convert catchfleet IF statements to lookup table

[Rick-Methot-NOAA]

code like:
for (g=1;g<=gmorph;g++)
if(use_morph(g)>0)

better as:
for (g1=1;g1<=gmorph_use;g1++)
{g=g_use(g1);
.......

similar for fishing fleets

[Rick-Methot-NOAA]

for fishing fleets, I have started making changes using the following syntax:
A. in readdata, create a vector containing fleet number for each fishing fleet:
if(fleet_type(f)<=2) {
N_catchfleets++;
fish_fleet(N_catchfleets)=f; // to find the original fleet index
...........

Then anywhere we want to process only the fishing fleets:
for(int ff=1;ff<=N_catchfleets;ff++) {
f=fish_fleet(ff);
<any code that uses "f" >

This should be more efficient than:
for(f=1;f<=Nfleet;f++)
if(fleet_type(f)<=2){
...........

unfortunately, this cannot be done by a global search and replace.

It may be useful to have an additional index:
imatrix fish_fleet_area(1,pop,1,N_fleet)
because many of the F calculations are only done for the fleets in an area, so more efficient to only loop over those fleets.

[Rick-Methot-NOAA]

There are 105 instances of
if(use_morph(g)==1)
in the code.

My review of the logic beginning on line 356 of readcontrol.ss is that use_morph(g) is ALWAYS true.
use_morph(g) becomes true if:
if(recr_dist_pattern(gp,settle,p)==1)

which means that if a gp (morph) settles in any area (p), then it is "used".
If it is not used, then a warning should be issued.

I suspect use_morph(g) is a relic from 3.24 birthseason logic and that all 105 of those IF statements can be removed from SS, hopefully for some execution time speed-up.

Before I do so, can some of you review my logic.
@k-doering-NOAA
@kellijohnson-NOAA
@iantaylor-NOAA
@chantelwetzel-noaa

[Rick-Methot-NOAA]

for the fleet IF statements the replacement protocol is:
if dealing with agnostic areas:
for(int ff=1;ff<=N_catchfleets(0);ff++)
{
f=fish_fleet_area(0,ff);

or for a specific area, p:
for(int ff=1;ff<=N_catchfleets(p);ff++)
{
f=fish_fleet_area(p,ff);

[iantaylor-NOAA]

I took a quick glance through https://github.com/nmfs-stock-synthesis/stock-synthesis/blob/main/SS_readcontrol_330.tpl#L356-L391. I can't claim to understand all the different possibilities of birth season, settlement event, morph, growth-pattern, etc. but your explanation makes sense so I support you removing the if(use_morph(g)==1) statements and adding a warning somewhere conditioned on if(use_morph(g)!=1) or similar.

[Rick-Methot-NOAA]

create useobs_l(1,nfleet,1,nobs) with values 0 or 1 depending on header_l(f,i,3)
then
if(header_l(f,i,3)>0) length_like_tot(f)+=length_like(f,i)
can be replaced by vector multiply
length_like_tot(f)+=length_like(f)*use_obs(f);

[Rick-Methot-NOAA]

I have created a scenario in which there are some "g" indexes that are not used.
This input to a 2 area, 2 seas, 2 GP model:
3 # recr_dist_method for parameters: 2=main effects for GP, Area, Settle timing; 3=each Settle entity; 4=none (only when N_GPNsettlepop==1)
1 # not yet implemented; Future usage: Spawner-Recruitment: 1=global; 2=by area
3 # number of recruitment settlement assignments
0 # unused option
#GPattern month area age (for each settlement assignment)
1 1 1 0
2 1 2 0
2 2 2 0

produces this g indexing scheme (as shown in echoinput):
MORPH_INDEXING
g Sex GP Settlement Birth_Seas Platoon Platoon% SexGP SexGP*settle_time Used(0/1) SettleTime_frac_yr
1 1 1 1 1 1 100 1 1 1 0
2 1 1 2 1 1 100 1 2 0 0.0833333
3 1 2 1 1 1 100 2 3 1 0
4 1 2 2 1 1 100 2 4 1 0.0833333
5 2 1 1 1 1 100 3 5 1 0
6 2 1 2 1 1 100 3 6 0 0.0833333
7 2 2 1 1 1 100 4 7 1 0
8 2 2 2 1 1 100 4 8 1 0.0833333

Take-away:

1. area does not matter, the indexing is just for biological entities which can be distributed across and move between areas.
2. the "not used" occurs because morph 2 has 2 settlement times and morph 1 has only 1.
3. this is a 3.24 relic in which recruitment distribution was by main effects. I should be able to make this go away if I can internally switch all the old inputs of main effects to the new recruitment distribution method 3. Right now, even if user selects method 3 for each settlement event, SS3 still creates the blank "g".
4. Emphasizing that area does not matter: if the above example was changed to put both of morph 2's settlements in month 1 but separate areas, then SS3 only creates two biological settlement timings (for morphs 1 and 2) with morph 2's settlement split between areas.
5. All of this was one of the most vexing parts of the 3.30 creation as I was late in fully grasping the difference between settlement timings and settlement areas. Timings require a "g", but areas just split a "g" among areas.
6. I now see that this would have been way easier if I had created "3=each settlement assignment" when I created 3.30 and forced conversion at that time. I may look further into this, but suspect that the current indexing is so entwined into fabric of the code that cost/benefit may not be favorable. But I cannot pull plug on use_morph(g) without doing that conversion.

[Rick-Methot-NOAA]

I have completed transition of the fleet IF to Lookup on popdyn and benchmark.

I will merge into Main, then leave this issue open for now, but suspect that I will not get back to doing same for the gmorph.