NSTimer Accuracy - objective-c

I am currently working on an app that requires a sound file to be played at exact intervals, which are variable in duration.
I seem to remember being told that NSTimer simply places an operation onto the stack after the specified duration, rather than the operation being run after the specified duration. This would mean that if there were lots of other operations on the stack before it, it would not be called on time.
I would like to know if this is indeed correct, and if so, is there any way of getting an operation to be guaranteed to run after a specified duration.
Before any makes the comment that the sound files may be delayed the first time, they are already pre-loaded to avoid this.

In the first paragraph of Apple's NSTimer documentation you'll find this:
A timer is not a real-time mechanism; it fires only when one of the
run loop modes to which the timer has been added is running and able
to check if the timer’s firing time has passed. Because of the various
input sources a typical run loop manages, the effective resolution of
the time interval for a timer is limited to on the order of 50-100
milliseconds. If a timer’s firing time occurs during a long callout or
while the run loop is in a mode that is not monitoring the timer, the
timer does not fire until the next time the run loop checks the timer.
Therefore, the actual time at which the timer fires potentially can be
a significant period of time after the scheduled firing time.
So you will definitely run (or play your sound, in your case) after your scheduled firing time, but not necessarily exactly at your time.

Related

How to determine what's blocking the main thread

So I restructured a central part in my Cocoa application (I really had to!) and I am running into issues since then.
Quick outline: my application controls the playback of QuickTime movies so that they are in sync with external timecode.
Thus, external timecode arrives on a CoreMIDI callback thread and gets posted to the application about 25 times per sec. The sync is then checked and adjusted if it needs to be.
All this checking and adjusting is done on the main thread.
Even if I put all the processing on a background thread it would be a ton of work as I'm currently using a lot of GCD blocks and I would need to rewrite a lot of functions so that they can be called from NSThread. So I would like to make sure first if it will solve my problem.
The problem
My Core MIDI callback is always called in time, but the GCD block that is dispatched to the main queue is sometimes blocked for up to 500 ms. Understandable that adjusting the sync does not quite work if that happens. I couldn't find a reason for it, so I'm guessing that I'm doing something that blocks the main thread.
I'm familiar with Instruments, but I couldn't find the right mode to see what keeps my messages from being processed in time.
I would appreciate if anyone could help.
Don't know what I can do about it.
Thanks in advance!
Watchdog
You can use watch dog that stop when the main thread stopped for time
https://github.com/wojteklu/Watchdog
you can install it using cocoapod
pod 'Watchdog'
You may be blocking the main thread or you might be flooding it with events.
I would suggest three things:
Grab a timestamp for when the timecode arrives in the CoreMIDI callback thread (see mach_absolute_time(). Then grab the current time when your main thread work is being done. You can then adjust accordingly based on how much time has elapsed between posting to the main thread and it actually being processed.
create some kind of coalescing mechanism such that when your main thread is blocked, interim timecode events (that are now out of date) are tossed. This can be as simple as a global NSUInteger that is incremented every time an event is received. The block dispatched to the main queue could capture the current value on creation, then check it when it is processed. If it differs by more than N (N for you to determine), then toss the event because more are in flight.
consider not sending an event to the main thread for every timecode notification. 25 adjustments per second is a lot of work. If processing only 5 per second yields a "good enough" perceptual experience, then that is an awful lot of work saved.
In general, instrumenting the main event loop is a bit tricky. The CPU profiler in Instruments can be quite helpful. It may come as a surprise, but so can the Allocations instrument. In particular, you can use the Allocations instrument to measure memory throughput. If there are tons of transient (short lived) allocations, it'll chew up a ton of CPU time doing all those allocations/deallocations.

More Accurate StopWatch (VB.NET)

Is there any sort of "sleep" method that is more accurate than the stopwatch? Or, is there a way to make the stopwatch class more accurate? It doesn't have to be in .NET, it can be in c++, but whatever language it is in has to have exactly 1ms accuracy; I don't need more then that. Say if I want my program to "sleep" for 300ms, I would like it to sleep for 300ms at least most of the time.
Currently I use:
Dim StopWatch As New StopWatch
StopWatch.Start
Do
Loop Until StopWatch.ELapsed.Milliseconds >= 300
StopWatch.Stop
My results running it 5 times were: 306, 305, 315, 327, 304.
It stayed like that if I ran it more.
I put my thread and process priority on "Realtime" / "High".
The Stopwatch class has a property IsHighResolution. If it returns ´true´ you are using the High Performance Event Timer (HPET) - availability depends on hardware and OS. Using this, you can measure times very accurate. BUT! Windows (as usual Linuxes) is NOT a realtime OS, but uses preemptive multitasking. Whenever the OS thinks, that it needs to, it will put your current thread on hold to do other work and after some time, it will return to your thread and let it continue. If this switch happens somewhere inside your loop, you still measure the correct time, but it contains an amount of inactivty time.
Since a time slice under Windows is something between 15 and 30 ms, you(r thread) might be suspended after 299 ms and 15-30 ms later you will get back. And that's the effect you see. The Stopwatch IS accurate. It just measures stuff you didn't expect.
How to overcome: You can't. As said: Windoes IS NOT a realtime OS! Even if you assign priority "realtime" to your process.
What you are seeing is completely normal. Delay will never be exactly 300ms, it will always be more than that. Sleep itself is accurate, but the actual delay depends on your operating system, and other processes running in parallel to yours.
If you want a more accurate timer, you need to use the current date and time as a reference. Here is a simple equation that you can run every millisecond:
currentTime - startTime = elapsedTime
...where currentTime is System.DateTime.Now, startTime is the time that the timer was started, and elapsedTime is a System.DateTime.TimeSpan.
For more details on how to do this, check out the source of a program I made in VB.Net, E-Tech Timer: http://etechtimer.codeplex.com

iOS: recreating countdown timers after clock change?

I'm working on an iOS app that uses an NSTimer for a countdown. This is prone to user tampering: if, for example, the user switches out of the app, closes the app manually, changes the device clock, and comes back in, the timer will have to be recreated. Another scenario: the user locks the device, it goes into low-power mode (which requires timers to be recreated), and the clock auto-sets before the game is opened again. If that happens, I won't have an accurate way of determining how much time has passed since the app was closed, since the device clock has changed.
Tl;dr: countdown timers sometimes have to be recreated after a device clock change. How is this problem usually handled?
Any time you're relying on the system clock for accurate timing you're going to have troubles, even if the user isn't deliberately tampering with the clock. Typically clock drift is corrected by slightly increasing or decreasing the length of a second to allow the clock to drift back into alignment over a period of minutes. If you need accurate timing, you can either use something like mach_absolute_time() which is related to the system uptime rather than the system clock, or you can use Grand Central Dispatch. The dispatch_after() function takes a dispatch_time_t which can either be expressed using wall time (e.g. system clock) or as an offset against DISPATCH_TIME_NOW (which ignores wall clock).
For future reference, in regard to different systems of timekeeping in OSX (and consequently iOS):
One way to measure the speed of any operation, including launch times,
is to use system routines to get the current time at the beginning and
end of the operation. Once you have the two time values, you can take
the difference and log the results.
The advantage of this technique is that it lets you measure the
duration of specific blocks of code. Mac OS X includes several
different ways to get the current time:
mach_absolute_time reads the CPU time base register and is the
basis for other time measurement functions.
The Core Services UpTime function provides nanosecond resolution
for time measurements.
The BSD gettimeofday function (declared in <sys/time.h>) provides
microsecond resolution. (Note, this function incurs some overhead but
is still accurate for most uses.)
In Cocoa, you can create an NSDate object with the current time at
the beginning of the operation and then use the
timeIntervalSinceDate: method to get the time difference.
Source: Launch Time Performance Guidelines, "Gathering Launch Time Metrics".

Calculating number of seconds between two points in time, in Cocoa, even when system clock has changed mid-way

I'm writing a Cocoa OS X (Leopard 10.5+) end-user program that's using timestamps to calculate statistics for how long something is being displayed on the screen. Time is calculated periodically while the program runs using a repeating NSTimer. [NSDate date] is used to capture timestamps, Start and Finish. Calculating the difference between the two dates in seconds is trivial.
A problem occurs if an end-user or ntp changes the system clock. [NSDate date] relies on the system clock, so if it's changed, the Finish variable will be skewed relative to the Start, messing up the time calculation significantly. My question:
1. How can I accurately calculate the time between Start and Finish, in seconds, even when the system clock is changed mid-way?
I'm thinking that I need a non-changing reference point in time so I can calculate how many seconds has passed since then. For example, system uptime. 10.6 has - (NSTimeInterval)systemUptime, part of NSProcessInfo, which provides system uptime. However, this won't work as my app must work in 10.5.
I've tried creating a time counter using NSTimer, but this isn't accurate. NSTimer has several different run modes and can only run one at a time. NSTimer (by default) is put into the default run mode. If a user starts manipulating the UI for a long enough time, this will enter NSEventTrackingRunLoopMode and skip over the default run mode, which can lead to NSTimer firings being skipped, making it an inaccurate way of counting seconds.
I've also thought about creating a separate thread (NSRunLoop) to run a NSTimer second-counter, keeping it away from UI interactions. But I'm very new to multi-threading and I'd like to stay away from that if possible. Also, I'm not sure if this would work accurately in the event the CPU gets pegged by another application (Photoshop rendering a large image, etc...), causing my NSRunLoop to be put on hold for long enough to mess up its NSTimer.
I appreciate any help. :)
Depending on what's driving this code, you have 2 choices:
For absolute precision, use mach_absolute_time(). It will give the time interval exactly between the points at which you called the function.
But in a GUI app, this is often actually undesirable. Instead, you want the time difference between the events that started and finished your duration. If so, compare [[NSApp currentEvent] timestamp]
Okay so this is a long shot, but you could try implementing something sort of like NSSystemClockDidChangeNotification available in Snow Leopard.
So bear with me here, because this is a strange idea and is definitely non-derterministic. But what if you had a watchdog thread running through the duration of your program? This thread would, every n seconds, read the system time and store it. For the sake of argument, let's just make it 5 seconds. So every 5 seconds, it compares the previous reading to the current system time. If there's a "big enough" difference ("big enough" would need to definitely be greater than 5, but not too much greater, to account for the non-determinism of process scheduling and thread prioritization), post a notification that there has been a significant time change. You would need to play around with fuzzing the value that constitutes "big enough" (or small enough, if the clock was reset to an earlier time) for your accuracy needs.
I know this is kind of hacky, but barring any other solution, what do you think? Might that, or something like that, solve your issue?
Edit
Okay so you modified your original question to say that you'd rather not use a watchdog thread because you are new to multithreading. I understand the fear of doing something a bit more advanced than you are comfortable with, but this might end up being the only solution. In that case, you might have a bit of reading to do. =)
And yeah, I know that something such as Photoshop pegging the crap out of the processor is a problem. Another (even more complicated) solution would be to, instead of having a watchdog thread, have a separate watchdog process that has top priority so it is a bit more immune to processor pegging. But again, this is getting really complicated.
Final Edit
I'm going to leave all my other ideas above for completeness' sake, but it seems that using the system's uptime will also be a valid way to deal with this. Since [[NSProcessInfo processInfo] systemUptime] only works in 10.6+, you can just call mach_absolute_time(). To get access to that function, just #include <mach/mach_time.h>. That should be the same value as returned by NSProcessInfo.
I figured out a way to do this using the UpTime() C function, provided in <CoreServices/CoreServices.h>. This returns Absolute Time (CPU-specific), which can easily be converted into Duration Time (milliseconds, or nanoseconds). Details here: http://www.meandmark.com/timingpart1.html (look under part 3 for UpTime)
I couldn't get mach_absolute_time() to work properly, likely due to my lack of knowledge on it, and not being able to find much documentation on the web about it. It appears to grab the same time as UpTime(), but converting it into a double left me dumbfounded.
[[NSApp currentEvent] timestamp] did work, but only if the application was receiving NSEvents. If the application went into the foreground, it wouldn't receive events, and [[NSApp currentEvent] timestamp] would simply continue to return the same old timestamp again and again in an NSTimer firing method, until the end-user decided to interact with the app again.
Thanks for all your help Marc and Mike! You both definitely sent me in the right direction leading to the answer. :)

NSTimer interval less than sleep command in fired code

I have a snippet of code I want to execute repeatedly, but with the ability to pause and resume it. To do this, I have utilised an NSTimer, which I can stop and start as required.
Within the snippet, I use a sleep command to wait for something to update (0.3 seconds). The timer is firing every 0.5 seconds.
What would be ideal is to keep the stop and start functionality, and be firing every 0.3 seconds, but to not have to explicitly say I want it to fire every x seconds. The 0.5 is completely arbitrary, and is just set to be > 0.3.
If I set the timer to fire every 0.01 seconds, but keep the sleep command within the code fired to 0.3 seconds, will I get the desired behaviour? Or will the timer back up all the events, with unexpected results (for example multiple firings after I have stopped it)? This way I could make the 0.3 sec sleep a variable, and not have to change the timer whenever I increase it over 0.5.
Or is there a better way to get this functionality?
One of the biggest problems with this I see is that NSTimer will be consuming time on whatever thread it's registered with.
In most cases for me this is my main thread, sleeping inside of your main thread would be a bad thing as far as I can see.
A few alternative designs that may be better for you.
Don't sleep. Have the selector called by the timer do a 'am I paused' check and exit early if it is.
Invalidate the NSTimer when pausing and recreate it when unpausing (Note: you can't just reschedule the old timer).
Spawn a background thread with it's own RunLoop or even your own while loop that just handles the rescheduling with sleeping.
Those are also roughly in the order I'd do them, as No. 1 seems the cleanest to me, followed by No.2. No.3 while viable could introduce a lot of potential nastiness (threading issues, clean shutdowns, etc)