36 #define FREEZE_INTERVAL 128 40 #define MAX_FRAME_SIZE 32768 45 #define MAX_TRELLIS 16 51 for (i = 0; i < 2; i++) {
80 "allowed. Using %d instead of %d\n", new_frame_size,
96 "allowed. Using %d instead of %d\n", new_trellis,
101 int frontier = 1 << avctx->
trellis;
104 for (
int i = 0;
i < 2;
i++) {
120 35, 72, 110, 150, 190, 233, 276, 323,
121 370, 422, 473, 530, 587, 650, 714, 786,
122 858, 940, 1023, 1121, 1219, 1339, 1458, 1612,
123 1765, 1980, 2195, 2557, 2919
127 int *xlow,
int *xhigh)
133 *xlow = xout[0] + xout[1] >> 14;
134 *xhigh = xout[0] - xout[1] >> 14;
146 int pred = 141 * state->scale_factor >> 8;
148 return ((diff ^ (diff >> (
sizeof(diff)*8-1))) < pred) + 2*(diff >= 0);
155 int limit = diff ^ (diff >> (
sizeof(
diff)*8-1));
157 limit = limit + 1 << 10;
158 if (limit >
low_quant[8] * state->scale_factor)
160 while (i < 29 && limit >
low_quant[i] * state->scale_factor)
162 return (diff < 0 ? (i < 2 ? 63 : 33) : 61) - i;
170 int frontier = 1 << trellis;
173 int pathn[2] = {0, 0}, froze = -1;
176 for (i = 0; i < 2; i++) {
181 nodes[
i][0]->
ssd = 0;
182 nodes[
i][0]->
path = 0;
183 nodes[
i][0]->state = c->
band[
i];
186 for (i = 0; i < nb_samples >> 1; i++) {
189 int heap_pos[2] = {0, 0};
191 for (j = 0; j < 2; j++) {
192 next[j] = c->
node_buf[j] + frontier*(i & 1);
193 memset(nodes_next[j], 0, frontier *
sizeof(**nodes_next));
198 for (j = 0; j < frontier && nodes[0][j]; j++) {
204 int range = j < frontier/2 ? 4 : 0;
207 int ilow =
encode_low(&cur_node->state, xlow);
209 for (k = ilow - range; k <= ilow + range && k <= 63; k += 4) {
210 int decoded, dec_diff,
pos;
219 + cur_node->state.s_predictor, 14);
220 dec_diff = xlow - decoded;
222 #define STORE_NODE(index, UPDATE, VALUE)\ 223 ssd = cur_node->ssd + dec_diff*dec_diff;\ 226 if (ssd < cur_node->ssd)\ 228 if (heap_pos[index] < frontier) {\ 229 pos = heap_pos[index]++;\ 230 av_assert2(pathn[index] < FREEZE_INTERVAL * frontier);\ 231 node = nodes_next[index][pos] = next[index]++;\ 232 node->path = pathn[index]++;\ 236 pos = (frontier>>1) + (heap_pos[index] & ((frontier>>1) - 1));\ 237 if (ssd >= nodes_next[index][pos]->ssd)\ 240 node = nodes_next[index][pos];\ 243 node->state = cur_node->state;\ 245 c->paths[index][node->path].value = VALUE;\ 246 c->paths[index][node->path].prev = cur_node->path;\ 250 int parent = (pos - 1) >> 1;\ 251 if (nodes_next[index][parent]->ssd <= ssd)\ 253 FFSWAP(struct TrellisNode*, nodes_next[index][parent],\ 254 nodes_next[index][pos]);\ 261 for (j = 0; j < frontier && nodes[1][j]; j++) {
269 for (ihigh = 0; ihigh < 4; ihigh++) {
270 int dhigh, decoded, dec_diff,
pos;
274 dhigh = cur_node->state.scale_factor *
276 decoded =
av_clip_intp2(dhigh + cur_node->state.s_predictor, 14);
277 dec_diff = xhigh - decoded;
283 for (j = 0; j < 2; j++) {
286 if (nodes[j][0]->
ssd > (1 << 16)) {
287 for (k = 1; k < frontier && nodes[j][k]; k++)
288 nodes[j][k]->
ssd -= nodes[j][0]->
ssd;
289 nodes[j][0]->ssd = 0;
294 p[0] = &c->
paths[0][nodes[0][0]->path];
295 p[1] = &c->
paths[1][nodes[1][0]->path];
296 for (j = i; j > froze; j--) {
297 dst[j] = p[1]->value << 6 | p[0]->value;
302 pathn[0] = pathn[1] = 0;
303 memset(nodes[0] + 1, 0, (frontier - 1)*
sizeof(**nodes));
304 memset(nodes[1] + 1, 0, (frontier - 1)*
sizeof(**nodes));
310 for (j = i; j > froze; j--) {
311 dst[j] = p[1]->value << 6 | p[0]->value;
315 c->
band[0] = nodes[0][0]->state;
316 c->
band[1] = nodes[1][0]->state;
322 int xlow, xhigh, ilow, ihigh;
329 *dst = ihigh << 6 | ilow;
337 for (i = 0; i < nb_samples; i += 2)
345 const int16_t *
samples = (
const int16_t *)frame->
data[0];
360 if (nb_samples < frame->nb_samples) {
361 int16_t last_samples[2] = { samples[nb_samples], samples[nb_samples] };
#define FF_CODEC_CAP_INIT_CLEANUP
The codec allows calling the close function for deallocation even if the init function returned a fai...
struct G722Context::TrellisNode ** nodep_buf[2]
This structure describes decoded (raw) audio or video data.
#define AV_LOG_WARNING
Something somehow does not look correct.
static av_cold int init(AVCodecContext *avctx)
struct G722Context::TrellisPath * paths[2]
static void filter_samples(G722Context *c, const int16_t *samples, int *xlow, int *xhigh)
static void g722_encode_no_trellis(G722Context *c, uint8_t *dst, int nb_samples, const int16_t *samples)
void * av_calloc(size_t nmemb, size_t size)
Non-inlined equivalent of av_mallocz_array().
void(* apply_qmf)(const int16_t *prev_samples, int xout[2])
int ff_alloc_packet2(AVCodecContext *avctx, AVPacket *avpkt, int64_t size, int64_t min_size)
Check AVPacket size and/or allocate data.
static int encode_high(const struct G722Band *state, int xhigh)
Undefined Behavior In the C some operations are like signed integer dereferencing freed accessing outside allocated Undefined Behavior must not occur in a C it is not safe even if the output of undefined operations is unused The unsafety may seem nit picking but Optimizing compilers have in fact optimized code on the assumption that no undefined Behavior occurs Optimizing code based on wrong assumptions can and has in some cases lead to effects beyond the output of computations The signed integer overflow problem in speed critical code Code which is highly optimized and works with signed integers sometimes has the problem that often the output of the computation does not c
#define PREV_SAMPLES_BUF_SIZE
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
static av_cold int g722_encode_init(AVCodecContext *avctx)
struct G722Context::TrellisNode * node_buf[2]
const int16_t ff_g722_low_inv_quant6[64]
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
int initial_padding
Audio only.
int16_t prev_samples[PREV_SAMPLES_BUF_SIZE]
memory of past decoded samples
simple assert() macros that are a bit more flexible than ISO C assert().
AVCodec ff_adpcm_g722_encoder
const char * name
Name of the codec implementation.
struct G722Context::G722Band band[2]
#define AV_CODEC_CAP_SMALL_LAST_FRAME
Codec can be fed a final frame with a smaller size.
static void g722_encode_trellis(G722Context *c, int trellis, uint8_t *dst, int nb_samples, const int16_t *samples)
static av_cold int g722_encode_close(AVCodecContext *avctx)
static int encode_low(const struct G722Band *state, int xlow)
these buffered frames must be flushed immediately if a new input produces new the filter must not call request_frame to get more It must just process the frame or queue it The task of requesting more frames is left to the filter s request_frame method or the application If a filter has several the filter must be ready for frames arriving randomly on any input any filter with several inputs will most likely require some kind of queuing mechanism It is perfectly acceptable to have a limited queue and to drop frames when the inputs are too unbalanced request_frame For filters that do not use the this method is called when a frame is wanted on an output For a it should directly call filter_frame on the corresponding output For a if there are queued frames already one of these frames should be pushed If the filter should request a frame on one of its repeatedly until at least one frame has been pushed Return or at least make progress towards producing a frame
void ff_g722_update_low_predictor(struct G722Band *band, const int ilow)
static const float pred[4]
int frame_size
Number of samples per channel in an audio frame.
static const int16_t low_quant[33]
Libavcodec external API header.
AVSampleFormat
Audio sample formats.
main external API structure.
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
common internal api header.
common internal and external API header
int prev_samples_pos
the number of values in prev_samples
int trellis
trellis RD quantization
#define STORE_NODE(index, UPDATE, VALUE)
static av_always_inline int diff(const uint32_t a, const uint32_t b)
const int16_t ff_g722_high_inv_quant[4]
av_cold void ff_g722dsp_init(G722DSPContext *c)
static av_always_inline void encode_byte(G722Context *c, uint8_t *dst, const int16_t *samples)
static enum AVSampleFormat sample_fmts[]
Filter the word “frame” indicates either a video frame or a group of audio samples
static int g722_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, const AVFrame *frame, int *got_packet_ptr)
static av_always_inline int64_t ff_samples_to_time_base(AVCodecContext *avctx, int64_t samples)
Rescale from sample rate to AVCodecContext.time_base.
#define FFSWAP(type, a, b)
void ff_g722_update_high_predictor(struct G722Band *band, const int dhigh, const int ihigh)
Filter the word “frame” indicates either a video frame or a group of audio as stored in an AVFrame structure Format for each input and each output the list of supported formats For video that means pixel format For audio that means channel sample they are references to shared objects When the negotiation mechanism computes the intersection of the formats supported at each end of a all references to both lists are replaced with a reference to the intersection And when a single format is eventually chosen for a link amongst the remaining all references to the list are updated That means that if a filter requires that its input and output have the same format amongst a supported all it has to do is use a reference to the same list of formats query_formats can leave some formats unset and return AVERROR(EAGAIN) to cause the negotiation mechanism toagain later.That can be used by filters with complex requirements to use the format negotiated on one link to set the formats supported on another.Frame references ownership and permissions
#define AV_CH_LAYOUT_MONO
This structure stores compressed data.
int16_t scale_factor
delayed quantizer scale factor
int nb_samples
number of audio samples (per channel) described by this frame
int64_t pts
Presentation timestamp in AVStream->time_base units; the time at which the decompressed packet will b...
#define AV_NOPTS_VALUE
Undefined timestamp value.