I'm working on an iOS music app (written in C++) and my model looks more or less like this:
--Song
----Track
----Track
------Pattern
------Pattern
--------Note
--------Note
--------Note
So basically a Song has multiple Tracks, a Track can have multiple Patterns and a Pattern has multiple Notes. Each one of those things is represented by a class and except for the Song object, they're all stored inside vectors.
Each Note has a "frame" parameter so that I can calculate when a note should be played. For example, if I have 44100 samples / second and the frame for a particular note is 132300 I know that I need that Note at the start of the third second.
My question is how I should represent those notes for best performance? Right now I'm thinking of storing the notes in a vector datamember of each pattern and than loop all the Tracks of the Song, than look the Patterns and than loop the Notes to see which one has a frame datamember that is greater than 132300 and smaller than 176400 (start of 4th second).
As you can tell, that's a lot of loops and a song could be as long as 10 minutes. So I'm wondering if this will be fast enough to calculate all the frames and send them to the buffer on time.
One thing you should remember is that to improve performance, normally memory consumption would have to increase. It is also relevant (and justified) in this case, because I believe you want to store the same data twice, in different ways.
First of all, you should have this basic structure for a song:
map<Track, vector<Pattern>> tracks;
It maps each Track to a vector of Patterns. Map is fine, because you don't care about the order of tracks.
Traversing through Tracks and Patterns should be fast, as their amounts will not be high (I assume). The main performance concern is to loop through thousands of notes. Here's how I suggest to solve it:
First of all, for each Pattern object you should have a vector<Note> as your main data storage. You will write all the changes on the Pattern's contents to this vector<Note> first.
vector<Note> notes;
And for performance considerations, you can have a second way of storing notes:
map<int, vector<Notes>> measures;
This one will map each measure (by its number) in a Pattern to the vector of Notes contained in this measure. Every time data changes in the main notes storage, you will apply the same changes to data in measures. You could also do it only once every time before the playback, or even while playback, in a separate thread.
Of course, you could only store notes in measures, without having to sync two sources of data. But it may be not so convenient to work with when you have to apply mass operations on bunches of notes.
During the playback, before the next measure starts, the following algorithm would happen (roughly):
In every track, find all patterns, for which pattern->startTime <= [current playback second] <= pattern->endTime.
For each pattern, calculate current measure number and get vector<Notes> for the corresponding measure from the measures map.
Now, until the next measure (second?) starts, you only have to loop through current measure's notes.
Just keep those vectors sorted.
During playback, you can just keep a pointer (index) into each vector for the last note player. To search for new notes, you check have to check the following note in each vector, no looping through notes required.
Keep your vectors sorted, and try things out - that is more important and any answer you can receive here.
For all of your questions you should seek to answer then with tests and prototypes, then you will know if you even have a problem. And also while trying it out you will see things that you wouldn't normally see with just the theory alone.
and my model looks more or less like this:
Several critically important concepts are missing from your model:
Tempo.
Dynamics.
Pedal
Instrument
Time signature.
(Optional) Tonality.
Effect (Reverberation/chorus, pitch wheel).
Stereo positioning.
Lyrics.
Chord maps.
Composer information/Title.
Each Note has a "frame" parameter so that I can calculate when a note should be played.
Several critically important concepts are missing from your model:
Articulation.
Aftertouch.
Note duration.
I'd advise to take a look at lilypond. It is typesetting software, but it is also one of the most precise way to represent music in human-readable text format.
My question is how I should represent those notes for best performance?
Put them all into std::map<Timestamp, Note> and find segment you want to playing using lower_bound/upper_bound. Alternatively you could binary search them in flat std::vector as long as data is sorted.
Unless you want to make a "beeper", making music application is much more difficult than you think. I'd strongly recommend to try another project.
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 5 years ago.
Improve this question
Recently, I found myself having to write up some concerns I have about race conditions in an application that is in development (not by me). This will likely be brought to the attention of stakeholders who are non-technical and with whom I do not have a direct line of communication, so my explanation needs to be in written form.
I have already made an attempt at this write-up. I gloss over the technical specifics as best I can, give an example of how a race condition would occur in the application, and describe its impact. I feel I did pretty well, but it's far from perfect.
The problem is, as much as I try to shield the reader from computer science, I have still found it difficult to eliminate phrases like "threads of execution" and "mutual exclusion" without losing correctness and substance. The risk is that, with too much hand-waving, these concerns could be dismissed as a made-up boogeyman.
Anyway, my question to you is this: How would you explain race conditions to a non-technical audience? Would you dare to explain CPU scheduling? Would you invoke the dining philosophers?
You don't have to work within the constraints of my situation (but it would be awesomely helpful if you did).
Company X has $1,000 in the bank. X pays a rent of $2,000 and received a payment of $10,000 for services rendered to company Y. However, due to a race condition, X is in deficit of $1,000 and is now applying for bankruptcy. =(
You might want to explain how the bank handles company X's account in this way: Bank staff A takes the current value of $1,000 and adds $10,000 to it. Bank staff B takes the current value of $1,000 and subtracts $2,000 from it. Bank staff A updates the value to $11,000. Bank staff B updates the value to -$1,000.
I think bank transactions might be a good example, both because it's easy to see that an incorrect result is bad and because race conditions are easy to create in such an environment.
I have $500 on my account.
Someone transfers $200 to me at the same time that I withdraw $50.
Now, if the bank doesn't handle race conditions properly, they will do the following (assuming the transactions are handled manually, of course)
Clerk A will see the request to add $200 to my balance, and note that my balance is currently $500.
Clerk B will see the request to subtract $50 from my balance, and note that my balance is currently $500 (clerk A hasn't yet transferred the money).
Clerk A finishes the paperwork and sets my account balance to $700 (500 + the 200 he was supposed to add).
And then, a minute later (because clerk B just had to grab a cup of coffee), clerk B finishes up the other transaction and sets my balance to $450 (the 500 I had when he checked, minus the 50 he was meant to subtract).
My balance is now $450, when it should have been $650, because of a race condition. The outcome depended on the order in which different parts of the two transactions were performed.
That's the general description of how race conditions are bad. Now say that instead of clerks, we have our application processing two separate tasks at the same time (that's your 'threads of execution'), and just like above, they both read a value, modify the value that they read, and then write it back. One of the modifications may then be lost if this happens in the order shown above.
That should relate it to the specific problems in your app.
I would go for a Dining Philosopher's-esque approach, but depending on my audience, I would try to analogize it to the context of my audience. Are you speaking to business executives? Then analogize it to something like allocate a meeting room or a corporate car or booking a hotel room or whatever. Are you talking to average people? Then the dining philosopher's example is fine, or you can think up a similar situation involving caring for farm animals or sitting in chairs or whatever.
Whether you hijack the dining philosopher's example, or make up your own, definitely use a metaphor.
If you are writing to a non technical audience, you'll want to simplify your explanations and relate it to something they can understand. One explanation taken from the paper Analogies for teaching parallel computing to inexperienced programmers (http://portal.acm.org/citation.cfm?doid=1189136.1189172) explains it in terms of a pen game:
We’re going to play a game called the
Pen Game. The rules are simple: I’m
going to hold a pen in my hand, and
then I’ll say “One, two, three, go.”
When I say “go,” take the pen from my
hand. Whoever gets the pen wins.
Ready? One, two, three, go.
You then ask if the outcome of this game can be predicted in advance. If it can't be predicted, can we guarantee a correct outcome? This should lead to the realization that it's possible to get incorrect results for simultaneous writes to the same piece of memory.
I was going to recommend the dining philosophers, but I see you have already found that one. So, as an alternative, how about using gridlock as an analogy?
Imagine normal traffic driving along the four streets next to a single city block (North ave, South ave, East street and West street). When there are only one or two cars on the road, everything moves smoothly. When there is steady traffic, some cars will have to stop and wait for other cars to move past, but this is a manageable problem. One car stops to wait for another car to go by, and then continues on its merry way.
Now, picture rush-hour traffic at the same location. Let's say that one car driving South on West street can't make it all the way through the intersection at the NorthWest corner of our city block. That car now blocks all of the Westbound cross traffic on North ave. It doesn't take long before a Westbound car tries to make it through the NorthEast corner intersection and gets stuck, blocking all of the Northbound traffic on East street. When this situation makes it all the way around the four intersections, no cars can move! Each one is waiting for the cars in front of it to move ahead, but there is no way for the gridlock to be releived without pulling cars out backwards.
The comparison to computing should be straightforward. Cars are threads or processes, streets and avenues are processors, buffers, or cores. The concept of blocking can be described using traffic lights or stop signs, and the whole thing starts to make intuitive sense, even to non-programmers.
Write a program:
Wait for salary.
Go to shop.
Buy food.
Turn on the plate.
Put food on the plate.
Keep plate for 20 minutes.
Eat.
Go to bed.
Now try to have two threads (you, wife) execute it without syncronization.
You: Wait for salary.
Wife: Go to the shop without money, crash
You: Turn on the plate.
You: Keep plate for 20 minutes.
You: Go to bed.
Wife: Eat at someone else's place.
Wife: Go to bed.
Peter wants to pull out of his driveway. He checks that nothing is in the way of his car, then gets in. His son Frank then hides behind the car. Peter cannot see him and runs him over.
The important thing here is that for a computer, "inspect" and "modify" tend to be two separate actions, so an example where you can't check something when you modify it is a good one.
How about the plain obvious?
A race condition is literally a race between two people.
A company is bidding on a project. Two employees working independently on bids submit them to the customer, but one of the employees has outdated information. Neither employee know that the other is in the process of submitting a bid, therefore depending on who is faster, the first bid may be replaced with the slower employee. This will cause confusion as the bid may have changed over time.
There needs to be communication between the two employees to either work together or stop one of them.
One difficulty in explaining the general concept is that race conditions manifest themselves in a wide variety of situations. If your goal is give your non-technical audience the sense that this is a generic problem type, you should try to offer more than one example.
A picture is worth a 1000 words. Its true. If you draw a timeline and put some entity on it, and show its state changes as time progresses you can demonstrate a race-condition pretty easily in one diagram. It may take a few redos to get the picture just right, but I've always found that drawing it out gets my point across must faster than describing it.
I think it's hard to explain this in a simple way, because thinking about concurrency is inherently hard. The basic idea of a financial transaction might be a good place to start, since people will have some familiarity with them from real life.
In any kind of transaction, you need to make simultaneous entries in two places - debits and credits. If the transaction gets interrupted in the middle by someone else trying to perform another transaction, they will see the wrong balance in one or the other of the accounts.
There's a great example in Structured Concurrent Programming With Operating Systems Applications (as I recall)
In the impoverished country of Bezerkistan, two lines merge onto a single track in a tunnel. There have been collisions and the ruling junta needs a solution.
The issue is that it's mountainous and the engineers are blind. There's very little advance warning of two trains about to collide in the tunnel.
Here's the plan.
Put a big bowl at the juncture.
Give each engineer a little brass monkey.
When you're about to enter the tunnel, you stop your train. You pat around in the bowl to see if a brass monkey is in the bowl.
If there's a monkey, someone else is using the tunnel, so you have to wait until their train is entirely in the tunnel, at which time the conductor gets out of the caboose and grabs the monkey from the bowl.
If there's no monkey, no one else is using the tunnel. So, you can grab your monkey from the engine compartment, put it in the bowl and drive through the tunnel, knowing you have acquired exclusive access to the track. Of course, you stop briefly to allow the conductor to retrieve the brass monkey.
Guess what?
They still had collisions!
Why? What's the situation or sequence of actions that causes this to fail?
That's a race condition.
In a written document, you can explain how the race condition leads to an accident.
In a presentation, you can coach the audience through reasoning about concurrency and locking.
i would use a shared memory bank account example of a data race condition.
explain that the computer does something like: load balance; add 1; store balance;. consider two threads that are modifying your bank account balance (you and your wife are both depositing one dollar at the same time).
if both threads get interuupted after the: load balance; and then resume, you can lose one dollar.
see: http://wasp.cs.washington.edu/atomeclipse/handouts.pdf
As you mentioned, you often need to introduce other concepts (mutual exclusion, threads of execution) to accurately describe race conditions, even in a metaphor. So try defining these terms (or at least getting the idea across) first, using metaphor.
As a simple example, let's use a 4-way intersection (set in a country where you drive on the right). Divide the intersection into 4 quadrants: North-West, North-East, South-East, and South-West. Now call each quadrant a resource, and call each car a thread of execution. These cars only respect traffic systems, and since there are no stop signs or traffic lights at this intersection, the cars barrel right on through without slowing or considering traffic.
You can easily show that simultaneous use of one of these quadrants by more than one car is bad, and results in a car crash. One obvious solution is to install a traffic system. The system ensures that no more than one car is passing through a quadrant at the same time. It can do this intricately, without tying up all the resources. For example, letting cars coming from the South make a left turn to head West (using south-east and north-west quadrants), while letting cars coming from the West make a right turn to head South (using the south-west quadrant). The traffic system is providing mutual exclusion, or preventing simultaneous use (by multiple cars) of a common resource (the quadrant of road in the intersection).
This at least provides the ideas behind these definitions, the idea that simultaneously accessing shared resources can be bad, and that mutual exclusion can solve this problem. After this is established, you'll need to map these to a more appropriate metaphor to show what a race condition is and how it's one of those bad things that results from lack of mutual exclusion for a common resource.
It takes a bit longer, but it grants some familiarity with terms and the big picture before drilling down into a more complex metaphor.
Talking about money to your stakeholders might send them into panic mode especially if they assume they are losing actual money because of this, which is not exactly ideal if the problem does not specifically result in a net loss of profits, so here's a less financially oriented story on how you can explain a race condition to anyone.
This story does not address the concept of deadlock, but the more traditional race condition scenario and consequences.
STORY STARTS HERE:
The Setting: There are 3 cities connected by a railway network. The trains do not have any signs on them indicating which city they are coming from and which city they are going to because they are being used between all 3 cities and the railway network didn't want to deal with the hassle of changing signs all the time. Since the network is small there is no concrete schedule on when trains arrive and leave. The station overseers just get a call from the other city station overseers when a train departs, the overseer takes a note of the time when it has left and since all trains are the same models they drive at the same speed, so when the overseer receives a call from the other cities they announce to the people in the station that: "The next train will be heading to city C". So the people who wish to travel to city C await the train, hop on and merrily ride to city C.
The Problem: But one day, as a train was planning its route from A to B to C, it broke down half-way between A and B. Luckily the technicians are very skilled and would be able to repair the train in a short while. However that same day another train was also planning a different route from C to B to A. The overseer at station B received a call from A that a train is coming, and shortly after received another call from C that another train was also coming. The station overseer then announced to the passengers awaiting in the station: "The first train arriving will be heading to station C, and shortly after the train after that will be heading to station A." As the passengers gathered their luggage and went to their respective platforms. The overseer saw a train coming and redirected the rails to the platform where people were planning to head to city C. Little did they know that the train was actually going to city A instead. The other train, after having fixed its' mechanical problems also arrived at the station and the overseer happily directed it to the platform containing passengers wishing to go to city A. Needless to say none of the passengers arrived where they planned to, all because the overseer assumed that they would arrive in order as usual.
The problem with race conditions and many many computer science constructs is that people are not computers. Every time I explain an algorithm to my students they say "but it doesn't make sense to do it that way", to which I reply "computers don't have common sense, all they have are instructions". That aside, you should explain a race condition as a race, and it makes most sense to let people actually try the race, if they can. That way they can see how things go wrong. But... they are not allowed to use common sense.
So let's assume we have a game where 2 persons fill up stacks of colored blocks in order Red, Orange, Yellow. They have many red, orange and yellow blocks. All stacks need to be exactly three blocks high.
In the first game both try to do this as fast as possible, but they only work on their own stacks.
In the second game they try to work together by allowing themselves to also stack blocks on each other's stacks. However they are not allowed to change the block they have in their hand, and they have to place a planned block.
You can imagine a situation like this occurs in stack 1:
player 1 grabs a red block
player 1 places red block - player 2 grabs an orange block
player 1 grabs an orange block - player 2 places an orange block
player 1 places an orange block
So now we have a stack with two orange blocks. It's obvious that with a human game this would never happen, because people have common sense: they see that the orange block is already placed, and revert their decision to also place an orange block.
Also you can show them this video: https://www.youtube.com/watch?v=TcGwNdbsAbc
Let's use a whiteboard to do a trivial accounting task. We've got $100 on hand - write it on the whiteboard.
Alice has dozens of invoices that add up to $100, so she's going to note that $100, go and add up her list and come back in 5 minutes and write $200 on the board.
Bob's been shopping. He's going to take that number from the whiteboard and go and subtract $50 worth of purchases, and then he's going to write $50 on the board.
If Bob gets back first, we'll see $200 after Alice writes her result. If Alice gets back first we'll see $50, also wrong. What we want to see is $150, and we need to add some precautions somewhere to make that happen.
That should be enough to scaffold a discussion of technical solutions with reasonable intuitions.
For example, a mutex means you lock the door to the room with the whiteboard in it, and make them do their work in there. An optimistic solution means you get them both to check and start over if the number changed while they were away. If you want to talk about deadlocks, you can laugh about Bob calling Alice from inside the locked room to ask her to hurry up.
Send them to Race Condition on Wikipedia.
The first part will make some sense, and the rest (not shown below) will make you look smart since they will assume you understand it.
"A race condition or race hazard is a flaw in a system or process whereby the output and/or result of the process is unexpectedly and critically dependent on the sequence or timing of other events. The term originates with the idea of two signals racing each other to influence the output first."
I think the key point to get across is that its most frequently a timing issue that can be unpredictable because the timing something takes differs from time to time.