Building sst with gem5 failed

32 views
Skip to first unread message

Michael Dimopoulos

unread,
May 25, 2012, 6:47:57 AM5/25/12
to Structural Simulation Toolkit
Hi.
I am trying to build sst with gem5 support but the compilation of sst
fails in the following step:

..................
CXX libm5C_la-debug.lo
debug.cpp: In function ‘void enableDebugFlags(std::string)’:
debug.cpp:63: error: ‘FlagsMap’ has not been declared
debug.cpp:63: error: expected ‘;’ before ‘i’
debug.cpp:64: error: ‘FlagsMap’ has not been declared
debug.cpp:64: error: expected ‘;’ before ‘end’
debug.cpp:65: error: ‘i’ was not declared in this scope
debug.cpp:65: error: ‘end’ was not declared in this scope
debug.cpp: At global scope:
debug.cpp:31: warning: ‘void enableDebugFlags(std::string)’ defined
but not used
make[4]: *** [libm5C_la-debug.lo] Error 1
.......................

I have successfully build all prerequisite packages.
The configuration I am using is:

sst version 2830
gcc v4.4.5.8
x86_64 Debian squeeze


and the options to configure sst are:
./configure --prefix=/mike/mdimop/local --with-boost=/home/mike/local
--with-zoltan=/home/mike/local --with-parmetis=/home/mike/local --with-
gem5=/home/mike/gem5-patched-v004/build/X86_SE --with-McPAT=/home/mike/
local/lib --with-hotspot=/home/mike/local/lib CFLAGS=-I/usr/include/
python2.6 CXXFLAGS=-I/usr/include/python2.6

Damien Couroussé

unread,
Jul 16, 2012, 3:52:19 AM7/16/12
to SST-si...@googlegroups.com
Hello!

I do experience the same build error on Debian 6.0.5 / amd64.
May I ask you how did you solve this problem ?

regards,

Simon Hammond

unread,
Jul 16, 2012, 10:11:13 AM7/16/12
to sst-si...@googlegroups.com
Hmm, this is interesting I am using Debian stable internally and we
don't see any of these errors.

Is this a completely clean SVN pull from the repository?

--
Si Hammond

Damien Couroussé

unread,
Jul 16, 2012, 11:23:43 AM7/16/12
to SST-si...@googlegroups.com, sst-si...@googlegroups.com
Yes,
I get this bug with revision 3040 and gem5-patched-v004

I did other compilations with revisions 2579 (with gem5-patched-v004) and 2564 (with gem5-patched-v003), and did not get the same problem.

The buguous code in debug.cpp looks as a copy of the same code in $gem5/debug.cc.
Hence if I delete this code from debug.cpp, the compilation succeeds.

It looks like a problem that is internal to SST's code.

Could it be a problem of configuration ?

BTW, I used as much as possible the development libraries included in Debian as long as the version numbers were OK, instead of recompiling from scratch as explained in the documentation.
I did only compile zoltan which is not included in Debian packages.
I am currently using the following Debian packages :
$ dpkg -l *dev | grep ^ii
ii  dpkg-dev                         1.15.8.12                    Debian package development tools
ii  libboost-dev                     1.42.0.1                     Boost C++ Libraries development files (default version)
ii  libboost-mpi-dev                 1.42.0.1                     C++ interface to the Message Passing Interface (MPI) (default version)
ii  libboost-mpi1.42-dev             1.42.0-4                     C++ interface to the Message Passing Interface (MPI)
ii  libboost-program-options-dev     1.42.0.1                     program options library for C++ (default version)
ii  libboost-program-options1.42-dev 1.42.0-4                     program options library for C++
ii  libboost-serialization1.42-dev   1.42.0-4                     serialization library for C++
ii  libboost1.42-dev                 1.42.0-4                     Boost C++ Libraries development files
ii  libc6-dev                        2.11.3-3                     Embedded GNU C Library: Development Libraries and Header Files
ii  libibverbs-dev                   1.1.3-2                      Development files for the libibverbs library
ii  libopenmpi-dev                   1.4.2-4                      high performance message passing library -- header files
ii  libparmetis-dev                  3.1.1-3                      Parallel Graph Partitioning and Sparse Matrix Ordering Libs: Devel
ii  libssl-dev                       0.9.8o-4squeeze12            SSL development libraries, header files and documentation
ii  libstdc++6-4.4-dev               4.4.5-8                      The GNU Standard C++ Library v3 (development files)
ii  linux-libc-dev                   2.6.32-45                    Linux support headers for userspace development
ii  manpages-dev                     3.27-1                       Manual pages about using GNU/Linux for development
ii  mpi-default-dev                  0.6                          Standard MPI development files
ii  python-dev                       2.6.6-3+squeeze7             header files and a static library for Python (default)
ii  python2.6-dev                    2.6.6-8+b1                   Header files and a static library for Python (v2.6)
ii  udev                             164-3                        /dev/ and hotplug management daemon
ii  zlib1g-dev                       1:1.2.3.4.dfsg-3             compression library - development

I can send a full package list if needed.

Damien


Le lundi 16 juillet 2012 16:11:13 UTC+2, Si Hammond a écrit :
Hmm, this is interesting I am using Debian stable internally and we
don't see any of these errors.

Is this a completely clean SVN pull from the repository?

--
Si Hammond


Simon Hammond

unread,
Jul 16, 2012, 11:45:45 AM7/16/12
to sst-si...@googlegroups.com
The M5 element (where the compile error happens) is an SST component
like you say.

I can't seem to replicate this error despite being on Debian stable
with pretty much the same configuration as you which is kind of odd so
it feels like it might be something about all of the code versions
lining up.

Instead of using the GEM5 v004 tar file can you get a mercurial clone
from (http://code.google.com/p/sst-gem5-devel/) and use this as your
GEM5 install. Our next release of SST will move over to use this code
base, its compilation is tested much more aggressively in our
overnight builds and has a ton of fixes in it.

We have the above running on lots of platforms including Debian,
Ubuntu and CentOS.


--
Si Hammond

Simon Hammond

unread,
Jul 16, 2012, 11:42:14 AM7/16/12
to sst-si...@googlegroups.com
--
Si Hammond


On 16 July 2012 09:23, Damien Couroussé <damien.c...@gmail.com> wrote:

Simon Hammond

unread,
Jul 16, 2012, 11:52:22 AM7/16/12
to sst-si...@googlegroups.com
Did you try Boost 1.43? You need to build it using the instructions on
the SST Wiki or else it will not build MPI.



--
Si Hammond

Damien Couroussé

unread,
Jul 24, 2012, 1:05:54 PM7/24/12
to Structural Simulation Toolkit
Hello Simon,

This configuration enables the code in #defined M5_WITH_POWER
sections, which are believe are no more compiled in the nightly
builds. It is true ?

At least for me, using the last versions of sst-trunk and sst-gem5
(devel branch), this code does not build, as I describe in this
thread : http://groups.google.com/group/SST-simulator/browse_frm/thread/b9ea05f0297058f.

Maybe I do not use the right configuration ?


regards,
Damien


On Jul 16, 5:42 pm, Simon Hammond <simon.hamm...@gmail.com> wrote:
> --
> Si Hammond
>
Reply all
Reply to author
Forward
0 new messages