SandBox/Pavel: media improvement.patch
| File media improvement.patch, 11.5 KB (added by pap, 15 years ago) |
|---|
-
src/media_decoder/media_decoder.cpp
### Eclipse Workspace Patch 1.0 #P sophie2-native
48 48 Response res = MediaResponse::createRequestBytes(s->pos, size); 49 49 BRIDGE.sendResponse(res); 50 50 Command *cmd = BRIDGE.readCommand(); 51 if (cmd->getId() == STOP_COMMAND_ID) { 52 BRIDGE.sendResponse(BRIDGE.doStop()); 53 exit(0); 54 } 51 55 assert(cmd->getId() == SEND_BYTES_COMMAND); 52 56 SendBytesCommand *bytesCmd = (SendBytesCommand *) cmd; 53 57 assert(s->pos == bytesCmd->filePos); … … 141 145 DECODER.initAudio(); 142 146 } 143 147 DECODER.isOpen = true; 148 DECODER.shifted = true; 149 DECODER.lastVideoMillis = -1; 150 DECODER.lastVideoDur = -1; 144 151 145 152 //insufficient info 146 153 fprintf(stderr, "\n"); -
src/media_decoder/media_decoder.h
20 20 //#define TRACE do{} while(false) 21 21 //#undef DUMP 22 22 //#define DUMP(x) do{} while(false) 23 24 23 static const int SAMPLE_RATE = 44100; 25 24 static const int CHANNELS = 2; 26 25 static const int CHUNK_SAMPLES = SAMPLE_RATE / 100; /* 1/100s = 10ms */ … … 31 30 static const int BUFFER_EXTRA = 1000; //ms 32 31 static const int BUFFER_LEN = BUFFER_STEP + BUFFER_EXTRA; //ms 33 32 34 35 33 class Decoder { 36 34 public: 37 35 Decoder() { 38 36 isOpen = false; 39 37 } 40 41 38 virtual ~Decoder() { 42 39 if (isOpen) { 43 40 if (videoStream > -1) { … … 52 49 isOpen = false; 53 50 } 54 51 } 55 56 52 int seek(int millis, int flags) { 57 53 assert(isOpen); 58 //AVSEEK_FLAG_ANY59 54 if (av_seek_frame(pFormatContext, -1, millis * 1000, flags) >= 0) { 60 55 return pFormatContext->data_offset; 61 56 } 62 fprintf(stderr, "Seeking ERROR!\tSeeking for : %d millisecond\n", 63 millis); 57 fprintf(stderr, "Seeking ERROR!\tSeeking for : %d millisecond\n", millis); 64 58 return -1; 65 59 66 60 } … … 71 65 if (av_seek_frame(pFormatContext, stream, rescaleMillis, flags) >= 0) { 72 66 return pFormatContext->data_offset; 73 67 } 74 fprintf(stderr, 75 "Seeking Stream ERROR!\tSeeking for : %d millisecond\n", 76 millis); 68 fprintf(stderr, "Seeking Stream ERROR!\tSeeking for : %d millisecond\n", millis); 77 69 return -1; 78 70 79 71 } 80 81 72 int rescaleToStream(int millis, int stream) { 82 73 int num = pFormatContext->streams[stream]->time_base.num; 83 74 int den = pFormatContext->streams[stream]->time_base.den; … … 93 84 94 85 vector<byte> lastAudioBuffer; 95 86 int lastAudioStep; 87 bool shifted; 88 vector<byte> lastVideoFrame; 89 int lastVideoMillis; 90 int lastVideoDur; 96 91 97 92 vector<byte> const& getAudioBuffer(int step) { 98 //step is in STEP_SIZE 99 assert(isOpen); 100 if (lastAudioStep != step) { 101 AVPacket packet; 102 lastAudioBuffer.clear(); 103 int millis = step * BUFFER_STEP; 104 seekStream(millis, audioStream, AVSEEK_FLAG_ANY); //AVSEEK_FLAG_BACKWARD 105 int currentTime = rescaleToStream(millis, audioStream); 106 int targetTime = rescaleToStream(millis + BUFFER_LEN, audioStream); 107 bool first = true; 108 while (currentTime <= targetTime && av_read_frame(pFormatContext, 109 &packet) >= 0) { 110 if (first) { 111 double firstSeconds = rescaleToSeconds(packet.pts, 112 audioStream); 113 double delta = firstSeconds - millis / 1000.0; 114 assert(delta >= 0); 115 int padding = int(delta * 1000 / CHUNK_MILLIS 116 * CHUNK_SAMPLES + 0.5) * 4; 117 lastAudioBuffer.insert(lastAudioBuffer.begin(), padding, 0); 118 //DUMP(lastAudioBuffer.size()); 119 first = false; 93 //step is in STEP_SIZE 94 assert(isOpen); 95 if (lastAudioStep != step) { 96 AVPacket packet; 97 lastAudioBuffer.clear(); 98 int millis = step * BUFFER_STEP; 99 //seekStream(millis, audioStream, AVSEEK_FLAG_ANY); //AVSEEK_FLAG_BACKWARD 100 seek(millis, AVSEEK_FLAG_ANY | AVSEEK_FLAG_BACKWARD); 101 shifted = true; 102 int currentTime = rescaleToStream(millis, audioStream); 103 int targetTime = rescaleToStream(millis + BUFFER_LEN, audioStream); 104 bool first = true; 105 while (currentTime <= targetTime && av_read_frame(pFormatContext, 106 &packet) >= 0) { 107 if (first && packet.stream_index == audioStream) { 108 double firstSeconds = rescaleToSeconds(packet.pts, 109 audioStream); 110 double delta = firstSeconds - millis / 1000.0; 111 if (delta < 0) { 112 DUMP(millis); 113 DUMP(firstSeconds * 1000); 114 av_free_packet(&packet); 115 continue; 116 } 117 int padding = int(delta * 1000 / CHUNK_MILLIS * CHUNK_SAMPLES + 0.5) * 4; 118 if (padding > 0) { 119 DUMP(padding); 120 DUMP(lastAudioBuffer.size()); 121 } 122 lastAudioBuffer.insert(lastAudioBuffer.begin(), padding, 0); 123 first = false; 124 } 125 if (packet.stream_index == audioStream) { 126 vector<byte> packetSamples = decodeAudio(packet); 127 lastAudioBuffer.insert(lastAudioBuffer.end(), 128 packetSamples.begin(), packetSamples.end()); 129 currentTime = packet.pts + packet.duration; 130 //currentTime += packet.duration; 131 packetSamples.clear(); 132 } 133 if (packet.data) { 134 av_free_packet(&packet); 135 } 120 136 } 121 if (packet.stream_index == audioStream) { 122 vector<byte> packetSamples = decodeAudio3(packet); 123 lastAudioBuffer.insert(lastAudioBuffer.end(), 124 packetSamples.begin(), packetSamples.end()); 125 currentTime = packet.pts + packet.duration; 126 //currentTime += packet.duration; 127 packetSamples.clear(); 128 } 129 if (packet.data) { 130 av_free_packet(&packet); 131 } 137 //double audioEndSeconds = rescaleToSeconds(currentTime, audioStream); 138 //assert(audioEndSeconds * 1000 > millis + BUFFER_LEN); 139 //DUMP(millis / 1000.0); 140 //DUMP(audioEndSeconds); 141 //DUMP(double(lastAudioBuffer.size()) / CHUNK_BYTES / 100); 142 lastAudioStep = step; 132 143 } 133 //double audioEndSeconds = rescaleToSeconds(currentTime, audioStream); 134 //assert(audioEndSeconds * 1000 > millis + BUFFER_LEN); 135 //DUMP(millis / 1000.0); 136 //DUMP(audioEndSeconds); 137 //DUMP(double(lastAudioBuffer.size()) / CHUNK_BYTES / 100); 138 lastAudioStep = step; 139 } 140 return lastAudioBuffer; 144 return lastAudioBuffer; 141 145 } 142 146 vector<byte> getAudio(int millis) { 143 147 assert(isOpen); … … 151 155 } else { 152 156 stepNum = millis / BUFFER_STEP; 153 157 } 154 //DUMP(millis);155 //DUMP(stepNum);156 158 vector<byte> const& v = getAudioBuffer(stepNum); 157 159 int index = (millis - stepNum * BUFFER_STEP) / CHUNK_MILLIS; 158 160 return vector<byte> (v.begin() + index * CHUNK_BYTES, v.begin() … … 217 219 assert(isOpen == true); 218 220 avpicture_fill((AVPicture*) pFrameRGB, buffer, pixFMT, 219 221 pVideoCodecCtx->width, pVideoCodecCtx->height); 220 int frameFinished, result;222 int frameFinished, packetCnt; 221 223 AVPacket packet; 222 224 223 225 int64_t pktTime; 224 226 AVRational pktTimeBase; 225 227 int pktMillis; 226 seek(millis, AVSEEK_FLAG_BACKWARD); 227 228 result = 0; 228 if (shifted) { 229 seekStream(millis, videoStream, AVSEEK_FLAG_BACKWARD); 230 shifted = false; 231 } 232 if (millis >= lastVideoMillis && millis <= lastVideoMillis + lastVideoDur){ 233 data = lastVideoFrame; 234 DUMP(lastVideoFrame.size()); 235 return lastVideoFrame.size(); 236 } 237 packetCnt = 0; 229 238 while (av_read_frame(pFormatContext, &packet) >= 0) { 230 239 if (packet.stream_index == videoStream) { 240 packetCnt++; 231 241 pktTimeBase = pFormatContext->streams[videoStream]->time_base; 232 pktTime = av_rescale(packet.pts, AV_TIME_BASE 233 * (int64_t) pktTimeBase.num, pktTimeBase.den); 234 242 pktTime = av_rescale(packet.dts, AV_TIME_BASE * (int64_t) pktTimeBase.num, pktTimeBase.den); 235 243 pktMillis = pktTime * 1000.0 / AV_TIME_BASE + 0.5; 236 244 237 int ret = avcodec_decode_video2(pVideoCodecCtx, pFrame, 238 &frameFinished, &packet); 245 int64_t pktDur = av_rescale(packet.duration, AV_TIME_BASE * (int64_t) pktTimeBase.num, pktTimeBase.den); 246 pktDur = pktDur * 1000.0 / AV_TIME_BASE + 0.5; 247 int ret = avcodec_decode_video2(pVideoCodecCtx, pFrame, &frameFinished, &packet); 239 248 assert(ret >= 0); 240 if (pktMillis > millis) { 249 // if (pktMillis + pktDur < millis){ 250 // av_free_packet(&packet); 251 // continue; 252 // } 253 if (pktMillis >= millis) { 241 254 if (frameFinished) { 242 return decodeFrame(data); 255 DUMP(packetCnt); 256 int res = decodeFrame(data); 257 lastVideoMillis = millis; 258 lastVideoDur = pktDur; 259 lastVideoFrame = data; 260 return res; 243 261 } 244 return -1;245 262 } 246 263 } 247 264 av_free_packet(&packet); … … 249 266 throw runtime_error("err_corrupted_frame: Frame can't be decoded"); 250 267 } 251 268 252 vector<byte> decodeAudio 3(AVPacket& packet) {269 vector<byte> decodeAudio(AVPacket& packet) { 253 270 ReSampleContext *resmpCtx = av_audio_resample_init(CHANNELS, 254 271 pAudioCodecCtx->channels, SAMPLE_RATE, 255 272 pAudioCodecCtx->sample_rate, SAMPLE_FMT_S16, … … 258 275 vector<byte> audioBuf(BUFFER_SIZE, 0); 259 276 int audioSize = 0; 260 277 int len; 261 262 278 uint8_t* oldPacketData = packet.data; 279 int oldPacketSize = packet.size; 263 280 264 281 do { 265 282 int tempAudioSize = BUFFER_SIZE; 266 283 len = avcodec_decode_audio3(pAudioCodecCtx, 267 (int16_t*) (&audioBuf[0] + audioSize), &tempAudioSize, &packet); 284 (int16_t*) (&audioBuf[0] + audioSize), &tempAudioSize, 285 &packet); 268 286 if (len <= 0) { 269 287 throw runtime_error("err_damaged_packet: length <= 0"); 270 288 } 271 272 289 audioSize += tempAudioSize; 273 if (audioBuf.size() - audioSize < BUFFER_SIZE / 2) {290 if (audioBuf.size() - audioSize < BUFFER_SIZE / 2) { 274 291 audioBuf.resize(audioBuf.size() + BUFFER_SIZE); 275 292 } 276 277 293 packet.size -= len; 294 packet.data += len; 278 295 279 } while (packet.size > 0);296 } while (packet.size > 0); 280 297 packet.data = oldPacketData; 281 298 packet.size = oldPacketSize; 282 299 … … 349 366 GetFrameCommand(vector<byte> const& data, int& pos) : 350 367 Command(data, pos) { 351 368 timeMillis = readInt(data, pos); 352 assert(pos == (int) data.size());369 assert(pos == (int) data.size()); 353 370 assert(getId() == GET_FRAME_COMMAND_ID); 354 371 } 355 372 … … 364 381 GetAudioCommand(vector<byte> const& data, int& pos) : 365 382 Command(data, pos) { 366 383 timeMillis = readInt(data, pos); 367 assert(pos == (int) data.size());384 assert(pos == (int) data.size()); 368 385 assert(getId() == GET_AUDIO_COMMAND_ID); 369 386 } 370 387 … … 378 395 public: 379 396 GetInfoCommand(vector<byte> const& data, int& pos) : 380 397 Command(data, pos) { 381 assert(pos == (int) data.size());398 assert(pos == (int) data.size()); 382 399 assert(getId() == GET_INFO_COMMAND_ID); 383 400 } 384 401 … … 391 408 OpenCommand(vector<byte> const& data, int& pos) : 392 409 Command(data, pos) { 393 410 fileSize = readInt(data, pos); 394 assert(pos == (int) data.size());411 assert(pos == (int) data.size()); 395 412 assert(getId() == OPEN_COMMAND_ID); 396 413 } 397 414 virtual Response execute(); … … 407 424 Command(data, pos) { 408 425 filePos = readInt(data, pos); 409 426 fileData = readBinData(data, pos); 410 assert(pos == (int) data.size());427 assert(pos == (int) data.size()); 411 428 assert(getId() == SEND_BYTES_COMMAND); 412 429 } 413 430 virtual Response execute() { … … 455 472 } 456 473 static Response createAudio(vector<byte> chunk, int pos) { 457 474 vector<byte> data; 458 assert((int) chunk.size() == CHUNK_BYTES);475 assert((int) chunk.size() == CHUNK_BYTES); 459 476 Message::writeInt(data, AUDIO_RESPONSE); 460 477 Message::writeInt(data, pos); 461 478 Message::writeBinData(data, chunk); … … 467 484 Message::writeInt(data, millis); 468 485 Message::writeInt(data, width); 469 486 Message::writeInt(data, height); 470 Message::writeInt(data, (int) hasAudio);487 Message::writeInt(data, (int) hasAudio); 471 488 return Response(data); 472 489 } 473 490 };
