[molpro-user] Parallel Molpro 2002.6 on ALTIX 350

Ross Nobes Ross.Nobes at uk.fujitsu.com
Thu Sep 15 11:15:04 BST 2005


Hi Jacek,

I am using the same compilers as you on Itanium hardware, but not on an
SGI Altix system. There is some special code in armci/src/copy.h for
Altix systems only, and hence I do not run into this fastbcopy problem.

If you really do not have access to fastbcopy, then you could disable
the following lines in armci/src/copy.h and then build the TCGMSG
version as I suggested earlier:

#if defined(LINUX64) && defined(SGIALTIX)
/* fastbcopy from Wayne Vieira and Gerardo Cisneros */
#define MEMCPY
#define armci_copy(src, dst, len) _fastbcopy(src, dst, len)
#define memcpy(dst, src, len)  _fastbcopy(src, dst, len)
#define bcopy(src, dst, len) _fastbcopy(src, dst, len)
#endif

e.g. delete the lines above and then do

make TARGET=LINUX64 CC=icc FC=ifort

Cheers,
Ross

> -----Original Message-----
> From: Jacek Klos [mailto:jakl at tiger.chem.uw.edu.pl]
> Sent: 14 September 2005 17:59
> To: molpro-user at molpro.net
> Cc: Ross Nobes
> Subject: RE: [molpro-user] Parallel Molpro 2002.6 on ALTIX 350
> 
> Hi Ross,
> I tried this gmake options as you wrote and I get the following error
> during gmake of GA:
> IPO Error: unresolved : _fastbcopy
>          Referenced in libarmci.a(armci.o)
>          Referenced in libarmci.a(message.o)
>          Referenced in libarmci.a(mutex.o)
>          Referenced in libarmci.a(strided.o)
>          Referenced in libarmci.a(vector.o)
>          Referenced in libarmci.a(rmw.o)
>          Referenced in libarmci.a(memlock.o)
>          Referenced in libarmci.a(request.o)
> ../../lib/LINUX64/libarmci.a(strided.o)(.text+0x24b2): In function
> `armci_copy_2D':
> : undefined reference to `_fastbcopy'
> ../../lib/LINUX64/libarmci.a(strided.o)(.text+0x2532): In function
> `armci_copy_2D':
> : undefined reference to `_fastbcopy'
> ../../lib/LINUX64/libarmci.a(strided.o)(.text+0x2572): In function
> `armci_copy_2D':
> : undefined reference to `_fastbcopy'
> ../../lib/LINUX64/libarmci.a(strided.o)(.text+0x2632): In function
> `armci_copy_2D':
> : undefined reference to `_fastbcopy'
> ../../lib/LINUX64/libarmci.a(strided.o)(.text+0x2692): In function
> `armci_copy_2D':
> : undefined reference to `_fastbcopy'
> ../../lib/LINUX64/libarmci.a(strided.o)(.text+0x2712): more undefined
> references to `_fastbcopy' follow
> gmake[2]: *** [test.x] Error 1
> rm util.o ffflush.o test.o
> gmake[2]: Leaving directory `/opt/GlobalArray3.3.3/g/global/testing'
> gmake[1]: *** [test.x] Error 1
> gmake[1]: Leaving directory `/opt/GlobalArray3.3.3/g/global'
> gmake: *** [test] Error 2
> 
> Did you run into this error?
> Thank you
> Jacek
> 
> Ross wrote------
> Hi Jacek,
> 
> If you want to use the TCGMSG-based version of the GA, you will need
to
> build with
> 
> make TARGET=LINUX64 CC=icc FC=ifort
> 
> Cheers,
> Ross
> 
> -----Original Message-----
> From: owner-molpro-user at molpro1.chem.cf.ac.uk [mailto:owner-molpro-
> user at molpro1.chem.cf.ac.uk] On Behalf Of Jacek Klos
> Sent: 14 September 2005 17:28
> To: molpro-user at molpro.net
> Subject: [molpro-user] Parallel Molpro 2002.6 on ALTIX 350
> 
> Dear Molpro users,
> I am trying to install parallel version of molpro 2002.6
> on SGI ALTIX 350 Itanium2 system with NUMAlink interconnect.
> I use ifort 9 and icc 9 Intel compiler. I succesfully compiled
> and linked serial version of Molpro, but I run into problems
> with parallel version.
> I am seeking advise on how  to properly compile Global Arrays
> and link it with Molpro installation. If I compile GA
> using TARGET=LINUX it compiles, but molpro linker fails
> with a lot of unresolved symbols, for example:
> IPO Error: unresolved : pbeginf_
>           Referenced in libmolpro.a(mpputil.o)
> IPO Error: unresolved : mitob_
>           Referenced in libmolpro.a(mpputil.o)
>           Referenced in libmolpro.a(globalop.o)
> IPO Error: unresolved : pend_
>           Referenced in libmolpro.a(mpputil.o)
> IPO Error: unresolved : nodeid_
>           Referenced in libmolpro.a(mpputil.o)
> IPO Error: unresolved : nnodes_
>           Referenced in libmolpro.a(mpputil.o)
> IPO Error: unresolved : mdtob_
>           Referenced in libmolpro.a(mpputil.o)
>           Referenced in libmolpro.a(globalop.o)
>           Referenced in libmolpro.a(getbs.o)
> 
> /opt/GlobalArray3.3.3/g/lib/LINUX/libarmci.a(fence.o)(.text+0xaa1): In
> function `ARMCI_Barrier':
> : undefined reference to `SYNCH_'
> /opt/GlobalArray3.3.3/g/lib/LINUX/libarmci.a(message.o)(.text+0x15a2):
> In
> function `armci_msg_barrier':
> _sbuf/message.c:377: undefined reference to `SYNCH_'
> /opt/GlobalArray3.3.3/g/lib/LINUX/libarmci.a(message.o)(.text+0x15f2):
> In
> function `armci_msg_me':
> _sbuf/message.c:391: undefined reference to `NODEID_'
> /opt/GlobalArray3.3.3/g/lib/LINUX/libarmci.a(message.o)(.text+0x1652):
> In
> function `armci_msg_nproc':
> _sbuf/message.c:405: undefined reference to `NNODES_'
> /opt/GlobalArray3.3.3/g/lib/LINUX/libarmci.a(message.o)(.text+0x16b2):
> In
> function `armci_timer':
> _sbuf/message.c:419: undefined reference to `tcgtime_'
> /opt/GlobalArray3.3.3/g/lib/LINUX/libarmci.a(message.o)(.text+0x1752):
> In
> function `armci_msg_abort':
> 
> and so on.
> I think my GA installation is not full and I will be glad to hear from
> experts in this matter.
> Thank you in advance
> Jacek
> 
> Jacek A. Klos
> ---------------------------------
> Quantum Chemistry Laboratory     |
> Department of Chemistry          |
> University of Warsaw             |
> Pasteura 1 Street                |
> 02-093 Warsaw                    |
> Poland                           |
> (+4822)8220211 ext. 213          |
> e-mail: jakl at tiger.chem.uw.edu.pl|
> ---------------------------------
> 
> 
> Jacek A. Klos
> ---------------------------------
> Quantum Chemistry Laboratory     |
> Department of Chemistry          |
> University of Warsaw             |
> Pasteura 1 Street                |
> 02-093 Warsaw                    |
> Poland                           |
> (+4822)8220211 ext. 213          |
> e-mail: jakl at tiger.chem.uw.edu.pl|
> ---------------------------------




More information about the Molpro-user mailing list