[aosd-discuss] Performance Studies on AspectJ

[Laurie Hendren]

In recent work we have started the process of defining appropriate
dynamic measurements and dynamic metrics to evaluate the performance
and behaviour of AspectJ programs.    We have applied our approach on
a selection of benchmarks to illustrate our ideas,  and we can show
some potential performance bottlenecks.

A significant portion of the bottlenecks can be avoided by adopters
who have an appropriate understanding of the current AspectJ best practice,
others can be improved via compilation techniques. Our tools are intended
to help extend such understanding for particular applications by exposing the
bottlenecks to the users and compiler writers.

Clearly this is only a start,  and we continue to work on our ideas
and make a more comprehensive set of benchmarks and metrics.   We are
welcome to comments,  and also would greatly appreciate more benchmark
programs.

A technical report version of our work is available at:

http://www.sable.mcgill.ca/publications/techreports/#report2003-8

Cheers, Laurie

+----------------------------------------------------------------+
| Laurie Hendren, Professor, Computer Science, McGill University |
| Visiting Fellow, Oxford University Computing Laboratory,       |
| Wolfson Building, Parks Road         Fax: +44 (0) 1865 273839  |
| Oxford, UK  OX1 3QD    Tel: +44 1865 283551, 01865 283551 (UK) |
| hendren at cs.mcgill.ca       http://www.sable.mcgill.ca/~hendren |
+----------------------------------------------------------------+

On Thu, 15 Jan 2004, Pan-Wei Ng wrote:

> Any performance studies on AspectJ to show potential adopters that
> performance is not a worry?
> Is there any effort to make AspectJ into JSR or into some standard. Folks
> who are building large systems would need to be convinced about the
> longevity of AspectJ
>
> /pan-wei
>