[mpich-discuss] Internal memory allocation error?

[Brian Harker]

Hi Rajeev-

Yes, the sends and receives are in lock-step.  The code finishes a
pixel, then the master sends out the next available pixel to the slave
that just returned a result, but during the slave's MPI_Recv to
receive that next available pixel is where the original error comes
up.  I know, I couldn't imagine how it would fail either, but lo and
behold my imagination isn't as good as the MPI error conditions  :)

Hi Peter-

I'm just running my code on a single workstation with a dual-core
processor.  The messages where the code fails are just 2-element
integer arrays, so my messages aren't large by any standards, and I'd
been thinking too about the large number of sends/receives being the
problem.  I too have tried decreasing the message size while
increasing the number of sends/receives, but still to no avail.


-- 
Cheers,
Brian
brian.harker at gmail.com


"In science, there is only physics; all the rest is stamp-collecting."

-Ernest Rutherford