Segmentation fault (Boozer spectrum)

[TacitusQ]

Hi Caoxiang,

Thank you for maintaining this code.

Today morning I had an issue
$ mpirun -np 8 xfocus ellipse

gave the following output

focus : Optimization took 4 M 20 S.
-----------POST-PROCESSING-----------------------------------
-----------COIL DIAGNOSTICS----------------------------------
diagnos : Bnormal ; Bmn harmonic ; tor. flux ; coil length ; c-s sep. ;
: 1.46056E-03 ; 0.00000E+00 ; 0.00000E+00 ; 7.24710E-01 ; 8.73693E+01 ;
: Maximum curvature of all the coils is : 1.048072925732205E+00 ; at coil 2
: Average length of the coils is : 7.713683172599356E+00
: The minimum coil-coil distance is : 7.238058243779492E-01
: The minimum coil-plasma distance is : 7.033711649841820E-01 ; at coil 1
: Average relative absolute Bn error is : 7.682967043717293E-03
findaxis: Finding axis at phi = 0.00000E+00 with (R,Z) = ( 3.00988E+00,-9.43841E-07 ).
findaxis: info=1, relative error between two consecutive iterates is at most xtol.
poinplot: following fieldlines between ( 3.00988E+00,-9.43841E-07 ) and ( 3.24000E+00, 0.00000E+00 )
: order= 1 ; masterid= 0 ; (R,Z)=( 3.03289E+00,-8.49456E-07 ) ; iota=-7.57725E-02 ; niter= 1000 .
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
xfocus 00000000004EE8D3 Unknown Unknown Unknown
libpthread-2.12.s 00007EFEB92517E0 Unknown Unknown Unknown
xfocus 00000000004A4BED boozsurf_ 213 boozer_m.F90
xfocus 00000000004A1817 poinplot_ 228 poinplot_m.F90
xfocus 00000000004ABEC7 MAIN__ 142 focus_m.F90
xfocus 000000000040BF22 Unknown Unknown Unknown
libc-2.12.so 00007EFEB7FF8D20 __libc_start_main Unknown Unknown
xfocus 000000000040BE29 Unknown Unknown Unknown

---

Primary job terminated normally, but 1 process returned
a non-zero exit code. Per user-direction, the job has been aborted.

---

---

mpirun detected that one or more processes exited with non-zero status, thus causing
the job to be terminated. The first process to do so was:
Process name: [[51155,1],0]
Exit code: 174

---

It looks like the problem is with boozer_m.f90 in post processing.

You suggested I switch the flag
case_postproc = 3
in input file, because Boozer spectrum is under development.

I will let you know if this works.

Tony

[zhucaoxiang]

@TacitusQ Thanks for reporting this. For some reason, I cannot repeat the error on PPPL cluster. But I do have two comments.

1. I recalled that I saw a similar error once. My solution at that time is to increase memory or reduce the resolution. When FOCUS call the Boozer spectrum subroutine, it will store a 3D array (N_filed_lines, XYZB, N_points_per_line). For the example case, it is a (10, 4, 360000) array. It is not trivial. So can you run it with higher memory? You can add #SBATCH --mem-per-cpu=2GB.

2. I did notice the boozer.f90 file had some bug/flaw that will write a lot unrelated information. I have corrected it. You can test it by checkouting the develop branch (git fetch; git checkout develop; make clean; make) or just use module load focus/develop at PPPL.

Let me know if the bug still exists.

[zhucaoxiang]

@TacitusQ Do you have any updates on this issue?

[TacitusQ]

@zhucaoxiang

Thank you for your follow up. First I wish to confirm that the work around you suggest works. Just change
case_postproc = 3

Second I would like to report that setting = 4 still produces a segmentation fault. I have tried increasing the memory per cpu, then using the focus/develop module. The bug still exists.

For your reference here is a copy of the error readout:
seg-fault-dec-6.txt

[zhucaoxiang]

@TacitusQ Thanks for reporting. I will keep this issue until the bug is fixed. Before that, I would suggest not using Boozer spectrum calculations.