EPCCed / wee_archlet

Instructions for setting up a Raspberry Pi cluster.
https://epcced.github.io/wee_archlet/
Other
16 stars 5 forks source link

Problems compiling Linpack #4

Closed simonhennessey closed 6 years ago

simonhennessey commented 6 years ago

Hi, I've got the cluster working but I can't get linpack to compile. I'm hitting two problems and having no success in fixing them. I'm using the latest version of Raspbian on new Pi3 Model B.
I get two errors from make arch=rpi: 'linker input file unused' and 'undefined reference to symbol 'pthread_create@@GLIBC_2.4'': See:

mpicc -o HPLpddriver.o -c -DAdd -DF77_INTEGER=int -DStringSunStyle -I/home/shared_dir/linpack/hpl-2.2/include -I/home/shared_dir/linpack/hpl-2.2/include/r$ gcc: warning: /usr/lib/atlas-base/libf77blas.a: linker input file unused because linking not done gcc: warning: /usr/lib/atlas-base/libatlas.a: linker input file unused because linking not done mpicc -o HPLpdinfo.o -c -DAdd -DF77_INTEGER=int -DStringSunStyle -I/home/shared_dir/linpack/hpl-2.2/include -I/home/shared_dir/linpack/hpl-2.2/include/rpi$ gcc: warning: /usr/lib/atlas-base/libf77blas.a: linker input file unused because linking not done gcc: warning: /usr/lib/atlas-base/libatlas.a: linker input file unused because linking not done mpicc -o HPLpdtest.o -c -DAdd -DF77_INTEGER=int -DStringSunStyle -I/home/shared_dir/linpack/hpl-2.2/include -I/home/shared_dir/linpack/hpl-2.2/include/rpi$ gcc: warning: /usr/lib/atlas-base/libf77blas.a: linker input file unused because linking not done gcc: warning: /usr/lib/atlas-base/libatlas.a: linker input file unused because linking not done mpif77 -o /home/shared_dir/linpack/hpl-2.2/bin/rpi/xhpl HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o /home/shared_dir/linpack/hpl-2.2/lib/rpi/$ /usr/bin/ld: /usr/lib/arm-linux-gnueabihf/libmpich.a(lib_libmpich_la-mpiu_thread_posix.o): undefined reference to symbol 'pthread_create@@GLIBC_2.4' //lib/arm-linux-gnueabihf/libpthread.so.0: error adding symbols: DSO missing from command line collect2: error: ld returned 1 exit status Makefile:76: recipe for target 'dexe.grd' failed make[2]: [dexe.grd] Error 1 make[2]: Leaving directory '/home/shared_dir/linpack/hpl-2.2/testing/ptest/rpi' Make.top:64: recipe for target 'build_tst' failed make[1]: [build_tst] Error 2 make[1]: Leaving directory '/home/shared_dir/linpack/hpl-2.2' Makefile:72: recipe for target 'build' failed make: *** [build] Error 2

I had previously had problems with the LAdir setting giving filenames of the form /usr/lib/atlas-base//libf77blas.a (with // ) so have set LAdir to /usr/lib/atlas-base (no final / ).

Any help would be much appreciated. Simon

gpsgibb commented 6 years ago

Hi Simon,

Were you building Linpack on one of the Pis, or on an external shared drive? That sometimes causes linking errors.

It could also be a case of some libraries not being installed. Could you try sudo apt-get install mpich-dev because it may be that your version of raspbian separates the MPI build into two packages (mpich, and mpich-dev) which for some reason are not both installed during sudo apt-get install mpich.

Let me know if any of that makes a difference.

Regards,

Gordon

simonhennessey commented 6 years ago

Hi Gordon

Thanks for your prompt response.

I was building Linpack following the instructions in https://epcced.github.io/wee_archlet/.

I tried sudo apt-get install mpich-dev and got an error ‘unable to locate package mpich-dev’.

I have successfully run the test:

mpiexec -n 20 -f hostfile ./hello

It is just the setup of Linpack that is failing.

Any help would be much appreciated.

Best regards

Simon

From: Gordon Gibb [mailto:notifications@github.com] Sent: 21 March 2018 12:15 To: EPCCed/wee_archlet wee_archlet@noreply.github.com Cc: simonhennessey simon@shennessey.co.uk; Author author@noreply.github.com Subject: Re: [EPCCed/wee_archlet] Problems compiling Linpack (#4)

Hi Simon,

Were you building Linpack on one of the Pis, or on an external shared drive? That sometimes causes linking errors.

It could also be a case of some libraries not being installed. Could you try sudo apt-get install mpich-dev because it may be that your version of raspbian separates the MPI build into two packages (mpich, and mpich-dev) which for some reason are not both installed during sudo apt-get install mpich.

Let me know if any of that makes a difference.

Regards,

Gordon

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/EPCCed/wee_archlet/issues/4#issuecomment-374917137 , or mute the thread https://github.com/notifications/unsubscribe-auth/AjVH3pGcRdqVXlyY0x9g39kcl1IJXOc8ks5tgkRZgaJpZM4SyeNv .


This email has been checked for viruses by AVG. http://www.avg.com

gpsgibb commented 6 years ago

Hi Simon,

I assume you're using Raspbian Stretch? We haven't tested our instructions on this version so there may be issues. We'll try to get our hands on a spare board and install stretch on it and see if we can reproduce your problem.

Hopefully we'll have some answers by next week.

Cheers,

Gordon

simonhennessey commented 6 years ago

Hi Gordon

Yes, that’s right. I’m new to this so didn’t initially appreciate the differences between Stretch and Jessie etc.

I’ve also had problems with the dhcp server but have been using the IP addresses from my router successfully, so far.

I think that stretch changed the way network connections were named but on my system they still seem to be eth0 and wlan0, which I think can be an option.

If you could try and stretch and let me know what you find that would be great.

Many thanks

Simon

From: Gordon Gibb [mailto:notifications@github.com] Sent: 22 March 2018 16:01 To: EPCCed/wee_archlet wee_archlet@noreply.github.com Cc: simonhennessey simon@shennessey.co.uk; Author author@noreply.github.com Subject: Re: [EPCCed/wee_archlet] Problems compiling Linpack (#4)

Hi Simon,

I assume you're using Raspbian Stretch? We haven't tested our instructions on this version so there may be issues. We'll try to get our hands on a spare board and install stretch on it and see if we can reproduce your problem.

Hopefully we'll have some answers by next week.

Cheers,

Gordon

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/EPCCed/wee_archlet/issues/4#issuecomment-375359909 , or mute the thread https://github.com/notifications/unsubscribe-auth/AjVH3vR-glVjszytWaqLeYZI66f-S6Vrks5tg8qsgaJpZM4SyeNv .


This email has been checked for viruses by AVG. http://www.avg.com

agngrant commented 6 years ago

Hi,

Apologies for the late reply - the error appears to be centred around this problem: /usr/bin/ld: /usr/lib/arm-linux-gnueabihf/libmpich.a(lib_libmpich_la-mpiu_thread_posix.o): undefined reference to symbol 'pthread_create@@GLIBC_2.4'

In the Make.rpi file - add to the line: HPL_LIBS -lpthread It should now read: HPL_LIBS = $(HPLlib) $(LAlib) $(MPlib) -lpthread

If you can try this and let us know. This is what solved it on our test.

regardsm alistair.

simonhennessey commented 6 years ago

Hi Alistair

That seems to have fixed it – many thanks for your help.

Regards

Simon

From: agngrant [mailto:notifications@github.com] Sent: 20 April 2018 12:05 To: EPCCed/wee_archlet wee_archlet@noreply.github.com Cc: simonhennessey simon@shennessey.co.uk; Author author@noreply.github.com Subject: Re: [EPCCed/wee_archlet] Problems compiling Linpack (#4)

Hi,

Apologies for the late reply - the error appears to be centred around this problem: /usr/bin/ld: /usr/lib/arm-linux-gnueabihf/libmpich.a(lib_libmpich_la-mpiu_thread_posix.o): undefined reference to symbol 'pthread_create@@GLIBC_2.4'

In the Make.rpi file - add to the line: HPL_LIBS -lpthread It should now read: HPL_LIBS = $(HPLlib) $(LAlib) $(MPlib) -lpthread

If you can try this and let us know. This is what solved it on our test.

regardsm alistair.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/EPCCed/wee_archlet/issues/4#issuecomment-383062823 , or mute the thread https://github.com/notifications/unsubscribe-auth/AjVH3jftmn2R6Yox2GrXav9SljyAi94aks5tqcDDgaJpZM4SyeNv . https://github.com/notifications/beacon/AjVH3rqG6utOeEc4fCe7mfZFqVhcuv98ks5tqcDDgaJpZM4SyeNv.gif


This email has been checked for viruses by AVG. http://www.avg.com

agngrant commented 6 years ago

Issue appears to have been addressed - closing 03/05/2018