Cache and Squeak Performance
Patrick Logan
patrickl at servio.gemstone.com
Fri Feb 13 19:02:00 UTC 1998
>>>>> "JohnM" == Maloney <johnm at wdi.disney.com> writes:
JohnM> ...I believe much of the speed improvement is due to the
JohnM> new "backside" cache design and faster memory bus. ...the
JohnM> 68040 chip run Squeak over four times faster than those
JohnM> based on the 68030 or 68020 chips... the 68040 has 8K of
JohnM> on-chip cache whereas the 68030 has only 64 *words* of
JohnM> cache and the 68020 has no cache at all...
JohnM> The moral of the story is that Squeak is extremely hungry
JohnM> for memory cycles. If you have a machine without a second
JohnM> level cache, it will probably run Squeak faster if you add
JohnM> one. No guarantees, though...
All of this reminds me of an article by Dave Thomas in a magazine like
JOOP maybe eight years ago. In it he predicted that software systems
built using relatively small and simple virtual machines would be more
and more practical as the entire VM fit in cache along with enough of
the application and data.
Well the newer Pentiums this year will have what? Two megabytes of
level two cache?
What does the Apple Mac G3 have?
--
Patrick Logan mailto:patrickl at gemstone.com
Voice 503-533-3365 Fax 503-629-8556
Gemstone Systems, Inc http://www.gemstone.com
More information about the Squeak-dev
mailing list
|