a4b787f4b8
git-svn-id: svn://kolibrios.org@6147 a494cfbc-eb01-0410-851d-a64ba20cac60
381 lines
11 KiB
C
381 lines
11 KiB
C
/*
|
|
* Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com>
|
|
*
|
|
* 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
|
|
*/
|
|
|
|
#include <stdint.h>
|
|
#include <string.h>
|
|
|
|
#include "libavutil/mem.h"
|
|
#include "audio_data.h"
|
|
|
|
static const AVClass audio_data_class = {
|
|
.class_name = "AudioData",
|
|
.item_name = av_default_item_name,
|
|
.version = LIBAVUTIL_VERSION_INT,
|
|
};
|
|
|
|
/*
|
|
* Calculate alignment for data pointers.
|
|
*/
|
|
static void calc_ptr_alignment(AudioData *a)
|
|
{
|
|
int p;
|
|
int min_align = 128;
|
|
|
|
for (p = 0; p < a->planes; p++) {
|
|
int cur_align = 128;
|
|
while ((intptr_t)a->data[p] % cur_align)
|
|
cur_align >>= 1;
|
|
if (cur_align < min_align)
|
|
min_align = cur_align;
|
|
}
|
|
a->ptr_align = min_align;
|
|
}
|
|
|
|
int ff_sample_fmt_is_planar(enum AVSampleFormat sample_fmt, int channels)
|
|
{
|
|
if (channels == 1)
|
|
return 1;
|
|
else
|
|
return av_sample_fmt_is_planar(sample_fmt);
|
|
}
|
|
|
|
int ff_audio_data_set_channels(AudioData *a, int channels)
|
|
{
|
|
if (channels < 1 || channels > AVRESAMPLE_MAX_CHANNELS ||
|
|
channels > a->allocated_channels)
|
|
return AVERROR(EINVAL);
|
|
|
|
a->channels = channels;
|
|
a->planes = a->is_planar ? channels : 1;
|
|
|
|
calc_ptr_alignment(a);
|
|
|
|
return 0;
|
|
}
|
|
|
|
int ff_audio_data_init(AudioData *a, uint8_t **src, int plane_size, int channels,
|
|
int nb_samples, enum AVSampleFormat sample_fmt,
|
|
int read_only, const char *name)
|
|
{
|
|
int p;
|
|
|
|
memset(a, 0, sizeof(*a));
|
|
a->class = &audio_data_class;
|
|
|
|
if (channels < 1 || channels > AVRESAMPLE_MAX_CHANNELS) {
|
|
av_log(a, AV_LOG_ERROR, "invalid channel count: %d\n", channels);
|
|
return AVERROR(EINVAL);
|
|
}
|
|
|
|
a->sample_size = av_get_bytes_per_sample(sample_fmt);
|
|
if (!a->sample_size) {
|
|
av_log(a, AV_LOG_ERROR, "invalid sample format\n");
|
|
return AVERROR(EINVAL);
|
|
}
|
|
a->is_planar = ff_sample_fmt_is_planar(sample_fmt, channels);
|
|
a->planes = a->is_planar ? channels : 1;
|
|
a->stride = a->sample_size * (a->is_planar ? 1 : channels);
|
|
|
|
for (p = 0; p < (a->is_planar ? channels : 1); p++) {
|
|
if (!src[p]) {
|
|
av_log(a, AV_LOG_ERROR, "invalid NULL pointer for src[%d]\n", p);
|
|
return AVERROR(EINVAL);
|
|
}
|
|
a->data[p] = src[p];
|
|
}
|
|
a->allocated_samples = nb_samples * !read_only;
|
|
a->nb_samples = nb_samples;
|
|
a->sample_fmt = sample_fmt;
|
|
a->channels = channels;
|
|
a->allocated_channels = channels;
|
|
a->read_only = read_only;
|
|
a->allow_realloc = 0;
|
|
a->name = name ? name : "{no name}";
|
|
|
|
calc_ptr_alignment(a);
|
|
a->samples_align = plane_size / a->stride;
|
|
|
|
return 0;
|
|
}
|
|
|
|
AudioData *ff_audio_data_alloc(int channels, int nb_samples,
|
|
enum AVSampleFormat sample_fmt, const char *name)
|
|
{
|
|
AudioData *a;
|
|
int ret;
|
|
|
|
if (channels < 1 || channels > AVRESAMPLE_MAX_CHANNELS)
|
|
return NULL;
|
|
|
|
a = av_mallocz(sizeof(*a));
|
|
if (!a)
|
|
return NULL;
|
|
|
|
a->sample_size = av_get_bytes_per_sample(sample_fmt);
|
|
if (!a->sample_size) {
|
|
av_free(a);
|
|
return NULL;
|
|
}
|
|
a->is_planar = ff_sample_fmt_is_planar(sample_fmt, channels);
|
|
a->planes = a->is_planar ? channels : 1;
|
|
a->stride = a->sample_size * (a->is_planar ? 1 : channels);
|
|
|
|
a->class = &audio_data_class;
|
|
a->sample_fmt = sample_fmt;
|
|
a->channels = channels;
|
|
a->allocated_channels = channels;
|
|
a->read_only = 0;
|
|
a->allow_realloc = 1;
|
|
a->name = name ? name : "{no name}";
|
|
|
|
if (nb_samples > 0) {
|
|
ret = ff_audio_data_realloc(a, nb_samples);
|
|
if (ret < 0) {
|
|
av_free(a);
|
|
return NULL;
|
|
}
|
|
return a;
|
|
} else {
|
|
calc_ptr_alignment(a);
|
|
return a;
|
|
}
|
|
}
|
|
|
|
int ff_audio_data_realloc(AudioData *a, int nb_samples)
|
|
{
|
|
int ret, new_buf_size, plane_size, p;
|
|
|
|
/* check if buffer is already large enough */
|
|
if (a->allocated_samples >= nb_samples)
|
|
return 0;
|
|
|
|
/* validate that the output is not read-only and realloc is allowed */
|
|
if (a->read_only || !a->allow_realloc)
|
|
return AVERROR(EINVAL);
|
|
|
|
new_buf_size = av_samples_get_buffer_size(&plane_size,
|
|
a->allocated_channels, nb_samples,
|
|
a->sample_fmt, 0);
|
|
if (new_buf_size < 0)
|
|
return new_buf_size;
|
|
|
|
/* if there is already data in the buffer and the sample format is planar,
|
|
allocate a new buffer and copy the data, otherwise just realloc the
|
|
internal buffer and set new data pointers */
|
|
if (a->nb_samples > 0 && a->is_planar) {
|
|
uint8_t *new_data[AVRESAMPLE_MAX_CHANNELS] = { NULL };
|
|
|
|
ret = av_samples_alloc(new_data, &plane_size, a->allocated_channels,
|
|
nb_samples, a->sample_fmt, 0);
|
|
if (ret < 0)
|
|
return ret;
|
|
|
|
for (p = 0; p < a->planes; p++)
|
|
memcpy(new_data[p], a->data[p], a->nb_samples * a->stride);
|
|
|
|
av_freep(&a->buffer);
|
|
memcpy(a->data, new_data, sizeof(new_data));
|
|
a->buffer = a->data[0];
|
|
} else {
|
|
av_freep(&a->buffer);
|
|
a->buffer = av_malloc(new_buf_size);
|
|
if (!a->buffer)
|
|
return AVERROR(ENOMEM);
|
|
ret = av_samples_fill_arrays(a->data, &plane_size, a->buffer,
|
|
a->allocated_channels, nb_samples,
|
|
a->sample_fmt, 0);
|
|
if (ret < 0)
|
|
return ret;
|
|
}
|
|
a->buffer_size = new_buf_size;
|
|
a->allocated_samples = nb_samples;
|
|
|
|
calc_ptr_alignment(a);
|
|
a->samples_align = plane_size / a->stride;
|
|
|
|
return 0;
|
|
}
|
|
|
|
void ff_audio_data_free(AudioData **a)
|
|
{
|
|
if (!*a)
|
|
return;
|
|
av_free((*a)->buffer);
|
|
av_freep(a);
|
|
}
|
|
|
|
int ff_audio_data_copy(AudioData *dst, AudioData *src, ChannelMapInfo *map)
|
|
{
|
|
int ret, p;
|
|
|
|
/* validate input/output compatibility */
|
|
if (dst->sample_fmt != src->sample_fmt || dst->channels < src->channels)
|
|
return AVERROR(EINVAL);
|
|
|
|
if (map && !src->is_planar) {
|
|
av_log(src, AV_LOG_ERROR, "cannot remap packed format during copy\n");
|
|
return AVERROR(EINVAL);
|
|
}
|
|
|
|
/* if the input is empty, just empty the output */
|
|
if (!src->nb_samples) {
|
|
dst->nb_samples = 0;
|
|
return 0;
|
|
}
|
|
|
|
/* reallocate output if necessary */
|
|
ret = ff_audio_data_realloc(dst, src->nb_samples);
|
|
if (ret < 0)
|
|
return ret;
|
|
|
|
/* copy data */
|
|
if (map) {
|
|
if (map->do_remap) {
|
|
for (p = 0; p < src->planes; p++) {
|
|
if (map->channel_map[p] >= 0)
|
|
memcpy(dst->data[p], src->data[map->channel_map[p]],
|
|
src->nb_samples * src->stride);
|
|
}
|
|
}
|
|
if (map->do_copy || map->do_zero) {
|
|
for (p = 0; p < src->planes; p++) {
|
|
if (map->channel_copy[p])
|
|
memcpy(dst->data[p], dst->data[map->channel_copy[p]],
|
|
src->nb_samples * src->stride);
|
|
else if (map->channel_zero[p])
|
|
av_samples_set_silence(&dst->data[p], 0, src->nb_samples,
|
|
1, dst->sample_fmt);
|
|
}
|
|
}
|
|
} else {
|
|
for (p = 0; p < src->planes; p++)
|
|
memcpy(dst->data[p], src->data[p], src->nb_samples * src->stride);
|
|
}
|
|
|
|
dst->nb_samples = src->nb_samples;
|
|
|
|
return 0;
|
|
}
|
|
|
|
int ff_audio_data_combine(AudioData *dst, int dst_offset, AudioData *src,
|
|
int src_offset, int nb_samples)
|
|
{
|
|
int ret, p, dst_offset2, dst_move_size;
|
|
|
|
/* validate input/output compatibility */
|
|
if (dst->sample_fmt != src->sample_fmt || dst->channels != src->channels) {
|
|
av_log(src, AV_LOG_ERROR, "sample format mismatch\n");
|
|
return AVERROR(EINVAL);
|
|
}
|
|
|
|
/* validate offsets are within the buffer bounds */
|
|
if (dst_offset < 0 || dst_offset > dst->nb_samples ||
|
|
src_offset < 0 || src_offset > src->nb_samples) {
|
|
av_log(src, AV_LOG_ERROR, "offset out-of-bounds: src=%d dst=%d\n",
|
|
src_offset, dst_offset);
|
|
return AVERROR(EINVAL);
|
|
}
|
|
|
|
/* check offsets and sizes to see if we can just do nothing and return */
|
|
if (nb_samples > src->nb_samples - src_offset)
|
|
nb_samples = src->nb_samples - src_offset;
|
|
if (nb_samples <= 0)
|
|
return 0;
|
|
|
|
/* validate that the output is not read-only */
|
|
if (dst->read_only) {
|
|
av_log(dst, AV_LOG_ERROR, "dst is read-only\n");
|
|
return AVERROR(EINVAL);
|
|
}
|
|
|
|
/* reallocate output if necessary */
|
|
ret = ff_audio_data_realloc(dst, dst->nb_samples + nb_samples);
|
|
if (ret < 0) {
|
|
av_log(dst, AV_LOG_ERROR, "error reallocating dst\n");
|
|
return ret;
|
|
}
|
|
|
|
dst_offset2 = dst_offset + nb_samples;
|
|
dst_move_size = dst->nb_samples - dst_offset;
|
|
|
|
for (p = 0; p < src->planes; p++) {
|
|
if (dst_move_size > 0) {
|
|
memmove(dst->data[p] + dst_offset2 * dst->stride,
|
|
dst->data[p] + dst_offset * dst->stride,
|
|
dst_move_size * dst->stride);
|
|
}
|
|
memcpy(dst->data[p] + dst_offset * dst->stride,
|
|
src->data[p] + src_offset * src->stride,
|
|
nb_samples * src->stride);
|
|
}
|
|
dst->nb_samples += nb_samples;
|
|
|
|
return 0;
|
|
}
|
|
|
|
void ff_audio_data_drain(AudioData *a, int nb_samples)
|
|
{
|
|
if (a->nb_samples <= nb_samples) {
|
|
/* drain the whole buffer */
|
|
a->nb_samples = 0;
|
|
} else {
|
|
int p;
|
|
int move_offset = a->stride * nb_samples;
|
|
int move_size = a->stride * (a->nb_samples - nb_samples);
|
|
|
|
for (p = 0; p < a->planes; p++)
|
|
memmove(a->data[p], a->data[p] + move_offset, move_size);
|
|
|
|
a->nb_samples -= nb_samples;
|
|
}
|
|
}
|
|
|
|
int ff_audio_data_add_to_fifo(AVAudioFifo *af, AudioData *a, int offset,
|
|
int nb_samples)
|
|
{
|
|
uint8_t *offset_data[AVRESAMPLE_MAX_CHANNELS];
|
|
int offset_size, p;
|
|
|
|
if (offset >= a->nb_samples)
|
|
return 0;
|
|
offset_size = offset * a->stride;
|
|
for (p = 0; p < a->planes; p++)
|
|
offset_data[p] = a->data[p] + offset_size;
|
|
|
|
return av_audio_fifo_write(af, (void **)offset_data, nb_samples);
|
|
}
|
|
|
|
int ff_audio_data_read_from_fifo(AVAudioFifo *af, AudioData *a, int nb_samples)
|
|
{
|
|
int ret;
|
|
|
|
if (a->read_only)
|
|
return AVERROR(EINVAL);
|
|
|
|
ret = ff_audio_data_realloc(a, nb_samples);
|
|
if (ret < 0)
|
|
return ret;
|
|
|
|
ret = av_audio_fifo_read(af, (void **)a->data, nb_samples);
|
|
if (ret >= 0)
|
|
a->nb_samples = ret;
|
|
return ret;
|
|
}
|