>You have a point here. Just for experimentation I hacked a little CS
>(attached) that should handle all cases pretty well - it makes use of a
>#timerPending flag which is set if we encounter situations that make the
>call to the high-res clock worthwhile (including GCs and some other obscure
>situations).
>
>While the tinyBenchmarks are hardly affected at all, I did find a consistent
>speedup of ~5% within the macro benchmarks. I'd be interested if others (in
>particular on U**x and Mac platforms) find similar results. Please let me
>know - if the results are the same then it seems pretty worthwhile.
>

ok, If I use the lowresmsec timer under os-x then I get

Total time = 257118 milliseconds.  (your changes plus use lowres 1/60 timer)
versus
Total time = 260310 milliseconds.  (3.2.1B4 I sent out yesterday to testers)
versus
Total time = 290056 milliseconds. (your changes but using the high res timer)

So it's more like 1% better for me under OS-X I'll look at classic 
mac later, but this seems like a ok change to make.

-- 
--
===========================================================================
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
===========================================================================