/* ** Copyright 2003-2010, VisualOn, Inc. ** ** Licensed under the Apache License, Version 2.0 (the "License"); ** you may not use this file except in compliance with the License. ** You may obtain a copy of the License at ** ** http://www.apache.org/licenses/LICENSE-2.0 ** ** Unless required by applicable law or agreed to in writing, software ** distributed under the License is distributed on an "AS IS" BASIS, ** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. ** See the License for the specific language governing permissions and ** limitations under the License. */ /************************************************************************ * File: c2t64fx.c * * * * Description:Performs algebraic codebook search for 6.60kbits mode* * * *************************************************************************/ #include "typedef.h" #include "basic_op.h" #include "math_op.h" #include "acelp.h" #include "cnst.h" #define NB_TRACK 2 #define STEP 2 #define NB_POS 32 #define MSIZE 1024 /************************************************************************* * Function: ACELP_2t64_fx() * * * * 12 bits algebraic codebook. * * 2 tracks x 32 positions per track = 64 samples. * * * * 12 bits --> 2 pulses in a frame of 64 samples. * * * * All pulses can have two (2) possible amplitudes: +1 or -1. * * Each pulse can have 32 possible positions. * **************************************************************************/ void ACELP_2t64_fx( Word16 dn[], /* (i) <12b : correlation between target x[] and H[] */ Word16 cn[], /* (i) <12b : residual after long term prediction */ Word16 H[], /* (i) Q12: impulse response of weighted synthesis filter */ Word16 code[], /* (o) Q9 : algebraic (fixed) codebook excitation */ Word16 y[], /* (o) Q9 : filtered fixed codebook excitation */ Word16 * index /* (o) : index (12): 5+1+5+1 = 11 bits. */ ) { Word32 i, j, k, i0, i1, ix, iy, pos, pos2; Word16 ps, psk, ps1, ps2, alpk, alp1, alp2, sq; Word16 alp, val, exp, k_cn, k_dn; Word16 *p0, *p1, *p2, *psign; Word16 *h, *h_inv, *ptr_h1, *ptr_h2, *ptr_hf; Word16 sign[L_SUBFR], vec[L_SUBFR], dn2[L_SUBFR]; Word16 h_buf[4 * L_SUBFR] = {0}; Word16 rrixix[NB_TRACK][NB_POS]; Word16 rrixiy[MSIZE]; Word32 s, cor; /*----------------------------------------------------------------* * Find sign for each pulse position. * *----------------------------------------------------------------*/ alp = 8192; /* alp = 2.0 (Q12) */ /* calculate energy for normalization of cn[] and dn[] */ /* set k_cn = 32..32767 (ener_cn = 2^30..256-0) */ #ifdef ASM_OPT /* asm optimization branch */ s = Dot_product12_asm(cn, cn, L_SUBFR, &exp); #else s = Dot_product12(cn, cn, L_SUBFR, &exp); #endif Isqrt_n(&s, &exp); s = L_shl(s, add1(exp, 5)); if (s > INT_MAX - 0x8000) { s = INT_MAX - 0x8000; } k_cn = vo_round(s); /* set k_dn = 32..512 (ener_dn = 2^30..2^22) */ #ifdef ASM_OPT /* asm optimization branch */ s = Dot_product12_asm(dn, dn, L_SUBFR, &exp); #else s = Dot_product12(dn, dn, L_SUBFR, &exp); #endif Isqrt_n(&s, &exp); k_dn = vo_round(L_shl(s, (exp + 8))); /* k_dn = 256..4096 */ k_dn = vo_mult_r(alp, k_dn); /* alp in Q12 */ /* mix normalized cn[] and dn[] */ p0 = cn; p1 = dn; p2 = dn2; for (i = 0; i < L_SUBFR/4; i++) { s = (k_cn* (*p0++))+(k_dn * (*p1++)); *p2++ = s >> 7; s = (k_cn* (*p0++))+(k_dn * (*p1++)); *p2++ = s >> 7; s = (k_cn* (*p0++))+(k_dn * (*p1++)); *p2++ = s >> 7; s = (k_cn* (*p0++))+(k_dn * (*p1++)); *p2++ = s >> 7; } /* set sign according to dn2[] = k_cn*cn[] + k_dn*dn[] */ for (i = 0; i < L_SUBFR; i ++) { val = dn[i]; ps = dn2[i]; if (ps >= 0) { sign[i] = 32767; /* sign = +1 (Q12) */ vec[i] = -32768; } else { sign[i] = -32768; /* sign = -1 (Q12) */ vec[i] = 32767; dn[i] = -val; } } /*------------------------------------------------------------* * Compute h_inv[i]. * *------------------------------------------------------------*/ /* impulse response buffer for fast computation */ h = h_buf + L_SUBFR; h_inv = h + (L_SUBFR<<1); for (i = 0; i < L_SUBFR; i++) { h[i] = H[i]; h_inv[i] = vo_negate(h[i]); } /*------------------------------------------------------------* * Compute rrixix[][] needed for the codebook search. * * Result is multiplied by 0.5 * *------------------------------------------------------------*/ /* Init pointers to last position of rrixix[] */ p0 = &rrixix[0][NB_POS - 1]; p1 = &rrixix[1][NB_POS - 1]; ptr_h1 = h; cor = 0x00010000L; /* for rounding */ for (i = 0; i < NB_POS; i++) { cor += ((*ptr_h1) * (*ptr_h1) << 1); ptr_h1++; *p1-- = (extract_h(cor) >> 1); cor += ((*ptr_h1) * (*ptr_h1) << 1); ptr_h1++; *p0-- = (extract_h(cor) >> 1); } /*------------------------------------------------------------* * Compute rrixiy[][] needed for the codebook search. * *------------------------------------------------------------*/ pos = MSIZE - 1; pos2 = MSIZE - 2; ptr_hf = h + 1; for (k = 0; k < NB_POS; k++) { p1 = &rrixiy[pos]; p0 = &rrixiy[pos2]; cor = 0x00008000L; /* for rounding */ ptr_h1 = h; ptr_h2 = ptr_hf; for (i = (k + 1); i < NB_POS; i++) { cor += ((*ptr_h1) * (*ptr_h2))<<1; ptr_h1++; ptr_h2++; *p1 = extract_h(cor); cor += ((*ptr_h1) * (*ptr_h2))<<1; ptr_h1++; ptr_h2++; *p0 = extract_h(cor); p1 -= (NB_POS + 1); p0 -= (NB_POS + 1); } cor += ((*ptr_h1) * (*ptr_h2))<<1; ptr_h1++; ptr_h2++; *p1 = extract_h(cor); pos -= NB_POS; pos2--; ptr_hf += STEP; } /*------------------------------------------------------------* * Modification of rrixiy[][] to take signs into account. * *------------------------------------------------------------*/ p0 = rrixiy; for (i = 0; i < L_SUBFR; i += STEP) { psign = sign; if (psign[i] < 0) { psign = vec; } for (j = 1; j < L_SUBFR; j += STEP) { *p0 = vo_mult(*p0, psign[j]); p0++; } } /*-------------------------------------------------------------------* * search 2 pulses: * * ~@~~~~~~~~~~~~~~ * * 32 pos x 32 pos = 1024 tests (all combinaisons is tested) * *-------------------------------------------------------------------*/ p0 = rrixix[0]; p1 = rrixix[1]; p2 = rrixiy; psk = -1; alpk = 1; ix = 0; iy = 1; for (i0 = 0; i0 < L_SUBFR; i0 += STEP) { ps1 = dn[i0]; alp1 = (*p0++); pos = -1; for (i1 = 1; i1 < L_SUBFR; i1 += STEP) { ps2 = add1(ps1, dn[i1]); alp2 = add1(alp1, add1(*p1++, *p2++)); sq = vo_mult(ps2, ps2); s = vo_L_mult(alpk, sq) - ((psk * alp2)<<1); if (s > 0) { psk = sq; alpk = alp2; pos = i1; } } p1 -= NB_POS; if (pos >= 0) { ix = i0; iy = pos; } } /*-------------------------------------------------------------------* * Build the codeword, the filtered codeword and index of codevector.* *-------------------------------------------------------------------*/ for (i = 0; i < L_SUBFR; i++) { code[i] = 0; } i0 = (ix >> 1); /* pos of pulse 1 (0..31) */ i1 = (iy >> 1); /* pos of pulse 2 (0..31) */ if (sign[ix] > 0) { code[ix] = 512; /* codeword in Q9 format */ p0 = h - ix; } else { code[ix] = -512; i0 += NB_POS; p0 = h_inv - ix; } if (sign[iy] > 0) { code[iy] = 512; p1 = h - iy; } else { code[iy] = -512; i1 += NB_POS; p1 = h_inv - iy; } *index = add1((i0 << 6), i1); for (i = 0; i < L_SUBFR; i++) { y[i] = vo_shr_r(add1((*p0++), (*p1++)), 3); } return; }