It would probably take less work to marshal timer functions to a single
timer thread than reworking this to other interfaces. I didn't do this
originally because I know that it's deadlock city :) There's a lot of subtle
deadlocks you end up with by having a single thread with a rendezvous with
many others to synchronize start/stop and execute timer operations.

Still, it's worth doing, and when it's done it'll be more efficient than all
the other implementations.

2008/10/3 Michael Nordman <[EMAIL PROTECTED]>

> This may be a good candidate for a CPAPI function for use on the 'plugin
> thread'. Another candidate for CPAPI could be thread message
> sending/dispatching, when sending to the 'plugin thread'.
>
>
> On Fri, Oct 3, 2008 at 9:29 AM, Matt Perry <[EMAIL PROTECTED]> wrote:
>
>>
>> Hmm.. in that case, I'd prefer that we didn't use it except as a
>> stopgap for --gears-in-renderer.
>>
>> Zach, can we use the win32 timers if OFFICIAL_BUILD==1, and android's
>> otherwise?
>>
>> On Fri, Oct 3, 2008 at 8:04 AM, John Ripley <[EMAIL PROTECTED]> wrote:
>> > [reviewlog]
>> > Just so you know - the Android version of timers is extremely
>> inefficient
>> > with thread usage. It starts one thread per timer. Ideally we should be
>> > using a single thread maintaining all timers and marshaling through
>> that.
>> > Just a todo for future.
>> >
>> > 2008/10/2 Zach Kuznia <[EMAIL PROTECTED]>
>> >>
>> >> Hello mpcomplete,
>> >>
>> >> I'd like you to do a code review.  Please execute
>> >>        g4 diff -c 8469672
>> >>
>> >> or point your web browser to
>> >>        http://mondrian/8469672
>> >>
>> >> to review the following code:
>> >>
>> >> Change 8469672 by [EMAIL PROTECTED] on 2008/10/02 15:18:19 *pending*
>> >>
>> >>        Switch Chrome build to using the Android version of timers, and
>> fix
>> >> a bug with sync messages.
>> >>
>> >>        R=mpcomplete
>> >>        [EMAIL PROTECTED]
>> >>        DELTA=15  (9 added, 4 deleted, 2 changed)
>> >>        OCL=8469672
>> >>
>> >> Affected files ...
>> >>
>> >> ...
>> //depot/googleclient/gears/opensource/gears/base/chrome/module_cr.cc#2
>> >> edit
>> >> ... //depot/googleclient/gears/opensource/gears/timer/timer.cc#3 edit
>> >> ... //depot/googleclient/gears/opensource/gears/timer/timer.h#5 edit
>> >>
>> >> 15 delta lines: 9 added, 4 deleted, 2 changed
>> >>
>> >> Also consider running:
>> >>        g4 lint -c 8469672
>> >>
>> >> which verifies that the changelist doesn't introduce new style
>> violations.
>> >>
>> >> If you can't do the review, please let me know as soon as possible.
>> >>  During
>> >> your review, please ensure that all new code has corresponding unit
>> tests
>> >> and
>> >> that existing unit tests are updated appropriately.  Visit
>> >> http://www/eng/code_review.html for more information.
>> >>
>> >> This is a semiautomated message from "g4 mail".  Complaints or
>> >> suggestions?
>> >> Mail [EMAIL PROTECTED]
>> >
>> >
>> >
>> > --
>> > Google UK Limited, Registered Office: Belgrave House, 76 Buckingham
>> Palace
>> > Road, London SW1 9TQ
>> > Registered in England Number: 3977902
>> >
>> >
>>
>
>


-- 
Google UK Limited, Registered Office: Belgrave House, 76 Buckingham Palace
Road, London SW1 9TQ
Registered in England Number: 3977902

Reply via email to