src/third_party/icu/source/test/perf/collationperf/readme.html

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>

<HEAD>
	<META HTTP-EQUIV="Content-Type" CONTENT="text/html;CHARSET=iso-8859-1">
        <meta name="Copyright" content="Copyright (c) 2002-2008, International Business Machines Corporation and others. All Rights Reserved.">
	<META NAME="GENERATOR" Content="Visual Page 2.0 for Windows">
	<TITLE>collperf</TITLE>
</HEAD>

<BODY BGCOLOR="white">

<H2>collperf</H2>
<P>collperf is a test program for comparing collation performance and key lengths of ICU, Windows native collation
and Unix/POSIX collation. It operates on a file of lines (names, for example), and performs one of three tests:</P>

<OL>
	<LI>Sort Key generation. Report on key lengths and key generation times.
	<LI>Binary search. Report the average time required to look up each of the names (file lines) from the file in
	a sorted list of all of the names.
	<LI>Quick Sort. Report the time required to sort the file in memory, using the C library qsort function. The file
	order is randomized prior to the sort.
</OL>

<P>
<P><B>Usage Summary</B></P>

<BLOCKQUOTE>
	<P>
	<TABLE BORDER="0" WIDTH="100%" BGCOLOR="#FFFBF0">
		<TR>
			<TD WIDTH="100%">
				<P><TT>collperf -help</TT>
				<BLOCKQUOTE>
					<PRE><TT>Usage: strperf options...
-help                  Display this message.
-file file_name        utf-16 format file of names
-locale name           ICU locale to use. Default is en_US
-langid 0x1234         Windows Language ID number. Default 0x409 (en_US)
                       see http://msdn.microsoft.com/library/psdk/winbase/nls_8xo3.htm
-win                   Run test using Windows native services. (ICU is default)
-unix                  Run test using Unix strxfrm, strcoll services.
-uselen                Use API with string lengths. Default is null-terminated strings
-usekeys               Run tests using sortkeys rather than strcoll
-loop nnnn             Loopcount for test. Adjust for reasonable total running time.
-terse                 Terse numbers-only output. Intended for use by scripts.
-french                French accent ordering
-norm                  Normalizing mode on
-shifted               Shifted mode
-lower                 Lower case first
-upper                 Upper case first
-case                  Enable separate case level
-level n               Sort level, 1 to 5, for Primary, Secndary, Tertiary, Quaternary, Identical
-binsearch             Binary Search timing test
-keygen                Sort Key Generation timing test
-qsort                 Quicksort timing test</TT></PRE>
				</BLOCKQUOTE>
			</TD>
		</TR>
	</TABLE>
</P>
</BLOCKQUOTE>

<P><B>Example</B></P>

<BLOCKQUOTE>
	<P>
	<TABLE BORDER="0" WIDTH="100%" BGCOLOR="#FFFBF0">
		<TR>
			<TD WIDTH="100%"><TT>C:\&gt;collperf -loop 200 -file latin.txt -keygen -shifted -level 4<BR>
				file &quot;latin.txt&quot;, 7604 lines.<BR>
				Sort Key Generation: total # of keys = 197704<BR>
				Sort Key Generation: time per key = 4253 ns<BR>
				Key Length / character = 1.730054</TT></TD>
		</TR>
	</TABLE>

</BLOCKQUOTE>

<P>

</BODY>

</HTML>