[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [ptp-dev] Update on what I'm working on
|
Just to give everyone a bit more head's up: the model now works and
fills in correctly as I outlined in the previous message. The UI, as I
expected, is not updating. I'm going to start digging through the UI
code to figure out what's going on but I expect it's not getting updates
or isn't being clean enough to let itself expand.
See, the machine's view has to load up with no entries and then expand
as nodes fill in - the components have to literally grow. Getting that
to work will be the next task.
-- Nathan
Correspondence
---------------------------------------------------------------------
Nathan DeBardeleben, Ph.D.
Los Alamos National Laboratory
Parallel Tools Team
High Performance Computing Environments
phone: 505-667-3428
email: ndebard@xxxxxxxx
---------------------------------------------------------------------
Nathan DeBardeleben wrote:
I'm still redesigning in the Model Manager and the interface to the
Monitoring and Control systems. This change was necessitated by major
misunderstandings with the Open MPI team, but they're going to end up
for the better for us, making our system more compatible with MPICH,
for example.
I've had to make changes to most of the base classes, Machine, Node,
Job, Process for instance. I've also made many changes to the
interfaces to the Control and Monitoring systems. What I've got now
works about 80% of the way, but it's not ready yet to commit. I'm
also scared about what will happen when I do, given the number of
files I've touched in a major way.
The internal model now fills up asynchronously, instead of blocking as
it did before until it knew about the entire parallel machine. This
has lots of ramifications on how the UI code works, as it has to be
able to handle having a model that may not be there or may have
entries that aren't entirely populated - like knowing there are 6
nodes, but only 3 of the nodes have names, and 5 of the nodes have
their states... it's going to be challenging but rewarding in the end.
This also has effects on things that Clement put in, like monitors, as
many of the operations are instant and not blocking.
We'll see how it turns out.
