annotate doc/EFR-rationale @ 202:0faf23a9286f

libgsmefr/reorder.c: perf opt
author Mychaela Falconia <falcon@freecalypso.org>
date Wed, 04 Jan 2023 04:54:02 +0000
parents 3a0ee08a3b9d
children 69b9a1eeb5a2
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
122
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
1 Problem in need of solving
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
2 ==========================
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
3
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
4 At the time of the undertaking of Themyscira libgsmefr project (late 2022),
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
5 there did not exist any readily available library solution for GSM EFR codec.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
6 The community of FOSS offers classic libgsm from 1990s for FR1 codec (it's an
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
7 implementation of GSM 06.10, on top of which we had to implement our own Rx DTX
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
8 handler) and opencore-amrnb for AMR (based on Android OpenCORE framework) - but
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
9 nothing for EFR. This situation creates a problem for anyone seeking to deploy
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
10 their own GSM network with a voice interface to PSTN or other networks: such
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
11 voice interface generally requires implementing a transcoder, and doing the
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
12 latter in turn requires a library that implements the codec to be supported.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
13 In the present situation, anyone who wishes to implement a speech transcoder
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
14 for GSM networks can easily support FR1 and AMR codecs, but not EFR.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
15
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
16 EFR is more than just 12k2 mode of AMR!
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
17 =======================================
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
18
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
19 There is a common misconception in the GSM hacker community that EFR is nothing
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
20 but the highest 12k2 mode of AMR, and that any library that implements AMR,
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
21 such as opencore-amrnb, is thus sufficient to support EFR as well. However,
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
22 the reality is more complex:
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
23
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
24 * If an AMR encoder operates with DTX disabled, such that the output contains
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
25 only speech frames and no SID, and the mode is forced to 12k2, then indeed a
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
26 simple reshuffling of bits will produce speech frames that can be fed to an
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
27 EFR decoder on the other end. Note that the two encoders (EFR and AMR 12k2)
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
28 will produce *different* encoded speech parameters from the same input, and
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
29 the decoded speech output on the other end will also be different, but the
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
30 two versions are expected to be equally good for human ears.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
31
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
32 * In the other direction, if an EFR input stream contains only good speech
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
33 frames (no SID and no lost, FACCH-stolen or DTX-suppressed frames), one can
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
34 likewise do a simple bit reordering and feed these frames to an AMR decoder.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
35 The output of this AMR decoder will once again be different from a proper
181
3a0ee08a3b9d doc/EFR-rationale: typo fix
Mychaela Falconia <falcon@freecalypso.org>
parents: 127
diff changeset
36 (bit-exact) EFR decoder for the same speech parameter inputs, but as long as
122
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
37 the EFR input stream is all good speech frames, the output will be good enough
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
38 for human ears.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
39
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
40 * The real problem occurs when the EFR input stream contains SID frames and BFI
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
41 frame gaps, as will always happen in reality if this stream is an uplink from
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
42 a GSM call. AMR SID mechanism is different from that of EFR, and an AMR
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
43 decoder will NOT recognize EFR SID frames. A quick experiment confirms that
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
44 when a real GSM EFR uplink RTP capture is converted to AMR by non-SID-aware
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
45 bit reshuffling and then fed to amrnb-dec from opencore-amrnb, unpleasant
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
46 sounds appear in the output whenever GSM uplink goes into SID.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
47
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
48 EFR reference code from ETSI
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
49 ============================
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
50
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
51 A published-source bit-exact implementation of GSM EFR encoder and decoder,
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
52 complete with all beyond-speech functions of DTX, VAD, comfort noise generation,
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
53 error concealment etc does exist in the form of reference code from ETSI.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
54 However, this code has never been turned into a usable codec library by anyone
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
55 prior to us (at least not by anyone who freely published their work), and doing
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
56 such librification (producing an EFR analogue to what Android OpenCORE people
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
57 did with AMR) is no easy feat! The original EFR code from ETSI exhibits two
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
58 problems which need to be remedied in the librification project:
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
59
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
60 1) The original code maintains all codec state in global variables (lots of
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
61 them) that are scattered throughout. 3GPP reference code for AMR (naturally
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
62 later than EFR in chronological order) is better in this regard (in the AMR
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
63 version they gathered their global vars into structs and pass pointers to
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
64 these structs, although still many separately-malloc'ed structs instead of
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
65 single unified encoder state and decoder state), but we need the EFR version
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
66 for correct handling of all beyond-speech aspects, and this version is all
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
67 global vars.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
68
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
69 2) These reference codes from ETSI/3GPP (both EFR and AMR versions, it seems)
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
70 were intended to serve as simulations, not as production code, and the code
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
71 is very inefficient.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
72
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
73 Themyscira libgsmefr
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
74 ====================
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
75
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
76 Libgsmefr presented in this code repository is our current solution for EFR.
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
77 It is a library styled after classic libgsm for FR1, but its guts consist of a
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
78 librified derivative of ETSI EFR code. The problem of global vars has been
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
79 solved in this library version - they've been gathered into one unified struct
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
80 for encoder state and another unified struct for decoder state - but the problem
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
81 of poor performance (significantly worse than opencore-amrnb) still remains for
b33f2168fdec doc/EFR-rationale article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
82 now.
127
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
83
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
84 Future roadmap
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
85 ==============
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
86
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
87 If someone is implementing a DSP vocoder block for a GSM MS or a network-side
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
88 speech transcoder that needs to support all standard GSM codecs, at some point
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
89 they will need to implement both EFR and AMR. Given the close relation between
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
90 these two codecs (they are not perfectly compatible as we started out saying,
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
91 but they are still very closely related), keeping two entirely separate library
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
92 implementations for AMR and EFR will be very inefficient in the long run, and a
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
93 nightmare to get them to perform equally well. It seems to me (Mother Mychaela)
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
94 that the correct solution will be to produce a single codec library that
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
95 implements both AMR and EFR, probably by starting with an AMR library and
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
96 extending it with special modes to handle those aspects where EFR differs. It
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
97 is my forecast that we are going to end up doing something along these lines in
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
98 Themyscira - but it will be much later down the road; for the time being, our
4af99bf8671a doc/EFR-rationale: add future roadmap section
Mychaela Falconia <falcon@freecalypso.org>
parents: 122
diff changeset
99 initial version of ThemWi will only support FR and EFR, but not AMR.