Ignore:
Timestamp:
Nov 17, 2017, 10:56:16 AM (4 years ago)
Author:
Rob Schluntz <rschlunt@…>
Branches:
aaron-thesis, arm-eh, cleanup-dtors, deferred_resn, demangler, jacob/cs343-translation, jenkins-sandbox, master, new-ast, new-ast-unique-expr, new-env, no_list, persistent-indexer, resolv-new, with_gc
Children:
cdbfab0
Parents:
f5c3b6c (diff), b7f8cb4 (diff)
Note: this is a merge changeset, the changes displayed below correspond to the merge itself.
Use the (diff) links above to see all the changes relative to each parent.
Message:

Merge branch 'master' into fix-bug-51

File:
1 edited

Legend:

Unmodified
Added
Removed
  • doc/proposals/concurrency/text/basics.tex

    rf5c3b6c r0fe4e62  
    99At its core, concurrency is based on having multiple call-stacks and scheduling among threads of execution executing on these stacks. Concurrency without parallelism only requires having multiple call stacks (or contexts) for a single thread of execution.
    1010
    11 Indeed, while execution with a single thread and multiple stacks where the thread is self-scheduling deterministically across the stacks is called coroutining, execution with a single and multiple stacks but where the thread is scheduled by an oracle (non-deterministic from the thread perspective) across the stacks is called concurrency.
    12 
    13 Therefore, a minimal concurrency system can be achieved by creating coroutines, which instead of context switching among each other, always ask an oracle where to context switch next. While coroutines can execute on the caller's stack-frame, stackfull coroutines allow full generality and are sufficient as the basis for concurrency. The aforementioned oracle is a scheduler and the whole system now follows a cooperative threading-model \cit. The oracle/scheduler can either be a stackless or stackfull entity and correspondingly require one or two context switches to run a different coroutine. In any case, a subset of concurrency related challenges start to appear. For the complete set of concurrency challenges to occur, the only feature missing is preemption. Indeed, concurrency challenges appear with non-determinism. Using mutual-exclusion or synchronisation are ways of limiting the lack of determinism in a system. A scheduler introduces order of execution uncertainty, while preemption introduces uncertainty about where context-switches occur. Now it is important to understand that uncertainty is not undesireable; uncertainty can often be used by systems to significantly increase performance and is often the basis of giving a user the illusion that tasks are running in parallel. Optimal performance in concurrent applications is often obtained by having as much non-determinism as correctness allows\cit.
     11Execution with a single thread and multiple stacks where the thread is self-scheduling deterministically across the stacks is called coroutining. Execution with a single and multiple stacks but where the thread is scheduled by an oracle (non-deterministic from the thread perspective) across the stacks is called concurrency.
     12
     13Therefore, a minimal concurrency system can be achieved by creating coroutines, which instead of context switching among each other, always ask an oracle where to context switch next. While coroutines can execute on the caller's stack-frame, stackfull coroutines allow full generality and are sufficient as the basis for concurrency. The aforementioned oracle is a scheduler and the whole system now follows a cooperative threading-model (a.k.a non-preemptive scheduling). The oracle/scheduler can either be a stackless or stackfull entity and correspondingly require one or two context switches to run a different coroutine. In any case, a subset of concurrency related challenges start to appear. For the complete set of concurrency challenges to occur, the only feature missing is preemption.
     14
     15A scheduler introduces order of execution uncertainty, while preemption introduces uncertainty about where context-switches occur. Mutual-exclusion and synchronisation are ways of limiting non-determinism in a concurrent system. Now it is important to understand that uncertainty is desireable; uncertainty can be used by runtime systems to significantly increase performance and is often the basis of giving a user the illusion that tasks are running in parallel. Optimal performance in concurrent applications is often obtained by having as much non-determinism as correctness allows.
    1416
    1517\section{\protect\CFA 's Thread Building Blocks}
    16 One of the important features that is missing in C is threading. On modern architectures, a lack of threading is unacceptable\cite{Sutter05, Sutter05b}, and therefore modern programming languages must have the proper tools to allow users to write performant concurrent and/or parallel programs. As an extension of C, \CFA needs to express these concepts in a way that is as natural as possible to programmers familiar with imperative languages. And being a system-level language means programmers expect to choose precisely which features they need and which cost they are willing to pay.
     18One of the important features that is missing in C is threading. On modern architectures, a lack of threading is unacceptable\cite{Sutter05, Sutter05b}, and therefore modern programming languages must have the proper tools to allow users to write performant concurrent programs to take advantage of parallelism. As an extension of C, \CFA needs to express these concepts in a way that is as natural as possible to programmers familiar with imperative languages. And being a system-level language means programmers expect to choose precisely which features they need and which cost they are willing to pay.
    1719
    1820\section{Coroutines: A stepping stone}\label{coroutine}
    19 While the main focus of this proposal is concurrency and parallelism, it is important to address coroutines, which are actually a significant building block of a concurrency system. Coroutines need to deal with context-switchs and other context-management operations. Therefore, this proposal includes coroutines both as an intermediate step for the implementation of threads, and a first class feature of \CFA. Furthermore, many design challenges of threads are at least partially present in designing coroutines, which makes the design effort that much more relevant. The core \acrshort{api} of coroutines revolve around two features: independent call stacks and \code{suspend}/\code{resume}.
    20 
    21 A good example of a problem made easier with coroutines is genereting the fibonacci sequence. This problem comes with the challenge of decoupling how a sequence is generated and how it is used. Figure \ref{fig:fibonacci-c} shows conventional approaches to writing generators in C. All three of these approach suffer from strong coupling. The left and center approaches require that the generator have knowledge of how the sequence will be used, while the rightmost approach requires to user to hold internal state between calls on behalf of th sequence generator and makes it much harder to handle corner cases like the Fibonacci seed.
     21While the main focus of this proposal is concurrency and parallelism, it is important to address coroutines, which are actually a significant building block of a concurrency system. Coroutines need to deal with context-switches and other context-management operations. Therefore, this proposal includes coroutines both as an intermediate step for the implementation of threads, and a first class feature of \CFA. Furthermore, many design challenges of threads are at least partially present in designing coroutines, which makes the design effort that much more relevant. The core \acrshort{api} of coroutines revolve around two features: independent call stacks and \code{suspend}/\code{resume}.
     22
    2223\begin{figure}
    23 \label{fig:fibonacci-c}
    2424\begin{center}
    2525\begin{tabular}{c @{\hskip 0.025in}|@{\hskip 0.025in} c @{\hskip 0.025in}|@{\hskip 0.025in} c}
     
    4545        }
    4646}
     47
     48int main() {
     49        void print_fib(int n) {
     50                printf("%d\n", n);
     51        }
     52
     53        fibonacci_func(
     54                10, print_fib
     55        );
     56
     57
     58
     59}
    4760\end{ccode}&\begin{ccode}[tabsize=2]
    4861//Using output array
     
    6275                        f2 = next;
    6376                }
    64                 *array = next;
    65                 array++;
    66         }
     77                array[i] = next;
     78        }
     79}
     80
     81
     82int main() {
     83        int a[10];
     84
     85        fibonacci_func(
     86                10, a
     87        );
     88
     89        for(int i=0;i<10;i++){
     90                printf("%d\n", a[i]);
     91        }
     92
    6793}
    6894\end{ccode}&\begin{ccode}[tabsize=2]
     
    7096typedef struct {
    7197        int f1, f2;
    72 } iterator_t;
     98} Iterator_t;
    7399
    74100int fibonacci_state(
    75         iterator_t * it
     101        Iterator_t * it
    76102) {
    77103        int f;
    78104        f = it->f1 + it->f2;
    79105        it->f2 = it->f1;
    80         it->f1 = f;
     106        it->f1 = max(f,1);
    81107        return f;
    82108}
     
    87113
    88114
     115
     116int main() {
     117        Iterator_t it={0,0};
     118
     119        for(int i=0;i<10;i++){
     120                printf("%d\n",
     121                        fibonacci_state(
     122                                &it
     123                        );
     124                );
     125        }
     126
     127}
    89128\end{ccode}
    90129\end{tabular}
    91130\end{center}
    92131\caption{Different implementations of a fibonacci sequence generator in C.}
     132\label{lst:fibonacci-c}
    93133\end{figure}
    94134
    95 
    96 Figure \ref{fig:fibonacci-cfa} is an example of a solution to the fibonnaci problem using \CFA coroutines, using the coroutine stack to hold sufficient state for the generation. This solution has the advantage of having very strong decoupling between how the sequence is generated and how it is used. Indeed, this version is a easy to use as the \code{fibonacci_state} solution, while the imlpementation is very similar to the \code{fibonacci_func} example.
     135A good example of a problem made easier with coroutines is generators, like the fibonacci sequence. This problem comes with the challenge of decoupling how a sequence is generated and how it is used. Figure \ref{lst:fibonacci-c} shows conventional approaches to writing generators in C. All three of these approach suffer from strong coupling. The left and center approaches require that the generator have knowledge of how the sequence is used, while the rightmost approach requires holding internal state between calls on behalf of the generator and makes it much harder to handle corner cases like the Fibonacci seed.
     136
     137Figure \ref{lst:fibonacci-cfa} is an example of a solution to the fibonnaci problem using \CFA coroutines, where the coroutine stack holds sufficient state for the generation. This solution has the advantage of having very strong decoupling between how the sequence is generated and how it is used. Indeed, this version is as easy to use as the \code{fibonacci_state} solution, while the imlpementation is very similar to the \code{fibonacci_func} example.
    97138
    98139\begin{figure}
    99 \label{fig:fibonacci-cfa}
    100140\begin{cfacode}
    101141coroutine Fibonacci {
     
    108148
    109149//main automacically called on first resume
    110 void main(Fibonacci & this) {
     150void main(Fibonacci & this) with (this) {
    111151        int fn1, fn2;           //retained between resumes
    112         this.fn = 0;
    113         fn1 = this.fn;
     152        fn = 0;
     153        fn1 = fn;
    114154        suspend(this);          //return to last resume
    115155
    116         this.fn = 1;
     156        fn = 1;
    117157        fn2 = fn1;
    118         fn1 = this.fn;
     158        fn1 = fn;
    119159        suspend(this);          //return to last resume
    120160
    121161        for ( ;; ) {
    122                 this.fn = fn1 + fn2;
     162                fn = fn1 + fn2;
    123163                fn2 = fn1;
    124                 fn1 = this.fn;
     164                fn1 = fn;
    125165                suspend(this);  //return to last resume
    126166        }
     
    140180\end{cfacode}
    141181\caption{Implementation of fibonacci using coroutines}
     182\label{lst:fibonacci-cfa}
    142183\end{figure}
    143184
    144 \subsection{Construction}
    145 One important design challenge for coroutines and threads (shown in section \ref{threads}) is that the runtime system needs to run code after the user-constructor runs to connect the object into the system. In the case of coroutines, this challenge is simpler since there is no non-determinism from preemption or scheduling. However, the underlying challenge remains the same for coroutines and threads.
    146 
    147 The runtime system needs to create the coroutine's stack and more importantly prepare it for the first resumption. The timing of the creation is non-trivial since users both expect to have fully constructed objects once execution enters the coroutine main and to be able to resume the coroutine from the constructor. As regular objects, constructors can leak coroutines before they are ready. There are several solutions to this problem but the chosen options effectively forces the design of the coroutine.
    148 
    149 Furthermore, \CFA faces an extra challenge as polymorphic routines create invisible thunks when casted to non-polymorphic routines and these thunks have function scope. For example, the following code, while looking benign, can run into undefined behaviour because of thunks:
    150 
    151 \begin{cfacode}
    152 //async: Runs function asynchronously on another thread
    153 forall(otype T)
    154 extern void async(void (*func)(T*), T* obj);
    155 
    156 forall(otype T)
    157 void noop(T *) {}
    158 
    159 void bar() {
    160         int a;
    161         async(noop, &a);
    162 }
    163 \end{cfacode}
    164 
    165 The generated C code\footnote{Code trimmed down for brevity} creates a local thunk to hold type information:
    166 
    167 \begin{ccode}
    168 extern void async(/* omitted */, void (*func)(void *), void *obj);
    169 
    170 void noop(/* omitted */, void *obj){}
    171 
    172 void bar(){
    173         int a;
    174         void _thunk0(int *_p0){
    175                 /* omitted */
    176                 noop(/* omitted */, _p0);
    177         }
    178         /* omitted */
    179         async(/* omitted */, ((void (*)(void *))(&_thunk0)), (&a));
    180 }
    181 \end{ccode}
    182 The problem in this example is a storage management issue, the function pointer \code{_thunk0} is only valid until the end of the block. This extra challenge limits which solutions are viable because storing the function pointer for too long causes undefined behavior; i.e. the stack based thunk being destroyed before it was used. This challenge is an extension of challenges that come with second-class routines. Indeed, GCC nested routines also have the limitation that the routines cannot be passed outside of the scope of the functions these were declared in. The case of coroutines and threads is simply an extension of this problem to multiple call-stacks.
    183 
    184 \subsection{Alternative: Composition}
    185 One solution to this challenge is to use composition/containement, where uses add insert a coroutine field which contains the necessary information to manage the coroutine.
    186 
    187 \begin{cfacode}
    188 struct Fibonacci {
    189         int fn; //used for communication
    190         coroutine c; //composition
    191 };
    192 
    193 void ?{}(Fibonacci & this) {
    194         this.fn = 0;
    195         (this.c){}; //Call constructor to initialize coroutine
    196 }
    197 \end{cfacode}
    198 There are two downsides to this approach. The first, which is relatively minor, made aware of the main routine pointer. This information must either be store in the coroutine runtime data or in its static type structure. When using composition, all coroutine handles have the same static type structure which means the pointer to the main needs to be part of the runtime data. This requirement means the coroutine data must be made larger to store a value that is actually a compile time constant (address of the main routine). The second problem, which is both subtle and significant, is that now users can get the initialisation order of coroutines wrong. Indeed, every field of a \CFA struct is constructed but in declaration order, unless users explicitly write otherwise. This semantics means that users who forget to initialize the coroutine handle may resume the coroutine with an uninitilized object. For coroutines, this is unlikely to be a problem, for threads however, this is a significant problem. Figure \ref{fig:fmt-line} shows the \code{Format} coroutine which rearranges text in order to group characters into blocks of fixed size. This is a good example where the control flow is made much simpler from being able to resume the coroutine from the constructor and highlights the idea that interesting control flow can occor in the constructor.
     185Figure \ref{lst:fmt-line} shows the \code{Format} coroutine which rearranges text in order to group characters into blocks of fixed size. The example takes advantage of resuming coroutines in the constructor to simplify the code and highlights the idea that interesting control flow can occur in the constructor.
     186
    199187\begin{figure}
    200 \label{fig:fmt-line}
    201188\begin{cfacode}[tabsize=3]
    202189//format characters into blocks of 4 and groups of 5 blocks per line
     
    244231\end{cfacode}
    245232\caption{Formatting text into lines of 5 blocks of 4 characters.}
     233\label{lst:fmt-line}
    246234\end{figure}
    247235
     236\subsection{Construction}
     237One important design challenge for coroutines and threads (shown in section \ref{threads}) is that the runtime system needs to run code after the user-constructor runs to connect the fully constructed object into the system. In the case of coroutines, this challenge is simpler since there is no non-determinism from preemption or scheduling. However, the underlying challenge remains the same for coroutines and threads.
     238
     239The runtime system needs to create the coroutine's stack and more importantly prepare it for the first resumption. The timing of the creation is non-trivial since users both expect to have fully constructed objects once execution enters the coroutine main and to be able to resume the coroutine from the constructor. As regular objects, constructors can leak coroutines before they are ready. There are several solutions to this problem but the chosen options effectively forces the design of the coroutine.
     240
     241Furthermore, \CFA faces an extra challenge as polymorphic routines create invisible thunks when casted to non-polymorphic routines and these thunks have function scope. For example, the following code, while looking benign, can run into undefined behaviour because of thunks:
     242
     243\begin{cfacode}
     244//async: Runs function asynchronously on another thread
     245forall(otype T)
     246extern void async(void (*func)(T*), T* obj);
     247
     248forall(otype T)
     249void noop(T*) {}
     250
     251void bar() {
     252        int a;
     253        async(noop, &a); //start thread running noop with argument a
     254}
     255\end{cfacode}
     256
     257The generated C code\footnote{Code trimmed down for brevity} creates a local thunk to hold type information:
     258
     259\begin{ccode}
     260extern void async(/* omitted */, void (*func)(void *), void *obj);
     261
     262void noop(/* omitted */, void *obj){}
     263
     264void bar(){
     265        int a;
     266        void _thunk0(int *_p0){
     267                /* omitted */
     268                noop(/* omitted */, _p0);
     269        }
     270        /* omitted */
     271        async(/* omitted */, ((void (*)(void *))(&_thunk0)), (&a));
     272}
     273\end{ccode}
     274The problem in this example is a storage management issue, the function pointer \code{_thunk0} is only valid until the end of the block, which limits the viable solutions because storing the function pointer for too long causes undefined behavior; i.e., the stack-based thunk being destroyed before it can be used. This challenge is an extension of challenges that come with second-class routines. Indeed, GCC nested routines also have the limitation that nested routine cannot be passed outside of the declaration scope. The case of coroutines and threads is simply an extension of this problem to multiple call-stacks.
     275
     276\subsection{Alternative: Composition}
     277One solution to this challenge is to use composition/containement, where coroutine fields are added to manage the coroutine.
     278
     279\begin{cfacode}
     280struct Fibonacci {
     281        int fn; //used for communication
     282        coroutine c; //composition
     283};
     284
     285void FibMain(void *) {
     286        //...
     287}
     288
     289void ?{}(Fibonacci & this) {
     290        this.fn = 0;
     291        //Call constructor to initialize coroutine
     292        (this.c){myMain};
     293}
     294\end{cfacode}
     295The downside of this approach is that users need to correctly construct the coroutine handle before using it. Like any other objects, doing so the users carefully choose construction order to prevent usage of unconstructed objects. However, in the case of coroutines, users must also pass to the coroutine information about the coroutine main, like in the previous example. This opens the door for user errors and requires extra runtime storage to pass at runtime information that can be known statically.
    248296
    249297\subsection{Alternative: Reserved keyword}
     
    255303};
    256304\end{cfacode}
    257 This mean the compiler can solve problems by injecting code where needed. The downside of this approach is that it makes coroutine a special case in the language. Users who would want to extend coroutines or build their own for various reasons can only do so in ways offered by the language. Furthermore, implementing coroutines without language supports also displays the power of the programming language used. While this is ultimately the option used for idiomatic \CFA code, coroutines and threads can both be constructed by users without using the language support. The reserved keywords are only present to improve ease of use for the common cases.
     305The \code{coroutine} keyword means the compiler can find and inject code where needed. The downside of this approach is that it makes coroutine a special case in the language. Users wantint to extend coroutines or build their own for various reasons can only do so in ways offered by the language. Furthermore, implementing coroutines without language supports also displays the power of the programming language used. While this is ultimately the option used for idiomatic \CFA code, coroutines and threads can still be constructed by users without using the language support. The reserved keywords are only present to improve ease of use for the common cases.
    258306
    259307\subsection{Alternative: Lamda Objects}
    260308
    261 For coroutines as for threads, many implementations are based on routine pointers or function objects\cit. For example, Boost implements coroutines in terms of four functor object types:
     309For coroutines as for threads, many implementations are based on routine pointers or function objects\cite{Butenhof97, ANSI14:C++, MS:VisualC++, BoostCoroutines15}. For example, Boost implements coroutines in terms of four functor object types:
    262310\begin{cfacode}
    263311asymmetric_coroutine<>::pull_type
     
    268316Often, the canonical threading paradigm in languages is based on function pointers, pthread being one of the most well known examples. The main problem of this approach is that the thread usage is limited to a generic handle that must otherwise be wrapped in a custom type. Since the custom type is simple to write in \CFA and solves several issues, added support for routine/lambda based coroutines adds very little.
    269317
    270 A variation of this would be to use an simple function pointer in the same way pthread does for threads :
     318A variation of this would be to use a simple function pointer in the same way pthread does for threads :
    271319\begin{cfacode}
    272320void foo( coroutine_t cid, void * arg ) {
     
    281329}
    282330\end{cfacode}
    283 This semantic is more common for thread interfaces than coroutines but would work equally well. As discussed in section \ref{threads}, this approach is superseeded by static approaches in terms of expressivity.
     331This semantics is more common for thread interfaces than coroutines works equally well. As discussed in section \ref{threads}, this approach is superseeded by static approaches in terms of expressivity.
    284332
    285333\subsection{Alternative: Trait-based coroutines}
     
    350398\end{cfacode}
    351399
    352 In this example, threads of type \code{foo} start execution in the \code{void main(foo &)} routine, which prints \code{"Hello World!"}. While this thesis encourages this approach to enforce strongly-typed programming, users may prefer to use the routine-based thread semantics for the sake of simplicity. With these semantics it is trivial to write a thread type that takes a function pointer as a parameter and executes it on its stack asynchronously
     400In this example, threads of type \code{foo} start execution in the \code{void main(foo &)} routine, which prints \code{"Hello World!"}. While this thesis encourages this approach to enforce strongly-typed programming, users may prefer to use the routine-based thread semantics for the sake of simplicity. With the static semantics it is trivial to write a thread type that takes a function pointer as a parameter and executes it on its stack asynchronously.
    353401\begin{cfacode}
    354402typedef void (*voidFunc)(int);
     
    361409void ?{}(FuncRunner & this, voidFunc inFunc, int arg) {
    362410        this.func = inFunc;
     411        this.arg  = arg;
    363412}
    364413
    365414void main(FuncRunner & this) {
     415        //thread starts here and runs the function
    366416        this.func( this.arg );
    367417}
    368418\end{cfacode}
    369419
    370 An consequence of the strongly typed approach to main is that memory layout of parameters and return values to/from a thread are now explicitly specified in the \acrshort{api}.
     420A consequence of the strongly-typed approach to main is that memory layout of parameters and return values to/from a thread are now explicitly specified in the \acrshort{api}.
    371421
    372422Of course for threads to be useful, it must be possible to start and stop threads and wait for them to complete execution. While using an \acrshort{api} such as \code{fork} and \code{join} is relatively common in the literature, such an interface is unnecessary. Indeed, the simplest approach is to use \acrshort{raii} principles and have threads \code{fork} after the constructor has completed and \code{join} before the destructor runs.
     
    389439\end{cfacode}
    390440
    391 This semantic has several advantages over explicit semantics: a thread is always started and stopped exaclty once and users cannot make any progamming errors and it naturally scales to multiple threads meaning basic synchronisation is very simple
     441This semantic has several advantages over explicit semantics: a thread is always started and stopped exaclty once, users cannot make any progamming errors, and it naturally scales to multiple threads meaning basic synchronisation is very simple.
    392442
    393443\begin{cfacode}
     
    411461\end{cfacode}
    412462
    413 However, one of the drawbacks of this approach is that threads now always form a lattice, that is they are always destroyed in opposite order of construction because of block structure. This restriction is relaxed by using dynamic allocation, so threads can outlive the scope in which they are created, much like dynamically allocating memory lets objects outlive the scope in which they are created
     463However, one of the drawbacks of this approach is that threads now always form a lattice, that is they are always destroyed in the opposite order of construction because of block structure. This restriction is relaxed by using dynamic allocation, so threads can outlive the scope in which they are created, much like dynamically allocating memory lets objects outlive the scope in which they are created.
    414464
    415465\begin{cfacode}
Note: See TracChangeset for help on using the changeset viewer.