#if !defined(HAVE_FAST_UNALIGNED)
>> > +
>> >  #ifdef __GNUC__
>> >
>> >  struct unaligned_64 { uint64_t l; } __attribute__((packed));
>> > @@ -37,7 +39,19 @@ struct unaligned_16 { uint16_t l; } __attribute__((packed));
>> >  #define AV_WN32(a, b) (((struct unaligned_32 *) (a))->l) = (b)
>> >  #define AV_WN64(a, b) (((struct unaligned_64 *) (a))->l) = (b)
>> >
>> > -#else /* __GNUC__ */
>> > +#elif defined(__DECC)
>> > +
>> > +#define AV_RN16(a) (*((const __unaligned uint16_t*)(a)))
>> 
>> Yes, these ifdefs should be replaced with checks for the required
>> compiler features in configure.  If we agree to this change, I'll
>> add such checks, and update the patch.
>
> can the current __GNUC__ be checked reliably? That
> __attribute__((packed)) acts as "unaligned marker" for structs with only
> one element is something that on first sight I would consider a compiler
> bug, in which case it would be really hard to check for...

GCC has always behaved like this.  However, the documentation for the
'packed' attribute makes no mention of this.

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