Just a reminder that getting the OOPSLA paper written, proofread and
polished is VERY IMPORTANT -- there's less than a week now.
Oege summarised the plan of action a few days ago -- see attached mail
below. I think the distribution of 'who does what' is still unchanged.
I have just committed a new version of experiments.tex with an almost
complete figure 7. The only numbers missing is reweave with TMNoLeak,
which I will supply as soon as possible, and the rows for
hashcode/aprove and jigsaw/lor, which Eric will hopefully fill in in the
very near future.
To this end, Eric, I am willing to give up my lock on Cardinal. I will
concentrate on producing a draft of section 4 with Julian, and so you
can use Cardinal for the (near) future. Please still talk to me on IRC
before starting any big jobs, as I still have benchmarks running that I
would have to stop for you.
The only other number that I think I haven't supplied yet is the effect
of -debug thisJoinPointObject on Jigsaw/LUnInMeth. This is what is
currently running on musketeer, but I think the numbers Eric needs to
gether are more important -- hence the statement above.
Note that while Julian's optimisation for simplifying skip transitions
seems to mainly benefit TMNoIndex, and seems to have a rather less
pronounced effect when indexing is used. This makes sense, since with
indexing adding negative bindings is already greatly simplified (we pick
out the relevant set of disjuncts and immediately discard everything
that obviously becomes false) -- thus, further reducing the work done
doesn't show a very marked improvement. I'm still working on gathering
numbers for the jhotdraw/safeenum variations with this optimisation, but
it doesn't look like it will be very dramatic, so, again, Eric's
benchmarks will take precedence.
So, once again -- here is the plan of action:
Oege de Moor wrote:
>
> Here's a rough plan of how we might proceed to produce a draft.
> Please use LOCKs on irc when editing.
>
> Laurie: paper intro, general story (section 1)
>
> Eric: collect numbers for lor+jigsaw and hashcode+aprove on
> cardinal;
> paper systems (section 2)
>
> Oege: paper experiments (section 3)
> [in particular fleshing out section 3.3 to determine
> what numbers exactly are needed for that]
> - try to explain jhotdraw/ajhotdraw discrepancy.
>
> Pavel & Julian: paper techniques (section 4)
>
> Ondrej: paper future work (section 5)
>
> Neil: [polish AOSD talk]
> - paper experiments (section 3), in particular
> benchmark descriptions [liaise with oege to avoid
> cvs conflicts]
- P
Received on Sat Mar 11 19:34:59 2006
This archive was generated by hypermail 2.1.8 : Tue Mar 06 2007 - 16:13:27 GMT