MPI fail when using 1 sized grid in one dimension

[jdonners]

The code should operate as a pure 2D code when using grid size n1m = 1 (n1 = 2). This works in serial, but with MPI it produces wrong results and crashes on dt too small or residue too large.

[jdonners]

Example bou.in

 N1  N2  N3   NSST(>1=RK,else AB) NWRIT NREAD
 2 513 257 3                0  0
 NTST   WALLTIMEMAX    TPIN      TMAX     IRESET
 10000     83000          1e-2      30.0      0
 ALX3D      ISTR3   STR3
 1.0000       6     12.0
 YLEN ZLEN
 2.0  2.0
   RA   PRA      DT      RESID    CFLMAX
 1e+6  1.0   1e-8   1.e-2    1.2
 STATON  BALANCEON  TSTA  STAREAD
 1        1            0    0
 inslws   inslwn
  1          1
 IDTV(0=dt fixed,1=var.)   DTMIN(for dt var.)  DTMAX(for dt var.)   CFLLIM(for fixed dt)  VLIM
   1                       1e-8                0.01                 1.0                   1e3
 TFRAME  SLABDUMP(STST3)
 0.02        0

Makefile:

FC = h5pfc -r8 -O0 -fpp -g -traceback -fpe0 -debug all -check all
LINKS = -lfftw3 -lmkl_intel_lp64 -lmkl_sequential -lmkl_core

and then run with:

srun -N 1 -c 1 ./boutnp

This reproduces the bug.

[jdonners]

the problem is that the global DT value is only calculated on rank 0, since MpiMaxRealScalar? use MPI_REDUCE. The other ranks need to know the correct DT value as well. One solution is to replace MPI_REDUCE with MPI_Allreduce and it works again.