[Ffmpeg-devel-irc] ffmpeg-devel.log.20170717

[burek]

[13:07:12 CEST] <J_Darnley> atomnuker: the vc2 encoder does not have a constant quantiser mode, right?
[13:09:24 CEST] <J_Darnley> To add that I would need to "remove" the rate control, select the quant I will use, but I still need to calculate how many bytes the bistream will write for every slice, right?
[16:17:28 CEST] <J_Darnley> Okay I am almost equally lost in the quantising as I am in the rest of the code.
[16:17:53 CEST] <J_Darnley> I can see the encoder counting bits
[16:18:10 CEST] <J_Darnley> I can see it using quantiser(s)
[16:19:00 CEST] <J_Darnley> but I can't see where they come from
[16:20:36 CEST] <J_Darnley> There seems to be tables of them.  One for every level and maybe one for each orientation or subband
[16:53:55 CEST] <kierank> J_Darnley: comes from ratecontrol I guess
[16:58:53 CEST] <atomnuker> J_Darnley: quantization _indices_ range from 0 to 116-ish(check the size of the table)
[16:59:01 CEST] <atomnuker> they look up quantizers
[16:59:09 CEST] <atomnuker> the quantizers are used to quantize
[16:59:32 CEST] <atomnuker> they're fairly large non-linear numbers
[16:59:53 CEST] <atomnuker> so to save a few instructions they're pre-generated in a lookup table
[17:00:34 CEST] <atomnuker> btw notice the spec says nothing about wavelet depth limit or coefficient intermediates sizes
[17:00:45 CEST] <atomnuker> they can be arbitrarily big
[17:01:34 CEST] <atomnuker> but for all practical purposes qindices above 90ish are useless with 32 bit wide coeffs and wavelet depth of 5
[17:01:51 CEST] <atomnuker> (the greater the depth the larger the coefficients)
[17:02:49 CEST] <atomnuker> rip out the rate control and the bit spreading function and make each slice use a constant one, shouldn't be hard at all
[17:05:17 CEST] <J_Darnley> okay thank you
[20:42:50 CEST] <J_Darnley> Oh fun!
[20:43:18 CEST] <J_Darnley> "Internal error, put_bits buffer too small"
[20:49:41 CEST] <J_Darnley> aaaaahhhh!  Is that why that memset is there?
[20:50:57 CEST] <kierank> yes to zero out the rest of the coeffs
[20:51:12 CEST] <J_Darnley> No, this is a different memset
[20:51:17 CEST] <kierank> ah
[20:51:41 CEST] <J_Darnley> I am using a cache of slice sizes that I think is uninitialised
[20:51:51 CEST] <J_Darnley> or might be cleared every frame
[20:52:29 CEST] <atomnuker> J_Darnley: no, just increase the slice size in bits
[20:52:47 CEST] <atomnuker> you're on a constant quantizer, you'd need to encode twice to get the bit cost
[20:53:07 CEST] <atomnuker> have huge amounts of fun figuring out how much to pad each cost in bits
[20:54:05 CEST] <J_Darnley> Then what the heck is this stuff I copy-pasted?
[20:54:41 CEST] <J_Darnley> ... out of count_hq_slice
[20:54:52 CEST] <J_Darnley> Is that not the bit cost?
[21:09:04 CEST] <J_Darnley> ugh.  removing the use of that cache just changed the problem
[21:10:29 CEST] <J_Darnley> oh that might be because I made a different bug by commenting out a line.
[21:12:36 CEST] <J_Darnley> ha ha ha
[21:12:47 CEST] <J_Darnley> quantiser index 25 looks pretty bad
[21:12:57 CEST] <J_Darnley> and the colour is washed out
[21:13:09 CEST] <J_Darnley> now didn't I see an option for that?
[21:31:58 CEST] <atomnuker> J_Darnley: you might also notice qindices of 1-4 actually take _more_ bits to encode rather than less
[21:33:23 CEST] <J_Darnley> I didn't see that yet but now I will look for it
[22:06:00 CEST] <iive> atomnuker: do you have some free time to run some tests?
[22:06:34 CEST] <atomnuker> in an hour or so
[22:06:53 CEST] <iive> great :D
[22:12:43 CEST] <durandal_170> atomnuker: will you finish audio denoiser?
[22:37:02 CEST] <atomnuker> I need to deal with a lot of things, I'm still not feeling great nor in a great place but if I feel like it I'll finish it
[22:39:00 CEST] <iive> you are not well?!
[22:39:20 CEST] <atomnuker> first my throat went bad for 3 weeks
[22:40:02 CEST] <atomnuker> then just as it started to get well after a course of antibiotics one of my ear tubes got inflamated and it hurt a bit and didn't sound right
[22:40:32 CEST] <atomnuker> after treating it with cold water and pills for sore throat the fucker went after my right ear
[22:41:19 CEST] <atomnuker> so I had to suffer through another round of ear pain (which only occured just before going to bed) and high frequency muffling
[22:42:52 CEST] <atomnuker> and now I think I have a slight tooth ache which is making me paranoid
[22:43:22 CEST] <atomnuker> I can't get a fucking break from feeling bad
[22:43:40 CEST] <iive> well, at least uk have good and free healthacre
[22:43:53 CEST] <iive> but yeh, these are quite bad symptoms
[22:43:54 CEST] <atomnuker> they did fuck all
[22:44:14 CEST] <atomnuker> no, they're not, so I was told not to bother doing much and that it'll pass
[22:44:15 CEST] <iive> you think antibiotics made it worse?
[22:44:24 CEST] <atomnuker> no, they made it better, it just spread
[22:46:53 CEST] <iive> yeh, it sounds like infection, so antibiotics should handle it well. Just don't forget to take some pro-biotics too, if theya re not already combined.
[22:49:52 CEST] <Compnn> atomnuker : i remember reading something about taking ear wax from an ear that didnt have an ear ache to an ear that did ...
[22:49:56 CEST] <Compnn> dunno if it will help :D
[22:49:58 CEST] <J_Darnley> atomnuker: I now feel a little bad for bothering you with my stupid dirac questions
[22:51:06 CEST] <atomnuker> Compnn: while my right ear was inflamated I couldn't get any out, after it went away a few days ago I pulled out quite a lot
[22:52:57 CEST] <atomnuker> J_Darnley: bah, its fine, at least I haven't had anything actually serious in a while *knocks on wood*
[22:53:23 CEST] <J_Darnley> Well that's something good then
[22:59:12 CEST] <iive> ear wax has some anti-microbial properties, so cleaning it too much can actually be bad thing. but that's not the case here
[23:00:06 CEST] <iive> it sounds more like infection going into the sinuses and internal ear. and that's why i said it is bad symptoms
[23:00:11 CEST] <Compnn> never stick anything into your ear :(
[23:00:15 CEST] <Compnn> sinus infection yes :(
[00:00:00 CEST] --- Tue Jul 18 2017