yading@10: /* yading@10: * Copyright (c) 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 "aacpsdsp.h" yading@10: yading@10: static void ps_add_squares_c(float *dst, const float (*src)[2], int n) yading@10: { yading@10: int i; yading@10: for (i = 0; i < n; i++) yading@10: dst[i] += src[i][0] * src[i][0] + src[i][1] * src[i][1]; yading@10: } yading@10: yading@10: static void ps_mul_pair_single_c(float (*dst)[2], float (*src0)[2], float *src1, yading@10: int n) yading@10: { yading@10: int i; yading@10: for (i = 0; i < n; i++) { yading@10: dst[i][0] = src0[i][0] * src1[i]; yading@10: dst[i][1] = src0[i][1] * src1[i]; yading@10: } yading@10: } yading@10: yading@10: static void ps_hybrid_analysis_c(float (*out)[2], float (*in)[2], yading@10: const float (*filter)[8][2], yading@10: int stride, int n) yading@10: { yading@10: int i, j; yading@10: yading@10: for (i = 0; i < n; i++) { yading@10: float sum_re = filter[i][6][0] * in[6][0]; yading@10: float sum_im = filter[i][6][0] * in[6][1]; yading@10: yading@10: for (j = 0; j < 6; j++) { yading@10: float in0_re = in[j][0]; yading@10: float in0_im = in[j][1]; yading@10: float in1_re = in[12-j][0]; yading@10: float in1_im = in[12-j][1]; yading@10: sum_re += filter[i][j][0] * (in0_re + in1_re) - yading@10: filter[i][j][1] * (in0_im - in1_im); yading@10: sum_im += filter[i][j][0] * (in0_im + in1_im) + yading@10: filter[i][j][1] * (in0_re - in1_re); yading@10: } yading@10: out[i * stride][0] = sum_re; yading@10: out[i * stride][1] = sum_im; yading@10: } yading@10: } yading@10: yading@10: static void ps_hybrid_analysis_ileave_c(float (*out)[32][2], float L[2][38][64], yading@10: int i, int len) yading@10: { yading@10: int j; yading@10: yading@10: for (; i < 64; i++) { yading@10: for (j = 0; j < len; j++) { yading@10: out[i][j][0] = L[0][j][i]; yading@10: out[i][j][1] = L[1][j][i]; yading@10: } yading@10: } yading@10: } yading@10: yading@10: static void ps_hybrid_synthesis_deint_c(float out[2][38][64], yading@10: float (*in)[32][2], yading@10: int i, int len) yading@10: { yading@10: int n; yading@10: yading@10: for (; i < 64; i++) { yading@10: for (n = 0; n < len; n++) { yading@10: out[0][n][i] = in[i][n][0]; yading@10: out[1][n][i] = in[i][n][1]; yading@10: } yading@10: } yading@10: } yading@10: yading@10: static void ps_decorrelate_c(float (*out)[2], float (*delay)[2], yading@10: float (*ap_delay)[PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2], yading@10: const float phi_fract[2], float (*Q_fract)[2], yading@10: const float *transient_gain, yading@10: float g_decay_slope, yading@10: int len) yading@10: { yading@10: static const float a[] = { 0.65143905753106f, yading@10: 0.56471812200776f, yading@10: 0.48954165955695f }; yading@10: float ag[PS_AP_LINKS]; yading@10: int m, n; yading@10: yading@10: for (m = 0; m < PS_AP_LINKS; m++) yading@10: ag[m] = a[m] * g_decay_slope; yading@10: yading@10: for (n = 0; n < len; n++) { yading@10: float in_re = delay[n][0] * phi_fract[0] - delay[n][1] * phi_fract[1]; yading@10: float in_im = delay[n][0] * phi_fract[1] + delay[n][1] * phi_fract[0]; yading@10: for (m = 0; m < PS_AP_LINKS; m++) { yading@10: float a_re = ag[m] * in_re; yading@10: float a_im = ag[m] * in_im; yading@10: float link_delay_re = ap_delay[m][n+2-m][0]; yading@10: float link_delay_im = ap_delay[m][n+2-m][1]; yading@10: float fractional_delay_re = Q_fract[m][0]; yading@10: float fractional_delay_im = Q_fract[m][1]; yading@10: float apd_re = in_re; yading@10: float apd_im = in_im; yading@10: in_re = link_delay_re * fractional_delay_re - yading@10: link_delay_im * fractional_delay_im - a_re; yading@10: in_im = link_delay_re * fractional_delay_im + yading@10: link_delay_im * fractional_delay_re - a_im; yading@10: ap_delay[m][n+5][0] = apd_re + ag[m] * in_re; yading@10: ap_delay[m][n+5][1] = apd_im + ag[m] * in_im; yading@10: } yading@10: out[n][0] = transient_gain[n] * in_re; yading@10: out[n][1] = transient_gain[n] * in_im; yading@10: } yading@10: } yading@10: yading@10: static void ps_stereo_interpolate_c(float (*l)[2], float (*r)[2], yading@10: float h[2][4], float h_step[2][4], yading@10: int len) yading@10: { yading@10: float h0 = h[0][0]; yading@10: float h1 = h[0][1]; yading@10: float h2 = h[0][2]; yading@10: float h3 = h[0][3]; yading@10: float hs0 = h_step[0][0]; yading@10: float hs1 = h_step[0][1]; yading@10: float hs2 = h_step[0][2]; yading@10: float hs3 = h_step[0][3]; yading@10: int n; yading@10: yading@10: for (n = 0; n < len; n++) { yading@10: //l is s, r is d yading@10: float l_re = l[n][0]; yading@10: float l_im = l[n][1]; yading@10: float r_re = r[n][0]; yading@10: float r_im = r[n][1]; yading@10: h0 += hs0; yading@10: h1 += hs1; yading@10: h2 += hs2; yading@10: h3 += hs3; yading@10: l[n][0] = h0 * l_re + h2 * r_re; yading@10: l[n][1] = h0 * l_im + h2 * r_im; yading@10: r[n][0] = h1 * l_re + h3 * r_re; yading@10: r[n][1] = h1 * l_im + h3 * r_im; yading@10: } yading@10: } yading@10: yading@10: static void ps_stereo_interpolate_ipdopd_c(float (*l)[2], float (*r)[2], yading@10: float h[2][4], float h_step[2][4], yading@10: int len) yading@10: { yading@10: float h00 = h[0][0], h10 = h[1][0]; yading@10: float h01 = h[0][1], h11 = h[1][1]; yading@10: float h02 = h[0][2], h12 = h[1][2]; yading@10: float h03 = h[0][3], h13 = h[1][3]; yading@10: float hs00 = h_step[0][0], hs10 = h_step[1][0]; yading@10: float hs01 = h_step[0][1], hs11 = h_step[1][1]; yading@10: float hs02 = h_step[0][2], hs12 = h_step[1][2]; yading@10: float hs03 = h_step[0][3], hs13 = h_step[1][3]; yading@10: int n; yading@10: yading@10: for (n = 0; n < len; n++) { yading@10: //l is s, r is d yading@10: float l_re = l[n][0]; yading@10: float l_im = l[n][1]; yading@10: float r_re = r[n][0]; yading@10: float r_im = r[n][1]; yading@10: h00 += hs00; yading@10: h01 += hs01; yading@10: h02 += hs02; yading@10: h03 += hs03; yading@10: h10 += hs10; yading@10: h11 += hs11; yading@10: h12 += hs12; yading@10: h13 += hs13; yading@10: yading@10: l[n][0] = h00 * l_re + h02 * r_re - h10 * l_im - h12 * r_im; yading@10: l[n][1] = h00 * l_im + h02 * r_im + h10 * l_re + h12 * r_re; yading@10: r[n][0] = h01 * l_re + h03 * r_re - h11 * l_im - h13 * r_im; yading@10: r[n][1] = h01 * l_im + h03 * r_im + h11 * l_re + h13 * r_re; yading@10: } yading@10: } yading@10: yading@10: av_cold void ff_psdsp_init(PSDSPContext *s) yading@10: { yading@10: s->add_squares = ps_add_squares_c; yading@10: s->mul_pair_single = ps_mul_pair_single_c; yading@10: s->hybrid_analysis = ps_hybrid_analysis_c; yading@10: s->hybrid_analysis_ileave = ps_hybrid_analysis_ileave_c; yading@10: s->hybrid_synthesis_deint = ps_hybrid_synthesis_deint_c; yading@10: s->decorrelate = ps_decorrelate_c; yading@10: s->stereo_interpolate[0] = ps_stereo_interpolate_c; yading@10: s->stereo_interpolate[1] = ps_stereo_interpolate_ipdopd_c; yading@10: yading@10: if (ARCH_ARM) yading@10: ff_psdsp_init_arm(s); yading@10: if (ARCH_MIPS) yading@10: ff_psdsp_init_mips(s); yading@10: }