New ATI Driver for Radeon 7500?

[Roland Scheidegger]

Ken Mandelberg wrote:
> Dave Airlie wrote:
>> On 9/13/07, Ken Mandelberg <km at mathcs.emory.edu> wrote:
>>> Dave Airlie wrote:
>>>> On 9/13/07, Ken Mandelberg <km at mathcs.emory.edu> wrote:
>>>>> Does the driver ATI announced do anything for a laptop with
>>>>> 
>>>>> ATI Technologies Inc Radeon Mobility M7 LW [Radeon Mobility 
>>>>> 7500]
>>>> We already have a driver for that card since like it came out..
>>>>  I've no idea why another driver might be useful..
>>>> 
>>>> Dave.
>>> Of course I use the current one.
>>> 
>>> I just thought ATI might know some undocumented tricks to improve
>>>  it.
>>> 
>>> For example the existing driver effectively lowers resolution to 
>>> get around the pink bar problem.
>> Which pink bar problem (I haven't used my M7 in years, but ajax has
>>  one in his laptop..)
>> 
>> Dave.
> 
> The pink bar is explained in
> 
> https://bugs.freedesktop.org/show_bug.cgi?id=1462
> 
> the workaround is to handle 1920x1080 as 960x1080 and scale it a 
> factor of 2 horizontally.
As far as I can tell, there is no possible better workaround for that,
if you use the overlay.
If you'd use textured video, you could do better. There is nothing
stopping someone from implementing that, this is just basic texturing
(you can actually even do planar yuv, using all 3 texturing units, there
is some special bit for that, though I've got no idea if it actually
works, but I guess it should.). You don't need the specs for doing this,
the 3d driver (and even 2d render) code has all information you need.

> Besides that at least on my 2GHZ Pentium 4 Mobility laptop its not 
> fast enough to keep up at HD. An Intel 915 on a 2GHZ Pentium M can.
A 2Ghz P-M is around 1.5 times as fast as a P4-2Ghz for general purpose
code, so no surprise there (but possibly not faster for actual video
decoding using sse, not sure). What would help is if you could use the
idct/mc unit of these chips, but I've got no idea if it could handle
hdtv resolutions. This stuff was designed to handle dvds...

> Again the existing driver works reasonably, I just thought if ATI was
>  releasing a driver they might have some other optimizations and 
> workarounds.
This chip is just too old. The 3d part could need some improvements, but
it's all things you could do without needing AMD's help (e.g. use ttm,
implement 3d textures, implement occlusion queries) - only for cleaning
up some hyperz issues AMD's assistance would probably make things easier.

Roland