summaryrefslogtreecommitdiffstats
path: root/media/ffvpx/libavcodec/resample.c
diff options
context:
space:
mode:
authorMoonchild <mcwerewolf@gmail.com>2018-10-01 15:25:04 +0200
committerGitHub <noreply@github.com>2018-10-01 15:25:04 +0200
commit45c24f05d023a2cd8289ed40a13708392ce2e6a4 (patch)
treefef75d382fc6216a093eeaf80560473dff19d883 /media/ffvpx/libavcodec/resample.c
parent79b00fc33b5cb6d56d29b50efac6d62ce3a89018 (diff)
downloadUXP-45c24f05d023a2cd8289ed40a13708392ce2e6a4.tar
UXP-45c24f05d023a2cd8289ed40a13708392ce2e6a4.tar.gz
UXP-45c24f05d023a2cd8289ed40a13708392ce2e6a4.tar.lz
UXP-45c24f05d023a2cd8289ed40a13708392ce2e6a4.tar.xz
UXP-45c24f05d023a2cd8289ed40a13708392ce2e6a4.zip
Revert "Update ffvpx code to 4.0.2"
Diffstat (limited to 'media/ffvpx/libavcodec/resample.c')
-rw-r--r--media/ffvpx/libavcodec/resample.c439
1 files changed, 439 insertions, 0 deletions
diff --git a/media/ffvpx/libavcodec/resample.c b/media/ffvpx/libavcodec/resample.c
new file mode 100644
index 000000000..4c5eb9f10
--- /dev/null
+++ b/media/ffvpx/libavcodec/resample.c
@@ -0,0 +1,439 @@
+/*
+ * samplerate conversion for both audio and video
+ * Copyright (c) 2000 Fabrice Bellard
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * samplerate conversion for both audio and video
+ */
+
+#include <string.h>
+
+#include "avcodec.h"
+#include "audioconvert.h"
+#include "libavutil/opt.h"
+#include "libavutil/mem.h"
+#include "libavutil/samplefmt.h"
+
+#if FF_API_AVCODEC_RESAMPLE
+FF_DISABLE_DEPRECATION_WARNINGS
+
+#define MAX_CHANNELS 8
+
+struct AVResampleContext;
+
+static const char *context_to_name(void *ptr)
+{
+ return "audioresample";
+}
+
+static const AVOption options[] = {{NULL}};
+static const AVClass audioresample_context_class = {
+ "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT
+};
+
+struct ReSampleContext {
+ struct AVResampleContext *resample_context;
+ short *temp[MAX_CHANNELS];
+ int temp_len;
+ float ratio;
+ /* channel convert */
+ int input_channels, output_channels, filter_channels;
+ AVAudioConvert *convert_ctx[2];
+ enum AVSampleFormat sample_fmt[2]; ///< input and output sample format
+ unsigned sample_size[2]; ///< size of one sample in sample_fmt
+ short *buffer[2]; ///< buffers used for conversion to S16
+ unsigned buffer_size[2]; ///< sizes of allocated buffers
+};
+
+/* n1: number of samples */
+static void stereo_to_mono(short *output, short *input, int n1)
+{
+ short *p, *q;
+ int n = n1;
+
+ p = input;
+ q = output;
+ while (n >= 4) {
+ q[0] = (p[0] + p[1]) >> 1;
+ q[1] = (p[2] + p[3]) >> 1;
+ q[2] = (p[4] + p[5]) >> 1;
+ q[3] = (p[6] + p[7]) >> 1;
+ q += 4;
+ p += 8;
+ n -= 4;
+ }
+ while (n > 0) {
+ q[0] = (p[0] + p[1]) >> 1;
+ q++;
+ p += 2;
+ n--;
+ }
+}
+
+/* n1: number of samples */
+static void mono_to_stereo(short *output, short *input, int n1)
+{
+ short *p, *q;
+ int n = n1;
+ int v;
+
+ p = input;
+ q = output;
+ while (n >= 4) {
+ v = p[0]; q[0] = v; q[1] = v;
+ v = p[1]; q[2] = v; q[3] = v;
+ v = p[2]; q[4] = v; q[5] = v;
+ v = p[3]; q[6] = v; q[7] = v;
+ q += 8;
+ p += 4;
+ n -= 4;
+ }
+ while (n > 0) {
+ v = p[0]; q[0] = v; q[1] = v;
+ q += 2;
+ p += 1;
+ n--;
+ }
+}
+
+/*
+5.1 to stereo input: [fl, fr, c, lfe, rl, rr]
+- Left = front_left + rear_gain * rear_left + center_gain * center
+- Right = front_right + rear_gain * rear_right + center_gain * center
+Where rear_gain is usually around 0.5-1.0 and
+ center_gain is almost always 0.7 (-3 dB)
+*/
+static void surround_to_stereo(short **output, short *input, int channels, int samples)
+{
+ int i;
+ short l, r;
+
+ for (i = 0; i < samples; i++) {
+ int fl,fr,c,rl,rr;
+ fl = input[0];
+ fr = input[1];
+ c = input[2];
+ // lfe = input[3];
+ rl = input[4];
+ rr = input[5];
+
+ l = av_clip_int16(fl + (0.5 * rl) + (0.7 * c));
+ r = av_clip_int16(fr + (0.5 * rr) + (0.7 * c));
+
+ /* output l & r. */
+ *output[0]++ = l;
+ *output[1]++ = r;
+
+ /* increment input. */
+ input += channels;
+ }
+}
+
+static void deinterleave(short **output, short *input, int channels, int samples)
+{
+ int i, j;
+
+ for (i = 0; i < samples; i++) {
+ for (j = 0; j < channels; j++) {
+ *output[j]++ = *input++;
+ }
+ }
+}
+
+static void interleave(short *output, short **input, int channels, int samples)
+{
+ int i, j;
+
+ for (i = 0; i < samples; i++) {
+ for (j = 0; j < channels; j++) {
+ *output++ = *input[j]++;
+ }
+ }
+}
+
+static void ac3_5p1_mux(short *output, short *input1, short *input2, int n)
+{
+ int i;
+ short l, r;
+
+ for (i = 0; i < n; i++) {
+ l = *input1++;
+ r = *input2++;
+ *output++ = l; /* left */
+ *output++ = (l / 2) + (r / 2); /* center */
+ *output++ = r; /* right */
+ *output++ = 0; /* left surround */
+ *output++ = 0; /* right surroud */
+ *output++ = 0; /* low freq */
+ }
+}
+
+#define SUPPORT_RESAMPLE(ch1, ch2, ch3, ch4, ch5, ch6, ch7, ch8) \
+ ch8<<7 | ch7<<6 | ch6<<5 | ch5<<4 | ch4<<3 | ch3<<2 | ch2<<1 | ch1<<0
+
+static const uint8_t supported_resampling[MAX_CHANNELS] = {
+ // output ch: 1 2 3 4 5 6 7 8
+ SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 0, 0, 0), // 1 input channel
+ SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 1, 0, 0), // 2 input channels
+ SUPPORT_RESAMPLE(0, 0, 1, 0, 0, 0, 0, 0), // 3 input channels
+ SUPPORT_RESAMPLE(0, 0, 0, 1, 0, 0, 0, 0), // 4 input channels
+ SUPPORT_RESAMPLE(0, 0, 0, 0, 1, 0, 0, 0), // 5 input channels
+ SUPPORT_RESAMPLE(0, 1, 0, 0, 0, 1, 0, 0), // 6 input channels
+ SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 1, 0), // 7 input channels
+ SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 0, 1), // 8 input channels
+};
+
+ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
+ int output_rate, int input_rate,
+ enum AVSampleFormat sample_fmt_out,
+ enum AVSampleFormat sample_fmt_in,
+ int filter_length, int log2_phase_count,
+ int linear, double cutoff)
+{
+ ReSampleContext *s;
+
+ if (input_channels > MAX_CHANNELS) {
+ av_log(NULL, AV_LOG_ERROR,
+ "Resampling with input channels greater than %d is unsupported.\n",
+ MAX_CHANNELS);
+ return NULL;
+ }
+ if (!(supported_resampling[input_channels-1] & (1<<(output_channels-1)))) {
+ int i;
+ av_log(NULL, AV_LOG_ERROR, "Unsupported audio resampling. Allowed "
+ "output channels for %d input channel%s", input_channels,
+ input_channels > 1 ? "s:" : ":");
+ for (i = 0; i < MAX_CHANNELS; i++)
+ if (supported_resampling[input_channels-1] & (1<<i))
+ av_log(NULL, AV_LOG_ERROR, " %d", i + 1);
+ av_log(NULL, AV_LOG_ERROR, "\n");
+ return NULL;
+ }
+
+ s = av_mallocz(sizeof(ReSampleContext));
+ if (!s) {
+ av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n");
+ return NULL;
+ }
+
+ s->ratio = (float)output_rate / (float)input_rate;
+
+ s->input_channels = input_channels;
+ s->output_channels = output_channels;
+
+ s->filter_channels = s->input_channels;
+ if (s->output_channels < s->filter_channels)
+ s->filter_channels = s->output_channels;
+
+ s->sample_fmt[0] = sample_fmt_in;
+ s->sample_fmt[1] = sample_fmt_out;
+ s->sample_size[0] = av_get_bytes_per_sample(s->sample_fmt[0]);
+ s->sample_size[1] = av_get_bytes_per_sample(s->sample_fmt[1]);
+
+ if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
+ if (!(s->convert_ctx[0] = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1,
+ s->sample_fmt[0], 1, NULL, 0))) {
+ av_log(s, AV_LOG_ERROR,
+ "Cannot convert %s sample format to s16 sample format\n",
+ av_get_sample_fmt_name(s->sample_fmt[0]));
+ av_free(s);
+ return NULL;
+ }
+ }
+
+ if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
+ if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1,
+ AV_SAMPLE_FMT_S16, 1, NULL, 0))) {
+ av_log(s, AV_LOG_ERROR,
+ "Cannot convert s16 sample format to %s sample format\n",
+ av_get_sample_fmt_name(s->sample_fmt[1]));
+ av_audio_convert_free(s->convert_ctx[0]);
+ av_free(s);
+ return NULL;
+ }
+ }
+
+ s->resample_context = av_resample_init(output_rate, input_rate,
+ filter_length, log2_phase_count,
+ linear, cutoff);
+
+ *(const AVClass**)s->resample_context = &audioresample_context_class;
+
+ return s;
+}
+
+/* resample audio. 'nb_samples' is the number of input samples */
+/* XXX: optimize it ! */
+int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
+{
+ int i, nb_samples1;
+ short *bufin[MAX_CHANNELS];
+ short *bufout[MAX_CHANNELS];
+ short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS];
+ short *output_bak = NULL;
+ int lenout;
+
+ if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
+ int istride[1] = { s->sample_size[0] };
+ int ostride[1] = { 2 };
+ const void *ibuf[1] = { input };
+ void *obuf[1];
+ unsigned input_size = nb_samples * s->input_channels * 2;
+
+ if (!s->buffer_size[0] || s->buffer_size[0] < input_size) {
+ av_free(s->buffer[0]);
+ s->buffer_size[0] = input_size;
+ s->buffer[0] = av_malloc(s->buffer_size[0]);
+ if (!s->buffer[0]) {
+ av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
+ return 0;
+ }
+ }
+
+ obuf[0] = s->buffer[0];
+
+ if (av_audio_convert(s->convert_ctx[0], obuf, ostride,
+ ibuf, istride, nb_samples * s->input_channels) < 0) {
+ av_log(s->resample_context, AV_LOG_ERROR,
+ "Audio sample format conversion failed\n");
+ return 0;
+ }
+
+ input = s->buffer[0];
+ }
+
+ lenout= 2*s->output_channels*nb_samples * s->ratio + 16;
+
+ if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
+ int out_size = lenout * av_get_bytes_per_sample(s->sample_fmt[1]) *
+ s->output_channels;
+ output_bak = output;
+
+ if (!s->buffer_size[1] || s->buffer_size[1] < out_size) {
+ av_free(s->buffer[1]);
+ s->buffer_size[1] = out_size;
+ s->buffer[1] = av_malloc(s->buffer_size[1]);
+ if (!s->buffer[1]) {
+ av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
+ return 0;
+ }
+ }
+
+ output = s->buffer[1];
+ }
+
+ /* XXX: move those malloc to resample init code */
+ for (i = 0; i < s->filter_channels; i++) {
+ bufin[i] = av_malloc_array((nb_samples + s->temp_len), sizeof(short));
+ bufout[i] = av_malloc_array(lenout, sizeof(short));
+
+ if (!bufin[i] || !bufout[i]) {
+ av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
+ nb_samples1 = 0;
+ goto fail;
+ }
+
+ memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short));
+ buftmp2[i] = bufin[i] + s->temp_len;
+ }
+
+ if (s->input_channels == 2 && s->output_channels == 1) {
+ buftmp3[0] = output;
+ stereo_to_mono(buftmp2[0], input, nb_samples);
+ } else if (s->output_channels >= 2 && s->input_channels == 1) {
+ buftmp3[0] = bufout[0];
+ memcpy(buftmp2[0], input, nb_samples * sizeof(short));
+ } else if (s->input_channels == 6 && s->output_channels ==2) {
+ buftmp3[0] = bufout[0];
+ buftmp3[1] = bufout[1];
+ surround_to_stereo(buftmp2, input, s->input_channels, nb_samples);
+ } else if (s->output_channels >= s->input_channels && s->input_channels >= 2) {
+ for (i = 0; i < s->input_channels; i++) {
+ buftmp3[i] = bufout[i];
+ }
+ deinterleave(buftmp2, input, s->input_channels, nb_samples);
+ } else {
+ buftmp3[0] = output;
+ memcpy(buftmp2[0], input, nb_samples * sizeof(short));
+ }
+
+ nb_samples += s->temp_len;
+
+ /* resample each channel */
+ nb_samples1 = 0; /* avoid warning */
+ for (i = 0; i < s->filter_channels; i++) {
+ int consumed;
+ int is_last = i + 1 == s->filter_channels;
+
+ nb_samples1 = av_resample(s->resample_context, buftmp3[i], bufin[i],
+ &consumed, nb_samples, lenout, is_last);
+ s->temp_len = nb_samples - consumed;
+ s->temp[i] = av_realloc_array(s->temp[i], s->temp_len, sizeof(short));
+ memcpy(s->temp[i], bufin[i] + consumed, s->temp_len * sizeof(short));
+ }
+
+ if (s->output_channels == 2 && s->input_channels == 1) {
+ mono_to_stereo(output, buftmp3[0], nb_samples1);
+ } else if (s->output_channels == 6 && s->input_channels == 2) {
+ ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
+ } else if ((s->output_channels == s->input_channels && s->input_channels >= 2) ||
+ (s->output_channels == 2 && s->input_channels == 6)) {
+ interleave(output, buftmp3, s->output_channels, nb_samples1);
+ }
+
+ if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
+ int istride[1] = { 2 };
+ int ostride[1] = { s->sample_size[1] };
+ const void *ibuf[1] = { output };
+ void *obuf[1] = { output_bak };
+
+ if (av_audio_convert(s->convert_ctx[1], obuf, ostride,
+ ibuf, istride, nb_samples1 * s->output_channels) < 0) {
+ av_log(s->resample_context, AV_LOG_ERROR,
+ "Audio sample format conversion failed\n");
+ return 0;
+ }
+ }
+
+fail:
+ for (i = 0; i < s->filter_channels; i++) {
+ av_free(bufin[i]);
+ av_free(bufout[i]);
+ }
+
+ return nb_samples1;
+}
+
+void audio_resample_close(ReSampleContext *s)
+{
+ int i;
+ av_resample_close(s->resample_context);
+ for (i = 0; i < s->filter_channels; i++)
+ av_freep(&s->temp[i]);
+ av_freep(&s->buffer[0]);
+ av_freep(&s->buffer[1]);
+ av_audio_convert_free(s->convert_ctx[0]);
+ av_audio_convert_free(s->convert_ctx[1]);
+ av_free(s);
+}
+
+FF_ENABLE_DEPRECATION_WARNINGS
+#endif