CASINO manual - Theory of Condensed Matter
CASINO manual - Theory of Condensed Matter
CASINO manual - Theory of Condensed Matter
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Set the verbosity level <strong>of</strong> the machine set-up process to . By<br />
default is 0. ’-v’ increases the verbosity level by 1, and<br />
’-q’ decreases it by 1.<br />
--info | -i<br />
Report back various machine-dependent parameters whose value one might like<br />
to know when deciding how to run a job, then stop. This is useful<br />
for inquiring what runqmc believes to be, e.g., the number <strong>of</strong> cores per node,<br />
according to the currently activated arch file.<br />
Options available on workstations<br />
--background | -B<br />
Run <strong>CASINO</strong> in the background, returning control to the shell after starting<br />
the run. This has the same effect as ’${0##*/} & disown’, whereby the<br />
<strong>CASINO</strong> process is detached from the shell, so if one wants to stop the run<br />
’kill’ or ’killall’ must be used. It is safe to log out after running with<br />
this option, the calculation will continue---no need for nohup/disown.<br />
Running multiple jobs causes them to run in the background whether this<br />
option is specified or not.<br />
--print-out | -P<br />
Print out the output <strong>of</strong> <strong>CASINO</strong> as it is being run. Implies --background.<br />
[CTRL]-[C] will stop the print-out, and the <strong>CASINO</strong> job will remain in the<br />
background. This option is ignored when running multiple jobs.<br />
--debugger= | --gdb|-g | --valgrind<br />
Run the code through the a debugger. This automatically sets to<br />
’debug’ if no version had been selected. --gdb and -g are equivalent to<br />
--debugger="gdb -q". --valgrind is equivalent to --debugger="valgrind"<br />
Note that specific debuggers tend to work better with specific compilers,<br />
for example the gdb debugger tends works better with GCC’s gfortran than<br />
with Intel’s ifort.<br />
Options available on parallel workstations and clusters<br />
--no-mpi | -1<br />
Run the binary directly without invoking mpirun etc. This option is applied<br />
before any others, and makes this script behave as if the machine was a<br />
single-core machine.<br />
--nproc= | -p <br />
Set the number <strong>of</strong> MPI processes to . This will have to be consistent<br />
with , , and the machine information in the relevant<br />
.arch file.<br />
--ppn=<br />
Set the number <strong>of</strong> MPI processes per physical, multi-core node to .<br />
This will have to be consistent with , , , and the<br />
machine information in the relevant .arch file.<br />
--shm[=] | --shmem[=] | -s<br />
Enable shared memory (which affects, e.g., storage <strong>of</strong> orbital<br />
coefficients in blip or Gaussian mode and some other large arrays).<br />
If is provided, set the<br />
number <strong>of</strong> processes among which to share memory to , otherwise<br />
shared memory is used across all cores on the same node by default. This<br />
option requires having compiled the code with shared memory support, e.g.,<br />
with ’make shm’ or ’make openmpshm.<br />
--diagram | -D<br />
Draw a diagram <strong>of</strong> the processes and threads on each node to the terminal<br />
during set-up.<br />
Options available on clusters<br />
27