[Starpu-devel] Bug when dealing with a huge number of tiles during distributed executions

[Marc Sergent <marc.sergent@inria.fr>]

This is a multi-part message in MIME format.
Hello,

I've done some runs using the distributed Cholesky factorization of 
Magmamorse (new_magmamorse r1729) linked to the trunk of StarPU (r12795) 
on a homogeneous cluster (4 Fourmi machines of PlaFRIM).
I got the following issue when doing some runs with a tile size of 192 
for a matrix size of 86400.

I joined you the config.log file of the StarPU build and a gdb trace 
that I got when I attached it on running MPI processes.

Let me know if you need everything else.

Best regards,
Marc

--
Marc Sergent
Ph.D Student at Inria Bordeaux Sud-Ouest
Runtime Team
Phone: (+33|0) 5 24 57 40 71

This file contains any messages produced by compilers while
running configure, to aid debugging if configure makes a mistake.

It was created by StarPU configure 1.2.0, which was
generated by GNU Autoconf 2.63.  Invocation command line was

  $ ../configure --prefix=/home/sergent/softs/Morse_StarPU/MORSE_build_dirs/starpu_builds/cpu-only-openmpi --with-fxt --disable-build-doc --disable-build-examples --disable-gcc-extensions --disable-cuda --disable-socl --disable-opencl --with-mkl-ldflags=-lgomp --enable-calibration-heuristic=100 CC=gcc CXX=g++ F77=gfortran --no-create --no-recursion

## --------- ##
## Platform. ##
## --------- ##

hostname = fourmi017
uname -m = x86_64
uname -r = 2.6.27.39-0.3-perfctr
uname -s = Linux
uname -v = #1 SMP 2009-11-23 12:57:38 +0100

/usr/bin/uname -p = unknown
/bin/uname -X     = unknown

/bin/arch              = x86_64
/usr/bin/arch -k       = unknown
/usr/convex/getsysinfo = unknown
/usr/bin/hostinfo      = unknown
/bin/machine           = unknown
/usr/bin/oslevel       = unknown
/bin/universe          = unknown