Removing XAA render accel in radeon
alexdeucher at gmail.com
Wed Jan 6 15:51:59 PST 2010
On Wed, Jan 6, 2010 at 6:15 PM, Richard Wilbur <richard.wilbur at gmail.com> wrote:
> On Mon, 2009-12-14 at 16:22 -0500, Alex Deucher wrote:
>> On Mon, Dec 14, 2009 at 4:17 PM, Roland Scheidegger
>> <sroland at tungstengraphics.com> wrote:
>> > On 14.12.2009 20:33, Alex Deucher wrote:
>> >> On Mon, Dec 14, 2009 at 2:21 PM, Roland Scheidegger
>> >> <sroland at tungstengraphics.com> wrote:
>> >>> On 14.12.2009 15:22, Alex Deucher wrote:
>> >>>> XAA render support seems to be broken (fdo bug 22055, possibly
>> >>>> others). It never really did much to begin with, and it only
>> >>>> supported r1xx and r2xx. Any objections to me dumping XAA render
>> >>>> accel in the radeon driver?
>> >>> Hmm, personally I was never able to get good performance out of EXA with
>> >>> a 32MB r100 card (read: completely unusable). Maybe that has changed
>> >>> with more recent xorg server versions, and I guess it should have
>> >>> changed with KMS (I certainly used to have problems even with a 64MB
>> >>> card without KMS, if the resolution was more than XGA), but unless
>> >>> you're willing to ditch non-KMS too that might be a reason to keep it
>> >>> for now, though I guess that would mean fixing...
>> >> I think the XAA render problem is due to bitrot in the XAA code in the
>> >> xserver as the XAA render code in the driver hasn't been touched in
>> >> years and only recently have we started getting reports of it being
>> >> broken. Does the XAA render code actually make much of a difference
>> >> performance-wise? I never really noticed much difference with it.
>> >> It's especially useless since offscreen pixmap support for XAA broke a
>> >> year or two ago. We could just default render accel to false with
>> >> XAA, but leave the code there, but I don't see anyone fixing XAA, so
>> >> there's not much point in keeping it.
>> > Oh atually I misread that I thought you were going to nuke xaa
>> > completely from the driver. I guess it's ok to to remove xaa render
>> > code. Is radeon actually the only driver left with xaa render accel?
>> Could be. I'm not sure what other drivers ever supported it.
> 1. What kind of bitrot?
No one has touched XAA in years other than to disable offscreen
pixmaps because it broke at some point. Things slowly break as
changes are made to the xserver.
> 2. Are the XAA render code bugs tagged similarly?
What are you asking? I suppose we could disable XAA render support in
the server like we did for XAA offscreen pixmap support. However, I
don't think any other drivers implemented XAA render support.
> 3. What kind of functionality did XAA render code provide?
Basic acceleration for the render protocol, but mostly useless since
offscreen memory in XAA is broken at the moment, and XAA offscreen
pixmaps had to be the same depth as the scan-out buffer.
> 4. What level of effort do you think it would require to repair?
No idea. It's been ages since I looked at XAA.
> PGP public key at
> xorg-driver-ati mailing list
> xorg-driver-ati at lists.x.org
More information about the xorg-driver-ati