[Ncep.list.nems.announce] Compiling NEMS regression test on Theia

[Raghu Reddy]

Hi Ratko,

 

I wanted some user application as a test for compilation and decided to try
the latest NEMS trunk on Theia.  After making a few minor changes I am able
to get it to compile on Theia using nceplibs from Moorthi's directory.

 

Since I don't have access to the WCOSS there is no way for me to do any
verification and checking on other platforms, but some of the things I found
may be helpful when you need to get everything working on Theia.

 

-          I checked out the NEMS trunk yesterday.

-          Everything I am doing is in the following directory:
/scratch4/NESCCINFRA/nesccmgmt/Raghu.Reddy/nems-trunk-2015-05-28

-          All the changes that have made are Theia specific, so it should
not impact anything else.

 

I have only verified the compilation part, because actually running the
tests requires more changes, particularly to the  *.IN and *.sh files that I
have not had time to do.

 

You can see the changes I have made by doing "svn diff".

 

The following files were modified:

tfe10.% svn diff|grep "Index:"

Index: tests/default_vars.sh

Index: tests/rt.sh

Index: tests/rt.conf

Index: src/configure

Index: src/conf/configure.nems.Theia.intel

Index: src/chem/gocart/src/Config/ESMA_arch.mk

tfe10.%

 

And the following two new files were added:

 

tfe10.% ll src/conf/modules.nems.Theia_ESMF_630*

-rwxr-xr-x 2 Raghu.Reddy nesccmgmt 163 May 28 20:25
src/conf/modules.nems.Theia_ESMF_630rAPI_gsm*

-rwxr-xr-x 2 Raghu.Reddy nesccmgmt 163 May 28 20:25
src/conf/modules.nems.Theia_ESMF_630rAPI_nmm*

tfe10.%

 

There is one specific thing I can point to that was causing the problem and
will be good to address which is the following:

 

The machine information as detected by detect_machine.sh is fairly robust
and is good.  

However, in the file: src/chem/gocart/src/Config/ESMA_arch.mk

 

The way machine is detected is very fragile:

 

  SITE := $(patsubst fe%,zeus,$(SITE))

  SITE := $(patsubst t%,wcoss,$(SITE))

  SITE := $(patsubst g%,wcoss,$(SITE))

 

It looks like the frontends for WCOSS starts with the letter "t", and so do
the frontends for Theia!  

 

Since the host name for theia also begins with the same letter it causes
some problems unless the host names in that makefile are capitalized etc.,
However there is a much simpler solution, which is to pass SITE on the gmake
command and the ./tests/rt.sh file:

 

227c241

<       gmake ${NEMS_VER} J=-j2                                     >>
$COMPILE_LOG 2>&1

---

>       gmake ${NEMS_VER} J=-j2 SITE=$MACHINE_ID                    >>
$COMPILE_LOG 2>&1

 

I just wanted to share this information in case it's useful.  Please let me
know if you need additional information.

 

Thanks,

Raghu

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://www.lstsrv.ncep.noaa.gov/pipermail/ncep.list.nems.announce/attachments/20150529/85999d2a/attachment.html