Jim LaGrone <jlagr...@cs.uh.edu> wrote on 08/27/2009 06:43:57 PM:
>
> On Aug 27, 2009, at 3:55 PM, Igor Peshansky wrote:
>
> > You can also use the place-local idiom, as follows:
> >
> > val norms = Array.make[Double](Dist.makeUnique(), (Point)=>0.0);
> > finish ateach ( (p) in distS.places() ) {
> > for ( (i, j) in distS | p ) { // at this point, p == here
> > norms(p.id) += Math.sqrt(S(i,j).real*S(i,j).real +
> > S(i,j).image*S(i,j).image);
> > }
> > }
> > rawNormPower2 = norms.sum(); // a distributed reduction
> >
> > Here, norms is an array with a unique distribution (which has one
> > point
> > per place). The ateach loop again spawns parallel iterations on each
> > place (but now there's one per place covered by the distribution).
> > The
> > body of the ateach has a sequential loop that runs over the part of
> > distS
> > that lives in that place. Each iteration of that sequential loop
> > updates
> > the local element of norms. There is no need for atomic, since all
> > updates happen sequentially. Once the ateach loop is done, a
> > (distributed)
> > reduction is performed over the norms array, and the sum is stored
> > in the
> > final result location.
>
>
> Using this method and Java backend, I have
>
> rawNormPower2 = 0;
> val norms = Array.make[Double](Dist.makeUnique(), (Point)=>0.0);
> finish ateach( (p) in distS.places() ){ //line 327
> for( (i,j) in distS | p ){
> norms(p.id) += Math.sqrt(S(i1,j1).real*S(i1,j1).real +
> S(i1,j1).image*S(i1,j1).image);
> }
> }
> rawNormPower2 = norms.sum()/( N*Mc ); //line 333
>
>
> but get this
>
> /Users/jlagrone/GradSchool/Research/workspace/X10-code/SSCA3-X10/src-
> par/State.x10:327: Could not infer type for formal parameter.
> /Users/jlagrone/GradSchool/Research/workspace/X10-code/SSCA3-X10/src-
> par/State.x10:327: Could not infer type for formal parameter.
> /Users/jlagrone/GradSchool/Research/workspace/X10-code/SSCA3-X10/src-
> par/State.x10:333: No valid method call found for sum() in
> x10.lang.Array[x10.lang.Double].
Sorry, this will teach me to actually test the code before sending it out.
Line 327 should be "finish for ( p in distS.places() ) async (p) {".
In line 33, replace "sum()" by "reduce((a:Int,b:Int)=>a+b, 0)".
> Also, is it safe to assume this will work with Java backend on dual-
> core machine? Or should I do development on cluster from here on out?
> How should the program be executed. We usually use slurm, but is there
> some special method?
If you are working with the 1.7.* releases, the Java backend should work.
If you want to use a cluster, you should use the C++ backend.
To execute the program on a JVM, use the "x10" script supplied with the
distribution. For running on the cluster of Linux machines, you can use
mpiexec from an OpenMPI installation.
Igor
--
Igor Peshansky (note the spelling change!)
IBM T.J. Watson Research Center
XJ: No More Pain for XML's Gain (http://www.research.ibm.com/xj/)
X10: Parallel Productivity and Performance (http://x10.sf.net/)
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users
