Keith Packard wrote:

> Around 8 o'clock on Aug 21, Peter Kaczowka wrote:
>
> > If MMX is already being used in Render or has already tried, I'm sorry to
> > imply that no one has thought of this before.
>
> It has been tried with no effect on performance.  As I said, current
> processors have no optimizations for AGP/PCI read cycles when running in
> write-combining mode, each one is a completely separate synchronous
> transaction.
>

I guess that's true for reads, but my experience is that quadword writes are
faster to some PCI devices.  That is, a blt from memory to the screen is
faster if done with MMX.  But that was as/of 2-3 years ago, e.g. on 400 MHz
Pentium II machines with PCI bus and our company's own cards e.g. the
DOME imaging Md5/PCI card; other cards might handle burst under different
conditions.  I did not see any improvement writing to other AGP-based cards;
DOME doesn't make AGP cards.

In any case it wouldn't be a big speedup and there's no substitute for real
hardware support.  I would be interested in hearing which cards have what kind
of hardware Render support if people feel like discussing it.  It doesn't
relate to what I do now but in a former life I was a 2D graphics type.

Keith, if you recall we were both on the XIE working group (me from
HP/Apollo).  That was a while ago, and after that I never thought I would be
using X as part of my job since Windows machines were kicking Unix workstation
butt.   Who would have thought that X would live on as XFree86?  I guess Keith
among others did; congratulations on keeping X alive.  It's still a far better
architecture than Windows and GDI.  Sorry to change the subject; I'm not
trying to start a separate thread here - but thanks, Keith.

Peter Kaczowka

>
> There is a small speedup available on 16-bit frame buffers by reading
> 32-bits at a time, but that will become less important as more people
> migrate to 24-bit displays.
>
> Some people are successfully implementing significant Render acceleration
> with hardware; that's clearly the right direction to go.
>
> Keith Packard        XFree86 Core Team        HP Cambridge Research Lab

_______________________________________________
Render mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/render

Reply via email to