1 09/22/83  trace_meters, tmt
 2 
 3 Syntax:  tmt {-control_args}
 4 
 5 
 6 Function:  formats and displays the cpu time and page fault meters of
 7 entry points that are traced with "-meter on" (see the trace command).
 8 
 9 
10 Control arguments:
11 -global_percent N, -gpct N, -g% N
12    reports only entry points whose global cpu time or global page
13    faults exceed N% of the total.  N must be a whole number.
14 -output_file path, -of path
15    causes output to be directed to the file specified by path.  The
16    file is overwritten if it already exists, or created if it does not.
17    The ".tmt" suffix is added to path if it is not given.
18 -percent N, -pct N, -% N
19    reports only entry points whose local cpu time or local page faults
20    exceed N% of the total.  N must be a whole number.
21 -report_reset, -rr
22    displays the report and zeroes the meters.
23 -reset, -rs
24    set the meters of every entry point in the trace table to zero and
25    does not display the report.
26 
27 
28 Notes:  If no arguments are supplied, the report is displayed and the
29 meters are not reset.
30 
31 Entry points that have not been called since the last reset are not
32 reported.
33 
34 Some programs are on a special list of programs that can never be
35 traced (see the trace command).  Some entry points can not be traced
36 because they are written in alm and are added to the trace table when
37 "-alm off" is in effect.  The tracing of some entry points may have
38 been turned off by the "trace -off" command.  Entry points that are not
39 metered either because they can not be traced or because they have been
40 turned off are, nevertheless, counted and are reported with empty meter
41 columns.
42 
43 
44 If the local virtual cpu time reported for a procedure that has been
45 called only a few times is unbelievably large, consider the possibility
46 that it snapped some dynamic links during its run.  This metering error
47 can be eliminated by reseting the meters and repeating the run within
48 the same process.  It can be completely avoided by making a trial run
49 before metering.
50 
51 
52 Notes on column headings:  The column headings of the report are
53 interpreted as follows:
54 
55        GREAL       global real time
56        GVCPU       global virtual cpu time
57        GPF         global page faults
58        LREAL       local real time
59        LVCPU       local virtual cpu time
60        LPF         local page faults
61        LVCPU/CALL  local virtual cpu time per call
62        LVCPU%      local vcpu time as a percentage of total vcpu time
63        CALLS       number of calls
64 
65 Global means the resources used by an entry point and everything it
66 calls.  Local means the resources used by an entry point less the
67 resources used by all traced entry points that it calls.