Easier threading with RASCAL

One more thing that most developers will implement in their RASCAL based 
applications is threading. The original specification required the 
application to call the rascal_work function periodically, telling it 
how much time it can spend waiting for events. This would require the 
developer to organize the main program loop so that the application 
periodically quits it to serve the network. This can not only be 
ineficient, but can also be impossible, like when the application spends 
most time in a third party library call and is unable to requently 
interrupt it. This could possibly require to rewrite significant parts 
of the application code, leaving ground for new bugs.

The alternative is to spawn a thread dedicated to working with the 
network, leaving the main program thread (the one with the GUI, for 
example) alone. The best recommended threading model is a thread per 
CPU, which is what I'm about to do: spawn this number of dedicated 
rascal threads from within the library startup code. The only thing that 
I'm still curious about is a way to synchronize threads. Perhaps the C 
version will expose internally used structures and functions related to 
the internally used mutex class, while the C++ version will also expose 
the class itself.


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click