On Sun, Feb 06, 2011 at 06:00:01PM -0600, Rajeev Thakur wrote:

In Example 11.12, don't you want the unlock to be after the read value on process A?

In Example 11.15, why is a Win_sync needed after the lock_all?

In Example 11.16, why is a flush needed immediately before the unlock_all?

In Example 11.17, why is the sync needed after the lock_all, and why is the flush needed before the unlock_all? Correct, I simply added the lock_all, unlock_all because it was defined to be inside such an epoch. This is is correct but might be simplified as Rajeev pointed out :-).

All instances are fixed now (see new attached documents). I also fixed a bug in the MPI_Get_accumulate() C interface which Adam pointed out to me a second ago (we need more reviews). Thanks & All the Best, Torsten -- bash$ :(){ :|:&};: --------------------- http://www.unixer.de/ ----- Torsten Hoefler | Performance Modeling and Simulation Lead Blue Waters Directorate | University of Illinois (UIUC) 1205 W Clark Street | Urbana, IL, 61801 NCSA Building | +01 (217) 244-7736