freebsd-skq/usr.bin/gprof/PSD.doc/profiling.me
1994-05-27 12:33:43 +00:00

116 lines
5.6 KiB
Plaintext

.\" Copyright (c) 1982, 1993
.\" The Regents of the University of California. All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\" 3. All advertising materials mentioning features or use of this software
.\" must display the following acknowledgement:
.\" This product includes software developed by the University of
.\" California, Berkeley and its contributors.
.\" 4. Neither the name of the University nor the names of its contributors
.\" may be used to endorse or promote products derived from this software
.\" without specific prior written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" @(#)profiling.me 8.1 (Berkeley) 6/8/93
.\"
.sh 1 "Types of Profiling"
.pp
There are several different uses for program profiles,
and each may require different information from the profiles,
or different presentation of the information.
We distinguish two broad categories of profiles:
those that present counts of statement or routine invocations,
and those that display timing information about statements
or routines.
Counts are typically presented in tabular form,
often in parallel with a listing of the source code.
Timing information could be similarly presented;
but more than one measure of time might be associated with each
statement or routine.
For example,
in the framework used by \fBgprof\fP
each profiled segment would display two times:
one for the time used by the segment itself, and another for the
time inherited from code segments it invokes.
.pp
Execution counts are used in many different contexts.
The exact number of times a routine or statement is activated
can be used to determine if an algorithm is performing as
expected.
Cursory inspection of such counters may show algorithms whose
complexity is unsuited to the task at hand.
Careful interpretation of counters can often suggest
improvements to acceptable algorithms.
Precise examination can uncover subtle errors in an
algorithm.
At this level, profiling counters are similar to
debugging statements whose purpose is to show the number of times
a piece of code is executed.
Another view of such counters is as boolean values.
One may be interested that a portion of code has executed at
all, for exhaustive testing, or to check that one implementation
of an abstraction completely replaces a previous one.
.pp
Execution counts are not necessarily proportional to the amount
of time required to execute the routine or statement.
Further, the execution time of a routine will not be the same for
all calls on the routine.
The criteria for establishing execution time
must be decided.
If a routine implements an abstraction by invoking other abstractions,
the time spent in the routine will not accurately reflect the
time required by the abstraction it implements.
Similarly, if an abstraction is implemented by several
routines the time required by the abstraction will be distributed
across those routines.
.pp
Given the execution time of individual routines,
\fBgprof\fP accounts to each routine the time spent
for it by the routines it invokes.
This accounting is done by assembling a \fIcall graph\fP with nodes that
are the routines of the program and directed arcs that represent
calls from call sites to routines.
We distinguish among three different call graphs for a program.
The \fIcomplete call graph\fP incorporates all routines and all
potential arcs,
including arcs that represent calls to functional parameters
or functional variables.
This graph contains the other two graphs as subgraphs.
The \fIstatic call graph\fP includes all routines and all possible arcs
that are not calls to functional parameters or variables.
The \fIdynamic call graph\fP includes only those routines and
arcs traversed by the profiled execution of the program.
This graph need not include all routines, nor need it include all
potential arcs between the routines it covers.
It may, however, include arcs to functional parameters or
variables that the static call graph may omit.
The static call graph can be determined from the (static) program text.
The dynamic call graph is determined only by profiling an
execution of the program.
The complete call graph for a monolithic program could be determined
by data flow analysis techniques.
The complete call graph for programs that change
during execution, by modifying themselves or dynamically loading
or overlaying code, may never be determinable.
Both the static call graph and the dynamic call graph are used
by \fBgprof\fP, but it does not search for the complete call
graph.