search for: qconst16

Gents, In Version 8.1, QCONST16 is defined as follows in arch.h within the FLOATING_POINT section: #define QCONST16(x,bits) (x) However, in both celt.c and quant_bands.c, QCONST16 is referenced without a conditional FLOATING_POINT define. So when I compile as FIXED_POINT I get an "undefined identifier" for QCONS...

Greetings, I have found that QCONST16(32.f, 10) yields different result (from gcc 4.4.7 ) when the code is compiled for TI C55 or C64 DSPs. gcc - result is 32767 TI compiler result is -32768 (C55 compiler version 4.4.1 and C6x compiler version 7.4.2) Although not in the current code, QCONST32(32.f, 26) results differ in similar fashi...

...ging the default definition to take advantage of this hardware changes the result of that computation. Please consider the change below for inclusion in the opus code base. /* *tf_estimate = 1 + MIN16(1, sqrt(MAX16(0, tf_max-30))/20); */ - *tf_estimate = celt_sqrt(MAX16(0, SHL32(MULT16_16(QCONST16(0.0069,14),MIN16(163,tf_max)),14)-QCONST32(0.139,28))); + *tf_estimate = celt_sqrt(MAX32(0, SHL32(MULT16_16(QCONST16(0.0069,14),MIN16(163,tf_max)),14)-QCONST32(0.139,28))); /*printf("%d %f\n", tf_max, mask_metric);*/ --

...1666,7 +1666,7 @@ int celt_encode_with_ec(CELTEncoder * OP prefilter_tapset = st->tapset_decision; pf_on = run_prefilter(st, in, prefilter_mem, CC, N, prefilter_tapset, &pitch_index, &gain1, &qg, enabled, nbAvailableBytes, &st->analysis); - if ((gain1 > QCONST16(.4f,15) || st->prefilter_gain > QCONST16(.4f,15)) && (!st->analysis.valid || st->analysis.tonality > .3) + if ((gain1 > QCONST16(.4f,15) || st->prefilter_gain > QCONST16(.4f,15)) && (!st->analysis.valid || st->analysis.tonality > .3f)...

...t C = CHANNELS(m); @@ -234,9 +235,11 @@ void compute_pitch_gain(const CELTMode *m, const celt_norm_t *X, const celt_norm Sxy = Sxx; /* We need to be a bit conservative (multiply gain by 0.9), otherwise the residual doesn't quantise well */ - Sxy = MULT16_32_Q15(QCONST16(.9f, 15), Sxy); + Sxy = MULT16_32_Q15(QCONST16(.99f, 15), Sxy); /* gain = Sxy/Sxx */ gains[i] = EXTRACT16(celt_div(Sxy,ADD32(SHR32(Sxx, PGAIN_SHIFT),EPSILON))); + if (gains[i]>QCONST16(.5,15)) + gain_sum++; /*printf ("%f ", 1-sqrt(1-gain*gain));*/...

Hi, It is better to replace this line in function filterbank_new: max_mel = toBARK(EXTRACT16(MULT16_16_Q15(QCONST16(.5f,15),sampling))); to max_mel = toBARK(EXTRACT16(sampling/2)); It gives the same but it seems to be faster and avoids overflow on 44100 kHz that prevents denoiser to process 44100 streams. (Yes I know that Speex should not pack 44100 streams but it does now and I use it). Best Regards, Dmitry...

...d of lax lately and have only now tried to reconcile it with the state structures in the 0.7.1 drop. I think (and feel free to contradict me here) that the following lines should be added to CELT_RESET_STATE in celt_encoder_ctl: st->fold_decision = 1; st->tonal_average = QCONST16(1.,8); st->gain_prod = 0; st->vbr_reservoir = 0; st->vbr_drift = 0; st->vbr_offset = 0; st->vbr_count = 0; st->xmem = 0; CELT_MEMSET(st->pitch_buf, 0, (MAX_PERIOD>>1)+2); and the following lines should be add...

...lculating 'beta' within 'speex_preprocess_run', I received a floating point exception even when compiled with -DFIXED_POINT. The reason is as follows: <original beta calculation within speex_preprocess_run> ... st->nb_adapt++; st->min_count++; beta = MAX16(QCONST16(.03,15),DIV32_16(Q15_ONE,st->nb_adapt)); beta_1 = Q15_ONE-beta; ... </> On my architecture(x86) at least the DIV32_16 is defined as follows. typedef short spx_int16_t; typedef spx_int16_t spx_word16_t; #define DIV32_16(a,b) ((spx_word16_t)(((spx_word32_t)(a))/((spx_word16_t)(b)))) Th...

...When I patch the value 0x4006 into the C55 build, the output matches > the C6x. The problem is that 2^15 evaluates to -32768 on the C55 and 32768 > on the C6x. Right on! > Applying our friend EXTEND32 causes the constant to evaluate correctly. In > fixed_generic.h, > #define QCONST16(x,bits) > ((spx_word16_t)((x)*((EXTEND32(1))<<(bits))+((EXTEND32(1))<<((bits)-1)))) Actually, this is a case for a simple cast to (spx_word32_t) because QCONST can be used in a static initialization and EXTEND32 *can* be defined as a function (e.g. for fixed-point debug). > La...

..., 0, /*LSP quantization*/ lsp_quant_lbr, lsp_unquant_lbr, /*Pitch quantization*/ pitch_search_3tap, pitch_unquant_3tap, &ltp_params_med, /*Innovation quantization*/ split_cb_search_shape_sign, split_cb_shape_sign_unquant, &split_cb_nb_med, QCONST16(.45,15), 220 }; /* Parameters for Long-Term Prediction (LTP)*/ static const ltp_params ltp_params_lbr = { gain_cdbk_lbr, 5, 7 }; /* Split-VQ innovation parameters narrowband */ static const split_cb_params split_cb_nb_med = { 8, /*subvect_size*/ 5,...

Hi all, I am the developer of the llcon software (llcon.sf.net) which is a software making it possible for musicians to play in real-time over the internet. Up to now I have used ADPCM or no audio coding. Gregory Maxwell was pointing me to the great CELT project. Using CELT has the advantage to be able to use higher sample rates, getting lower code rate and better error concealment. I have

Hello Jean-Marc, When you count the adaptation step size in case where the filter has already had minimal adaptation, you use the following expression: r = (0.7*r + 0.3*15*RER*e)/e*(power[i] + 10) = (0.7*leak_estimate*Yf[i] + 0.3*15*RER*(Rf[i] + 1))/((Rf[i] + 1)*(power[i] + 10)). Why do we need this weighted sum and the component 0.3*15*RER*(Rf[i] + 1)? Why use the correlation-based RER if

...en you count the adaptation step size in case where the filter has already had > minimal adaptation, you use the following expression: > > r = (0.7*r + 0.3*15*RER*e)/e*(power[i] + 10) = (0.7*leak_estimate*Yf[i] + > 0.3*15*RER*(Rf[i] + 1))/((Rf[i] + 1)*(power[i] + 10)). Actually, the QCONST16(.3,15) doesn't mean (.3*15), but the value .3 in Q15 format (fixed-point). > Why do we need this weighted sum and the component 0.3*15*RER*(Rf[i] + 1)? First, the reason for multiplying by (Rf[i]+1) is because I'll be dividing by it in the next line (and I didn't want to do divisi...

...nt64)(b),31)) +#else #define MULT32_32_Q31(a,b) ADD32(ADD32(SHL(MULT16_16(SHR((a),16),SHR((b),16)),1), SHR(MULT16_16SU(SHR((a),16),((b)&0x0000ffff)),15)), SHR(MULT16_16SU(SHR((b),16),((a)&0x0000ffff)),15)) +#endif /** Compile-time conversion of float constant to 16-bit value */ #define QCONST16(x,bits) ((opus_val16)(.5+(x)*(((opus_val32)1)<<(bits)))) -- 2.4.9 (Apple Git-60)

...tput >> matches >> the C6x. The problem is that 2^15 evaluates to -32768 on the C55 and >> 32768 >> on the C6x. > >Right on! > >> Applying our friend EXTEND32 causes the constant to evaluate correctly. >> In >> fixed_generic.h, >> #define QCONST16(x,bits) >> ((spx_word16_t)((x)*((EXTEND32(1))<<(bits))+((EXTEND32(1))<<((bits)-1)))) > >Actually, this is a case for a simple cast to (spx_word32_t) because >QCONST can be used in a static initialization and EXTEND32 *can* be >defined as a function (e.g. for fixed-poin...

...Pitch quantization*/ > > pitch_search_3tap, > > pitch_unquant_3tap, > > &ltp_params_med, > > /*Innovation quantization*/ > > split_cb_search_shape_sign, > > split_cb_shape_sign_unquant, > > &split_cb_nb_med, > > QCONST16(.45,15), > > 220 > > }; > > > > /* Parameters for Long-Term Prediction (LTP)*/ > > static const ltp_params ltp_params_lbr = { > > gain_cdbk_lbr, > > 5, > > 7 > > }; > > > > /* Split-VQ innovation parameters...

...CELTEncoder * restrict st, const celt_sig * pcm, i ec_enc_icdf(enc, prefilter_tapset, tapset_icdf, 2); else prefilter_tapset = 0; + fprintf(stderr,"pitch_index: %4i gain %0.2f tapset: %i\n",pitch_index,gain1,prefilter_tapset); gain1 = QCONST16(0.09375f,15)*(qg+1); pf_on = 1; } @@ -1967,17 +1968,26 @@ static void celt_decode_lost(CELTDecoder * restrict st, celt_word16 * restrict p log2Amp(st->mode, st->start, st->end, bandE, backgroundLogE, C); seed = st->rng; - for (i=0;i<C*N;i++) +...

...wnsample(celt_sig * OPUS_RESTRICT x[], opus_val16 * OPUS_RESTRICT x_lp, - int len, int C) + int len, int C, const int arch) { int i; opus_val32 ac[5]; @@ -167,11 +167,11 @@ void pitch_downsample(celt_sig * OPUS_RESTRICT x[], opus_val16 * OPUS_RESTRICT x tmp = MULT16_16_Q15(QCONST16(.9f,15), tmp); lpc[i] = MULT16_16_Q15(lpc[i], tmp); } - celt_fir(x_lp, lpc, x_lp, len>>1, 4, mem); + celt_fir[arch&OPUS_ARCHMASK](x_lp, lpc, x_lp, len>>1, 4, mem); mem[0]=0; lpc[0]=QCONST16(.8f,12); - celt_fir(x_lp, lpc, x_lp, len>>1, 1, mem); + ce...

...9 on the C55 and 0x4006 on the C6x. When I patch the value 0x4006 into the C55 build, the output matches the C6x. The problem is that 2^15 evaluates to -32768 on the C55 and 32768 on the C6x. Applying our friend EXTEND32 causes the constant to evaluate correctly. In fixed_generic.h, #define QCONST16(x,bits) ((spx_word16_t)((x)*((EXTEND32(1))<<(bits))+((EXTEND32(1))<<((bits)-1)))) Later I will check if this change makes these two builds match in the latest SVN code. >> Finally, in the simulator I measured the peak MIPs for the C55x as 29.4, >> where it was 41.5 in Sp...

--- celt/arch.h | 5 +++++ silk/macros.h | 4 +--- 2 files changed, 6 insertions(+), 3 deletions(-) diff --git a/celt/arch.h b/celt/arch.h index 9f74ddd..670527b 100644 --- a/celt/arch.h +++ b/celt/arch.h @@ -78,6 +78,11 @@ static OPUS_INLINE void _celt_fatal(const char *str, const char *file, int line) #define UADD32(a,b) ((a)+(b)) #define USUB32(a,b) ((a)-(b)) +/* Set this if opus_int64