Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] OpenMPI 1.4.2 and PathScale 3.2 C++
From: Paul H. Hargrove (PHHargrove_at_[hidden])
Date: 2010-08-28 00:48:48


I have tried again w/ flags nearly identical to Larry's:

$ [path_to]/configure --with-contrib-vt-flags=--with-platform=linux
--with-sge \
    CC="pathcc-3.2 -m64" CFLAGS="-g -O3 -march=anyx86" \
    CXX="pathCC-3.2 -m64" CXXFLAGS="-g -O1 -march=anyx86" \
    FC="pathf90-3.2 -m64" FCFLAGS="-g -O3 -march=anyx86
-fno-second-underscore" \
    F77="pathf90-3.2 -m64" FFLAGS="-g -O3 -march=anyx86
-fno-second-underscore"

And got success with both 1.4.3rc1 and 1.5rc5, due I assume to the
explict -O1 in CXXFLAGS.
However, if I change that to an explicit -O3 I SILL have no problems.

So, at this point I suspect the difference is between the gcc3 and gcc4
modes of pathcc.
However, my attempts to configure with CC="pathcc-3.2 -m64 -gnu3" fails
configure with

checking for functional offsetof macro... no
configure: WARNING: Your compiler does not support offsetof macro
configure: error: Configure: Cannot continue

So, I suspect the gcc3 mode is just not available/usable on my installation.

In short: I cannot repoduce Larry's failure, but that is probably
because I also cannot reproduce his compiler.

-Paul

Paul H. Hargrove wrote:
> I have built both 1.4.3rc1 and 1.5rc5 using the 3.2 PathScale compilers.
>
> I did not encounter the problem Larry reported, but there are (at
> least) 2 differences between my build and Larry's
> 1) I didn't pass and explicit {C,CXX,F,FC}FLAGS
> 2) My pathcc install is using the gcc4 toolchain, not gcc3.
> I plan a few more builds to narrow down the differences.
>
> Some notes:
>
> To build 1.5rc5 with this compiler I needed a 1-line change
> (http://www.open-mpi.org/community/lists/devel/2010/08/8413.php)
>
> My configure args for both 1.4.3rc1 and 1.5rc5:
>
> $ [path_to]/configure \
> configure --with-contrib-vt-flags=--with-platform=linux \
> CC=pathcc-3.2 CXX=pathCC-3.2 \
> F77=pathf90-3.2 FC=pathf90-3.2
>
> Note that the "--with-contrib-vt-flags=--with-platform=linux" is
> because the machine is a front-end to a Cray XT and I wanted to build
> VT for the front-end, not for the Cray.
>
>
> Paul H. Hargrove wrote:
>> Among the tests I conducted but did not report was a successful build
>> of 1.5rc5 with the PathScale "3.2.99" compilers.
>> $ pathcc --version
>> PathScale(TM) Compiler Suite: Version 3.2.99
>> Built on: 2009-08-21 13:23:57 -0500
>> Thread model: posix
>> GNU gcc version 4.2.0 (PathScale 3.2.99 driver)
>>
>> Copyright 2000, 2001 Silicon Graphics, Inc. All Rights Reserved.
>> Copyright 2002, 2003, 2004, 2005, 2006 PathScale, Inc. All Rights
>> Reserved.
>> Copyright 2006, 2007 QLogic Corporation. All Rights Reserved.
>> Copyright 2007, 2008 PathScale LLC. All Rights Reserved.
>> See complete copyright, patent and legal notices in the
>> /opt/pathscale/share/doc/pathscale-compilers-3.2.99/LEGAL.pdf file.
>>
>> This was a beta release of the 3.3 (never released?) provided to some
>> customers, and not a general release.
>>
>> I also note that my PathScale install is using gcc-4, rather than
>> gcc-3, which I understand contains different bugs.
>>
>> I will test the "stock" 3.2 with 1.5rc5 and 1.4.3rc1 and report my
>> findings.
>>
>> -Paul
>>
>> Larry Baker wrote:
>>> The PathScale 3.2 C++ compiler segment faults for optimization
>>> levels higher than -O1 (-O2 is the default). This is for OpenMPI
>>> 1.4.2 — my first attempt to compile using the PathScale compilers.
>>> I could not find any -WOPT options to eliminate the error. I don't
>>> understand the current state of the PathScale compilers. I think
>>> the company changed hands since we bought the product and the last
>>> owner went bankrupt. I think the name has been resurrected by the
>>> people that wrote the compiler. They use the same name, but I get
>>> the impression that our license is with a company that is gone now,
>>> not with them, so they want us to buy a new license. Anyway, there
>>> may be a newer compiler than 3.2 that does not have this problem.
>>>
>>> [root_at_hydra vtfilter]# pathCC -v
>>> PathScale(TM) Compiler Suite: Version 3.2
>>> Built on: 2008-06-16 16:45:36 -0700
>>> Thread model: posix
>>> GNU gcc version 3.3.1 (PathScale 3.2 driver)
>>>
>>> [root_at_hydra vtfilter]# ./configure >configure.log 2>&1 \
>>> --prefix=$PATHSCALE_DIR/openmpi --with-sge \
>>> CC="pathcc -m64" \
>>> CFLAGS="-g -O3 -march=anyx86" \
>>> CXX="pathCC -m64" \
>>> CXXFLAGS="-g -O1 -march=anyx86" \
>>> FC="pathf90 -m64" \
>>> FCFLAGS="-g -O3 -march=anyx86 -fno-second-underscore" \
>>> F77="pathf90 -m64" \
>>> FFLAGS="-g -O3 -march=anyx86 -fno-second-underscore"
>>>
>>> [root_at_hydra vtfilter]# make
>>>
>>> <snip>
>>>
>>> Making all in vtfilter
>>> make[6]: Entering directory
>>> `/usr/local/src/openmpi-1.4.2/ompi/contrib/vt/vt/tools/vtfilter'
>>>
>>> <snip>
>>>
>>> pathCC -m64 -DHAVE_CONFIG_H -I. -I../.. -I../../extlib/otf/otflib
>>> -I../../extlib/otf/otflib -I../../vtlib/ -I../../vtlib
>>> -DINSIDE_OPENMPI -D_GNU_SOURCE -mp -DVT_OMP -g -O3 -march=anyx86
>>> -MT vtfilter-vt_tracefilter.o -MD -MP -MF
>>> .deps/vtfilter-vt_tracefilter.Tpo -c -o vtfilter-vt_tracefilter.o
>>> `test -f 'vt_tracefilter.cc' || echo './'`vt_tracefilter.cc
>>> Signal: Segmentation fault in Global Optimization -- Dead Store
>>> Elimination phase.
>>> Error: Signal Segmentation fault in phase Global Optimization --
>>> Dead Store Elimination -- processing aborted
>>> *** Internal stack backtrace:
>>> pathCC INTERNAL ERROR: /opt/pathscale/lib/3.2/be died due to signal 4
>>> Please report this problem to <support_at_[hidden]>.
>>> Problem report saved as /root/.ekopath-bugs/pathCC_error_weMqHF.ii
>>> Please review the above file and, if possible, attach it to your
>>> problem report.
>>> make[6]: *** [vtfilter-vt_tracefilter.o] Error 1
>>>
>>> <snip>
>>>
>>> Larry Baker
>>> US Geological Survey
>>> 650-329-5608
>>> baker_at_[hidden]
>>>
>>>
>>> _______________________________________________
>>> devel mailing list
>>> devel_at_[hidden]
>>> http://www.open-mpi.org/mailman/listinfo.cgi/devel
>>
>

-- 
Paul H. Hargrove                          PHHargrove_at_[hidden]
Future Technologies Group
HPC Research Department                   Tel: +1-510-495-2352
Lawrence Berkeley National Laboratory     Fax: +1-510-486-6900