Prev: mmreader gives incorrect intensity values
Next: Access Neuroscan via TCP
From: Matt on 5 Aug 2010 17:30
Dear all,

I have written some mexed routines for speeding up my code in Windows, and everything compiles fine and works beautifully in Windows, but when moving my code to Linux I'm having trouble with lapack routines. To simplify things I'm using the MATLAB example file matrixDivide.c [from http://www.mathworks.com/access/helpdesk/help/techdoc/matlab_external/br_2m24-1.html].

Compiling and running the test example proceeds as follows:

>> mex -v matrixDivide.c -llapack

-> mexopts.sh sourced from directory (DIR = $HOME/.matlab/$REL_VERSION)
FILE = /home/mpuecker/.matlab/R2010a/mexopts.sh
----------------------------------------------------------------
-> MATLAB = /share/apps/matlabr2010a
-> CC = gcc
-> CC flags:
CFLAGS = -ansi -D_GNU_SOURCE -fexceptions -fPIC -fno-omit-frame-pointer -pthread
CDEBUGFLAGS = -g
COPTIMFLAGS = -O -DNDEBUG
CLIBS = -Wl,-rpath-link,/share/apps/matlabr2010a/bin/glnxa64 -L/share/apps/matlabr2010a/bin/glnxa64 -lmx -lmex -lmat -lm -lstdc++
arguments = -DMX_COMPAT_32
-> CXX = g++
-> CXX flags:
CXXFLAGS = -ansi -D_GNU_SOURCE -fPIC -fno-omit-frame-pointer -pthread
CXXDEBUGFLAGS = -g
CXXOPTIMFLAGS = -O -DNDEBUG
CXXLIBS = -Wl,-rpath-link,/share/apps/matlabr2010a/bin/glnxa64 -L/share/apps/matlabr2010a/bin/glnxa64 -lmx -lmex -lmat -lm
arguments = -DMX_COMPAT_32
-> FC = g95
-> FC flags:
FFLAGS = -fexceptions -fPIC -fno-omit-frame-pointer
FDEBUGFLAGS = -g
FOPTIMFLAGS = -O
FLIBS = -Wl,-rpath-link,/share/apps/matlabr2010a/bin/glnxa64 -L/share/apps/matlabr2010a/bin/glnxa64 -lmx -lmex -lmat -lm
arguments = -DMX_COMPAT_32
-> LD = gcc
-> Link flags:
LDFLAGS = -pthread -shared -Wl,--version-script,/share/apps/matlabr2010a/extern/lib/glnxa64/mexFunction.map -Wl,--no-undefined
LDDEBUGFLAGS = -g
LDOPTIMFLAGS = -O
LDEXTENSION = .mexa64
arguments = -llapack
-> LDCXX =
-> Link flags:
LDCXXFLAGS =
LDCXXDEBUGFLAGS =
LDCXXOPTIMFLAGS =
LDCXXEXTENSION =
arguments = -llapack
----------------------------------------------------------------

-> gcc -c -I/share/apps/matlabr2010a/extern/include -I/share/apps/matlabr2010a/simulink/include -DMATLAB_MEX_FILE -ansi -D_GNU_SOURCE -fexceptions -fPIC -fno-omit-frame-pointer -pthread -DMX_COMPAT_32 -O -DNDEBUG "matrixDivide.c"

In file included from matrixDivide.c:21:
/share/apps/matlabr2010a/extern/include/lapack.h:6498:1: warning: "dgesv" redefined
matrixDivide.c:17:1: warning: this is the location of the previous definition
matrixDivide.c: In function 'mexFunction':
matrixDivide.c:64: warning: incompatible implicit declaration of built-in function 'memcpy'
matrixDivide.c:74: warning: passing argument 1 of 'dgesv_' from incompatible pointer type
matrixDivide.c:74: warning: passing argument 2 of 'dgesv_' from incompatible pointer type
matrixDivide.c:74: warning: passing argument 4 of 'dgesv_' from incompatible pointer type
matrixDivide.c:74: warning: passing argument 5 of 'dgesv_' from incompatible pointer type
matrixDivide.c:74: warning: passing argument 7 of 'dgesv_' from incompatible pointer type
matrixDivide.c:74: warning: passing argument 8 of 'dgesv_' from incompatible pointer type
-> gcc -O -pthread -shared -Wl,--version-script,/share/apps/matlabr2010a/extern/lib/glnxa64/mexFunction.map -Wl,--no-undefined -o "matrixDivide.mexa64" matrixDivide.o -llapack -Wl,-rpath-link,/share/apps/matlabr2010a/bin/glnxa64 -L/share/apps/matlabr2010a/bin/glnxa64 -lmx -lmex -lmat -lm -lstdc++

>> A = [1 2; 3 4];
>> B = [5; 6];
>> X = matrixDivide(A,B)
** On entry to DGESV parameter number 1 had an illegal value

At which point MATLAB quits. A few notes:
- Sometimes I get parameter number 3, or number 7 instead of number 1 as the error message
- I have also gotten a segmentation fault as the error, but have not been able to reproduce that lately
- The exact same code compiles/runs flawlessly on Windows with microsoft visual express compiler
- I've also tried using the pathscale compiler pathcc instead with no luck
- You'll notice I compiled with -llapack instead of -lmwlapack
+ This is because we have a 64 bit linux system.
+ When getting blas to work I found compiling with -lblas instead of -lmwblas would avoid segmentation faults
+ I've also tried compiling against -lacml, but I get a similar error

At this point I'm stumped. Any help would be great! [ps. I'm a newb when it comes to linking to precompiled libraries etc.]

THANKS!
 | 
Pages: 1
Prev: mmreader gives incorrect intensity values
Next: Access Neuroscan via TCP