On Tue, 2002-07-09 at 01:39, Mike Mestnik wrote: 
> --- Michel Dänzer <[EMAIL PROTECTED]> wrote:
> > On Mon, 2002-07-08 at 20:17, Tim Smith wrote: 
> > > On Monday 08 Jul 2002 12:49 am, Michel Dänzer scribed numinously:"
> > > 
> > > > The scratch register values need to be read with DRM_READ32(), which
> > > > accounts both for endianness and memory barriers. So it would be
> > > >
> > > >                 u32 done_age = DRM_READ32(&dev_priv->scratch[1]);
> > > 
> > > That's good to know; I'll file that a little closer to my forebrain. I'd 
> > > noticed the macros before but not taken enough notice. I thought the card 
> > > took care of that when it wrote the value back (I believe it can) but maybe 
> > > not.
> > 
> > It can, but that would mean extra code to set the control registers
> > according to endianness and wouldn't really buy us anything as reading
> > little endian data is free with a decent big endian CPU and the memory
> > barriers would still have to be dealt with.
> > 
> That surprises me, WHY would you go throught all the trouble

Is it really that much trouble?

> of making avalibul a control register that changes byte ordering?

In order to keep Apple as a customer? ;)

> Is the GPU big endian, and thay are just letting you turn off there
> endianness schem?  I don't know mutch about modern technology but I
> allways thought that in the olden days endianness was an optimisation,
> like the way singed intergers are handeled.  Could tweaking this have
> any effect, even /w little endian CPUs or do any thing?

The Radeon can swap the bytes of pretty much anything in all possible
ways. I assume this is primarily intended to make things easier for a
big endian host. We're only using few of these features (on big endian
machines, none otherwise of course) for various reasons, e.g. because
PCI is defined to be little endian, for the sake of code consistency,
because it simply isn't worth the effort, ...


-- 
Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member   /  CS student, Free Software enthusiast


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Stuff, things, and much much more.
http://thinkgeek.com/sf
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel

Reply via email to