[regression] drm/i915/execbuffer: Reorder binding of objects to favour restrictions

[regression] drm/i915/execbuffer: Reorder binding of objects to favour restrictions

[Alexey Fisher]

[-- Attachment #1: Type: text/plain, Size: 1152 bytes --]

Hi Chris,

i have new regression after this patch. Like usual, artifakts,
disappearing icons, with only one difference gl seems to work ok.

see screencast in attachment.

My HW:
board Intel DG45ID, graphic:
Vendor: 0x8086, Device: 0x2e22, Revision: 0x03 (A3)

xserver-xorg-video-intel
2:2.14.0+git20110110.fd9235eb-0ubuntu0sarvatt~maverick

xserver-xorg
1:7.5+6+xserver1.9

commit 6fe4f14044f181e146cdc15485428f95fa541ce8
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Mon Jan 10 17:35:37 2011 +0000

drm/i915/execbuffer: Reorder binding of objects to favour restrictions

As the mappable portion of the aperture is always a small subset at the
start of the GTT, it is allocated preferentially by drm_mm. This is
useful in case we ever need to map an object later. However, if you have
a large object that can consume the entire mappable region of the
GTT this prevents the batchbuffer from fitting and so causing an error.
Instead allocate all those that require a mapping up front in order to
improve the likelihood of finding sufficient space to bind them.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
-- 
Regards,
        Alexey

[-- Attachment #2: bild-1.ogv --]
[-- Type: video/ogg, Size: 1705930 bytes --]

[-- Attachment #3: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [regression] drm/i915/execbuffer: Reorder binding of objects to favour restrictions

[Alexey Fisher]

Hi Chris,
did you got my regression report?

Am Freitag, den 14.01.2011, 16:11 +0100 schrieb Alexey Fisher:
> Hi Chris,
> 
> i have new regression after this patch. Like usual, artifakts,
> disappearing icons, with only one difference gl seems to work ok.
> 
> see screencast in attachment.
> 
> My HW:
> board Intel DG45ID, graphic:
> Vendor: 0x8086, Device: 0x2e22, Revision: 0x03 (A3)
> 
> xserver-xorg-video-intel
> 2:2.14.0+git20110110.fd9235eb-0ubuntu0sarvatt~maverick
> 
> xserver-xorg
> 1:7.5+6+xserver1.9
> 
> commit 6fe4f14044f181e146cdc15485428f95fa541ce8
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date: Mon Jan 10 17:35:37 2011 +0000
> 
> drm/i915/execbuffer: Reorder binding of objects to favour restrictions
> 
> As the mappable portion of the aperture is always a small subset at the
> start of the GTT, it is allocated preferentially by drm_mm. This is
> useful in case we ever need to map an object later. However, if you have
> a large object that can consume the entire mappable region of the
> GTT this prevents the batchbuffer from fitting and so causing an error.
> Instead allocate all those that require a mapping up front in order to
> improve the likelihood of finding sufficient space to bind them.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Regards,
        Alexey

Re: [regression] drm/i915/execbuffer: Reorder binding of objects to favour restrictions

[Chris Wilson]

On Sun, 16 Jan 2011 15:38:34 +0100, Alexey Fisher <bug-track@fisher-privat.net> wrote:
> Hi Chris,
> did you got my regression report?

It appears you didn't get my response saying Linus had already pulled the
fix.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

Re: [regression] drm/i915/execbuffer: Reorder binding of objects to favour restrictions

[Alexey Fisher]

Am Sonntag, den 16.01.2011, 14:53 +0000 schrieb Chris Wilson:
> On Sun, 16 Jan 2011 15:38:34 +0100, Alexey Fisher <bug-track@fisher-privat.net> wrote:
> > Hi Chris,
> > did you got my regression report?
> 
> It appears you didn't get my response saying Linus had already pulled the
> fix.
> -Chris
> 

Ah, ok.

Then the fix is not in drm-intel-next

-- 
Regards,
        Alexey

Re: [regression] drm/i915/execbuffer: Reorder binding of objects to favour restrictions

[Chris Wilson]

On Sun, 16 Jan 2011 15:55:57 +0100, Alexey Fisher <bug-track@fisher-privat.net> wrote:
> Am Sonntag, den 16.01.2011, 14:53 +0000 schrieb Chris Wilson:
> > On Sun, 16 Jan 2011 15:38:34 +0100, Alexey Fisher <bug-track@fisher-privat.net> wrote:
> > > Hi Chris,
> > > did you got my regression report?
> > 
> > It appears you didn't get my response saying Linus had already pulled the
> > fix.
> > -Chris
> > 
> 
> Ah, ok.
> 
> Then the fix is not in drm-intel-next

No, I haven't started pulling in stuff for -next just yet. I've a few
tagged but my focus is currently on making sure -fixes/rc1 has no further
regressions. Once -rc1 is released, I'll sync up.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre