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 | <header> |
---|
5 | <title>Hoard Frequently Asked Questions</title> |
---|
6 | </header> |
---|
7 | |
---|
8 | <body> |
---|
9 | <section> |
---|
10 | <title> |
---|
11 | What kind of applications will Hoard speed up? |
---|
12 | </title> |
---|
13 | <p> |
---|
14 | Hoard will always improve the performance of multithreaded programs |
---|
15 | running on multiprocessors that make frequent use of the heap (calls |
---|
16 | to malloc/free or new/delete, as well as many STL functions). Because |
---|
17 | Hoard avoids false sharing, Hoard also speeds up programs that only |
---|
18 | occasionally call heap functions but access these objects frequently. |
---|
19 | </p> |
---|
20 | </section> |
---|
21 | |
---|
22 | |
---|
23 | <section> |
---|
24 | <title> |
---|
25 | I'm using the STL but not seeing any performance improvement. Why not? |
---|
26 | </title> |
---|
27 | <p> |
---|
28 | In order to benefit from Hoard, you have to tell STL to use malloc instead of its internal custom memory allocator: |
---|
29 | </p> |
---|
30 | |
---|
31 | <source> |
---|
32 | typedef list<unsigned int, malloc_alloc> mylist; |
---|
33 | </source> |
---|
34 | |
---|
35 | </section> |
---|
36 | |
---|
37 | |
---|
38 | <section><title> |
---|
39 | What systems does Hoard work on? |
---|
40 | </title> |
---|
41 | <p> |
---|
42 | Hoard has been successfully tested on numerous Windows, Linux and |
---|
43 | Solaris systems, including a 4-processor x86 box running Windows |
---|
44 | NT/2000, a 4-processor x86 box running RedHat Linux 6.0 and 6.1, and a |
---|
45 | 16-processor Sun Enterprise server running Solaris. |
---|
46 | </p> |
---|
47 | </section> |
---|
48 | |
---|
49 | |
---|
50 | <section><title> |
---|
51 | Have you compared Hoard with SmartHeap SMP? |
---|
52 | </title> |
---|
53 | <p> |
---|
54 | We tried SmartHeap SMP but it did not work on our Suns (due to an |
---|
55 | apparent race condition in the code). |
---|
56 | </p> |
---|
57 | </section> |
---|
58 | |
---|
59 | |
---|
60 | <section> |
---|
61 | <title> |
---|
62 | Have you compared Hoard against mtmalloc or libumem? |
---|
63 | </title> |
---|
64 | <p> |
---|
65 | Yes. Hoard is much faster than either. For example, here's an |
---|
66 | execution of threadtest on Solaris: |
---|
67 | </p> |
---|
68 | <table> |
---|
69 | <tr> |
---|
70 | <td>Default:</td> |
---|
71 | <td>4.60 seconds</td> |
---|
72 | </tr> |
---|
73 | <tr> |
---|
74 | <td>Libmtmalloc:</td> |
---|
75 | <td>6.23 seconds</td> |
---|
76 | </tr> |
---|
77 | <tr> |
---|
78 | <td>Libumem:</td> |
---|
79 | <td>5.47 seconds</td> |
---|
80 | </tr> |
---|
81 | <tr> |
---|
82 | <td>Hoard 3.2:</td> |
---|
83 | <td>1.99 seconds</td> |
---|
84 | </tr> |
---|
85 | </table> |
---|
86 | </section> |
---|
87 | |
---|
88 | <section> |
---|
89 | <title> |
---|
90 | Can I use Hoard with a commercial application? |
---|
91 | </title> |
---|
92 | <p> |
---|
93 | Yes. Hoard is available under two licenses. The first license is the |
---|
94 | GNU General Public License. That license is free, but it requires you |
---|
95 | to open-source your application. The second option is to purchase a |
---|
96 | license from the University of Texas at Austin. Click the "Licensing Info" |
---|
97 | link on the left for more information. |
---|
98 | </p> |
---|
99 | |
---|
100 | </section> |
---|
101 | |
---|
102 | </body> |
---|
103 | </document> |
---|