Charm++ parallel programming system

[Nikhil Jain <nikhil.jain AT acm.org>]

Hi Jingjie,

On most BG/Q system,  the following command should list the content of the folder we are interested in:

ls -lat /bgsys/drivers/

The ppcfloor folder in this directory is the default target used in all Charm++ (and many other software) for compiling on BG/Q. It is typically symlinked to one of the driver releases. An example output from a system I have access to looks like this:

ls -lat /bgsys/drivers/
total 32
drwxr-xr-x 13 root root 4096 Apr  6 02:33 toolchain
drwxr-xr-x  8 root root 4096 Apr  1 16:02 .
lrwxrwxrwx  1 root root   27 Apr  1 16:02 ppcfloor -> /bgsys/drivers/V1R2M4/ppc64
drwxr-xr-x  3 root root 4096 Feb 24 01:39 V1R2M4
drwxr-xr-x  3 root root 4096 Jul 29  2015 V1R2M2
drwxr-xr-x  3 root root 4096 Jul  7  2015 V1R2M3
drwxr-xr-x 18 root root 4096 Jun 30  2014 ..
drwxr-xr-x  3 root root 4096 Jun 23  2014 V1R2M1
drwxr-xr-x  3 root root 4096 Aug 12  2013 V1R2M0

So that system is currently using V1R2M4.

In any case, the related revert will not affect NAMD's compilation or basic runtime behavior.

--Nikhil

Jingjie Yeo wrote:
Hi Nikhil,

Thank you very much for your prompt response. May I know what information about the system can I provide you with? I’m currently a user of the system, not the administrator, so I am unfamiliar with the setup. I’m trying to compile on the BGQ in Singapore’s ACRC.

May I know whether reverting this commit will affect the compilation of the latest version NAMD (if you are familiar with it)?

Best Regards,

JJ

On 2 May 2016, at 2:45 PM, Nikhil Jain <nikhil.jain AT acm.org> wrote:

Hello Jingjie,
Kernel_GetMapping function was released as part of BG/Q system driver version V1R2M1 more than an year ago. It seems like the Blue Gene/Q system you are on was not updated for this release. Can you tell us which BG/Q system are you running on?

More history on this issue can be found here:
https://charm.cs.illinois.edu/redmine/issues/374#change-5023
The commit that enabled these changes can be viewed here:
https://charm.cs.illinois.edu/gerrit/#/c/422/
Easiest way to compile Charm++ if this is the only issue would be by reverting this commit:
git revert e223cf64a5f1c08b00e17d3b8dfd938062fbbbba

--Nikhil

Jingjie Yeo

May 2, 2016 at 09:48

Dear Admin,

I am having some issues compiling on Bluegene/Q, I’ve tried compiling with:
./build charm++ pamilrts-bluegeneq xlc -j16 --with-production
But I ended up with the error:
"lz4.c", line 157.16: 1506-508 (W) Option push for pragma pack is not supported.
"lz4.c", line 157.1: 1506-224 (W) Incorrect pragma ignored.
"lz4.c", line 165.16: 1506-726 (W) Attempting to pop an empty pack stack. Current pack may change.
"BGQTorus.C", line 57.3: 1540-0274 (S) The name lookup for "Kernel_GetMapping" did not find a declaration.
../bin/charmc -optimize -production -c pup_f.f90 && ../bin/charmc -optimize -production -cpmod ../include pupmod.M || touch pup_f.o
Fatal Error by charmc in directory /home/ihpcjyeo/scratch/NAMD_2.11_Source/charm-6.7.0/pamilrts-bluegeneq-xlc/tmp Command bgxlC_r -qhalt=e -qnokeyword=__int128 -qtls=local-exec -I../bin/../include -I/bgsys/drivers/ppcfloor/comm/sys-fast/include -I/bgsys/drivers/ppcfloor/spi/include -I/bgsys/drivers/ppcfloor -I/bgsys/drivers/ppcfloor/spi/include/kernel/cnk -I/soft/libraries/alcf/current/xl/ZLIB//include -D__CHARMC__=1 -I. -O3 -U_FORTIFY_SOURCE -qlanglvl=extended0x -c BGQTorus.C -o BGQTorus.o returned error code 1
charmc exiting...
../bin/charmc -optimize -production -o ../lib/libldb-rand.a cldb.rand.c gmake: *** [BGQTorus.o] Error 1
gmake: *** Waiting for unfinished jobs....
1517-020: (E) Source file is empty.
** pup_f.f90 === End of Compilation 1 ===
"cmimemcpy_qpx.c", line 104.22: 1506-280 (W) Function argument assignment between types "char*" and "const char*" is not allowed. "pup_f.f90", 1500-030 (I) INFORMATION: (block data)__1: Additional optimization may be attained by recompiling and specifying MAXMEM option with a value greater than 0.
1501-510 Compilation successful for file pup_f.f90.
cp: cannot stat `PUPMOD': No such file or directory
Fatal Error by charmc in directory /home/ihpcjyeo/scratch/NAMD_2.11_Source/charm-6.7.0/pamilrts-bluegeneq-xlc/tmp Command cp -p PUPMOD ../include/ returned error code 1
charmc exiting...
/bgsys/drivers/ppcfloor/gnu-linux/bin/powerpc64-bgq-linux-ar: creating ../lib/libldb-rand.a
../bin/charmc -optimize -production -o ../lib/libconv-cplus-y.a machine.o
/bgsys/drivers/ppcfloor/gnu-linux/bin/powerpc64-bgq-linux-ar: creating ../lib/libconv-cplus-y.a
../bin/charmc -optimize -production -I. -o ../lib/libconv-cplus-n.a -DFOR_CPLUS=0 machine.c
1500-036: (I) The NOSTRICT option (default at OPT(3)) has the potential to alter the semantics of a program. Please refer to documentation on the STRICT/NOSTRICT option for more information.
/bgsys/drivers/ppcfloor/gnu-linux/bin/powerpc64-bgq-linux-ar: creating ../lib/libconv-cplus-n.a
1500-036: (I) The NOSTRICT option (default at OPT(3)) has the potential to alter the semantics of a program. Please refer to documentation on the STRICT/NOSTRICT option for more information.
1500-036: (I) The NOSTRICT option (default at OPT(3)) has the potential to alter the semantics of a program. Please refer to documentation on the STRICT/NOSTRICT option for more information.

May I know how can I resolve this issue? Thank you very much for your kind assistance.

Best Regards,
Jingjie

--
Nikhil Jain, Ph.D.
Postdoctoral Fellow, Lawrence Livermore National Laboratory
nikhil.jain AT acm.org, http://nikhil-jain.github.io/

--
Nikhil Jain, Ph.D.
Postdoctoral Fellow, Lawrence Livermore National Laboratory
nikhil.jain AT acm.org, http://nikhil-jain.github.io/