/*** This file is part of PulseAudio. Copyright 2004-2006 Lennart Poettering PulseAudio 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. PulseAudio 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 General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with PulseAudio; if not, see . ***/ #ifdef HAVE_CONFIG_H #include #endif #include #include #include #include #include #include #include #include #include "resampler.h" /* Number of samples of extra space we allow the resamplers to return */ #define EXTRA_FRAMES 128 struct ffmpeg_data { /* data specific to ffmpeg */ struct AVResampleContext *state; }; static int copy_init(pa_resampler *r); static void setup_remap(const pa_resampler *r, pa_remap_t *m, bool *lfe_remixed); static void free_remap(pa_remap_t *m); static int (* const init_table[])(pa_resampler *r) = { #ifdef HAVE_LIBSAMPLERATE [PA_RESAMPLER_SRC_SINC_BEST_QUALITY] = pa_resampler_libsamplerate_init, [PA_RESAMPLER_SRC_SINC_MEDIUM_QUALITY] = pa_resampler_libsamplerate_init, [PA_RESAMPLER_SRC_SINC_FASTEST] = pa_resampler_libsamplerate_init, [PA_RESAMPLER_SRC_ZERO_ORDER_HOLD] = pa_resampler_libsamplerate_init, [PA_RESAMPLER_SRC_LINEAR] = pa_resampler_libsamplerate_init, #else [PA_RESAMPLER_SRC_SINC_BEST_QUALITY] = NULL, [PA_RESAMPLER_SRC_SINC_MEDIUM_QUALITY] = NULL, [PA_RESAMPLER_SRC_SINC_FASTEST] = NULL, [PA_RESAMPLER_SRC_ZERO_ORDER_HOLD] = NULL, [PA_RESAMPLER_SRC_LINEAR] = NULL, #endif [PA_RESAMPLER_TRIVIAL] = pa_resampler_trivial_init, #ifdef HAVE_SPEEX [PA_RESAMPLER_SPEEX_FLOAT_BASE+0] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+1] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+2] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+3] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+4] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+5] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+6] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+7] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+8] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+9] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FLOAT_BASE+10] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+0] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+1] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+2] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+3] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+4] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+5] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+6] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+7] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+8] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+9] = pa_resampler_speex_init, [PA_RESAMPLER_SPEEX_FIXED_BASE+10] = pa_resampler_speex_init, #else [PA_RESAMPLER_SPEEX_FLOAT_BASE+0] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+1] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+2] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+3] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+4] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+5] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+6] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+7] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+8] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+9] = NULL, [PA_RESAMPLER_SPEEX_FLOAT_BASE+10] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+0] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+1] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+2] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+3] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+4] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+5] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+6] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+7] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+8] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+9] = NULL, [PA_RESAMPLER_SPEEX_FIXED_BASE+10] = NULL, #endif [PA_RESAMPLER_FFMPEG] = pa_resampler_ffmpeg_init, [PA_RESAMPLER_AUTO] = NULL, [PA_RESAMPLER_COPY] = copy_init, [PA_RESAMPLER_PEAKS] = pa_resampler_peaks_init, #ifdef HAVE_SOXR [PA_RESAMPLER_SOXR_MQ] = pa_resampler_soxr_init, [PA_RESAMPLER_SOXR_HQ] = pa_resampler_soxr_init, [PA_RESAMPLER_SOXR_VHQ] = pa_resampler_soxr_init, #else [PA_RESAMPLER_SOXR_MQ] = NULL, [PA_RESAMPLER_SOXR_HQ] = NULL, [PA_RESAMPLER_SOXR_VHQ] = NULL, #endif }; static void calculate_gcd(pa_resampler *r) { unsigned gcd, n; pa_assert(r); gcd = r->i_ss.rate; n = r->o_ss.rate; while (n != 0) { unsigned tmp = gcd; gcd = n; n = tmp % n; } r->gcd = gcd; } static pa_resample_method_t choose_auto_resampler(pa_resample_flags_t flags) { pa_resample_method_t method; if (pa_resample_method_supported(PA_RESAMPLER_SPEEX_FLOAT_BASE + 1)) method = PA_RESAMPLER_SPEEX_FLOAT_BASE + 1; else if (flags & PA_RESAMPLER_VARIABLE_RATE) method = PA_RESAMPLER_TRIVIAL; else method = PA_RESAMPLER_FFMPEG; return method; } static pa_resample_method_t fix_method( pa_resample_flags_t flags, pa_resample_method_t method, const uint32_t rate_a, const uint32_t rate_b) { pa_assert(pa_sample_rate_valid(rate_a)); pa_assert(pa_sample_rate_valid(rate_b)); pa_assert(method >= 0); pa_assert(method < PA_RESAMPLER_MAX); if (!(flags & PA_RESAMPLER_VARIABLE_RATE) && rate_a == rate_b) { pa_log_info("Forcing resampler 'copy', because of fixed, identical sample rates."); method = PA_RESAMPLER_COPY; } if (!pa_resample_method_supported(method)) { pa_log_warn("Support for resampler '%s' not compiled in, reverting to 'auto'.", pa_resample_method_to_string(method)); method = PA_RESAMPLER_AUTO; } switch (method) { case PA_RESAMPLER_COPY: if (rate_a != rate_b) { pa_log_info("Resampler 'copy' cannot change sampling rate, reverting to resampler 'auto'."); method = PA_RESAMPLER_AUTO; break; } /* Else fall through */ case PA_RESAMPLER_FFMPEG: if (flags & PA_RESAMPLER_VARIABLE_RATE) { pa_log_info("Resampler '%s' cannot do variable rate, reverting to resampler 'auto'.", pa_resample_method_to_string(method)); method = PA_RESAMPLER_AUTO; } break; /* The Peaks resampler only supports downsampling. * Revert to auto if we are upsampling */ case PA_RESAMPLER_PEAKS: if (rate_a < rate_b) { pa_log_warn("The 'peaks' resampler only supports downsampling, reverting to resampler 'auto'."); method = PA_RESAMPLER_AUTO; } break; default: break; } if (method == PA_RESAMPLER_AUTO) method = choose_auto_resampler(flags); #ifdef HAVE_SPEEX /* At this point, method is supported in the sense that it * has an init function and supports the required flags. However, * speex-float implementation in PulseAudio relies on the * assumption that is invalid if speex has been compiled with * --enable-fixed-point. Besides, speex-fixed is more efficient * in this configuration. So use it instead. */ if (method >= PA_RESAMPLER_SPEEX_FLOAT_BASE && method <= PA_RESAMPLER_SPEEX_FLOAT_MAX) { if (pa_speex_is_fixed_point()) { pa_log_info("Speex appears to be compiled with --enable-fixed-point. " "Switching to a fixed-point resampler because it should be faster."); method = method - PA_RESAMPLER_SPEEX_FLOAT_BASE + PA_RESAMPLER_SPEEX_FIXED_BASE; } } #endif return method; } /* Return true if a is a more precise sample format than b, else return false */ static bool sample_format_more_precise(pa_sample_format_t a, pa_sample_format_t b) { pa_assert(pa_sample_format_valid(a)); pa_assert(pa_sample_format_valid(b)); switch (a) { case PA_SAMPLE_U8: case PA_SAMPLE_ALAW: case PA_SAMPLE_ULAW: return false; break; case PA_SAMPLE_S16LE: case PA_SAMPLE_S16BE: if (b == PA_SAMPLE_ULAW || b == PA_SAMPLE_ALAW || b == PA_SAMPLE_U8) return true; else return false; break; case PA_SAMPLE_S24LE: case PA_SAMPLE_S24BE: case PA_SAMPLE_S24_32LE: case PA_SAMPLE_S24_32BE: if (b == PA_SAMPLE_ULAW || b == PA_SAMPLE_ALAW || b == PA_SAMPLE_U8 || b == PA_SAMPLE_S16LE || b == PA_SAMPLE_S16BE) return true; else return false; break; case PA_SAMPLE_FLOAT32LE: case PA_SAMPLE_FLOAT32BE: case PA_SAMPLE_S32LE: case PA_SAMPLE_S32BE: if (b == PA_SAMPLE_FLOAT32LE || b == PA_SAMPLE_FLOAT32BE || b == PA_SAMPLE_S32LE || b == PA_SAMPLE_S32BE) return false; else return true; break; default: return false; } } static pa_sample_format_t choose_work_format( pa_resample_method_t method, pa_sample_format_t a, pa_sample_format_t b, bool map_required) { pa_sample_format_t work_format; pa_assert(pa_sample_format_valid(a)); pa_assert(pa_sample_format_valid(b)); pa_assert(method >= 0); pa_assert(method < PA_RESAMPLER_MAX); if (method >= PA_RESAMPLER_SPEEX_FIXED_BASE && method <= PA_RESAMPLER_SPEEX_FIXED_MAX) method = PA_RESAMPLER_SPEEX_FIXED_BASE; switch (method) { /* This block is for resampling functions that only * support the S16 sample format. */ case PA_RESAMPLER_SPEEX_FIXED_BASE: case PA_RESAMPLER_FFMPEG: work_format = PA_SAMPLE_S16NE; break; /* This block is for resampling functions that support * any sample format. */ case PA_RESAMPLER_COPY: case PA_RESAMPLER_TRIVIAL: if (!map_required && a == b) { work_format = a; break; } /* If both input and output are using S32NE and we don't * need any resampling we can use S32NE directly, avoiding * converting back and forth between S32NE and * FLOAT32NE. */ if ((a == PA_SAMPLE_S32NE) && (b == PA_SAMPLE_S32NE)) { work_format = PA_SAMPLE_S32NE; break; } /* Else fall through */ case PA_RESAMPLER_PEAKS: /* PEAKS, COPY and TRIVIAL do not benefit from increased * working precision, so for better performance use s16ne * if either input or output fits in it. */ if (a == PA_SAMPLE_S16NE || b == PA_SAMPLE_S16NE) { work_format = PA_SAMPLE_S16NE; break; } /* Else fall through */ case PA_RESAMPLER_SOXR_MQ: case PA_RESAMPLER_SOXR_HQ: case PA_RESAMPLER_SOXR_VHQ: /* Do processing with max precision of input and output. */ if (sample_format_more_precise(a, PA_SAMPLE_S16NE) || sample_format_more_precise(b, PA_SAMPLE_S16NE)) work_format = PA_SAMPLE_FLOAT32NE; else work_format = PA_SAMPLE_S16NE; break; default: work_format = PA_SAMPLE_FLOAT32NE; } return work_format; } pa_resampler* pa_resampler_new( pa_mempool *pool, const pa_sample_spec *a, const pa_channel_map *am, const pa_sample_spec *b, const pa_channel_map *bm, unsigned crossover_freq, pa_resample_method_t method, pa_resample_flags_t flags) { pa_resampler *r = NULL; bool lfe_remixed = false; pa_assert(pool); pa_assert(a); pa_assert(b); pa_assert(pa_sample_spec_valid(a)); pa_assert(pa_sample_spec_valid(b)); pa_assert(method >= 0); pa_assert(method < PA_RESAMPLER_MAX); method = fix_method(flags, method, a->rate, b->rate); r = pa_xnew0(pa_resampler, 1); r->mempool = pool; r->method = method; r->flags = flags; r->in_frames = 0; r->out_frames = 0; /* Fill sample specs */ r->i_ss = *a; r->o_ss = *b; calculate_gcd(r); if (am) r->i_cm = *am; else if (!pa_channel_map_init_auto(&r->i_cm, r->i_ss.channels, PA_CHANNEL_MAP_DEFAULT)) goto fail; if (bm) r->o_cm = *bm; else if (!pa_channel_map_init_auto(&r->o_cm, r->o_ss.channels, PA_CHANNEL_MAP_DEFAULT)) goto fail; r->i_fz = pa_frame_size(a); r->o_fz = pa_frame_size(b); r->map_required = (r->i_ss.channels != r->o_ss.channels || (!(r->flags & PA_RESAMPLER_NO_REMAP) && !pa_channel_map_equal(&r->i_cm, &r->o_cm))); r->work_format = choose_work_format(method, a->format, b->format, r->map_required); r->w_sz = pa_sample_size_of_format(r->work_format); if (r->i_ss.format != r->work_format) { if (r->work_format == PA_SAMPLE_FLOAT32NE) { if (!(r->to_work_format_func = pa_get_convert_to_float32ne_function(r->i_ss.format))) goto fail; } else { pa_assert(r->work_format == PA_SAMPLE_S16NE); if (!(r->to_work_format_func = pa_get_convert_to_s16ne_function(r->i_ss.format))) goto fail; } } if (r->o_ss.format != r->work_format) { if (r->work_format == PA_SAMPLE_FLOAT32NE) { if (!(r->from_work_format_func = pa_get_convert_from_float32ne_function(r->o_ss.format))) goto fail; } else { pa_assert(r->work_format == PA_SAMPLE_S16NE); if (!(r->from_work_format_func = pa_get_convert_from_s16ne_function(r->o_ss.format))) goto fail; } } if (r->o_ss.channels <= r->i_ss.channels) { /* pipeline is: format conv. -> remap -> resample -> format conv. */ r->work_channels = r->o_ss.channels; /* leftover buffer is remap output buffer (before resampling) */ r->leftover_buf = &r->remap_buf; r->leftover_buf_size = &r->remap_buf_size; r->have_leftover = &r->leftover_in_remap; } else { /* pipeline is: format conv. -> resample -> remap -> format conv. */ r->work_channels = r->i_ss.channels; /* leftover buffer is to_work output buffer (before resampling) */ r->leftover_buf = &r->to_work_format_buf; r->leftover_buf_size = &r->to_work_format_buf_size; r->have_leftover = &r->leftover_in_to_work; } r->w_fz = pa_sample_size_of_format(r->work_format) * r->work_channels; pa_log_debug("Resampler:"); pa_log_debug(" rate %d -> %d (method %s)", a->rate, b->rate, pa_resample_method_to_string(r->method)); pa_log_debug(" format %s -> %s (intermediate %s)", pa_sample_format_to_string(a->format), pa_sample_format_to_string(b->format), pa_sample_format_to_string(r->work_format)); pa_log_debug(" channels %d -> %d (resampling %d)", a->channels, b->channels, r->work_channels); /* set up the remap structure */ if (r->map_required) setup_remap(r, &r->remap, &lfe_remixed); if (lfe_remixed && crossover_freq > 0) { pa_sample_spec wss = r->o_ss; wss.format = r->work_format; /* FIXME: For now just hardcode maxrewind to 3 seconds */ r->lfe_filter = pa_lfe_filter_new(&wss, &r->o_cm, (float)crossover_freq, b->rate * 3); pa_log_debug(" lfe filter activated (LR4 type), the crossover_freq = %uHz", crossover_freq); } /* initialize implementation */ if (init_table[method](r) < 0) goto fail; return r; fail: if (r->lfe_filter) pa_lfe_filter_free(r->lfe_filter); pa_xfree(r); return NULL; } void pa_resampler_free(pa_resampler *r) { pa_assert(r); if (r->impl.free) r->impl.free(r); else pa_xfree(r->impl.data); if (r->lfe_filter) pa_lfe_filter_free(r->lfe_filter); if (r->to_work_format_buf.memblock) pa_memblock_unref(r->to_work_format_buf.memblock); if (r->remap_buf.memblock) pa_memblock_unref(r->remap_buf.memblock); if (r->resample_buf.memblock) pa_memblock_unref(r->resample_buf.memblock); if (r->from_work_format_buf.memblock) pa_memblock_unref(r->from_work_format_buf.memblock); free_remap(&r->remap); pa_xfree(r); } void pa_resampler_set_input_rate(pa_resampler *r, uint32_t rate) { pa_assert(r); pa_assert(rate > 0); pa_assert(r->impl.update_rates); if (r->i_ss.rate == rate) return; /* Recalculate delay counters */ r->in_frames = pa_resampler_get_delay(r, false); r->out_frames = 0; r->i_ss.rate = rate; calculate_gcd(r); r->impl.update_rates(r); } void pa_resampler_set_output_rate(pa_resampler *r, uint32_t rate) { pa_assert(r); pa_assert(rate > 0); pa_assert(r->impl.update_rates); if (r->o_ss.rate == rate) return; /* Recalculate delay counters */ r->in_frames = pa_resampler_get_delay(r, false); r->out_frames = 0; r->o_ss.rate = rate; calculate_gcd(r); r->impl.update_rates(r); if (r->lfe_filter) pa_lfe_filter_update_rate(r->lfe_filter, rate); } /* pa_resampler_request() and pa_resampler_result() should be as exact as * possible to ensure that no samples are lost or duplicated during rewinds. * Ignore the leftover buffer, the value appears to be wrong for ffmpeg * and 0 in all other cases. If the resampler is NULL it means that no * resampling is necessary and the input length equals the output length. * FIXME: These functions are not exact for the soxr resamplers because * soxr uses a different algorithm. */ size_t pa_resampler_request(pa_resampler *r, size_t out_length) { size_t in_length; if (!r || out_length == 0) return out_length; /* Convert to output frames */ out_length = out_length / r->o_fz; /* Convert to input frames. The equation matches exactly the * behavior of the used resamplers and will calculate the * minimum number of input frames that are needed to produce * the given number of output frames. */ in_length = (out_length - 1) * r->i_ss.rate / r->o_ss.rate + 1; /* Convert to input length */ return in_length * r->i_fz; } size_t pa_resampler_result(pa_resampler *r, size_t in_length) { size_t out_length; if (!r) return in_length; /* Convert to intput frames */ in_length = in_length / r->i_fz; /* soxr processes samples in blocks, depending on the ratio. * Therefore samples that do not fit into a block must be * ignored. */ if (r->method == PA_RESAMPLER_SOXR_MQ || r->method == PA_RESAMPLER_SOXR_HQ || r->method == PA_RESAMPLER_SOXR_VHQ) { double ratio; size_t block_size; int k; ratio = (double)r->i_ss.rate / (double)r->o_ss.rate; for (k = 0; k < 7; k++) { if (ratio < pow(2, k + 1)) break; } block_size = pow(2, k); in_length = in_length - in_length % block_size; } /* Convert to output frames. This matches exactly the algorithm * used by the resamplers except for the soxr resamplers. */ out_length = in_length * r->o_ss.rate / r->i_ss.rate; if ((double)in_length * (double)r->o_ss.rate / (double)r->i_ss.rate - out_length > 0) out_length++; /* The libsamplerate resamplers return one sample more if the result is integral and the ratio is not integral. */ else if (r->method >= PA_RESAMPLER_SRC_SINC_BEST_QUALITY && r->method <= PA_RESAMPLER_SRC_SINC_FASTEST && r->i_ss.rate > r->o_ss.rate && r->i_ss.rate % r->o_ss.rate > 0 && (double)in_length * (double)r->o_ss.rate / (double)r->i_ss.rate - out_length <= 0) out_length++; else if (r->method == PA_RESAMPLER_SRC_ZERO_ORDER_HOLD && r->i_ss.rate > r->o_ss.rate && (double)in_length * (double)r->o_ss.rate / (double)r->i_ss.rate - out_length <= 0) out_length++; /* Convert to output length */ return out_length * r->o_fz; } size_t pa_resampler_max_block_size(pa_resampler *r) { size_t block_size_max; pa_sample_spec max_ss; size_t max_fs; size_t frames; pa_assert(r); block_size_max = pa_mempool_block_size_max(r->mempool); /* We deduce the "largest" sample spec we're using during the * conversion */ max_ss.channels = (uint8_t) (PA_MAX(r->i_ss.channels, r->o_ss.channels)); /* We silently assume that the format enum is ordered by size */ max_ss.format = PA_MAX(r->i_ss.format, r->o_ss.format); max_ss.format = PA_MAX(max_ss.format, r->work_format); max_ss.rate = PA_MAX(r->i_ss.rate, r->o_ss.rate); max_fs = pa_frame_size(&max_ss); frames = block_size_max / max_fs - EXTRA_FRAMES; pa_assert(frames >= (r->leftover_buf->length / r->w_fz)); if (*r->have_leftover) frames -= r->leftover_buf->length / r->w_fz; block_size_max = ((uint64_t) frames * r->i_ss.rate / max_ss.rate) * r->i_fz; if (block_size_max > 0) return block_size_max; else /* A single input frame may result in so much output that it doesn't * fit in one standard memblock (e.g. converting 1 Hz to 44100 Hz). In * this case the max block size will be set to one frame, and some * memory will be probably be allocated with malloc() instead of using * the memory pool. * * XXX: Should we support this case at all? We could also refuse to * create resamplers whose max block size would exceed the memory pool * block size. In this case also updating the resampler rate should * fail if the new rate would cause an excessive max block size (in * which case the stream would probably have to be killed). */ return r->i_fz; } void pa_resampler_reset(pa_resampler *r) { pa_assert(r); if (r->impl.reset) r->impl.reset(r); if (r->lfe_filter) pa_lfe_filter_reset(r->lfe_filter); *r->have_leftover = false; r->in_frames = 0; r->out_frames = 0; } /* This function runs amount bytes of data from the history queue through the * resampler and discards the result. The history queue is unchanged after the * call. This is used to preload a resampler after a reset. Returns the number * of frames produced by the resampler. */ size_t pa_resampler_prepare(pa_resampler *r, pa_memblockq *history_queue, size_t amount) { size_t history_bytes, max_block_size, out_size; int64_t to_run; pa_assert(r); if (!history_queue || amount == 0) return 0; /* Rewind the LFE filter by the amount of history data. */ history_bytes = pa_resampler_result(r, amount); if (r->lfe_filter) pa_lfe_filter_rewind(r->lfe_filter, history_bytes); pa_memblockq_rewind(history_queue, amount); max_block_size = pa_resampler_max_block_size(r); to_run = amount; out_size = 0; while (to_run > 0) { pa_memchunk in_chunk, out_chunk; size_t current; current = PA_MIN(to_run, (int64_t) max_block_size); /* Get data from memblockq */ if (pa_memblockq_peek_fixed_size(history_queue, current, &in_chunk) < 0) { pa_log_warn("Could not read history data for resampler."); /* Restore queue to original state and reset resampler */ pa_memblockq_drop(history_queue, to_run); pa_resampler_reset(r); return out_size; } /* Run the resampler */ pa_resampler_run(r, &in_chunk, &out_chunk); /* Discard result */ if (out_chunk.length != 0) { out_size += out_chunk.length; pa_memblock_unref(out_chunk.memblock); } pa_memblock_unref(in_chunk.memblock); pa_memblockq_drop(history_queue, current); to_run -= current; } return out_size; } size_t pa_resampler_rewind(pa_resampler *r, size_t out_bytes, pa_memblockq *history_queue, size_t amount) { pa_assert(r); /* For now, we don't have any rewindable resamplers, so we just reset * the resampler if we cannot rewind using pa_resampler_prepare(). */ if (r->impl.reset && !history_queue) r->impl.reset(r); if (r->lfe_filter) pa_lfe_filter_rewind(r->lfe_filter, out_bytes); if (!history_queue) { *r->have_leftover = false; r->in_frames = 0; r->out_frames = 0; } if (history_queue && amount > 0) return pa_resampler_prepare(r, history_queue, amount); return 0; } pa_resample_method_t pa_resampler_get_method(pa_resampler *r) { pa_assert(r); return r->method; } const pa_channel_map* pa_resampler_input_channel_map(pa_resampler *r) { pa_assert(r); return &r->i_cm; } const pa_sample_spec* pa_resampler_input_sample_spec(pa_resampler *r) { pa_assert(r); return &r->i_ss; } const pa_channel_map* pa_resampler_output_channel_map(pa_resampler *r) { pa_assert(r); return &r->o_cm; } const pa_sample_spec* pa_resampler_output_sample_spec(pa_resampler *r) { pa_assert(r); return &r->o_ss; } static const char * const resample_methods[] = { "src-sinc-best-quality", "src-sinc-medium-quality", "src-sinc-fastest", "src-zero-order-hold", "src-linear", "trivial", "speex-float-0", "speex-float-1", "speex-float-2", "speex-float-3", "speex-float-4", "speex-float-5", "speex-float-6", "speex-float-7", "speex-float-8", "speex-float-9", "speex-float-10", "speex-fixed-0", "speex-fixed-1", "speex-fixed-2", "speex-fixed-3", "speex-fixed-4", "speex-fixed-5", "speex-fixed-6", "speex-fixed-7", "speex-fixed-8", "speex-fixed-9", "speex-fixed-10", "ffmpeg", "auto", "copy", "peaks", "soxr-mq", "soxr-hq", "soxr-vhq" }; const char *pa_resample_method_to_string(pa_resample_method_t m) { if (m < 0 || m >= PA_RESAMPLER_MAX) return NULL; return resample_methods[m]; } int pa_resample_method_supported(pa_resample_method_t m) { if (m < 0 || m >= PA_RESAMPLER_MAX) return 0; #ifndef HAVE_LIBSAMPLERATE if (m <= PA_RESAMPLER_SRC_LINEAR) return 0; #endif #ifndef HAVE_SPEEX if (m >= PA_RESAMPLER_SPEEX_FLOAT_BASE && m <= PA_RESAMPLER_SPEEX_FLOAT_MAX) return 0; if (m >= PA_RESAMPLER_SPEEX_FIXED_BASE && m <= PA_RESAMPLER_SPEEX_FIXED_MAX) return 0; #endif #ifndef HAVE_SOXR if (m >= PA_RESAMPLER_SOXR_MQ && m <= PA_RESAMPLER_SOXR_VHQ) return 0; #endif return 1; } pa_resample_method_t pa_parse_resample_method(const char *string) { pa_resample_method_t m; pa_assert(string); for (m = 0; m < PA_RESAMPLER_MAX; m++) if (pa_streq(string, resample_methods[m])) return m; if (pa_streq(string, "speex-fixed")) return PA_RESAMPLER_SPEEX_FIXED_BASE + 1; if (pa_streq(string, "speex-float")) return PA_RESAMPLER_SPEEX_FLOAT_BASE + 1; return PA_RESAMPLER_INVALID; } static bool on_left(pa_channel_position_t p) { return p == PA_CHANNEL_POSITION_FRONT_LEFT || p == PA_CHANNEL_POSITION_REAR_LEFT || p == PA_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER || p == PA_CHANNEL_POSITION_SIDE_LEFT || p == PA_CHANNEL_POSITION_TOP_FRONT_LEFT || p == PA_CHANNEL_POSITION_TOP_REAR_LEFT; } static bool on_right(pa_channel_position_t p) { return p == PA_CHANNEL_POSITION_FRONT_RIGHT || p == PA_CHANNEL_POSITION_REAR_RIGHT || p == PA_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER || p == PA_CHANNEL_POSITION_SIDE_RIGHT || p == PA_CHANNEL_POSITION_TOP_FRONT_RIGHT || p == PA_CHANNEL_POSITION_TOP_REAR_RIGHT; } static bool on_center(pa_channel_position_t p) { return p == PA_CHANNEL_POSITION_FRONT_CENTER || p == PA_CHANNEL_POSITION_REAR_CENTER || p == PA_CHANNEL_POSITION_TOP_CENTER || p == PA_CHANNEL_POSITION_TOP_FRONT_CENTER || p == PA_CHANNEL_POSITION_TOP_REAR_CENTER; } static bool on_lfe(pa_channel_position_t p) { return p == PA_CHANNEL_POSITION_LFE; } static bool on_front(pa_channel_position_t p) { return p == PA_CHANNEL_POSITION_FRONT_LEFT || p == PA_CHANNEL_POSITION_FRONT_RIGHT || p == PA_CHANNEL_POSITION_FRONT_CENTER || p == PA_CHANNEL_POSITION_TOP_FRONT_LEFT || p == PA_CHANNEL_POSITION_TOP_FRONT_RIGHT || p == PA_CHANNEL_POSITION_TOP_FRONT_CENTER || p == PA_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER || p == PA_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER; } static bool on_rear(pa_channel_position_t p) { return p == PA_CHANNEL_POSITION_REAR_LEFT || p == PA_CHANNEL_POSITION_REAR_RIGHT || p == PA_CHANNEL_POSITION_REAR_CENTER || p == PA_CHANNEL_POSITION_TOP_REAR_LEFT || p == PA_CHANNEL_POSITION_TOP_REAR_RIGHT || p == PA_CHANNEL_POSITION_TOP_REAR_CENTER; } static bool on_side(pa_channel_position_t p) { return p == PA_CHANNEL_POSITION_SIDE_LEFT || p == PA_CHANNEL_POSITION_SIDE_RIGHT || p == PA_CHANNEL_POSITION_TOP_CENTER; } enum { ON_FRONT, ON_REAR, ON_SIDE, ON_OTHER }; static int front_rear_side(pa_channel_position_t p) { if (on_front(p)) return ON_FRONT; if (on_rear(p)) return ON_REAR; if (on_side(p)) return ON_SIDE; return ON_OTHER; } /* Fill a map of which output channels should get mono from input, not including * LFE output channels. (The LFE output channels are mapped separately.) */ static void setup_oc_mono_map(const pa_resampler *r, float *oc_mono_map) { unsigned oc; unsigned n_oc; bool found_oc_for_mono = false; pa_assert(r); pa_assert(oc_mono_map); n_oc = r->o_ss.channels; if (!(r->flags & PA_RESAMPLER_NO_FILL_SINK)) { /* Mono goes to all non-LFE output channels and we're done. */ for (oc = 0; oc < n_oc; oc++) oc_mono_map[oc] = on_lfe(r->o_cm.map[oc]) ? 0.0f : 1.0f; return; } else { /* Initialize to all zero so we can select individual channels below. */ for (oc = 0; oc < n_oc; oc++) oc_mono_map[oc] = 0.0f; } for (oc = 0; oc < n_oc; oc++) { if (r->o_cm.map[oc] == PA_CHANNEL_POSITION_MONO) { oc_mono_map[oc] = 1.0f; found_oc_for_mono = true; } } if (found_oc_for_mono) return; for (oc = 0; oc < n_oc; oc++) { if (r->o_cm.map[oc] == PA_CHANNEL_POSITION_FRONT_CENTER) { oc_mono_map[oc] = 1.0f; found_oc_for_mono = true; } } if (found_oc_for_mono) return; for (oc = 0; oc < n_oc; oc++) { if (r->o_cm.map[oc] == PA_CHANNEL_POSITION_FRONT_LEFT || r->o_cm.map[oc] == PA_CHANNEL_POSITION_FRONT_RIGHT) { oc_mono_map[oc] = 1.0f; found_oc_for_mono = true; } } if (found_oc_for_mono) return; /* Give up on finding a suitable map for mono, and just send it to all * non-LFE output channels. */ for (oc = 0; oc < n_oc; oc++) oc_mono_map[oc] = on_lfe(r->o_cm.map[oc]) ? 0.0f : 1.0f; } static void setup_remap(const pa_resampler *r, pa_remap_t *m, bool *lfe_remixed) { unsigned oc, ic; unsigned n_oc, n_ic; bool ic_connected[PA_CHANNELS_MAX]; pa_strbuf *s; char *t; pa_assert(r); pa_assert(m); pa_assert(lfe_remixed); n_oc = r->o_ss.channels; n_ic = r->i_ss.channels; m->format = r->work_format; m->i_ss = r->i_ss; m->o_ss = r->o_ss; memset(m->map_table_f, 0, sizeof(m->map_table_f)); memset(m->map_table_i, 0, sizeof(m->map_table_i)); memset(ic_connected, 0, sizeof(ic_connected)); *lfe_remixed = false; if (r->flags & PA_RESAMPLER_NO_REMAP) { for (oc = 0; oc < PA_MIN(n_ic, n_oc); oc++) m->map_table_f[oc][oc] = 1.0f; } else if (r->flags & PA_RESAMPLER_NO_REMIX) { for (oc = 0; oc < n_oc; oc++) { pa_channel_position_t b = r->o_cm.map[oc]; for (ic = 0; ic < n_ic; ic++) { pa_channel_position_t a = r->i_cm.map[ic]; /* We shall not do any remixing. Hence, just check by name */ if (a == b) m->map_table_f[oc][ic] = 1.0f; } } } else { /* OK, we shall do the full monty: upmixing and downmixing. Our * algorithm is relatively simple, does not do spacialization, or delay * elements. LFE filters are done after the remap step. Patches are always * welcome, though. Oh, and it doesn't do any matrix decoding. (Which * probably wouldn't make any sense anyway.) * * This code is not idempotent: downmixing an upmixed stereo stream is * not identical to the original. The volume will not match, and the * two channels will be a linear combination of both. * * This is loosely based on random suggestions found on the Internet, * such as this: * http://www.halfgaar.net/surround-sound-in-linux and the alsa upmix * plugin. * * The algorithm works basically like this: * * 1) Connect all channels with matching names. * This also includes fixing confusion between "5.1" and * "5.1 (Side)" layouts, done by mpv. * * 2) Mono Handling: * S:Mono: See setup_oc_mono_map(). * D:Mono: Avg all S:channels * * 3) Mix D:Left, D:Right (if PA_RESAMPLER_NO_FILL_SINK is clear): * D:Left: If not connected, avg all S:Left * D:Right: If not connected, avg all S:Right * * 4) Mix D:Center (if PA_RESAMPLER_NO_FILL_SINK is clear): * If not connected, avg all S:Center * If still not connected, avg all S:Left, S:Right * * 5) Mix D:LFE * If not connected, avg all S:* * * 6) Make sure S:Left/S:Right is used: S:Left/S:Right: If not * connected, mix into all D:left and all D:right channels. Gain is * 1/9. * * 7) Make sure S:Center, S:LFE is used: * * S:Center, S:LFE: If not connected, mix into all D:left, all * D:right, all D:center channels. Gain is 0.5 for center and 0.375 * for LFE. C-front is only mixed into L-front/R-front if available, * otherwise into all L/R channels. Similarly for C-rear. * * 8) Normalize each row in the matrix such that the sum for each row is * not larger than 1.0 in order to avoid clipping. * * S: and D: shall relate to the source resp. destination channels. * * Rationale: 1, 2 are probably obvious. For 3: this copies front to * rear if needed. For 4: we try to find some suitable C source for C, * if we don't find any, we avg L and R. For 5: LFE is mixed from all * channels. For 6: the rear channels should not be dropped entirely, * however have only minimal impact. For 7: movies usually encode * speech on the center channel. Thus we have to make sure this channel * is distributed to L and R if not available in the output. Also, LFE * is used to achieve a greater dynamic range, and thus we should try * to do our best to pass it to L+R. */ unsigned ic_left = 0, ic_right = 0, ic_center = 0, ic_unconnected_left = 0, ic_unconnected_right = 0, ic_unconnected_center = 0, ic_unconnected_lfe = 0; bool ic_unconnected_center_mixed_in = 0; float oc_mono_map[PA_CHANNELS_MAX]; for (ic = 0; ic < n_ic; ic++) { if (on_left(r->i_cm.map[ic])) ic_left++; if (on_right(r->i_cm.map[ic])) ic_right++; if (on_center(r->i_cm.map[ic])) ic_center++; } setup_oc_mono_map(r, oc_mono_map); for (oc = 0; oc < n_oc; oc++) { bool oc_connected = false; pa_channel_position_t b = r->o_cm.map[oc]; for (ic = 0; ic < n_ic; ic++) { pa_channel_position_t a = r->i_cm.map[ic]; if (a == b) { m->map_table_f[oc][ic] = 1.0f; oc_connected = true; ic_connected[ic] = true; } else if (a == PA_CHANNEL_POSITION_MONO && oc_mono_map[oc] > 0.0f) { m->map_table_f[oc][ic] = oc_mono_map[oc]; oc_connected = true; ic_connected[ic] = true; } else if (b == PA_CHANNEL_POSITION_MONO) { m->map_table_f[oc][ic] = 1.0f / (float) n_ic; oc_connected = true; ic_connected[ic] = true; } } if (!oc_connected) { /* Maybe it is due to 5.1 rear/side confustion? */ for (ic = 0; ic < n_ic; ic++) { pa_channel_position_t a = r->i_cm.map[ic]; if (ic_connected[ic]) continue; if ((a == PA_CHANNEL_POSITION_REAR_LEFT && b == PA_CHANNEL_POSITION_SIDE_LEFT) || (a == PA_CHANNEL_POSITION_SIDE_LEFT && b == PA_CHANNEL_POSITION_REAR_LEFT) || (a == PA_CHANNEL_POSITION_REAR_RIGHT && b == PA_CHANNEL_POSITION_SIDE_RIGHT) || (a == PA_CHANNEL_POSITION_SIDE_RIGHT && b == PA_CHANNEL_POSITION_REAR_RIGHT)) { m->map_table_f[oc][ic] = 1.0f; oc_connected = true; ic_connected[ic] = true; } } } if (!oc_connected) { /* Try to find matching input ports for this output port */ if (on_left(b) && !(r->flags & PA_RESAMPLER_NO_FILL_SINK)) { /* We are not connected and on the left side, let's * average all left side input channels. */ if (ic_left > 0) for (ic = 0; ic < n_ic; ic++) if (on_left(r->i_cm.map[ic])) { m->map_table_f[oc][ic] = 1.0f / (float) ic_left; ic_connected[ic] = true; } /* We ignore the case where there is no left input channel. * Something is really wrong in this case anyway. */ } else if (on_right(b) && !(r->flags & PA_RESAMPLER_NO_FILL_SINK)) { /* We are not connected and on the right side, let's * average all right side input channels. */ if (ic_right > 0) for (ic = 0; ic < n_ic; ic++) if (on_right(r->i_cm.map[ic])) { m->map_table_f[oc][ic] = 1.0f / (float) ic_right; ic_connected[ic] = true; } /* We ignore the case where there is no right input * channel. Something is really wrong in this case anyway. * */ } else if (on_center(b) && !(r->flags & PA_RESAMPLER_NO_FILL_SINK)) { if (ic_center > 0) { /* We are not connected and at the center. Let's average * all center input channels. */ for (ic = 0; ic < n_ic; ic++) if (on_center(r->i_cm.map[ic])) { m->map_table_f[oc][ic] = 1.0f / (float) ic_center; ic_connected[ic] = true; } } else if (ic_left + ic_right > 0) { /* Hmm, no center channel around, let's synthesize it * by mixing L and R.*/ for (ic = 0; ic < n_ic; ic++) if (on_left(r->i_cm.map[ic]) || on_right(r->i_cm.map[ic])) { m->map_table_f[oc][ic] = 1.0f / (float) (ic_left + ic_right); ic_connected[ic] = true; } } /* We ignore the case where there is not even a left or * right input channel. Something is really wrong in this * case anyway. */ } else if (on_lfe(b) && (r->flags & PA_RESAMPLER_PRODUCE_LFE)) { /* We are not connected and an LFE. Let's average all * channels for LFE. */ for (ic = 0; ic < n_ic; ic++) m->map_table_f[oc][ic] = 1.0f / (float) n_ic; /* Please note that a channel connected to LFE doesn't * really count as connected. */ *lfe_remixed = true; } } } for (ic = 0; ic < n_ic; ic++) { pa_channel_position_t a = r->i_cm.map[ic]; if (ic_connected[ic]) continue; if (on_left(a)) ic_unconnected_left++; else if (on_right(a)) ic_unconnected_right++; else if (on_center(a)) ic_unconnected_center++; else if (on_lfe(a)) ic_unconnected_lfe++; } for (ic = 0; ic < n_ic; ic++) { pa_channel_position_t a = r->i_cm.map[ic]; if (ic_connected[ic]) continue; for (oc = 0; oc < n_oc; oc++) { pa_channel_position_t b = r->o_cm.map[oc]; if (on_left(a) && on_left(b)) m->map_table_f[oc][ic] = (1.f/9.f) / (float) ic_unconnected_left; else if (on_right(a) && on_right(b)) m->map_table_f[oc][ic] = (1.f/9.f) / (float) ic_unconnected_right; else if (on_center(a) && on_center(b)) { m->map_table_f[oc][ic] = (1.f/9.f) / (float) ic_unconnected_center; ic_unconnected_center_mixed_in = true; } else if (on_lfe(a) && (r->flags & PA_RESAMPLER_CONSUME_LFE)) m->map_table_f[oc][ic] = .375f / (float) ic_unconnected_lfe; } } if (ic_unconnected_center > 0 && !ic_unconnected_center_mixed_in) { unsigned ncenter[PA_CHANNELS_MAX]; bool found_frs[PA_CHANNELS_MAX]; memset(ncenter, 0, sizeof(ncenter)); memset(found_frs, 0, sizeof(found_frs)); /* Hmm, as it appears there was no center channel we could mix our center channel in. In this case, mix it into left and right. Using .5 as the factor. */ for (ic = 0; ic < n_ic; ic++) { if (ic_connected[ic]) continue; if (!on_center(r->i_cm.map[ic])) continue; for (oc = 0; oc < n_oc; oc++) { if (!on_left(r->o_cm.map[oc]) && !on_right(r->o_cm.map[oc])) continue; if (front_rear_side(r->i_cm.map[ic]) == front_rear_side(r->o_cm.map[oc])) { found_frs[ic] = true; break; } } for (oc = 0; oc < n_oc; oc++) { if (!on_left(r->o_cm.map[oc]) && !on_right(r->o_cm.map[oc])) continue; if (!found_frs[ic] || front_rear_side(r->i_cm.map[ic]) == front_rear_side(r->o_cm.map[oc])) ncenter[oc]++; } } for (oc = 0; oc < n_oc; oc++) { if (!on_left(r->o_cm.map[oc]) && !on_right(r->o_cm.map[oc])) continue; if (ncenter[oc] <= 0) continue; for (ic = 0; ic < n_ic; ic++) { if (!on_center(r->i_cm.map[ic])) continue; if (!found_frs[ic] || front_rear_side(r->i_cm.map[ic]) == front_rear_side(r->o_cm.map[oc])) m->map_table_f[oc][ic] = .5f / (float) ncenter[oc]; } } } } for (oc = 0; oc < n_oc; oc++) { float sum = 0.0f; for (ic = 0; ic < n_ic; ic++) sum += m->map_table_f[oc][ic]; if (sum > 1.0f) for (ic = 0; ic < n_ic; ic++) m->map_table_f[oc][ic] /= sum; } /* make an 16:16 int version of the matrix */ for (oc = 0; oc < n_oc; oc++) for (ic = 0; ic < n_ic; ic++) m->map_table_i[oc][ic] = (int32_t) (m->map_table_f[oc][ic] * 0x10000); s = pa_strbuf_new(); pa_strbuf_printf(s, " "); for (ic = 0; ic < n_ic; ic++) pa_strbuf_printf(s, " I%02u ", ic); pa_strbuf_puts(s, "\n +"); for (ic = 0; ic < n_ic; ic++) pa_strbuf_printf(s, "------"); pa_strbuf_puts(s, "\n"); for (oc = 0; oc < n_oc; oc++) { pa_strbuf_printf(s, "O%02u |", oc); for (ic = 0; ic < n_ic; ic++) pa_strbuf_printf(s, " %1.3f", m->map_table_f[oc][ic]); pa_strbuf_puts(s, "\n"); } pa_log_debug("Channel matrix:\n%s", t = pa_strbuf_to_string_free(s)); pa_xfree(t); /* initialize the remapping function */ pa_init_remap_func(m); } static void free_remap(pa_remap_t *m) { pa_assert(m); pa_xfree(m->state); } /* check if buf's memblock is large enough to hold 'len' bytes; create a * new memblock if necessary and optionally preserve 'copy' data bytes */ static void fit_buf(pa_resampler *r, pa_memchunk *buf, size_t len, size_t *size, size_t copy) { pa_assert(size); if (!buf->memblock || len > *size) { pa_memblock *new_block = pa_memblock_new(r->mempool, len); if (buf->memblock) { if (copy > 0) { void *src = pa_memblock_acquire(buf->memblock); void *dst = pa_memblock_acquire(new_block); pa_assert(copy <= len); memcpy(dst, src, copy); pa_memblock_release(new_block); pa_memblock_release(buf->memblock); } pa_memblock_unref(buf->memblock); } buf->memblock = new_block; *size = len; } buf->length = len; } static pa_memchunk* convert_to_work_format(pa_resampler *r, pa_memchunk *input) { unsigned in_n_samples, out_n_samples; void *src, *dst; bool have_leftover; size_t leftover_length = 0; pa_assert(r); pa_assert(input); pa_assert(input->memblock); /* Convert the incoming sample into the work sample format and place them * in to_work_format_buf. The leftover data is already converted, so it's * part of the output buffer. */ have_leftover = r->leftover_in_to_work; r->leftover_in_to_work = false; if (!have_leftover && (!r->to_work_format_func || !input->length)) return input; else if (input->length <= 0) return &r->to_work_format_buf; in_n_samples = out_n_samples = (unsigned) ((input->length / r->i_fz) * r->i_ss.channels); if (have_leftover) { leftover_length = r->to_work_format_buf.length; out_n_samples += (unsigned) (leftover_length / r->w_sz); } fit_buf(r, &r->to_work_format_buf, r->w_sz * out_n_samples, &r->to_work_format_buf_size, leftover_length); src = pa_memblock_acquire_chunk(input); dst = (uint8_t *) pa_memblock_acquire(r->to_work_format_buf.memblock) + leftover_length; if (r->to_work_format_func) r->to_work_format_func(in_n_samples, src, dst); else memcpy(dst, src, input->length); pa_memblock_release(input->memblock); pa_memblock_release(r->to_work_format_buf.memblock); return &r->to_work_format_buf; } static pa_memchunk *remap_channels(pa_resampler *r, pa_memchunk *input) { unsigned in_n_samples, out_n_samples, in_n_frames, out_n_frames; void *src, *dst; size_t leftover_length = 0; bool have_leftover; pa_assert(r); pa_assert(input); pa_assert(input->memblock); /* Remap channels and place the result in remap_buf. There may be leftover * data in the beginning of remap_buf. The leftover data is already * remapped, so it's not part of the input, it's part of the output. */ have_leftover = r->leftover_in_remap; r->leftover_in_remap = false; if (!have_leftover && (!r->map_required || input->length <= 0)) return input; else if (input->length <= 0) return &r->remap_buf; in_n_samples = (unsigned) (input->length / r->w_sz); in_n_frames = out_n_frames = in_n_samples / r->i_ss.channels; if (have_leftover) { leftover_length = r->remap_buf.length; out_n_frames += leftover_length / r->w_fz; } out_n_samples = out_n_frames * r->o_ss.channels; fit_buf(r, &r->remap_buf, out_n_samples * r->w_sz, &r->remap_buf_size, leftover_length); src = pa_memblock_acquire_chunk(input); dst = (uint8_t *) pa_memblock_acquire(r->remap_buf.memblock) + leftover_length; if (r->map_required) { pa_remap_t *remap = &r->remap; pa_assert(remap->do_remap); remap->do_remap(remap, dst, src, in_n_frames); } else memcpy(dst, src, input->length); pa_memblock_release(input->memblock); pa_memblock_release(r->remap_buf.memblock); return &r->remap_buf; } static void save_leftover(pa_resampler *r, void *buf, size_t len) { void *dst; pa_assert(r); pa_assert(buf); pa_assert(len > 0); /* Store the leftover data. */ fit_buf(r, r->leftover_buf, len, r->leftover_buf_size, 0); *r->have_leftover = true; dst = pa_memblock_acquire(r->leftover_buf->memblock); memmove(dst, buf, len); pa_memblock_release(r->leftover_buf->memblock); } static pa_memchunk *resample(pa_resampler *r, pa_memchunk *input) { unsigned in_n_frames, out_n_frames, leftover_n_frames; pa_assert(r); pa_assert(input); /* Resample the data and place the result in resample_buf. */ if (!r->impl.resample || !input->length) return input; in_n_frames = (unsigned) (input->length / r->w_fz); out_n_frames = ((in_n_frames*r->o_ss.rate)/r->i_ss.rate)+EXTRA_FRAMES; fit_buf(r, &r->resample_buf, r->w_fz * out_n_frames, &r->resample_buf_size, 0); leftover_n_frames = r->impl.resample(r, input, in_n_frames, &r->resample_buf, &out_n_frames); if (leftover_n_frames > 0) { void *leftover_data = (uint8_t *) pa_memblock_acquire_chunk(input) + (in_n_frames - leftover_n_frames) * r->w_fz; save_leftover(r, leftover_data, leftover_n_frames * r->w_fz); pa_memblock_release(input->memblock); } r->resample_buf.length = out_n_frames * r->w_fz; return &r->resample_buf; } static pa_memchunk *convert_from_work_format(pa_resampler *r, pa_memchunk *input) { unsigned n_samples, n_frames; void *src, *dst; pa_assert(r); pa_assert(input); /* Convert the data into the correct sample type and place the result in * from_work_format_buf. */ if (!r->from_work_format_func || !input->length) return input; n_samples = (unsigned) (input->length / r->w_sz); n_frames = n_samples / r->o_ss.channels; fit_buf(r, &r->from_work_format_buf, r->o_fz * n_frames, &r->from_work_format_buf_size, 0); src = pa_memblock_acquire_chunk(input); dst = pa_memblock_acquire(r->from_work_format_buf.memblock); r->from_work_format_func(n_samples, src, dst); pa_memblock_release(input->memblock); pa_memblock_release(r->from_work_format_buf.memblock); return &r->from_work_format_buf; } void pa_resampler_run(pa_resampler *r, const pa_memchunk *in, pa_memchunk *out) { pa_memchunk *buf; pa_assert(r); pa_assert(in); pa_assert(out); pa_assert(in->length); pa_assert(in->memblock); pa_assert(in->length % r->i_fz == 0); buf = (pa_memchunk*) in; r->in_frames += buf->length / r->i_fz; buf = convert_to_work_format(r, buf); /* Try to save resampling effort: if we have more output channels than * input channels, do resampling first, then remapping. */ if (r->o_ss.channels <= r->i_ss.channels) { buf = remap_channels(r, buf); buf = resample(r, buf); } else { buf = resample(r, buf); buf = remap_channels(r, buf); } if (r->lfe_filter) buf = pa_lfe_filter_process(r->lfe_filter, buf); if (buf->length) { buf = convert_from_work_format(r, buf); *out = *buf; r->out_frames += buf->length / r->o_fz; if (buf == in) pa_memblock_ref(buf->memblock); else pa_memchunk_reset(buf); } else pa_memchunk_reset(out); } /* Get delay in input frames. Some resamplers may have negative delay. */ double pa_resampler_get_delay(pa_resampler *r, bool allow_negative) { double frames; frames = r->out_frames * r->i_ss.rate / r->o_ss.rate; if (frames >= r->in_frames && !allow_negative) return 0; return r->in_frames - frames; } /* Get delay in usec */ pa_usec_t pa_resampler_get_delay_usec(pa_resampler *r) { if (!r) return 0; return (pa_usec_t) (pa_resampler_get_delay(r, false) * PA_USEC_PER_SEC / r->i_ss.rate); } /* Get GCD of input and output rate. */ unsigned pa_resampler_get_gcd(pa_resampler *r) { pa_assert(r); return r->gcd; } /* Get maximum resampler history. The resamplers have finite impulse response, so really old * data (more than 2x the resampler latency) cannot affect the output. This means, that in an * ideal case, we should re-run 2 - 3 times the resampler delay through the resampler when it * is rewound. On the other hand this would mean for high sample rates that more than 25000 * samples would need to be used (384k * 33ms). Therefore limit the history to 1.5 times the * maximum resampler delay, which should be fully sufficient in most cases and allows to run * at least more than one delay through the resampler in case of high rates. */ size_t pa_resampler_get_max_history(pa_resampler *r) { if (!r) return 0; return (uint64_t) PA_RESAMPLER_MAX_DELAY_USEC * r->i_ss.rate * 3 / PA_USEC_PER_SEC / 2; } /*** copy (noop) implementation ***/ static int copy_init(pa_resampler *r) { pa_assert(r); pa_assert(r->o_ss.rate == r->i_ss.rate); return 0; }