runloop_mod.f

65 lines · 1 definitions · on GitHub (7be4133)

Python version: runloop_mod.py, 19 lines, on the right (switch to Python above the code).

Variables on this page (4)

iresca, lendag, lsetdt, rescal

Purpose

RUNLOOP_MOD holds the state of the run loop that is not in the namelist: for each of up to 50 runs (the kind cards), its kind, its rescaling targets and its stop switches. Shared by 10 routines. It was the first block made a module (the pilot, MHP via Claude, 10/14/26), and replaces COMMON/CKIND/ and COMMON/SETT/ of rp_1, first merged into COMMON/RUNLOOP/ (10/13/26); their namelist members (NMODLS, LFIRST, NUMRUN, ENDAGE, SETDT, END_DCEN, END_XCEN, END_YCEN) moved to NMLPAR_MOD.

What its variables hold, who sets them, who reads them. Each is dimensioned on 50, the maximum number of runs, and indexed by the run NK.

IRESCA(50): the run's kind, copied by PARMIN from the namelist's KINDRN (1 evolve, 2 rescale, 3 rescale and evolve); reset to 1 by SETCAL and SETSCAL for the repeated runs of a calibration. Read by the main program, STARIN (which calls RSCALE when it is not 1) and WRTHEAD.
RESCAL(4,50): the rescaling targets, in order the mass, $X$, $Z$ and the core mass, copied by PARMIN from RSCLM, RSCLX, RSCLZ, RSCLCM; changed by the calibrations (SETCAL, SETSCAL, CHKCAL, CHKSCAL, which adjust $X$ and $Z$ between runs). Read by RSCALE (which applies them), PARMIN and the main program.
LENDAG(50): evolve until a stop condition (an age ENDAGE, or a central D, X or Y), set by PARMIN and by the calibrations; read by the main program, PARMIN and HTIMER (which keeps the step from passing the age).
LSETDT(50): a fixed time step (SETDT > 0), set by PARMIN, SETCAL, SETSCAL; read by HTIMER and PARMIN.

Design

A blanket USE RUNLOOP_MOD in each routine. The module's header explains the convention of the whole conversion (a blanket USE is safe because every block was declared identically everywhere, and no member's name is in two blocks or names a procedure) and points to KNOWN_ISSUES.md issue 22 for the risk of later edits; since every file now has IMPLICIT NONE, a USE that would capture a local is a compile error, so the script shadow_census.py that checked for it was removed. IRESCA is now INTEGER*4 everywhere; at rp_1 only PARMIN declared it so, and the other files left it to the default integer kind.

Definitions

moduleRUNLOOP_MOD

lines 55–65 · GitHub

Used by (10 units)

(file level), CHKCAL, CHKSCAL, HTIMER, PARMIN, RSCALE, SETCAL, SETSCAL, STARIN, WRTHEAD

Change log

The changes since rp_1

Changed since rp_1. New file: COMMON/CKIND/ and COMMON/SETT/ merged as a module, with their namelist members moved to NMLPAR_MOD.

Changed since a8a5950. The header's advice to run shadow_census.py before adding a USE is replaced by a note that IMPLICIT NONE everywhere makes such a capture a compile error; comments only.

1C 2C 3C$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$ 4C RUNLOOP_MOD 5C$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$ 6C 10/14/26 MHP via Claude 7C code changed to convert COMMON/RUNLOOP/ into this module. First module 8C in the code, and the pattern the other blocks follow. 9C 10C RUNLOOP was itself the 10/13/26 merge of COMMON/CKIND/ and COMMON/SETT/: 11C what was left of the per-run-loop control state once their namelist 12C members moved to the consolidated block. All four members are 13C dimensioned on 50, the maximum number of runs -- not a shell or table 14C extent -- and they are written by parmin.f, setcal.f, setscal.f and 15C chkscal.f. These are defined per large loop, not per model. 16C 17C The types restate what the old implicit typing gave each name, so the 18C storage is unchanged: 19C RESCAL R is in O-Z -> IMPLICIT REAL*8(A-H,O-Z) 20C IRESCA I is in no rule -> default implicit INTEGER 21C LENDAG L -> IMPLICIT LOGICAL*4(L) 22C LSETDT L -> IMPLICIT LOGICAL*4(L) 23C 24C IRESCA is worth a note. Only parmin.f carried IMPLICIT INTEGER*4(I,...); 25C the other nine files left I to the default implicit rule, which is 26C INTEGER*4 as built here but would become INTEGER*8 under 27C -fdefault-integer-8, silently disagreeing with parmin.f. Declaring 28C INTEGER*4 here removes that latent inconsistency rather than preserving 29C it. 30C 31C No member has a DATA statement, which is what makes this block a safe 32C pilot: DATA on a USE-associated variable is illegal ("Cannot change 33C attributes of USE-associated symbol"), so blocks that initialize their 34C members need their initialization moved into the module itself. 35C 36C SAVE is stated explicitly rather than relying on module variables being 37C implicitly saved, since that became implicit only in Fortran 2008 and 38C ifort Classic predates it. This matches COMMON's semantics exactly. 39C 40C Importers use a blanket USE, which is the convention for the whole 41C conversion. It is safe because of three properties that hold across the 42C code as of 10/14/26: every block is declared identically in every file 43C that holds it, no member name appears in two blocks, and no member name 44C is also a procedure name. A blanket USE therefore imports exactly the 45C names the COMMON declaration already made global in that scoping unit, 46C so it cannot change meaning. 47C 48C The residual risk is the next edit, not this one: adding a USE to a file 49C that already uses one of these names as an implicitly typed local would 50C capture it silently. See KNOWN_ISSUES.md issue 22. 51C 11/02/26 MHP via Claude 52C Every file now has IMPLICIT NONE, so such a capture is a compile error 53C (a USE'd name clashing with a local declaration). shadow_census.py, 54C which checked for it, was removed; it is in git history before this date. 55 MODULE RUNLOOP_MOD 56 57 IMPLICIT NONE 58 SAVE 59 60 REAL*8 :: RESCAL(4,50) 61 INTEGER*4 :: IRESCA(50) 62 LOGICAL*4 :: LENDAG(50) 63 LOGICAL*4 :: LSETDT(50) 64 65 END MODULE RUNLOOP_MOD

1"""Python version of runloop_mod.f (YREC 7be4133): per-run state of the run 2loop that is not in the namelist, formerly COMMON/CKIND/ and COMMON/SETT/. 3 4Translates: runloop_mod.f#RUNLOOP_MOD (7be4133) 5 6Module-level names, as in the Fortran module, in lower case. Routines that 7USE RUNLOOP_MOD import this module and write ``runloop_mod.<name>``. 8Each array is indexed by the run (1 to 50; index 0 unused). 9""" 10 11import numpy as np 12 13# The module's variables (Fortran lines 60–63). 14# Rescaling targets: rescal[1..4, nk] = mass, X, Z, core mass (parmin; 15# changed by the calibrations setcal, setscal, chkcal, chkscal). 16rescal = np.zeros((5, 51)) 17iresca = np.zeros(51, dtype=np.int32) # run kind: 1 evolve, 2 rescale, 3 both 18lendag = np.zeros(51, dtype=bool) # evolve until a stop condition 19lsetdt = np.zeros(51, dtype=bool) # fixed time step