Helpful info for running the Kotschenreuther-Dorland-Liu Gyrokinetic code GS2
*****************************************************************************

Main references:

"KRT": Mike Kotschenreuther, G. Rewoldt, W.M. Tang, "Comparison of initial
value and eigenvalue codes for kinetic toroidal plasma instabilities",
Computer Physics Communications 88, 128 (1995)

and various other papers and reports by Bill Dorland, Mike K., et.al. 


********************************************************************

Preferred sample template file at the moment is gs_templateNN.in (where NN
is the largest version number available).

Should produce a standard template file that uses the Miller local
equilibrium, and 4 species (deuterium, electrons, carbon, beams).  Ask
Dorland how the "temperature gradient" for the beam slowing down
distribution is treated.  Varying Einj isn't realistic, while varying Ecrit
doesn't have a straightforward mapping to a beam "temperature gradient"?
Better to just use equivalent hot Maxwellian for fast ion species?

To compile the code, just type "make".

********************************************************************
To plot fields do:

gnuplot

set data style lines
set data s l           ! acceptable abbreviation
plot 'waltz05.fields' u 1:4  ! using columns 1 and 4 for x and y axes.

column 1 is theta
column 2 is theta_0
column 3 is k_theta
column 4 is Re(Phi)
column 5 is Im(Phi)
column 6 is Re(A_par)
column 7 is Im(A_par)
column 8 is Re(delta B_par)
column 9 is Im(delta B_par)
column 10 is shifted theta (only for nonlinear runs with theta connections)


There is a blank line in the *.fields file separating different k_theta,
theta_0 modes, which gnuplot interprets as starting a new curve.

You can get a normalized plot of the eigenfunctions by choosing

write_eigenfunc = .true.  

in the gs2_diagnostics_knobs namelist.  The output file is runname.eigenfunc.
The columns are the same as in runname.fields.

************************************

The quantity phtot that is printed in the output is proportional to a sum
of |phi|^2 integrated over the eigenfunction.

************************************

On the NERSC T3E, you can just use Dorland's standard version of gs2 (and
related codes at ~bdorland/dist/gs2.

To build your own version of GS2 on the NERSC T3E, move to your gs2
directory.  Type "module list" to make sure you have the following modules,
or add them by doing

module add GNU
module add tools
module add netcdf

(GS2 needs GNU make to work right, and may need some of the others to load
the right libraries).  Then just type 

make

************************************

For optimal speed of linear runs on the T3E, do:

mpprun -n NCPUS ../gs2 runname

where NCPUS = (# of species)*(# of k's).  You can run with even more NCPUS
than this (though this will start to require more communication) by
increasing it by an integer which is a factor of negrid (the next dimension
which is parallelized).  You can use the command

ps -fMPa 

to see how many PE's are presently being used interactively (in the eighth
column of output).  The total number available during the day is 132.

************************************

on the Origin-2000 hecate.princeton.edu:

mpirun -np NCPUS ../gs2 runname

On the Origin-2000's at acl.lanl.gov:

bsub -I -n 10 -R "span[ptile=10]" mpirun ../gs2 runname

where the -R "span[ptile=10]" command insures that all 10 processors are in
the same machine (to avoid hippi communication).


************************************************************