Re: [Lcd4linux-devel] Synchronised widget updates

[Michael R. <mi...@re...>]

Hi Martin,

> I think you have gotten me wrong.  The timers run pretty fine.  The
> problem is that each widget has its own timer with its own starting
> time, so the widgets aren't synchronised to begin with.

Ok, thanks for your explanation, now I got you!

BUT (you know, there's alwys a BUT from my side ;-)

I know this "timers getting out of sync" problem for years, never had the time to dig into that.

While your modification surely solves this for widget updates, I want to get rid of it fundamentally.

Apart from that, one (future) extension would be impossible: think of two widgets, both having an update timer of say
500 msec, but the second one is being pre-delayed by 250 msec (there is no such thing like a initial delay at the
moment, but you never can tell with the future..)

I think a clean solution would be to re-sync the timers. Given the fact that there should be no timer less that 10 msec,
and that computers are fast these days, there shouldn't be a difference of mor than 1 or 2 msec. But they sum up with
the current code.

My idea would be to process not only the timer(s) that expired *now* but also timers that expire at now+5 msec, and set
the "last time fired" timestamp for all these timers to *now*.

the last point should be already in there: the timer_process() gets the current timestamp only once, at the very
beginning. So "slow" callbacks would not affect the timing.

Another needed functionality is already inside: the loop not only looks for timers that expire *now*, but also for
timers that expired in the past.


I can think of several possible solutions:

a) when looking for expired timers, look a short period time into the future

b) call timer_process() not exactly at the timeout of the next timer, but a bit later, so more timers would have expired.

c) do some rounding of timeout values, probably depending on the delay timne (e.g. 10% of delay time)


especially c) sounds interesting to me: if a timer has a timeout of 200 msec, round the next start time to 20msec, so it
expires at 0:0200, 0:0400, 0:0600, ....


I hope you all do understand what I'm talking about....



bye, Michael

-- 
Michael Reinelt <mi...@re...>
http://home.pages.at/reinelt
GPG-Key 0xDF13BA50
ICQ #288386781