Related
I need to be extremely concerned with speed/latency in my current multi-threaded project.
Cache access is something I'm trying to understand better. And I'm not clear on how lock-free queues (such as the boost::lockfree::spsc_queue) access/use memory on a cache level.
I've seen queues used where the pointer of a large object that needs to be operated on by the consumer core is pushed into the queue.
If the consumer core pops an element from the queue, I presume that means the element (a pointer in this case) is already loaded into the consumer core's L2 and L1 cache. But to access the element, does it not need to access the pointer itself by finding and loading the element either from either the L3 cache or across the interconnect (if the other thread is on a different cpu socket)? If so, would it maybe be better to simply send a copy of the object that could be disposed of by the consumer?
Thank you.
C++ principally a pay-for-what-you-need eco-system.
Any regular queue will let you choose the storage semantics (by value or by reference).
However, this time you ordered something special: you ordered a lock free queue.
In order to be lock free, it must be able to perform all the observable modifying operations as atomic operations. This naturally restricts the types that can be used in these operations directly.
You might doubt whether it's even possible to have a value-type that exceeds the system's native register size (say, int64_t).
Good question.
Enter Ringbuffers
Indeed, any node based container would just require pointer swaps for all modifying operations, which is trivially made atomic on all modern architectures.
But does anything that involves copying multiple distinct memory areas, in non-atomic sequence, really pose an unsolvable problem?
No. Imagine a flat array of POD data items. Now, if you treat the array as a circular buffer, one would just have to maintain the index of the buffer front and end positions atomically. The container could, at leisure update in internal 'dirty front index' while it copies ahead of the external front. (The copy can use relaxed memory ordering). Only as soon as the whole copy is known to have completed, the external front index is updated. This update needs to be in acq_rel/cst memory order[1].
As long as the container is able to guard the invariant that the front never fully wraps around and reaches back, this is a sweet deal. I think this idea was popularized in the Disruptor Library (of LMAX fame). You get mechanical resonance from
linear memory access patterns while reading/writing
even better if you can make the record size aligned with (a multiple) physical cache lines
all the data is local unless the POD contains raw references outside that record
How Does Boost's spsc_queue Actually Do This?
Yes, spqc_queue stores the raw element values in a contiguous aligned block of memory: (e.g. from compile_time_sized_ringbuffer which underlies spsc_queue with statically supplied maximum capacity:)
typedef typename boost::aligned_storage<max_size * sizeof(T),
boost::alignment_of<T>::value
>::type storage_type;
storage_type storage_;
T * data()
{
return static_cast<T*>(storage_.address());
}
(The element type T need not even be POD, but it needs to be both default-constructible and copyable).
Yes, the read and write pointers are atomic integral values. Note that the boost devs have taken care to apply enough padding to avoid False Sharing on the cache line for the reading/writing indices: (from ringbuffer_base):
static const int padding_size = BOOST_LOCKFREE_CACHELINE_BYTES - sizeof(size_t);
atomic<size_t> write_index_;
char padding1[padding_size]; /* force read_index and write_index to different cache lines */
atomic<size_t> read_index_;
In fact, as you can see, there are only the "internal" index on either read or write side. This is possible because there's only one writing thread and also only one reading thread, which means that there could only be more space at the end of write operation than anticipated.
Several other optimizations are present:
branch prediction hints for platforms that support it (unlikely())
it's possible to push/pop a range of elements at once. This should improve throughput in case you need to siphon from one buffer/ringbuffer into another, especially if the raw element size is not equal to (a whole multiple of) a cacheline
use of std::unitialized_copy where possible
The calling of trivial constructors/destructors will be optimized out at instantiation time
the unitialized_copy will be optimized into memcpy on all major standard library implementations (meaning that e.g. SSE instructions will be employed if your architecture supports it)
All in all, we see a best-in-class possible idea for a ringbuffer
What To Use
Boost has given you all the options. You can elect to make your element type a pointer to your message type. However, as you already raised in your question, this level of indirection reduces locality of reference and might not be optimal.
On the other hand, storing the complete message type in the element type could become expensive if copying is expensive. At the very least try to make the element type fit nicely into a cache line (typically 64 bytes on Intel).
So in practice you might consider storing frequently used data right there in the value, and referencing the less-of-used data using a pointer (the cost of the pointer will be low unless it's traversed).
If you need that "attachment" model, consider using a custom allocator for the referred-to data so you can achieve memory access patterns there too.
Let your profiler guide you.
[1] I suppose say for spsc acq_rel should work, but I'm a bit rusty on the details. As a rule, I make it a point not to write lock-free code myself. I recommend anyone else to follow my example :)
I'm pretty sure this is a silly newbie question but I didn't know it so I had to ask...
Why do we use data structures, like Linked List, Binary Search Tree, etc? (when no dynamic allocation is needed)
I mean: wouldn't it be faster if we kept a single variable for a single object? Wouldn't that speed up access time? Eg: BST possibly has to run through some pointers first before it gets to the actual data.
Except for when dynamic allocation is needed, is there a reason to use them?
Eg: using linked list/ BST / std::vector in a situation where a simple (non-dynamic) array could be used.
Each thing you are storing is being kept in it's own variable (or storage location). Data structures apply organization to your data. Imagine if you had 10,000 things you were trying to track. You could store them in 10,000 separate variables. If you did that, then you'd always be limited to 10,000 different things. If you wanted more, you'd have to modify your program and recompile it each time you wanted to increase the number. You might also have to modify the code to change the way in which the calculations are done if the order of the items changes because the new one is introduced in the middle.
Using data structures, from simple arrays to more complex trees, hash tables, or custom data structures, allows your code to both be more organized and extensible. Using an array, which can either be created to hold the required number of elements or extended to hold more after it's first created keeps you from having to rewrite your code each time the number of data items changes. Using an appropriate data structure allows you to design algorithms based on the relationships between the data elements rather than some fixed ordering, giving you more flexibility.
A simple analogy might help to understand. You could, for example, organize all of your important papers by putting each of them into separate filing cabinet. If you did that you'd have to memorize (i.e., hard-code) the cabinet in which each item can be found in order to use them effectively. Alternatively, you could store each in the same filing cabinet (like a generic array). This is better in that they're all in one place, but still not optimum, since you have to search through them all each time you want to find one. Better yet would be to organize them by subject, putting like subjects in the same file folder (separate arrays, different structures). That way you can look for the file folder for the correct subject, then find the item you're looking for in it. Depending on your needs you can use different filing methods (data structures/algorithms) to better organize your information for it's intended use.
I'll also note that there are times when it does make sense to use individual variables for each data item you are using. Frequently there is a mixture of individual variables and more complex structures, using the appropriate method depending on the use of the particular item. For example, you might store the sum of a collection of integers in a variable while the integers themselves are stored in an array. A program would need to be pretty simple though before the introduction of data structures wouldn't be appropriate.
Sorry, but you didn't just find a great new way of doing things ;) There are several huge problems with this approach.
How could this be done without requring programmers to massively (and nontrivially) rewrite tons of code as soon as the number of allowed items changes? Even when you have to fix your data structure sizes at compile time (e.g. arrays in C), you can use a constant. Then, changing a single constant and recompiling is sufficent for changes to that size (if the code was written with this in mind). With your approach, we'd have to type hundreds or even thousands of lines every time some size changes. Not to mention that all this code would be incredibly hard to read, write, maintain and verify. The old truism "more lines of code = more space for bugs" is taken up to eleven in such a setting.
Then there's the fact that the number is almost never set in stone. Even when it is a compile time constant, changes are still likely. Writing hundreds of lines of code for a minor (if it exists at all) performance gain is hardly ever worth it. This goes thrice if you'd have to do the same amount of work again every time you want to change something. Not to mention that it isn't possible at all once there is any remotely dynamic component in the size of the data structures. That is to say, it's very rarely possible.
Also consider the concept of implicit and succinct data structures. If you use a set of hard-coded variables instead of abstracting over the size, you still got a data structure. You merely made it implicit, unrolled the algorithms operating on it, and set its size in stone. Philosophically, you changed nothing.
But surely it has a performance benefit? Well, possible, although it will be tiny. But it isn't guaranteed to be there. You'd save some space on data, but code size would explode. And as everyone informed about inlining should know, small code sizes are very useful for performance to allow the code to be in the cache. Also, argument passing would result in excessive copying unless you'd figure out a trick to derive the location of most variables from a few pointers. Needless to say, this would be nonportable, very tricky to get right even on a single platform, and liable to being broken by any change to the code or the compiler invocation.
Finally, note that a weaker form is sometimes done. The Wikipedia page on implicit and succinct data structures has some examples. On a smaller scale, some data structures store much data in one place, such that it can be accessed with less pointer chasing and is more likely to be in the cache (e.g. cache-aware and cache-oblivious data structures). It's just not viable for 99% of all code and taking it to the extreme adds only a tiny, if any, benefit.
The main benefit to datastructures, in my opinion, is that you are relationally grouping them. For instance, instead of having 10 separate variables of class MyClass, you can have a datastructure that groups them all. This grouping allows for certain operations to be performed because they are structured together.
Not to mention, having datastructures can potentially enforce type security, which is powerful and necessary in many cases.
And last but not least, what would you rather do?
string string1 = "string1";
string string2 = "string2";
string string3 = "string3";
string string4 = "string4";
string string5 = "string5";
Console.WriteLine(string1);
Console.WriteLine(string2);
Console.WriteLine(string3);
Console.WriteLine(string4);
Console.WriteLine(string5);
Or...
List<string> myStringList = new List<string>() { "string1", "string2", "string3", "string4", "string5" };
foreach (string s in myStringList)
Console.WriteLine(s);
given:
-record(foo, {a, b, c}).
I do something like this:
Thing = #foo{a={1,2}, b={3,4}, c={5,6}},
Thing1 = Thing#foo{a={7,8}}.
From a semantic view, Thing and Thing1 are unique entities. However, from a language implementation standpoint, making a full copy of Thing to generate Thing1 would be intensely wasteful. For example, if the record were a megabyte in size and I made a thousand "copies," each modifying a couple of bytes, I've just burned a gigabyte. If the internal structure kept track of a representation of the parent structure and each derivative marked up that parent in a way that indicated its own change but preserved everyone elses' versions, the derivates could be created with a minimum of memory overhead.
My question is this: is erlang doing anything clever - internally - to keep the overhead of the usual erlang scribble;
Thing = #ridiculously_large_record,
Thing1 = make_modified_copy(Thing),
Thing2 = make_modified_copy(Thing1),
Thing3 = make_modified_copy(Thing2),
Thing4 = make_modified_copy(Thing3),
Thing5 = make_modified_copy(Thing4)
...to a minimum?
I ask because there would be a number of changes to the way that I did cross-process communications if this were the case.
The exact workings of the garbage collection and memory allocation is only known to a few. Thankfully, they are very happy to share their knowledge and the following is based on what I have learnt from the erlang-questions mailing list and by discussing with OTP developers.
When messaging between processes, the content is always copied as there is no shared heap between processes. The only exception is binaries bigger than 64 bytes, where only a reference is copied.
When executing code in one process, only parts are updated. Let's analyze tuples, as that is the example you provided.
A tuple is actually a structure that keeps references to the actual data somewhere on the heap (except for small integers and maybe one more data type which I can't remember). When you update a tuple, using for example setelement/3, a new tuple is created with the given element replaced, however for all other elements only the reference is copied. There is one exception which I have never been able to take advantage of.
The garbage collector keeps track of each tuple and understands when it is safe to reclaim any tuple that is no longer used. It might be that the data referenced by the tuple is still in use, in which case the data itself is not collected.
As always, Erlang gives you some tools to understand exactly what is going on. The efficiency guide details how to use erts_debug:size/1 and erts_debug:flat_size/1 to understand the size of the data structure when used internally in a process and when copied. The trace tools also allows you to understand when, what and how much was garbage collected.
The record foo is of arity four (holding four words), but the whole structure is 14 words in size. Any immediate (pids, ports, small integers, atoms, catch and nil) can be stored directly in the tuples array. Any other term which can't fit into a word, such as other tuples, are not stored directly but referenced by boxed pointers (a boxed pointer is an erlang term with a forwarding address to the real eterm ... just internals).
In your case a new tuple of same arity is created and the atom foo and all the pointers are copied from the previous tuple except for index two, a, which points to the new tuple {7,8} which constitutes 3 words. In all 5 + 3 new words are created on the heap and only 3 words are copied from the old tuple the other 9 words are not touched.
Excessively large tuples are not recommended. When updating a tuple, the whole tuple, i.e the array and not the deep content, needs to copied and then updated in other to preserve a persistent data structure. This will also generate increased garbage, forcing the garbage collector to heat up which also hurts performance. The dict and array modules avoids using large tuples for this reason and have a shallow tuple tree instead.
I can definitely verify what people have already pointed out:
a record is just a tuple with the record name as the first element and all the fields just the following tuple element
when an element of a tuple is changed, updating a field in a record in your case, only the top level tuple is new, all the elements are just reused
This works just because we have immutable data. So in your example each time you update a value in a #foo record none of the data in the elements are copied and only a new 4-element tuple (5 words) is created. Erlang will never does a deep copy in this type of operation or when passing arguments in function calls.
In conclusion:
Thing = #foo{a={1,2}, b={3,4}, c={5,6}},
Thing1 = Thing#foo{a={7,8}}.
Here, if Thing is not used again, it will probably be updated in place and copying of the tuple will be avoided, as the Efficiency Guide says. (tuple and record syntax is complied into something like setelement, I think)
Thing = #ridiculously_large_record,
Thing1 = make_modified_copy(Thing),
Thing2 = make_modified_copy(Thing1),
...
Here the tuples are actually copied every time.
I guess that it would be theoretically possible make an interesting optimization to this. If the compiler could perform escape analysis on the return value of make_modified_copy and detect that the only reference to it is the one returned, in could save this information about the function. When it encounter a call the that function it would know that it is safe to modify the return value in place.
This would only be possible to do on inter module calls, because of the code replace feature.
Maybe one day we will have it.
Warning! possibly a very dumb question
Does functional programming eat up more memory than procedural programming?
I mean ... if your objects(data structures whatever) are all imutable. Don't you end up having more object in the memory at a given time.
Doesn't this eat up more memory?
It depends on what you're doing. With functional programming you don't have to create defensive copies, so for certain problems it can end up using less memory.
Many functional programming languages also have good support for laziness, which can further reduce memory usage as you don't create objects until you actually use them. This is arguably something that's only correlated with functional programming rather than a direct cause, however.
Persistent values, that functional languages encourage but which can be implemented in an imperative language, make sharing a no-brainer.
Although the generally accepted idea is that with a garbage collector, there is some amount of wasted space at any given time (already unreachable but not yet collected blocks), in this context, without a garbage collector, you end up very often copying values that are immutable and could be shared, just because it's too much of a mess to decide who is responsible for freeing the memory after use.
These ideas are expanded on a bit in this experience report which does not claim to be an objective study but only anecdotal evidence.
Apart from avoiding defensive copies by the programmer, a very smart implementation of pure functional programming languages like Haskell or Standard ML (which lack physical pointer equality) can actively recover sharing of structurally equal values in memory, e.g. as part of the memory management and garbage collection.
Thus you can have automatic hash consing provided by your programming language runtime-system.
Compare this with objects in Java: object identity is an integral part of the language definition. Even just exchanging one immutable String for another poses semantic problems.
There is indeed at least a tendency to regard memory as affluent ressource (which, in fact, it really is in most cases), but this applies to modern programming as a whole.
With multiple cores, parallel garbage collectors and available RAM in the gigabytes, one used to concentrate on different aspects of a program than in earlier times, when every byte one could save counted. Remember when Bill Gates said "640K should be enough for every program"?
I know that I'm a lot late on this question.
Functional languages does not in general use more memory than imperative or OO languages. It depends more on the code you write. Yes F#, SML, Haskell and such has immutable values (not variables), but for all of them it goes without saying that if you update f.x. a single linked list, it re-compute only what is necessary.
Say you got a list of 5 elements, and you are removing the first 3 and adding a new one in front of it. it will simply get the pointer that points to the fourth element and let the new list point to that point of data i.e. reusing data. as seen below.
old list
[x0,x1,x2]
\
[x3,x4]
new list /
[y0,y1]
If it was an imperative language we could not do this because the values x3 and x4 could very well change over time, the list [x3,x4] could change too. Say that the 3 elements removed are not used afterward, the memory they use can be cleaned up right away, in contrast to unused space in an array.
That all data are immutable (except IO) are a strength. It simplifies the data flow analysis from a none trivial computation to a trivial one. This combined with a often very strong type system, will give the compiler a bunch of information about the code it can use to do optimization it normally could not do because of indicability. Most often the compiler turn values that are re-computed recursively and discarded from each iteration (recursion) into a mutable computation. These two things gives you the proof that if your program compile it will work. (with some assumptions)
If you look at the language Rust (not functional) just by learning about "borrow system" you will understand more about how and when things can be shared safely. it is a language that is painful to write code in unless you like to see your computer scream at you that your are an idiot. Rust is for the most part the combination of all the study made of programming language and type theory for more than 40 years. I mention Rust, because it despite the pain of writing in it, has the promise that if your program compile, there will be NO memory leaking, dead locking, dangling pointers, even in multi processing programs. This is because it uses much of the research of functional programming language that has been done.
For a more complex example of when functional programming uses less memory, I have made a lexer/parser interpreter (the same as generator but without the need to generate a code file) when computing the states of the DFA (deterministic finite automata) it uses immutable sets, because it compute new sets of already computed sets, my code allocate less memory simply because it borrow already known data points instead of copying it to a new set.
To wrap it up, yes functional programming can use more memory than imperative once. Most likely it is because you are using the wrong abstraction to mirror the problem. i.e. If you try to do it the imperative way in a functional language it will hurt you.
Try this book, it has not much on memory management but is a good book to start with if you will learn about compiler theory and yes it is legal to download. I have ask Torben, he is my old professor.
http://hjemmesider.diku.dk/~torbenm/Basics/
I'll throw my hat in the ring here. The short answer to the question is no, and this is because immutability does not mean the same thing as stored in memory. For example, let's take this toy program :
x = 2
x = x * 3
x = x * 2
print(x)
Which uses mutation to compute new values. Compare this to the same program which does not use mutation:
x = 2
y = x * 3
z = y * 2
print(z)
At first glance, it appears this requires 3x the memory of the first program! However, just because a value is immutable doesn't mean it needs to be stored in memory. In the case of the second program, after y is computed, x is no longer necessary, because it isn't used for the rest of the program, and can be garbage collected, or removed from memory. Similarly, after z is computed, y can be garbage collected. So, in principle, with a perfect garbage collector, after we execute the third line of code, I only need to have stored z in memory.
Another oft-worried about source of memory consumption in functional languages is deep recursion. For example, calculating a large Fibonacci number.
calc_fib(x):
if x > 1:
return x * calc_fib(x-1)
else:
return x
If I run calc_fib(100000), I could implement this in a way which requires storing 100000 values in memory, or I could use Tail-Call Elimination (basically storing only the most-recently computed value in memory instead of all function calls). For less straightforward recursion you can resort to trampolining. So for functional languages which support this, recursion does not need to be a source of massive memory consumption, either. However, not all nominally functional languages do (for example, JavaScript does not).
I've been looking into how programming languages work, and some of them have a so-called virtual machines. I understand that this is some form of emulation of the programming language within another programming language, and that it works like how a compiled language would be executed, with a stack. Did I get that right?
With the proviso that I did, what bamboozles me is that many non-compiled languages allow variables with "liberal" type systems. In Python for example, I can write this:
x = "Hello world!"
x = 2**1000
Strings and big integers are completely unrelated and occupy different amounts of space in memory, so how can this code even be represented in a stack-based environment? What exactly happens here? Is x pointed to a new place on the stack and the old string data left unreferenced? Do these languages not use a stack? If not, how do they represent variables internally?
Probably, your question should be titled as "How do dynamic languages work?."
That's simple, they store the variable type information along with it in memory. And this is not only done in interpreted or JIT compiled languages but also natively-compiled languages such as Objective-C.
In most VM languages, variables can be conceptualized as pointers (or references) to memory in the heap, even if the variable itself is on the stack. For languages that have primitive types (int and bool in Java, for example) those may be stored on the stack as well, but they can not be assigned new types dynamically.
Ignoring primitive types, all variables that exist on the stack have their actual values stored in the heap. Thus, if you dynamically reassign a value to them, the original value is abandoned (and the memory cleaned up via some garbage collection algorithm), and the new value is allocated in a new bit of memory.
The VM has nothing to do with the language. Any language can run on top of a VM (the Java VM has hundreds of languages already).
A VM enables a different kind of "assembly language" to be run, one that is more fit to adapting a compiler to. Everything done in a VM could be done in a CPU, so think of the VM like a CPU. (Some actually are implemented in hardware).
It's extremely low level, and in many cases heavily stack based--instead of registers, machine-level math is all relative to locations relative to the current stack pointer.
With normal compiled languages, many instructions are required for a single step. a + might look like "Grab the item from a point relative to the stack pointer into reg a, grab another into reg b. add reg a and b. put reg a into a place relative to the stack pointer.
The VM does all this with a single, short instruction, possibly one or two bytes instead of 4 or 8 bytes PER INSTRUCTION in machine language (depending on 32 or 64 bit architecture) which (guessing) should mean around 16 or 32 bytes of x86 for 1-2 bytes of machine code. (I could be wrong, my last x86 coding was in the 80286 era.)
Microsoft used (probably still uses) VMs in their office products to reduce the amount of code.
The procedure for creating the VM code is the same as creating machine language, just a different processor type essentially.
VMs can also implement their own security, error recovery and memory mechanisms that are very tightly related to the language.
Some of my description here is summary and from memory. If you want to explore the bytecode definition yourself, it's kinda fun:
http://java.sun.com/docs/books/jvms/second_edition/html/Instructions2.doc.html
The key to many of the 'how do VMs handle variables like this or that' really comes down to metadata... The meta information stored and then updated gives the VM a much better handle on how to allocate and then do the right thing with variables.
In many cases this is the type of overhead that can really get in the way of performance. However, modern day implementations, etc have come a long way in doing the right thing.
As for your specific questions - treating variables as vanilla objects / etc ... comes down to reassigning / reevaluating meta information on new assignments - that's why x can look one way and then the next.
To answer a part of your questions, I'd recommend a google tech talk about python, where some of your questions concerning dynamic languages are answered; for example what a variable is (it is not a pointer, nor a reference, but in case of python a label).