[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

2.9.6 build fails on an Opteron RHEL4

From: John W. Eaton
Subject: 2.9.6 build fails on an Opteron RHEL4
Date: Mon, 12 Jun 2006 14:34:04 -0400

On 12-Jun-2006, Fredrik Lingvall wrote:

| My Octave 2.9.6 (and 2.9.5) build fails every time on the file.
| The machine is an 2 cpu dual-core machine running RedHat EL4.
| How should I configure octave to avoid this?

I don't know the answer to that question, but I have a few for you.

| My current setup:
| -I$HOME/work/OPTERON/usr/local/include -O3 -m64 -Wall -funroll-loops 
| -march=opteron -mtune=opteron -pthread

Why these options?  In particular, why -pthread?  That won't make
Octave thread safe.

| -L$HOME/work/OPTERON/usr/local/lib

Also, are you setting these in the environment?  If so, I would
recommend against that and instead suggest that you add them to the
configure command line:

  configure CFLAGS="..." CXXFLAGS="..." ...

That way, the options will be recorded in the config.status file.

| ./configure --prefix=$HOME/work/OPTERON/usr/local 
| --with-blas=$HOME/work/OPTERON/usr/local/lib/ 
| --with-lapack=$HOME/work/OPTERON/usr/local/lib/ --with-f77 
| --enable-64

Are you sure the --with-blas and --with-lapack options are doing what
you want?  I think the proper way to specify this is to just set
LDFLAGS (again, on the configure command line) so the linker knows
where to find the libraries.  Then I think the configure script will
find your custom libraries and you won't need to specify --with-lapack
or --with-blas at all.

| --enable-shared --enable-dl --with-umfpack --with-colamd 
| --with-ccolamd --with-cholmod --with-cxsparse --with-fftw

These are all defaults, so you don't need to specify them.

|   Fortran compiler:     f77 

Precisely what compiler is this?  When you specify --enable-64, Octave
is compiled to assume that Fortran integers are 8 bytes wide.  Does
this compiler do that automatically, or have you provided the
appropriate flags for that?  Also, are your custom BLAS and LAPACK
libraries compiled so that the integers are 8 bytes wide?

Finally, --enable-64 is still somewhat experimental, so unless you
really need to work with Arrays larger than 2GB, or you would like to
help improve Octave's 64-bit capabilities, I'd recommend against using
this configure option.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]