With write_outputs_to_HDF5=true (the default), the writer evaluates the ffit spline fits on the full ψ grid and stores ForceFreeStates/EulerLagrangeMatrices/Ideal/{A,B,C,D,E,H,F,K,G} as (npsi, np, np) complex arrays (_eval_mat_spline in GeneralizedPerturbedEquilibrium.jl). At 852 knots and N=33 that is ≈130 MB per run, and there is no finer-grained flag — a user who wants Delta_prime_matrix and the per-surface quantities in the file must also carry the operator matrices. For time-series Δ′ work (hundreds of slices) this is a disk and I/O problem; the legacy mat_flag is now deprecated/ignored, so there is no opt-out short of turning the file off and consuming GPEC.main's return value in-process (which works well, for what it's worth).
Suggestion: a write_operator_matrices::Bool = false (or an output-level selector) in ForceFreeStatesControl, with the scalar/per-surface groups written regardless. Happy to open a PR.
With
write_outputs_to_HDF5=true(the default), the writer evaluates theffitspline fits on the full ψ grid and storesForceFreeStates/EulerLagrangeMatrices/Ideal/{A,B,C,D,E,H,F,K,G}as(npsi, np, np)complex arrays (_eval_mat_splineinGeneralizedPerturbedEquilibrium.jl). At 852 knots and N=33 that is ≈130 MB per run, and there is no finer-grained flag — a user who wantsDelta_prime_matrixand the per-surface quantities in the file must also carry the operator matrices. For time-series Δ′ work (hundreds of slices) this is a disk and I/O problem; the legacymat_flagis now deprecated/ignored, so there is no opt-out short of turning the file off and consumingGPEC.main's return value in-process (which works well, for what it's worth).Suggestion: a
write_operator_matrices::Bool = false(or an output-level selector) inForceFreeStatesControl, with the scalar/per-surface groups written regardless. Happy to open a PR.