1272822 - Consider reducing AMD blacklisting from 8.62.0.0 to 8.56.15.1

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Jeff Muizelaar [:jrmuizel]]

One of the top reasons for not getting WebGL is the AMD blacklists.
Here's a histogram of blacklisted versions from WebGL failures:
            {u'6.1.7600.16385': 3,
             u'8.33.0.0': 1,
             u'8.420.0.0': 1,
             u'8.471.0.0': 1,
             u'8.479.0.0': 1,
             u'8.522.0.0': 1,
             u'8.56.1.15': 3,
             u'8.56.1.16': 5,
             u'8.593.100.0': 5})

I think we should consider reducing the blacklist to 8.56.1.15.

The current choice is AFAIK arbitrary and comes from bug 792190

Comment 1

[Milan Sreckovic [:milan] (needinfo for best results)]

That sounds reasonable.  Anthony, if we were to do this, we'd want to track the overall crash numbers for the drivers that we'd allow in (8.56.1.16 and 8.593.100.0), perhaps some others, and track to see what effect, if any.

Jeff, would you want to start by just enabling WebGL on these drivers, blocking the rest of the features?

Comment 2

[Jeff Muizelaar [:jrmuizel]]

(In reply to Milan Sreckovic [:milan] from comment #1)
> That sounds reasonable.  Anthony, if we were to do this, we'd want to track
> the overall crash numbers for the drivers that we'd allow in (8.56.1.16 and
> 8.593.100.0), perhaps some others, and track to see what effect, if any.
> 
> Jeff, would you want to start by just enabling WebGL on these drivers,
> blocking the rest of the features?

I think I would unblock everything, but I don't feel very strongly about that position.

Comment 3

[u279076]

(In reply to Milan Sreckovic [:milan] from comment #1)
> That sounds reasonable.  Anthony, if we were to do this, we'd want to track
> the overall crash numbers for the drivers that we'd allow in (8.56.1.16 and
> 8.593.100.0), perhaps some others, and track to see what effect, if any.

You want to track overall crash volume or volume of specific crashes?

Comment 4

[Milan Sreckovic [:milan] (needinfo for best results)]

Overall, and what we'd consider "graphics" crashes.

Comment 5

[Jeff Muizelaar [:jrmuizel]]

It's worth noting that 8.56.1.16 is a version that shipped with Windows so if you installed windows and never enabled updates you'd have this driver.

Comment 6

[Jeff Muizelaar [:jrmuizel]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #5)
> It's worth noting that 8.56.1.16 is a version that shipped with Windows so
> if you installed windows and never enabled updates you'd have this driver.

8.56.1.16 is the Win7 SP1 driver
8.56.1.15 is the Win7 RTM driver

Comment 7

[u279076]

(In reply to Milan Sreckovic [:milan] from comment #4)
> Overall, and what we'd consider "graphics" crashes.

I've added prototype that allows graphing any driver version (all vs graphics):
http://ashughes1.github.io/metrics-graphics-gfx/#driver_lookup

Simply enter the driver version and click 'graph'. With this we should be able to graph *any* driver version upon request, not just the ones above. Hopefully this serves what's needed here. 

Let me know.

Comment 8

[Jeff Muizelaar [:jrmuizel]]

A reasonable driver to compare to with is probably 8.631.0.0. This is one of the windows update drivers that we currently allow that came out later that year.

Comment 9

[Jeff Muizelaar [:jrmuizel]]

(In reply to Anthony Hughes (:ashughes) [GFX][QA][Mentor] from comment #7)
> 
> I've added prototype that allows graphing any driver version (all vs
> graphics):
> http://ashughes1.github.io/metrics-graphics-gfx/#driver_lookup

Is it possible to have these graphs normalized by # of users on a particular driver?

Comment 10

[u279076]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #9)
> Is it possible to have these graphs normalized by # of users on a particular driver?

Done.

Comment 11

[Jeff Muizelaar [:jrmuizel]]

Attachment: Block less AMD

The reduces the blocklisting to the version of the AMD driver that shipped with Windows 7 RTM. I've tested out that driver on hardware in Toronto and everything seems to run fine. I ran the WebGL conformance test suite and it only failed a couple of tests. I expect it will fail the same tests with our currently accepted driver but I will double check.

Comment 12

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/5dd1d10d4c99

Comment 13

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/5dd1d10d4c99

Comment 14

[Milan Sreckovic [:milan] (needinfo for best results)]

Release Note Request (optional, but appreciated)
[Why is this notable]:
[Suggested wording]: Windows computers with AMD graphics and driver versions between 8.56.1.15 and 8.62.0.0, previously blocking graphics hardware acceleration, are now allowing it.
[Links (documentation, blog post, etc)]:

Comment 15

[Liz Henry (:lizzard) (relman/hg->git project)]

Milan - Here I think this would be nice to say, but without linking to the bug or some explanation it is hard to express in a short release note.   Maybe some of these perf improvements and other things you are suggesting for relnotes could be explained in a blog post and then we could note and link to it from one general release note.