[vorbis-dev] Re: Different average bitrates on different machines

[Segher Boessenkool]

Simon Hosie wrote:
> 
> On Tue, Dec 11, 2001 at 10:55:45PM +0100, Segher Boessenkool wrote:
> > Post-increments are evil, and not useful on nearly any processor (nowadays)
> > (the exception is ARM).
> 
> And PPC, and 68k, and AVR, and some of the 680x lot that are still in use,

No, on PPC, only pre-increments have cpu support.

> and ...  I don't think they express much that the compiler couldn't figure
> out on its own, in any case.
> 
> > But they are _so_ very handy for the programmer :-)
> 
> They're perfectly safe in the hands of responsible adults.  But I do
> think they could be better checked by the compiler.

Sure.  But they _are_ weird.

> > The standard _has_ to be vague about a lot of things.  Not everything uses
> > two's complement, for example (although most programs assume all the world
> > is like that, and because of that, are not portable C).
> 
> I'd have to challenge that, too.  Although the spec is vague, it does

The C99 spec says in so-many words that integer representation should be
either ones- or twos-complement.

> assure something really peculiar along the lines of negative values cast
> to unsigned same-data-type are larger than any positive value
> expressible in the signed type.  However it specifies it, it's
> sufficient to satisfy the only assumption I've ever seen made, that
> (unsigned)x < 10 is true when 0 <= x < 10.  I can't think of any other
> assumptions that would be made regularly, except when writing binary
> files, which isn't too common in the greater scheme of things.

Say "bit-fiddling".  We have a lot of bit-fiddling going on, in Vorbis...

Segher

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.