Charm++ parallel programming system

[Kristian Rune Larsen <s991029 AT student.dtu.dk>]

Hi 

I have some charm++ questions.

The first question is just to clarify array initiation. I have tried to run 
my 
program on four processes on two nodes , but I get seg-fault when trying to 
initiate chare arrays on node 2. The arrays are initiated by chares, one on 
each processor, and the seg-fault happens when the chare on the second 
processor on node two tries in initiate a chare array. If I create the arrays 
in advance and pass their proxies to the chares in the constructor everything 
works fine. Does arrays need to be initialised on PE 0 ? 

The default map object for a chare array is a round-robin map. I can't figure 
out, is this a kind of load-balancing ?. Is the "home" PE the processor where 
the array element is located or is just responsible for knowing where the 
element is ?  

I have tried to profile my program using the sun profiler called "analyzer". 
I 
think it is a java frontend for the collect program. My program is called 
pprospect. If I use
$ collect -<..> charmrun pprospect args ++<..>
I get the profiling data for charmrun, and if I try to run the stand alone 
version
$collect -<opt> pprospect args  
It ends prematurely, but with exit success. The stand alone version works 
fine 
outside of the collect program. Has anyone experienced this ?  


best wishes 
Kristian Rune Larsen