Increased latency from removal of GEM_THROTTLE in 2D driver

Increased latency from removal of GEM_THROTTLE in 2D driver

[Peter Clifton]

I've been playing with my app again, and whilst it seems a tiny bit
faster in benchmarks, since the 2D driver change:

commit 8ff37667bf864b771d16a58fc5041cb48408b6a8
Author: Eric Anholt <eric@anholt.net>
Date:   Tue Nov 2 10:36:03 2010 -0700

    Remove the intermittent GEM_THROTTLE call.


I've noticed the latency / interactivity is WAY down, and I've also had
intermittent GPU hangs whilst benchmarking which I'd never seen before. 

It might be unrelated (as I'm running a back-ported drm-intel-next, and
just pulled it up to date), but for a quick check, I just reverted to a
previous 2D driver just prior to the above commit, and interactivity has
improved a lot (I don't have a number for it, just from testing).

Presumably this says something bad about my app, or its benchmarking
code. Any clues as to what that is?

Without the commit, I can use other X11 apps whilst my PCB program is
benchmarking. With it, the X11 server graphics almost completely freeze
up.

-- 
Peter Clifton

Electrical Engineering Division,
Engineering Department,
University of Cambridge,
9, JJ Thomson Avenue,
Cambridge
CB3 0FA

Tel: +44 (0)7729 980173 - (No signal in the lab!)
Tel: +44 (0)1223 748328 - (Shared lab phone, ask for me)

Re: Increased latency from removal of GEM_THROTTLE in 2D driver

[Chris Wilson]

On Sat, 06 Nov 2010 01:44:59 +0000, Peter Clifton <pcjc2@cam.ac.uk> wrote:
> Presumably this says something bad about my app, or its benchmarking
> code. Any clues as to what that is?
> 
> Without the commit, I can use other X11 apps whilst my PCB program is
> benchmarking. With it, the X11 server graphics almost completely freeze
> up.

Do you get a similar effect with the THROTTLE in the ddx and a recent
kernel (2.6.37-rc1 or drm-intel-next)? Those kernels should only penalize
the caller of THROTTLE and so X should not end up blocking your app.

I can feel a new tracepoint coming: throttle-begin, throttle-end.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

Re: Increased latency from removal of GEM_THROTTLE in 2D driver

[Eric Anholt]

[-- Attachment #1.1: Type: text/plain, Size: 1272 bytes --]

On Sat, 06 Nov 2010 01:44:59 +0000, Peter Clifton <pcjc2@cam.ac.uk> wrote:
> I've been playing with my app again, and whilst it seems a tiny bit
> faster in benchmarks, since the 2D driver change:
> 
> commit 8ff37667bf864b771d16a58fc5041cb48408b6a8
> Author: Eric Anholt <eric@anholt.net>
> Date:   Tue Nov 2 10:36:03 2010 -0700
> 
>     Remove the intermittent GEM_THROTTLE call.
> 
> 
> I've noticed the latency / interactivity is WAY down, and I've also had
> intermittent GPU hangs whilst benchmarking which I'd never seen before. 
> 
> It might be unrelated (as I'm running a back-ported drm-intel-next, and
> just pulled it up to date), but for a quick check, I just reverted to a
> previous 2D driver just prior to the above commit, and interactivity has
> improved a lot (I don't have a number for it, just from testing).
> 
> Presumably this says something bad about my app, or its benchmarking
> code. Any clues as to what that is?
> 
> Without the commit, I can use other X11 apps whilst my PCB program is
> benchmarking. With it, the X11 server graphics almost completely freeze
> up.

I'm seeing this too, with vblank_mode=0 and various apps.  Something
appears to be broken with the swapbuffers waits in the GL driver,
perhaps.

[-- Attachment #1.2: Type: application/pgp-signature, Size: 197 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

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