ERROR running application example of CBLAST

[bibi951]

I am a very beginner of ROMS.When I run the application example of Calast I get this error
makefile:233: INCLUDING FILE Compilers/make_macros.mk WHICH CONTAINS APPLICATION-DEPENDENT MAKE DEFINITIONS
cp -f /usr/g95soft/netcdf/include/netcdf.mod Build
cp: cannot stat `/usr/g95soft/netcdf/include/netcdf.mod': No such file or directory
I dont understand that sentence in the makefile.please give me some guide.
Thanks in advance!

[kate]

cp -f /usr/g95soft/netcdf/include/netcdf.mod Build
cp: cannot stat `/usr/g95soft/netcdf/include/netcdf.mod': No such file or directory

This is coming from a file in the Compilers directory. What system are you on and what compiler are you using? That combination tells make which of those files to use. In any case, you need to tell it where to look for your netcdf library and module files, whether in that file, in the build script, or in your environment. It's currently looking for the module files in the wrong place.

Have you checked the wiki?

[bibi951]

HI,I have use :
export NETCDF_INCDIR=/usr/local/include
export NETCDF_LIBDIR=/usr/local/lib
before cd ${MY_ROMS_SRC}
in build_cblast.sh script.
It also leads to this error mentioned above.
But if I use
NETCDF_INCDIR ?=/usr/local/include
NETCDF_LIBDIR ?=/usr/local/lib
in the CYGWIN-g95.mk
It can compile.
But saying "/cygwindrive/d/windows/system32/export:permision denied."
That is OK,right?
Thanks.

[bibi951]

$ ./oceanS < cblast_stations.in ocean_cblast.in
Error:
Resolution, Grid 01: 0000x0000x000, Parallel Threads: 1, Tiling: 000x000

ROMS/TOMS: Wrong choice of domain 001 partition or number of parallel threads.
NtileI*NtileJ must be a positive multiple of the number of threads.
Change number of threads (environment variable) or
change domain partition in input script.
But in ocean_cblast.in, I use
NtileI == 1 ! I-direction partition
NtileJ == 1 ! J-direction partition
What is wrong? Thanks in advance.

[wilkin]

$ ./oceanS < cblast_stations.in ocean_cblast.in
Error:
Resolution, Grid 01: 0000x0000x000, Parallel Threads: 1, Tiling: 000x000

You are giving ROMS the cblast_stations.in file as standard input, so it can't make any sense of it. (The first thing it fails on is the tile partition - which it must know immediately so it can parse out the work of reading the ocean.in in parallel.)

You need to issue the command:

Code: Select all

$ ./oceanS < ocean_cblast.in

The stations file will be named inside ocean_cblast.in (see the STANAME input).

John.

[bibi951]

First,I would like to thank all your replies.
I now use SGI to run this model example cblast in parallel.This error occur:
cp: cannot stat `/usr/local/netcdf3.6.2/include/netcdf.mod': No such file or directory
make: *** [/lun/home/lujiuyou/roms/Projects/cblast/Forward/Build/netcdf.mod] Error 1
But I use "cp /usr/local/netcdf3.6.2/include/netcdf.mod ",It works.
In makefile
NETCDF_MODFILE := netcdf.mod
$(SCRATCH_DIR)/$(NETCDF_MODFILE): | $(SCRATCH_DIR)
cp -f $(NETCDF_INCDIR)/$(NETCDF_MODFILE) $(SCRATCH_DIR)

$(SCRATCH_DIR)/$(TYPESIZES_MODFILE): | $(SCRATCH_DIR)
cp -f $(NETCDF_INCDIR)/$(TYPESIZES_MODFILE) $(SCRATCH_DIR)

$(SCRATCH_DIR)/MakeDepend: makefile \
$(SCRATCH_DIR)/$(NETCDF_MODFILE) \
$(SCRATCH_DIR)/$(TYPESIZES_MODFILE) \
| $(SCRATCH_DIR)
$(SFMAKEDEPEND) $(MDEPFLAGS) $(sources) > $(SCRATCH_DIR)/MakeDepend
cp -p $(CURDIR)/$(MAKE_MACROS) $(SCRATCH_DIR)
In build_cblast.sh
setenv NETCDF_INCDIR /usr/local/netcdf3.6.2/include
setenv NETCDF_LIBDIR /usr/local/netcdf3.6.2/lib
How to solve this problem?
Thanks.

[wilkin]

First,I would like to thank all your replies.
I now use SGI to run this model example cblast in parallel.This error occur:
cp: cannot stat `/usr/local/netcdf3.6.2/include/netcdf.mod': No such file or directory
make: *** [/lun/home/lujiuyou/roms/Projects/cblast/Forward/Build/netcdf.mod] Error 1
But I use "cp /usr/local/netcdf3.6.2/include/netcdf.mod ",It works.

This is a very common "getting started" problem described in the tutorial I am preparing for next week's user meeting in Sydney. Does this work? :

Code: Select all

ls -l /usr/local/netcdf3.6.2/include/netcdf.mod

If not, see the wiki entry here:
https://www.myroms.org/wiki/index.php/R ... r_messages

I advise against messing with the makefile. It's better to work within the framework provided by build.bash if you can. The tricks are recognizing that you need to set USE_MY_LIBS in order to over-ride the default include and library path, and paying close attention to where you make the changes to NETCDF_INCDIR for your particular compiler choice.

The output from the compilation that is typed to your screen also tells you what the compiler is actually doing, as opposed to what you think you asked for. Look at it closely to see what netcdf include path is being used - is it what you wanted?

Before attempting using MPI, I suggest you compile and run in serial first. The library and include paths are the same regardless of serial/MPI choice, but you tend to have fewer compiler compatibility issues with serial and at run-time sometimes get more helpful error messages. The compile in serial commenting out these lines in build.bash:

Code: Select all

#export USE_MPI=on
#export USE_MPIF90=on 

John.

[bibi951]

I now use SGI to run this model(Serial).But I am not the administrator,and I canot install Cygwin as I wish.So I use SSH (installed) and netcdf(installed).
What can I only do is to modify my build_cblast.sh to be adapted to the installed system.So
In build_cblast.sh
setenv NETCDF_INCDIR /usr/local/netcdf3.6.2/include
setenv NETCDF_LIBDIR /usr/local/netcdf3.6.2/lib
I am sure this is the correct netcdf path.
I use "cp /usr/local/netcdf3.6.2/include/netcdf.mod ",It works.
But In makefile I did not modify this file.I just show the context.
NETCDF_MODFILE := netcdf.mod
$(SCRATCH_DIR)/$(NETCDF_MODFILE): | $(SCRATCH_DIR)
cp -f $(NETCDF_INCDIR)/$(NETCDF_MODFILE) $(SCRATCH_DIR)
$(SCRATCH_DIR)/$(TYPESIZES_MODFILE): | $(SCRATCH_DIR)
cp -f $(NETCDF_INCDIR)/$(TYPESIZES_MODFILE) $(SCRATCH_DIR)
This error occur:
cp: cannot stat `/usr/local/netcdf3.6.2/include/netcdf.mod': No such file or directory
Any possibility for this?
Thanks in advance.
Best regards!

[wilkin]

Cygwin is software that emulates UNIX on a Windows PC. Is your SGI actually an Intel box running Windows? If it has native UNIX, then forget the whole Cygwin thing.

You say "cp /usr/local/netcdf3.6.2/include/netcdf.mod" works but it can't. That's not a vaild UNIX command.

What is the result of:

ls -l /usr/local/netcdf3.6.2/include/netcdf.mod

[bibi951]

lu@a330:/> ls -l /usr/local/netcdf3.6.2/include/netcdf.mod
-rw-r--r-- 1 root root 263985 2007-08-03 11:21 /usr/local/netcdf3.6.2/include/netcdf.mod

"cp /usr/local/netcdf3.6.2/include/netcdf.mod" Sorry, I did not complete it with its destinition(Build directory for scratch).

My SGI OS is unix. But how to solve this?
Thanks.

[wilkin]

Look carefully in build.bash.

If you wish to over-ride the default netcdf include and lib entries you need to set...

export USE_MY_LIBS=on

...and then you need to make the appropriate changes in the section of build.bash code that follows. Where you make the changes depends on your compiler.

Once you have made the changes, and re-run build.bash, you will will see build.bash report what it is trying to do. It echoes each actual UNIX command to your terminal.

Check that what it is doing is actually what you intended. If you have not set a library or include path correctly, it should be clear from the command echoed to your screen shortly before the compile fails that build.bash is doing something different.

[bibi951]

Thank you for your replies.
I've just realized having made a simple mistake
qsub a job in a incorrect way on my SGI.Not the problem of this model.
Many Thanks and Best Regards

[bibi951]

Hi.
I now use SGI to run this example cblast in parallel.
Its compilation ends with no error.BUT when execute oceanM:
MPI: On host a3700, Program /lun/home/lujiuyou/roms/Projects/cblast/Forward/oceanM, Rank 0, Process 32711 received signal SIGSEGV(11)

MPI: --------stack traceback-------
sh: line 1: idb: command not found

MPI: -----stack traceback ends-----
MPI: On host a3700, Program /lun/home/lujiuyou/roms/Projects/cblast/Forward/oceanM, Rank 0, Process 32711: Dumping core on signal SIGSEGV(11) into directory /lun/home/lujiuyou/roms/Projects/cblast/Forward
MPI: MPI_COMM_WORLD rank 0 has terminated without calling MPI_Finalize()
MPI: aborting job
MPI: Received signal 11
What lead to this error.Your guide would be appreciated very much!
Best regards!

[wilkin]

Did you use "mpirun" ?

[bibi951]

Oh ,
I use
mpirun -np 4 ocean < ocean.in

not
mpirun -np 4 ocean ocean.in

A silly mistake