cache limit is exceeded - a request is allocated from the global pool. A flag
mca_pml_base_send_request.req_cached has been added to the request so that the
ptl can determine wether the request came from the cache and hence has been
initialized via the ptls ptl_request_init method. If not - the ptl is responsible
for allocating the resources required to handle the request - or potentially queue
the request until resources become available.
This commit was SVN r2570.
Added a field to the ompi_rte_node_schedule_t structure to keep track of the number of items on the environ list, thus making it easier to append more things to it. Adjusted the mca_pcm_base_build_base_env function correspondingly to take that field as an additional argument.
Changed mpirun2 to a .c program for convenience since it wasn't using any c++ features anyway.
This commit was SVN r2561.
RMS and RSH pcms. The RSH pcm was hard-coded to start from 0. The RMS
pcm always used the name server. UNfortunately, due to some interesting
interactions, using the name server causes failures in the PTL logic
because vpids start at 1, not 0 when the reservation code is used.
Right now, the default is to start vpids at 0 so that everything still
runs out of the box. But this should make it possible for Tim to look
at the PTL problems a bit more easily.
This commit was SVN r2560.
doesn't do what it was designed for, and therefore wasn't useful, so
per discussion with Ralph last night, we decided to scrap it.
This commit was SVN r2539.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
-- Make code compatiable with expanded libelan interface to free cmdq.
-- Put the pure device_open() related code into a static routine,
to be replaced with provided elan_int().
This commit was SVN r2533.
pcm for RMS:
- support setting vpids from an offset, in case that's what the ns
gave us (ie, not starting from 0)
- if we have it, use the given jobid from the ns rather than the
RMS jobid
* grab the cellid from the environment, if its there. In reality, this is
probably just another way of setting the cellid to 0, but seems like
a good thing to do...
This commit was SVN r2530.
