[molpro-user] {SPAM}? Re: {SPAM}? Re: Catastrophic error: internal compiler error on src/local/geminal.f

Andy May MayAJ1 at cardiff.ac.uk
Wed May 6 16:28:35 BST 2009


Jacek,

Further to my previous email, I have discovered that when using full
optimization with ifort 11 on code containing deeply nested loops the
code can sometimes go badly wrong. The solution is to add:

drv2el_l3ext_lmp2g.f drv2el_l3ext_cen.f to F_OPT1
basis_integral_shells.f to F_OPT2

in CONFIG, and then obviously delete the corresponding object files and
run make. This allowed all testjobs to run with ifort 11.

The above changes will be patched into Molpro 2008.1 in the next few days.

Best wishes,

Andy


Andy May wrote:
> Jacek,
> 
> It's good news that many of the jobs are running.
> 
> The 'some files are not ...' message was because configure had been
> updated in recent patches resulting in your CONFIG file being out of
> date; re-running configure would have been the solution.
> 
> I notice that you built GA with TARGET=MACX, should this not be MACX64
> for 64-bit mac? This could well be the cause of the problem if the size
> of data types does not match between Molpro and GA. If you compile and
> run serial Molpro do you see the same failures?
> 
> We have only recently gained access to the Intel 11 compilers so it is
> also possible there may be problems with Molpro exposed only by this
> compiler, problems with the compiler itself (like the known compilation
> bug for geminal.f or over aggressive optimization) or a mixture of both.
>   I will be testing extensively in the coming weeks and will patch in
> any necessary changes.
> 
> Best wishes,
> 
> Andy
> 
> jacek klos wrote:
>> Dear Andy, 
>>
>>  I used 'make patch" to apply patches but on my Mac Pro after
>> application of patches and issuing 'make' command it was complaining
>> that 'some files are not on the FOPT list' or something like that.
>> So I downloaded new source with the latest patches included and
>> compilation of parallel molpro with mpich2 and GA-4.1.1 went with no
>> problem. I could start mpp test with 8 cpus and all  go OK except the
>> following one with restart:
>>
>> make MOLPRO_OPTIONS=-n8 test
>> make[1]: Nothing to be done for `default'.
>> make[2]: `../lib/libmolpro.a' is up to date.
>> make[2]: `../bin/molpro.exe' is up to date.
>> make[1]: Nothing to be done for `default'.
>> Running job h2o_restart2.test
>> Read error in iow_direct_read; fd=12, l=20, p=114290; read returns 0
>> Read error in iow_direct_read; fd=12, l=20, p=114290; read returns 0
>> Read error in iow_direct_read; fd=12, l=20, p=114290; read returns 0
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:40342):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:40343):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:40341):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> ERROR in 'expr-expr' length of vectors don't match
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:40340):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> ERROR in 'expr-expr' length of vectors don't match
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:40337):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> **** PROBLEMS WITH JOB h2o_restart2.test
>> **** For further information, look in the output file
>> **** /Users/jklos/Molpro/MPP/molpro2008.1/testjobs/h2o_restart2.errout
>>
>>
>> Running job system.test
>> **** PROBLEMS WITH JOB system.test
>> Binary file system.out matches
>> **** For further information, look in the output file
>> **** /Users/jklos/Molpro/MPP/molpro2008.1/testjobs/system.errout
>>
>>
>> Running job c6h6_freq_restart.test
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:66298):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:66299):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> Read error in iow_direct_read; fd=12, l=16, p=300349; read returns 0
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:66297):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> Read error in iow_direct_read; fd=12, l=1, p=300349; read returns 0
>> Read error in iow_direct_read; fd=12, l=1, p=300349; read returns 0
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:66296):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>> 0:0:fehler on processor   0:: -1
>> (rank:0 hostname: pid:66301):ARMCI DASSERT fail.
>> armci.c:ARMCI_Error():260 cond:0
>>
>>
>> I used following commands to compile mpich2 and GA-4.1.1 on MacPro with
>> ifort 11.064: 
>>
>> for mpich2 64bit:
>>
>> export F77=ifort
>> export F90=ifort
>> export CFLAGS="-arch x86_64 -mmacosx-version-min=10.5 -m64 -O3  -DMACOS"
>>  ./configure --prefix=/usr/local/mpich2_64bit
>>
>> for GA-4.1.1 64bit 
>>
>> make TARGET=MACX FC=ifort CC=gcc COPT="-arch x86_64
>> -mmacosx-version-min=10.5 -m64 -O3  -DMACOS"
>> MPI_INCLUDE=/usr/local/mpich2_64bit/include
>> MPI_LIB=/usr/local/mpich2_64bit/lib LIBMPI="-lmpich -lpmpich" USE_MPI=yes
>>
>>
>> Many thanks
>>
>> Jacek
>>
>>
>>
>>
>> On Apr 27, 2009, at 8:09 PM, Andy May wrote:
>>
>>> Jacek,
>>>
>>> This is a known problem (see mailing list archives for February) with
>>> the Intel compiler, which Molpro automatically avoids if you have all of
>>> the latest patches applied.
>>>
>>> Best wishes,
>>>
>>> Andy
>>>
>>> jacek klos wrote:
>>>> Dear Molpro Users and Developers
>>>>
>>>> We got new machines Mac Pro from Apple with dual quad i7 2.66GHz CPUs
>>>> with Mac OS system Leopard:
>>>> Darwin hostname 9.6.3 Darwin Kernel Version 9.6.3: Tue Jan 20 18:26:40
>>>> PST 2009; root:xnu-1228.10.33~1/RELEASE_I386 i386
>>>>
>>>> We got the newest ifort compiler for Mac OS X and MKL:
>>>> /opt/intel/Compiler/11.0/059/bin/intel64/
>>>>
>>>> When trying to compile molpro 2008.1 with ./configure -i8 option
>>>> compilation goes, but at the src/local/geminal.f file it crashes:
>>>>
>>>> geminal.f(3066): (col. 9) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(2971): (col. 9) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(2852): (col. 12) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(2892): (col. 17) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(2751): (col. 9) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(2628): (col. 25) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(2165): (col. 20) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(3137): (col. 11) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(553): (col. 7) remark: PARTIAL LOOP WAS VECTORIZED.
>>>> geminal.f(192): (col. 15) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(226): (col. 13) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(354): (col. 12) remark: PARTIAL LOOP WAS VECTORIZED.
>>>> geminal.f(500): (col. 9) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(850): (col. 17) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(1752): (col. 14) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(1758): (col. 14) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(1853): (col. 14) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(1950): (col. 15) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(2014): (col. 11) remark: LOOP WAS VECTORIZED.
>>>> geminal.f(2055): (col. 13) remark: LOOP WAS VECTORIZED0_1871
>>>>
>>>> : catastrophic error: **Internal compiler error: internal abort** Please
>>>> report this error along with the circumstances in which it occurred in a
>>>> Software Problem Report.  Note: File and line given may not be explicit
>>>> cause of this error.
>>>> in file geminal.f, line 2348, column 17
>>>>
>>>> .
>>>> compilation aborted for geminal.f (code 3)
>>>>
>>>> When one uses the older ifort version 10.1, the compilation goes OK and
>>>> molpro executables are linked and this
>>>> error do not appear.
>>>>
>>>>
>>>> Here is CONFIG file:
>>>>
>>>> CONFIGURE_OPTIONS="-i8"
>>>>
>>>> AR=ar
>>>> ARCHNAME=Darwin/x86_64
>>>> ARFLAGS=-rS
>>>> AWK=awk
>>>> BIBTEX=bibtex
>>>> BLASLIB=-L/Library/Frameworks/Intel_MKL.framework/Libraries/em64t
>>>> -lmkl_intel_ilp64 -lmkl_sequential -lmkl_core
>>>> BUILD=s
>>>> CAT=cat
>>>> CC=/usr/bin/gcc
>>>> CCVERSION=4.0.1
>>>> CC_FRONT=
>>>> CDEBUG=-g
>>>> CFLAGS=-m64 -DINT64 -DZLIB
>>>> CLEARSPEEDLIB=
>>>> CMPPINCLUDE=
>>>> COPT=-O3
>>>> COPT0=-O0
>>>> COPT1=-O1
>>>> COPT2=-O2
>>>> COPT3=-O3
>>>> CP=cp -p
>>>> CPROFILE=-p
>>>> CSCN=cscn
>>>> CSFLAGS=-O3 -I. --dynamic
>>>> C_OPT0=
>>>> C_OPT1=
>>>> C_OPT2=
>>>> C_OPT3=
>>>> EXPORT=export
>>>> F90FLAGS=
>>>> FC=/opt/intel/Compiler/11.0/059/bin/intel64/ifort
>>>> FCVERSION=11.0.059
>>>> FDEBUG=-g
>>>> FFLAGS=-pc 64 -auto -autodouble -Vaxlib -i8
>>>> FMPPINCLUDE=
>>>> FOPT=-O3
>>>> FOPT0=-O0
>>>> FOPT1=-O1
>>>> FOPT2=-O2
>>>> FOPT3=-O3
>>>> FPP=-fpp
>>>> FPROFILE=-p
>>>> FSTATIC=
>>>> FTCFLAGS=unix-i8 Darwin lapack blas1 blas2 blas3
>>>> F_OPT0=util/iom.F explicit/explicit_util.f
>>>> F_OPT1=
>>>> F_OPT2=mrci/cicon.f sapt/integrals.f90 mrci/cirs2iic.f nevpt2/koopro4.F
>>>> F_OPT3=
>>>> GROUPPOS=4
>>>> HDF5=
>>>> HOSTFILE_FORMAT=
>>>> INSTBIN=/usr/local/bin
>>>> INSTHTML=/Users/jklos/public_html/molpro/molpro2008.1
>>>> INSTLIB=/usr/local/lib/molpros_2008_1_Darwin_x86_64_i8
>>>> INTEGER=8
>>>> LAPACKLIB=
>>>> LATEX2HTML=latex2html
>>>> LAUNCHER=
>>>> LD_ENV=/Library/Frameworks/Intel_MKL.framework/Libraries/em64t
>>>> LD_ENVNAME=DYLD_LIBRARY_PATH
>>>> LIBRARY_SUFFIX=a
>>>> LIBS=-lz
>>>> LIBS_FRONT=
>>>> LINKOPT=-framework CoreFoundation -framework IOKit
>>>> LINKOPT_FRONT=
>>>> LN=ln -s
>>>> LSFULL=ls -l
>>>> MAKEDEPEND_OPTIONS=
>>>> MAKEINDEX=makeindex
>>>> MAPLE=
>>>> MKDIR=mkdir -p
>>>> MODULE_FLAG=-I
>>>> MODULE_SUFFIX=mod
>>>> MOLPRO_CACHE=12288
>>>> MOLPRO_FLOPDGM=0386
>>>> MOLPRO_FLOPDGV=5432
>>>> MOLPRO_FLOPMXM=5247
>>>> MOLPRO_FLOPMXV=2369
>>>> MOLPRO_MINDGC=12
>>>> MOLPRO_MINDGL=3
>>>> MOLPRO_MINDGM=16
>>>> MOLPRO_MINDGR=1
>>>> MOLPRO_MINDGV=12
>>>> MOLPRO_MXMBLK=64
>>>> MOLPRO_MXMBLN=64
>>>> MOLPRO_SHAREDLIB=0
>>>> MOLPRO_UNROLL=2
>>>> MPILIB=
>>>> MPPLIB=
>>>> OBJECT_SUFFIX=o
>>>> OWNERPOS=3
>>>> PAPER=a4paper
>>>> PARSE=parse-Darwin-x86_64-i8.o
>>>> PATCHER=patcher.exe
>>>> PDFLATEX=pdflatex
>>>> PNAME=molpros_2008_1_Darwin_x86_64_i8
>>>> PTSIZE=11
>>>> RANLIB=ranlib -c
>>>> RM=rm -rf
>>>> SHAREDLIBFLAGS=
>>>> SHAREDLIBSUFFIX=dylib
>>>> SHELL=/bin/sh
>>>> STRIP=echo WARNING, strip not used; true
>>>> TAR=tar -cf
>>>> UNTAR=tar -xf
>>>> VERSION=2008.1
>>>> XSD=/usr/bin/xmllint --noout --schema
>>>> XSLT=/usr/bin/xsltproc
>>>> YACC=bison -b y
>>>>
>>>> .SUFFIXES:
>>>> MAKEFLAGS+=-r
>>>>
>>>>
>>>> Has anyone have experience with new ifort 11 on Mac OS X and with molpro
>>>> 2008.1 ?
>>>>
>>>>
>>>> I think it maybe bug in the new ifort Compiler version, but I guess I
>>>> can't submit this crash report to Intel including geminal.f source file
>>>> due to license agreements.
>>>>
>>>> I will be glad if someone has some experience with it and can share.
>>>>
>>>> Best regards
>>>> Jacek Klos
>>>>
>>>>
>>>> __________________________________
>>>> Jacek Klos, Ph. D.
>>>> Department of Chemistry and Biochemistry
>>>> University of Maryland
>>>> College Park, MD 20742 USA
>>>> http://www.researcherid.com/rid/A-6457-2008
>>>> __________________________________
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> ------------------------------------------------------------------------
>>>>
>>>> _______________________________________________
>>>> Molpro-user mailing list
>>>> Molpro-user at molpro.net <mailto:Molpro-user at molpro.net>
>>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>> _______________________________________________
>>> Molpro-user mailing list
>>> Molpro-user at molpro.net <mailto:Molpro-user at molpro.net>
>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>
>>
>> __________________________________
>> Jacek Klos, Ph. D.
>> Department of Chemistry and Biochemistry
>> University of Maryland
>> College Park, MD 20742 USA
>> http://www.researcherid.com/rid/A-6457-2008
>> __________________________________
>>
>>
>>
>>
>>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user



More information about the Molpro-user mailing list