GitLab

[svn-r7416] Major improvement to RMA performance for long lists of operations, an immediate mode accumulate for single ints, store the MPID_Comm within the window, and added a basic performance instrumentation interface that was extensively used to improve the RMA performance (enabled with --enable-g=instr).  With these fixes, MPICH2 can run the one-sided version of the Graph500 benchmark at a respectable if not great rate

[svn-r7413] Possible fixes for #1117, need to build shared library for C++ files with the C++ compiler rather than the C compiler

[svn-r7408] well this sort of got out of hand... the venerable john may testmpio test fails if access modes are declared differently on the processors, so now we catch this error.

found.

[svn-r7400] The vc overridden probe function is blocking.  Unless internally the allfunc mutex is released, this could lead to a deadlock.  Instead, we call iprobe in a loop, and release the mutex in the loop.

RESTART_FAIL, PROCS_DEAD.

events.

perfectly valid flattened representations end up with one big block and do not
need an additional upper bound or lower bound.   Thanks Michael Raymond
<mraymond@sgi.com>

and the unexpected callback might be called and set a NULL vc ptr to the ADI request.
This test fixes this case. Ideally, connecting to the remote endpoint and attaching
the vc ptr info should be atomic.

[svn-r7378] finally, after far too many years, map PVFS error codes to something like MPI error codes

some cases.

No reviewer.

No reviewer.

No reviewer.