On Fri, 14 Dec 2001, Tim Rowledge wrote:

> John M McIntosh <johnmci at smalltalkconsulting.com> is widely believed to have written:
>
> > Ok, 5% is quite different from 1%, it's hard to decide if it's noise then.
> > So leave it out.

> I disagree; 5% or 1% is still an improvement and although Andreas _demo_
> changes are a little complicated I think that implementing the original
> suggestion (indexed prims are assumed short by default, named prims are

I don't know.....

Here's what I did, the machinery was already in sqXWindow.c (and is
activated with #define ITIMER)

	int ioLowResMSecs(void)
	{
	#ifdef USE_ITIMER
	  return lowResMSecs;
	#else
	  return ioMSecs();
	#endif
	}

Plus some machinery in to set the alarm handler up in:
	void SetUpTimers(void);

I got a few percent speedup as part of this, and I didn't have to have any
'this primitive might be slow, this primitive is fast' nonsense. I say to
reuse this idiom for the other platforms? Assume that the OS can give us a
wake-up call every 10ms, which we use to decrement a counter?

Also, this does not entirely cut down on the invocation of ioMSecs(), I
still have something on the order of 1000-some invocations of ioMSecs()
per second. Those are on the Smalltalk side. If anyone cares, I'll let the
list quash those.

Scott