[FFmpeg-devel] [PATCH 2/6] Frame-based multithreading framework using pthreads

[Alexander Strange]

On Feb 6, 2011, at 10:57 AM, Justin Ruggles wrote:

> On 02/05/2011 06:38 PM, Alexander Strange wrote:
> 
>> On Sat, Feb 5, 2011 at 1:10 PM, Justin Ruggles <justin.ruggles at gmail.com> wrote:
>>> On 02/05/2011 12:28 AM, Alexander Strange wrote:
>>> 
>>>> +/**
>>>> + * Context used by codec threads and stored in their AVCodecContext thread_opaque.
>>>> + */
>>>> +typedef struct PerThreadContext {
>>>> +    struct FrameThreadContext *parent;
>>>> +
>>>> +    pthread_t      thread;
>>>> +    pthread_cond_t input_cond;      ///< Used to wait for a new frame from the main thread.
>>>> +    pthread_cond_t progress_cond;   ///< Used by child threads to wait for progress to change.
>>>> +    pthread_cond_t output_cond;     ///< Used by the main thread to wait for frames to finish.
>>>> +
>>>> +    pthread_mutex_t mutex;          ///< Mutex used to protect the contents of the PerThreadContext.
>>>> +    pthread_mutex_t progress_mutex; ///< Mutex used to protect frame progress values and progress_cond.
>>>> +
>>>> +    AVCodecContext *avctx;          ///< Context used to decode frames passed to this thread.
>>>> +
>>>> +    AVPacket       avpkt;           ///< Input frame (for decoding) or output (for encoding).
>>>> +    int            allocated_buf_size; ///< Size allocated for avpkt.data
>>>> +
>>>> +    AVFrame picture;                ///< Output picture (for decoding) or input (for encoding).
>>>> +    int     got_picture;            ///< The output of got_picture_ptr from the last avcodec_decode_video() call.
>>>> +    int     result;                 ///< The result of the last codec decode/encode() call.
>>> 
>>> 
>>> Here and many other place use the term "picture" a lot for the AVFrame.
>>> I understand the current code is focused on video decoding, but it
>>> would be nice if the framework (especially the public parts) was more
>>> neutral by using the term "frame" instead of "picture" so it will make
>>> sense for audio at some point as well.
>> 
>> I tried to use "frame" for encoded data and "picture" for decoded data.
>> I can use "packet" for the first, but using "frame" for the second
>> made me uncomfortable because field pictures aren't "frames". That's
>> why I stuck to "picture".
>> 
>> I don't mind changing it to "packet" for encoded and "frame" for
>> decoded though, it matches the data types. Is that OK?
> 
> 
> Yes, that sounds better.
> 
> Thanks,
> Justin

Updated patch.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Frame-based-multithreading-framework-using-pthreads.patch
Type: application/octet-stream
Size: 47607 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20110206/c7228f3c/attachment.obj>