[FFmpeg-devel] [PATCH v3 2/3] libavcodec/utils.c: simplify avcodec locking with atomics

Rostislav Pehlivanov atomnuker at gmail.com
Sun Nov 26 04:56:55 EET 2017


On 25 November 2017 at 19:48, Rostislav Pehlivanov <atomnuker at gmail.com>
wrote:

>
>
> On 25 November 2017 at 18:40, James Almer <jamrial at gmail.com> wrote:
>
>> On 11/25/2017 2:01 PM, Rostislav Pehlivanov wrote:
>> > Also makes it more robust than using volatiles.
>> >
>> > Signed-off-by: Rostislav Pehlivanov <atomnuker at gmail.com>
>> > ---
>> >  libavcodec/internal.h |  1 -
>> >  libavcodec/utils.c    | 12 ++++++------
>> >  2 files changed, 6 insertions(+), 7 deletions(-)
>> >
>> > diff --git a/libavcodec/internal.h b/libavcodec/internal.h
>> > index d3310b6afe..1c54966f37 100644
>> > --- a/libavcodec/internal.h
>> > +++ b/libavcodec/internal.h
>> > @@ -246,7 +246,6 @@ int ff_init_buffer_info(AVCodecContext *s, AVFrame
>> *frame);
>> >
>> >  void ff_color_frame(AVFrame *frame, const int color[4]);
>> >
>> > -extern volatile int ff_avcodec_locked;
>> >  int ff_lock_avcodec(AVCodecContext *log_ctx, const AVCodec *codec);
>> >  int ff_unlock_avcodec(const AVCodec *codec);
>> >
>> > diff --git a/libavcodec/utils.c b/libavcodec/utils.c
>> > index 3a0f3c11f5..96bc9ff4a4 100644
>> > --- a/libavcodec/utils.c
>> > +++ b/libavcodec/utils.c
>> > @@ -114,7 +114,7 @@ static int (*lockmgr_cb)(void **mutex, enum
>> AVLockOp op) = NULL;
>> >  #endif
>> >
>> >
>> > -volatile int ff_avcodec_locked;
>> > +static atomic_bool ff_avcodec_locked;
>> >  static atomic_int entangled_thread_counter = ATOMIC_VAR_INIT(0);
>> >  static void *codec_mutex;
>> >  static void *avformat_mutex;
>> > @@ -1937,6 +1937,7 @@ int av_lockmgr_register(int (*cb)(void **mutex,
>> enum AVLockOp op))
>> >
>> >  int ff_lock_avcodec(AVCodecContext *log_ctx, const AVCodec *codec)
>> >  {
>> > +    _Bool exp = 1;
>> >      if (codec->caps_internal & FF_CODEC_CAP_INIT_THREADSAFE ||
>> !codec->init)
>> >          return 0;
>> >
>> > @@ -1952,22 +1953,21 @@ int ff_lock_avcodec(AVCodecContext *log_ctx,
>> const AVCodec *codec)
>> >                 atomic_load(&entangled_thread_counter));
>> >          if (!lockmgr_cb)
>> >              av_log(log_ctx, AV_LOG_ERROR, "No lock manager is set,
>> please see av_lockmgr_register()\n");
>> > -        ff_avcodec_locked = 1;
>> > +        atomic_store(&ff_avcodec_locked, 1);
>> >          ff_unlock_avcodec(codec);
>> >          return AVERROR(EINVAL);
>> >      }
>> > -    av_assert0(!ff_avcodec_locked);
>> > -    ff_avcodec_locked = 1;
>> > +    av_assert0(atomic_compare_exchange_strong(&ff_avcodec_locked,
>> &exp, 1));
>>
>> _Bool atomic_compare_exchange_strong( volatile A* obj,
>>                                       C* expected, C desired );
>>
>> "Atomically compares the value pointed to by obj with the value pointed
>> to by expected, and if those are equal, replaces the former with desired
>> (performs read-modify-write operation).
>> Return value: The result of the comparison: true if *obj was equal to
>> *exp, false otherwise."
>>
>> exp should be 0. You need to assert that ff_avcodec_locked == 0, then
>> set it to 1.
>>
>
> The whole experssion seems fine to me as-is and works as expected.
>
>
>>
>> >      return 0;
>> >  }
>> >
>> >  int ff_unlock_avcodec(const AVCodec *codec)
>> >  {
>> > +    _Bool exp = 0;
>> >      if (codec->caps_internal & FF_CODEC_CAP_INIT_THREADSAFE ||
>> !codec->init)
>> >          return 0;
>> >
>> > -    av_assert0(ff_avcodec_locked);
>> > -    ff_avcodec_locked = 0;
>> > +    av_assert0(atomic_compare_exchange_strong(&ff_avcodec_locked,
>> &exp, 0));
>>
>> And here exp should be 1.
>>
>> >      atomic_fetch_add(&entangled_thread_counter, -1);
>> >      if (lockmgr_cb) {
>> >          if ((*lockmgr_cb)(&codec_mutex, AV_LOCK_RELEASE))
>> >
>>
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel at ffmpeg.org
>> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>
>
> Seems like I sent an older version of the patch (had to rebase to fix
> older commit), attached patch fixed both.
>

Pushed, thanks


More information about the ffmpeg-devel mailing list