[d-burg]

Release note

• Audience: users
• Numerical impact: none (harness @ 6bf955b)
• Migration: none.

GPEC splines a g-file's FPOL and differentiates it, so when FPOL was written in single precision its rounding becomes noise in Δ′: about ±2–3 on Δ′(2/1) at 257 radial nodes on DIII-D-like files. Runs with the riccati or galerkin integrator on such a file now warn and suggest regenerating it in double precision. This is the alternative to #506 agreed there.

Changes

• Equilibrium.gfile_fpol_single_precision(filename) (in Equilibrium/Sources/GFile.jl) reads only the header and the FPOL block, and returns true when every value lies within 1e-8 (relative) of its Float32 rounding.
  • Why 1e-8. Measured on the repo's g-files, single-precision FPOL sits at ≤ 1.5e-9 from its Float32 rounding. Double-precision values printed to the format's ten digits sit at a median of 1.6e-8, and a whole column below 1e-8 does not happen by chance.
  • Test: the shipped TokaMaker example is flagged, while the double-precision g65.geqdsk and EQDSK_COCOS_02 are not.
• main_from_inputs warns once, after the equilibrium is built, for EFIT-type equilibria whose integrator computes Δ′. The warning names the file.

Regression report

regress --cases diiid_n1_riccati,solovev_n1,efit_fixedbdy_separatrix --refs 0b4b9fbfb,6bf955b94 --force on feynman (SLURM, 8 threads, manifest pinned), after merging develop at 0b4b9fbfb: every case is unchanged. runtests_equil.jl passes (332/332).

diiid_n1_riccati

Regression Report: diiid_n1_riccati
============================================================================
Ref 1: 0b4b9fbfb  @ 0b4b9fbfb (2026-10-27)
       env: julia 1.11.9, x86_64-linux-gnu, manifest f241dbfc (pinned), 8 threads/8 BLAS, jit build
Ref 2: 6bf955b94  @ 6bf955b94 (2026-10-27)
       env: julia 1.11.9, x86_64-linux-gnu, manifest f241dbfc (pinned), 8 threads/8 BLAS, jit build
----------------------------------------------------------------------------
Quantity                       0b4b9fbfb      6bf955b94      Diff     Status
----------------------------------------------------------------------------
delta prime (BVP diagonal)     [5 elem]       [5 elem]       0.0e+00  OK
delta prime (raw side-major)   [10 elem]      [10 elem]      0.0e+00  OK
edge coil response delta_coil  [10 elem]      [10 elem]      0.0e+00  OK
total energy Re(et[1])         8.031977e-01   8.031977e-01   0.0e+00  OK
plasma energy Re(ep[1])        -1.368533e+00  -1.368533e+00  0.0e+00  OK
vacuum energy Re(ev[1])        2.171731e+00   2.171731e+00   0.0e+00  OK
total energy (all)             [35 elem]      [35 elem]      0.0e+00  OK
# singular surfaces            5              5              0.0e+00  OK
singular psi locations         [5 elem]       [5 elem]       0.0e+00  OK
singular q values              [5 elem]       [5 elem]       0.0e+00  OK
ODE steps (saved)              96             96             0.0e+00  OK
ODE steps (total)              2643           2643           0.0e+00  OK
mpert                          35             35             0.0e+00  OK
npert                          1              1              0.0e+00  OK
q0                             1.204212e+00   1.204212e+00   0.0e+00  OK
q95                            4.781723e+00   4.781723e+00   0.0e+00  OK
beta_n                         1.372511e+00   1.372511e+00   0.0e+00  OK
Runtime (s)                    240.3s         234.5s                  --
============================================================================
Summary: 17 unchanged

solovev_n1

Regression Report: solovev_n1
=================================================================================
Ref 1: 0b4b9fbfb  @ 0b4b9fbfb (2026-10-27)
       env: julia 1.11.9, x86_64-linux-gnu, manifest f241dbfc (pinned), 8 threads/8 BLAS, jit build
Ref 2: 6bf955b94  @ 6bf955b94 (2026-10-27)
       env: julia 1.11.9, x86_64-linux-gnu, manifest f241dbfc (pinned), 8 threads/8 BLAS, jit build
---------------------------------------------------------------------------------
Quantity                      0b4b9fbfb        6bf955b94        Diff       Status
---------------------------------------------------------------------------------
total energy Re(et[1])        1.090571e-02     1.090571e-02     0.0e+00    OK
total energy Im(et[1])        -3.987034e-10    -3.987034e-10    0.0e+00    OK
plasma energy Re(ep[1])       -1.041640e+01    -1.041640e+01    0.0e+00    OK
vacuum energy Re(ev[1])       1.042730e+01     1.042730e+01     0.0e+00    OK
vacuum matrix min eigenvalue  2.170703e+00     2.170703e+00     0.0e+00    OK
plasma energy (all)           [32 elem]        [32 elem]        0.0e+00    OK
vacuum energy (all)           [32 elem]        [32 elem]        0.0e+00    OK
total energy (all)            [32 elem]        [32 elem]        0.0e+00    OK
ODE steps (saved)             289              289              0.0e+00    OK
ODE steps (total)             287              287              0.0e+00    OK
q0                            1.900006e+00     1.900006e+00     0.0e+00    OK
q95                           3.147422e+00     3.147422e+00     0.0e+00    OK
beta_t                        4.621027e-02     4.621027e-02     0.0e+00    OK
beta_n                        3.217775e+00     3.217775e+00     0.0e+00    OK
# singular surfaces           2                2                0.0e+00    OK
singular psi locations        [2 elem]         [2 elem]         0.0e+00    OK
singular q values             [2 elem]         [2 elem]         0.0e+00    OK
mpert                         32               32               0.0e+00    OK
npert                         1                1                0.0e+00    OK
q profile (checksum)          260c3629d39c...  260c3629d39c...  identical  OK
pressure profile (checksum)   3dc9cd459e40...  3dc9cd459e40...  identical  OK
ca_left (checksum)            a784076f8ecc...  a784076f8ecc...  identical  OK
Runtime (s)                   226.3s           223.3s                      --
=================================================================================
Summary: 22 unchanged

efit_fixedbdy_separatrix

Regression Report: efit_fixedbdy_separatrix
=======================================================================
Ref 1: 0b4b9fbfb  @ 0b4b9fbfb (2026-10-27)
       env: julia 1.11.9, x86_64-linux-gnu, manifest f241dbfc (pinned), 8 threads/8 BLAS, jit build
Ref 2: 6bf955b94  @ 6bf955b94 (2026-10-27)
       env: julia 1.11.9, x86_64-linux-gnu, manifest f241dbfc (pinned), 8 threads/8 BLAS, jit build
-----------------------------------------------------------------------
Quantity                    0b4b9fbfb     6bf955b94     Diff     Status
-----------------------------------------------------------------------
psio (total poloidal flux)  3.344390e-01  3.344390e-01  0.0e+00  OK
q0                          1.155202e+00  1.155202e+00  0.0e+00  OK
q95                         5.741022e+00  5.741022e+00  0.0e+00  OK
beta_t                      3.945193e-05  3.945193e-05  0.0e+00  OK
beta_n                      1.009510e-01  1.009510e-01  0.0e+00  OK
Runtime (s)                 59.0s         58.6s                  --
=======================================================================
Summary: 5 unchanged

Notes for reviewers

• The shipped examples trigger the warning. TkMkr_D3Dlike_Hmode.geqdsk, used by most DIII-D-like examples and harness cases, and the efit_fixedbdy example file are both single precision, so riccati and galerkin runs on them now warn. That is accurate. Regenerating them with OFT's double-precision writer would remove the noise, but it would move every pinned Δ′, so it belongs in its own PR.
• Deck path only. The check sits in main, so a script that calls the solver API directly does not get it.

---

🤖 Generated with Claude Code

[matt-pharr]

Merged develop and ported gfile_fpol_single_precision from the removed ReadEquilibrium.jl into Sources/GFile.jl. The helper, driver warning, and detection tests retain their original behavior.

[matt-pharr]

@d-burg do you want review on this now? It is marked as a draft still

[d-burg]

@matt-pharr yes please, and thanks for porting the check into Sources/GFile.jl. I've merged develop again on top of your merge (6bf955b94) and rerun the harness: diiid_n1_riccati, solovev_n1 and efit_fixedbdy_separatrix are unchanged, and runtests_equil passes 332/332. The description is updated, and the PR is now marked ready for review.

[matt-pharr]

Useful change and something we should all be more aware of in Delta' world :)