[Webkit-unassigned] [Bug 231352] New: IPC should not have client side message hysteresis
bugzilla-daemon at webkit.org
bugzilla-daemon at webkit.org
Wed Oct 6 23:18:39 PDT 2021
https://bugs.webkit.org/show_bug.cgi?id=231352
Bug ID: 231352
Summary: IPC should not have client side message hysteresis
Product: WebKit
Version: WebKit Local Build
Hardware: Unspecified
OS: Unspecified
Status: NEW
Severity: Normal
Priority: P2
Component: WebKit2
Assignee: webkit-unassigned at lists.webkit.org
Reporter: kkinnunen at apple.com
CC: kkinnunen at apple.com, simon.fraser at apple.com,
wenson_hsieh at apple.com
IPC should not have client side message hysteresis
Bug 231304 added the hysteresis to stream implementation.
The messages may have action at distance that is detectable via out-of-band observations.
At best, the hysteresis needs to be disabled for interfaces with such message behaviours.
At worst, this is not understood and results in hard to understand bugs.
Example:
- Current webgl would benefit from hysteresis
- Current webgl has prepareForDisplay() message
- If prepareForDisplay() is buffered due to hysteresis, no frame gets displayed.
Leaving a message in the client side buffer is poor form.
The problem the hysteresis is solving is;
- client side produces messages fast
- server side processes messages faster
- server side goes to sleep fast
- client side ends up spending significant time in semaphore notification
This should be solved for example with server side hysteresis wait:
When server runs out of messages, it waits with a deadline on the stream semaphore but does not set the "server is waiting flag".
To match the current hysteresis implementation, the wait time could be tuned to be roughly (current hysteresis item count * average time to produce 1 item).
When server does not have new messages messages after the hysteresis wait, it sets the "server is waiting flag" and waits on the semaphore.
--
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20211007/0cdad831/attachment.htm>
More information about the webkit-unassigned
mailing list