Compiling the front branch

[tomchor]

Hi, everyone, I'm trying to compile the front branch on NCAR's Cheyenne cluster. The reason for that is that I'm trying to contribute on the code organization (I've talked to @johnryantaylor about it).
I was able to compile previous on this cluster successfully, but I haven't been able to compile the front branch specifically because it fails with the following error:

diablo_io.o periodic.o channel.o courant.o les.o \
duct.o cavity.o fft.o rand.o set_ics.o save_stats.o filter.o user_rhs.o \
mpi.o -L/glade/u/home/tomasc/opt/fftw-2.1.5/lib -lrfftw -lfftw hdf5.o 
/usr/lib64/gcc/x86_64-suse-linux/4.8/../../../../x86_64-suse-linux/bin/ld: les.o: in function `compute_strain_chan_':
/glade/work/tomasc/frontcomp/for/les.f:557: undefined reference to `ghost_csij2_mpi_'
Makefile:83: recipe for target 'diablo' failed
make: *** [diablo] Error 1

Which (I think) means that ghost_csij2_mpi is a function/subroutine that was called, but was never defined. And indeed if I try to find that function is the code, I can only find a call to it, but not it's definition:

frontcomp/for$ grep -i ghost_csi *.f
les.f:      call ghost_CSij2_mpi

Does anyone know what happened here?

Thanks!

[johnryantaylor]

Hi Tomas, Thanks for letting me know.  I think that subroutine is something that we added for parallelizing the AMD model (LES_MODEL_TYPE=5 in input_chan.dat).  I’ll have a look and update the relevant file.  In the meantime you could comment out the call to it since I’m pretty sure that it doesn’t show up in the Smagorinsky model which you’re using. John

…

On Nov 20, 2020, 4:03 PM +0000, Tomas Chor ***@***.***>, wrote: Hi, everyone, I'm trying to compile the front branch on NCAR's Cheyenne cluster. The reason for that is that I'm trying to contribute on the code organization (I've talked to @johnryantaylor about it). I was able to compile previous on this cluster successfully, but I haven't been able to compile the front branch specifically because it fails with the following error: diablo_io.o periodic.o channel.o courant.o les.o \ duct.o cavity.o fft.o rand.o set_ics.o save_stats.o filter.o user_rhs.o \ mpi.o -L/glade/u/home/tomasc/opt/fftw-2.1.5/lib -lrfftw -lfftw hdf5.o /usr/lib64/gcc/x86_64-suse-linux/4.8/../../../../x86_64-suse-linux/bin/ld: les.o: in function `compute_strain_chan_': /glade/work/tomasc/frontcomp/for/les.f:557: undefined reference to `ghost_csij2_mpi_' Makefile:83: recipe for target 'diablo' failed make: *** [diablo] Error 1 Which (I think) means that ghost_csij2_mpi is a function/subroutine that was called, but was never defined. And indeed if I try to find that function is the code, I can only find a call to it, but not it's definition: frontcomp/for$ grep -i ghost_csi *.f les.f: call ghost_CSij2_mpi Does anyone know what happened here? Thanks! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[tomchor]

So, I've been using the version that Jacob gave me because I could never make the github one work and that one doesn't compile this part of the code. My question about this is because I'd like to start using the github version so that I can take advantage of the improvements and bug corrections and vice versa.

I think I'll wait until this bug is fixed and continue using Jacob's version of the code in the mean time! But I'll try to migrate as soon as this issue is resolved and then try to merge some of my improvements on the code (if you're okay with them, of course).