Steve
----------------------------------------------------------------------
Dear all,
since I don't know whom to send suggested corrections to Section 1 - 3 of the
MPI-2 draft of May 8, 1997, here they are.
7/2-4 update marks are in the wrong margin
9/29 (void *) is bracketed not everywhere in the text (e.g., 11/26)
9/35,41 different fonts are used for "INTEGER (KIND=... )"
10/3,4,10-12 would look better in 2.6.1
10/10-12 these two sentences seem to contradict each other. Why avoiding
only the lowercase prefixes?
11/34 not all C++ compilers can do namespacing yet
11/43,44 IMHO, a C compiler is unlikely to see anything that has been
discarded by the preprocessor. How can the compiler "require
that the source protected this way be legal C code"? What
particular compiler was caught doing this?
12/4 "semantic descriptions"
13/36 would do better before 13/32
15/43.5 syntax error. Better:
MPI_Init(&argc,&argv)
18/40-43 these lines echo 18/21-24 almost literally
19/6 closing ) is missing
21/32 "implementation"
22/20-23 strictly speaking, the mentioned output argument has type
(int *) rather than just (int)
24/22.5 reference to "this document" from within MPI 1.1 would
look strange unless MPI 1.1 and MPI-2 documents were going to
be merged; perhaps, "MPI-2 specs" are meant here
25/24, 26/2 see 24/22.5 above
26/35-41 I found that at least on IBM SP2 a jth process cannot send
data of length 0; perhaps, a clarification for all the
collective calls of this kind is necessary
Generally, it's strange (although historically clear) to see "I/O" chapter
behind "Miscellany". I'd suggest the "I/O" to be put before "Language
bindings". Hopefully, it's not too late.
Best regards.
Alexander
Dr. Alexander Supalov
GMD -- German National Research Center for Information Technology
SCAI -- Institute for Algorithms and Scientific Computing
Schloss Birlinghoven phone: +49 2241 14 2371
53754 Sankt Augustin fax: +49 2241 14 2102
Germany e-mail: supalov@gmd.de