diff options
Diffstat (limited to 'celt/rate.c')
-rw-r--r-- | celt/rate.c | 640 |
1 files changed, 640 insertions, 0 deletions
diff --git a/celt/rate.c b/celt/rate.c new file mode 100644 index 00000000..987dffda --- /dev/null +++ b/celt/rate.c @@ -0,0 +1,640 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR + CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include <math.h> +#include "modes.h" +#include "cwrs.h" +#include "arch.h" +#include "os_support.h" + +#include "entcode.h" +#include "rate.h" + +static const unsigned char LOG2_FRAC_TABLE[24]={ + 0, + 8,13, + 16,19,21,23, + 24,26,27,28,29,30,31,32, + 32,33,34,34,35,36,36,37,37 +}; + +#ifdef CUSTOM_MODES + +/*Determines if V(N,K) fits in a 32-bit unsigned integer. + N and K are themselves limited to 15 bits.*/ +static int fits_in32(int _n, int _k) +{ + static const opus_int16 maxN[15] = { + 32767, 32767, 32767, 1476, 283, 109, 60, 40, + 29, 24, 20, 18, 16, 14, 13}; + static const opus_int16 maxK[15] = { + 32767, 32767, 32767, 32767, 1172, 238, 95, 53, + 36, 27, 22, 18, 16, 15, 13}; + if (_n>=14) + { + if (_k>=14) + return 0; + else + return _n <= maxN[_k]; + } else { + return _k <= maxK[_n]; + } +} + +void compute_pulse_cache(CELTMode *m, int LM) +{ + int C; + int i; + int j; + int curr=0; + int nbEntries=0; + int entryN[100], entryK[100], entryI[100]; + const opus_int16 *eBands = m->eBands; + PulseCache *cache = &m->cache; + opus_int16 *cindex; + unsigned char *bits; + unsigned char *cap; + + cindex = opus_alloc(sizeof(cache->index[0])*m->nbEBands*(LM+2)); + cache->index = cindex; + + /* Scan for all unique band sizes */ + for (i=0;i<=LM+1;i++) + { + for (j=0;j<m->nbEBands;j++) + { + int k; + int N = (eBands[j+1]-eBands[j])<<i>>1; + cindex[i*m->nbEBands+j] = -1; + /* Find other bands that have the same size */ + for (k=0;k<=i;k++) + { + int n; + for (n=0;n<m->nbEBands && (k!=i || n<j);n++) + { + if (N == (eBands[n+1]-eBands[n])<<k>>1) + { + cindex[i*m->nbEBands+j] = cindex[k*m->nbEBands+n]; + break; + } + } + } + if (cache->index[i*m->nbEBands+j] == -1 && N!=0) + { + int K; + entryN[nbEntries] = N; + K = 0; + while (fits_in32(N,get_pulses(K+1)) && K<MAX_PSEUDO) + K++; + entryK[nbEntries] = K; + cindex[i*m->nbEBands+j] = curr; + entryI[nbEntries] = curr; + + curr += K+1; + nbEntries++; + } + } + } + bits = opus_alloc(sizeof(unsigned char)*curr); + cache->bits = bits; + cache->size = curr; + /* Compute the cache for all unique sizes */ + for (i=0;i<nbEntries;i++) + { + unsigned char *ptr = bits+entryI[i]; + opus_int16 tmp[MAX_PULSES+1]; + get_required_bits(tmp, entryN[i], get_pulses(entryK[i]), BITRES); + for (j=1;j<=entryK[i];j++) + ptr[j] = tmp[get_pulses(j)]-1; + ptr[0] = entryK[i]; + } + + /* Compute the maximum rate for each band at which we'll reliably use as + many bits as we ask for. */ + cache->caps = cap = opus_alloc(sizeof(cache->caps[0])*(LM+1)*2*m->nbEBands); + for (i=0;i<=LM;i++) + { + for (C=1;C<=2;C++) + { + for (j=0;j<m->nbEBands;j++) + { + int N0; + int max_bits; + N0 = m->eBands[j+1]-m->eBands[j]; + /* N=1 bands only have a sign bit and fine bits. */ + if (N0<<i == 1) + max_bits = C*(1+MAX_FINE_BITS)<<BITRES; + else + { + const unsigned char *pcache; + opus_int32 num; + opus_int32 den; + int LM0; + int N; + int offset; + int ndof; + int qb; + int k; + LM0 = 0; + /* Even-sized bands bigger than N=2 can be split one more + time. */ + if (N0 > 2 && !(N0&1)) + { + N0>>=1; + LM0--; + } + /* N0=1 bands can't be split down to N<2. */ + else if (N0 <= 1) + { + LM0=IMIN(i,1); + N0<<=LM0; + } + /* Compute the cost for the lowest-level PVQ of a fully split + band. */ + pcache = bits + cindex[(LM0+1)*m->nbEBands+j]; + max_bits = pcache[pcache[0]]+1; + /* Add in the cost of coding regular splits. */ + N = N0; + for(k=0;k<i-LM0;k++){ + max_bits <<= 1; + /* Offset the number of qtheta bits by log2(N)/2 + + QTHETA_OFFSET compared to their "fair share" of + total/N */ + offset = ((m->logN[j]+((LM0+k)<<BITRES))>>1)-QTHETA_OFFSET; + /* The number of qtheta bits we'll allocate if the remainder + is to be max_bits. + The average measured cost for theta is 0.89701 times qb, + approximated here as 459/512. */ + num=459*(opus_int32)((2*N-1)*offset+max_bits); + den=((opus_int32)(2*N-1)<<9)-459; + qb = IMIN((num+(den>>1))/den, 57); + celt_assert(qb >= 0); + max_bits += qb; + N <<= 1; + } + /* Add in the cost of a stereo split, if necessary. */ + if (C==2) + { + max_bits <<= 1; + offset = ((m->logN[j]+(i<<BITRES))>>1)-(N==2?QTHETA_OFFSET_TWOPHASE:QTHETA_OFFSET); + ndof = 2*N-1-(N==2); + /* The average measured cost for theta with the step PDF is + 0.95164 times qb, approximated here as 487/512. */ + num = (N==2?512:487)*(opus_int32)(max_bits+ndof*offset); + den = ((opus_int32)ndof<<9)-(N==2?512:487); + qb = IMIN((num+(den>>1))/den, (N==2?64:61)); + celt_assert(qb >= 0); + max_bits += qb; + } + /* Add the fine bits we'll use. */ + /* Compensate for the extra DoF in stereo */ + ndof = C*N + ((C==2 && N>2) ? 1 : 0); + /* Offset the number of fine bits by log2(N)/2 + FINE_OFFSET + compared to their "fair share" of total/N */ + offset = ((m->logN[j] + (i<<BITRES))>>1)-FINE_OFFSET; + /* N=2 is the only point that doesn't match the curve */ + if (N==2) + offset += 1<<BITRES>>2; + /* The number of fine bits we'll allocate if the remainder is + to be max_bits. */ + num = max_bits+ndof*offset; + den = (ndof-1)<<BITRES; + qb = IMIN((num+(den>>1))/den, MAX_FINE_BITS); + celt_assert(qb >= 0); + max_bits += C*qb<<BITRES; + } + max_bits = (4*max_bits/(C*((m->eBands[j+1]-m->eBands[j])<<i)))-64; + celt_assert(max_bits >= 0); + celt_assert(max_bits < 256); + *cap++ = (unsigned char)max_bits; + } + } + } +} + +#endif /* CUSTOM_MODES */ + +#define ALLOC_STEPS 6 + +static inline int interp_bits2pulses(const CELTMode *m, int start, int end, int skip_start, + const int *bits1, const int *bits2, const int *thresh, const int *cap, opus_int32 total, opus_int32 *_balance, + int skip_rsv, int *intensity, int intensity_rsv, int *dual_stereo, int dual_stereo_rsv, int *bits, + int *ebits, int *fine_priority, int _C, int LM, ec_ctx *ec, int encode, int prev) +{ + opus_int32 psum; + int lo, hi; + int i, j; + int logM; + const int C = CHANNELS(_C); + int stereo; + int codedBands=-1; + int alloc_floor; + opus_int32 left, percoeff; + int done; + int balance; + SAVE_STACK; + + alloc_floor = C<<BITRES; + stereo = C>1; + + logM = LM<<BITRES; + lo = 0; + hi = 1<<ALLOC_STEPS; + for (i=0;i<ALLOC_STEPS;i++) + { + int mid = (lo+hi)>>1; + psum = 0; + done = 0; + for (j=end;j-->start;) + { + int tmp = bits1[j] + (mid*(opus_int32)bits2[j]>>ALLOC_STEPS); + if (tmp >= thresh[j] || done) + { + done = 1; + /* Don't allocate more than we can actually use */ + psum += IMIN(tmp, cap[j]); + } else { + if (tmp >= alloc_floor) + psum += alloc_floor; + } + } + if (psum > total) + hi = mid; + else + lo = mid; + } + psum = 0; + /*printf ("interp bisection gave %d\n", lo);*/ + done = 0; + for (j=end;j-->start;) + { + int tmp = bits1[j] + (lo*bits2[j]>>ALLOC_STEPS); + if (tmp < thresh[j] && !done) + { + if (tmp >= alloc_floor) + tmp = alloc_floor; + else + tmp = 0; + } else + done = 1; + /* Don't allocate more than we can actually use */ + tmp = IMIN(tmp, cap[j]); + bits[j] = tmp; + psum += tmp; + } + + /* Decide which bands to skip, working backwards from the end. */ + for (codedBands=end;;codedBands--) + { + int band_width; + int band_bits; + int rem; + j = codedBands-1; + /* Never skip the first band, nor a band that has been boosted by + dynalloc. + In the first case, we'd be coding a bit to signal we're going to waste + all the other bits. + In the second case, we'd be coding a bit to redistribute all the bits + we just signaled should be cocentrated in this band. */ + if (j<=skip_start) + { + /* Give the bit we reserved to end skipping back. */ + total += skip_rsv; + break; + } + /*Figure out how many left-over bits we would be adding to this band. + This can include bits we've stolen back from higher, skipped bands.*/ + left = total-psum; + percoeff = left/(m->eBands[codedBands]-m->eBands[start]); + left -= (m->eBands[codedBands]-m->eBands[start])*percoeff; + rem = IMAX(left-(m->eBands[j]-m->eBands[start]),0); + band_width = m->eBands[codedBands]-m->eBands[j]; + band_bits = (int)(bits[j] + percoeff*band_width + rem); + /*Only code a skip decision if we're above the threshold for this band. + Otherwise it is force-skipped. + This ensures that we have enough bits to code the skip flag.*/ + if (band_bits >= IMAX(thresh[j], alloc_floor+(1<<BITRES))) + { + if (encode) + { + /*This if() block is the only part of the allocation function that + is not a mandatory part of the bitstream: any bands we choose to + skip here must be explicitly signaled.*/ + /*Choose a threshold with some hysteresis to keep bands from + fluctuating in and out.*/ +#ifdef FUZZING + if ((rand()&0x1) == 0) +#else + if (band_bits > ((j<prev?7:9)*band_width<<LM<<BITRES)>>4) +#endif + { + ec_enc_bit_logp(ec, 1, 1); + break; + } + ec_enc_bit_logp(ec, 0, 1); + } else if (ec_dec_bit_logp(ec, 1)) { + break; + } + /*We used a bit to skip this band.*/ + psum += 1<<BITRES; + band_bits -= 1<<BITRES; + } + /*Reclaim the bits originally allocated to this band.*/ + psum -= bits[j]+intensity_rsv; + if (intensity_rsv > 0) + intensity_rsv = LOG2_FRAC_TABLE[j-start]; + psum += intensity_rsv; + if (band_bits >= alloc_floor) + { + /*If we have enough for a fine energy bit per channel, use it.*/ + psum += alloc_floor; + bits[j] = alloc_floor; + } else { + /*Otherwise this band gets nothing at all.*/ + bits[j] = 0; + } + } + + celt_assert(codedBands > start); + /* Code the intensity and dual stereo parameters. */ + if (intensity_rsv > 0) + { + if (encode) + { + *intensity = IMIN(*intensity, codedBands); + ec_enc_uint(ec, *intensity-start, codedBands+1-start); + } + else + *intensity = start+ec_dec_uint(ec, codedBands+1-start); + } + else + *intensity = 0; + if (*intensity <= start) + { + total += dual_stereo_rsv; + dual_stereo_rsv = 0; + } + if (dual_stereo_rsv > 0) + { + if (encode) + ec_enc_bit_logp(ec, *dual_stereo, 1); + else + *dual_stereo = ec_dec_bit_logp(ec, 1); + } + else + *dual_stereo = 0; + + /* Allocate the remaining bits */ + left = total-psum; + percoeff = left/(m->eBands[codedBands]-m->eBands[start]); + left -= (m->eBands[codedBands]-m->eBands[start])*percoeff; + for (j=start;j<codedBands;j++) + bits[j] += ((int)percoeff*(m->eBands[j+1]-m->eBands[j])); + for (j=start;j<codedBands;j++) + { + int tmp = (int)IMIN(left, m->eBands[j+1]-m->eBands[j]); + bits[j] += tmp; + left -= tmp; + } + /*for (j=0;j<end;j++)printf("%d ", bits[j]);printf("\n");*/ + + balance = 0; + for (j=start;j<codedBands;j++) + { + int N0, N, den; + int offset; + int NClogN; + int excess; + + celt_assert(bits[j] >= 0); + N0 = m->eBands[j+1]-m->eBands[j]; + N=N0<<LM; + bits[j] += balance; + + if (N>1) + { + excess = IMAX(bits[j]-cap[j],0); + bits[j] -= excess; + + /* Compensate for the extra DoF in stereo */ + den=(C*N+ ((C==2 && N>2 && !*dual_stereo && j<*intensity) ? 1 : 0)); + + NClogN = den*(m->logN[j] + logM); + + /* Offset for the number of fine bits by log2(N)/2 + FINE_OFFSET + compared to their "fair share" of total/N */ + offset = (NClogN>>1)-den*FINE_OFFSET; + + /* N=2 is the only point that doesn't match the curve */ + if (N==2) + offset += den<<BITRES>>2; + + /* Changing the offset for allocating the second and third + fine energy bit */ + if (bits[j] + offset < den*2<<BITRES) + offset += NClogN>>2; + else if (bits[j] + offset < den*3<<BITRES) + offset += NClogN>>3; + + /* Divide with rounding */ + ebits[j] = IMAX(0, (bits[j] + offset + (den<<(BITRES-1))) / (den<<BITRES)); + + /* Make sure not to bust */ + if (C*ebits[j] > (bits[j]>>BITRES)) + ebits[j] = bits[j] >> stereo >> BITRES; + + /* More than that is useless because that's about as far as PVQ can go */ + ebits[j] = IMIN(ebits[j], MAX_FINE_BITS); + + /* If we rounded down or capped this band, make it a candidate for the + final fine energy pass */ + fine_priority[j] = ebits[j]*(den<<BITRES) >= bits[j]+offset; + + /* Remove the allocated fine bits; the rest are assigned to PVQ */ + bits[j] -= C*ebits[j]<<BITRES; + + } else { + /* For N=1, all bits go to fine energy except for a single sign bit */ + excess = IMAX(0,bits[j]-(C<<BITRES)); + bits[j] -= excess; + ebits[j] = 0; + fine_priority[j] = 1; + } + + /* Fine energy can't take advantage of the re-balancing in + quant_all_bands(). + Instead, do the re-balancing here.*/ + if(excess > 0) + { + int extra_fine; + int extra_bits; + extra_fine = IMIN(excess>>(stereo+BITRES),MAX_FINE_BITS-ebits[j]); + ebits[j] += extra_fine; + extra_bits = extra_fine*C<<BITRES; + fine_priority[j] = extra_bits >= excess-balance; + excess -= extra_bits; + } + balance = excess; + + celt_assert(bits[j] >= 0); + celt_assert(ebits[j] >= 0); + } + /* Save any remaining bits over the cap for the rebalancing in + quant_all_bands(). */ + *_balance = balance; + + /* The skipped bands use all their bits for fine energy. */ + for (;j<end;j++) + { + ebits[j] = bits[j] >> stereo >> BITRES; + celt_assert(C*ebits[j]<<BITRES == bits[j]); + bits[j] = 0; + fine_priority[j] = ebits[j]<1; + } + RESTORE_STACK; + return codedBands; +} + +int compute_allocation(const CELTMode *m, int start, int end, const int *offsets, const int *cap, int alloc_trim, int *intensity, int *dual_stereo, + opus_int32 total, opus_int32 *balance, int *pulses, int *ebits, int *fine_priority, int _C, int LM, ec_ctx *ec, int encode, int prev) +{ + int lo, hi, len, j; + const int C = CHANNELS(_C); + int codedBands; + int skip_start; + int skip_rsv; + int intensity_rsv; + int dual_stereo_rsv; + VARDECL(int, bits1); + VARDECL(int, bits2); + VARDECL(int, thresh); + VARDECL(int, trim_offset); + SAVE_STACK; + + total = IMAX(total, 0); + len = m->nbEBands; + skip_start = start; + /* Reserve a bit to signal the end of manually skipped bands. */ + skip_rsv = total >= 1<<BITRES ? 1<<BITRES : 0; + total -= skip_rsv; + /* Reserve bits for the intensity and dual stereo parameters. */ + intensity_rsv = dual_stereo_rsv = 0; + if (C==2) + { + intensity_rsv = LOG2_FRAC_TABLE[end-start]; + if (intensity_rsv>total) + intensity_rsv = 0; + else + { + total -= intensity_rsv; + dual_stereo_rsv = total>=1<<BITRES ? 1<<BITRES : 0; + total -= dual_stereo_rsv; + } + } + ALLOC(bits1, len, int); + ALLOC(bits2, len, int); + ALLOC(thresh, len, int); + ALLOC(trim_offset, len, int); + + for (j=start;j<end;j++) + { + /* Below this threshold, we're sure not to allocate any PVQ bits */ + thresh[j] = IMAX((C)<<BITRES, (3*(m->eBands[j+1]-m->eBands[j])<<LM<<BITRES)>>4); + /* Tilt of the allocation curve */ + trim_offset[j] = C*(m->eBands[j+1]-m->eBands[j])*(alloc_trim-5-LM)*(end-j-1) + <<(LM+BITRES)>>6; + /* Giving less resolution to single-coefficient bands because they get + more benefit from having one coarse value per coefficient*/ + if ((m->eBands[j+1]-m->eBands[j])<<LM==1) + trim_offset[j] -= C<<BITRES; + } + lo = 1; + hi = m->nbAllocVectors - 1; + do + { + int done = 0; + int psum = 0; + int mid = (lo+hi) >> 1; + for (j=end;j-->start;) + { + int bitsj; + int N = m->eBands[j+1]-m->eBands[j]; + bitsj = C*N*m->allocVectors[mid*len+j]<<LM>>2; + if (bitsj > 0) + bitsj = IMAX(0, bitsj + trim_offset[j]); + bitsj += offsets[j]; + if (bitsj >= thresh[j] || done) + { + done = 1; + /* Don't allocate more than we can actually use */ + psum += IMIN(bitsj, cap[j]); + } else { + if (bitsj >= C<<BITRES) + psum += C<<BITRES; + } + } + if (psum > total) + hi = mid - 1; + else + lo = mid + 1; + /*printf ("lo = %d, hi = %d\n", lo, hi);*/ + } + while (lo <= hi); + hi = lo--; + /*printf ("interp between %d and %d\n", lo, hi);*/ + for (j=start;j<end;j++) + { + int bits1j, bits2j; + int N = m->eBands[j+1]-m->eBands[j]; + bits1j = C*N*m->allocVectors[lo*len+j]<<LM>>2; + bits2j = hi>=m->nbAllocVectors ? + cap[j] : C*N*m->allocVectors[hi*len+j]<<LM>>2; + if (bits1j > 0) + bits1j = IMAX(0, bits1j + trim_offset[j]); + if (bits2j > 0) + bits2j = IMAX(0, bits2j + trim_offset[j]); + if (lo > 0) + bits1j += offsets[j]; + bits2j += offsets[j]; + if (offsets[j]>0) + skip_start = j; + bits2j = IMAX(0,bits2j-bits1j); + bits1[j] = bits1j; + bits2[j] = bits2j; + } + codedBands = interp_bits2pulses(m, start, end, skip_start, bits1, bits2, thresh, cap, + total, balance, skip_rsv, intensity, intensity_rsv, dual_stereo, dual_stereo_rsv, + pulses, ebits, fine_priority, C, LM, ec, encode, prev); + RESTORE_STACK; + return codedBands; +} + |