Re: Re; sparse merge

[John W. Eaton]

On 28-Feb-2005, Dmitri A. Sergatskov <address@hidden> wrote:

| David Bateman wrote:
| >> Built from CVS 2005-02-28 on iMac G4/1.25GHz with OS X 10.3.8 using 
| >> gcc 3.3 20030304 (Apple Computer, Inc. build 1671) and XLF fortran 
| >> compiler (setting F77=xlf and FFLAGS=-qextname). Configured with 
| >> --enable-shared --enable-dl --disable-static.
| >> Passes all tests.
| > 
| > 
| > By all tests do you mean "make check"? As there are no sparse tests 
| > there yet, thats
| > not that complete. I'll forward my test code off-line...
| > 
| 
| You both lucky to be able to compile it. :)
| Look what I get on Solaris:
| 
| On 27-Feb-2005, Dmitri A. Sergatskov <address@hidden> wrote:
| 
| | I tried to rebuild with built-in blas/lapack. Still the same problem:
| |
| | g++  -I. -I.. -I../liboctave -I../src -I../libcruft/misc  -I../glob 
-I../glob -DHAVE_CONFIG_H  -Wall -W -Wshadow -O3 -mcpu=ultrasparc 
-mtune=ultrasparc -pipe -rdynamic \
| | -L..  -fPIC  -o octave \
| | main.o  \
| | -L../liboctave -L../libcruft -L../src -R /usr/local/lib/octave-2.1.cvs \
| | -loctinterp -loctave  -lcruft  ../glob/glob.o ../glob/fnmatch.o  \
| |    -lfftw3 -lreadline  -lncurses -ldl -lm  -L/usr/ccs/lib -L/usr/lib 
-L/usr/local/lib/gcc/sparc-sun-solaris2.8/3.4.3
| | 
-L/usr/local/lib/gcc/sparc-sun-solaris2.8/3.4.3/../../../../sparc-sun-solaris2.8/lib
 -L/usr/ccs/bin -L/usr/local/lib/gcc/sparc-sun-solaris2.8/3.4.3/../../.. 
-lfrtbegin
| | -lg2c -lm -lgcc_s
| | g++: unrecognized option `-rdynamic'
| | Undefined                       first referenced
| |   symbol                             in file
| | dger                                ../liboctave/liboctave.so
| | dgemm                               ../liboctave/liboctave.so
| | dgemv                               ../liboctave/liboctave.so
| | dscal                               ../liboctave/liboctave.so
| | dtrsm                               ../liboctave/liboctave.so
| | dtrsv                               ../liboctave/liboctave.so
| | zgemm                               ../liboctave/liboctave.so
| | zgemv                               ../liboctave/liboctave.so
| | zgeru                               ../liboctave/liboctave.so
| | ztrsm                               ../liboctave/liboctave.so
| | ztrsv                               ../liboctave/liboctave.so
| | ld: fatal: Symbol referencing errors. No output written to octave
| | collect2: ld returned 1 exit status
| | make[2]: *** [octave] Error 1
| | make[2]: Leaving directory `/espresso/dima/src/sun/octave/src'
| | make[1]: *** [src] Error 2
| | make[1]: Leaving directory `/espresso/dima/src/sun/octave'
| | make: *** [all] Error 2
| |
| |
| | Blas is definitely being built -- there are *.o files in
| | blas/pic...
| |
| | I wonder if this is underscore handling problem. I have:
| |   address@hidden pic]$ nm dgemm.o
| | 00000000 r .LLC0
| | 00000008 r .LLC1
| | 00000018 r .LLC2
| | 00000010 r .LLC3
| | 00000020 r .LLC4
| | 00000028 r .LLC5
| |           U _GLOBAL_OFFSET_TABLE_
| | 00000020 T dgemm_
| |           U lsame_
| |           U xerbla_
| |
| | So this is dgemm vs dgemm_
| 
| (the same problem if I use an external blas/atlas)

Did you see my earlier message (not to the list) about this?  I think
I said that it looks like UMFPACK uses a lame method for dealing with
calls to BLAS.  Now I think it must be expecting to link with -lcblas
or some other blas library.  In any case, I don't have time to deal
with this problem right now.

jwe