commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kohsuke Kawaguchi <kohsuke.kawagu...@sun.com>
Subject Re: [Javaflow] Continuation questions
Date Sun, 29 Jan 2006 00:54:03 GMT
Jean Morissette wrote:
> Hi developers,
>    I'm the lead developer of the JCyclone project
> http://jcyclone.sourceforge.net/ and I'm very interested by Java
> continuation.  I want to use continuations in JCyclone as logical
> threads and create a fine-grained predictable thread scheduler.  You
> may want to take a look in our CVS to see our continuation
> implementation prototype, under the sandbox directory.

Hi! I'm one of the developers of the project.

> Is there some limitations of your implementation?  By exemple, do you
> instrument synchronized statements?  I was thinking to just replace
> them with ReentrantReadWriteLock. Also, there can be some issues with
> finally statements; do you handle them?

Yes, there are a few issues. For example, monitors acquired through 
'monitorenter' instruction isn't released when the stack capture 
happens, causing IllegalMonitorStateException.

Finally block has a similar problem. We've been thinking about just 
inlining JSR/RET blocks to avoid the problem altogether.

> I would be interested to give my contribution in your project. 
> Please, let me know if I can help.

Thanks for the offer. I attached the TODO file for your consideration. 
We've been using BCEL for the instrumentation, but trying to migrate to 
ASM now (I think BCEL version is still more stable, but that's where we 
could use some help.) I think ASM has a better dataflow analysis 
package, which hopefully would let us attack the abovementioned 
instrumentation problems.


-- 
Kohsuke Kawaguchi
Sun Microsystems                   kohsuke.kawaguchi@sun.com

Mime
View raw message