Spent much of this weekend looking at houses again.
Found two great houses over by Janna, and we were all set to sit down and slog through the details of deciding which to get, and then found out that neither of them have DSL availability.
Arrgh!!
I'm doing a bunch of LAM work right now to enable Ron and Brian start the release process for LAM 6.5 (did I mention in the journal already that we're going to call the next release 6.5 instead of 6.3.3? It's a long complicated story [e.g., where's 6.4?], but there are definite reasons for everything. To summarize: there's been major changes since 6.3.2 such that we didn't feel that an increase in the release number was sufficient to describe the enormity of the change. It isn't quite as revolutionary as should indicate a major number change, so we settled for a minor number change. There.).
I've added a whole schload of programs to the lamtests test suite, and added a few more canonical example programs to our "examples/" directory --
something I've been meaning to do for a while. We have some good examples already, but none are the "standard" examples that are typically used in MPI, like the pi approximation program and the ring program.
Now I'm briefly diverting to write a few man pages (we have a bunch left unwritten for MPI-2 functions, so we've divided them up into groups and assigned them to various Llamas. Tackling them a few at a time is a good way to whittle the number of unfinished pages down to a small number, as the limit goes to 0). Mostly MPI-2 dynamic functions for me.
After that, I'll finally get around to fixing MPI_COMM_SPAWN and MPI_COMM_SPAWN_MULTIPLE
-- there's something wrong with using app schemas such that you still have to give a process count on the root or something (you shouldn't have to). And I think the error code reporting is futzed up somehow (lamteam advised me of this about a month ago or something. Not a huge deal since errors typically cause aborts, but it is possible that someone could set the error handlers to return and expect to get valid error codes back).
Then if all else looks good (oops... looks like I have a seg fault in one of the new test programs...), I'll hand the tarballs over to Ron/Brian to begin the release process.
Long live LAM!
There are currently 144 copies of xmms running on queeg, which is 63% of all processes.