yading@10: /* yading@10: * AAC Spectral Band Replication decoding functions yading@10: * Copyright (c) 2008-2009 Robert Swain ( rob opendot cl ) yading@10: * Copyright (c) 2009-2010 Alex Converse yading@10: * yading@10: * This file is part of Libav. yading@10: * yading@10: * Libav is free software; you can redistribute it and/or yading@10: * modify it under the terms of the GNU Lesser General Public yading@10: * License as published by the Free Software Foundation; either yading@10: * version 2.1 of the License, or (at your option) any later version. yading@10: * yading@10: * Libav is distributed in the hope that it will be useful, yading@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of yading@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU yading@10: * Lesser General Public License for more details. yading@10: * yading@10: * You should have received a copy of the GNU Lesser General Public yading@10: * License along with Libav; if not, write to the Free Software yading@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA yading@10: */ yading@10: yading@10: #include "config.h" yading@10: #include "libavutil/attributes.h" yading@10: #include "libavutil/intfloat.h" yading@10: #include "sbrdsp.h" yading@10: yading@10: static void sbr_sum64x5_c(float *z) yading@10: { yading@10: int k; yading@10: for (k = 0; k < 64; k++) { yading@10: float f = z[k] + z[k + 64] + z[k + 128] + z[k + 192] + z[k + 256]; yading@10: z[k] = f; yading@10: } yading@10: } yading@10: yading@10: static float sbr_sum_square_c(float (*x)[2], int n) yading@10: { yading@10: float sum0 = 0.0f, sum1 = 0.0f; yading@10: int i; yading@10: yading@10: for (i = 0; i < n; i += 2) yading@10: { yading@10: sum0 += x[i + 0][0] * x[i + 0][0]; yading@10: sum1 += x[i + 0][1] * x[i + 0][1]; yading@10: sum0 += x[i + 1][0] * x[i + 1][0]; yading@10: sum1 += x[i + 1][1] * x[i + 1][1]; yading@10: } yading@10: yading@10: return sum0 + sum1; yading@10: } yading@10: yading@10: static void sbr_neg_odd_64_c(float *x) yading@10: { yading@10: union av_intfloat32 *xi = (union av_intfloat32*)x; yading@10: int i; yading@10: for (i = 1; i < 64; i += 4) yading@10: { yading@10: xi[i+0].i ^= 1U<<31; yading@10: xi[i+2].i ^= 1U<<31; yading@10: } yading@10: } yading@10: yading@10: static void sbr_qmf_pre_shuffle_c(float *z) yading@10: { yading@10: union av_intfloat32 *zi = (union av_intfloat32*)z; yading@10: int k; yading@10: zi[64].i = zi[0].i; yading@10: zi[65].i = zi[1].i; yading@10: for (k = 1; k < 31; k+=2) { yading@10: zi[64+2*k+0].i = zi[64 - k].i ^ (1U<<31); yading@10: zi[64+2*k+1].i = zi[ k + 1].i; yading@10: zi[64+2*k+2].i = zi[63 - k].i ^ (1U<<31); yading@10: zi[64+2*k+3].i = zi[ k + 2].i; yading@10: } yading@10: zi[64+2*31+0].i = zi[64 - 31].i ^ (1U<<31); yading@10: zi[64+2*31+1].i = zi[31 + 1].i; yading@10: } yading@10: yading@10: static void sbr_qmf_post_shuffle_c(float W[32][2], const float *z) yading@10: { yading@10: const union av_intfloat32 *zi = (const union av_intfloat32*)z; yading@10: union av_intfloat32 *Wi = (union av_intfloat32*)W; yading@10: int k; yading@10: for (k = 0; k < 32; k+=2) { yading@10: Wi[2*k+0].i = zi[63-k].i ^ (1U<<31); yading@10: Wi[2*k+1].i = zi[k+0].i; yading@10: Wi[2*k+2].i = zi[62-k].i ^ (1U<<31); yading@10: Wi[2*k+3].i = zi[k+1].i; yading@10: } yading@10: } yading@10: yading@10: static void sbr_qmf_deint_neg_c(float *v, const float *src) yading@10: { yading@10: const union av_intfloat32 *si = (const union av_intfloat32*)src; yading@10: union av_intfloat32 *vi = (union av_intfloat32*)v; yading@10: int i; yading@10: for (i = 0; i < 32; i++) { yading@10: vi[ i].i = si[63 - 2*i ].i; yading@10: vi[63 - i].i = si[63 - 2*i - 1].i ^ (1U<<31); yading@10: } yading@10: } yading@10: yading@10: static void sbr_qmf_deint_bfly_c(float *v, const float *src0, const float *src1) yading@10: { yading@10: int i; yading@10: for (i = 0; i < 64; i++) { yading@10: v[ i] = src0[i] - src1[63 - i]; yading@10: v[127 - i] = src0[i] + src1[63 - i]; yading@10: } yading@10: } yading@10: yading@10: static av_always_inline void autocorrelate(const float x[40][2], yading@10: float phi[3][2][2], int lag) yading@10: { yading@10: int i; yading@10: float real_sum = 0.0f; yading@10: float imag_sum = 0.0f; yading@10: if (lag) { yading@10: for (i = 1; i < 38; i++) { yading@10: real_sum += x[i][0] * x[i+lag][0] + x[i][1] * x[i+lag][1]; yading@10: imag_sum += x[i][0] * x[i+lag][1] - x[i][1] * x[i+lag][0]; yading@10: } yading@10: phi[2-lag][1][0] = real_sum + x[ 0][0] * x[lag][0] + x[ 0][1] * x[lag][1]; yading@10: phi[2-lag][1][1] = imag_sum + x[ 0][0] * x[lag][1] - x[ 0][1] * x[lag][0]; yading@10: if (lag == 1) { yading@10: phi[0][0][0] = real_sum + x[38][0] * x[39][0] + x[38][1] * x[39][1]; yading@10: phi[0][0][1] = imag_sum + x[38][0] * x[39][1] - x[38][1] * x[39][0]; yading@10: } yading@10: } else { yading@10: for (i = 1; i < 38; i++) { yading@10: real_sum += x[i][0] * x[i][0] + x[i][1] * x[i][1]; yading@10: } yading@10: phi[2][1][0] = real_sum + x[ 0][0] * x[ 0][0] + x[ 0][1] * x[ 0][1]; yading@10: phi[1][0][0] = real_sum + x[38][0] * x[38][0] + x[38][1] * x[38][1]; yading@10: } yading@10: } yading@10: yading@10: static void sbr_autocorrelate_c(const float x[40][2], float phi[3][2][2]) yading@10: { yading@10: #if 0 yading@10: // This code is slower because it multiplies memory accesses. yading@10: // It is left as eucational purpose and because it may offer yading@10: // a better reference for writing arch-specific dsp functions. yading@10: autocorrelate(x, phi, 0); yading@10: autocorrelate(x, phi, 1); yading@10: autocorrelate(x, phi, 2); yading@10: #else yading@10: float real_sum2 = x[ 0][0] * x[ 2][0] + x[ 0][1] * x[ 2][1]; yading@10: float imag_sum2 = x[ 0][0] * x[ 2][1] - x[ 0][1] * x[ 2][0]; yading@10: float real_sum1 = 0.f, imag_sum1 = 0.f, real_sum0 = 0.0f; yading@10: int i; yading@10: for (i = 1; i < 38; i++) { yading@10: real_sum0 += x[i][0] * x[i ][0] + x[i][1] * x[i ][1]; yading@10: real_sum1 += x[i][0] * x[i+1][0] + x[i][1] * x[i+1][1]; yading@10: imag_sum1 += x[i][0] * x[i+1][1] - x[i][1] * x[i+1][0]; yading@10: real_sum2 += x[i][0] * x[i+2][0] + x[i][1] * x[i+2][1]; yading@10: imag_sum2 += x[i][0] * x[i+2][1] - x[i][1] * x[i+2][0]; yading@10: } yading@10: phi[2-2][1][0] = real_sum2; yading@10: phi[2-2][1][1] = imag_sum2; yading@10: phi[2 ][1][0] = real_sum0 + x[ 0][0] * x[ 0][0] + x[ 0][1] * x[ 0][1]; yading@10: phi[1 ][0][0] = real_sum0 + x[38][0] * x[38][0] + x[38][1] * x[38][1]; yading@10: phi[2-1][1][0] = real_sum1 + x[ 0][0] * x[ 1][0] + x[ 0][1] * x[ 1][1]; yading@10: phi[2-1][1][1] = imag_sum1 + x[ 0][0] * x[ 1][1] - x[ 0][1] * x[ 1][0]; yading@10: phi[0 ][0][0] = real_sum1 + x[38][0] * x[39][0] + x[38][1] * x[39][1]; yading@10: phi[0 ][0][1] = imag_sum1 + x[38][0] * x[39][1] - x[38][1] * x[39][0]; yading@10: #endif yading@10: } yading@10: yading@10: static void sbr_hf_gen_c(float (*X_high)[2], const float (*X_low)[2], yading@10: const float alpha0[2], const float alpha1[2], yading@10: float bw, int start, int end) yading@10: { yading@10: float alpha[4]; yading@10: int i; yading@10: yading@10: alpha[0] = alpha1[0] * bw * bw; yading@10: alpha[1] = alpha1[1] * bw * bw; yading@10: alpha[2] = alpha0[0] * bw; yading@10: alpha[3] = alpha0[1] * bw; yading@10: yading@10: for (i = start; i < end; i++) { yading@10: X_high[i][0] = yading@10: X_low[i - 2][0] * alpha[0] - yading@10: X_low[i - 2][1] * alpha[1] + yading@10: X_low[i - 1][0] * alpha[2] - yading@10: X_low[i - 1][1] * alpha[3] + yading@10: X_low[i][0]; yading@10: X_high[i][1] = yading@10: X_low[i - 2][1] * alpha[0] + yading@10: X_low[i - 2][0] * alpha[1] + yading@10: X_low[i - 1][1] * alpha[2] + yading@10: X_low[i - 1][0] * alpha[3] + yading@10: X_low[i][1]; yading@10: } yading@10: } yading@10: yading@10: static void sbr_hf_g_filt_c(float (*Y)[2], const float (*X_high)[40][2], yading@10: const float *g_filt, int m_max, intptr_t ixh) yading@10: { yading@10: int m; yading@10: yading@10: for (m = 0; m < m_max; m++) { yading@10: Y[m][0] = X_high[m][ixh][0] * g_filt[m]; yading@10: Y[m][1] = X_high[m][ixh][1] * g_filt[m]; yading@10: } yading@10: } yading@10: yading@10: static av_always_inline void sbr_hf_apply_noise(float (*Y)[2], yading@10: const float *s_m, yading@10: const float *q_filt, yading@10: int noise, yading@10: float phi_sign0, yading@10: float phi_sign1, yading@10: int m_max) yading@10: { yading@10: int m; yading@10: yading@10: for (m = 0; m < m_max; m++) { yading@10: float y0 = Y[m][0]; yading@10: float y1 = Y[m][1]; yading@10: noise = (noise + 1) & 0x1ff; yading@10: if (s_m[m]) { yading@10: y0 += s_m[m] * phi_sign0; yading@10: y1 += s_m[m] * phi_sign1; yading@10: } else { yading@10: y0 += q_filt[m] * ff_sbr_noise_table[noise][0]; yading@10: y1 += q_filt[m] * ff_sbr_noise_table[noise][1]; yading@10: } yading@10: Y[m][0] = y0; yading@10: Y[m][1] = y1; yading@10: phi_sign1 = -phi_sign1; yading@10: } yading@10: } yading@10: yading@10: static void sbr_hf_apply_noise_0(float (*Y)[2], const float *s_m, yading@10: const float *q_filt, int noise, yading@10: int kx, int m_max) yading@10: { yading@10: sbr_hf_apply_noise(Y, s_m, q_filt, noise, 1.0, 0.0, m_max); yading@10: } yading@10: yading@10: static void sbr_hf_apply_noise_1(float (*Y)[2], const float *s_m, yading@10: const float *q_filt, int noise, yading@10: int kx, int m_max) yading@10: { yading@10: float phi_sign = 1 - 2 * (kx & 1); yading@10: sbr_hf_apply_noise(Y, s_m, q_filt, noise, 0.0, phi_sign, m_max); yading@10: } yading@10: yading@10: static void sbr_hf_apply_noise_2(float (*Y)[2], const float *s_m, yading@10: const float *q_filt, int noise, yading@10: int kx, int m_max) yading@10: { yading@10: sbr_hf_apply_noise(Y, s_m, q_filt, noise, -1.0, 0.0, m_max); yading@10: } yading@10: yading@10: static void sbr_hf_apply_noise_3(float (*Y)[2], const float *s_m, yading@10: const float *q_filt, int noise, yading@10: int kx, int m_max) yading@10: { yading@10: float phi_sign = 1 - 2 * (kx & 1); yading@10: sbr_hf_apply_noise(Y, s_m, q_filt, noise, 0.0, -phi_sign, m_max); yading@10: } yading@10: yading@10: av_cold void ff_sbrdsp_init(SBRDSPContext *s) yading@10: { yading@10: s->sum64x5 = sbr_sum64x5_c; yading@10: s->sum_square = sbr_sum_square_c; yading@10: s->neg_odd_64 = sbr_neg_odd_64_c; yading@10: s->qmf_pre_shuffle = sbr_qmf_pre_shuffle_c; yading@10: s->qmf_post_shuffle = sbr_qmf_post_shuffle_c; yading@10: s->qmf_deint_neg = sbr_qmf_deint_neg_c; yading@10: s->qmf_deint_bfly = sbr_qmf_deint_bfly_c; yading@10: s->autocorrelate = sbr_autocorrelate_c; yading@10: s->hf_gen = sbr_hf_gen_c; yading@10: s->hf_g_filt = sbr_hf_g_filt_c; yading@10: yading@10: s->hf_apply_noise[0] = sbr_hf_apply_noise_0; yading@10: s->hf_apply_noise[1] = sbr_hf_apply_noise_1; yading@10: s->hf_apply_noise[2] = sbr_hf_apply_noise_2; yading@10: s->hf_apply_noise[3] = sbr_hf_apply_noise_3; yading@10: yading@10: if (ARCH_ARM) yading@10: ff_sbrdsp_init_arm(s); yading@10: if (ARCH_X86) yading@10: ff_sbrdsp_init_x86(s); yading@10: if (ARCH_MIPS) yading@10: ff_sbrdsp_init_mips(s); yading@10: }