gen_be_stage2

Issues with running 3DVAR.

gen_be_stage2

Postby hnlim » Tue Jul 15, 2008 5:12 pm

Hi

I know these messages sound vague, but I can't find more hint to it.
I had ran through gen_be stage 0, stage1 and stage2a with no error messages. However when run stage2, I get 0 eigenvalue and NaN cumulative variance as shown.
-------------------------------------------------------------------------------------------------------
[3] Calculate eigenvectors, eigenvalues and inverse for psi/psi covariance
Mode Eigenvalue Cumulative Variance e(k,k)
1 0.0000E+00 NaN 0.0000E+00
2 0.0000E+00 NaN 0.0000E+00
---------------------------------------------------------------------------------------------------------
No clue at all where went wrong.

gen_be_stage2_nl.nl looks like this :
===================================
&gen_be_stage2_nl
start_date = '2003061700',
end_date = '2003062300',
interval = 24,
ne = 1,
testing_eofs = .true. /
=====================================
hnlim
 
Posts: 74
Joined: Wed Apr 16, 2008 11:51 am

Re: gen_be_stage2

Postby hclin » Wed Jul 16, 2008 11:35 am

Are you running on a linux machine?
gen_be_stage2 uses subroutine dsyev from LAPACK for eof_decomposition.
How were your BLAS/LAPACK compiled?
On this web page: http://www.mmm.ucar.edu/wrf/WG4/wrfvar/wrfvar-setup.htm
it says those external libraries need to be compiled with -r8. However, I don't think
it is necessary as the libraries have two versions of subroutines (one for double precision
starting with d*.f, the other one for single precision starting with s*.f), double_precision
ones are called in WRF-Var.
Not sure if this will solve the problem, but maybe you can try re-compile BLAS and LAPACK
without -r8, then re-compile all_wrfvar.
hclin
 
Posts: 68
Joined: Thu Apr 24, 2008 7:21 pm

Re: gen_be_stage2

Postby hnlim » Wed Jul 16, 2008 6:58 pm

Hi

Yes I am running on Linux and I do have those two libraries compiled with -r8.
It does not give enough error messages for me to determine what was wrong.
hnlim
 
Posts: 74
Joined: Wed Apr 16, 2008 11:51 am

Re: gen_be_stage2

Postby hclin » Fri Jul 18, 2008 6:08 pm

More details:
The failure in gen_be_stage2 is probably from lapack/*lamc*.f where
comparisons between two floating point numbers are used in if statements.

When compiling LAPACK,
if using g95, adding -ffloat-store,
if using ifort, adding -fltconsistency,
might help.
hclin
 
Posts: 68
Joined: Thu Apr 24, 2008 7:21 pm

Re: gen_be_stage2

Postby lzhang » Mon Aug 04, 2008 9:54 pm

I also have the same problem when I using wrf 3dvar V3 on linux, I have compiled the lapack library with pgf90 -r8 , anyone have ideas?
Thanks!
lzhang
 
Posts: 16
Joined: Thu Jul 17, 2008 1:44 pm

Re: gen_be_stage2

Postby hclin » Wed Aug 06, 2008 11:23 am

Does your gen_be_stage2.log also show message as below?
hnlim wrote:-------------------------------------------------------------------------------------------------------
[3] Calculate eigenvectors, eigenvalues and inverse for psi/psi covariance
Mode Eigenvalue Cumulative Variance e(k,k)
1 0.0000E+00 NaN 0.0000E+00
2 0.0000E+00 NaN 0.0000E+00
---------------------------------------------------------------------------------------------------------

If so, that means the stage0 was not run correctly (all the forecast difference values are zero).
hclin
 
Posts: 68
Joined: Thu Apr 24, 2008 7:21 pm

Re: gen_be_stage2

Postby lzhang » Thu Aug 07, 2008 12:31 pm

Thank you for your reply.
It's true, but I don't know why

my data is arranged as following:
2006082112wrfout_d01_2006-08-21_12:00:00
2006082112wrfout_d01_2006-08-22_00:00:00
2006082112wrfout_d01_2006-08-22_12:00:00
2006082200wrfout_d01_2006-08-22_12:00:00
2006082200wrfout_d01_2006-08-23_00:00:00
2006082200wrfout_d01_2006-08-23_12:00:00
.
.
2006082412wrfout_d01_2006-08-24_12:00:00
2006082412wrfout_d01_2006-08-25_00:00:00
2006082412wrfout_d01_2006-08-25_12:00:00

START_DATE=2006082212
END_DATE=2006082500
NUM_LEVELS=27 (in wrf model, n_vert=28)

anything wrong with that?

Thanks !
lzhang
 
Posts: 16
Joined: Thu Jul 17, 2008 1:44 pm

Re: gen_be_stage2

Postby lzhang » Thu Aug 07, 2008 12:33 pm

sorry, the data should be like this:2006082112wrfout_d01_2006-08-21_12:00:00
2006082112/wrfout_d01_2006-08-22_00:00:00
2006082112/wrfout_d01_2006-08-22_12:00:00
2006082200/wrfout_d01_2006-08-22_12:00:00
2006082200/wrfout_d01_2006-08-23_00:00:00
2006082200/wrfout_d01_2006-08-23_12:00:00
.
.
2006082412/wrfout_d01_2006-08-24_12:00:00
2006082412/wrfout_d01_2006-08-25_00:00:00
2006082412/wrfout_d01_2006-08-25_12:00:00
lzhang
 
Posts: 16
Joined: Thu Jul 17, 2008 1:44 pm

Re: gen_be_stage2

Postby lzhang » Sat Aug 09, 2008 6:32 pm

the problem has been resolved
lzhang
 
Posts: 16
Joined: Thu Jul 17, 2008 1:44 pm

Re: gen_be_stage2

Postby beckyselin » Mon Aug 18, 2008 2:43 pm

I have a somewhat similar problem, although not exactly the same. I am attempting to run the "Running gen_be" portion of the WRF-VAR tutorial in the WRF user's guide. I have using WRF-VAR v3.0.1, compiled for an AIX system in serial. Gen_be_wrapper.ksh fails in Stage 2a, with the following error:

Code: Select all
"gen_be_stage2a.f", line 85: 1525-001 The READ statement on the file gen_be_stage2.dat cannot be completed because the end of the file was reached.  The program will stop.
Stage 2a failed with error 1


However, examination of gen_be_stage2.log shows the error is actually in Stage 2, with the following message in the log:

Code: Select all
 [3] Calculate eigenvectors, eigenvalues and inverse for psi/psi covariance
 ** On entry to DSYTRD parameter number  1 had an illegal value


DSYTRD.f is part of the lapack-3.1.1 library. I've tried compiling the lapack, blas, bufrlib libraries both with and without the -qrealsize=8 option (I'm compiling with xlf), but both ways produce the same error.

The first parameter passed to dsytrd.f is UPLO, a single character variable. However, I cannot find where the dsytrd function is called in the code, so I don't know what is being passed that would cause such an error. I'm at a loss and would appreciate any help - thanks in advance!

Becky
beckyselin
 
Posts: 13
Joined: Thu Jul 31, 2008 3:54 pm

Next

Return to Runtime Problems

Who is online

Users browsing this forum: No registered users and 3 guests