[comment {-*- tcl -*- doctools manpage}]
[manpage_begin bench_lang_spec n 1.0]
[copyright {2007 Andreas Kupries <andreas_kupries@users.sourceforge.net>}]
[moddesc   {Documentation tools}]
[titledesc {bench language specification}]
[category  {Benchmark tools}]
[description]
[para]

This document specifies both names and syntax of all the commands
which together are the bench language, version 1.

As this document is intended to be a reference the commands are listed
in alphabetical order, and the descriptions are relatively short.

A beginner should read the more informally written
[term {bench language introduction}] first.


[section Commands]
[list_begin definitions]

[call [cmd bench_rm] [arg path]...]

This command silently removes the files specified as its arguments and
then returns the empty string as its result.

The command is [emph trusted], there is no checking if the specified
files are outside of whatever restricted area the benchmarks are run
in.


[call [cmd bench_tmpfile]]

This command returns the path to a bench specific unique temporary
file. The uniqueness means that multiple calls will return different
paths. While the path may exist from previous runs, the command itself
does [emph not] create aynthing.

[para]

The base location of the temporary files is platform dependent:

[list_begin definitions]
[def {Unix, and indeterminate platform}]
[file /tmp]
[def Windows]
[var \$TEMP]
[def {Anything else}]
The current working directory.
[list_end]
[para]


[call [cmd bench] [arg options]...]

This command declares a single benchmark. Its result is the empty
string. All parts of the benchmark are declared via options, and their
values. The options can occur in any order. The accepted options are:

[list_begin options]
[opt_def -body script]

The argument of this option declares the body of the benchmark, the
Tcl script whose performance we wish to measure. This option, and
[option -desc], are the two required parts of each benchmark.

[opt_def -desc msg]

The argument of this option declares the name of the benchmark. It has
to be unique, or timing data from different benchmarks will be mixed
together.

[para]

[emph Beware!] This requirement is not checked when benchmarks are
executed, and the system will silently produce bogus data. This
option, and [option -body], are the two required parts of each
benchmark.

[opt_def -ipost script]

The argument of this option declares a script which is run immediately
[emph after] each iteration of the body. Its responsibility is to
release resources created by the body, or [option -ipre]-bodym which
we do not wish to live into the next iteration.

[opt_def -ipre script]

The argument of this option declares a script which is run immediately
[emph before] each iteration of the body. Its responsibility is to
create the state of the system expected by the body so that we measure
the right thing.

[opt_def -iterations num]

The argument of this option declares the maximum number of times to
run the [option -body] of the benchmark. During execution this and the
global maximum number of iterations are compared and the smaller of
the two values is used.

[para]

This option should be used only for benchmarks which are expected or
known to take a long time per run. I.e. reduce the number of times
they are run to keep the overall time for the execution of the whole
benchmark within manageable limits.

[opt_def -post script]

The argument of this option declares a script which is run
[emph after] all iterations of the body have been run. Its
responsibility is to release resources created by the body,
or [option -pre]-body.

[opt_def -pre script]

The argument of this option declares a script which is run
[emph before] any of the iterations of the body are run. Its
responsibility is to create whatever resources are needed by the body
to run without failing.

[list_end]
[list_end]

[section {BUGS, IDEAS, FEEDBACK}]

This document, and the package it describes, will undoubtedly contain
bugs and other problems.

Please report such in the category [emph bench] of the
[uri {http://sourceforge.net/tracker/?group_id=12883} {Tcllib SF Trackers}].

Please also report any ideas for enhancements you may have for either
package and/or documentation.

[see_also bench_intro]
[see_also bench_lang_intro]
[keywords testing performance benchmark {bench language} specification]
[manpage_end]