[BUG] why is frame rate in 3.2.5 slower?
John M McIntosh
johnmci at smalltalkconsulting.com
Wed Mar 13 17:54:23 UTC 2002
>But then I'm back to hitting the clock every 1/1000 of a second.
>Sigh but I'll go away and consider what this really costs.
ps if we are going to force a check every 1ms or so, could we
streamline checkForInterrupts a bit to remove the trottle logic, and
forgo decrementing that counter, just check to see if it's zero.
I'm also assuming in the external prim call logic we could remove the
clock checks because certainly on a return from a long external prim
we'll have reset the counter because surely the timer must interrupt
most system calls etc to guarantee some degree of accuracy.
say in primitiveResponse
(timerPending and: [interruptCheckCounter == 0 & successFlag])
ifTrue: [checkForInterrupts()].
mmm in checkForInterrupts I'm sure we could only look at the high
resolution clock when nextWakeupTick != 0, and use the low resolution
clock to deal with nextPollTick. Mind if I knew I was getting called
here every about 1/1000 of a second I could just have a counter and
count to 500 and not screw with the clock call.
Well that doesn't answer what to do if there isn't timer support in
your Operating system. But maybe we have two execution paths here,
and the VM implementer gets to set a flag when he builds?
--
--
===========================================================================
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com
===========================================================================
More information about the Squeak-dev
mailing list
|