[MITgcm-devel] some numbers

[Jean-Michel Campin]

Hi,

I recently (Aug 30th) starting to count+list all the CPP options that are
used somewhere in the MITgcm code (including in the verification*/code*/
 dir), and came up with an impressive number of ...  677
But many of these options (291 at that time) are not "documented" anywhere
in the standard source code (i.e., excluding  verification*/code*/)
meaning that a defined/undef statement was not found in any *_OPTIONS.h files
nor locally in  *.F *.h files, and was never defined/undef by genmake2 or in
any build_options files.

I think it would help to specify, in the appropriate *_OPTIONS.h,
an #undef or #define for any option that is added somewhere in the code
(at least it would be easy to know where we are supposed to turned it on
or off).

Mostly unrelated: I counted how many "unused variable" warnings I get
at compile time when running testreport (done with g77):
Counting all forward experiment tests (50 built, with or without mnc),
it's about 80, mostly coming from fizhi experiments (especially the
gridalt one), and the number is not very different for an MPI testreport.
But for an adjoint testreport, it's currently more than 86,000
for all the 15 mitgcmuv_ad that are produced, an average of more than
5,700 per built (few days ago, was only 85000, but I guess it hard to
stop the progression).
Also it's worth noticing that I did not find any unused variable warning 
coming from ad_taf_output.f compilation (TAF generated code is clean).
It's not a serious issue (I think most compiler will discard these
unused var, so it will not take any memory space), but it makes
thing more difficult to debug (since some debugging options are not
longer of any use).

Cheers,
Jean-Michel