Help building ATLAST+LAPACK on Solaris 10

[smontanaro]

Is this the right place to ask for help getting ATLAS to build?  I didn't
see anything which looked like an atlas-users mailing list.  Hopefully I
haven't guessed wrong.

I'm trying to build LAPACK according to the ATLAS build instructions:

    http://math-atlas.sourceforge.net/atlas_install/atlas_install.html#SECTION00041000000000000000

It seems straightforward enough, and I came up with a reasonable make.inc
file:

    FORTRAN  = gfortran
    OPTS     = -fomit-frame-pointer -mfpmath=387 -O2 -falign-loops=4 -fPIC -m32
    DRVOPTS  = $(OPTS)
    NOOPT    = -O0 -fPIC -m32
    LOADER   = $(FORTRAN)
    LOADOPTS = $(OPTS)
    TIMER    = INT_ETIME

When I run make it barfed at one point which I interpreted to mean that it
couldn't find the BLAS routines.  I simply executed

    (cd BLAS/SRC ; make)

followed by

    make

After generating lots of compilation output I get a core dump during a
test.  Here's the tail end of the make output:

    gfortran -fomit-frame-pointer -mfpmath=387 -O2 -falign-loops=4 -fPIC -m32 -c ztzt01.f -o ztzt01.o
    gfortran -fomit-frame-pointer -mfpmath=387 -O2 -falign-loops=4 -fPIC -m32 -c ztzt02.f -o ztzt02.o
    gfortran -fomit-frame-pointer -mfpmath=387 -O2 -falign-loops=4 -fPIC -m32  aladhd.o alaerh.o alaesm.o alahd.o alareq.o alasum.o alasvm.o chkxer.o icopy.o ilaenv.o xlaenv.o xerbla.o dlaord.o  zchkaa.o zchkeq.o zchkgb.o zchkge.o zchkgt.o zchkhe.o zchkhp.o zchklq.o zchkpb.o zchkpo.o zchkpp.o zchkpt.o zchkq3.o zchkql.o zchkqp.o zchkqr.o zchkrq.o zchksp.o zchksy.o zchktb.o zchktp.o zchktr.o zchktz.o zdrvgb.o zdrvge.o zdrvgt.o zdrvhe.o zdrvhp.o zdrvls.o zdrvpb.o zdrvpo.o zdrvpp.o zdrvpt.o zdrvsp.o zdrvsy.o zerrge.o zerrgt.o zerrhe.o zerrlq.o zerrls.o zerrpo.o zerrql.o zerrqp.o zerrqr.o zerrrq.o zerrsy.o zerrtr.o zerrtz.o zerrvx.o zgbt01.o zgbt02.o zgbt05.o zgelqs.o zgeqls.o zgeqrs.o zgerqs.o zget01.o zget02.o zget03.o zget04.o zget07.o zgtt01.o zgtt02.o zgtt05.o zhet01.o zhpt01.o zlaipd.o zlaptm.o zlarhs.o zlatb4.o zlatsp.o zlatsy.o zlattb.o zlattp.o zlattr.o zlavhe.o zlavhp.o zlavsp.o zlavsy.o zlqt01.o zlqt02.o zlqt03.o zpbt01.o zpbt02.o zpbt05.o zpot01.o zpot02.o zpot03.o zpo
 t05.o zppt01.o zppt02.o zppt03.o zppt05.o zptt01.o zptt02.o zptt05.o zqlt01.o zqlt02.o zqlt03.o zqpt01.o zqrt01.o zqrt02.o zqrt03.o zqrt11.o zqrt12.o zqrt13.o zqrt14.o zqrt15.o zqrt16.o zqrt17.o zrqt01.o zrqt02.o zrqt03.o zrzt01.o zrzt02.o zsbmv.o  zspt01.o zspt02.o zspt03.o zsyt01.o zsyt02.o zsyt03.o ztbt02.o ztbt03.o ztbt05.o ztbt06.o ztpt01.o ztpt02.o ztpt03.o ztpt05.o ztpt06.o ztrt01.o ztrt02.o ztrt03.o ztrt05.o ztrt06.o ztzt01.o ztzt02.o dget06.o \
            ../../tmglib_SOL10.a ../../lapack_SOL10.a ../../blas_SOL10.a  -o ../xlintstz
    make[2]: Leaving directory `/home/tuba/skipm/src/lapack-3.1.1/TESTING/LIN'
    Testing COMPLEX16 LAPACK linear equation routines
    ./xlintstz < ztest.in > ztest.out 2>&1
    Segmentation Fault - core dumped
    make[1]: *** [ztest.out] Error 139
    make[1]: Leaving directory `/home/tuba/skipm/src/lapack-3.1.1/TESTING'
    make: *** [lapack_testing] Error 2

My environment is:

    Solaris 10 on Intel plus gcc/gfortran 4.2.2:
    % gcc -v
    Reading specs from /opt/app/g++lib6/gcc-4.2/lib/gcc/i386-pc-solaris2.10/4.2.2/specs
    Target: i386-pc-solaris2.10
    Configured with: ../configure --prefix=/opt/app/g++lib6/gcc-4.2 --enable-languages=c,c++,fortran,objc --disable-nls --with-included-gettext --with-gnu-as --with-as=/usr/sfw/bin/gas --with-target-tools=/usr/sfw/bin/ --with-gmp=/opt/app/nonc++/gmp-4.2 --with-mpfr=/opt/app/nonc++/mpfr-2.3
    Thread model: posix
    gcc version 4.2.2
    % gfortran -v
    Reading specs from /opt/app/g++lib6/gcc-4.2/lib/gcc/i386-pc-solaris2.10/4.2.2/specs
    Target: i386-pc-solaris2.10
    Configured with: ../configure --prefix=/opt/app/g++lib6/gcc-4.2 --enable-languages=c,c++,fortran,objc --disable-nls --with-included-gettext --with-gnu-as --with-as=/usr/sfw/bin/gas --with-target-tools=/usr/sfw/bin/ --with-gmp=/opt/app/nonc++/gmp-4.2 --with-mpfr=/opt/app/nonc++/mpfr-2.3
    Thread model: posix
    gcc version 4.2.2

I went ahead and slapped a -k onto the make command, but I'm probably just
sticking my head in the sand.  In fact, core dumping seems to be the test
programs' favorite thing to do:

    ...
    make[2]: Leaving directory `/home/tuba/skipm/src/lapack-3.1.1/TESTING/EIG'
    NEP: Testing Nonsymmetric Eigenvalue Problem routines
    ./xeigtstz < nep.in > znep.out 2>&1
    Segmentation Fault - core dumped
    make[1]: *** [znep.out] Error 139
    SEP: Testing Symmetric Eigenvalue Problem routines
    ./xeigtstz < sep.in > zsep.out 2>&1
    Segmentation Fault - core dumped
    make[1]: *** [zsep.out] Error 139
    SVD: Testing Singular Value Decomposition routines
    ./xeigtstz < svd.in > zsvd.out 2>&1
    Segmentation Fault - core dumped
    make[1]: *** [zsvd.out] Error 139
    ZEC: Testing COMPLEX16 Eigen Condition Routines
    ./xeigtstz < zec.in > zec.out 2>&1
    ^Cmake[1]: *** Deleting file `zec.out'
    Segmentation Fault - core dumped
    make[1]: *** [zec.out] Error 139
    make: *** [lapack_testing] Error 130
    ...

Here's the gdb backtrace from xlintstz:

    #0  0xfeeba2c3 in cexp () from /lib/libm.so.2
    #1  0x0813b9b3 in zlarnv_ ()
    #2  0x00000000 in ?? ()

Any suggestions appreciated.

--
Skip Montanaro - skip@... - http://www.webfast.com/~skip/
"Be different, express yourself like everyone else."
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel

[Dmitri A. Sergatskov]

On Tue, May 20, 2008 at 2:53 PM,  <skip@...> wrote:

>
>    Solaris 10 on Intel plus gcc/gfortran 4.2.2:

This is just a guess, but I would suspect the gcc version.
Try upgrading to gcc 4.2.3...

Sincerely,

Dmitri.
--

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel

[smontanaro]

    >> Solaris 10 on Intel plus gcc/gfortran 4.2.2:

    Dmitri> This is just a guess, but I would suspect the gcc version.  Try
    Dmitri> upgrading to gcc 4.2.3...

Thanks, that got me further along the path to Nirvana.  I still got one or
two test errors, but nothing like what I saw yesterday with gcc 4.2.2.

Now, about your "name": "List for developer discussion, NOT SUPPORT."  That
implies I should have asked my question elsewhere.  Where is that other
list?

--
Skip Montanaro - skip@... - http://www.webfast.com/~skip/
"Be different, express yourself like everyone else."

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel

[Clint Whaley]

Skip,

>Now, about your "name": "List for developer discussion, NOT SUPPORT."  That
>implies I should have asked my question elsewhere.  Where is that other
>list?

Directions for getting support are at:
   http://math-atlas.sourceforge.net/faq.html#help

Sound like you might also be interested in:
   http://math-atlas.sourceforge.net/errata.html#lptest

Essentially, you get some lapack failures even when using the reference BLAS.

Cheers,
Clint

**************************************************************************
** R. Clint Whaley, PhD ** Assist Prof, UTSA ** www.cs.utsa.edu/~whaley **
**************************************************************************

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel