Charm++ parallel programming system

[Phil Miller <mille121 AT illinois.edu>]

On Wed, Jun 30, 2010 at 09:43, ion.mironescu 
<ion.mironescu AT ulbsibiu.ro>
 wrote:
> Hello everybody,
>
> Tanks to the quick response from Phil Miller now i have  a functional
> charm++ distribution on a CellBE blade. Now that my program is compiling and
> working i want to make a performance analysis trough  Projection. So i have
> same questions
>
> So first of all
>
> - is there any documentation about Projection and accelerated methods for
> programs running on Cell BE  other than the article
>
> "Towards a Framework for Abstracting Accelerators in  Parallel Applications:
> Experience with Cell"  by David M. Kunzman and Laxmikant V. Kalé.

I don't think there is, aside from some example code under
examples/charm++/cell in the Charm++ distribution.

> If not then:
>
> - how much from the functionality  presented in article is implemented in
> the 6.2.0 Charm++standard edition

All of the existing code to support accelerated entry methods is
included in that release.

> - is there a special option for tracemode linking or runtime to visualise
> the timlines (and other statistics) for the SPE-s associated with a PPE
> (like in figure 6 of the cited article ).  I observed that they are 2
> different .sts files but i get only one from the standard  .proj executable.

The paper suggests that the usual "-tracemode projections" should
suffice to get information about what's happening on the SPEs, but I
don't know what the interface in Projections is to get at that. Dave
might be able to answer that more helpfully.

Phil