lamtask(1)

1MPITASK(1)                       LAM COMMANDS                       MPITASK(1)
2
3
4

NAME

6       mpitask - Monitor MPI processes under LAM.
7

SYNTAX

9       mpitask [-cdh] [-gps] [<nodes>] [<processes>]
10

OPTIONS

12       -c      Print communicator descriptions.  See "Communicators".
13
14       -d      Print datatype descriptions.  See "Datatypes".
15
16       -gps    Print process information in GPS format.
17
18       -h      Print useful information on this command.
19
20       The -c and -d options are mutually exclusive.
21

DESCRIPTION

23       The  mpitask  command displays information on processes which are using
24       MPI.  One line is printed for each reported MPI process.  With no  pro‐
25       cesses  or nodes explicitly specified on the command line, all MPI pro‐
26       cesses on all nodes are reported.
27
28% mpitask
29TASK (G/L)           FUNCTION      PEER|ROOT  TAG    COMM   COUNT   DATATYPE
300/0 trivial          Ssend         1/1        123    WORLD  64      INT
311/1 trivial          Recv          0/0        456    WORLD  64      INT
32
33       For each process mpitask normally prints the following information:
34
35       TASK      an identification of the process - If the process is current‐
36                 ly communicating, a `/' followed by the process's rank within
37                 the current communicator is also displayed.   The  executable
38                 name,  if  available,  is  also  displayed.  See "MPI Process
39                 Identification".
40
41       FUNCTION  an abbreviated form of the function name if  the  process  is
42                 blocked inside an MPI function - Otherwise one of the follow‐
43                 ing execution states is printed:
44
45                 <running> free to run on the underlying OS
46
47                 <paused>  blocked on lam_kpause(2)
48
49                 <stopped> stopped by the  LAM  signal,  LAM_SIGARREST  -  See
50                           doom(1).
51
52                 <blocked> blocked  in a LAM function - In general this should
53                           be a transitory state.  Further  information  on  a
54                           LAM process's state can be obtained with state(1).
55
56       PEER|ROOT the  source  or destination of a point-to-point communication
57                 or the root process  of  certain  collective  communications,
58                 followed  by  a '/' and the process's rank within the current
59                 communicator -
60
61       TAG       the message tag, if any, which was specified as  a  parameter
62                 to the current MPI function
63
64       COMM      the communicator ID, if any, which was specified as a parame‐
65                 ter to the current MPI function - Communicators used in  col‐
66                 lective  calls are displayed with a * suffix.  Further infor‐
67                 mation on the communicator may be obtained with  the  -c  op‐
68                 tion.
69
70       COUNT     the element count, if any, which was specified as a parameter
71                 to the current MPI function
72
73       DATATYPE  the element datatype, if any, which was specified as a param‐
74                 eter to the current MPI function - For intrinsic datatypes, a
75                 shortened version of the datatype name is displayed.  For de‐
76                 rived  datatypes, a datatype label is displayed.  Further in‐
77                 formation on the datatype may be obtained with the -d option.
78
79   MPI Process Identification
80       By  default,  MPI  processes  are   identified   by   their   rank   in
81       MPI_COMM_WORLD.   We  refer to this rank more concisely as the "global"
82       rank (G).  The rank within the currently employed communicator  is  re‐
83       ferred to as the "local" rank (L).
84
85       Since  processes  may  be  dynamically  spawned (see MPIL_Spawn(2)) and
86       since multiple concurrent MPI applications are allowed, it is  possible
87       for  multiple MPI_COMM_WORLD communicators to coexist.  In these situa‐
88       tions, the global rank is no longer globally unique and the identifica‐
89       tion  is ambiguous.  Thus, LAM provides an alternate way of identifying
90       MPI processes, the GPS (Global Positioning System).   A  process's  GPS
91       consists  of the nodeid the process is running on and the process's LAM
92       index  on  that  node.   It  is  displayed  in  mpitask  as  the   pair
93       n<node>,i<index>.   If the -gps option is given then the GPS is substi‐
94       tuted for the global rank (G).
95
96   Communicators
97       If the -c option is given then information is no  longer  displayed  in
98       the  horizontal  format  described  above.   Instead  for each selected
99       process currently using a communicator, the information from  the  TASK
100       column,  described  above, is given followed by an expanded description
101       of the communicator.  This description includes the size of the  commu‐
102       nicator  group(s)  and  the  global  identifiers  of all members of the
103       group(s).
104
105   Datatypes
106       If the -d option is given then information is no  longer  displayed  in
107       the  horizontal  format  described  above.   Instead  for each selected
108       process currently using a communicator, the information from  the  TASK
109       column, described above, is given followed by the datatype's type map.
110

EXAMPLES

112       mpitask
113           Display the status of all MPI processes on all nodes.
114
115       mpitask -c n0 i9
116           Display the communicator (if any) of process index 9 on node 0.
117

DIAGNOSTICS

119       If no MPI processes are found, only the title line is displayed.
120

BUGS

122       Sometimes  processes  may be shown as blocked inside a non-blocking MPI
123       function.  This occurs when the process is blocked inside the MPI func‐
124       tion on some internal LAM event.  Such states are highly transitory.
125