Christoph Pospiech <christoph.pospi...@de.ibm.com> wrote on 09/07/2010
08:15:58 AM:
> On Tuesday, September 07, 2010, Igor Peshansky wrote:
> > If you do not get the C++ line numbers in the trace, you can also run
your
> > program under gdb and set a breakpoint in
> > 'x10aux::throwException(x10aux::ref<x10::lang::Throwable>)'. When you
hit
> > this breakpoint, go up a frame and you should end up on the C++ line
that
> > caused the exception to be thrown. Again, the "//#line" comment above
> > that
> > C++ line should point you to the corresponding X10 source line.
>
> Igor,
>
> following your suggestion above, I was able to pin the exception down to
the
> following single X10 statement.
>
>
> finish ateach (pt in v_src ) {
> /* ... some lines left out ... */
> /**
> * Scatter the vector v
> */
> v_src(pt).copyFrom(j_min(pt),
> v,j_min(pt),j_len(pt));
> /* ... some more lines left out ... */
> }
>
> In this code snippet, we have the following data types.
>
> static type Array1D = Array[Double]{rank==1};
> var v: Array1D;
> global val v_src: DistArray[Array1D]{rank==1};
>
> /**
> * v_src is generated from v by scattering
> * the data across the MPI tasks. The following
> * two distributed arrays describe which parts
> * of v get where.
> */
> global val j_min: DistArray[Int]{rank==1};
> global val j_len: DistArray[Int]{rank==1};
>
> All objects of class DistArray are build on the same distribution.
> val Dv_tmp:Dist = Dist.makeUnique(D.places());
>
> Do you see what is wrong with this copyFrom() ? Actually, the exception
is
> thrown only when run on more than one place, and it is thrown when the
place
> v_src.dist()(pt) is not v.home(). It somehow smells like a placement
error of
> some sort.
Hi, Christoph,
You still did not say what *kind* of exception you're getting, but from
your description it looks like you get a ClassCastException from a
dynamic check. I suggest you try compiling with -STATIC_CALLS and see
what compilation error you get on that line. My suspicion is that it's
not "v" that's a problem, but rather one of "j_min" or "j_len"...
> BTW, I was using the following statement to start the gdb debugger.
>
> mpiexec -np 2 ddd ./matmul
You can also do "mpiexec -np 2 -gdb ./matmul", and it will spawn an xterm
with a gdb session for each process.
> This gave me a GUI for each MPI task. Do you know whether the PTP
parallel
> debugger could be used for this as well ? At least for debugging the C++
code.
> It is also based on gdb, as far as I know.
We have used the PTP debugger successfully in the past with X10, but
it took a fair bit of setup. The problem is finding
You could try to use the X10DT C++ backend
launch for your application in Debug mode, or even create a regular PTP
MPI launch config for your remote X10 application, and then use the
Debug mode with that, but you may have to tweak stuff by hand.
Igor
--
Igor Peshansky (note the spelling change!)
IBM T.J. Watson Research Center
X10: Parallel Productivity and Performance (http://x10-lang.org/)
XJ: No More Pain for XML's Gain (http://www.research.ibm.com/xj/)
"I hear and I forget. I see and I remember. I do and I understand" --
Confucius
------------------------------------------------------------------------------
This SF.net Dev2Dev email is sponsored by:
Show off your parallel programming skills.
Enter the Intel(R) Threading Challenge 2010.
http://p.sf.net/sfu/intel-thread-sfd
_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users
