[molpro-user] Failing tests on Opteron System.

[Fabio Mariotti]

	Dear Molpros,

	I have a version of molpro compiled on AMD Opteron
	and the first set of tests (make test) went ok
	(except that I had to increase the memory for the form_*
	tests).

	Here details:

	> uname -a
Linux hostname 2.6.8-24.11.qsnetp1-smp #1 SMP Fri Jan 14 14:01:26 CET 2005 x86_64 x86_64 x86_64 GNU/Linux
(I'm not administrator of the machine and don't know exact details on the copiled kernel)

	> head -5 /proc/cpuinfo
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 15
model           : 5
model name      : AMD Opteron(tm) Processor 250

	> pgf90 -V
pgf90 6.0-5 64-bit target on x86-64 Linux

	extracted CONFIG lines:

ARCHITECTURE="unix unix-i8 unix-linux unix-linux-x86_64 mpp"
# Compilers ..
CC="cc -Dpgf90 -DI64 -m64"
FC="/opt/pgi/linux86-64/6.0/bin/pgf90 -pc 64 -i8 "
F90="/opt/pgi/linux86-64/6.0/bin/pgf90 -pc 64 -i8 "
# BLAS library
BLASLIB="-lacml"
# LAPACK library
LAPACKLIB=""
MPPNAME="tcgmsg"
ARCHNAME="amd64"
BLASLIB_amd64="-lacml"

	Hope it helps
	Best
	Fabio



Holger Naundorf wrote:
> Hello,
> I am just trying to get molpro2002.6 installed on an AMD Opteron system
> and would be interested in any experiences other people might have had
> with this.
> 
> Currently I have the following setup:
> 
> 4CPU Opteron System with 32GB of RAM
> 
> OS is Red Hat Enterprise Linux AS release 4 (Nahant Update 1)
> 
> uname -srvp
> Linux 2.6.9-11.ELsmp #1 SMP Fri May 20 18:25:30 EDT 2005 x86_64
> 
> Compiler used at the moment:
> Intel(R) C Compiler for 32-bit applications, Version 9.0    Build
> 20050430Z Package ID: l_cc_p_9.0.021
> Intel(R) Fortran Compiler for 32-bit applications, Version 9.0    Build
> 20050430Z Package ID: l_fc_p_9.0.021
> 
> The symptoms:
> Trying to build with the 64bit versions of the compilers failed already
> at the ./configure stage, so I am using the 32bit at the moment.
> 
> Compilation went fine using the RedHat supplied BLAS/Lapack libs.
> 
> But doing a 'make test' results in each test failing with something
> like:
> 
> --------------------------------------------------------------------
>  ERROR WRITING           89 WORDS AT OFFSET       4096. TO FILE 1 
> IMPLEMENTATION=df   FILE HANDLE=  8  IERR=    -1
>  ? Error
>  ? I/O error
>  ? The problem occurs in writew
>  
>  ERROR EXIT
>  CURRENT STACK:      MAIN
> Bad seek in iow_direct_write; fd=8, p=4096
>  
> 
> 
> **********************************************************************************************************************************
>  
>  ERROR WRITING          349 WORDS AT OFFSET       4185. TO FILE 1 
> IMPLEMENTATION=df   FILE HANDLE=  8  IERR=    -1
>  ? Error
>  ? I/O error
>  ? The problem occurs in writew
>  
>  ERROR EXIT
>  CURRENT STACK:      MAIN
> Bad seek in iow_direct_write; fd=8, p=4185
>  
>  RECURSIVE CALL OF ERROR HANDLER, FILE INFORMATION MAY BE LOST
>  
>  EMERGENCY STOP
> tl_so_i.test: ERRORS DETECTED: non-zero return code ... inspect output
> 
> -----------------------------------------------------------------
> 
> right after startup/allocating memory.
> So ther is some trouble with file access/writing...
> 
> All the directories given in 'molpro.rc' exist and are writable and
> after the run I end up with a huge list of files called 'fort??????' in
> the dir stored in $TMPDIR (either empty or containing snippets of input,
> like:
>             SET,'ORBITAL',
>             SPD,O,AUG-CC-PVDZ,
>             C,
>             S,H,AVDZ,
>             C,
>             SPD,AR,ECP10MHF,
>             C,
>             D,AR,AVDZ,
>             ECP,AR,ECP10MHF,
>             END
> )
> 
> As I have no immediate idea where the failure might come from I would
> like to know if anyone else has encountered something similar, or has a
> similar setup and knows a combination of compiler/libs that work...
> 
> Thanks in advance,
> Holger N.

-- 
----------------------------------------------
Dr. F.Mariotti - mariotti at ir.phys.chem.ethz.ch
ETH Zurich - Honggerberg, Zurich, Switzerland
----------------------------------------------