[FFmpeg-devel] [PATCH] avcodec/mpeg4_unpack_bframes: cache input packets directly

James Almer jamrial at gmail.com
Wed Apr 25 22:50:50 EEST 2018


On 3/21/2018 11:21 AM, James Almer wrote:
> Avoids unnecessary allocs+copies and makes the code slightly simpler.
> 
> Signed-off-by: James Almer <jamrial at gmail.com>
> ---
>  libavcodec/mpeg4_unpack_bframes_bsf.c | 55 +++++++++++++----------------------
>  1 file changed, 20 insertions(+), 35 deletions(-)
> 
> diff --git a/libavcodec/mpeg4_unpack_bframes_bsf.c b/libavcodec/mpeg4_unpack_bframes_bsf.c
> index ba970794c5..e2e73446d8 100644
> --- a/libavcodec/mpeg4_unpack_bframes_bsf.c
> +++ b/libavcodec/mpeg4_unpack_bframes_bsf.c
> @@ -24,8 +24,7 @@
>  #include "mpeg4video.h"
>  
>  typedef struct UnpackBFramesBSFContext {
> -    uint8_t *b_frame_buf;
> -    int      b_frame_buf_size;
> +    AVPacket *b_frame;
>  } UnpackBFramesBSFContext;
>  
>  /* search next start code */
> @@ -71,18 +70,6 @@ static void scan_buffer(const uint8_t *buf, int buf_size,
>      }
>  }
>  
> -/* allocate new buffer and copy size bytes from src */
> -static uint8_t *create_new_buffer(const uint8_t *src, int size) {
> -    uint8_t *dst = av_malloc(size + AV_INPUT_BUFFER_PADDING_SIZE);
> -
> -    if (dst) {
> -        memcpy(dst, src, size);
> -        memset(dst + size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
> -    }
> -
> -    return dst;
> -}
> -
>  static int mpeg4_unpack_bframes_filter(AVBSFContext *ctx, AVPacket *out)
>  {
>      UnpackBFramesBSFContext *s = ctx->priv_data;
> @@ -97,20 +84,18 @@ static int mpeg4_unpack_bframes_filter(AVBSFContext *ctx, AVPacket *out)
>      av_log(ctx, AV_LOG_DEBUG, "Found %d VOP startcode(s) in this packet.\n", nb_vop);
>  
>      if (pos_vop2 >= 0) {
> -        if (s->b_frame_buf) {
> +        if (s->b_frame->data) {
>              av_log(ctx, AV_LOG_WARNING,
>                     "Missing one N-VOP packet, discarding one B-frame.\n");
> -            av_freep(&s->b_frame_buf);
> -            s->b_frame_buf_size = 0;
> +            av_packet_unref(s->b_frame);
>          }
>          /* store the packed B-frame in the BSFContext */
> -        s->b_frame_buf_size = in->size - pos_vop2;
> -        s->b_frame_buf      = create_new_buffer(in->data + pos_vop2, s->b_frame_buf_size);
> -        if (!s->b_frame_buf) {
> -            s->b_frame_buf_size = 0;
> -            ret = AVERROR(ENOMEM);
> +        ret = av_packet_ref(s->b_frame, in);
> +        if (ret < 0) {
>              goto fail;
>          }
> +        s->b_frame->size -= pos_vop2;
> +        s->b_frame->data += pos_vop2;
>      }
>  
>      if (nb_vop > 2) {
> @@ -118,29 +103,23 @@ static int mpeg4_unpack_bframes_filter(AVBSFContext *ctx, AVPacket *out)
>         "Found %d VOP headers in one packet, only unpacking one.\n", nb_vop);
>      }
>  
> -    if (nb_vop == 1 && s->b_frame_buf) {
> +    if (nb_vop == 1 && s->b_frame->data) {
>          /* use frame from BSFContext */
> +        av_packet_move_ref(out, s->b_frame);
> +
> +        /* use properties from current input packet */
>          ret = av_packet_copy_props(out, in);
>          if (ret < 0) {
>              goto fail;
>          }
>  
> -        ret = av_packet_from_data(out, s->b_frame_buf, s->b_frame_buf_size);
> -        if (ret < 0) {
> -            goto fail;
> -        }
>          if (in->size <= MAX_NVOP_SIZE) {
>              /* N-VOP */
>              av_log(ctx, AV_LOG_DEBUG, "Skipping N-VOP.\n");
> -            s->b_frame_buf      = NULL;
> -            s->b_frame_buf_size = 0;
>          } else {
>              /* copy packet into BSFContext */
> -            s->b_frame_buf_size = in->size;
> -            s->b_frame_buf      = create_new_buffer(in->data, in->size);
> -            if (!s->b_frame_buf) {
> -                s->b_frame_buf_size = 0;
> -                ret = AVERROR(ENOMEM);
> +            ret = av_packet_ref(s->b_frame, in);
> +            if (ret < 0) {
>                  goto fail;
>              }
>          }
> @@ -168,6 +147,12 @@ fail:
>  
>  static int mpeg4_unpack_bframes_init(AVBSFContext *ctx)
>  {
> +    UnpackBFramesBSFContext *s = ctx->priv_data;
> +
> +    s->b_frame = av_packet_alloc();
> +    if (!s->b_frame)
> +        return AVERROR(ENOMEM);
> +
>      if (ctx->par_in->extradata) {
>          int pos_p_ext = -1;
>          scan_buffer(ctx->par_in->extradata, ctx->par_in->extradata_size, &pos_p_ext, NULL, NULL);
> @@ -184,7 +169,7 @@ static int mpeg4_unpack_bframes_init(AVBSFContext *ctx)
>  static void mpeg4_unpack_bframes_close(AVBSFContext *bsfc)
>  {
>      UnpackBFramesBSFContext *ctx = bsfc->priv_data;
> -    av_freep(&ctx->b_frame_buf);
> +    av_packet_free(&ctx->b_frame);
>  }
>  
>  static const enum AVCodecID codec_ids[] = {

Pushed.


More information about the ffmpeg-devel mailing list