Showing the changes to libavfilter/anlms_template.c (1 of 3 files). Show all
Unified · Split
| @@ -0,0 +1,141 @@ | ||
| 1 | +/* | |
| 2 | + * This file is part of FFmpeg. | |
| 3 | + * | |
| 4 | + * FFmpeg is free software; you can redistribute it and/or | |
| 5 | + * modify it under the terms of the GNU Lesser General Public | |
| 6 | + * License as published by the Free Software Foundation; either | |
| 7 | + * version 2.1 of the License, or (at your option) any later version. | |
| 8 | + * | |
| 9 | + * FFmpeg is distributed in the hope that it will be useful, | |
| 10 | + * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 11 | + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 12 | + * Lesser General Public License for more details. | |
| 13 | + * | |
| 14 | + * You should have received a copy of the GNU Lesser General Public | |
| 15 | + * License along with FFmpeg; if not, write to the Free Software | |
| 16 | + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
| 17 | + */ | |
| 18 | + | |
| 19 | +#undef ONE | |
| 20 | +#undef ftype | |
| 21 | +#undef SAMPLE_FORMAT | |
| 22 | +#if DEPTH == 32 | |
| 23 | +#define SAMPLE_FORMAT float | |
| 24 | +#define ftype float | |
| 25 | +#define ONE 1.f | |
| 26 | +#else | |
| 27 | +#define SAMPLE_FORMAT double | |
| 28 | +#define ftype double | |
| 29 | +#define ONE 1.0 | |
| 30 | +#endif | |
| 31 | + | |
| 32 | +#define fn3(a,b) a##_##b | |
| 33 | +#define fn2(a,b) fn3(a,b) | |
| 34 | +#define fn(a) fn2(a, SAMPLE_FORMAT) | |
| 35 | + | |
| 36 | +#if DEPTH == 64 | |
| 37 | +static double scalarproduct_double(const double *v1, const double *v2, int len) | |
| 38 | +{ | |
| 39 | + double p = 0.0; | |
| 40 | + | |
| 41 | + for (int i = 0; i < len; i++) | |
| 42 | + p += v1[i] * v2[i]; | |
| 43 | + | |
| 44 | + return p; | |
| 45 | +} | |
| 46 | +#endif | |
| 47 | + | |
| 48 | +static ftype fn(fir_sample)(AudioNLMSContext *s, ftype sample, ftype *delay, | |
| 49 | + ftype *coeffs, ftype *tmp, int *offset) | |
| 50 | +{ | |
| 51 | + const int order = s->order; | |
| 52 | + ftype output; | |
| 53 | + | |
| 54 | + delay[*offset] = sample; | |
| 55 | + | |
| 56 | + memcpy(tmp, coeffs + order - *offset, order * sizeof(ftype)); | |
| 57 | + | |
| 58 | +#if DEPTH == 32 | |
| 59 | + output = s->fdsp->scalarproduct_float(delay, tmp, s->kernel_size); | |
| 60 | +#else | |
| 61 | + output = scalarproduct_double(delay, tmp, s->kernel_size); | |
| 62 | +#endif | |
| 63 | + | |
| 64 | + if (--(*offset) < 0) | |
| 65 | + *offset = order - 1; | |
| 66 | + | |
| 67 | + return output; | |
| 68 | +} | |
| 69 | + | |
| 70 | +static ftype fn(process_sample)(AudioNLMSContext *s, ftype input, ftype desired, | |
| 71 | + ftype *delay, ftype *coeffs, ftype *tmp, int *offsetp) | |
| 72 | +{ | |
| 73 | + const int order = s->order; | |
| 74 | + const ftype leakage = s->leakage; | |
| 75 | + const ftype mu = s->mu; | |
| 76 | + const ftype a = ONE - leakage; | |
| 77 | + ftype sum, output, e, norm, b; | |
| 78 | + int offset = *offsetp; | |
| 79 | + | |
| 80 | + delay[offset + order] = input; | |
| 81 | + | |
| 82 | + output = fn(fir_sample)(s, input, delay, coeffs, tmp, offsetp); | |
| 83 | + e = desired - output; | |
| 84 | + | |
| 85 | +#if DEPTH == 32 | |
| 86 | + sum = s->fdsp->scalarproduct_float(delay, delay, s->kernel_size); | |
| 87 | +#else | |
| 88 | + sum = scalarproduct_double(delay, delay, s->kernel_size); | |
| 89 | +#endif | |
| 90 | + norm = s->eps + sum; | |
| 91 | + b = mu * e / norm; | |
| 92 | + if (s->anlmf) | |
| 93 | + b *= e * e; | |
| 94 | + | |
| 95 | + memcpy(tmp, delay + offset, order * sizeof(ftype)); | |
| 96 | + | |
| 97 | +#if DEPTH == 32 | |
| 98 | + s->fdsp->vector_fmul_scalar(coeffs, coeffs, a, s->kernel_size); | |
| 99 | + s->fdsp->vector_fmac_scalar(coeffs, tmp, b, s->kernel_size); | |
| 100 | +#else | |
| 101 | + s->fdsp->vector_dmul_scalar(coeffs, coeffs, a, s->kernel_size); | |
| 102 | + s->fdsp->vector_dmac_scalar(coeffs, tmp, b, s->kernel_size); | |
| 103 | +#endif | |
| 104 | + | |
| 105 | + memcpy(coeffs + order, coeffs, order * sizeof(ftype)); | |
| 106 | + | |
| 107 | + switch (s->output_mode) { | |
| 108 | + case IN_MODE: output = input; break; | |
| 109 | + case DESIRED_MODE: output = desired; break; | |
| 110 | + case OUT_MODE: output = desired - output; break; | |
| 111 | + case NOISE_MODE: output = input - output; break; | |
| 112 | + case ERROR_MODE: break; | |
| 113 | + } | |
| 114 | + return output; | |
| 115 | +} | |
| 116 | + | |
| 117 | +static int fn(filter_channels)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) | |
| 118 | +{ | |
| 119 | + AudioNLMSContext *s = ctx->priv; | |
| 120 | + AVFrame *out = arg; | |
| 121 | + const int start = (out->ch_layout.nb_channels * jobnr) / nb_jobs; | |
| 122 | + const int end = (out->ch_layout.nb_channels * (jobnr+1)) / nb_jobs; | |
| 123 | + | |
| 124 | + for (int c = start; c < end; c++) { | |
| 125 | + const ftype *input = (const ftype *)s->frame[0]->extended_data[c]; | |
| 126 | + const ftype *desired = (const ftype *)s->frame[1]->extended_data[c]; | |
| 127 | + ftype *delay = (ftype *)s->delay->extended_data[c]; | |
| 128 | + ftype *coeffs = (ftype *)s->coeffs->extended_data[c]; | |
| 129 | + ftype *tmp = (ftype *)s->tmp->extended_data[c]; | |
| 130 | + int *offset = (int *)s->offset->extended_data[c]; | |
| 131 | + ftype *output = (ftype *)out->extended_data[c]; | |
| 132 | + | |
| 133 | + for (int n = 0; n < out->nb_samples; n++) { | |
| 134 | + output[n] = fn(process_sample)(s, input[n], desired[n], delay, coeffs, tmp, offset); | |
| 135 | + if (ctx->is_disabled) | |
| 136 | + output[n] = input[n]; | |
| 137 | + } | |
| 138 | + } | |
| 139 | + | |
| 140 | + return 0; | |
| 141 | +} | |