Error compiling openblas

57 views
Skip to first unread message

mmarco

unread,
Dec 4, 2019, 11:09:56 AM12/4/19
to sage-devel
I am trying to setup several virtual machines to run patchbots (for different distros). I am getting the following error when trying to compile Sage from source:

n not be used when making a shared object; recompile with -fPIC
/usr/lib64/gcc/x86_64-suse-linux/7/../../../../x86_64-suse-linux/bin/ld: ../libopenblas_atom-r0.3.6.a(zsymv_U.o): relocation R_X86_64_32S against undefined symbol `PREFETCHSIZE' ca
n not be used when making a shared object; recompile with -fPIC
/usr/lib64/gcc/x86_64-suse-linux/7/../../../../x86_64-suse-linux/bin/ld: ../libopenblas_atom-r0.3.6.a(zsymv_L.o): relocation R_X86_64_32S against undefined symbol `PREFETCHSIZE' ca
n not be used when making a shared object; recompile with -fPIC
/usr/lib64/gcc/x86_64-suse-linux/7/../../../../x86_64-suse-linux/bin/ld: ../libopenblas_atom-r0.3.6.a(zhemv_U.o): relocation R_X86_64_32S against undefined symbol `PREFETCHSIZE' ca
n not be used when making a shared object; recompile with -fPIC
/usr/lib64/gcc/x86_64-suse-linux/7/../../../../x86_64-suse-linux/bin/ld: ../libopenblas_atom-r0.3.6.a(zhemv_L.o): relocation R_X86_64_32S against undefined symbol `PREFETCHSIZE' ca
n not be used when making a shared object; recompile with -fPIC
/usr/lib64/gcc/x86_64-suse-linux/7/../../../../x86_64-suse-linux/bin/ld: final link failed: nonrepresentable section on output
collect2: error: ld returned 1 exit status
make[5]: *** [Makefile:153: ../libopenblas_atom-r0.3.6.so] Error 1
make[5]: Leaving directory '/home/patchbot/sage/local/var/tmp/sage/build/openblas-0.3.6.p0/src/exports'
make[4]: *** [Makefile:101: shared] Error 2
make[4]: Leaving directory '/home/patchbot/sage/local/var/tmp/sage/build/openblas-0.3.6.p0/src'
********************************************************************************



The error has appeared both in a OpenSuse 15 and Debian 10 installs.

Someone has an idea about what could be going on?

Dima Pasechnik

unread,
Dec 4, 2019, 11:41:07 AM12/4/19
to sage-devel
I suppose openblas fails to recongnise the CPU, resorts to TARGET=ATOM
and fails there.
You probaby can try setting OPENBLAS_CONFIGURE="$OPENBLAS_CONFIGURE TARGET=???"
replacing ??? with the correct value, and rebuild...
> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/08b6df64-6698-4d7c-92aa-e123f71a57d4%40googlegroups.com.

mmarco

unread,
Dec 4, 2019, 12:51:05 PM12/4/19
to sage-devel
Thanks, using

OPENBLAS_CONFIGURE="TARGET=ZEN"  make 

seemed to work.

Now, since It will be running a patchbot, it might need to recompile openblas depending on the ticket that is being checked at each time. Is there a way to make sure that this flag is used always?
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-...@googlegroups.com.

mmarco

unread,
Dec 4, 2019, 12:54:13 PM12/4/19
to sage-devel
Great, now fflas_ffpack complains about not finding blas

Dima Pasechnik

unread,
Dec 4, 2019, 1:53:24 PM12/4/19
to sage-...@googlegroups.com


On Wed, 4 Dec 2019 at 17:51, mmarco <mma...@unizar.es> wrote:
Thanks, using

OPENBLAS_CONFIGURE="TARGET=ZEN"  make 

seemed to work.

Now, since It will be running a patchbot, it might need to recompile openblas depending on the ticket that is being checked at each time. Is there a way to make sure that this flag is used always?

I suppose you will be running the bot under a dedicated username.
The probably adding 

export OPENBLAS_CONFIGURE="TARGET=ZEN"

to .bashrc or something like that will do
To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/a883de7f-8af4-421d-b90c-233f7deff502%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages