Problem regarding mlpack installation

[Pankaj]

Hi,

I have been trying to install mlpack on Ubuntu but it hangs after a
while. My gcc version is 4.3.2. I am running Ubuntu inside VMware.

The last few lines generated while compiling lapack look something
like this -

/home/pankaj/ML/fastlib2/fastlib/data/dataset.h:76: instantiated
from here
/home/pankaj/ML/fastlib2/fastlib/data/dataset.h:78: warning:
dereferencing type-punned pointer will break strict-aliasing rules
... Making bin/i686_Linux_check_gcc4_-DDISABLE_DISK_MATRIX/
libexternal_home_pankaj_ML_spectral_clustering_spec__auto.a
ar: creating /home/pankaj/ML/fastlib2/bin/i686_Linux_check_gcc4_-
DDISABLE_DISK_MATRIX/
libexternal_home_pankaj_ML_spectral_clustering_spec__auto.a

I guess there isn't much you can infer from the above lines. Also I
see 'testslamch' taking most of my cpu at this point.

Thanks,
Pankaj

[Pankaj]

Also I tried compiling mlpack on some COC machine and it throws me a
bunch of linking errors like this -

/net/hc295/pchawla3/CDA/mlpack/fastlib2/bin_keep/
i686_Linux_COMMON_gcc_COMMON/liblapack.a(dhseqr.o)(.text+0x620): In
function `dhseqr_':
/net/hc295/pchawla3/CDA/mlpack/fastlib2/bin_keep/
i686_Linux_COMMON_gcc_COMMON/netlib_workspace2/lapack-3.2/SRC/dhseqr.f:
350: undefined reference to `_gfortran_concat_string'
/net/hc295/pchawla3/CDA/mlpack/fastlib2/bin_keep/
i686_Linux_COMMON_gcc_COMMON/liblapack.a(dormbr.o)(.text+0x34c): In
function `dormbr_':
/net/hc295/pchawla3/CDA/mlpack/fastlib2/bin_keep/
i686_Linux_COMMON_gcc_COMMON/netlib_workspace2/lapack-3.2/SRC/dormbr.f:
183: undefined reference to `_gfortran_concat_string'
/net/hc295/pchawla3/CDA/mlpack/fastlib2/bin_keep/
i686_Linux_COMMON_gcc_COMMON/liblapack.a(dormbr.o)(.text+0x3e3):/net/
hc295/pchawla3/CDA/mlpack/fastlib2/bin_keep/
i686_Linux_COMMON_gcc_COMMON/netlib_workspace2/lapack-3.2/SRC/dormbr.f:
186: undefined reference to `_gfortran_concat_string'

[Nishant]

I ran into this problem on my laptop as well. The problem lies in
testdlamch and testslamch (used by lapack for some architecture
specific profiling) going into finite loops when they are compiling
with -O3 optimization (which is the default optimization level we
use). The fix is to switch those from -O3 to -Os. Here is how:

in .../fastlib/la/build.py, make the following changes

on line 202: change -O3 to -Os
on line 203: change -O3 to -Os


The infinite loop problem was resolved for me after doing the above
changes. This is of course a temporary fix until we change the -O
level

-Nishant

[byungki]

Hey, I had the same situation. As far as I figured out, it's because
'testslamch' didn't spit out any results. So while lapack was
installed, it was just locked.
I don't remember exactly how I fix that, but try to compile things in /
INSTALL and see if 'testslamch' works fine first. If it works fine,
it'll give you some of the numerical information about your machine
like precision, etc.

On Dec 6, 11:12 pm, Pankaj <pankaj.cha...@gmail.com> wrote:

[Pankaj]

Thanks, changing the optimization level solved the problem...