source: proiecte/swift/trunk/lib/hoard-371/doc/src/documentation/content/xdocs/screenshot.xml @ 176

Last change on this file since 176 was 176, checked in by (none), 14 years ago
  • imported repo from "guagal"
File size: 2.2 KB
Line 
1<?xml version="1.0" encoding="UTF-8"?>
2<!DOCTYPE document PUBLIC "-//APACHE//DTD Documentation V2.0//EN" "http://forrest.apache.org/dtd/document-v20.dtd">
3<document> 
4
5  <header> 
6    <title>Screenshots</title> 
7  </header> 
8
9<body>
10
11  <p>
12  Because Hoard is a transparent plug-in replacement, ordinary
13  screenshots don't make much sense, since the applications look and
14  work the same! However, it's easy to visualize Hoard's impact on
15  application performance. Here are a few example speedup
16  graphs. Ideally, the speedup as the number of processors increases
17  is a straight diagonal line going from the lower left of each graph
18  to the upper right. These benchmarks compare Hoard to Ptmalloc (the
19  Linux allocator), mtmalloc version 8, and the Solaris default
20  allocator, and are run on a 14-processor Sun.
21</p>
22
23  <table>
24  <tr>
25    <td>
26    <!-- <center> -->
27    <img alt="Threadtest graph" src="images/plot-threadtest-speedup.gif"/>
28    <!-- </center> -->
29    </td>
30  </tr>
31  <tr>
32    <td>
33    <!-- <center> -->
34<p>
35    This benchmark has multiple threads, each allocating and freeing
36    objects. It stresses raw scalability of allocation in individual
37    threads. In this case, Hoard achieves a linear speedup. Ptmalloc
38    scales to four processors but then runs out of steam.
39</p>
40    <!-- </center> -->
41    </td>
42  </tr>
43  <tr>
44    <td>
45    <!-- <center> -->
46    <img alt="Larson graph" src="images/plot-larson-speedup.gif"/>
47    <!-- </center> -->
48    </td>
49  </tr>
50  <tr>
51    <td>
52    <!-- <center> -->
53<p>
54    In this benchmark, multiple threads allocate objects and pass some
55    objects to other threads that free them. This stresses the ability
56    of the memory manager to handle object sharing. Only Hoard scales
57    for this case.
58</p>
59    <!-- </center> -->
60    </td>
61  </tr>
62  <tr>
63    <td>
64    <!-- <center> -->
65    <img alt="Cache-scratch graph" src="images/plot-cache-scratch-speedup.gif"/>
66    <!-- </center> -->
67    </td>
68  </tr>
69  <tr>
70    <td>
71    <!-- <center> -->
72<p>
73    A test of whether a memory manager avoids causing false sharing of
74    cache lines (an insidious effect that destroys scalability). The
75    linear speedup shows that Hoard avoids false sharing completely.
76</p>
77    <!-- </center> -->
78    </td>
79  </tr>
80  </table>
81
82</body>
83</document>
Note: See TracBrowser for help on using the repository browser.