[abc] OOPSLA paper

The OOPSLA deadline is on March 18. We must do our
trace-based experiments paper!

http://www.oopsla.org/2006/researchPapers.html

Thanks to Julian for mailing round a to-do list.

Can everyone summarise the state of the benchmarks they
have worked on for the not-a-cc-and-not-an-ecoop papers.
In particular:

JigSaw - some j2j bug, but not isolated [Neil - please
          describe what you're seeing, and where people
          can find the benchmark to try and reproduce the
          errant behaviour]. Please include instructions
          for building just with Soot, so if some Soot
          (but not abc) expert is feeling courageous, they
          can give a hand.

Weka - not sure what the story was here?

Approve - Eric, this sounds a great benchmark. Can you
           check into tmbenches, or is it too sensitive for
           that? Please summarise the problems you're having.

We also had some results with very disappointing numbers, at
least on the face of it. Can someone dig these out, so we
can try to attack the performance difficulties?

On the bright side, during my visit to IBM last week, John
Field and others were quite impressed by even slowdowns of
a factor of 30... a while ago they tried similar instrumentation
and found it unworkable.

Only 7 weeks or so to go, so we need to finalise at least one
benchmark per week :-)

-O
Received on Tue Jan 24 17:09:03 2006