octave-maintainers
[Top][All Lists]
Advanced

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

Re: Octave 5.2.0 release


From: Dmitri A. Sergatskov
Subject: Re: Octave 5.2.0 release
Date: Fri, 17 Jan 2020 12:44:14 -0500



On Fri, Jan 17, 2020 at 12:27 PM John W. Eaton <address@hidden> wrote:
On 1/17/20 10:25 AM, Dmitri A. Sergatskov wrote:
>
>
> On Fri, Jan 17, 2020 at 11:08 AM John W. Eaton <address@hidden
> <mailto:address@hidden>> wrote:
>
>     On 1/16/20 9:26 PM, Nicholas Jankowski wrote:
>
>      > side question: now that w32 is fading into antiquity, what are the
>      > drawbacks to having w64-64 just become the default build for w64?
>
>     As stated previously, the default is to build with 64-bit indexing on
>     64-bit systems.  It is only the BLAS and other Fortran libraries that
>     are an issue.  We provide Windows binaries with 32-bit and 64-bit BLAS
>     because we are building all the dependencies for those binary
>     distributions.  But for Linux, the default system BLAS libraries are
>     still using 32-bit integers.  If those change, then the Linux
>     distributions could start to distribute fully 64-bit aware versions of
>     Octave as well.
>
>
> Fedora ships openblas with 64-bit integers, e.g.:
> <<<<
> rpm -qi openblas-threads64-0.3.7-1.fc31.x86_64
> Name        : openblas-threads64
> Version     : 0.3.7
> Release     : 1.fc31
> Architecture: x86_64
> <...deleted...>
> Summary     : An optimized BLAS library based on GotoBLAS2, pthreads version
> Description :
>
> OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD
> version. The project is supported by the Lab of Parallel Software and
> Computational Science, ISCAS. http://www.rdcps.ac.cn
>
> This package contains the library compiled with threading support and
> 64-bit integer interface.
>
>  >>>>
>
> I could not figure out how to use it though. Running configure
> --with-blas=/usr/lib64/libopenblasp64.so.0
> results in:
>
> <<<<
> checking BLAS library integer size... 8
> checking for sgemm_ in /usr/lib64/libopenblasp64.so.0... yes
> checking whether LSAME is called correctly from Fortran... yes
> checking whether ISAMAX is called correctly from Fortran... no
> checking whether SDOT is called correctly from Fortran... no
> checking whether DDOT is called correctly from Fortran... yes
> checking whether CDOTU is called correctly from Fortran... no
> checking whether ZDOTU is called correctly from Fortran... no
> checking BLAS library integer size... 8
> configure: error: BLAS and LAPACK libraries are required
>  >>>>

You need to specify the fortran compiler option to make the default
integer size 8 bytes:

   configure F77_INTEGER_8_FLAG=-fdefault-integer-8 ...

The last time we looked at this problem, I recall that we decided it
wasn't worth the effort to attempt to detect both the option to use for
the Fortran compiler and the size of the integers used for the BLAS library.

Also, just having the BLAS library using 64-bit integers isn't enough.
You will also need to have lapack, suitesparse, qrupdate, and arpack
built to use 64-bit integers.

Thanks, this works. I understand about other libraries.
(BTW, isn't openblas provides lapack libraries as well?)

jwe


Dmitri.

reply via email to

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