[MLton] More on Parallel Runtime

[Eric McCorkle]

On Oct 20, 2007, at 3:40 PM, Philip Schatz wrote:

(clipped)

Another point which I need to address is the write barrier which the  
Cheng-Blelloch algorithm needs, and I believe any semispace algorithm  
also requires.

My first intuition was to have a global variable which holds the  
state of affairs in the heap (use slot 1, use slot 2, use both  
slots).  I don't like that this makes writes more expensive, though.   
Also, it has to be modified for generational collection.

Have you dealt with this at all (or does anyone else have any ideas)?

-- 
Eric McCorkle
Brown University
CS Graduate Student