AV1代码学习：解码端aomdec.c的main函数

aom2.0的工程中，aomdec.c是解码端工程，主要完成码流的解析，并输出解码后的YUV，其入口是amdec.c中的main函数

在main函数里面做的工作很简单，首先解析了命令行参数loops，获取其解码文件的次数，然后开始进入解码循环main_loop函数中

其代码如下：

int main(int argc, const char **argv_) {
  unsigned int loops = 1, i;
  char **argv, **argi, **argj;
  struct arg arg;
  int error = 0;

  argv = argv_dup(argc - 1, argv_ + 1);
  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
    memset(&arg, 0, sizeof(arg));
    arg.argv_step = 1;

    if (arg_match(&arg, &looparg, argi)) {
      loops = arg_parse_uint(&arg);
      break;
    }
  }
  free(argv);
  for (i = 0; !error && i < loops; i++) error = main_loop(argc, argv_);
  return error;
}

main_loop函数，主要流程如下：

1. 初始化变量， 主要包括一些解码器的上下文、读取帧数、解码时间等
2. 解析命令行参数
3. 打开输入文件，判断输入文件类型（IVF/OBU/WEBM/RAW），打开输出的yuv文件
4. 解码器类型检查
5. 进入解码循环
   1. 通过read_frame从码流中读取一帧数据
   2. 调用aom_codec_decode解码当前帧
   3. 调用aom_codec_get_frame获得解码一帧后的数据，然后输出该帧（如果需要可以对当前帧进行缩放）
6. 释放变量

代码及注释如下：

static int main_loop(int argc, const char **argv_) {
  aom_codec_ctx_t decoder;
  char *fn = NULL;
  int i;
  int ret = EXIT_FAILURE;
  uint8_t *buf = NULL;
  size_t bytes_in_buffer = 0, buffer_size = 0;
  FILE *infile;
  int frame_in = 0, frame_out = 0, flipuv = 0, noblit = 0;
  int do_md5 = 0, progress = 0;
  int stop_after = 0, summary = 0, quiet = 1;
  int arg_skip = 0;
  int keep_going = 0;
  uint64_t dx_time = 0;
  struct arg arg;
  char **argv, **argi, **argj;

  int single_file;
  int use_y4m = 1;
  int opt_yv12 = 0;
  int opt_i420 = 0;
  int opt_raw = 0;
  aom_codec_dec_cfg_t cfg = { 0, 0, 0, !FORCE_HIGHBITDEPTH_DECODING };
  unsigned int fixed_output_bit_depth = 0;
  unsigned int is_annexb = 0;
  int frames_corrupted = 0;
  int dec_flags = 0;
  int do_scale = 0;
  int operating_point = 0;
  int output_all_layers = 0;
  int skip_film_grain = 0;
  int enable_row_mt = 0;
  aom_image_t *scaled_img = NULL;
  aom_image_t *img_shifted = NULL;
  int frame_avail, got_data, flush_decoder = 0;
  int num_external_frame_buffers = 0;
  struct ExternalFrameBufferList ext_fb_list = { 0, NULL };

  const char *outfile_pattern = NULL;
  char outfile_name[PATH_MAX] = { 0 };
  FILE *outfile = NULL;

  FILE *framestats_file = NULL;

  MD5Context md5_ctx;
  unsigned char md5_digest[16];

  struct AvxDecInputContext input = { NULL, NULL, NULL };
  struct AvxInputContext aom_input_ctx;
  memset(&aom_input_ctx, 0, sizeof(aom_input_ctx));
#if CONFIG_WEBM_IO
  struct WebmInputContext webm_ctx;
  memset(&webm_ctx, 0, sizeof(webm_ctx));
  input.webm_ctx = &webm_ctx;
#endif
  struct ObuDecInputContext obu_ctx = { NULL, NULL, 0, 0, 0 };
  int is_ivf = 0;

  obu_ctx.avx_ctx = &aom_input_ctx;
  input.obu_ctx = &obu_ctx;
  input.aom_input_ctx = &aom_input_ctx;

  /* Parse command line */
  exec_name = argv_[0];
  argv = argv_dup(argc - 1, argv_ + 1);

  aom_codec_iface_t *interface = NULL;
  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
    memset(&arg, 0, sizeof(arg));
    arg.argv_step = 1;

    if (arg_match(&arg, &help, argi)) {
      show_help(stdout, 0);
      exit(EXIT_SUCCESS);
    } else if (arg_match(&arg, &codecarg, argi)) {
      interface = get_aom_decoder_by_short_name(arg.val);
      if (!interface)
        die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
    } else if (arg_match(&arg, &looparg, argi)) {
      // no-op
    } else if (arg_match(&arg, &outputfile, argi)) { //输出文件类型
      outfile_pattern = arg.val;
    } else if (arg_match(&arg, &use_yv12, argi)) {
      use_y4m = 0;
      flipuv = 1;
      opt_yv12 = 1;
      opt_i420 = 0;
      opt_raw = 0;
    } else if (arg_match(&arg, &use_i420, argi)) {
      use_y4m = 0;
      flipuv = 0;
      opt_yv12 = 0;
      opt_i420 = 1;
      opt_raw = 0;
    } else if (arg_match(&arg, &rawvideo, argi)) {
      use_y4m = 0;
      opt_yv12 = 0;
      opt_i420 = 0;
      opt_raw = 1;
    } else if (arg_match(&arg, &flipuvarg, argi)) {
      flipuv = 1;
    } else if (arg_match(&arg, &noblitarg, argi)) {
      noblit = 1;
    } else if (arg_match(&arg, &progressarg, argi)) {
      progress = 1;
    } else if (arg_match(&arg, &limitarg, argi)) {
      stop_after = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &skiparg, argi)) {
      arg_skip = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &md5arg, argi)) {
      do_md5 = 1;
    } else if (arg_match(&arg, &framestatsarg, argi)) {
      framestats_file = fopen(arg.val, "w");
      if (!framestats_file) {
        die("Error: Could not open --framestats file (%s) for writing.\n",
            arg.val);
      }
    } else if (arg_match(&arg, &summaryarg, argi)) {
      summary = 1;
    } else if (arg_match(&arg, &threadsarg, argi)) {
      cfg.threads = arg_parse_uint(&arg);
#if !CONFIG_MULTITHREAD
      if (cfg.threads > 1) {
        die("Error: --threads=%d is not supported when CONFIG_MULTITHREAD = "
            "0.\n",
            cfg.threads);
      }
#endif
    } else if (arg_match(&arg, &rowmtarg, argi)) {
      enable_row_mt = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &verbosearg, argi)) {
      quiet = 0;
    } else if (arg_match(&arg, &scalearg, argi)) {
      do_scale = 1;
    } else if (arg_match(&arg, &fb_arg, argi)) {
      num_external_frame_buffers = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &continuearg, argi)) {
      keep_going = 1;
    } else if (arg_match(&arg, &outbitdeptharg, argi)) {
      fixed_output_bit_depth = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &isannexb, argi)) {
      is_annexb = 1;
      input.obu_ctx->is_annexb = 1;
    } else if (arg_match(&arg, &oppointarg, argi)) {
      operating_point = arg_parse_int(&arg);
    } else if (arg_match(&arg, &outallarg, argi)) {
      output_all_layers = 1;
    } else if (arg_match(&arg, &skipfilmgrain, argi)) {
      skip_film_grain = 1;
    } else {
      argj++;
    }
  }

  /* Check for unrecognized options */
  for (argi = argv; *argi; argi++)
    if (argi[0][0] == '-' && strlen(argi[0]) > 1)
      die("Error: Unrecognized option %s\n", *argi);

  /* Handle non-option arguments */
  fn = argv[0];

  if (!fn) {
    free(argv);
    fprintf(stderr, "No input file specified!\n");
    usage_exit();
  }
  /* Open file */
  infile = strcmp(fn, "-") ? fopen(fn, "rb") : set_binary_mode(stdin);

  if (!infile) {
    fatal("Failed to open input file '%s'", strcmp(fn, "-") ? fn : "stdin");
  }
#if CONFIG_OS_SUPPORT
  /* Make sure we don't dump to the terminal, unless forced to with -o - */
  if (!outfile_pattern && isatty(STDOUT_FILENO) && !do_md5 && !noblit) {
    fprintf(stderr,
            "Not dumping raw video to your terminal. Use '-o -' to "
            "override.\n");
    free(argv);
    return EXIT_FAILURE;
  }
#endif
  input.aom_input_ctx->filename = fn;
  input.aom_input_ctx->file = infile;
  // 判断输入文件类型
  if (file_is_ivf(input.aom_input_ctx)) {
    input.aom_input_ctx->file_type = FILE_TYPE_IVF;
    is_ivf = 1;
  }
#if CONFIG_WEBM_IO
  else if (file_is_webm(input.webm_ctx, input.aom_input_ctx))
    input.aom_input_ctx->file_type = FILE_TYPE_WEBM;
#endif
  else if (file_is_obu(&obu_ctx))
    input.aom_input_ctx->file_type = FILE_TYPE_OBU;
  else if (file_is_raw(input.aom_input_ctx))
    input.aom_input_ctx->file_type = FILE_TYPE_RAW;
  else {
    fprintf(stderr, "Unrecognized input file type.\n");
#if !CONFIG_WEBM_IO
    fprintf(stderr, "aomdec was built without WebM container support.\n");
#endif
    free(argv);
    return EXIT_FAILURE;
  }

  outfile_pattern = outfile_pattern ? outfile_pattern : "-";
  single_file = is_single_file(outfile_pattern);

  if (!noblit && single_file) {
    generate_filename(outfile_pattern, outfile_name, PATH_MAX,
                      aom_input_ctx.width, aom_input_ctx.height, 0);
    if (do_md5)
      MD5Init(&md5_ctx);
    else
      outfile = open_outfile(outfile_name);
  }

  if (use_y4m && !noblit) {
    if (!single_file) {
      fprintf(stderr,
              "YUV4MPEG2 not supported with output patterns,"
              " try --i420 or --yv12 or --rawvideo.\n");
      return EXIT_FAILURE;
    }

#if CONFIG_WEBM_IO
    if (aom_input_ctx.file_type == FILE_TYPE_WEBM) {
      if (webm_guess_framerate(input.webm_ctx, input.aom_input_ctx)) {
        fprintf(stderr,
                "Failed to guess framerate -- error parsing "
                "webm file?\n");
        return EXIT_FAILURE;
      }
    }
#endif
  }

  aom_codec_iface_t *fourcc_interface =
      get_aom_decoder_by_fourcc(aom_input_ctx.fourcc);

  if (is_ivf && !fourcc_interface)
    fatal("Unsupported fourcc: %x\n", aom_input_ctx.fourcc);

  if (interface && fourcc_interface && interface != fourcc_interface)
    warn("Header indicates codec: %s\n",
         aom_codec_iface_name(fourcc_interface));
  else
    interface = fourcc_interface;

  if (!interface) interface = get_aom_decoder_by_index(0);

  dec_flags = 0;
  if (aom_codec_dec_init(&decoder, interface, &cfg, dec_flags)) {
    fprintf(stderr, "Failed to initialize decoder: %s\n",
            aom_codec_error(&decoder));
    goto fail2;
  }

  if (!quiet) fprintf(stderr, "%s\n", decoder.name);

  if (AOM_CODEC_CONTROL_TYPECHECKED(&decoder, AV1D_SET_IS_ANNEXB, is_annexb)) {
    fprintf(stderr, "Failed to set is_annexb: %s\n", aom_codec_error(&decoder));
    goto fail;
  }

  if (AOM_CODEC_CONTROL_TYPECHECKED(&decoder, AV1D_SET_OPERATING_POINT,
                                    operating_point)) {
    fprintf(stderr, "Failed to set operating_point: %s\n",
            aom_codec_error(&decoder));
    goto fail;
  }

  if (AOM_CODEC_CONTROL_TYPECHECKED(&decoder, AV1D_SET_OUTPUT_ALL_LAYERS,
                                    output_all_layers)) {
    fprintf(stderr, "Failed to set output_all_layers: %s\n",
            aom_codec_error(&decoder));
    goto fail;
  }

  if (AOM_CODEC_CONTROL_TYPECHECKED(&decoder, AV1D_SET_SKIP_FILM_GRAIN,
                                    skip_film_grain)) {
    fprintf(stderr, "Failed to set skip_film_grain: %s\n",
            aom_codec_error(&decoder));
    goto fail;
  }

  if (AOM_CODEC_CONTROL_TYPECHECKED(&decoder, AV1D_SET_ROW_MT, enable_row_mt)) {
    fprintf(stderr, "Failed to set row multithreading mode: %s\n",
            aom_codec_error(&decoder));
    goto fail;
  }

  // 跳过前几个帧
  if (arg_skip) fprintf(stderr, "Skipping first %d frames.\n", arg_skip);
  while (arg_skip) {
    if (read_frame(&input, &buf, &bytes_in_buffer, &buffer_size)) break;
    arg_skip--;
  }

  if (num_external_frame_buffers > 0) {
    ext_fb_list.num_external_frame_buffers = num_external_frame_buffers;
    ext_fb_list.ext_fb = (struct ExternalFrameBuffer *)calloc(
        num_external_frame_buffers, sizeof(*ext_fb_list.ext_fb));
    if (aom_codec_set_frame_buffer_functions(&decoder, get_av1_frame_buffer,
                                             release_av1_frame_buffer,
                                             &ext_fb_list)) {
      fprintf(stderr, "Failed to configure external frame buffers: %s\n",
              aom_codec_error(&decoder));
      goto fail;
    }
  }

  frame_avail = 1;
  got_data = 0;

  if (framestats_file) fprintf(framestats_file, "bytes,qp\r\n");

  /* Decode file */
  while (frame_avail || got_data) {
    aom_codec_iter_t iter = NULL;
    aom_image_t *img;
    struct aom_usec_timer timer;
    int corrupted = 0;

    frame_avail = 0;
    if (!stop_after || frame_in < stop_after) {
      if (!read_frame(&input, &buf, &bytes_in_buffer, &buffer_size)) { //从码流中读取一帧数据
        frame_avail = 1;
        frame_in++;

        aom_usec_timer_start(&timer);//计时器开始

        if (aom_codec_decode(&decoder, buf, bytes_in_buffer, NULL)) { //进行解码
          const char *detail = aom_codec_error_detail(&decoder);
          warn("Failed to decode frame %d: %s", frame_in,
               aom_codec_error(&decoder));

          if (detail) warn("Additional information: %s", detail);
          if (!keep_going) goto fail;
        }

        if (framestats_file) { //打印解码时的数据
          int qp;
          if (AOM_CODEC_CONTROL_TYPECHECKED(&decoder, AOMD_GET_LAST_QUANTIZER,
                                            &qp)) {
            warn("Failed AOMD_GET_LAST_QUANTIZER: %s",
                 aom_codec_error(&decoder));
            if (!keep_going) goto fail;
          }
          fprintf(framestats_file, "%d,%d\r\n", (int)bytes_in_buffer, qp);
        }

        aom_usec_timer_mark(&timer);
        dx_time += aom_usec_timer_elapsed(&timer);//累计解码时间
      } else {
        flush_decoder = 1;//如果没有读取到数据就刷新解码器
      }
    } else {
      flush_decoder = 1;//停止读取帧数据，刷新解码器
    }

    aom_usec_timer_start(&timer);

    if (flush_decoder) {
      // Flush the decoder. 刷新解码器
      if (aom_codec_decode(&decoder, NULL, 0, NULL)) {
        warn("Failed to flush decoder: %s", aom_codec_error(&decoder));
      }
    }

    aom_usec_timer_mark(&timer);
    dx_time += aom_usec_timer_elapsed(&timer);

    got_data = 0;
    while ((img = aom_codec_get_frame(&decoder, &iter))) {
      ++frame_out;
      got_data = 1;

      if (AOM_CODEC_CONTROL_TYPECHECKED(&decoder, AOMD_GET_FRAME_CORRUPTED,
                                        &corrupted)) { //检查帧是否损坏
        warn("Failed AOM_GET_FRAME_CORRUPTED: %s", aom_codec_error(&decoder));
        if (!keep_going) goto fail;
      }
      frames_corrupted += corrupted;

      if (progress) show_progress(frame_in, frame_out, dx_time); //打印每一帧的解码信息

      if (!noblit) { //如果对解码帧进行处理
        const int PLANES_YUV[] = { AOM_PLANE_Y, AOM_PLANE_U, AOM_PLANE_V };
        const int PLANES_YVU[] = { AOM_PLANE_Y, AOM_PLANE_V, AOM_PLANE_U };
        const int *planes = flipuv ? PLANES_YVU : PLANES_YUV;

        if (do_scale) { //如果需要输出帧进行缩放
          if (frame_out == 1) {
            // If the output frames are to be scaled to a fixed display size
            // then use the width and height specified in the container. If
            // either of these is set to 0, use the display size set in the
            // first frame header. If that is unavailable, use the raw decoded
            // size of the first decoded frame.
			// 如果要将输出帧缩放为固定的显示大小，则使用容器中指定的宽度和高度。
			// 如果其中任何一个设置为0，则使用在第一个帧标题中设置的显示大小。
			// 如果不可用，则使用第一个解码帧的原始解码大小。
            int render_width = aom_input_ctx.width;
            int render_height = aom_input_ctx.height;
            if (!render_width || !render_height) {
              int render_size[2];
              if (AOM_CODEC_CONTROL_TYPECHECKED(&decoder, AV1D_GET_DISPLAY_SIZE,
                                                render_size)) {
                // As last resort use size of first frame as display size.
                render_width = img->d_w;
                render_height = img->d_h;
              } else {
                render_width = render_size[0];
                render_height = render_size[1];
              }
            }
            scaled_img =
                aom_img_alloc(NULL, img->fmt, render_width, render_height, 16);
            scaled_img->bit_depth = img->bit_depth;
            scaled_img->monochrome = img->monochrome;
            scaled_img->csp = img->csp;
          }

          if (img->d_w != scaled_img->d_w || img->d_h != scaled_img->d_h) {
#if CONFIG_LIBYUV
            libyuv_scale(img, scaled_img, kFilterBox);
            img = scaled_img;
#else
            fprintf(
                stderr,
                "Failed to scale output frame: %s.\n"
                "libyuv is required for scaling but is currently disabled.\n"
                "Be sure to specify -DCONFIG_LIBYUV=1 when running cmake.\n",
                aom_codec_error(&decoder));
            goto fail;
#endif
          }
        }
        // Default to codec bit depth if output bit depth not set 
		// 如果未设置输出位深度，则默认为编解码器位深度
        unsigned int output_bit_depth;
        if (!fixed_output_bit_depth && single_file) {
          output_bit_depth = img->bit_depth;
        } else {
          output_bit_depth = fixed_output_bit_depth;
        }
        // Shift up or down if necessary
        if (output_bit_depth != 0)
          aom_shift_img(output_bit_depth, &img, &img_shifted);

        aom_input_ctx.width = img->d_w;
        aom_input_ctx.height = img->d_h;

        int num_planes = (opt_raw && img->monochrome) ? 1 : 3;
        if (single_file) {
          if (use_y4m) { //如果输出y4m格式的视频
            char y4m_buf[Y4M_BUFFER_SIZE] = { 0 };
            size_t len = 0;
            if (frame_out == 1) {
              // Y4M file header
              len = y4m_write_file_header(
                  y4m_buf, sizeof(y4m_buf), aom_input_ctx.width,
                  aom_input_ctx.height, &aom_input_ctx.framerate,
                  img->monochrome, img->csp, img->fmt, img->bit_depth);
              if (img->csp == AOM_CSP_COLOCATED) {
                fprintf(stderr,
                        "Warning: Y4M lacks a colorspace for colocated "
                        "chroma. Using a placeholder.\n");
              }
              if (do_md5) {
                MD5Update(&md5_ctx, (md5byte *)y4m_buf, (unsigned int)len);
              } else {
                fputs(y4m_buf, outfile);
              }
            }

            // Y4M frame header
            len = y4m_write_frame_header(y4m_buf, sizeof(y4m_buf));
            if (do_md5) {
              MD5Update(&md5_ctx, (md5byte *)y4m_buf, (unsigned int)len);
              y4m_update_image_md5(img, planes, &md5_ctx);
            } else {
              fputs(y4m_buf, outfile);
              y4m_write_image_file(img, planes, outfile);
            }
          } else { //输出原始YUV视频
            if (frame_out == 1) {
              // Check if --yv12 or --i420 options are consistent with the
              // bit-stream decoded
              if (opt_i420) {
                if (img->fmt != AOM_IMG_FMT_I420 &&
                    img->fmt != AOM_IMG_FMT_I42016) {
                  fprintf(stderr,
                          "Cannot produce i420 output for bit-stream.\n");
                  goto fail;
                }
              }
              if (opt_yv12) {
                if ((img->fmt != AOM_IMG_FMT_I420 &&
                     img->fmt != AOM_IMG_FMT_YV12) ||
                    img->bit_depth != 8) {
                  fprintf(stderr,
                          "Cannot produce yv12 output for bit-stream.\n");
                  goto fail;
                }
              }
            }
            if (do_md5) {
              raw_update_image_md5(img, planes, num_planes, &md5_ctx);
            } else {
              raw_write_image_file(img, planes, num_planes, outfile);
            }
          }
        } else { //！singleFile
          generate_filename(outfile_pattern, outfile_name, PATH_MAX, img->d_w,
                            img->d_h, frame_in);
          if (do_md5) {
            MD5Init(&md5_ctx);
            if (use_y4m) {
              y4m_update_image_md5(img, planes, &md5_ctx);
            } else {
              raw_update_image_md5(img, planes, num_planes, &md5_ctx);
            }
            MD5Final(md5_digest, &md5_ctx);
            print_md5(md5_digest, outfile_name);
          } else {
            outfile = open_outfile(outfile_name);
            if (use_y4m) {
              y4m_write_image_file(img, planes, outfile);
            } else {
              raw_write_image_file(img, planes, num_planes, outfile);
            }
            fclose(outfile);
          }
        }
      }
    }// 解码帧
  }// Decode file

  if (summary || progress) {
    show_progress(frame_in, frame_out, dx_time);
    fprintf(stderr, "\n");
  }

  if (frames_corrupted) {
    fprintf(stderr, "WARNING: %d frames corrupted.\n", frames_corrupted);
  } else {
    ret = EXIT_SUCCESS;
  }

fail:

  if (aom_codec_destroy(&decoder)) {
    fprintf(stderr, "Failed to destroy decoder: %s\n",
            aom_codec_error(&decoder));
  }

fail2:

  if (!noblit && single_file) {
    if (do_md5) {
      MD5Final(md5_digest, &md5_ctx);
      print_md5(md5_digest, outfile_name);
    } else {
      fclose(outfile);
    }
  }

#if CONFIG_WEBM_IO
  if (input.aom_input_ctx->file_type == FILE_TYPE_WEBM)
    webm_free(input.webm_ctx);
#endif
  if (input.aom_input_ctx->file_type == FILE_TYPE_OBU)
    obudec_free(input.obu_ctx);

  if (input.aom_input_ctx->file_type != FILE_TYPE_WEBM) free(buf);

  if (scaled_img) aom_img_free(scaled_img);
  if (img_shifted) aom_img_free(img_shifted);

  for (i = 0; i < ext_fb_list.num_external_frame_buffers; ++i) {
    free(ext_fb_list.ext_fb[i].data);
  }
  free(ext_fb_list.ext_fb);

  fclose(infile);
  if (framestats_file) fclose(framestats_file);

  free(argv);

  return ret;
}