Could not we have a group of motivated guys really checking MC1.6

Hi guys

I spent 3 hours looking for improvement in MC1.5. (of course I got  
problem when loading the code - but this can be fixed)
When I could load it, it is indeed much faster.

Finally And matthew nicely pointed me to his improvement.
I will have a look at the specific speedup made by matthew and see if  
I can make that work.

Now instead of extracting a speedup from that code (and looking like a  
thief)
I would prefer that we do a task force and check MC1.6 improve it  
clean it if necessary.

So does anybody with knowledge want to help on that topic?

Stef



Here is my conversation with matthew

> Hi matthew
>
> could you give me some hints abotu what I should look for?
> Because I loaded these two files I found on
>
> http://installer.pbworks.com/LevelPlayingField-Monticello15
>
> But this is not that simple to spot changes.

I'd rather you just used MC1.5 rather than port it, but ok

There are two speedups in Monticello 1.6:

Scanning the class heiarchy looking for package elements is 10x
faster, due to a rewrite of the scanning code.

Loading is 3-6x faster, especially for large packages, due to
the atomic loader.

I believe you can't make much use of the second, as the atomic
loader doesn't support traits yet.


The rewritten scanning code is in class PackageInfo, category
single-pass. To merge it into Monticello, simply call
PackageInfo >> classesDo:methodsDo:displayingProgress: from
MCPackage >> snapshot. Attached is a changeset of the relevant
code, extracted from mc1.6. You may have to slightly modify
MCPackage >> snapshot, as it contains instructions for
interacting with the Orphanage, which I'm guessing your branch
of Monticello does not include.

--
Matthew Fulmer -- http://mtfulmer.wordpress.com/

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project