Here's the next variance of scheduler. Could you PTAL? Thanks!
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
virtual bool ShouldBlockBeginMainFrameWhenIdle() const;Maybe we should rename the function in a follow-up cl 😊 So that each cl does one small thing
// - Allows beginning `BeginMainFrame` while idle to begin preparing the next
// frameWe should follow-up on how this could work when cc scheduler is no longer responsible for sending the BMF signal.
Maybe we can have a second unit test that checks things like ACTION::COMMIT or ACTION::SEND_BEGIN_MAIN_FRAME or ACTION::ACTIVATE_SYNC_TREE
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
virtual bool ShouldBlockBeginMainFrameWhenIdle() const;Maybe we should rename the function in a follow-up cl 😊 So that each cl does one small thing
https://chromium-review.git.corp.google.com/c/chromium/src/+/7657486 will rebase once that's landed
Maybe we can have a second unit test that checks things like ACTION::COMMIT or ACTION::SEND_BEGIN_MAIN_FRAME or ACTION::ACTIVATE_SYNC_TREE
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Code-Review | +1 |
Every one of our problems can be solved with a new cc scheduler subclass.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Jonathan RossMaybe we can have a second unit test that checks things like ACTION::COMMIT or ACTION::SEND_BEGIN_MAIN_FRAME or ACTION::ACTIVATE_SYNC_TREE
Done
I don't see it!
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Jonathan RossMaybe we can have a second unit test that checks things like ACTION::COMMIT or ACTION::SEND_BEGIN_MAIN_FRAME or ACTION::ACTIVATE_SYNC_TREE
Stacy GaikovaiaDone
I don't see it!
Apaprently the local tree had a rebase warning that prevented the upload. My bad
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
+vmiura@ for thoughts on this variant of the scheduler. To prevent re-entrancy and to restrict to immediate/waitforscrolls
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Sorry for slow review. Will take a look by tomorrow.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Sorry for slow review. Will take a look by tomorrow.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Code-Review | +1 |
LGTM, thanks
This state machine's supports:nit: `This state machine supports` ?
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
nit: `This state machine supports` ?
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
6 is the latest approved patch-set.
No files were changed between the latest approved patch-set and the submitted one.
Introduce SlimSchedulerStateMachine to prevent re-entrancy
This state machine supports:
- Prevents re-entrancy during a VSync Interval. We will no longer repeat an `Action`
- Removed BeginImplFrameDeadlineMode::REGULAR and LATE. As these modes lead to the next VSync being IMMEDIATE. This can result in two frame submissions within less than a VSync interval. This leads to dropped frames or exhaustion of frame buffers, which pauses rendering.
- Restricts different `Action` to different `BeginImplFrameState`. So that we only attempt to Commit and Activate after submitting the current frame, when we are IDLE.
- Allows beginning `BeginMainFrame` while idle to begin preparing the next frame
- Allows `PrepareTiles` while idle, so that this doesn't slow down frame submission.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |