Re: MCL 2.0.1, ptable & egc

[bill@cambridge.apple.com (Bill St. Clair)]

At 12:39 PM 12/29/93 -0500, Doug Currie, Flavors Technology, Inc. wrote:
>At  4:27 PM 93.12.28 -0600, Bill St. Clair wrote:
>>When I timed compiling a file on my IIfx, it went about 10% faster
>>without ptable.
>
>How much of this slowdown is mitigated by using egc? It's difficult to
>quantify, I'm sure, but do you really recommend running a IIfx without
>ptable/egc? [I assume that Virtual Memory is disabled anyway, as this would
>be much slower!?]
>
>e

The answer to your questions depends a lot on what you're doing. As
Kalman said in one of his answers yesterday, turning on EGC makes the
average pause time less, but often makes the total time spent in GC
increase. Whether EGC (and ptable) will be a win for your application
depends on how big your heap is and how much ephemeral garbage your code
creates.

I doubt that it's ever worthwhile to use the EGC without ptable
(though a possible counterexample is a Mac Plus, which takes 18 seconds
to GC a 3 meg heap and 5 seconds to EGC ephemeral level 0),
but you'll have to decide on whether to enable EGC or not according
to the requirements and behavior of your application.