1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
|
-------
FFdecsa
-------
FFdecsa is a fast implementation of the CSA decryption algorithm for MPEG
TS packets.
Q: What does FF stands for?
A: FFdecsa means "Fucking Fast decsa".
Q: Why would you use such a rude name?
A: Because this code is fucking fast, more than 800% the speed of the best
implementation I'm able to find around at the moment.
Q: How it that possible? Are all other programmers stupid?
A: No, they just tried to save a cycle or two tweaking a fundamentally wrong
implementation. The algorithm has to be implemented in a totally different
way to achieve good speed.
Q: Do you use multimedia instructions?
A: I use every trick I could come up with, including multimedia instructions.
They are not fundamental in achieving speed, a version without them runs
at 6x the speed of the best implementation around (which uses MMX).
Q: So how did you do that?
A: By using a different approach for the implementation. This code is not
exploiting some new CSA vulnerability, it is just doing the same
calculations better. Think about replacing bubble sort with quick sort.
Q: You're joking, it's impossible to gain so much speed.
A: Speed test are available, technical documentation is available, source
code is available. Try it yourself.
If you want details, these are some of the documented tricks I used
(more details in the docs directory):
TRICK NUMBER 0: emulate the hardware
TRICK NUMBER 1: virtual shift registers
TRICK NUMBER 2: parallel bitslice
TRICK NUMBER 3: multimedia instructions
TRICK NUMBER 4: parallel byteslice
TRICK NUMBER 5: efficient bit permutation
TRICK NUMBER 6: efficient normal<->slice conversion
TRICK NUMBER 7: try hard to process packets together
TRICK NUMBER 8: try to avoid doing the same thing many times
TRICK NUMBER 9: compiler
TRICK NUMBER a: a lot of brain work
Q: How can be this code useful?
A: You can use this code in place of the old slow implementations and save a
lot of CPU power.
Q: Just that?
A: Well, new applications are possible.
Decrypting a whole transponder is easily doable now. Well, a $50 CPU can
decrypt four transponder at the same time if you have four DVB boards (but
I couldn't test that).
Q: You're cheating, this code is fake, I don't believe one word.
A: Go away. This is technical stuff for people with brains.
Q: This code is great, may I distribute your code in original or modified
form?
A: Only if you respect the license.
Q: May I use your code in my player/library/plugin...?
A: Again, you have to respect the license.
Q: Are you an extraterrestrial programmer?
A: No, just a Turkish guy with a PC to play with :-)
Q: Why did you spend your time doing this?
A: Because I thought that my approach was doable and I was sure it would
have been much faster, so I had to implement it to confirm I was right.
I got 8x the speed and that's enough to be proud of it. And I could not
just keep the code for myself only.
Q: What is the answer to the meaning of the universe?
A: 42,43,71,5f,65,85,f6,76,0d,13,28,96,...
|
