Hi All, I've been refining the linked list example we looked at several months ago for inclusion in the RMA chapter. Previously, we had two versions: one that did put/get under exclusive lock and another that did CAS under exclusive lock. This new version is lock free and demonstrates the following MPI-3 functionality: dynamic windows, lock all, flush, CAS, accumulate with replace, and get accumulate with no op. The code can be compiled to a .o using the attached mpi3.h (for OpenMPI you may get a warning from MPI_NO_OP, define it to NULL instead of 0). It's still a little on the long side; we're currently at roughly three pages and we were hoping for two. We can try to squeeze and remove non-essential parts (e.g. maintaining the allocation list) if we want to save space. If anybody can spare a few cycles, I would appreciate any feedback. Thanks, ~Jim.