[Ffmpeg-devel] [PATCH] MXF demuxer improvements

[Michael Niedermayer]

Hi

On Fri, Jul 28, 2006 at 11:08:55PM -0400, Rich Felker wrote:
> On Fri, Jul 28, 2006 at 07:16:16PM +0200, Michael Niedermayer wrote:
> > Hi
> > 
> > On Fri, Jul 28, 2006 at 12:47:50PM -0400, Rich Felker wrote:
> > > On Thu, Jul 27, 2006 at 10:54:16PM +0200, Baptiste Coudurier wrote:
> > > > Hi
> > > > 
> > > > I reworked MXF demuxer to prepare handling of other operationnal
> > > > patterns and streaming demuxing, with many changes, more accurate
> > > > parsing and simplified codec detection.
> > > > 
> > > > I'm not sure about the DECLARE_ALIGNED with typedef, is it right to do
> > > > that ?
> > > 
> > > A better question would be... why? Surely there will never be SSE2
> > > code processing demuxer fields. If you want to process them
> > > system-word-at-a-time, use an array of long instead of char.
> > 
> > the thing which is being aligned is an array of bytes and under some
> > circumstances 4 of these bytes have to be read as an 32bit int and
> > that requires alignment on some architectures, surely these could
> > be read in a way which doesn need any alignment but the original code
> > needed it ...
> 
> Yes, but what I said was: if it needs to be read as 32bit int, then
> declare it as an array of 32bit ints or as a union type with both byte
> array and 32bit int array...

yes, i agree a union is the best choice if we directly read a 32bit int
from the array, alternatively we could read it bytewise ...

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

In the past you could go to a library and read, borrow or copy any book
Today you'd get arrested for mere telling someone where the library is