On Friday 25 June 2004 00:12, Nicolas Souchu wrote:
> On Sun, Jun 20, 2004 at 05:52:21PM +0200, Rodolphe Ortalo wrote:
> > It's true that this non-null check does not seem useful. The map is
> > probably non-null at this stage.
> > However, maybe on the very first usage of an accel buffer, before all
> > structures are initialized, a null one can sneak in.
> > It would be worth a try in fact (on Linux).
> Or on FreeBSD once I get it working :)
Sure! Especially as I have a hardware problem on my good old Matrox-related
computer... again! (That's the problem when living in a sunny area...)
BTW, I'm using a brand new mail setup due to this hardware issue. Please
confirm if it works...
> BTW Rodolphe, how works the Matrox DMA? I'm surprised not to find any
> reference to OS-dependent code in the driver. How is served the DMA
Hmmm. IIRC, there is something OS-dependent. (Too bad I can't check the code
right now.) Oh, no, that a "irq claim line" thing, and that's certainly a
resource of the chipset thing. The only thing the driver does is program the
graphics engine in order to generate an interrupt with the "line" it chose.
(I hope this notion of "line" is not Linux dependent.)