1369541 - (omtp) [meta] Off Main Thread Painting Redux

Status: RESOLVED FIXED

(Core :: Graphics: Layers, enhancement, P3)

Description

[David Anderson [:dvander] - inactive, e-mail if emergency]

This is a meta bug for take 2 of OMTP. See bug 1256578 for Jerry's earlier work on OMTP. Most of the problems encountered there still apply.

Comment 1

[Mason Chang [Inactive] [:mchang]]

Latest try run - https://treeherder.mozilla.org/#/jobs?repo=try&revision=5ad3d9ca0348171d926a5fd86df7581f7b7302b6

Comment 2

[Mason Chang [Inactive] [:mchang]]

Latest try run with 1376859 - https://treeherder.mozilla.org/#/jobs?repo=try&revision=ad56bd3f69a9692b2c08658be10b48526e958d89

Comment 3

[Mason Chang [Inactive] [:mchang]]

Latest try run with bug 1378538 - https://treeherder.mozilla.org/#/jobs?repo=try&revision=13143a8b1b8db007bb87d0762b0dd223039b57a4

Comment 4

[Mason Chang [Inactive] [:mchang]]

Try run with bug 1379029 - https://treeherder.mozilla.org/#/jobs?repo=try&revision=5f95488ea02b5d1ebdc6b558752e29ac0b12cf84

Comment 5

[Mason Chang [Inactive] [:mchang]]

Try run with reviewed bug 1379029 - https://treeherder.mozilla.org/#/jobs?repo=try&revision=5a5c8ad7f180904f1ae5000450dc587223f0656d

Comment 6

[Mason Chang [Inactive] [:mchang]]

Try run with both bug 1379029 and 1379322 - https://treeherder.mozilla.org/#/jobs?repo=try&revision=5e1cfcb52ad41b7d26ffdc5d34e65f86855cc42a

Comment 7

[Mason Chang [Inactive] [:mchang]]

Updated try with bug 1380483 - https://treeherder.mozilla.org/#/jobs?repo=try&revision=0c7b88637c7165be1c1d26f20ab6a2fb69058672

Comment 8

[Mason Chang [Inactive] [:mchang]]

Try with bug 1380493 - https://treeherder.mozilla.org/#/jobs?repo=try&revision=c26af68ab1e88c1aa7482fe769a1254d942e1df4

Comment 9

[Mason Chang [Inactive] [:mchang]]

Latest try with Bug 1381172. https://treeherder.mozilla.org/#/jobs?repo=try&revision=e97392a21ec05f5c7835e25c549475109c1be5e3

Comment 10

[Mason Chang [Inactive] [:mchang]]

sync omtp: https://treeherder.mozilla.org/#/jobs?repo=try&revision=49764c0dc2aa2a253cfd45ad41221fa13674cb58
async omtp: https://treeherder.mozilla.org/#/jobs?repo=try&revision=50b187b6aaa70f7916ce66ad8d7fcbb0382ed104

Comment 11

[Mason Chang [Inactive] [:mchang]]

Latest try with bug 1388921 and bug 1387639.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=745f33fe63a5ccf292eb9e96fc3781f37dc473e1

Comment 12

[Ryan Hunt [:rhunt] (on leave until early May)]

Latest try with bug 1390295, bug 1390679, and bug 1390755.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=a3d6944410ea5ccbb32f7527489c7ef08d45567c

Comment 13

[Mason Chang [Inactive] [:mchang]]

Latest try - https://treeherder.mozilla.org/#/jobs?repo=try&revision=74c059a0033ad236c47a655aa8973eb7f8b2788e

Latest status since I'm on PTO. There are three main blocking issues left:

1) Direct2D performance still doesn't make sense, this is bug 1387327. Some amount of investigation went in bug 1392453. This seems to happen across multiple GPU types as well.
2) Skia OMTP still has some reftest failures where it looks like the buffer is getting corrupt. The tests are mostly /layout/reftests/position-dynamic-changes/relative/move-right-bottom.html
3) Leakcheck failures - OMTP seems to be leaking textures somehow but I haven't been able to track these down.

Comment 14

[Bas Schouten (:bas.schouten)]

Latest try - https://treeherder.mozilla.org/#/jobs?repo=try&revision=3aee6db3b881dd3f5b2d9990cc61c57bd9c17693

This contains the fix to 1396474.

Comment 15

[Bas Schouten (:bas.schouten)]

It looks like the latest try run is green except from a tiny rounding error in Skia on Linux (1 difference of 1 pixel), which should just be fuzzed. So except for possible performance considerations (I've fired off a Talos run) we should be good to go.

Comment 16

[Bas Schouten (:bas.schouten)]

(In reply to Bas Schouten (:bas.schouten) from comment #15)
> It looks like the latest try run is green except from a tiny rounding error
> in Skia on Linux (1 difference of 1 pixel), which should just be fuzzed. So
> except for possible performance considerations (I've fired off a Talos run)
> we should be good to go.

Looks like I was wrong, there's one intermittent test failure on Linux -debug- builds. Ryan, could you have a look at that? Note that this doesn't bar us from switching it on on windows.

Comment 17

[Ryan Hunt [:rhunt] (on leave until early May)]

I'm having a difficult time reproducing the intermittent test failure. I've filed bug 1397083 to track it.

Comment 18

[Ryan Hunt [:rhunt] (on leave until early May)]

Latest try run with OMTP enabled:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=f0df9bcbbd3d4cc9d182debdcd3aa6df58756d83

There are some new failures.
1. Crashes in mozilla::ipc::MessageChannel::Send (this is bug 1400299)
2. R8 failed reftest-no-paint
3. failure in layout/reftests/text/color-opacity-rtl-1.html and color-opacity-rtl-2.html
4. leakcheck | gpu process: 128 bytes leaked (TextureHost, TextureSource, TextureSourceProvider, nsTArray_base)

Some of them probably have the same root cause.

Comment 19

[Ryan Hunt [:rhunt] (on leave until early May)]

Latest try run with OMTP enabled:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=b66f20de76828074df3c1fa281d4d4237e358ec5

One issue carried over, the rest were resolved.

1. leakcheck | gpu process: 128 bytes leaked (TextureHost, TextureSource, TextureSourceProvider, nsTArray_base)