Mercurial > audio-send

comparison cleanup-message.txt @ 0:f9476ff7637e

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
initial forking of open-al to create multiple listeners
author Robert McIntyre <rlm@mit.edu>
date Tue, 25 Oct 2011 13:02:31 -0700
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:f9476ff7637e
1 My name is Robert McIntyre. I am seeking help packaging some changes
2 I've made to open-al.
3
4 * tl;dr how do I distribute changes to open-al which involve adding a
5 new device?
6
7 * Background / Motivation
8
9 I'm working on an AI simulation involving multiple listeners, where
10 each listener is a separate AI entity. Since each AI entity can move
11 independently, I needed to add multiple listener functionality to
12 open-al. Furthermore, my entire simulation allows time to dilate
13 depending on how hard the entities are collectively "thinking," so
14 that the entities can keep up with the simulation. Therefore, I
15 needed to have a system that renders sound on-demand instead of trying
16 to render in real time as open-al does.
17
18 I don't need any of the more advanced effects, just 3D positioning,
19 and I'm only using open-al from jMonkeyEngine3 that uses the LWJGL
20 bindings that importantly only allow access to one and only one
21 context.
22
23 Under these constraints, I made a new device which renders sound in
24 small, user-defined increments. It must be explicitly told to render
25 sound or it won't do anything. It maintains a separate "auxiliary
26 context" for every additional listener, and syncs the sources from the
27 LWJGL context whenever it is about to render samples. I've tested it
28 and it works quite well for my purposes. So far, I've gotten 1,000
29 separate listeners to work in a single simulation easily.
30
31 The code is here:
32 http://aurellem.localhost/audio-send/html/send.html
33 No criticism is too harsh!
34
35 Note that the java JNI bindings that are part of the device code right
36 now would be moved to a separate file the the same way that LWJGL does
37 it. I left them there for now to show how the device might be used.
38
39 Although I made this for my own AI work, it's ideal for recording
40 perfect audio from a video game to create trailers/demos, since the
41 computer doesn't have to try to record the sound in real time. This
42 device could be used to record audio in any system that wraps open-al
43 and only exposes one context, which is what many wrappers do.
44
45
46 * Actual Question
47
48 My question is about packaging --- how do you recommend I distribute
49 my new device? I got it to work by just grafting it on the open-al's
50 primitive object system, but this requires quite a few changes to main
51 open-al source files, and as far as I can tell requires me to
52 recompile open-al against my new device.
53
54 I also don't want the user to be able to hide my devices presence
55 using their ~/.alsoftrc file, since that gets in the way of easy
56 recording when the system is wrapped several layers deep, and they've
57 already implicitly requested my device anyway by using my code in the
58 first place.
59
60 The options I have thought of so far are:
61
62 1.) Create my own C-artifact, compiled against open-al, which somehow
63 "hooks in" my device to open-al and forces open-al to use it to the
64 exclusion of all other devices. This new artifact would have bindings
65 for java, etc. I don't know how to do this, since there doesn't seem
66 to be any way to access the list of devices in Alc/ALc.c for example.
67 In order to add a new device to open-al I had to modify 5 separate
68 files, documented here:
69
70 http://aurellem.localhost/audio-send/html/add-new-device.html
71
72 and there doesn't seem to be a way to get the same effect
73 programmatically.
74
75 2.) Strip down open-al to a minimal version that only has my device
76 and deal with selecting the right open-al library at a higher level,
77 depending on whether the user wants to record sound or actually hear
78 it. I don't like this because I can't easily benefit from
79 improvements in the main open-al distribution. It also would involve
80 more significant modification to jMonkeyEngine3's logic which selects
81 the appropriate C-artifact at runtime.
82
83 3.) Get this new device added to open-al, and provide a java wrapper
84 for it in a separate artifact. Problem here is that this device does
85 not have the same semantics as the other devices --- it must be told
86 to render sound, doesn't support multiple user-created contexts, and
87 it exposes extra functions for retrieving the rendered sounds. It also
88 might be too "niche" for open-al proper.
89
90 4.) Maybe abandon the device metaphor and use something better suited
91 to my problem that /can/ be done as in (1)?
92
93
94 I'm sure someone here knows enough about open-al's devices to give me
95 a better solution than these 4! All help would be most appreciated.
96
97 sincerely,
98 --Robert McIntyre
99
100