There is also a code tracer working in 2.0 that I maintain and use from
time to time, called Jejak.
It would require a small extension and recompiling a lot of methods, but
it could be made to trace and record all method entries (taking care of
not tracing its own code :)).
It will slow down the VM by a significant amount, of course.
Thierry
Le 24/09/2013 16:30, Alejandro Pulver a écrit :
Thanks for the idea, I'll try emulating shotgun sequencing by increasing
the sampling rate for short periods of time.
But a better alternative would be to enable and disable the simulator
(context's runSimulated) for exact sequences (even if shorter, to avoid
slowdowns). Is it possible with the current implementation?
It's not that I don't appreciate the ideas for the new reflection
framework. but I'm looking for something to use for this project. And we
have only a few weeks. Actually, the most important part is not the
system integration but analyzing the structure of those message
sequences (like word analysis in natural language processing).
On Tue, Sep 24, 2013 at 4:59 AM, Marcus Denker <marcus.den...@inria.fr
<mailto:marcus.den...@inria.fr>> wrote:
On Sep 24, 2013, at 9:41 AM, Alejandro Pulver <alepul...@gmail.com
<mailto:alepul...@gmail.com>> wrote:
> I see. But what about sampling? It shouldn't slow down the
program nearly as much as these "instrumentation" like mechanisms.
>
> And as I'm going to count them anyways, missing the most unused
ones won't be such a problem.
>
> That combined with complete capture for a small run (like a test
or block) would be enough for our needs.
>
>
Yes, one could just use some random selection (I really like the
idea of shotgun sequencing…
http://en.wikipedia.org/wiki/Shotgun_sequencing)
The resulting information should be quite good.
But the first step is that we need a really good, very dynamic, very
general and powerful basis… this is what we want to build
with Reflectivity2 (and Reflectivity3 will have to hook deeper into
the JIT of the VM, but that is another story).
Related to step 3), another source for interesting data is the VM
itself (e.g. caches).
But step by step…
The idea is to take the new possibilities we have in Pharo3 and make
a well engineered realization of partial behavioral reflection,
with two extensions:
-> meta links on instance variables. Very cool
possibilities now that instance variables are reified objects ("Slot")
-> object specific meta links by integrating with a good
proxy implementation
This will already bring amazing capabilities to the system (and
subsume MethodWrappers, simple CLOS style MOPs, perfect for
breakpoints, watchpoints and debugging in general, and it can be
the basis for AOP like things).
Marcus
--
Thierry Goubier
CEA list
Laboratoire des Fondations des Systèmes Temps Réel Embarqués
91191 Gif sur Yvette Cedex
France
Phone/Fax: +33 (0) 1 69 08 32 92 / 83 95
