Ralph Castain 390645ac2a Per patch from Tetsuya Mishima, do a nicer job of warning the user that we need to map to a higher level to get the number of requested cpus/rank. Also, change the mapping policy to "byslot" when falling back to that option. ...

cmr=v1.8:reviewer=rhc

This commit was SVN r31196.

2014-03-24 15:47:29 +00:00

..

dfs

Fix longstanding issue with our multi-project support. Rather than using

2014-01-07 22:11:15 +00:00

errmgr

Deal with the corner case where we encounter an error when attempting to launch a daemon. In this case, we will order abnormal termination before daemons callback to us, and thus any attempt to send them a "die" message will fail. Ensure that mpirun at least exits cleanly in this scenario, thereby allowing the remote daemons that did get launched to commit suicide when comm fails.

2014-03-14 15:32:30 +00:00

ess

Set the locality for remote procs even after a comm_spawn. Ensure we store our own local cpuset upon launch so it will be shared during comm_join.

2014-03-18 14:51:07 +00:00

filem

Protect array against crossing boundaries

2014-01-17 21:36:20 +00:00

grpcomm

Fully fix the PMI2 warning - turned out to be larger than originally thought due to the way the function was being handled across multiple files. Properly resolve the problem by not compiling the file if PMI2 is not desired, and then appropriately setting the visibility of the function within the module

2014-03-17 17:36:37 +00:00

iof

Cleanup some potential memory overruns

2014-01-19 16:31:26 +00:00

odls

More corrections w.r.t. process groups

2014-03-18 21:31:01 +00:00

oob

oob: delete events when destroying a peer

2014-03-19 22:15:49 +00:00

plm

Be consistent in our handling of the "only HNP in allocation" case when setting up the VM. Thanks to Tetsuya Mishima for the suggestion.

2014-03-24 15:28:09 +00:00

ras

Paul Hargrove has pointed out that some big SMP systems (e.g., from SGI) configure Torque differently - instead of listing each node name once/slot in the nodefile, they list the node only once and set an envar to indicate the number of procs/node being allocated. Add an MCA param users can set to indicate we are in such an environment, and then use the envar to set the slots. Error out if the mode flag is given, but (a) we don't find the PBS_PPN envar, or (b) we find a node actually listed more than once in the PBS_Nodefile.

2014-02-05 15:51:17 +00:00

rmaps

Per patch from Tetsuya Mishima, do a nicer job of warning the user that we need to map to a higher level to get the number of requested cpus/rank. Also, change the mapping policy to "byslot" when falling back to that option.

2014-03-24 15:47:29 +00:00

rml

use the newly created JOB_STATE_FT_* events

2014-03-12 12:37:14 +00:00

routed

Deal with the corner case where we encounter an error when attempting to launch a daemon. In this case, we will order abnormal termination before daemons callback to us, and thus any attempt to send them a "die" message will fail. Ensure that mpirun at least exits cleanly in this scenario, thereby allowing the remote daemons that did get launched to commit suicide when comm fails.

2014-03-14 15:32:30 +00:00

sensor

Do a little cleanup - only resusage needs the node/proc info, so remove it from the sensor base

2014-03-17 21:26:46 +00:00

snapc

Use unique collective ids for the checkpoint/restart code

2014-02-04 14:03:05 +00:00

sstore

fix "warning: 'sstore_stage_select' defined but not used"

2014-03-06 16:53:27 +00:00

state

Continue to resolve priority issues. Cleanup the case of forced termination in mpirun during launch processing by ensuring we can respond to socket closures, and ensuring that the remote daemons correctly close their sockets when terminating.

2014-03-13 04:02:24 +00:00