Here some more corrections. I tried to delete overlap with Rolf's list.
------------------------------------------------------------ %< cut here
Chapter 5, One-Sided Communication
65:45 "<type> base" should (?) be "<type> BASE(*)" to be consistent
with create_win.
66:41 "as specified in the init call" should (?) be "as specified in
the create_win calls"
67:12 "MPI_Win_create_win" should be "MPI_Comm_create_win" if 67:1 is
correct.
67:14 "MPI_WIN_CREATE" should be "MPI_COMM_CREATE_WIN" if 67:1 is
correct.
67:19 "MPI::Win::Create" should be "MPI::Comm::Create_win" if 67:1 is
correct.
67:31 "spawn" should be "span" (? :-)
68:16 } "MPI_WIN_INIT" then should be "MPI_COMM_CREATE_WIN"
68:20 }
73:18 }
74:42 }
76:27 }
68:18 } "MPI_MEM_ALLOC" should be "MPI_ALLOC_MEM"
71:41 }
73:37 "END DO" is missing
74:19 } assert in MPI_WIN_FENCE is missing in Fortran examples
74:24 } "_FENCE(" should be "_FENCE(0," for example
74:44 }
75:4 }
76:31 }
76:38 }
78:8 "This mechanism is more efficient" should (?) be "This mechanism
may be more efficient". MPI does not mandate relative efficiency.
------------------------------------------------------------ %< cut here
-- +-|epcc|----------------------------+----<l.clarke@epcc.ed.ac.uk>-+ | Lyndon J Clarke, PhD | Tel: [+ 44] (0)131 650 5021 | | EPCC, The University of Edinburgh | Fax: [+ 44] (0)131 650 6555 | +-----------------------------------+-----------------------------+