problem with inlet_test case

Discussion on computers, ROMS installation and compiling

Moderators: arango, robertson

Post Reply
Message
Author
abhisek.sac
Posts: 6
Joined: Fri Dec 31, 2010 4:25 pm
Location: ISRO

problem with inlet_test case

#1 Unread post by abhisek.sac »

In my PC (Ubuntu 11, ia32, Core 2 Duo) I have successfully compiled and run the 'INLET_TEST' case from latest ROMS (svn 596)using the following configurations:

(i) Foran Compiler: ifort (Non-commercial version 12.10)
(ii) netcdf version 3.6.3
(iii) mpich2 1.2.1
(iv)MCT 2.6

However, when I tried to compile the same case in my Lab. in SUSE Linux High End Server (Itanium, ia64 with 32 Dual Core CPUs)with the following config.
(i) Foran Compier: ifort (Commercial version 10.1)
(ii) netcdf version 3.6.3
(iii) mpich2 1.2.1/1.4.1
(iv)MCT 2.6.
a Warning Message "Optimization problem with read_pfypar.f90. Contact Intel Premire Support......" appears. The compilation took a longer time and produces the executable "oceanM". But when I ran the "oceanM" using the following cammand

mpiexec -n 2 ./oceanM coupling_inlet_test.in

the model stops showing configuration error.

To check the compatibility of my Lab. compiler (ifort 10.1) , I downloaded ROMS-3.4 (svn 599) and found the 'INLET_TEST' case compiles and runs successfully (34560 model time steps)giving the various output files.

Now, I have the following queries:

(1) Whether the optimization used in latest ROMS (particularly INLET_TEST) case is compatible to ifort 10.1? If not, can anyone help me to make it compatible to ifort10.1? (Because finally I have to run the model in my Lab. only with ifort 10.1)
(2) Is there (Physics + Math. wise) large difference in the Coupled ROMS+SWAN for for verions 3.4 and 3.6?

Thanking in advance.

User avatar
kate
Posts: 4091
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

Re: problem with inlet_test case

#2 Unread post by kate »

You can look through the ROMS changes here. I suspect ticket 519 is causing your trouble, but you can check that for yourself.

You are motivating me to get my code up to date and run it through some tests.

abhisek.sac
Posts: 6
Joined: Fri Dec 31, 2010 4:25 pm
Location: ISRO

Re: problem with inlet_test case

#3 Unread post by abhisek.sac »

Thnx Kate. As per your suggestion, I checked ticket 519 which led me to this forum where I found a statement

"read_phypar.F: reads and reports physical model input parameters (ocean.in). This routine is still too long and we may get optimization warnings from the ifort compiler. These warnings can be ignored because there is nothing to optimize and is called only during initialization."

made by Arango on the release of ROMS 3.6. Arango suggested to ignore the optimization warnings associated with "read_phypar.F" from ifort. I ignored too. But the resultant executable is not running with ifort 10.1.

Any help in this regard will be highly appreciated.

User avatar
kate
Posts: 4091
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

Re: problem with inlet_test case

#4 Unread post by kate »

You should be able to turn off optimization for that one file. Go into Compilers/Linux-ifort.mk (if it's Linux). At the bottom, there is a section with:

Code: Select all

$(SCRATCH_DIR)/mod_ncparam.o: FFLAGS += -free
In that section add:

Code: Select all

$(SCRATCH_DIR)/read_phypar.o: FFLAGS += -O0
and see if that helps. You are giving it -O3 -O0 which could confuse it if you're unlucky, but will be no optimization if you're lucky. If you're unlucky, try it with ":=" instead of "+=" (with all the other necessary FFLAGS).

abhisek.sac
Posts: 6
Joined: Fri Dec 31, 2010 4:25 pm
Location: ISRO

Re: problem with inlet_test case

#5 Unread post by abhisek.sac »

Dear Kate,

Many thanks for your suggestions which worked for me and I have successfully ran the Inlet_Test case with ROMS-3.6 (svn 600).

Thanks once again.

Post Reply