GETRF checking fails when multiple threads threads are bind to the same core.

[abouteiller]

Original report by Nuria Losada (Bitbucket: nuriallv, GitHub: nuriallv).

---

The result checking fails on GETRF when multiple threads threads are bind to the same core.
Setting info[2] = 1 on the body of getrf at zgetrf.jdf seems to fix it.

Small reproducer (running on saturn, nodes c0*) using the dplasma/tests

mpirun -np 1 ./testing_dgetrf -c 4 -N 20 -t 2 -x -- --parsec_bind file:binding_onecore

$ cat binding_onecore 
0

‌

[abouteiller]

replicated, PMIX_MCA_psec='' SLURM_TIMELIMIT=100 PARSEC_MCA_device_cuda_enabled=1 PARSEC_MCA_device_cuda_memory_use=10 OMPI_MCA_rmaps_base_oversubscribe=true salloc -wleconte -n 8 --gpus-per-task=1 --cpus-per-task=1 srun -n 1 --pty gdb --args tests/testing_dgetrf_1d -c 4 -N 20 -t 2 -x -- --mca bind_map file:binding_onecore --mca debug_verbose 0

cat binding_onecore
0,0,0,0