stellinstall/vmec__params_8f_source.html

      MODULE vmec_params

      USE stel_kinds, ONLY: rprec, dp

      USE vparams, ONLY: mpold

C-----------------------------------------------

C   L o c a l   P a r a m e t e r s

C-----------------------------------------------

      INTEGER, PARAMETER :: meven = 0, modd = 1

      INTEGER, PARAMETER :: ndamp = 10

      INTEGER, PARAMETER :: ns4 = 25


      INTEGER, PRIVATE :: ink

      INTEGER, PARAMETER, DIMENSION(0:mpold) ::

     1  jmin1 = (/ 1,1,(2,ink=2,mpold) /),        !starting js(m) values where R,Z are non-zero

     2  jmin2 = (/ 1,2,(2,ink=2,mpold) /),        !starting js(m) values for which R,Z are evolved

     3  jlam  = (/ 2,2,(2,ink=2,mpold) /)         !starting js(m) values for which Lambda is evolved


!  Besure to update werror in fileout.f when adding more error flags.

      INTEGER, PARAMETER :: norm_term_flag=0, bad_jacobian_flag=1,

     1                      more_iter_flag=2,

     2                      jac75_flag=4, input_error_flag=5,

     3                      phiedge_error_flag=7,

     4                      ns_error_flag=8,

     5                      misc_error_flag=9,

     6                      successful_term_flag=11, !ftol force criterion has been met

     7                      bsub_bad_js1_flag=12,

     8                      r01_bad_value_flag=13,

     9                      arz_bad_value_flag=14

      INTEGER, PARAMETER :: restart_flag=1, readin_flag=2,

     1                      timestep_flag=4,output_flag=8,

     2                      cleanup_flag=16, reset_jacdt_flag=32


      REAL(rprec), PARAMETER :: pdamp = 0.05_dp

      CHARACTER(LEN=*), PARAMETER :: version_ = '9.0'

!-----------------------------------------------

!   L o c a l   V a r i a b l e s

!-----------------------------------------------

      INTEGER :: ntmax, rcc, rss, rsc, rcs, zsc, zcs, zcc, zss

      INTEGER :: mnyq, nnyq

      INTEGER, ALLOCATABLE :: uminus(:)

      REAL(rprec), ALLOCATABLE :: mscale(:), nscale(:)

      REAL(rprec) :: signgs, lamscale

!-----------------------------------------------

!

!     VERSION INFORMATION HISTORY

!

!     8.00

!          a) added lforbal logical to fbal module to control whether to compute the flux-averaged

!             force balance equation printed in the threed1 file. This requires a modification of

!             the m=1,n=0 forces for R,Z in tomnsps subroutine. This works well, generally, and

!             yields an improved <EQUIF> in threed1 file. However, there have been some cases where

!             this non-variational departure fails to converge.

!          b) added "bias" to iotas in bcovar when determining preconditioner for very small iota

!             values. Seems to need this for improving convergence of preconditioned current-hole cases.

!             Eliminated in v8.20.

!          c) added "totzsps,a_hess" to recompute r,z,l rapidly for only those modes that are jogged during

!             Hessian calculation. NOTE: need to do this for lasym=true case, as well, eventually

!     8.20  (August, 2004)

!          a) removed 2-pt tie at origin for m=1 modes, to preserve tri-diagonal structure of Hessian.

!             This is needed for preconditioning, which assumes block-tridi structure of equations

!          b) fixed problem with free-boundary preconditioner, namely, ctor can not be extrapolated

!             at edge when computing preconditioner, because this breaks tri-diagonal structure

!          c) added new variables to input file to control preconditioning:

!             1) PRECON_TYPE: = 'default', default tri-di (block size = 1)

!                             = 'cg',      block tri-di, conjugate-gradient time-stepper

!                             = 'gmres',   "          ", gmres time-stepper

!                             = 'tfqmr',   "          ", transpose free qmr

!             2) PREC2D_THRESHOLD: value of (unpreconditioned) forces at which block (2D) preconditioner

!                                  is turned on (=0 block preconditioner never turned on); recommended

!                                  (default) value ~ 1.E-10, or smaller, if convergence is poor

!             3) LFORBAL: logical variable (default = .true.); when true, the force balance

!                         used in the threed1 file is used to evolve the m=1 R,Z components. This

!                         is a non-variational departure from the force equations for these modes,

!                         but generally does not have an unfavorable impact on convergence.

!          d) added new internal variable, ICTRL_PREC2D, to precon2d module. Replaces previous lprec2d

!             and iequi>1 variables.

!          e) removed lsweep_fast option from module precon2d. This slows the computation of the Hessian

!             by about 2/3, but is more accurate (includes pdamp, liota, lforbal correctly)

!          f) removed lflam logicals from bcovar and tomnsps, since now we must compute dFlam/dR,Z by

!             jogging

!          g) removed Compute_Hess_Flam_RZ from lamblks; this is now computed via jogging

!             (also removed Get_dFlam_dRZ, FFT2Hessian, Forbal_avg, GetGsqrtVar supporting routines)

!          h) removed internal liota logic, used to push iota profile rather than solving for it. Had

!             needed this for symmetric Hessian (lsweep_fast=true option), but no longer required. Also,

!             it was not implemented entirely correctly for lforbal=true case

!          i) for lasym m=1 constraint rsc = zcc, changed xc array so that R+ = .5*(rsc + zcc) is stored at

!             xc(rsc,m=1) and R- = .5*(rsc - zcc) is stored at xc(zcc,m=1). In residue, gcz(R-) == gcz(zcc)

!             is zeroed by "evolving" gcr(zcc) = azd*[xc(zcc)-xcint], and gcr(rsc) => .5*[gcr(rsc) + gcz(zcc)]

!             is evolved. In totzspa, the real rsc and zcc are computed from the internal representations

!             (check convert call, too) by calling a new routine convert_asym (also called from wrout before

!             writing out xc info). In readin, the original R+,R- are stored, so that for external "jogs",

!             there will be no change in forces. All these changes are needed to obtain an invertible Hessian.

!          j) added m=1 constraint for 3D case (similar to (i)), rss(n) = zcs(n), for n != 0. Imposed this

!             on forces by adding routine constrain_m1 in residue. Added convert_sym routine to totzsp to convert

!             from internal xc representation TO internal one.

!          k) Decreased exponent on pdamp factor r2 (in bcovar) from 2 to 1, to give better conditioning

!             especially for current hole cases

!          l) Eliminated iotas bias for determining preconditioner, previously added in v8.00 for stabilizing

!             current hole cases (not needed with corrected preconditioner)

!     8.30  (October, 2004)

!          a) Implemented flags for "reverse-communication" mode of vmec

!     8.40 a) Converted the m=1 constraints for 3D and asym back to old way; did not always

!             converge well with the new constraints introduced in 8.20 (i-j)

!     8.45  (December, 2005)

!          a) Added the lconm1 logical. If = True, new constraint; if = False, old m=1 constraint used

!          b) Added "perturbation" computation for lasym=TRUE case (totzspa_hess)

!     8.46  (June, 2009)

!          a) Added LRFP logical to allow easy switching of profiles between Stellarator/tokamak (PHIP=1, LRFP=F)

!             and RFP (CHIP=1, LRFP=T). When LRFP=T, AI coefficients are expansion of q = 1/iota. Added lrfp to

!             LIBSTELL/vmec_input module.

!     8.47  (July, 2010)

!          a) Rewrote magnetic field representation so that phip*lambda = new internal lambda. This greatly improves

!             the conditioning of the lambda equations which otherwise become singular at the RFP reversal point

!     8.48  (March 2012 - JDH)

!          a) Accumulated small changes from SPH & JDH

!          b) Modifications from J Geiger, March 2012

!             - to be able to get additional main iterations if the force tolerance is

!                  not met. Parameter MAX_MAIN_ITERATIONS specifies how many main iteration

!                  cycles should be run.

!             - to get a full output in the threed1-file if the force tolerance is not

!                  met. Specify the logical LFULL3D1OUT to true for this.

!             - if vmec2000 is compiled with netcdf, you can still get the ascii-output

!                  if you specify the logical LWOUTTXT as true.

!             - you get the output for diagno 1.0 and 1.5 if the logical LDIAGNO set true.

!             - you get a rather old fort.8-output if you specify LOLDOUT as true.

!

!             If none of these new variables is set, the behavior of vmec2000 is as

!             expected from the version without the changes.

!     8.49  (June, 2012)

!          a) Fixed bug in bcovar when averaging half-grid covariant components onto full grid: did not

!             zero components at (phantom) js=1 point, so edge force averages were incorrect

!          b) Added lamscale factor to scale lambda in contravariant B-components. Modify wrout to maintain

!             old-style lambda output

!          c) Set lbsubs=F in jxbforce by default to capture current sheets

!          d) Added lmove_axis INPUT logical (=T by default) so user can control whether or not the magnetic

!             axis can be initially shifted to improve the initial force residuals. It is important NOT to move

!             the helical axis for RFP equilibria requiring a helical seed (set l_moveaxis=F for this case!)

!

!     8.50   (Jan, 2013)

!          a) Improved scaling of lambda forces with respect to lamscale

!          b) Fixed fnorm1 scaling (remove hs dependence)

!          c) Added lgiveup logical (M. Drevlak/J. Geiger)

!-----------------------------------------------

      END MODULE vmec_params