similar to: Question on new release.

Speex tells me that the decoder is always 5 ms, but it says that the encoder is 5 ms for NB, 8.9375 ms for WB, and 10.90625 ms for UWB. Is there an extra frame of delay in the encoder that isn't otherwise accounted for? John Ridges Jean-Marc Valin wrote: > Quoting John Ridges <jridges at masque.com>: > >> I also need to know the precise delays from Speex but I used

I got the approximation from a Google book: http://books.google.com/books?id=2CAqsF-RebgC&pg=PA385 Page 392, formula (10.33) Using this formula, you're right, hypergeom_gain() would *not* converge to 1 for large x, but would instead be gamma(1.25)/sqrt(sqrt(x)) which would approach zero. Now if the formula for the hypergeometric gain were instead gamma(1.5) * M(-.5;1;-x) / sqrt(x)

By my reckoning the confluent hypergoemetric functions should have the following values: M(-.25;1;-.5) = 1.11433 M(-.25;1;-1) = 1.21088 M(-.25;1;-1.5) = 1.29385 M(-.25;1;-2) = 1.36627 M(-.25;1;-2.5) = 1.43038 M(-.25;1;-3) = 1.48784 M(-.25;1;-3.5) = 1.53988 M(-.25;1;-4) = 1.58747 M(-.25;1;-4.5) = 1.63134 M(-.25;1;-5) = 1.67206 M(-.25;1;-5.5) = 1.71009 M(-.25;1;-6) = 1.74579 M(-.25;1;-6.5) =

Thanks, I look forward to seeing what you find out. BTW, I was wondering if you tried replacing the SIG2WORD16 macro using the vqmovns_s32 intrinsic? I'm sure it would be faster than the C code, but in the grand scheme of things it might not make much difference. On 11/13/2015 12:15 PM, Jonathan Lennox wrote: >> On Nov 13, 2015, at 1:51 PM, John Ridges <jridges at masque.com>

Hi Jonathan, I'm sorry to bring this up again, and I don't want to beat a dead horse, but I was very surprised by your benchmarks so I took a little closer look. I think what's happening is that it's a little unfair to compare the ARM64 inline assembly to the C code, because looking at the C macros in "fixed_generic.h" for MULT16_32_Q16 and MULT16_32_Q15 you find

Thanks for the confirmation Jean-Marc. I kind of suspected from the comments that it was the confluent hypergoemetric function, which I was trying to evaluate using Kummer's equation, namely: M(a;b;x) is the sum from n=0 to infinity of (a)n*x^n / (b)n*n! where (a)n = a(a+1)(a+2) ... (a+n-1) But when I use Kummer's equation, I don't get the values in the "hypergeom_gain"

JM, I also need to know the precise delays from Speex but I used the SPEEX_GET_LOOKAHEAD control requests to determine them (plus the "speex_resampler_get_output_latency" function from the resampler). The returned values from the Speex lookahead request don't seem to match with the values you gave Alexander. Am I doing this wrong? Thanks, John Ridges speex-dev-request at

I?ve tried adding support for OPUS_FAST_INT64 to celt/arch.h, and I?ve found that this is indeed comparable in speed, if not a touch faster, than my inline assembly. I?ll submit patches for this. The inline assembly parts of my aarch64 patch set can thus be considered withdrawn. I haven?t yet tried replacing SIG2WORD16 (or silk_ADD_SAT32/silk_SUB_SAT32) with Neon intrinsics. That?s an obvious

> On Nov 19, 2015, at 5:47 PM, John Ridges <jridges at masque.com> wrote: > > Any speedup from the intrinsics may just be swamped by the rest of the encode/decode process. But I think you really want SIG2WORD16 to be (vqmovns_s32(PSHR32((x), SIG_SHIFT))) Yes, you?re right. I forgot to run the vectors under qemu with my previous version (oh, the embarrassment!) Fixed forthcoming

Hello everyone, I was curious if anyone has performed any listening tests or done work with solo instrument performances with the CELT codec? I experience quite a bit of noise coming up with a heavily tonal stimulus, the worst offender being a trumpet. I understand some of this behavior is to be expected, this is a perceptual codec afterall, but was curious if anyone was able to find an elegant

On Nov 23, 2015, at 12:04 PM, John Ridges <jridges at masque.com<mailto:jridges at masque.com>> wrote: Hi Jonathan. I really, really hate to bring this up this late in the game, but I just noticed that your NEON code doesn't use any of the "high" intrinsics for ARM64, e.g. instead of: int32x4_t coef1 = vmovl_s16(vget_high_s16(coef16)); you could use: int32x4_t coef1

Something looks odd without your values (or the doc) because hypergeom_gain() should really approach 1 as x goes to infinity. But in the end, an approximation is probably OK because denoising is anything but an exact science :-) Jean-Marc Quoting John Ridges <jridges at masque.com>: > By my reckoning the confluent hypergoemetric functions should have the > following values: >

Quoting John Ridges <jridges at masque.com>: > Speex tells me that the decoder is always 5 ms, but it says that the > encoder is 5 ms for NB, 8.9375 ms for WB, and 10.90625 ms for UWB. Is > there an extra frame of delay in the encoder that isn't otherwise > accounted for? Oh, delay = frame_size + lookahead If you have a frame size of 20 ms, then there's no choice but

It's been a while since I did the maths. M(-.5;1;-x) optimises something else, though it's not too far. I think it may be [M(-.25;1;-x)]^2 (or is it the sqrt?) that is supposed to be there. In any case, if there's a mismatch between the doc and the code, the code is the one likely to be correct. Jean-Marc John Ridges a ?crit : > I got the approximation from a Google book: >

Hi ? I used this "sample counting " method to?resample and put my audio signals in synch. It worked perfectly?in XP machines using a SoundMax?audio card, but it failed in other XPs using Realtek cards. As seen on http://lists.xiph.org/pipermail/speex-dev/2008-September/006889.html?my application?continously checked my AEC level to slighly modify resample frequency, but convergence was

One other minor thing: I notice that in the inline assembly the result (rd) is constrained as an earlyclobber operand. What was the reason for that?

Hi JM, When porting my project to the Mac, I found that the definition for "celt_int16" and "celt_uint16" are wrong for the MacOS X Framework in celt_types.h, and need to be changed to "int16_t" and "u_int16_t" respectively. Just thought you ought to know. John Ridges

Hi Jean-Marc, I could be way off base here, but it seems to me that line 115 in pitch.c in the function "pitch_downsample": x_lp[i] = SHR32(HALF32(HALF32(x[1][(2*i-1)]+x[1][(2*i+1)])+x[1][2*i]), SIG_SHIFT+2); should actually be: x_lp[i] += SHR32(HALF32(HALF32(x[1][(2*i-1)]+x[1][(2*i+1)])+x[1][2*i]), SIG_SHIFT+2); Sorry if I'm totally misreading things and just

> On Nov 13, 2015, at 1:51 PM, John Ridges <jridges at masque.com> wrote: > > Hi Jonathan, > > I'm sorry to bring this up again, and I don't want to beat a dead horse, but I was very surprised by your benchmarks so I took a little closer look. > > I think what's happening is that it's a little unfair to compare the ARM64 inline assembly to the C code,

Nit: in dual_inner_prod_sse, why not do both horizontal sums at the same time? As in: xsum1 = _mm_add_ps(_mm_movelh_ps(xsum1, xsum2), _mm_movehl_ps(xsum2, xsum1)); xsum1 = _mm_add_ps(xsum1, _mm_shuffle_ps(xsum1, xsum1, 0xf5)); _mm_store_ss(xy1, xsum1); _mm_store_ss(xy2, _mm_movehl_ps(xsum1, xsum1)); --John