Thanks a lot! I think both solution work but I am still wondering what's the
best way to copy entire distributed array to a local array in terms of
communication cost.
Could you give me some advice?
Kirak
On Fri, Dec 17, 2010 at 3:48 AM, Vijay Saraswat <vi...@saraswat.org> wrote:
> To add to Josh's comments. If you have an object o at a place p, and you
> need to go to a place q, and then come back and modify o (and not a copy
> of o), you have to use a GlobalRef. (You can also use a PlaceLocalHandle.)
>
> Thus:
>
> val a = makeArray();
> val globalA = GlobalRef(a);
> val home = here;
> at (here.next()) {
> val x = ...
> at (home) { // go back to the original place
> globalA()(0) = x; // assign x to a(0).
> }
> }
>
>
> at(here.next())
> > at(here)
> > {
> > for(p in src_array)
> > {
> > src_array(p) = 10000;
> > }
> >
> > }
> >
> > for(p in src_array)
> > Console.OUT.println(src_array(p));
>
>
>
> On 12/17/2010 1:14 AM, Josh Milthorpe wrote:
> > Hi Kirak,
> >
> > I'm assuming that src_array was created at (here).
> >
> > In the first example, the src_array that is modified is actually a copy
> > of a copy - it is copied to here.next() in the first line and then again
> > (to the same place). Note that at(here) is _not_ a noop - it copies all
> > variables in the closure environment! This is to make the behaviour of
> > the at statement consistent regardless of the target place.
> >
> > The second example doesn't actually work for me, etiher (against SVN
> > HEAD) - if I print the contents of src_array, they have not been
> > updated. Again, the src_array that is updated is a copy, so the updates
> > don't affect the original src_array.
> >
> > Have a look at the RemoteArray class and the Array.copy(...) and
> > Array.asyncCopy(...) methods. These support a variety of patterns of
> > local and remote updates - hopefully one of them matches your
> requirement.
> >
> > Cheers,
> >
> > Josh
> >
> > On 17/12/10 16:16, Kirak Hong wrote:
> >> Hi,
> >>
> >> I have another question regarding X10 array. I tried to modify a local
> array
> >> from remote place using following code. I noticed that there's some
> >> computation running, but the content of array was never changed.
> >>
> >> at(here.next())
> >> at(here)
> >> {
> >> for(p in src_array)
> >> {
> >> src_array(p) = 10000;
> >> }
> >>
> >> }
> >>
> >> for(p in src_array)
> >> Console.OUT.println(src_array(p));
> >>
> >> However, the following code seems to be working, although it is highly
> >> inefficient.
> >>
> >> at(here.next())
> >> for(p in src_array)
> >> {
> >> var x : int = 0;
> >> at(here)
> >> {
> >> x = 10000;
> >> }
> >> src_array(p) = x;
> >> }
> >>
> >> I looked at the X10 programming manual and it says it will implicitly
> copy
> >> an object to another place if that object is used within 'at' clause.
> >> However, I don't understand why above behavior happens. Can anybody tell
> me
> >> the reason?
> >>
> >> Thanks,
> >> Kirak
>
>
>
>
> ------------------------------------------------------------------------------
> Lotusphere 2011
> Register now for Lotusphere 2011 and learn how
> to connect the dots, take your collaborative environment
> to the next level, and enter the era of Social Business.
> http://p.sf.net/sfu/lotusphere-d2d
> _______________________________________________
> X10-users mailing list
> X10-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/x10-users
>
--
Kirak Hong
PhD Student, College of Computing
Georgia Institute of Technology
------------------------------------------------------------------------------
Lotusphere 2011
Register now for Lotusphere 2011 and learn how
to connect the dots, take your collaborative environment
to the next level, and enter the era of Social Business.
http://p.sf.net/sfu/lotusphere-d2d
_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users
