Compset B1850 with Clang compiler

[Shruti]

Hello,

I am trying to run compset B1850, res f09_g17 with clang.
As i understand clang is not supported yet.
I was trying to work around things to get this done.
Facing the following error:
ERROR: BUILD FAIL: clm.buildlib failed,

cime/scripts/caseB4/Tools/Makefile:888: recipe for target 'clmfates_paraminterfaceMod.o' failed

make: *** [clmfates_paraminterfaceMod.o] Error 1
make: *** Waiting for unfinished jobs....

ERROR: Command make complib -j 4 MODEL=clm COMPLIB=cime/projects/scratch/caseB4/bld/aocc/openmpi/nodebug/nothreads/mct/noesmf/lib/libclm.a -f cime/scripts/caseB4/Tools/Makefile MACFILE=cime/scripts/caseB4/Macros.amd failed with rc=2


Any help would be appreciated.

 

[fischer]

Hi Shruti,

I'm not familiar with Clang, but did a little research into it. It doesn't look like it supports FORTRAN, that should be something you look into.
Also, you should start with a simpler model configuration, try compset X, res f19_g17.

Chris

 

[Shruti]

Hi Chris,

I have already checked for X compset with f19_g17. That works for me.
Ok will check into it. If you get any idea, please let me know.
Thanks for the help!!

 

[fischer]

Hi Shruti,
Since your failure is happening in clm, you should try compset I2000Clm50BgcCrop and res f19_g17. This might make
your debugging go quicker. You can also try compset A to see what happens, this is a bit more of an involved build, but
not the the level of an I or B compset.

 

[Shruti]

Hi Chris,

I have already tried A compset and it works fine for me.
The one which you suggested "compset I2000Clm50BgcCrop and res f19_g17" is giving me the following error:
ERROR: Could not find all inputdata on any server

As per my read, this can occur if res is not supported. Is that correct?
Also is there any other resolution which is supported?
I have already tried with f09_g17, f19_g17 and f19_g17_gl4 and on 2 separate systems.
All give the same error as above.

 

[fischer]

What version of CESM are you using. All of the inputdata is available for I2000Clm50BgcCrop at res f19_g17.
You can run ./check_inputdata --download in your case directory to download any missing inputdata that's needed
for this case.

Chris

 

[Shruti]

Hello,

I am checking for CESM 2.1.2 , compset B1850 with clang compiler.
./case.build seems to be successful.
./case.submit is giving the following error :


ERROR:
GETFIL: FAILED to get cime/projects/cesm-inputdata/atm/cam/chem/emis/CMIP6_emissions_1750_2015/emissions-cmip6_DMS_other_surface_1750-2015_0.9x1.25_c20170322
[hpc-ethanolx-01:188003] PMIX ERROR: UNREACHABLE in file server/pmix_server.c at line 2079

Within "cime/projects/cesm-inputdata/atm/cam/chem/emis/CMIP6_emissions_1750_2015" folder, there seems to be a file with "cmip6_DMS_other_surface_1750-2015_0.9x1.25_c20170322" but with a .nc extension.
Presently was just trying to create files with names which errupts in ./case.sumbit error log, so that they are a copy of the files already present in the folder.
Not sure if this is the right method though.

Any ideas for the same?

 

[fischer]

Hi Shruti,

You can download the data using ./check_input_data --download. If you get errors, try ftp ftp.cgd.ucar.edu as anonymous to see
if you can access the server.

Chris

 

[Shruti]

Hi,

I was able to get past the above error.
Now i am facing another issue.
" POP aborting...
ERROR kmt inconsistency for overflows"

According to this link, ,PaleoResources for CESM1.2

Runtime Issue » ERROR: original kmt not equal to actual kmt; POP aborting... ERROR kmt inconsistency for overflows

• Problem: The POP2 model may be reading bad data from the topography/bathymetry (KMT) file.
• Solution: Check that the KMT file was compiled with a big_endian flag.

I am not able to understand how this would work.
Is there any particular input file which i should check.
Any other suggestion would be appreciated

 

[fischer]

Hi Shruti,

It looks like the suggestion is to set the big_endian compiler flag when you build the model.

I also found this older thread.

https://bb.cgd.ucar.edu/cesm/threads/kmt-inconsistency-in-pop2.1553/

Chris

 

[Shruti]

Hi Chris,

Thank you for the reply.
Unfortunately the link which you shared didn't work out for me.
I am still facing the " ERROR kmt inconsistency for overflows " issue.

I have even included big-endian flag as "-fconvert=big-endian" in FFLAGS.
But the error isn't resolved. Any other approach i could try?

 

[fischer]

Hi Shruti,
Are there any other messages about the kmt inconsistency in the log files?
Chris

 

[Shruti]

Hi Chris

PFA the cesm log file which reflects kmt error

 

[mlevy]

Can you please post your ocn.bldlog file? From the cesm.log file you posted above, it looks like binary files are still being read in little-endian:

 init_overflows_kmt: KMT = ***** at global (i,j) =   198     7  changed to    33
 init_overflows_kmt: KMT = ***** at global (i,j) =   199     7  changed to    33
 init_overflows_kmt: KMT = ***** at global (i,j) =   200     7  changed to    33
 init_overflows_kmt: kmt inconsistencies for   3 points
 original kmt not equal to actual kmt

So I wonder if the -fconvert=big-endian option is being used in the build. (The ***** is noting that 5 digits is not enough space to write KMT, which should be between 0 and 60...)

 

[Shruti]

Thank you for your response. PFA the ocn.bldlog.
I have included "-fconvert=big-endian" in FFLAGS.
Macros.make reflects it as below

ifeq ($(COMPILER),clang)
FFLAGS := -fconvert=big-endian -ffree-line-length-none
endif

Should it additionally be included in another flag? I've just added in FFLAGS.

 

[fischer]

You should just need it for FFLAGS.

 

[Shruti]

Hello,

I am checking for CESM 2.1.2 , compset B1850 with clang compiler.
./case.submit is giving the following error :

ERROR:
GETFIL: FAILED to get cime/projects/cesm-inputdata/atm/cam/chem/emis/CMIP6_emissions_1750_2015/emissions-cmip6_DMS_other_surface_1750-2015_0.9x1.25_c20170322

But this error doesn't occur with GCC. Is the input data downloading dependent on compilers?

 

[fischer]

The downloading of input data is independent of the compiler.

 

[Shruti]

Hello,

Yes. Sorry for the confusion.
As suggested by Edwards in another post, I checked the input data path length.
It was exceeding the limit. Reduced the path length(changed relative path to absolute path) and the error resolved.
Thank you for the support.!! :)

 

[sunxupeng]

[QUOTE =“ Shruti，帖子：37413，成员：3550”]
你好，

是。对困惑感到抱歉。
正如Edwards在另一篇文章中所建议的那样，我检查了输入数据路径的长度。
它超出了限制。减少了路径长度（将相对路径更改为绝对路径）并解决了错误。
谢谢你的支持。！！:)
[/引用]
您好，我正在运行/ case.submit遇到此错误时：AttributeError：“ NoneType”对象没有属性“ getfile”
我该怎么办