numderiv

A made-up case for checking suspected bugs by running YREC: testsuite/Test_m1p0feh+0p0GS98_base_TAMS at 7be4133, with the changes below as diffs against that case (removed lines red, added green, the words that differ shaded darker); besides these, only the head and the output names differ. Its head says what it is:

numderiv: eqstat.f bug 1.
From YREC 7be4133, testsuite/Test_m1p0feh+0p0GS98_base_TAMS: a 1 Msun star from the ZAMS to the TAMS.
Changed (each with its source):
LNumDeriv = .TRUE.: docs/source/microphysics.md line 33 and namelist_physics.md line 80 offer it ("a large (factor of 3) performance penalty, with little impact")
and the output names. Nothing else.

numderiv.nml1 (0 setting lines changed from testsuite/Test_m1p0feh+0p0GS98_base_TAMS.nml1; download the whole file)

Unchanged.

numderiv.nml2 (1 setting line changed from testsuite/Test_m1p0feh+0p0GS98_base_TAMS.nml2; download the whole file)

@@ YREC's line 54, this file's line 59 @@
  ! Equation of state flags
  LSCV = .TRUE. ! Use SCV EoS in domain where it is valid and OPAL is not
  LOPALE06 = .TRUE. ! Use OPAL tabulated EoS. Note - a simplified boltzman/saha solver, fully ionized at high T, is used when this is not enabled, or outside the table
- LNumDeriv = .FALSE. ! Use numerical derivatives of structural variables, rather than analytic ones. Induces dramatic code slow-down if adopted.
+ LNumDeriv = .TRUE. ! bug check: docs/source/microphysics.md line 33 and namelist_physics.md line 80 offer it ("a large (factor of 3) performance penalty, with little impact")
  LDH = .TRUE. ! Use Debye-Huckel correction for the Yale EoS
  ETADH0 = -2.0D0 ! Lower DH boundary
  ETADH1 = 1.0D0 ! Upper DH boundary