Timestamp:
Sep 2, 2010, 3:45:23 PM (14 years ago)
Author:
lguez
Message:

Replaced Numerical Recipes procedures for spline interpolation (not in
the public domain) by procedures from the Pchip package in the Slatec
library. This only affects the program "ce0l", not the program
"gcm". Tested on Brodie SX8 with "-debug" and "-prod", "-parallel
none" and "-parallel mpi". "start.nc" and "limit.nc" are
changed. "startphy.nc" is not changed. The relative change is of order
1e-7 or less. The revision makes the program faster (tested on Brodie
with "-prod -d 144x142x39", CPU time is 38 s, instead of 54
s). Procedures from Slatec are untouched, except for
"i1mach.F". Created procedures "pchfe_95" and "pchsp_95" which are
wrappers for "pchfe" and "pchsp" from Slatec. "pchfe_95" and
"pchsp_95" have a safer and simpler interface.

Replaced "make" by "sxgmake" in "arch-SX8_BRODIE.fcm". Added files for
compilation by FCM with "g95".

In "arch-linux-32bit.fcm", replaced "pgf90" by "pgf95". There was no
difference between "dev" and "debug" so added "-O1" to "dev". Added
debugging options. Removed "-Wl,-Bstatic
-L/usr/lib/gcc-lib/i386-linux/2.95.2", which usually produces an error
at link-time.

Bash is now ubiquitous while KornShell? is not so use Bash instead of
KornShell? in FCM.

Replaced some statements "write(6,*)" by "write(lunout,*)". Replaced
"stop" by "stop 1" in the case where "abort_gcm" is called with "ierr
/= 0". Removed "stop" statements at the end of procedures
"limit_netcdf" and main program "ce0l" (why not let the program end
normally?).

Made some arrays automatic instead of allocatable in "start_inter_3d".

Zeroed "wake_pe", "fm_therm", "entr_therm" and "detr_therm" in
"dyn3dpar/etat0_netcdf.F90". The parallel and sequential results of
"ce0l" are thus identical.

File:
1 added

Note: See TracChangeset for help on using the changeset viewer.