[FFmpeg-devel] [PATCH 3/3] mmaldec with plain yuv420p without copy

[Mark Thompson]

On 11/02/2021 21:48, Lluís Batlle i Rossel wrote:
> From: Lluís Batlle i Rossell <viric at viric.name>
> 
> ---
>   libavcodec/mmaldec.c | 48 ++++++++++++++++++++------------------------
>   1 file changed, 22 insertions(+), 26 deletions(-)
> 
> diff --git a/libavcodec/mmaldec.c b/libavcodec/mmaldec.c
> index 4dfaacbb41..097b990f92 100644
> --- a/libavcodec/mmaldec.c
> +++ b/libavcodec/mmaldec.c
> @@ -119,10 +119,11 @@ static void ffmmal_release_frame(void *opaque, uint8_t *data)
>   
>   // Setup frame with a new reference to buffer. The buffer must have been
>   // allocated from the given pool.
> -static int ffmmal_set_ref(AVFrame *frame, FFPoolRef *pool,
> -                          MMAL_BUFFER_HEADER_T *buffer)
> +static int ffmmal_set_ref(AVCodecContext *avctx, AVFrame *frame,
> +        FFPoolRef *pool, MMAL_BUFFER_HEADER_T *buffer)
>   {
>       FFBufferRef *ref = av_mallocz(sizeof(*ref));
> +

Random whitespace change?

>       if (!ref)
>           return AVERROR(ENOMEM);
>   
> @@ -140,8 +141,19 @@ static int ffmmal_set_ref(AVFrame *frame, FFPoolRef *pool,
>       atomic_fetch_add_explicit(&ref->pool->refcount, 1, memory_order_relaxed);
>       mmal_buffer_header_acquire(buffer);
>   
> -    frame->format = AV_PIX_FMT_MMAL;
> -    frame->data[3] = (uint8_t *)ref->buffer;
> +    frame->format = avctx->pix_fmt;
> +
> +    if (avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
> +        int w = FFALIGN(avctx->width, 32);
> +        int h = FFALIGN(avctx->height, 16);
> +
> +        av_image_fill_arrays(frame->data, frame->linesize,
> +                buffer->data + buffer->type->video.offset[0],
> +                avctx->pix_fmt, w, h, 1);
> +    } else {
> +        frame->data[3] = (uint8_t *)ref->buffer;
> +    }
> +
>       return 0;
>   }
>   
> @@ -633,30 +645,14 @@ static int ffmal_copy_frame(AVCodecContext *avctx,  AVFrame *frame,

The function name is very misleading after this change.

>       frame->interlaced_frame = ctx->interlaced_frame;
>       frame->top_field_first = ctx->top_field_first;
>   
> -    if (avctx->pix_fmt == AV_PIX_FMT_MMAL) {
> -        if (!ctx->pool_out)
> -            return AVERROR_UNKNOWN; // format change code failed with OOM previously
> -
> -        if ((ret = ff_decode_frame_props(avctx, frame)) < 0)
> -            goto done;
> -
> -        if ((ret = ffmmal_set_ref(frame, ctx->pool_out, buffer)) < 0)
> -            goto done;
> -    } else {
> -        int w = FFALIGN(avctx->width, 32);
> -        int h = FFALIGN(avctx->height, 16);
> -        uint8_t *src[4];
> -        int linesize[4];
> +    if (!ctx->pool_out)
> +        return AVERROR_UNKNOWN; // format change code failed with OOM previously
>   
> -        if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
> -            goto done;
> +    if ((ret = ff_decode_frame_props(avctx, frame)) < 0)
> +        goto done;
>   
> -        av_image_fill_arrays(src, linesize,
> -                             buffer->data + buffer->type->video.offset[0],
> -                             avctx->pix_fmt, w, h, 1);
> -        av_image_copy(frame->data, frame->linesize, src, linesize,
> -                      avctx->pix_fmt, avctx->width, avctx->height);
> -    }
> +    if ((ret = ffmmal_set_ref(avctx, frame, ctx->pool_out, buffer)) < 0)
> +        goto done;
>   
>       frame->pts = buffer->pts == MMAL_TIME_UNKNOWN ? AV_NOPTS_VALUE : buffer->pts;
>   #if FF_API_PKT_PTS
> 

What happens if the user holds on to multiple of the output frames, perhaps because they are being queued into some other operation?  The number of buffers appears fixed, so what happens if they run out?

- Mark