Sequential consistency?

John M May (johnmay@coral.llnl.gov)
Thu, 24 Oct 1996 15:08:13 -0700 (PDT)

I'm having trouble understanding Section 10.6.1 (pp234-236) of
the current draft of the I/O chapter. First, there seems to
be a formatting error: the descriptions of MPI_SET_ATOMICITY
and MPI_GET_ATOMICITY interrupt the discussion of file
consistency. Should't the paragraphs beginning on p 236 at
line 10 go right after the advice to implementors at p235
line 28?

But my real confusion is with the notion of sequential
consistency, or the lack of it, on *non-overlapping*
accesses. How would one detect non-sequentially-
consistent behavior between accesses that didn't overlap?

John