purpose ------- A proposal for proper syncing and clocking of a pipeline. Requirements ------------ - elements should be able to get the time and wait for a specific time - some elements should be able to control and adjust the clock. (clock master) - the application should be able to provide another clock. Clocks ------ A clock extends the abstract GstClock class. gst_clock_get_time should always report an equal or increasing value with each succesive call. Clock providers --------------- Clock providers call gst_element_provides_clock (element, clock) in their _init function. This will set some element flags. This also means that a clock provider cannot stop being a clock provider. Clock providers will update the clock at specific intervals. get_resolution/set_resolution can be used to control the resolution. When a clock provider is not going to update the clock anymore it should make sure elements still blocked on the clock get unblocked at the right time. This can be done by converting all blocking waits to a select call. All further waits on the clock should be estimated (using gettimeofday for example) Clock receivers --------------- An element needing a clock must implement the element receive_clock method. It has to use the clock received in this function or if the clock == NULL it should not do any waiting at all. Clocks and the scheduler ------------------------ The scheduler knows about the clocks, else a clock provider and receiver in the same scheduler could cause a deadlock. Only one clock provider is allowed per scheduler. multiple clock receivers are allowed per scheduler. Specifying clocks ----------------- on every bin with a scheduler, gst_bin_use_clock (bin, clock) can be used to force the use of this specific clock for all the elements in this bin. gst_bin_use_clock (bin, NULL) to disable all clocking for this bin. gst_bin_auto_clock (bin) to use the default algorithm to find a clock. bins that get a use clock set a flag and store the clock, they also call set_clock on all of their children. When other children are added, they all get the stored clock. Clock distribution ------------------ clock providers and receivers are collected in the bins on element add/remove, recursing bins if needed. the toplevel bin with a scheduler selects the global clock and calls set_clock on itself. This happens in the NULL->READY state. bins dispatch the set_clock to to all of the reveivers. Bins with a scheduler and another use_clock do nothing. Bins with a scheduler also notify the scheduler of the clock. Clock usage ----------- when an element wants to wait for a specific time, if calls gst_element_clock_wait (elements, clock, time). The call is dispatched to the scheduler of the element which can use the owner field of the clock to check if the elements are in the same scheduler. For elements waiting for a clock provided by an element in another scheduler there is no problem. When provider and receiver are in the same scheduler, a deadlock can occur since the scheduler will block on the wait without being able to schedule the provider again to update the clock. A solution would be to call the async notify of the clock and schedule some other element (the provider?). We probably need an event based scheduler for this. When the async event arrives we can reschedule the receiver and continue. The current scheduler will assert on this condition for now. Changing clocks --------------- The clock can only be changed when the bin is in the PAUSED state. State Changes ------------- When the pipeline is PAUSED, the clock is stopped with gst_clock_enable (clock, FALSE). The clock should unblock all waiting elements ASAP and return GST_CLOCK_STOP in the wait. Elements waiting for a clock and receiving the STOP should process the last buffer ASAP and break out of their loop. When the pipeline is brought to the READY state, the clock is set to 0. When the clock is enabled again, it should start counting from where it was last disabled. NULL->READY : distribute clock, clock_reset, READY->PAUSED : clock_activate (FALSE) PAUSED->PLAYING : clock_activate (TRUE) PLAYING->PAUSED : clock_activate (FALSE); PAUSED->READY : clock_reset READY->NULL : delete clock; automatic Clock selection ------------------------- Select a random clock from the Src elements. if no Src elements exist with a clock, select a random clock from the Sink elements. else use a default System Clock. Src elements with a clock are prefered because they usualy provide live audio/video. Issues ------ ossrc ! osssink can cause clock drift if osssink doesn't process the bytes at the same rate osssrc provides them (different hardware). Things will stutter and cracle if this is the case. QoS and a resampler could solve this. Use Cases --------- -- queue ! mpeg2dec ! videosink / filesrc ! mpegdemux \ -- queue ! mad ! osssink videosink is a receiver osssink is a provider osssink is selected as a clock provider since it is a Sink. The global pipeline distributes the clock to videosink and osssink. osssink sees that it receives its own clock. osssink uses the OSS ioctls to determine the number of bytes processed by the hardware. using the audio rate it can figure out the exact time and updates its clock with a resolution that matches the resolution as closely as possible. videosink blocks on the clock. with each update of the clock videosink is unblocked if the current time >= wait time and shows the frame. when osssink is PAUSED, the clock will not be updated anymore. osssink instructs its clock to convert all requests to select() calls. When it is set to PLAYING again, it resumes normal operation.