[Intel-gfx] [PATCH] Add a flag to signal possible origin of the gpu hang.

[Eugeni Dodonov]

On Mon, Oct 17, 2011 at 15:10, Ben Widawsky <ben at bwidawsk.net> wrote:

> On Mon, Oct 17, 2011 at 02:25:01PM -0200, Eugeni Dodonov wrote:
> > This adds a hint on where the call trace originated from. This is more of
> > a guess, but still, it seems to be quite accurate on the traces I tested.
> >
> > So it appears in the output like:
> >
> > ...
> > 0x06c470c8:      0x780e0002: 3DSTATE_CC_STATE_POINTERS
> > 0x06c470cc:      0x00007f81:    blend change 1
> > 0x06c470d0:      0x00007f01:    depth stencil change 1
> > 0x06c470d4:      0x00007f41:    cc change 1
> > 0x06c470d8:      0x78021302: 3DSTATE_SAMPLER_STATE_POINTERS: VS mod 1, GS
> > mod 1, PS mod 1
> > 0x06c470dc:      0x00000000:    VS sampler state
> > 0x06c470e0:      0x00000000:    GS sampler state
> > 0x06c470e4:      0x00007e00:    WM sampler state
> > 0x06c470e8:      0x78150003: Possible origin: mesa
> > 0x06c470e8:      0x78150003: 3DSTATE_CONSTANT_VS_STATE
> > 0x06c470ec:      0x00000000:    dword 1
> > 0x06c470f0:      0x00000000:    dword 2
> > 0x06c470f4:      0x00000000:    dword 3
> > 0x06c470f8:      0x00000000:    dword 4
> > ...
> >
> > Signed-off-by: Eugeni Dodonov <eugeni.dodonov at intel.com>
> > ---
>
> I have patches which I never posted which do something similar, but
> perhaps more useful. It gets a PID and a timestamp for each batch.
> Unfortunately those require kernel changes. If that sounds useful to
> you, I will clean them up and put them on the list.
>

Yep, it looks interesting for me. I also have a work in progress which grabs
the perf results and combines them with timestamps, but it is still far from
finished. I could try to adapt it to your patches if you think it would be
interesting.

-- 
Eugeni Dodonov
<http://eugeni.dodonov.net/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20111017/a8dc3f22/attachment.html>