Minutes of Power in 9.1.0 meeting

Benjamin M. Schwartz bmschwar at fas.harvard.edu
Wed Dec 17 19:06:19 EST 2008


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Chris Ball wrote:
> Hi,
>
>    > I think it would be good to clarify what the 9.1.0 plans are for
>    > #7958 (DCON shows old image with noise artifacts for the duration
>    > of suspend) and #8893 (image jumps vertically for a single frame
>    > during resume).  These two tickets seem to have been confused on
>    > the Improve_battery_life page.
>
> Thanks.  I have a feeling they're both caused by the same bug -- the
> kernel missing DCONLOAD at resume time.

I don't know much about the DCON, but this doesn't make sense to me.  I
thought DCONLOAD was only used at suspend time, not resume time.  Having
DCONLOAD fail or be skipped at suspend explains #7958, but for #8893 the
correct image is shown during suspend, indicating that DCONLOAD succeeded.

For #8893, the artifact is seen at resume time, and meshes perfectly with
an explanation given to me by Ajax: the GPU is not genlocked to the DCON,
so it starts displaying the frame at a totally random point in the refresh
cycle, leading to the top of the image being somewhere in the middle of
the screen.  Ajax suggested that the software workaround would be to give
the GPU a resolution of 1200x100 and then poll for the DCON's
vertical-blanking-interval signal before making the switch. I am not clear
as to whether this was implemented and then reverted, or never implemented.

>  If they aren't, #8893 seems
> much more important than #7958 (which I don't recall seeing in the past
> few months) to me, so we'd start out with #8893 and move on to #7958 if
> it still exists and we get time.  I've updated the wiki ("Requirement 12
> addressed by..") to say so.

OK.  I think your prioritization is reasonable, but #7958 definitely still
exists.  I've seen it recently.  It's just very infrequent, and seemingly
more frequent on some machines than others.  I recall one user who brought
in her G1G1 to our launch party at John Harvard's, showing #7958 on almost
every suspend.

Everybody keep your eyes open for a machine that gets weird noise on the
screen during suspend.

- --Ben
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)

iEYEARECAAYFAklJk/sACgkQUJT6e6HFtqTYSQCfTlztXwiy4YI5cmOAoOgCRnuN
8K8AoKB8bA8itjzP6wp2t2tBgtk2VDx5
=KLyj
-----END PGP SIGNATURE-----



More information about the Devel mailing list