I'm new to the list
Klaus D. Witzel
klaus.witzel at cobss.com
Sun Nov 6 18:48:16 CET 2005
Hi list,
I just read (in the #squeak IRC log) that this list is now served by
news.gmane.org, so it's time for me to subscribe :)
I have two interests here: tracking of news for The Weekly Squeak
( http://minnow.cc.gatech.edu/squeak/5740 ) and learning new ways (& tipps
& tricks :) on how performance constraints can and will influence
development of outstanding software (emphasis on but not limited to
Squeak).
I have just finished tweaking Squeaks's VM for allowing me to define sort
of "basic" ClassBox for which the VM can report:
- memory allocation and reclamation
- message sends entering the box
- message sends within the box
- message sends leaving the box
When the box (which can be redefined in Squeak without compiling the VM)
is empty, the above changes have less than 1% performance penalty on
bytecode execution and less than 10% performance penalty on message sends
(relative to '0 tinyBenchmarks'), so are almost acceptable even for HTTP
servers and multimedia applications.
My next steps are to brain storm and populate example boxes as
representable sets which exhibit interesting characteristics. For example,
"what Number methods are involved in Collection operations", "what
Magnitude methods are involved in BitBlt operations", just to give an
idea. The boxes I define have two configurable components: the out-of-box
behaviors and the in-the-box behaviors, that's what makes them flexible.
I'd be happy to receiving *any suggestions* for example boxes from members
of the list, for example what is currently defined for benchmarking
Exupery.
/Klaus
More information about the Exupery
mailing list