[Bug 47007] HDMI monitor polling causing 100ms rendering stalls

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Thu Mar 15 02:50:51 PDT 2012


https://bugs.freedesktop.org/show_bug.cgi?id=47007

Tvrtko Ursulin <tvrtko.ursulin at onelan.co.uk> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #58360|0                           |1
        is obsolete|                            |

--- Comment #16 from Tvrtko Ursulin <tvrtko.ursulin at onelan.co.uk> 2012-03-15 02:50:51 PDT ---
Created attachment 58481
  --> https://bugs.freedesktop.org/attachment.cgi?id=58481
Use shared HPD for VGA on R600+

I've come up with a new approach which seems to work well on hardware I have
here. However it makes one assumption which I cannot be certain is true and
therefore needs AMD opinion and review. Assumption is that if DDC lines are
shared between the VGA and DVI connectors, that also means HPD line is shared
(it is on our hardware).

Under that assumption I have reworked connector adding a bit so shared DDC is
not a boolean any more but a pointer to the shared connector instance. That
allows me to enable HPD for the VGA connector which makes the poll worker not
run every ten seconds (fixing a major source of pain for us). 

It also enables me to avoid fetching the VGA EDID if the shared connector is
connected - rationale being that only one of the pair can be connected at a
time. This logic is dependant on R600+ ASICs where we can rely on HPD lines
being present.

Furthermore, I've left in the bit which trusts HPD sense on R600+ ASICs,
meaning as long as HPD sense remained unchanged, connector detect functions
will bail out early.

Last two are only considered if detect is not triggered from userspace.
Otherwise full probe will be done as before.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.


More information about the dri-devel mailing list