Mac 3.0 benchmarks
Marcel Weiher
marcel at metaobject.com
Thu Feb 8 18:31:04 UTC 2001
> From: Tim Rowledge <tim at sumeru.stanford.edu>
> I see a similar slowdown on Acorn, and it seems to be almost entirely
> the cost of hitting the event polling prim(s) every 20mSec; change to a
> brain numbing 200mSec and get back most of the performance but
loose all
> control over the pointer....
Yes, changing the poll-period also changed things. However, I don't
think the problem is the (external, VM-based) expense of polling.
Why? Well, my VM is multi-threaded, so each poll just locks a
semaphore and checks a buffer, which is empty most of the time. Not
expensive.
Marcel
More information about the Squeak-dev
mailing list
|