/* * Copyright (c) 2012 Justin Ruggles * * This file is part of Libav. * * Libav is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * Libav is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with Libav; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include #include "libavutil/common.h" #include "libavutil/libm.h" #include "libavutil/samplefmt.h" #include "avresample.h" #include "internal.h" #include "audio_data.h" #include "audio_mix.h" static const char *coeff_type_names[] = { "q8", "q15", "flt" }; struct AudioMix { AVAudioResampleContext *avr; enum AVSampleFormat fmt; enum AVMixCoeffType coeff_type; uint64_t in_layout; uint64_t out_layout; int in_channels; int out_channels; int ptr_align; int samples_align; int has_optimized_func; const char *func_descr; const char *func_descr_generic; mix_func *mix; mix_func *mix_generic; int in_matrix_channels; int out_matrix_channels; int output_zero[AVRESAMPLE_MAX_CHANNELS]; int input_skip[AVRESAMPLE_MAX_CHANNELS]; int output_skip[AVRESAMPLE_MAX_CHANNELS]; int16_t *matrix_q8[AVRESAMPLE_MAX_CHANNELS]; int32_t *matrix_q15[AVRESAMPLE_MAX_CHANNELS]; float *matrix_flt[AVRESAMPLE_MAX_CHANNELS]; void **matrix; }; void ff_audio_mix_set_func(AudioMix *am, enum AVSampleFormat fmt, enum AVMixCoeffType coeff_type, int in_channels, int out_channels, int ptr_align, int samples_align, const char *descr, void *mix_func) { if (fmt == am->fmt && coeff_type == am->coeff_type && ( in_channels == am->in_matrix_channels || in_channels == 0) && (out_channels == am->out_matrix_channels || out_channels == 0)) { char chan_str[16]; am->mix = mix_func; am->func_descr = descr; am->ptr_align = ptr_align; am->samples_align = samples_align; if (ptr_align == 1 && samples_align == 1) { am->mix_generic = mix_func; am->func_descr_generic = descr; } else { am->has_optimized_func = 1; } if (in_channels) { if (out_channels) snprintf(chan_str, sizeof(chan_str), "[%d to %d] ", in_channels, out_channels); else snprintf(chan_str, sizeof(chan_str), "[%d to any] ", in_channels); } else if (out_channels) { snprintf(chan_str, sizeof(chan_str), "[any to %d] ", out_channels); } else { snprintf(chan_str, sizeof(chan_str), "[any to any] "); } av_log(am->avr, AV_LOG_DEBUG, "audio_mix: found function: [fmt=%s] " "[c=%s] %s(%s)\n", av_get_sample_fmt_name(fmt), coeff_type_names[coeff_type], chan_str, descr); } } #define MIX_FUNC_NAME(fmt, cfmt) mix_any_ ## fmt ##_## cfmt ##_c #define MIX_FUNC_GENERIC(fmt, cfmt, stype, ctype, sumtype, expr) \ static void MIX_FUNC_NAME(fmt, cfmt)(stype **samples, ctype **matrix, \ int len, int out_ch, int in_ch) \ { \ int i, in, out; \ stype temp[AVRESAMPLE_MAX_CHANNELS]; \ for (i = 0; i < len; i++) { \ for (out = 0; out < out_ch; out++) { \ sumtype sum = 0; \ for (in = 0; in < in_ch; in++) \ sum += samples[in][i] * matrix[out][in]; \ temp[out] = expr; \ } \ for (out = 0; out < out_ch; out++) \ samples[out][i] = temp[out]; \ } \ } MIX_FUNC_GENERIC(FLTP, FLT, float, float, float, sum) MIX_FUNC_GENERIC(S16P, FLT, int16_t, float, float, av_clip_int16(lrintf(sum))) MIX_FUNC_GENERIC(S16P, Q15, int16_t, int32_t, int64_t, av_clip_int16(sum >> 15)) MIX_FUNC_GENERIC(S16P, Q8, int16_t, int16_t, int32_t, av_clip_int16(sum >> 8)) /* TODO: templatize the channel-specific C functions */ static void mix_2_to_1_fltp_flt_c(float **samples, float **matrix, int len, int out_ch, int in_ch) { float *src0 = samples[0]; float *src1 = samples[1]; float *dst = src0; float m0 = matrix[0][0]; float m1 = matrix[0][1]; while (len > 4) { *dst++ = *src0++ * m0 + *src1++ * m1; *dst++ = *src0++ * m0 + *src1++ * m1; *dst++ = *src0++ * m0 + *src1++ * m1; *dst++ = *src0++ * m0 + *src1++ * m1; len -= 4; } while (len > 0) { *dst++ = *src0++ * m0 + *src1++ * m1; len--; } } static void mix_2_to_1_s16p_flt_c(int16_t **samples, float **matrix, int len, int out_ch, int in_ch) { int16_t *src0 = samples[0]; int16_t *src1 = samples[1]; int16_t *dst = src0; float m0 = matrix[0][0]; float m1 = matrix[0][1]; while (len > 4) { *dst++ = av_clip_int16(lrintf(*src0++ * m0 + *src1++ * m1)); *dst++ = av_clip_int16(lrintf(*src0++ * m0 + *src1++ * m1)); *dst++ = av_clip_int16(lrintf(*src0++ * m0 + *src1++ * m1)); *dst++ = av_clip_int16(lrintf(*src0++ * m0 + *src1++ * m1)); len -= 4; } while (len > 0) { *dst++ = av_clip_int16(lrintf(*src0++ * m0 + *src1++ * m1)); len--; } } static void mix_2_to_1_s16p_q8_c(int16_t **samples, int16_t **matrix, int len, int out_ch, int in_ch) { int16_t *src0 = samples[0]; int16_t *src1 = samples[1]; int16_t *dst = src0; int16_t m0 = matrix[0][0]; int16_t m1 = matrix[0][1]; while (len > 4) { *dst++ = (*src0++ * m0 + *src1++ * m1) >> 8; *dst++ = (*src0++ * m0 + *src1++ * m1) >> 8; *dst++ = (*src0++ * m0 + *src1++ * m1) >> 8; *dst++ = (*src0++ * m0 + *src1++ * m1) >> 8; len -= 4; } while (len > 0) { *dst++ = (*src0++ * m0 + *src1++ * m1) >> 8; len--; } } static void mix_1_to_2_fltp_flt_c(float **samples, float **matrix, int len, int out_ch, int in_ch) { float v; float *dst0 = samples[0]; float *dst1 = samples[1]; float *src = dst0; float m0 = matrix[0][0]; float m1 = matrix[1][0]; while (len > 4) { v = *src++; *dst0++ = v * m1; *dst1++ = v * m0; v = *src++; *dst0++ = v * m1; *dst1++ = v * m0; v = *src++; *dst0++ = v * m1; *dst1++ = v * m0; v = *src++; *dst0++ = v * m1; *dst1++ = v * m0; len -= 4; } while (len > 0) { v = *src++; *dst0++ = v * m1; *dst1++ = v * m0; len--; } } static void mix_6_to_2_fltp_flt_c(float **samples, float **matrix, int len, int out_ch, int in_ch) { float v0, v1; float *src0 = samples[0]; float *src1 = samples[1]; float *src2 = samples[2]; float *src3 = samples[3]; float *src4 = samples[4]; float *src5 = samples[5]; float *dst0 = src0; float *dst1 = src1; float *m0 = matrix[0]; float *m1 = matrix[1]; while (len > 0) { v0 = *src0++; v1 = *src1++; *dst0++ = v0 * m0[0] + v1 * m0[1] + *src2 * m0[2] + *src3 * m0[3] + *src4 * m0[4] + *src5 * m0[5]; *dst1++ = v0 * m1[0] + v1 * m1[1] + *src2++ * m1[2] + *src3++ * m1[3] + *src4++ * m1[4] + *src5++ * m1[5]; len--; } } static void mix_2_to_6_fltp_flt_c(float **samples, float **matrix, int len, int out_ch, int in_ch) { float v0, v1; float *dst0 = samples[0]; float *dst1 = samples[1]; float *dst2 = samples[2]; float *dst3 = samples[3]; float *dst4 = samples[4]; float *dst5 = samples[5]; float *src0 = dst0; float *src1 = dst1; while (len > 0) { v0 = *src0++; v1 = *src1++; *dst0++ = v0 * matrix[0][0] + v1 * matrix[0][1]; *dst1++ = v0 * matrix[1][0] + v1 * matrix[1][1]; *dst2++ = v0 * matrix[2][0] + v1 * matrix[2][1]; *dst3++ = v0 * matrix[3][0] + v1 * matrix[3][1]; *dst4++ = v0 * matrix[4][0] + v1 * matrix[4][1]; *dst5++ = v0 * matrix[5][0] + v1 * matrix[5][1]; len--; } } static int mix_function_init(AudioMix *am) { am->func_descr = am->func_descr_generic = "n/a"; am->mix = am->mix_generic = NULL; /* no need to set a mix function when we're skipping mixing */ if (!am->in_matrix_channels || !am->out_matrix_channels) return 0; /* any-to-any C versions */ ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT, 0, 0, 1, 1, "C", MIX_FUNC_NAME(FLTP, FLT)); ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_FLT, 0, 0, 1, 1, "C", MIX_FUNC_NAME(S16P, FLT)); ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_Q15, 0, 0, 1, 1, "C", MIX_FUNC_NAME(S16P, Q15)); ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_Q8, 0, 0, 1, 1, "C", MIX_FUNC_NAME(S16P, Q8)); /* channel-specific C versions */ ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT, 2, 1, 1, 1, "C", mix_2_to_1_fltp_flt_c); ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_FLT, 2, 1, 1, 1, "C", mix_2_to_1_s16p_flt_c); ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_Q8, 2, 1, 1, 1, "C", mix_2_to_1_s16p_q8_c); ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT, 1, 2, 1, 1, "C", mix_1_to_2_fltp_flt_c); ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT, 6, 2, 1, 1, "C", mix_6_to_2_fltp_flt_c); ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT, 2, 6, 1, 1, "C", mix_2_to_6_fltp_flt_c); if (ARCH_X86) ff_audio_mix_init_x86(am); if (!am->mix) { av_log(am->avr, AV_LOG_ERROR, "audio_mix: NO FUNCTION FOUND: [fmt=%s] " "[c=%s] [%d to %d]\n", av_get_sample_fmt_name(am->fmt), coeff_type_names[am->coeff_type], am->in_channels, am->out_channels); return AVERROR_PATCHWELCOME; } return 0; } AudioMix *ff_audio_mix_alloc(AVAudioResampleContext *avr) { AudioMix *am; int ret; am = av_mallocz(sizeof(*am)); if (!am) return NULL; am->avr = avr; if (avr->internal_sample_fmt != AV_SAMPLE_FMT_S16P && avr->internal_sample_fmt != AV_SAMPLE_FMT_FLTP) { av_log(avr, AV_LOG_ERROR, "Unsupported internal format for " "mixing: %s\n", av_get_sample_fmt_name(avr->internal_sample_fmt)); goto error; } am->fmt = avr->internal_sample_fmt; am->coeff_type = avr->mix_coeff_type; am->in_layout = avr->in_channel_layout; am->out_layout = avr->out_channel_layout; am->in_channels = avr->in_channels; am->out_channels = avr->out_channels; /* build matrix if the user did not already set one */ if (avr->mix_matrix) { ret = ff_audio_mix_set_matrix(am, avr->mix_matrix, avr->in_channels); if (ret < 0) goto error; av_freep(&avr->mix_matrix); } else { double *matrix_dbl = av_mallocz(avr->out_channels * avr->in_channels * sizeof(*matrix_dbl)); if (!matrix_dbl) goto error; ret = avresample_build_matrix(avr->in_channel_layout, avr->out_channel_layout, avr->center_mix_level, avr->surround_mix_level, avr->lfe_mix_level, avr->normalize_mix_level, matrix_dbl, avr->in_channels, avr->matrix_encoding); if (ret < 0) { av_free(matrix_dbl); goto error; } ret = ff_audio_mix_set_matrix(am, matrix_dbl, avr->in_channels); if (ret < 0) { av_log(avr, AV_LOG_ERROR, "error setting mix matrix\n"); av_free(matrix_dbl); goto error; } av_free(matrix_dbl); } return am; error: av_free(am); return NULL; } void ff_audio_mix_free(AudioMix **am_p) { AudioMix *am; if (!*am_p) return; am = *am_p; if (am->matrix) { av_free(am->matrix[0]); am->matrix = NULL; } memset(am->matrix_q8, 0, sizeof(am->matrix_q8 )); memset(am->matrix_q15, 0, sizeof(am->matrix_q15)); memset(am->matrix_flt, 0, sizeof(am->matrix_flt)); av_freep(am_p); } int ff_audio_mix(AudioMix *am, AudioData *src) { int use_generic = 1; int len = src->nb_samples; int i, j; /* determine whether to use the optimized function based on pointer and samples alignment in both the input and output */ if (am->has_optimized_func) { int aligned_len = FFALIGN(len, am->samples_align); if (!(src->ptr_align % am->ptr_align) && src->samples_align >= aligned_len) { len = aligned_len; use_generic = 0; } } av_dlog(am->avr, "audio_mix: %d samples - %d to %d channels (%s)\n", src->nb_samples, am->in_channels, am->out_channels, use_generic ? am->func_descr_generic : am->func_descr); if (am->in_matrix_channels && am->out_matrix_channels) { uint8_t **data; uint8_t *data0[AVRESAMPLE_MAX_CHANNELS]; if (am->out_matrix_channels < am->out_channels || am->in_matrix_channels < am->in_channels) { for (i = 0, j = 0; i < FFMAX(am->in_channels, am->out_channels); i++) { if (am->input_skip[i] || am->output_skip[i] || am->output_zero[i]) continue; data0[j++] = src->data[i]; } data = data0; } else { data = src->data; } if (use_generic) am->mix_generic(data, am->matrix, len, am->out_matrix_channels, am->in_matrix_channels); else am->mix(data, am->matrix, len, am->out_matrix_channels, am->in_matrix_channels); } if (am->out_matrix_channels < am->out_channels) { for (i = 0; i < am->out_channels; i++) if (am->output_zero[i]) av_samples_set_silence(&src->data[i], 0, len, 1, am->fmt); } ff_audio_data_set_channels(src, am->out_channels); return 0; } int ff_audio_mix_get_matrix(AudioMix *am, double *matrix, int stride) { int i, o, i0, o0; if ( am->in_channels <= 0 || am->in_channels > AVRESAMPLE_MAX_CHANNELS || am->out_channels <= 0 || am->out_channels > AVRESAMPLE_MAX_CHANNELS) { av_log(am->avr, AV_LOG_ERROR, "Invalid channel counts\n"); return AVERROR(EINVAL); } #define GET_MATRIX_CONVERT(suffix, scale) \ if (!am->matrix_ ## suffix[0]) { \ av_log(am->avr, AV_LOG_ERROR, "matrix is not set\n"); \ return AVERROR(EINVAL); \ } \ for (o = 0, o0 = 0; o < am->out_channels; o++) { \ for (i = 0, i0 = 0; i < am->in_channels; i++) { \ if (am->input_skip[i] || am->output_zero[o]) \ matrix[o * stride + i] = 0.0; \ else \ matrix[o * stride + i] = am->matrix_ ## suffix[o0][i0] * \ (scale); \ if (!am->input_skip[i]) \ i0++; \ } \ if (!am->output_zero[o]) \ o0++; \ } switch (am->coeff_type) { case AV_MIX_COEFF_TYPE_Q8: GET_MATRIX_CONVERT(q8, 1.0 / 256.0); break; case AV_MIX_COEFF_TYPE_Q15: GET_MATRIX_CONVERT(q15, 1.0 / 32768.0); break; case AV_MIX_COEFF_TYPE_FLT: GET_MATRIX_CONVERT(flt, 1.0); break; default: av_log(am->avr, AV_LOG_ERROR, "Invalid mix coeff type\n"); return AVERROR(EINVAL); } return 0; } static void reduce_matrix(AudioMix *am, const double *matrix, int stride) { int i, o; memset(am->output_zero, 0, sizeof(am->output_zero)); memset(am->input_skip, 0, sizeof(am->input_skip)); memset(am->output_skip, 0, sizeof(am->output_skip)); /* exclude output channels if they can be zeroed instead of mixed */ for (o = 0; o < am->out_channels; o++) { int zero = 1; /* check if the output is always silent */ for (i = 0; i < am->in_channels; i++) { if (matrix[o * stride + i] != 0.0) { zero = 0; break; } } /* check if the corresponding input channel makes a contribution to any output channel */ if (o < am->in_channels) { for (i = 0; i < am->out_channels; i++) { if (matrix[i * stride + o] != 0.0) { zero = 0; break; } } } if (zero) { am->output_zero[o] = 1; am->out_matrix_channels--; } } if (am->out_matrix_channels == 0) { am->in_matrix_channels = 0; return; } /* skip input channels that contribute fully only to the corresponding output channel */ for (i = 0; i < FFMIN(am->in_channels, am->out_channels); i++) { int skip = 1; for (o = 0; o < am->out_channels; o++) { if ((o != i && matrix[o * stride + i] != 0.0) || (o == i && matrix[o * stride + i] != 1.0)) { skip = 0; break; } } if (skip) { am->input_skip[i] = 1; am->in_matrix_channels--; } } /* skip input channels that do not contribute to any output channel */ for (; i < am->in_channels; i++) { int contrib = 0; for (o = 0; o < am->out_channels; o++) { if (matrix[o * stride + i] != 0.0) { contrib = 1; break; } } if (!contrib) { am->input_skip[i] = 1; am->in_matrix_channels--; } } if (am->in_matrix_channels == 0) { am->out_matrix_channels = 0; return; } /* skip output channels that only get full contribution from the corresponding input channel */ for (o = 0; o < FFMIN(am->in_channels, am->out_channels); o++) { int skip = 1; for (i = 0; i < am->in_channels; i++) { if ((o != i && matrix[o * stride + i] != 0.0) || (o == i && matrix[o * stride + i] != 1.0)) { skip = 0; break; } } if (skip) { am->output_skip[o] = 1; am->out_matrix_channels--; } } if (am->out_matrix_channels == 0) { am->in_matrix_channels = 0; return; } } int ff_audio_mix_set_matrix(AudioMix *am, const double *matrix, int stride) { int i, o, i0, o0, ret; char in_layout_name[128]; char out_layout_name[128]; if ( am->in_channels <= 0 || am->in_channels > AVRESAMPLE_MAX_CHANNELS || am->out_channels <= 0 || am->out_channels > AVRESAMPLE_MAX_CHANNELS) { av_log(am->avr, AV_LOG_ERROR, "Invalid channel counts\n"); return AVERROR(EINVAL); } if (am->matrix) { av_free(am->matrix[0]); am->matrix = NULL; } am->in_matrix_channels = am->in_channels; am->out_matrix_channels = am->out_channels; reduce_matrix(am, matrix, stride); #define CONVERT_MATRIX(type, expr) \ am->matrix_## type[0] = av_mallocz(am->out_matrix_channels * \ am->in_matrix_channels * \ sizeof(*am->matrix_## type[0])); \ if (!am->matrix_## type[0]) \ return AVERROR(ENOMEM); \ for (o = 0, o0 = 0; o < am->out_channels; o++) { \ if (am->output_zero[o] || am->output_skip[o]) \ continue; \ if (o0 > 0) \ am->matrix_## type[o0] = am->matrix_## type[o0 - 1] + \ am->in_matrix_channels; \ for (i = 0, i0 = 0; i < am->in_channels; i++) { \ double v; \ if (am->input_skip[i]) \ continue; \ v = matrix[o * stride + i]; \ am->matrix_## type[o0][i0] = expr; \ i0++; \ } \ o0++; \ } \ am->matrix = (void **)am->matrix_## type; if (am->in_matrix_channels && am->out_matrix_channels) { switch (am->coeff_type) { case AV_MIX_COEFF_TYPE_Q8: CONVERT_MATRIX(q8, av_clip_int16(lrint(256.0 * v))) break; case AV_MIX_COEFF_TYPE_Q15: CONVERT_MATRIX(q15, av_clipl_int32(llrint(32768.0 * v))) break; case AV_MIX_COEFF_TYPE_FLT: CONVERT_MATRIX(flt, v) break; default: av_log(am->avr, AV_LOG_ERROR, "Invalid mix coeff type\n"); return AVERROR(EINVAL); } } ret = mix_function_init(am); if (ret < 0) return ret; av_get_channel_layout_string(in_layout_name, sizeof(in_layout_name), am->in_channels, am->in_layout); av_get_channel_layout_string(out_layout_name, sizeof(out_layout_name), am->out_channels, am->out_layout); av_log(am->avr, AV_LOG_DEBUG, "audio_mix: %s to %s\n", in_layout_name, out_layout_name); av_log(am->avr, AV_LOG_DEBUG, "matrix size: %d x %d\n", am->in_matrix_channels, am->out_matrix_channels); for (o = 0; o < am->out_channels; o++) { for (i = 0; i < am->in_channels; i++) { if (am->output_zero[o]) av_log(am->avr, AV_LOG_DEBUG, " (ZERO)"); else if (am->input_skip[i] || am->output_skip[o]) av_log(am->avr, AV_LOG_DEBUG, " (SKIP)"); else av_log(am->avr, AV_LOG_DEBUG, " %0.3f ", matrix[o * am->in_channels + i]); } av_log(am->avr, AV_LOG_DEBUG, "\n"); } return 0; }