ABINIT Discussion Forums

Dear abinit users and developers,

I'm trying to use the MPI+OpenMP-compiled version of Abinit-7.10.2 on a BG/Q.
The calculations start correctly but the number of threads per MPI task doesn't
seem to be retrieved/set correctly:

==== OpenMP parallelism is ON ====
- Max_threads: 1
- Num_threads: 1
- Num_procs: 1
- Dynamic: F
- Nested: F

The calculations are launched as follows:

runjob --np 256 --ranks-per-node 4 --envs "OMP_NUM_THREADS=8" : abinit < calc.files > calc.log

I also noticed in the log file that the parameter max_nthreads is set to 1:

- outvars: echo of global parameters not present in the input file
- max_nthreads = 1

Am I missing something obvious? Thanks in advance for your help.

Best regards,
Max

PS. Please find below the biuld information

=== Build Information ===
Version : 7.10.2
Build target : powerpc64_linux_ibm14.1
Build date : 20150113

=== Compiler Suite ===
C compiler : ibm12.1
C++ compiler : ibm12.1
Fortran compiler : ibm14.1
CFLAGS : -g -O3 -qarch=auto -qtune=auto -qstrict -qspill=2000
CXXFLAGS : -g -O3 -qarch=auto -qtune=auto -qstrict -qspill=2000 -qessl
FCFLAGS : -g -qzerosize -F/home/lawson/apps/abinit/misc/xlf.cfg -qsmp
FC_LDFLAGS : -Wl,--allow-multiple-definition

=== Optimizations ===
Debug level : basic
Optimization level : standard
Architecture : ibm_powerpc64

=== Multicore ===
Parallel build : yes
Parallel I/O : yes
openMP support : yes
GPU support : no

=== Connectors / Fallbacks ===
Connectors on : yes
Fallbacks on : yes
DFT flavor : libxc-fallback+wannier90-fallback
FFT flavor : fftw3
LINALG flavor : custom
MATH flavor : none
TIMER flavor : abinit
TRIO flavor : none

=== Experimental features ===
Bindings : @enable_bindings@
Exports : no
GW double-precision : no

From what you say, the problem is likely coming from your run-time environment, not from Abinit. You might have to tune more parameters of your runjob script to fully enable OpenMP. Feel free to ask your system administrator should you have any doubt.