Re: [Dri-devel] issue with texmem and no irqs

[Ian R. <id...@us...>]

Ville Syrj=E4l=E4 wrote:
> On Mon, Apr 21, 2003 at 08:25:08AM -0700, Ian Romanick wrote:
>=20
>>>I don't think you can get away without changing the DDX. Maybe you cou=
ld
>>>check if linecomp has a reasonable value before the driver would claim=
 to
>>>have vsync support.
>>
>>Is there any reason the kernel module couldn't just program in a=20
>>reasonable value?  That is, other than than the difficulty of=20
>>determining what a reasonable value would be. :)
>=20
>=20
> linecomp too lives in the CRTC registers so it faces the same problem
> with mode switches as the vsync interrupt.
>=20
> I would call a reasonable value to be something between vdispend and
> vtotal.

I'm starting to *really* hate the G400.  We can't check the value of=20
line comp. in the 3D driver because we don't have permission to write to=20
MGAREG_CRTC_INDEX. :(  Unless I'm mistaken, the only other option is to=20
add code to the DRM to do this checking.  Perhaps an additional=20
GETPARAM?  Something like MGA_PARAM_LINECOMP_SANE?

I think this is the only issue holding up the merge (texmem-0-0-1 to=20
trunk).  Does anyone know of any other issues?  I plan to leave the=20
vsync behavior of the trunk as it is when I merge.