1092260 - Some AMD drivers seem to have an extra frame of delay

Status: RESOLVED FIXED

(Core :: Graphics: Layers, defect)

Description

[Jeff Muizelaar [:jrmuizel]]

In the reproduction we've seen of bug 1089183 we've discovered that the AMD driver doesn't seem to present the current frame until you call Present() for the next frame.

This is easily reproducible by typing into the input box at the following URL:
data:text/html,<textarea>

Comment 1

[Jeff Muizelaar [:jrmuizel]]

I was able to reproduce this with a simple D3D11 test case. I was able to reproduce the same thing with a port of the D3D11 test case to D3D10.

Comment 2

[Jeff Muizelaar [:jrmuizel]]

Turns out this was a wild goose chase. The same problem reproduces in IE11 as well as Firefox 25.

Comment 3

[Jeff Muizelaar [:jrmuizel]]

This looks it might be the same as https://code.google.com/p/chromium/issues/detail?id=402134

Comment 4

[Jeff Muizelaar [:jrmuizel]]

Attachment: Block D3D11 layers on AMD hardware that delays by a frame

Approval Request Comment
[Feature/regressing bug #]: OMTC
[User impact if declined]: Delayed by frame
[Risks and why]: Low. just modifying a block list

Comment 5

[Jeff Muizelaar [:jrmuizel]]

Attachment: amd

Approval Request Comment
[Feature/regressing bug #]: OMTC
[User impact if declined]: Delayed by frame
[Risks and why]: Low. just modifying a block list

Comment 6

[Jeff Muizelaar [:jrmuizel]]

Attachment: Block D3D11 layers on AMD hardware that delays by a frame

Approval Request Comment
[Feature/regressing bug #]:
[User impact if declined]:
[Describe test coverage new/current, TBPL]:
[Risks and why]: 
[String/UUID change made/needed]:

Approval Request Comment
[Feature/regressing bug #]:
[User impact if declined]:
[Describe test coverage new/current, TBPL]:
[Risks and why]: 
[String/UUID change made/needed]:

Approval Request Comment
[Feature/regressing bug #]: OMTC
[User impact if declined]: Delayed by frame
[Risks and why]: Low. just modifying a block list

Comment 7

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 8524064 [details] [diff] [review]
Block D3D11 layers on AMD hardware that delays by a frame

Review of attachment 8524064 [details] [diff] [review]:
-----------------------------------------------------------------

Bas verbally r+

Comment 8

[Lawrence Mandel [:lmandel] (use needinfo)]

Comment on attachment 8524064 [details] [diff] [review]
Block D3D11 layers on AMD hardware that delays by a frame

Discussed with Jeff. The current state is that the browser is pretty much unusable for affected gfx hardware as the current frame is only displayed when the subsequent from should be. This makes things like form entry very difficult.

Beta+
Aurora+

Comment 9

[Jeff Muizelaar [:jrmuizel]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/141ecbdcdb45

Comment 10

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/2a254799e303
https://hg.mozilla.org/releases/mozilla-beta/rev/14159808f8ca

Comment 11

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Seems like we could work around this if we wanted to.

Comment 12

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/141ecbdcdb45

Comment 13

[Jeff Muizelaar [:jrmuizel]]

Looks like this is fixed by passing the PREVENT_INTERNAL_THREADING_OPTIMIZATIONS flag on device creation:
https://tbpl.mozilla.org/?tree=Try&rev=815f50bb7696

Comment 14

[Jeff Muizelaar [:jrmuizel]]

Attachment: D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS

Comment 15

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 8524796 [details] [diff] [review]
D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS

Approval Request Comment
[Feature/regressing bug #]: OMTC
[User impact if declined]: black listed hardware acceleration, slow performance if the blacklist isn't correct
[Describe test coverage new/current, TBPL]: Any drawing will test this path
[Risks and why]: Who knows what D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS does.

Comment 16

[Lawrence Mandel [:lmandel] (use needinfo)]

Should this bug have been reopened cat comment 13? Does the latest patch need to land on m-c as well as aurora and beta?

Comment 17

[Lawrence Mandel [:lmandel] (use needinfo)]

Jeff - Can you handle the landings yourself on Wed?

Comment 18

[Lawrence Mandel [:lmandel] (use needinfo)]

Comment on attachment 8524796 [details] [diff] [review]
D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS

Beta+

Comment 19

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/406cce274929

Comment 20

[Lawrence Mandel [:lmandel] (use needinfo)]

Comment on attachment 8524796 [details] [diff] [review]
D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS

Resetting Beta flag. Jeff and I are going to review this fix Thu morning after it has a day on Nightly.

Comment 21

[Jeff Muizelaar [:jrmuizel]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/3e38373768ad

Comment 22

[Benoit Girard (:BenWa)]

https://hg.mozilla.org/releases/mozilla-beta/rev/a32832a5fe59

Comment 23

[Benoit Girard (:BenWa)]

Opps, I didn't set the author for that patch when rebasing =\.

Comment 24

[Lawrence Mandel [:lmandel] (use needinfo)]

Comment on attachment 8524796 [details] [diff] [review]
D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS

I reviewed this bug with Jeff this morning and gave verbal approval to land. The fix has been verified by people who were experiencing the issue. This has also been on m-c and m-a for a day or so with no known fallout. We're going to take this fix in beta11. Beta+