[directfb-dev] Re: Where is FBIO_WAITFORVSYNC supposed to getdefined for matrox cards?

["Brian J. Murrell" <brian@interlinx.bc.ca>]

On Sun, 2004-01-11 at 08:14, Brian J. Murrell wrote:
> Now given that I am looking at this "crtc2_wait_vsync"
...
>  I see the basic logic is to call the
> FBIO_WAITFORVSYNC, which makes the kernel put the process to sleep until
> the video card reports an irq, which is done when it has finished
> "drawing" a frame (or does the irq come after drawing fields?) to the
> television set and is on it's way to drawing the next frame.

> Now, after being woken up from waiting for the irq, the process then
> sits in a spin loop until the video card is drawing line number 242
> (NTSC here), which should be how many iterations (re: my "BTW" question
> above)?

A closer look at the ioctl code in the kernel and I see that the spin
loop never happens if the ioctl succeeds.  Indeed, the loop is a
fallback in case the ioctl fails.

> Can I safely remove this spin loop if I *know* that my directfb and
> kernel are properly processing interrupts from the G400 card?

With my new understanding, this is of course moot.

> Indeed, in my local directfb installation, I am perfectly happy for an
> application to _break_ if the kernel is not built to handle interrupts
> and will change my source locally to do so.

I think I might still do this however.

> I _do_not_ want to sit in
> spin loops to flip frames and would rather things broke than do that. 
> In this situation the kernel needs fixing, not directfb, IMHO.  :-)

Rather than a spin loop, I think I will error out.

b.

-- 
My other computer is your Microsoft Windows server.

Brian J. Murrell

Attachment: signature.asc
Description: This is a digitally signed message part