[RegCNET] RegCM3 on SGI Origin 2000

Fri Jul 21 19:50:41 CEST 2006

Hi Simon and others,

There were several things we needed to change to get the new version of
RegCM3 working on our SGI Origin 300.  For reference our compiler version
is: MIPSpro Compilers: Version 7.4.2m

* Each step of the preprocessing needs to be compiled with f90.  Also, ibyt=
e
must be set to '4' when compiling with f90 on an SGI, not '1'.

* To address item 2 below, in param.F we changed iutbc to '103', not '71' a=
s
is noted in the code.  That comment should be changed in the next release o=
f
the model.  You will also need to update ICBC.f in the ICBC directory, set
ifile =3D 103.

> 2. When running regcm.x the ICBC files (fort.101, fort.102...) are not
> recognized. It lloks like the SGI compiler does not accept XXX numbers.
> Somebody has apparently noticed it - the the papam.F there is a commented=
 line
> where it is suggested to use iutbc =3D 71 on SGIs. This does not help howev=
er -
> the code still expects the XXX form (' 71' and not '71').

* This is our FFLAGS line:  FFLAGS =3D -64 -O3 -static -TENV:check_div=3D3:X=3D0
-TARG:exec_min=3DUOZV:exec_max=3DUOZV -OPT:fast_nint=3DOFF:Olimit=3D4657 -IPA
-LANG:short_circuit_conditionals=3DON:recursive=3DON

You may need to make some modifications for your system, but these are
fairly general compile options.  I think this might help with the floating
point exceptions you are having.

* If you are trying to run with the parallel code you should change the
bottom of your Makefile to this:

all: regcm

.f.o:=20
        $(FC) -c $(FFLAGS) $*.f -lmpi

regcm:  $(OBJ)
        $(LD) -o $@ $(FFLAGS) $(OBJ) -lmpi

clean:
        rm -f $(OBJ) regcm *.o

I also agree with Jing about changing the lines below, it doesn't need to b=
e
done.  The problem was with MAKECODE (see my next point below)

> .F.o:
>         $(FC) -c $(FFLAGS) $*.F
> =20
> Instead of:
> .f.o:
>         $(FC) -c $(FFLAGS) $*.f

* I should also point out again what Jing said below, that you should check
the paths to 'cpp' in MAKECODE and change them to match your system, as wel=
l
as the path to 'csh'

>> I met exactly the same problem  like yours in the past few days. The rea=
son
>> is that the 'MAKECODE'file didn't  invoke *.F files into *.f. You may ch=
eck
>> if the path of 'cpp' in your  'MAKECODE' is right. I Hope this could be
>> helpful to you.
>>=20
>> In my case, I  checked the 'cpp' path, and also deleted the first line (=
i.e.,
>> '#!/lib/csh  -f'), and then the model seemed working well until it gave =
me
>> another error of  'floating exception'as the  following:

Let me know how things go,

Mark

--=20
Mark A. Snyder
Postdoctoral Researcher
Climate Change and Impacts Laboratory
Dept. of Earth and Planetary Sciences
UC Santa Cruz
(W) 831-459-3504
msnyder at es.ucsc.edu
http://es.ucsc.edu/~msnyder
--

On 7/21/06 1:45 AM, "Simon Krichak" <shimon at cyclone.tau.ac.il> wrote:

> Dear all,
> Hi Jing, Mark:=20
> =20
> Something looks wrong with the of the RegCM3 code on my (at least) SGI Or=
igin
> 2000.
> I see two - probably interrelated, problems.
> =20
> 1. To compile the Main on the SGI I had to modify two lines in the
> Makefile_SGI
> =20
> .F.o:
>         $(FC) -c $(FFLAGS) $*.F
> =20
> Instead of:
> .f.o:
>         $(FC) -c $(FFLAGS) $*.f
> =20
> =20
> 2. When running regcm.x the ICBC files (fort.101, fort.102...) are not
> recognized. It lloks like the SGI compiler does not accept XXX numbers.
> Somebody has apparently noticed it - the the papam.F there is a commented=
 line
> where it is suggested to use iutbc =3D 71 on SGIs. This does not help howev=
er -
> the code still expects the XXX form (' 71' and not '71').
> =20
> The run stops anyway:
> =20
> BCs are ready from  2001010100   to  2001010106
>  dt, dtau =3D  150.,  37.5,  75.
> 0 linearization about standard atmosphere (lstand=3D.t.)
> 0sigmaf      0.00      4.000E-02  0.100      0.170      0.250      0.350
> 0.
> 460      0.560      0.670      0.770      0.860
>            0.930      0.970      0.990       1.00
> 0t mean      218.       218.       218.       225.       236.       247.
> 2
> 56.       264.       272.       277.       282.
>             285.       286.       287.
> 0ps mean     100.
> 0 vertical mode problem completed for kx=3D 14     0 errors detected   (sho=
uld
> be
> 0)
>  m, fac =3D  4,  33.333333333333336
>  m, fac =3D  2,  60.
>  Writing output files in direct access format
> =20
>  ******* OPENING NEW OUTPUT FILES: 2001010100
>  OPENING NEW OUT FILE: output/ATM.2001010100
>  OPENING NEW BAT FILE: output/SRF.2001010100
>  OPENING NEW RAD FILE: output/RAD.2001010100
>  OUT-history written date =3D  2001010100.
> Floating Exception
> Abort
> =20
> Mark: Are you running the new RegCM3 code on the Origin 3000? Which compi=
ler
> version do you use?
> =20
> Any comment from the ICTP please.
> =20
> Best, Simon=20
> =20
>    =20
> =20
>> =20
>> ----- Original Message -----
>> =20
>> From:  =D6=A3=E6=BA <mailto:jzheng at mail.iap.ac.cn>
>> =20
>> To: Simon Krichak <mailto:shimon at cyclone.tau.ac.il>  ; Mark Snyder
>> <mailto:msnyder at pmc.ucsc.edu>
>> =20
>> Cc: regcnet at lists.ictp.it
>> =20
>> Sent: Wednesday, July 19, 2006 5:49  AM
>> =20
>> Subject: Re: [RegCNET] RegCM3 on an SGI  Origin 2000
>> =20
>>=20
>> Hi, Simon,
>>=20
>> I met exactly the same problem  like yours in the past few days. The rea=
son
>> is that the 'MAKECODE'file didn't  invoke *.F files into *.f. You may ch=
eck
>> if the path of 'cpp' in your  'MAKECODE' is right. I Hope this could be
>> helpful to you.
>>=20
>> In my case, I  checked the 'cpp' path, and also deleted the first line (=
i.e.,
>> '#!/lib/csh  -f'), and then the model seemed working well until it gave =
me
>> another error of  'floating exception'as the  following:
>>=20
>> ------------------------------
>> Writing output files in  direct access format
>>=20
>> ******* OPENING NEW OUTPUT FILES:  1998062500
>> OPENING NEW OUT FILE: output/ATM.1998062500
>> OPENING NEW BAT  FILE: output/SRF.1998062500
>> OPENING NEW RAD FILE:  output/RAD.1998062500
>> OUT-history written date =3D 1998062500.
>> Floating  Exception
>> Abort (core dumped)
>>=20
>> ---------------------------------
>>=20
>> And I'm also asking for help on this "floating exception" error,  becaus=
e the
>> model works well in Linux system with the same codes.
>>=20
>> Best  Regards,
>>=20
>>=20
>> Jing
>>>=20
>>>=20
>>> -----=D4=AD=CA=BC=D3=CA=BC=FE-----
>>> =B4=D3:  "Simon Krichak"
>>> =B5=BD: "Mark Snyder"
>>> =B3=AD=CB=CD: regcnet at lists.ictp.it
>>> =C8=D5=C6=DA:  2006/07/19(=C8=FD) 01:52am
>>> =D6=F7=CC=E2: Re: [RegCNET] RegCM3 on an SGI Origin  2000
>>>=20
>>> Hi Marc,
>>>=20
>>> I am using the SGI version of  the Makefile (see below).
>>> The diagnostic (repated below) is for  aermix.o only.
>>>=20
>>> Simon
>>>=20
>>>=20
>>>=20
>>> ld64:  FATAL 12 : Expecting n64 objects: aermix.o is n32.
>>>> *** Error code  2 (bu21)
>>>=20
>>>=20
>>> # Makefile for RegCM@
>>> # for SGI (by  Zhang Yan)
>>>=20
>>> FFLAGS =3D -64 -g -trapuv
>>> #FFLAGS =3D  -O2
>>> # DEBUG -- FFLAGS =3D -g -trapuv
>>> FC =3D f90
>>> LD =3D  $(FC)
>>>=20
>>> SRC =3D \
>>> aermix.f aeroppt.f aerout.f albedov.f  balanc.f balbak.f bconst.f bdyin=
.f
>>> bdyuv.f
>>> \
>>> bdyval.f  blhnew.f BLOCKDATA001.f bmpara.f bndry.f \
>>> carbon.f cdiv.f chdrydep.f  chemtap.f chrsetc.f chsrfem.f \
>>> cldefr.f cldems.f cldfrac.f co2.f  colmod3.f conadv.f condch.f \
>>> condcq.f condtq.f conmas.f cumtran.f  cupara.f \
>>> cuparan.f cupemandrv.f cupeman.f cup.f \
>>> depth.f  deriv.f diffu.f diffut.f \
>>> dragdn.f drag.f drip.f dsflx.f EDDY.f  elmhes.f eltran.f \
>>> eomb.f FINDDATE.f for_next.f frawat.f getdat.f  grads_stuf.f \
>>> hadv.f holtbl.f \
>>> hqr2.f hqr.f htdiff.f ice.f  inirad.f initb.f INITDATE.f init.f \
>>> inidust.f initlk.f interf.f  intmax.f invmtrx.f isamax.f isrchfgt.f \
>>> isrchfle.f lakedrv.f lake.f  lenchr.f lfdrag.f lftemp.f \
>>> ljchrn.f lshfch.f lutbl.f mapsmp.f  maximi.f minimi.f MIXER.f mkfile.f =
\
>>> nconvp.f nudge.f o3data.f  outprt.f output.f \
>>> outsav.f outsrf.f outsub.f outtap0.f outtap.f  param.f pcp.f \
>>> radabs.f radclr.f radclw.f radcsw.f radctl.f radded.f  radems.f \
>>> radini.f radinp.f radout.f radoz2.f radtap.f radtpl.f  reorder.f regcm.=
f \
>>> resetr.f rg.f root.f satur.f saxpy.f sgedi.f  sgefa.f slice.f slice3d.f=
 \
>>> snow.f soilbc.f solar1.f spinit.f spline.f  splitf.f \
>>>=20
>>>=20
>>> ----- Original Message -----
>>>  From: "Mark Snyder"
>>> To: "Simon Krichak"
>>> Sent: Tuesday, July  18, 2006 7:12 PM
>>> Subject: Re: [RegCNET] RegCM3 on an SGI Origin  2000
>>>=20
>>>=20
>>> Hi Simon,
>>>=20
>>> In your Makefile, I  would check to see if the ?4 flag is set in FFLAGS=
. I
>>> have a feeling  that aermix.f is being compiled as 32 bit when it shoul=
d be
>>> 64. Are  you getting this message for all the .f files, or just this on=
e?
>>> =20
>>> Mark
>>>=20
>>> --=20
>>> Mark A. Snyder
>>> Postdoctoral  Researcher
>>> Climate Change and Impacts Laboratory
>>> Dept. of  Earth and Planetary Sciences
>>> UC Santa Cruz
>>> (W)  831-459-3504
>>> msnyder at es.ucsc.edu
>>>  http://es.ucsc.edu/~msnyder
>>> --
>>>=20
>>>=20
>>> On 7/18/06  10:00 AM, "Simon Krichak" wrote:
>>>=20
>>>> Dear all,
>>>>=20
>>>> I have the following compilation problem in the Main  directory of the=
 new
>>>> RegCM3 on the SGI Origin 2000.
>>>>=20
>>>> ld64: FATAL 12 : Expecting n64 objects: aermix.o is  n32.
>>>> *** Error code 2 (bu21)
>>>> Any recommendation  please.
>>>>=20
>>>> Simon
>>>>=20
>>>> Dr. Simon O.  Krichak
>>>> Dept. of Geophysics and Planetary Sciences
>>>>  Faculty of Exact Sciences,
>>>> Tel Aviv University, Ramat Aviv, Tel  Aviv,
>>>> 69978, Israel
>>>> Tel: 972-3-6405694(o),  972-3-9515748
>>>> Fax: 972-3-6409282
>>>>=20
>>>>  E-mail: shimon at cyclone.tau.ac.il,
>>>> simon.krichak at gmail.com
>>>>=20
>>>>=20
>>>>  _______________________________________________
>>>> RegCNET mailing  list
>>>> RegCNET at lists.ictp.it
>>>>  https://lists.ictp.it/mailman/listinfo/regcnet
>>>=20
>>>=20
>>> =20
>>>=20
>>> +++++++++++++++++++++++++++++++++++++++++++
>>> This  Mail Was Scanned By Mail-seCure System
>>> at the Tel-Aviv University  CC.
>>>=20
>>>=20
>>>  _______________________________________________
>>> RegCNET mailing  list
>>> RegCNET at lists.ictp.it
>>>  https://lists.ictp.it/mailman/listinfo/regcnet
>>>=20
>>> =20
>>=20
>>=20
>> ***************************************
>> * Jing ZHENG, Ph.D.  Student *
>> * Institute of Atmospheric Physics *
>> * Chinese Academy of  Sciences *
>> * P.O.Box 9804, *
>> * Beijing 100029, China *
>> * E-mail:  jzheng at mail.iap.ac.cn *
>> * Tel: 86-10-62063912 (Office)  *
>> ***************************************
>>=20
>>=20
>>   r ++++++++++++++b h iW aeBMlerSt<>at
>=20