[gentoo-soc] Progress Report: MPI Overlay, Week 3

[Michael Gilroy]

Hi,

*Summary:*

This week I worked towards a much better mpi-providers eclass and started
the framework for the mpi-select.

Last week's plan:
1) 2ND ECLASS: solidify design, verify with mentors. I expect the 2nd
eclass to take a few weeks to complete.
2) properly implement mpi-providers eclass for more mpi implementations,
write patches accordingly
3) refactor/modularize 1st eclass
4) start proper documentation (formatted in markdown)

mpi-providers (1st eclass) is nearing completion- I believe just some more
error handling is needed. As per conversations/meetings with mentors, the
2nd eclass design is making a lot more sense, and I will begin writing
non-pseudocode this week.

Next week's plan:
1) Start working towards multiplexed & multilib installs for mpi-select.
2) Start testing new mpi-select against mpich and openmpi implementations,
seeing if both implementations can be installed against something like hpl.
3) Refine wiki/documentation for the project


*Day-by-day breakdown:*
*June 12th:*
1. What was my plan for today?
a) Discuss progress of mpi-select eclass on Monday, get design criticism.
b) Discuss & further test sysconfdir implementations.
2. Did I achieve it? If not, what do I intend to do about it?
a) Completed- I now have design guidance. I will be tinkering with
multilib/multilib-minimal to try and understand that implementation.
b) Testing. Will give details if this continues to be a pain.
 3. What is my plan for tomorrow?
a) Progress was made on safe_mv without the use of rsync. Test this
further. Currently in the dev branch. If properly tested, merge with
master.
b) Sysconfdir setting is still an issue. Test this, if it is still
failing, write a detailed email about the problems.

*June 13th:*
1. What was my plan for today?
a) Progress was made on safe_mv without the use of rsync. Test this
further. Currently in the dev branch. If properly tested, merge with
master.
b) Sysconfdir setting is still an issue. Test this, if it is still
failing, write a detailed email about the problems.
c) Continue exploring multilib/multilib-minimal implementation to get
inspiration for mpi-select design. Start setting up framework for
exported functions.
2. Did I achieve it? If not, what do I intend to do about it?
a) Completed. Changes are merged with master.
b) Completed. Sysconfdir is properly set, will add documentation on usage
to wiki.
c) Work in progress. This will take a bit to properly figure out.
 3. What is my plan for tomorrow?
a) Continue exploring multilib implementations, debug exactly what the
eclasses are doing and why they are doing it.
b) Try to have mentors review 1st eclass, as it is nearing completion.
c) Continue working on skeleton for 2nd eclass.
d) Expand upon the wiki page.

*June 14th:*
1. What was my plan for today?
a) Continue exploring multilib implementations, debug exactly what the
eclasses are doing and why they are doing it.
b) Try to have mentors review 1st eclass, as it is nearing completion.
c) Continue working on skeleton for 2nd eclass.
d) Expand upon the wiki page.
2. Did I achieve it? If not, what do I intend to do about it?
a) Work in progress. I will continue tinkering tomorrow.
b) Incomplete. I would like to clean up my repo with repoman tomorrow, and
stick to best practices from here on out. I will be doing this and emailing
the mentors tomorrow.
c) Complete. I hope to get heavy criticism.
d) Work in progress. Once I make it look prettier, there will be more to
add as I continue my work.
 3. What is my plan for tomorrow?
a) Make repoman happy.
b) Email mentors regarding 1st eclass review, among other things in our
email thread.
c) Continue multilib tinkering.
d) Receive feedback on mpi-select.eclass and start on new iteration of
eclass skeleton

*June 15th:*
*Today's plans:*

Make repoman happy.

- Work in progress. There has been some repository refactoring, but repoman
is still yelling about ebuild syntax. Hoping to wrap this up tomorrow.


Email mentors regarding 1st eclass review, among other things in our email
thread.

- Completed. As per my weekly meeting with Andrew, I have my work cut out
for me for the foreseeable future :) . Will continue to email with
questions.


Continue multilib tinkering.

Work in progress. Expect commits regarding debuggin output tomorrow.


Receive feedback on mpi-select.eclass and start on new iteration of eclass
skeleton.

Completed. The bulk of this work will likely start next week, but the
feedback I have received has been very helpful.

*Tomorrow's plans:*

Add more debugging statements to multilib to get a better sense of going
on, commit changes. Start documenting progress as needed.

Continue work on "find" statement of mpi-providers, as discussed in our
weekly meeting.

Finalize overlay structure, fix ebuild-specific syntax errors for repoman.


*June 16th:*
*Today's plans:*

Add more debugging statements to multilib to get a better sense of going
on, commit changes. Start documenting progress as needed.
Work in progress. Starting Monday I will be committing code to emulate the
multilib implementation.

Continue work on "find" statement of mpi-providers, as discussed in our
weekly meeting.
Work in progress. I have been playing with this but it requires further
testing.

Finalize overlay structure, fix ebuild-specific syntax errors for repoman.

Completed. I'm hoping to keep master repoman-friendly.

*Monday's plans:*

Clean up documentation (README, metadata.xml's) and migrate wiki page.

Test and add more implementation versions to the overlay.

Fix find statement in mpi-providers.


Start committing code to emulate multilib multiplexing with MPI installs.


Thanks,
Michael