raspberrypi.rst

[vlzambr2]

No description provided.

[cassella]

Hi @vlzambr2 , FYI the Chapel group doesn't get notifications on PRs being opened. To get their attention, mention someone here (maybe the person who suggested this doc), or ask for a reviewer in gitter or discourse.

[lydia-duncan]

Looks like you have two errors:

/__w/chapel/chapel/doc/rst/platforms/raspberrypi.rst:86: ERROR: Unknown target name: "gasnet documentation".

...
/__w/chapel/chapel/doc/rst/platforms/raspberrypi.rst: WARNING: document isn't included in any toctree

I suspect the former is due to an incorrect link. The latter can be resolved by adding your file to the appropriate place in doc/rst/platforms/index.rst

[cassella]

BTW, if you haven't discovered it yet, make docs may be a more convenient way to test. You may need to install doxygen and python to run that. See also doc/rst/developer/bestPractices/buildingdocs.rst for some tips about linking.

[vasslitvinov]

Perhaps not the compiler -- it is not needed on each node. How about "Ensure that the Chapel tree ...".

What if I build the Chapel tree on the shared file system? Perhaps mention it as a possibility?

[arezaii]

Hi @vlzambr2 , thanks for writing these instructions!
To fix the docs CI, you can add .. _GASNet documentation: https://gasnet.lbl.gov/dist/udp-conduit/README to the end of this file and add raspberrypi to the "Major Platforms" section of doc/rst/platforms/index.rst.
You can see these changes at main...arezaii:chapel:raspberrypi-docs.

[vlzambr2]

I am okay with the team updating this post-merge! Also, I haven't been able to build the compiler from main even on a Pi with 1GB of RAM. How was it done? Were any steps documented?

…

On Thu, Mar 16, 2023, 9:27 AM arezaii ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In doc/rst/platforms/raspberrypi.rst <#21356 (comment)>: > + +2. Set up GASNet environment variables + +Set the following GASNet environment variables: + +.. code-block:: bash + + GASNET_SPAWNFN=S # Job spawn mechanism, where 'S' means ssh/rsh-based spawn + GASNET_MASTERIP= # IP address of master node in your cluster, eg. 10.0.0.1 + GASNET_SSH_SERVERS= # Space-delimited list of server names, eg. 10.0.0.1 10.0.0.2 10.0.0.3 + + +Notes +----- + +The last version of the Chapel compiler recorded to successfully build on a Raspberry Pi is 1.28.0. @vlzambr2 <https://github.com/vlzambr2> - I believe this is more a limitation of the specific hardware used, not Chapel on Pis in general. For example, we have seen our latest main build on Pi hardware that includes >512MB of RAM. Additionally, we generally avoid including version numbers in the docs like this because they tend to be forgotten and then not updated/revisited later. Would you mind either updating this PR to remove this note, or acknowledging that you are OK with us updating the docs post-merge by responding here? — Reply to this email directly, view it on GitHub <#21356 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ATA6HNU5AEXWGK4I4AZVAOLW4M5QJANCNFSM6AAAAAATYRQIMQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[arezaii]

Also, I haven't been able to build the compiler from main even on a Pi with 1GB of RAM. How was it done? Were any steps documented?

This is good information to know, thank you!

It was a Pi 4 B model with 2GB of RAM. AFAIK, there was no special configuration or setup used, just a standard 'make' from the root of the source tree. Tagging @riftEmber, who did the build on their hardware, in case there are more questions.

[arezaii]

Thanks for the contribution, @vlzambr2!

[riftEmber]

I just redid my build on my Pi as Ahmad described and it succeeded, so I am fairly confident this is an amount of RAM issue vs something else about the Raspberry Pi platform.

1GB of RAM not being enough is surprising and unfortunate. For what it's worth, memory required to build chpl hasn't fallen off our radar, and we are looking into setting up automated testing to make sure we are aware when the memory requirement increases (so we can publish a memory requirement and/or take action to reduce it).