class FramePredictor

Computes the target presentation time for

|request.requested_presentation_time|, and a latch-point that is early

enough to apply one update and render a frame, in order to hit the

predicted presentation time.

Both |PredictedTimes.latch_point_time| and

|PredictedTimes.presentation_time| are guaranteed to be after

|request.now|. |PredictedTimes.presentation_time| is guaranteed to be later

than or equal to |request.requested_presentation_time|.

Used by the client to report a measured render duration. The render

duration is the CPU + GPU time it takes to build and render a frame. This

will be considered in subsequent calls to |GetPrediction|.

Used by the client to report a measured update duration. The update

duration is the time it takes to apply a batch of updates. This will be

considered in subsequent calls to |GetPrediction|.

Returns the |PredictionTimes| for a |PredictionRequest|. |frame_preparation_time| is the

a duration long enough to prepare a frame for display (i.e. to update the global scene graph,

optionally do GPU composition, and whatever else is necessary).

Conceptually, starts at |base_vsync_time| and repeatedly adds |vsync_interval| until arriving

at a time that is >= |min_vsync_time|, which is then returned. The actual implementation is

more efficient.

|base_vsync_time| Assumed to be close to a past or future vsync time.

|vsync_interval| The expected time between vsyncs.

|min_vsync_time| The minimum time allowed to return.