Dar Yoshi,

I don't have much experience with cross compilers, but I'll try to answer your questions below.

On 11/12/2017 14:32, Yoshi Horigome wrote:
Hello Kenneth,

I'm sorry, your report has been delayed.
I am sorry that the explanation is bad.

As Markus says, it is a cross compiler used in the login node in a
system using FX100 as a compute node and x86_64 as the login node. By
using this you will be able to generate code for SPARCv9.

I assume the login node is also going to be using as a build node too, i.e. that's where software will be compiled & installed, to then be used on the compute nodes?

One problem you will definitely run into is that lots of software configuration/build scripts try to run small tests during the installation. In this setup, this is going to fail pretty hard since SPARCv9 code won't run on the x86_64 build host.

And there's little that EasyBuild can do about this, if the software build procedure is not aware of cross-compilation...

I would like to incorporate this cross compiler itself as a toolchain
of Easybuild. Since linking with other libraries should not have much
to think about, is it close to intel compiler and so on as a
composition?

For the composition of a new toolchain, you  need three things:

* a toolchain definition that the EasyBuild framework is aware of;
  see the ones in [1]; you can include a toolchain definition dynamically via --include-toolchains, see [2]

* an easyconfig file for a particular version of that toolchain,
  which also specifies the version of the toolchain components;
  see for example the ones we have for the 'foss' toolchain [3]

* easyconfig files for each of the components, except for the ones you mark as 'external' modules [4]

In this case, is it better to refer to intel compiler's Easyblock file
and create a file as toolchain? Although it seems that options are set
as a constant as far as I confirmed a little, in this case, I did not
know the judgment criteria etc. as to what option to select in the
Easyblock file will be selected is.

Should I reprint an Easyblock file that describes the flow of
processing at the moment?
It seems you're mixing up terms here...
What do you mean when you say "intel compiler's easyblock file", can you point it out?


Also, can you outline which components (compiler, MPI library, BLAS/LAPACK libraries) you would be using in this cross-compiler toolchain?
That may help with discussing this.


regards,

Kenneth


[1] https://github.com/easybuilders/easybuild-framework/tree/master/easybuild/toolchains [2] http://easybuild.readthedocs.io/en/latest/Including_additional_Python_modules.html#including-additional-toolchains-include-toolchains [3] https://github.com/easybuilders/easybuild-easyconfigs/tree/master/easybuild/easyconfigs/f/foss
[4] http://easybuild.readthedocs.io/en/latest/Using_external_modules.html


Markus
Thank you for following me.


Best Regards.

2017-12-11 17:21 GMT+09:00 Markus Geimer <m.gei...@fz-juelich.de>:
Kenneth,

If it's about Fujitsu FX, then GCC is a real cross-compiler, i.e.,
running on an x86_64 frontend but producing code for SPARC compute
nodes.

Markus

On 12/10/2017 06:31 PM, Kenneth Hoste wrote:
Dear Yoshi,

Sorry for the lack of response to your question.

I guess your question is not entirely clear...

So, you want to build GCC on a Fujitsu system using EasyBuild?

Can you elaborate on what makes this a cross-compiler specifically?

Did you already get it working, or are you experiencing problems?

Are you asking if it makes sense to contribute back your efforts by
defining an additional toolchain that is specific to Fujitsu somehow?
If so, how is it a different from another GCC-based toolchain like
'foss' (see
http://easybuild.readthedocs.io/en/latest/Common-toolchains.html#foss-toolchain)?



regards,

Kenneth

On 08/12/2017 05:44, Yoshi Horigome wrote:
Hello Kenneth,

Please let me know about this matter.
Can I post as an issue in the Easybuild-framework repository?

"If you implement support for new compiler toolchains and / or
toolchain elements,
please consider contributing them back to the main easybuild-framework
repository."

Best regards.

2017-11-28 17:41 GMT+09:00 Yoshi Horigome <kome...@gmail.com>:
Hello, all

Please kindly advise.
I would like to make compiler such as GCC for Fujitsu available for
EasyBuild.

This cross-compiler, to use the software for performing the other
scientific computing, I think that it is necessary to be able to also
be used as a toolchain.

1. Does this idea match?
2. Is it enough to add the framework file in the following directory
by referring to it?


${HOME}/.local/easybuild/software/EasyBuild/3.4.1/lib/python2.6/site-packages/easybuild_framework-3.4.1-py2.6.egg/easybuild/toolchains


${HOME}/.local/easybuild/software/EasyBuild/3.4.1/lib/python2.6/site-packages/easybuild_framework-3.4.1-py2.6.egg/easybuild/toolchains/compiler


3. I've confirmed "https://easybuild.readthedocs.io/";, but are there
any documents to be helpful?


As a current situation, it is possible to create an easyconfig file
and an Easyblock file to make the cross compiler environment model
using EasyBuild.

Best regards.
--
Dr. Markus Geimer
Juelich Supercomputing Centre
Institute for Advanced Simulation
Forschungszentrum Juelich GmbH
52425 Juelich, Germany

Phone:  +49-2461-61-1773
Fax:    +49-2461-61-6656
E-Mail: m.gei...@fz-juelich.de
WWW:    http://www.fz-juelich.de/jsc


------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------
Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
Prof. Dr. Sebastian M. Schmidt
------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------


Reply via email to