Re: [Xpert] Re: [rtl] Re: Virge DX and accel. introduced latency?(was Re: [Xpert]XFree86in realtime applications)

[John Regehr]

On Sun, 26 Nov 2000 [EMAIL PROTECTED] wrote:

> On Sun, Nov 26, 2000 at 01:10:32PM -0500, Kevin Brosius wrote:
> > The idea is to simulate 3.3.6 behavior by forcing the accelerator to go
> > idle at the end of every accelerated graphics primitive.  4.0.x is able
> > to handle the accel. setup more efficiently, by offering 3 functions
> > used for a primitive.
> 
> It would perhaps be better to simply read the fifo first before issuing
> any output and force delays only in those cases.

Right - this is the only behavior that makes sense if one wants good X
performance and also good real-time response.

There are several things that the real-time Linux community can do:

1) Come up with patched drivers that always check the FIFO status
   before writing to it (and that don't fiddle with interrupts
   if that's also a problem).

2) Make driver developers aware of these patches, and why they
   matter.

3) Put a "black list" of known-bad driver/hardware combinations into a
   FAQ, and leave it there until the developers include the patches or
   something similar.

John Regehr



-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
---
For more information on Real-Time Linux see:
http://www.rtlinux.org/rtlinux/