[aspectj-dev] What does ajc do to speed up matching?

["Eric Bodden" <eric.bodden@xxxxxxxxxxxxxx>]

Dear ajc developers.

I am currently weaving quite a large application with our abc compiler
and especially the matching process takes quite some time. This is due
to the very naive matching strategy we have, which is basically "for
all advice for all statements generate the residue for this statement
and advice". If you have millions or more of statemements, this can
take quite some time. abc was never optimized for matching time but as
the size of our benchmarks increases, I feel that it's maybe
worthwhile to spend a little time on that.

I am pretty sure that ajc does something (much?) more clever, so I was
wondering whether this is documented anywhere. Could anybody give me a
high-level overview of the tricks you apply?

Cheers,
Eric

--
Eric Bodden
Sable Research Group
McGill University, Montréal, Canada