Files
seta75D ff309bfe1c Init
2021-10-11 18:37:13 -03:00

87 lines
1.7 KiB
Groff

.\" @(#)prof.3 1.1 94/10/31 SMI; from S5
.TH PROF 3 "6 October 1987"
.SH NAME
prof \- profile within a function
.SH SYNOPSIS
.LP
.nf
.B #define \s-1MARK\s+1
.B #include <prof.h>
.LP
.B void \s-1MARK\s+1 (name)
.fi
.SH DESCRIPTION
.IX "prof()" "" "\fLprof()\fR \(em profile within a function"
.IX "profiling, within a function" "" "\fLprof()\fR"
.LP
.SM MARK
introduces a mark called
.I name
that is treated
the same as a function entry point.
Execution of the
mark adds to a counter for that mark, and
program-counter time spent is accounted to the
immediately preceding mark or to the function if
there are no preceding marks within the active function.
.LP
.I name
may be any combination of up to six letters, numbers or
underscores. Each
.I name
in a single compilation must be unique, but may be the same as
any ordinary program symbol.
.LP
For marks to be effective, the symbol
.SM MARK
must be defined before the header file
.B <prof.h>
is included. This may be defined by a
preprocessor directive as in the synopsis, or by a command
line argument, such as:
.RS
.LP
.B cc \-p \-\s-1DMARK\s+1 foo.c
.RE
.LP
If
.SM MARK
is not defined, the
.SM MARK
.RB ( name )
statements may be left in the source files
containing them and will be ignored.
.SH EXAMPLE
In this example, marks
can be used to determine how much time is spent in each loop.
Unless this example is compiled with
.SM MARK
defined on the command line, the
marks
are ignored.
.IP
.ft B
.nf
#include <prof.h>
func( )
{
int i, j;
.
.
.
\s-1MARK\s+1 (loop1);
for (i = 0; i < 2000; i++) {
. . .
}
\s-1MARK\s+1 (loop2);
for (j = 0; j < 2000; j++) {
. . .
}
}
.fi
.ft R
.SH SEE ALSO
.BR prof (1),
.BR profil (2),
.BR monitor (3)