Can not run SWAN with multi processors in v3.8

[chunlei666]

Hi, john
sorry to bother you again. I tried to run Sandy case in v3.8, and an error that "MCT::m_SparseMatrixPlus::initFromRoot_:: FATAL--length of vector y different from row count of sMat.Length of y = 45240 Number of rows in sMat = 5655" is occured. 45240/5655=8, is just equal to the number of processors i assigned on the SWAN. the case can only run well with 1 swan processor. i was confused and believe that your advice is helpful. i attached my slurm*.out and the compile log. (in the compile log, it seems compiled successfully, but some errors exist for the swan as i attached in the picture)

compile_log.txt
slurm-8645149.out.txt

[jcwarner-usgs]

that does seem odd.
make sure you have the most recent version of the code.
if you do
ls -ltr
did the coawstM get compiled last ?

[chunlei666]

Hi John,
thanks for your response. the coawstM got complied. I think the problem comes from the "get_sparse_matrix" function is not linked in the "waves_coupler.f" of SWAN. but i truly find the get_sparse_matrix.f and successfully compiled to *.o file in the Build_roms, i do not know why it can not be linked to SWAN. do you have any ideas about this?

[jcwarner-usgs]

i think those build issues are just when it tries to make swan.exe, and we dont use swan.exe, so i dont think that is a problem.
I recently changed the build a little, and pushed out an update. those issues will go away but i dont think that is your problem.
Your issue is running the case.
Do you have the correct scrip file listed in teh coupling.in?
were there any Err* files created?

[chunlei666]

Looking forward to the update! In my test, I only changed the number of processors. I assigned 4*4 processors for both WRF and ROMS, and 8 for SWAN, no Error files for SWAN it just stoped. By the way, which version of libraries do you apply for the coawst?

[jcwarner-usgs]

the update is already out there. do a git pull.
this will most likely not fix your problem.
can i see the full stdout? the output you sent was not all of it.

[chunlei666]

Thanks, i'll update the coawst. Here is the sandy.out.
sandy.out.txt

[chunlei666]

Hi, John
i complied the updated code. i find that coawst can be perfectly compiled when the WRF is not coupled (e.g. Inlest_coupled), but seems the compiling errors in SWAN still exist when WRF is involved(e.g. Sandy). The complile_log is attached.
compile_log (2).txt

[jcwarner-usgs]

ok i will work on that.
But did you get a coawstM?
if so, that is all you need. the error is in building swan.exe and we dont need that, we just need to the libswan.