l4/pkg/libstdc++-v3/contrib/libstdc++-v3-4.4/doc/html/ext/pb_ds/tree_random_int_find_find_timing_test.html

   1 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
   2     "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
   3
   4 <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
   5 <head>
   6 <meta name="generator" content="HTML Tidy for Linux/x86 (vers 12 April 2005), see www.w3.org" />
   7 <title>Tree Text Find Timing Test</title>
   8 <meta http-equiv="Content-Type" content="text/html; charset=us-ascii" />
   9 </head>
  10 <body>
  11 <div id="page">
  12 <h1>Tree-Based and Trie-Based Text <tt>find</tt> Find Timing
  13     Test</h1>
  14 <h2><a name="description" id="description">Description</a></h2>
  15 <p>This test inserts a number of values with keys from an
  16     arbitrary text ([<a href="references.html#wickland96thirty">wickland96thirty</a>]) into
  17     a container, then performs a series of finds using
  18     <tt>find</tt>. It measures the average time for <tt>find</tt>
  19     as a function of the number of values inserted.</p>
  20 <p>(The test was executed with <a href="http://gcc.gnu.org/viewcvs/*checkout*/trunk/libstdc%2B%2B-v3/testsuite/performance/ext/pb_ds/text_find_timing.cc"><tt>text_find_timing_test</tt></a>
  21     thirty_years_among_the_dead_preproc.txt 200 200 2100)</p>
  22 <h2><a name="purpose" id="purpose">Purpose</a></h2>
  23 <p>The test checks the effect of different underlying
  24     data structures.</p>
  25 <h2><a name="results" id="results">Results</a></h2>
  26 <p>Figures <a href="#NTTG">NTTG</a>, <a href="#NTTM">NTTM</a>,
  27     and <a href="#NTTG">NTTL</a> show the results for the native,
  28     tree-based, and trie-based types in <a href="assoc_performance_tests.html#gcc"><u>g++</u></a>, <a href="assoc_performance_tests.html#local"><u>local</u></a>, and
  29     <a href="assoc_performance_tests.html#local"><u>local</u></a>,
  30     respectively.</p>
  31 <div id="NTTG_res_div">
  32 <div id="NTTG_gcc">
  33 <div id="NTTG_random_int_find_find_timing_test_tree">
  34 <div id="NTTG_assoc">
  35 <div id="NTTG_Native_456_tree-based_456_random_int_find_timing_test_using__tt_find_455tt_"><div style="border-style: dotted; border-width: 1px; border-color: lightgray"><h6 class="c1"><a name="NTTG" id="NTTG"><img src="random_int_find_find_timing_test_tree_gcc.png" alt="no image" /></a></h6>NTTG: Native, tree-based, random int find timing test using <tt>find</tt> - <a href="assoc_performance_tests.html#gcc">g++</a><p>In the above figure, the names in the legends have the following meaning:</p>
  36 <ol>
  37 <li>
  38 splay_tree_map-
  39 <a href="tree.html"><tt>tree</tt></a>
  40  with <tt>Tag</tt> = <a href="splay_tree_tag.html"><tt>splay_tree_tag</tt></a>
  41 , and <tt>Node_Update</tt> = <a href="null_tree_node_update.html"><tt>null_tree_node_update</tt></a>
  42 </li>
  43 <li>
  44 ov_tree_map-
  45 <a href="tree.html"><tt>tree</tt></a>
  46  with <tt>Tag</tt> = <a href="ov_tree_tag.html"><tt>ov_tree_tag</tt></a>
  47 , and <tt>Node_Update</tt> = <a href="null_tree_node_update.html"><tt>null_tree_node_update</tt></a>
  48 </li>
  49 <li>
  50 n_map-
  51 <tt>std::map</tt></li>
  52 <li>
  53 rb_tree_map-
  54 <a href="tree.html"><tt>tree</tt></a>
  55  with <tt>Tag</tt> = <a href="rb_tree_tag.html"><tt>rb_tree_tag</tt></a>
  56 , and <tt>Node_Update</tt> = <a href="null_tree_node_update.html"><tt>null_tree_node_update</tt></a>
  57 </li>
  58 </ol>
  59 </div><div style="width: 100%; height: 20px"></div></div>
  60 </div>
  61 </div>
  62 <div id="NTTM_res_div">
  63 <div id="NTTM_msvc">
  64 <div id="NTTM_random_int_find_find_timing_test_tree">
  65 <div id="NTTM_assoc">
  66 <div id="NTTM_Native_456_tree-based_456_random_int_find_timing_test_using__tt_find_455tt_"><div style="border-style: dotted; border-width: 1px; border-color: lightgray"><h6 class="c1"><a name="NTTM" id="NTTM"><img src="random_int_find_find_timing_test_tree_msvc.png" alt="no image" /></a></h6>NTTM: Native, tree-based, random int find timing test using <tt>find</tt> - <a href="assoc_performance_tests.html#msvc">msvc++</a><p>In the above figure, the names in the legends have the following meaning:</p>
  67 <ol>
  68 <li>
  69 splay_tree_map-
  70 <a href="tree.html"><tt>tree</tt></a>
  71  with <tt>Tag</tt> = <a href="splay_tree_tag.html"><tt>splay_tree_tag</tt></a>
  72 , and <tt>Node_Update</tt> = <a href="null_tree_node_update.html"><tt>null_tree_node_update</tt></a>
  73 </li>
  74 <li>
  75 ov_tree_map-
  76 <a href="tree.html"><tt>tree</tt></a>
  77  with <tt>Tag</tt> = <a href="ov_tree_tag.html"><tt>ov_tree_tag</tt></a>
  78 , and <tt>Node_Update</tt> = <a href="null_tree_node_update.html"><tt>null_tree_node_update</tt></a>
  79 </li>
  80 <li>
  81 n_map-
  82 <tt>std::map</tt></li>
  83 <li>
  84 rb_tree_map-
  85 <a href="tree.html"><tt>tree</tt></a>
  86  with <tt>Tag</tt> = <a href="rb_tree_tag.html"><tt>rb_tree_tag</tt></a>
  87 , and <tt>Node_Update</tt> = <a href="null_tree_node_update.html"><tt>null_tree_node_update</tt></a>
  88 </li>
  89 </ol>
  90 </div><div style="width: 100%; height: 20px"></div></div>
  91 </div>
  92 </div>
  93 <div id="NTTL_res_div">
  94 <div id="NTTL_local">
  95 <div id="NTTL_random_int_find_find_timing_test_tree">
  96 <div id="NTTL_assoc">
  97 <div id="NTTL_Native_456_tree-based_456_random_int_find_timing_test_using__tt_find_455tt_"><div style = "border-style: dotted; border-width: 1px; border-color: lightgray"><h6 class="c1"><a name="NTTL" id= "NTTL"><img src="random_int_find_find_timing_test_tree_local.png" alt="no image" /></a></h6>NTTL: Native, tree-based, random int find timing test using <tt>find</tt> - <a href = "assoc_performance_tests.html#local">local</a></div><div style = "width: 100%; height: 20px"></div></div>
  98 </div>
  99 </div>
 100 <h2><a name="observations" id="observations">Observations</a></h2>
 101 <p>For this setting, a splay tree (<a href="tree.html"><tt>tree</tt></a>
 102                 with <tt>Tag =</tt> <a href="splay_tree_tag.html"><tt>splay_tree_tag</tt></a>)
 103                 does not do well. This is possibly due to two
 104                 reasons:</p>
 105 <ol>
 106 <li>A splay tree is not guaranteed to be balanced
 107                   [<a href="references.html#motwani95random">motwani95random</a>].
 108                   If a splay tree contains <i>n</i> nodes, its
 109                   average root-leaf path can be <i>m &gt;&gt;
 110                   log(n)</i>.</li>
 111 <li>Assume a specific root-leaf search path has
 112                   length <i>m</i>, and the search-target node has
 113                   distance <i>m'</i> from the root. A red-black
 114                   tree will require <i>m + 1</i> comparisons to
 115                   find the required node; a splay tree will require
 116                   <i>2 m'</i> comparisons. A splay tree,
 117                   consequently, can perform many more comparisons
 118                   than a red-black tree.</li>
 119 </ol>
 120 <p>An ordered-vector tree (<a href="tree.html"><tt>tree</tt></a>
 121                 with <tt>Tag =</tt> <a href="ov_tree_tag.html"><tt>ov_tree_tag</tt></a>),
 122                 a red-black tree (<a href="tree.html"><tt>tree</tt></a>
 123                 with <tt>Tag =</tt> <a href="splay_tree_tag.html"><tt>rb_tree_tag</tt></a>),
 124                 and the native red-black tree all share
 125                 approximately the same performance.</p>
 126 <p>An ordered-vector tree is slightly slower than
 127                 red-black trees, since it requires, in order to
 128                 find a key, more math operations than they do.
 129                 Conversely, an ordered-vector tree requires far
 130                 lower space than the others. ([<a href="references.html#austern00noset">austern00noset</a>],
 131                 however, seems to have an implementation that is
 132                 also faster than a red-black tree).</p>
 133 <p>A PATRICIA trie (<a href="trie.html"><tt>trie</tt></a>
 134                 with <tt>Tag =</tt> <a href="pat_trie_tag.html"><tt>pat_trie_tag</tt></a>)
 135                 has good look-up performance, due to its large
 136                 fan-out in this case. In this setting, a PATRICIA
 137                 trie has lookup performance comparable to a hash
 138                 table (see <a href="hash_text_find_find_timing_test.html">Hash-Based
 139                 Text <tt>find</tt> Find Timing Test</a>), but it is
 140                 order preserving. This is not that surprising,
 141                 since a large fan-out PATRICIA trie works like a
 142                 hash table with collisions resolved by a sub-trie.
 143                 A large fan-out PATRICIA trie does not do well on
 144                 modifications (see <a href="tree_text_insert_timing_test.html">Tree-Based and
 145                 Trie-Based Text Insert Timing Test</a>). It is
 146                 possibly beneficial to semi-static settings,
 147                 therefore.</p>
 148 <p><a href="assoc_performance_tests.html#tree_like_based_types">
 149                 Observations::Tree-Like-Based Container Types</a>
 150                 summarizes some observations on tree-based and
 151                 trie-based containers.</p>
 152 </div>
 153 </div>
 154 </div>
 155 </div>
 156 </div>
 157 </div>
 158 </div>
 159 </body>
 160 </html>