source: doc/proposals/concurrency/text/concurrency.tex @ 21a1efb

ADTaaron-thesisarm-ehast-experimentalcleanup-dtorsdeferred_resndemanglerenumforall-pointer-decayjacob/cs343-translationjenkins-sandboxnew-astnew-ast-unique-exprnew-envno_listpersistent-indexerpthread-emulationqualifiedEnumresolv-newwith_gc
Last change on this file since 21a1efb was 21a1efb, checked in by Thierry Delisle <tdelisle@…>, 7 years ago

Sent a draft to peter

  • Property mode set to 100644
File size: 47.9 KB
Line 
1% ======================================================================
2% ======================================================================
3\chapter{Concurrency}
4% ======================================================================
5% ======================================================================
6Several tool can be used to solve concurrency challenges. Since many of these challenges appear with the use of mutable shared-state, some languages and libraries simply disallow mutable shared-state (Erlang~\cite{Erlang}, Haskell~\cite{Haskell}, Akka (Scala)~\cite{Akka}). In these paradigms, interaction among concurrent objects relies on message passing~\cite{Thoth,Harmony,V-Kernel} or other paradigms that closely relate to networking concepts (channels\cit for example). However, in languages that use routine calls as their core abstraction-mechanism, these approaches force a clear distinction between concurrent and non-concurrent paradigms (i.e., message passing versus routine call). This distinction in turn means that, in order to be effective, programmers need to learn two sets of designs patterns. While this distinction can be hidden away in library code, effective use of the librairy still has to take both paradigms into account.
7
8Approaches based on shared memory are more closely related to non-concurrent paradigms since they often rely on basic constructs like routine calls and shared objects. At the lowest level, concurrent paradigms are implemented as atomic operations and locks. Many such mechanisms have been proposed, including semaphores~\cite{Dijkstra68b} and path expressions~\cite{Campbell74}. However, for productivity reasons it is desireable to have a higher-level construct be the core concurrency paradigm~\cite{HPP:Study}.
9
10An approach that is worth mentionning because it is gaining in popularity is transactionnal memory~\cite{Dice10}[Check citation]. While this approach is even pursued by system languages like \CC\cit, the performance and feature set is currently too restrictive to be the main concurrency paradigm for general purpose language, which is why it was rejected as the core paradigm for concurrency in \CFA.
11
12One of the most natural, elegant, and efficient mechanisms for synchronization and communication, especially for shared memory systems, is the \emph{monitor}. Monitors were first proposed by Brinch Hansen~\cite{Hansen73} and later described and extended by C.A.R.~Hoare~\cite{Hoare74}. Many programming languages---e.g., Concurrent Pascal~\cite{ConcurrentPascal}, Mesa~\cite{Mesa}, Modula~\cite{Modula-2}, Turing~\cite{Turing:old}, Modula-3~\cite{Modula-3}, NeWS~\cite{NeWS}, Emerald~\cite{Emerald}, \uC~\cite{Buhr92a} and Java~\cite{Java}---provide monitors as explicit language constructs. In addition, operating-system kernels and device drivers have a monitor-like structure, although they often use lower-level primitives such as semaphores or locks to simulate monitors. For these reasons, this project proposes monitors as the core concurrency-construct.
13
14\section{Basics}
15Non-determinism requires concurrent systems to offer support for mutual-exclusion and synchronisation. Mutual-exclusion is the concept that only a fixed number of threads can access a critical section at any given time, where a critical section is a group of instructions on an associated portion of data that requires the restricted access. On the other hand, synchronization enforces relative ordering of execution and synchronization tools numerous mechanisms to establish timing relationships among threads.
16
17\subsection{Mutual-Exclusion}
18As mentionned above, mutual-exclusion is the guarantee that only a fix number of threads can enter a critical section at once. However, many solution exists for mutual exclusion which vary in terms of performance, flexibility and ease of use. Methods range from low-level locks, which are fast and flexible but require significant attention to be correct, to  higher-level mutual-exclusion methods, which sacrifice some performance in order to improve ease of use. Ease of use comes by either guaranteeing some problems cannot occur (e.g., being deadlock free) or by offering a more explicit coupling between data and corresponding critical section. For example, the \CC \code{std::atomic<T>} which offer an easy way to express mutual-exclusion on a restricted set of operations (.e.g: reading/writing large types atomically). Another challenge with low-level locks is composability. Locks are not composable because it takes careful organising for multiple locks to be used while preventing deadlocks. Easing composability is another feature higher-level mutual-exclusion mechanisms often offer.
19
20\subsection{Synchronization}
21As for mutual-exclusion, low level synchronisation primitive often offer good performance and good flexibility at the cost of ease of use. Again, higher-level mechanism often simplify usage by adding better coupling between synchronization and data, .eg., message passing, or offering simple solution to otherwise involved challenges. An example of this is barging. As mentionned above synchronization can be expressed as guaranteeing that event \textit{X} always happens before \textit{Y}. Most of the time synchronisation happens around a critical section, where threads most acquire said critical section in a certain order. However, it may also be desired to be able to guarantee that event \textit{Z} does not occur between \textit{X} and \textit{Y}. This is called barging, where event \textit{X} tries to effect event \textit{Y} but anoter thread races to grab the critical section and emits \textit{Z} before \textit{Y}. Preventing or detecting barging is an involved challenge with low-level locks, which can be made much easier by higher-level constructs.
22
23% ======================================================================
24% ======================================================================
25\section{Monitors}
26% ======================================================================
27% ======================================================================
28A monitor is a set of routines that ensure mutual exclusion when accessing shared state. This concept is generally associated with Object-Oriented Languages like Java~\cite{Java} or \uC~\cite{uC++book} but does not strictly require OO semantics. The only requirements is the ability to declare a handle to a shared object and a set of routines that act on it :
29\begin{cfacode}
30        typedef /*some monitor type*/ monitor;
31        int f(monitor & m);
32
33        int main() {
34                monitor m;  //Handle m
35                f(m);       //Routine using handle
36        }
37\end{cfacode}
38
39% ======================================================================
40% ======================================================================
41\subsection{Call semantics} \label{call}
42% ======================================================================
43% ======================================================================
44The above monitor example displays some of the intrinsic characteristics. First, it is necessary to use pass-by-reference over pass-by-value for monitor routines. This semantics is important because at their core, monitors are implicit mutual-exclusion objects (locks), and these objects cannot be copied. Therefore, monitors are implicitly non-copyable objects.
45
46Another aspect to consider is when a monitor acquires its mutual exclusion. For example, a monitor may need to be passed through multiple helper routines that do not acquire the monitor mutual-exclusion on entry. Pass through can occur for generic helper routines (\code{swap}, \code{sort}, etc.) or specific helper routines like the following to implement an atomic counter :
47
48\begin{cfacode}
49        monitor counter_t { /*...see section $\ref{data}$...*/ };
50
51        void ?{}(counter_t & nomutex this); //constructor
52        size_t ++?(counter_t & mutex this); //increment
53
54        //need for mutex is platform dependent
55        void ?{}(size_t * this, counter_t & mutex cnt); //conversion
56\end{cfacode}
57
58Here, the constructor(\code{?\{\}}) uses the \code{nomutex} keyword to signify that it does not acquire the monitor mutual-exclusion when constructing. This semantics is because an object not yet constructed should never be shared and therefore does not require mutual exclusion. The prefix increment operator uses \code{mutex} to protect the incrementing process from race conditions. Finally, there is a conversion operator from \code{counter_t} to \code{size_t}. This conversion may or may not require the \code{mutex} keyword depending on whether or not reading an \code{size_t} is an atomic operation.
59
60Having both \code{mutex} and \code{nomutex} keywords is redundant based on the meaning of a routine having neither of these keywords. For example, given a routine without qualifiers \code{void foo(counter_t & this)}, then it is reasonable that it should default to the safest option \code{mutex}, whereas assuming \code{nomutex} is unsafe and may cause subtle errors. In fact, \code{nomutex} is the "normal" parameter behaviour, with the \code{nomutex} keyword effectively stating explicitly that "this routine is not special". Another alternative is to make having exactly one of these keywords mandatory, which would provide the same semantics but without the ambiguity of supporting routines neither keyword. Mandatory keywords would also have the added benefit of being self-documented but at the cost of extra typing. While there are several benefits to mandatory keywords, they do bring a few challenges. Mandatory keywords in \CFA would imply that the compiler must know without a doubt wheter or not a parameter is a monitor or not. Since \CFA relies heavily on traits as an abstraction mechanism, the distinction between a type that is a monitor and a type that looks like a monitor can become blurred. For this reason, \CFA only has the \code{mutex} keyword.
61
62
63The next semantic decision is to establish when \code{mutex} may be used as a type qualifier. Consider the following declarations:
64\begin{cfacode}
65int f1(monitor & mutex m);
66int f2(const monitor & mutex m);
67int f3(monitor ** mutex m);
68int f4(monitor * mutex m []);
69int f5(graph(monitor*) & mutex m);
70\end{cfacode}
71The problem is to indentify which object(s) should be acquired. Furthermore, each object needs to be acquired only once. In the case of simple routines like \code{f1} and \code{f2} it is easy to identify an exhaustive list of objects to acquire on entry. Adding indirections (\code{f3}) still allows the compiler and programmer to indentify which object is acquired. However, adding in arrays (\code{f4}) makes it much harder. Array lengths are not necessarily known in C, and even then making sure objects are only acquired once becomes none-trivial. This can be extended to absurd limits like \code{f5}, which uses a graph of monitors. To keep everyone as sane as possible~\cite{Chicken}, this projects imposes the requirement that a routine may only acquire one monitor per parameter and it must be the type of the parameter with one level of indirection (ignoring potential qualifiers). Also note that while routine \code{f3} can be supported, meaning that monitor \code{**m} is be acquired, passing an array to this routine would be type safe and yet result in undefined behavior because only the first element of the array is acquired. This is specially true for non-copyable objects like monitors, where an array of pointers is simplest way to express a group of monitors. However, this ambiguity is part of the C type-system with respects to arrays. For this reason, \code{mutex} is disallowed in the context where arrays may be passed:
72
73\begin{cfacode}
74int f1(monitor & mutex m);   //Okay : recommanded case
75int f2(monitor * mutex m);   //Okay : could be an array but probably not
76int f3(monitor mutex m []);  //Not Okay : Array of unkown length
77int f4(monitor ** mutex m);  //Not Okay : Could be an array
78int f5(monitor * mutex m []); //Not Okay : Array of unkown length
79\end{cfacode}
80
81Unlike object-oriented monitors, where calling a mutex member \emph{implicitly} acquires mutual-exclusion, \CFA uses an explicit mechanism to acquire mutual-exclusion. A consequence of this approach is that it extends naturally to multi-monitor calls.
82\begin{cfacode}
83int f(MonitorA & mutex a, MonitorB & mutex b);
84
85MonitorA a;
86MonitorB b;
87f(a,b);
88\end{cfacode}
89The capacity to acquire multiple locks before entering a critical section is called \emph{\gls{group-acquire}}. In practice, writing multi-locking routines that do not lead to deadlocks is tricky. Having language support for such a feature is therefore a significant asset for \CFA. In the case presented above, \CFA guarantees that the order of aquisition is consistent across calls to routines using the same monitors as arguments. However, since \CFA monitors use multi-acquisition locks, users can effectively force the acquiring order. For example, notice which routines use \code{mutex}/\code{nomutex} and how this affects aquiring order:
90\begin{cfacode}
91        void foo(A & mutex a, B & mutex b) { //acquire a & b
92                ...
93        }
94
95        void bar(A & mutex a, B & /*nomutex*/ b) { //acquire a
96                ... foo(a, b); ... //acquire b
97        }
98
99        void baz(A & /*nomutex*/ a, B & mutex b) { //acquire b
100                ... foo(a, b); ... //acquire a
101        }
102\end{cfacode}
103The multi-acquisition monitor lock allows a monitor lock to be acquired by both \code{bar} or \code{baz} and acquired again in \code{foo}. In the calls to \code{bar} and \code{baz} the monitors are acquired in opposite order.
104
105However, such use leads the lock acquiring order problem. In the example above, the user uses implicit ordering in the case of function \code{foo} but explicit ordering in the case of \code{bar} and \code{baz}. This subtle mistake means that calling these routines concurrently may lead to deadlock and is therefore undefined behavior. As shown on several occasion\cit, solving this problem requires:
106\begin{enumerate}
107        \item Dynamically tracking of the monitor-call order.
108        \item Implement rollback semantics.
109\end{enumerate}
110While the first requirement is already a significant constraint on the system, implementing a general rollback semantics in a C-like language is prohibitively complex \cit. In \CFA, users simply need to be carefull when acquiring multiple monitors at the same time.
111
112Finally, for convenience, monitors support multiple acquiring, that is acquiring a monitor while already holding it does not cause a deadlock. It simply increments an internal counter which is then used to release the monitor after the number of acquires and releases match up. This is particularly usefull when monitor routines use other monitor routines as helpers or for recursions. For example:
113\begin{cfacode}
114monitor bank {
115        int money;
116        log_t usr_log;
117};
118
119void deposit( bank & mutex b, int deposit ) {
120        b.money += deposit;
121        b.usr_log | "Adding" | deposit | endl;
122}
123
124void transfer( bank & mutex mybank, bank & mutex yourbank, int me2you) {
125        deposit( mybank, -me2you );
126        deposit( yourbank, me2you );
127}
128\end{cfacode}
129
130% ======================================================================
131% ======================================================================
132\subsection{Data semantics} \label{data}
133% ======================================================================
134% ======================================================================
135Once the call semantics are established, the next step is to establish data semantics. Indeed, until now a monitor is used simply as a generic handle but in most cases monitors contain shared data. This data should be intrinsic to the monitor declaration to prevent any accidental use of data without its appropriate protection. For example, here is a complete version of the counter showed in section \ref{call}:
136\begin{cfacode}
137monitor counter_t {
138        int value;
139};
140
141void ?{}(counter_t & this) {
142        this.cnt = 0;
143}
144
145int ?++(counter_t & mutex this) {
146        return ++this.value;
147}
148
149//need for mutex is platform dependent here
150void ?{}(int * this, counter_t & mutex cnt) {
151        *this = (int)cnt;
152}
153\end{cfacode}
154
155This counter is used as follows:
156\begin{center}
157\begin{tabular}{c @{\hskip 0.35in} c @{\hskip 0.35in} c}
158\begin{cfacode}
159//shared counter
160counter_t cnt1, cnt2;
161
162//multiple threads access counter
163thread 1 : cnt1++; cnt2++;
164thread 2 : cnt1++; cnt2++;
165thread 3 : cnt1++; cnt2++;
166        ...
167thread N : cnt1++; cnt2++;
168\end{cfacode}
169\end{tabular}
170\end{center}
171Notice how the counter is used without any explicit synchronisation and yet supports thread-safe semantics for both reading and writting.
172
173% ======================================================================
174% ======================================================================
175\subsection{Implementation Details: Interaction with polymorphism}
176% ======================================================================
177% ======================================================================
178Depending on the choice of semantics for when monitor locks are acquired, interaction between monitors and \CFA's concept of polymorphism can be complex to support. However, it is shown that entry-point locking solves most of the issues.
179
180First of all, interaction between \code{otype} polymorphism and monitors is impossible since monitors do not support copying. Therefore, the main question is how to support \code{dtype} polymorphism. Since a monitor's main purpose is to ensure mutual exclusion when accessing shared data, this implies that mutual exclusion is only required for routines that do in fact access shared data. However, since \code{dtype} polymorphism always handles incomplete types (by definition), no \code{dtype} polymorphic routine can access shared data since the data requires knowledge about the type. Therefore, the only concern when combining \code{dtype} polymorphism and monitors is to protect access to routines.
181
182Before looking into complex control-flow, it is important to present the difference between the two acquiring options : callsite and entry-point locking, i.e. acquiring the monitors before making a mutex routine call or as the first operation of the mutex routine-call. For example:
183\begin{center}
184\setlength\tabcolsep{1.5pt}
185\begin{tabular}{|c|c|c|}
186Code & \gls{callsite-locking} & \gls{entry-point-locking} \\
187\CFA & pseudo-code & pseudo-code \\
188\hline
189\begin{cfacode}[tabsize=3]
190void foo(monitor& mutex a){
191
192
193
194        //Do Work
195        //...
196
197}
198
199void main() {
200        monitor a;
201
202
203
204        foo(a);
205
206}
207\end{cfacode} & \begin{pseudo}[tabsize=3]
208foo(& a) {
209
210
211
212        //Do Work
213        //...
214
215}
216
217main() {
218        monitor a;
219        //calling routine
220        //handles concurrency
221        acquire(a);
222        foo(a);
223        release(a);
224}
225\end{pseudo} & \begin{pseudo}[tabsize=3]
226foo(& a) {
227        //called routine
228        //handles concurrency
229        acquire(a);
230        //Do Work
231        //...
232        release(a);
233}
234
235main() {
236        monitor a;
237
238
239
240        foo(a);
241
242}
243\end{pseudo}
244\end{tabular}
245\end{center}
246
247\Gls{callsite-locking} is inefficient, since any \code{dtype} routine may have to obtain some lock before calling a routine, depending on whether or not the type passed is a monitor. However, with \gls{entry-point-locking} calling a monitor routine becomes exactly the same as calling it from anywhere else.
248
249Note the \code{mutex} keyword relies on the resolver, which means that in cases where a generic monitor routine is actually desired, writing a mutex routine is possible with the proper trait. This is possible because monitors are designed in terms a trait. For example:
250\begin{cfacode}
251//Incorrect
252//T is not a monitor
253forall(dtype T)
254void foo(T * mutex t);
255
256//Correct
257//this function only works on monitors
258//(any monitor)
259forall(dtype T | is_monitor(T))
260void bar(T * mutex t));
261\end{cfacode}
262
263
264% ======================================================================
265% ======================================================================
266\section{Internal scheduling} \label{insched}
267% ======================================================================
268% ======================================================================
269In addition to mutual exclusion, the monitors at the core of \CFA's concurrency can also be used to achieve synchronisation. With monitors, this is generally achieved with internal or external scheduling as in\cit. Since internal scheduling of single monitors is mostly a solved problem, this proposal concentraits on extending internal scheduling to multiple monitors at once. Indeed, like the \gls{group-acquire} semantics, internal scheduling extends to multiple monitors at once in a way that is natural to the user but requires additional complexity on the implementation side.
270
271First, here is a simple example of such a technique:
272
273\begin{cfacode}
274        monitor A {
275                condition e;
276        }
277
278        void foo(A & mutex a) {
279                ...
280                // Wait for cooperation from bar()
281                wait(a.e);
282                ...
283        }
284
285        void bar(A & mutex a) {
286                // Provide cooperation for foo()
287                ...
288                // Unblock foo at scope exit
289                signal(a.e);
290        }
291\end{cfacode}
292
293There are two details to note here. First, there \code{signal} is a delayed operation, it only unblocks the waiting thread when it reaches the end of the critical section. This is needed to respect mutual-exclusion. Second, in \CFA, \code{condition} have no particular need to be stored inside a monitor, beyond any software engineering reasons. Here routine \code{foo} waits for the \code{signal} from \code{bar} before making further progress, effectively ensuring a basic ordering.
294
295An important aspect to take into account here is that \CFA does not allow barging, which means that once function \code{bar} releases the monitor, foo is guaranteed to resume immediately after (unless some other thread waited on the same condition). This guarantees offers the benefit of not having to loop arount waits in order to guarantee that a condition is still met. The main reason \CFA offers this guarantee is that users can easily introduce barging if it becomes a necessity but adding barging prevention or barging avoidance is more involved without language support. Supporting barging prevention as well as extending internal scheduling to multiple monitors is the main source of complexity in the design of \CFA concurrency.
296
297% ======================================================================
298% ======================================================================
299\subsection{Internal Scheduling - multi monitor}
300% ======================================================================
301% ======================================================================
302It is easier to understand the problem of multi-monitor scheduling using a series of pseudo-code. Note that for simplicity in the following snippets of pseudo-code, waiting and signalling is done using an implicit condition variable, like Java built-in monitors.
303
304\begin{multicols}{2}
305thread 1
306\begin{pseudo}
307acquire A
308        wait A
309release A
310\end{pseudo}
311
312\columnbreak
313
314thread 2
315\begin{pseudo}
316acquire A
317        signal A
318release A
319\end{pseudo}
320\end{multicols}
321The example shows the simple case of having two threads (one for each column) and a single monitor A. One thread acquires before waiting (atomically blocking and releasing A) and the other acquires before signalling. There is an important thing to note here, both \code{wait} and \code{signal} must be called with the proper monitor(s) already acquired. This restriction is hidden on the user side in \uC, as it is a logical requirement for barging prevention.
322
323A direct extension of the previous example is the \gls{group-acquire} version:
324
325\begin{multicols}{2}
326\begin{pseudo}
327acquire A & B
328        wait A & B
329release A & B
330\end{pseudo}
331
332\columnbreak
333
334\begin{pseudo}
335acquire A & B
336        signal A & B
337release A & B
338\end{pseudo}
339\end{multicols}
340This version uses \gls{group-acquire} (denoted using the \& symbol), but the presence of multiple monitors does not add a particularly new meaning. Synchronization happens between the two threads in exactly the same way and order. The only difference is that mutual exclusion covers more monitors. On the implementation side, handling multiple monitors does add a degree of complexity as the next few examples demonstrate.
341
342While deadlock issues can occur when nesting monitors, these issues are only a symptom of the fact that locks, and by extension monitors, are not perfectly composable. However, for monitors as for locks, it is possible to write a program using nesting without encountering any problems if nested is done correctly. For example, the next pseudo-code snippet acquires monitors A then B before waiting while only acquiring B when signalling, effectively avoiding the nested monitor problem.
343
344\begin{multicols}{2}
345\begin{pseudo}
346acquire A
347        acquire B
348                wait B
349        release B
350release A
351\end{pseudo}
352
353\columnbreak
354
355\begin{pseudo}
356
357acquire B
358        signal B
359release B
360
361\end{pseudo}
362\end{multicols}
363
364The next example is where \gls{group-acquire} adds a significant layer of complexity to the internal signalling semantics.
365
366\begin{multicols}{2}
367Waiting thread
368\begin{pseudo}[numbers=left]
369acquire A
370        // Code Section 1
371        acquire A & B
372                // Code Section 2
373                wait A & B
374                // Code Section 3
375        release A & B
376        // Code Section 4
377release A
378\end{pseudo}
379
380\columnbreak
381
382Signalling thread
383\begin{pseudo}[numbers=left, firstnumber=10]
384acquire A
385        // Code Section 5
386        acquire A & B
387                // Code Section 6
388                signal A & B
389                // Code Section 7
390        release A & B
391        // Code Section 8
392release A
393\end{pseudo}
394\end{multicols}
395\begin{center}
396Listing 1
397\end{center}
398
399It is particularly important to pay attention to code sections 8 and 4, which are where the existing semantics of internal scheduling need to be extended for multiple monitors. The root of the problem is that \gls{group-acquire} is used in a context where one of the monitors is already acquired and is why it is important to define the behaviour of the previous pseudo-code. When the signaller thread reaches the location where it should "release A \& B" (line 16), it must actually transfer ownership of monitor B to the waiting thread. This ownership trasnfer is required in order to prevent barging. Since the signalling thread still needs the monitor A, simply waking up the waiting thread is not an option because it would violate mutual exclusion. There are three options:
400
401\subsubsection{Delaying signals}
402The first more obvious solution to solve the problem of multi-monitor scheduling is to keep ownership of all locks until the last lock is ready to be transferred. It can be argued that that moment is the correct time to transfer ownership when the last lock is no longer needed because this semantics fits most closely to the behaviour of single monitor scheduling. This solution has the main benefit of transferring ownership of groups of monitors, which simplifies the semantics from mutiple objects to a single group of object, effectively making the existing single monitor semantic viable by simply changing monitors to monitor collections.
403\begin{multicols}{2}
404Waiter
405\begin{pseudo}[numbers=left]
406acquire A
407        acquire A & B
408                wait A & B
409        release A & B
410release A
411\end{pseudo}
412
413\columnbreak
414
415Signaller
416\begin{pseudo}[numbers=left, firstnumber=6]
417acquire A
418        acquire A & B
419                signal A & B
420        release A & B
421        //Secretly keep B here
422release A
423//Wakeup waiter and transfer A & B
424\end{pseudo}
425\end{multicols}
426However, this solution can become much more complicated depending on what is executed while secretly holding B (at line 10). Indeed, nothing prevents a user from signalling monitor A on a different condition variable:
427\newpage
428\begin{multicols}{2}
429Thread 1
430\begin{pseudo}[numbers=left, firstnumber=1]
431acquire A
432        acquire A & B
433                wait A & B
434        release A & B
435release A
436\end{pseudo}
437
438Thread 2
439\begin{pseudo}[numbers=left, firstnumber=6]
440acquire A
441        wait A
442release A
443\end{pseudo}
444
445\columnbreak
446
447Thread 3
448\begin{pseudo}[numbers=left, firstnumber=10]
449acquire A
450        acquire A & B
451                signal A & B
452        release A & B
453        //Secretly keep B here
454        signal A
455release A
456//Wakeup thread 1 or 2?
457//Who wakes up the other thread?
458\end{pseudo}
459\end{multicols}
460
461The goal in this solution is to avoid the need to transfer ownership of a subset of the condition monitors. However, this goal is unreacheable in the previous example. Depending on the order of signals (line 12 and 15) two cases can happen.
462
463\paragraph{Case 1: thread 1 goes first.} In this case, the problem is that monitor A needs to be passed to thread 2 when thread 1 is done with it.
464\paragraph{Case 2: thread 2 goes first.} In this case, the problem is that monitor B needs to be passed to thread 1, which can be done directly or using thread 2 as an intermediate.
465\\
466
467Note that ordering is not determined by a race condition but by whether signalled threads are enqueued in FIFO or FILO order. However, regardless of the answer, users can move line 15 before line 11 and get the reverse effect.
468
469In both cases, the threads need to be able to distinguish on a per monitor basis which ones need to be released and which ones need to be transferred. Which means monitors cannot be handled as a single homogenous group.
470
471\subsubsection{Dependency graphs}
472In the Listing 1 pseudo-code, there is a solution which statisfies both barging prevention and mutual exclusion. If ownership of both monitors is transferred to the waiter when the signaller releases A and then the waiter transfers back ownership of A when it releases it then the problem is solved. Dynamically finding the correct order is therefore the second possible solution. The problem it encounters is that it effectively boils down to resolving a dependency graph of ownership requirements. Here even the simplest of code snippets requires two transfers and it seems to increase in a manner closer to polynomial. For example, the following code, which is just a direct extension to three monitors, requires at least three ownership transfer and has multiple solutions:
473
474\begin{multicols}{2}
475\begin{pseudo}
476acquire A
477        acquire B
478                acquire C
479                        wait A & B & C
480                release C
481        release B
482release A
483\end{pseudo}
484
485\columnbreak
486
487\begin{pseudo}
488acquire A
489        acquire B
490                acquire C
491                        signal A & B & C
492                release C
493        release B
494release A
495\end{pseudo}
496\end{multicols}
497Resolving dependency graph being a complex and expensive endeavour, this solution is not the preffered one.
498
499\subsubsection{Partial signalling} \label{partial-sig}
500Finally, the solution that is chosen for \CFA is to use partial signalling. Consider the following case:
501
502\begin{multicols}{2}
503\begin{pseudo}[numbers=left]
504acquire A
505        acquire A & B
506                wait A & B
507        release A & B
508release A
509\end{pseudo}
510
511\columnbreak
512
513\begin{pseudo}[numbers=left, firstnumber=6]
514acquire A
515        acquire A & B
516                signal A & B
517        release A & B
518        // ... More code
519release A
520\end{pseudo}
521\end{multicols}
522The partial signalling solution transfers ownership of monitor B at lines 10 but does not wake the waiting thread since it is still using monitor A. Only when it reaches line 11 does it actually wakeup the waiting thread. This solution has the benefit that complexity is encapsulated into only two actions, passing monitors to the next owner when they should be release and conditionnaly waking threads if all conditions are met. Contrary to the other solutions, this solution quickly hits an upper bound on complexity of implementation.
523
524% ======================================================================
525% ======================================================================
526\subsection{Signalling: Now or Later}
527% ======================================================================
528% ======================================================================
529An important note is that, until now, signalling a monitor was a delayed operation. The ownership of the monitor is transferred only when the monitor would have otherwise been released, not at the point of the \code{signal} statement. However, in some cases, it may be more convenient for users to immediately transfer ownership to the thread that is waiting for cooperation, which is achieved using the \code{signal_block} routine\footnote{name to be discussed}.
530
531For example here is an example highlighting the difference in behaviour:
532\begin{center}
533\begin{tabular}{|c|c|}
534\code{signal} & \code{signal_block} \\
535\hline
536\begin{cfacode}
537monitor M { int val; };
538
539void foo(M & mutex m ) {
540        m.val++;
541        sout| "Foo:" | m.val |endl;
542
543        wait( c );
544
545        m.val++;
546        sout| "Foo:" | m.val |endl;
547}
548
549void bar(M & mutex m ) {
550        m.val++;
551        sout| "Bar:" | m.val |endl;
552
553        signal( c );
554
555        m.val++;
556        sout| "Bar:" | m.val |endl;
557}
558\end{cfacode}&\begin{cfacode}
559monitor M { int val; };
560
561void foo(M & mutex m ) {
562        m.val++;
563        sout| "Foo:" | m.val |endl;
564
565        wait( c );
566
567        m.val++;
568        sout| "Foo:" | m.val |endl;
569}
570
571void bar(M & mutex m ) {
572        m.val++;
573        sout| "Bar:" | m.val |endl;
574
575        signal_block( c );
576
577        m.val++;
578        sout| "Bar:" | m.val |endl;
579}
580\end{cfacode}
581\end{tabular}
582\end{center}
583Assuming that \code{val} is initialized at 0, that each routine are called from seperate thread and that \code{foo} is always called first. The previous code would yield the following output:
584
585\begin{center}
586\begin{tabular}{|c|c|}
587\code{signal} & \code{signal_block} \\
588\hline
589\begin{pseudo}
590Foo: 0
591Bar: 1
592Bar: 2
593Foo: 3
594\end{pseudo}&\begin{pseudo}
595Foo: 0
596Bar: 1
597Foo: 2
598Bar: 3
599\end{pseudo}
600\end{tabular}
601\end{center}
602
603As mentionned, \code{signal} only transfers ownership once the current critical section exits, resulting in the second "Bar" line to be printed before the second "Foo" line. On the other hand, \code{signal_block} immediately transfers ownership to \code{bar}, causing an inversion of output. Obviously this means that \code{signal_block} is a blocking call, which will only be resumed once the signalled function exits the critical section.
604
605% ======================================================================
606% ======================================================================
607\subsection{Internal scheduling: Implementation} \label{inschedimpl}
608% ======================================================================
609% ======================================================================
610There are several challenges specific to \CFA when implementing internal scheduling. These challenges are direct results of \gls{group-acquire} and loose object definitions. These two constraints are to root cause of most design decisions in the implementation of internal scheduling. Furthermore, to avoid the head-aches of dynamically allocating memory in a concurrent environment, the internal-scheduling design is entirely free of mallocs and other dynamic memory allocation scheme. This is to avoid the chicken and egg problem of having a memory allocator that relies on the threading system and a threading system that relies on the runtime. This extra goal, means that memory management is a constant concern in the design of the system.
611
612The main memory concern for concurrency is queues. All blocking operations are made by parking threads onto queues. These queues need to be intrinsic\cit to avoid the need memory allocation. This entails that all the fields needed to keep track of all needed information. Since internal scheduling can use an unbound amount of memory (depending on \gls{group-acquire}) statically defining information information in the intrusive fields of threads is insufficient. The only variable sized container that does not require memory allocation is the callstack, which is heavily used in the implementation of internal scheduling. Particularly the GCC extension variable length arrays which is used extensively.
613
614Since stack allocation is based around scope, the first step of the implementation is to identify the scopes that are available to store the information, and which of these can have a variable length. In the case of external scheduling, the threads and the condition both allow a fixed amount of memory to be stored, while mutex-routines and the actual blocking call allow for an unbound amount (though adding too much to the mutex routine stack size can become expansive faster).
615
616The following figure is the traditionnal illustration of a monitor :
617
618\begin{center}
619{\resizebox{0.4\textwidth}{!}{\input{monitor}}}
620\end{center}
621
622For \CFA, the previous picture does not have support for blocking multiple monitors on a single condition. To support \gls{group-acquire} two changes to this picture are required. First, it doesn't make sense to tie the condition to a single monitor since blocking two monitors as one would require arbitrarily picking a monitor to hold the condition. Secondly, the object waiting on the conditions and AS-stack cannot simply contain the waiting thread since a single thread can potentially wait on multiple monitors. As mentionned in section \ref{inschedimpl}, the handling in multiple monitors is done by partially passing, which entails that each concerned monitor needs to have a node object. However, for waiting on the condition, since all threads need to wait together, a single object needs to be queued in the condition. Moving out the condition and updating the node types yields :
623
624\begin{center}
625{\resizebox{0.8\textwidth}{!}{\input{int_monitor}}}
626\end{center}
627
628\newpage
629
630This picture and the proper entry and leave algorithms is the fundamental implementation of internal scheduling.
631
632\begin{multicols}{2}
633Entry
634\begin{pseudo}[numbers=left]
635if monitor is free
636        enter
637elif I already own the monitor
638        continue
639else
640        block
641increment recursion
642
643\end{pseudo}
644\columnbreak
645Exit
646\begin{pseudo}[numbers=left, firstnumber=8]
647decrement recursion
648if recursion == 0
649        if signal_stack not empty
650                set_owner to thread
651                if all monitors ready
652                        wake-up thread
653
654        if entry queue not empty
655                wake-up thread
656\end{pseudo}
657\end{multicols}
658
659Some important things to notice about the exit routine. The solution discussed in \ref{inschedimpl} can be seen on line 11 of the previous pseudo code. Basically, the solution boils down to having a seperate data structure for the condition queue and the AS-stack, and unconditionally transferring ownership of the monitors but only unblocking the thread when the last monitor has trasnferred ownership. This solution is safe as well as preventing any potential barging.
660
661% ======================================================================
662% ======================================================================
663\section{External scheduling} \label{extsched}
664% ======================================================================
665% ======================================================================
666An alternative to internal scheduling is to use external scheduling.
667\begin{center}
668\begin{tabular}{|c|c|}
669Internal Scheduling & External Scheduling \\
670\hline
671\begin{ucppcode}
672_Monitor Semaphore {
673        condition c;
674        bool inUse;
675public:
676        void P() {
677                if(inUse) wait(c);
678                inUse = true;
679        }
680        void V() {
681                inUse = false;
682                signal(c);
683        }
684}
685\end{ucppcode}&\begin{ucppcode}
686_Monitor Semaphore {
687
688        bool inUse;
689public:
690        void P() {
691                if(inUse) _Accept(V);
692                inUse = true;
693        }
694        void V() {
695                inUse = false;
696
697        }
698}
699\end{ucppcode}
700\end{tabular}
701\end{center}
702This method is more constrained and explicit, which may help users tone down the undeterministic nature of concurrency. Indeed, as the following examples demonstrates, external scheduling allows users to wait for events from other threads without the concern of unrelated events occuring. External scheduling can generally be done either in terms of control flow (e.g., \uC) or in terms of data (e.g. Go). Of course, both of these paradigms have their own strenghts and weaknesses but for this project control-flow semantics were chosen to stay consistent with the rest of the languages semantics. Two challenges specific to \CFA arise when trying to add external scheduling with loose object definitions and multi-monitor routines. The previous example shows a simple use \code{_Accept} versus \code{wait}/\code{signal} and its advantages. Note that while other languages often use \code{accept} as the core external scheduling keyword, \CFA uses \code{waitfor} to prevent name collisions with existing socket APIs.
703
704In the case of internal scheduling, the call to \code{wait} only guarantees that \code{V} is the last routine to access the monitor. This entails that the routine \code{V} may have acquired mutual exclusion several times while routine \code{P} was waiting. On the other hand, external scheduling guarantees that while routine \code{P} was waiting, no routine other than \code{V} could acquire the monitor.
705
706% ======================================================================
707% ======================================================================
708\subsection{Loose object definitions}
709% ======================================================================
710% ======================================================================
711In \uC, monitor declarations include an exhaustive list of monitor operations. Since \CFA is not object oriented it becomes both more difficult to implement but also less clear for the user:
712
713\begin{cfacode}
714        monitor A {};
715
716        void f(A & mutex a);
717        void f(int a, float b);
718        void g(A & mutex a) {
719                waitfor(f); // Less obvious which f() to wait for
720        }
721\end{cfacode}
722
723Furthermore, external scheduling is an example where implementation constraints become visible from the interface. Indeed, since there is no hard limit to the number of threads trying to acquire a monitor concurrently, performance is a significant concern. Here is the pseudo code for the entering phase of a monitor:
724
725\begin{center}
726\begin{tabular}{l}
727\begin{pseudo}
728        if monitor is free
729                enter
730        elif I already own the monitor
731                continue
732        elif monitor accepts me
733                enter
734        else
735                block
736\end{pseudo}
737\end{tabular}
738\end{center}
739
740For the fist two conditions, it is easy to implement a check that can evaluate the condition in a few instruction. However, a fast check for \pscode{monitor accepts me} is much harder to implement depending on the constraints put on the monitors. Indeed, monitors are often expressed as an entry queue and some acceptor queue as in the following figure:
741
742\begin{center}
743{\resizebox{0.4\textwidth}{!}{\input{monitor}}}
744\end{center}
745
746There are other alternatives to these pictures but in the case of this picture implementing a fast accept check is relatively easy. Indeed simply updating a bitmask when the acceptor queue changes is enough to have a check that executes in a single instruction, even with a fairly large number (e.g. 128) of mutex members. This technique cannot be used in \CFA because it relies on the fact that the monitor type declares all the acceptable routines. For OO languages this does not compromise much since monitors already have an exhaustive list of member routines. However, for \CFA this is not the case; routines can be added to a type anywhere after its declaration. Its important to note that the bitmask approach does not actually require an exhaustive list of routines, but it requires a dense unique ordering of routines with an upper-bound and that ordering must be consistent across translation units.
747The alternative would be to have a picture more like this one:
748
749\begin{center}
750{\resizebox{0.4\textwidth}{!}{\input{ext_monitor}}}
751\end{center}
752
753Not storing the queues inside the monitor means that the storage can vary between routines, allowing for more flexibility and extensions. Storing an array of function-pointers would solve the issue of uniquely identifying acceptable routines. However, the single instruction bitmask compare has been replaced by dereferencing a pointer followed by a linear search. Furthermore, supporting nested external scheduling may now require additionnal searches on calls to waitfor to check if a routine is already queued in.
754
755At this point we must make a decision between flexibility and performance. Many design decisions in \CFA achieve both flexibility and performance, for example polymorphic routines add significant flexibility but inlining them means the optimizer can easily remove any runtime cost. Here however, the cost of flexibility cannot be trivially removed. In the end, the most flexible approach has been chosen since it allows users to write programs that would otherwise be prohibitively hard to write. This is based on the assumption that writing fast but inflexible locks is closer to a solved problems than writing locks that are as flexible as external scheduling in \CFA.
756
757Another aspect to consider is what happens if multiple overloads of the same routine are used. For the time being it is assumed that multiple overloads of the same routine are considered as distinct routines. However, this could easily be extended in the future.
758
759% ======================================================================
760% ======================================================================
761\subsection{Multi-monitor scheduling}
762% ======================================================================
763% ======================================================================
764
765External scheduling, like internal scheduling, becomes orders of magnitude more complex when we start introducing multi-monitor syntax. Even in the simplest possible case some new semantics need to be established:
766\begin{cfacode}
767        mutex struct A {};
768
769        mutex struct B {};
770
771        void g(A & mutex a, B & mutex b) {
772                waitfor(f); //ambiguous, which monitor
773        }
774\end{cfacode}
775
776The obvious solution is to specify the correct monitor as follows:
777
778\begin{cfacode}
779        mutex struct A {};
780
781        mutex struct B {};
782
783        void g(A & mutex a, B & mutex b) {
784                waitfor( f, b );
785        }
786\end{cfacode}
787
788This is unambiguous. Both locks will be acquired and kept, when routine \code{f} is called the lock for monitor \code{b} will be temporarily transferred from \code{g} to \code{f} (while \code{g} still holds lock \code{a}). This behavior can be extended to multi-monitor waitfor statment as follows.
789
790\begin{cfacode}
791        mutex struct A {};
792
793        mutex struct B {};
794
795        void g(A & mutex a, B & mutex b) {
796                waitfor( f, a, b);
797        }
798\end{cfacode}
799
800Note that the set of monitors passed to the \code{waitfor} statement must be entirely contained in the set of monitor already acquired in the routine. \code{waitfor} used in any other context is Undefined Behaviour.
801
802An important behavior to note is that what happens when set of monitors only match partially :
803
804\begin{cfacode}
805        mutex struct A {};
806
807        mutex struct B {};
808
809        void g(A & mutex a, B & mutex b) {
810                waitfor(f, a, b);
811        }
812
813        A a1, a2;
814        B b;
815
816        void foo() {
817                g(a1, b);
818        }
819
820        void bar() {
821                f(a2, b);
822        }
823\end{cfacode}
824
825While the equivalent can happen when using internal scheduling, the fact that conditions are branded on first use means that users have to use two different condition variables. In both cases, partially matching monitor sets will not wake-up the waiting thread. It is also important to note that in the case of external scheduling, as for routine calls, the order of parameters is important; \code{waitfor(f,a,b)} and \code{waitfor(f,b,a)} are to distinct waiting condition.
826
827% ======================================================================
828% ======================================================================
829\subsection{Implementation Details: External scheduling queues}
830% ======================================================================
831% ======================================================================
832To support multi-monitor external scheduling means that some kind of entry-queues must be used that is aware of both monitors. However, acceptable routines must be aware of the entry queues which means they must be stored inside at least one of the monitors that will be acquired. This in turn adds the requirement a systematic algorithm of disambiguating which queue is relavant regardless of user ordering. The proposed algorithm is to fall back on monitors lock ordering and specify that the monitor that is acquired first is the lock with the relevant entry queue. This assumes that the lock acquiring order is static for the lifetime of all concerned objects but that is a reasonable constraint. This algorithm choice has two consequences, the entry queue of the highest priority monitor is no longer a true FIFO queue and the queue of the lowest priority monitor is both required and probably unused. The queue can no longer be a FIFO queue because instead of simply containing the waiting threads in order arrival, they also contain the second mutex. Therefore, another thread with the same highest priority monitor but a different lowest priority monitor may arrive first but enter the critical section after a thread with the correct pairing. Secondly, since it may not be known at compile time which monitor will be the lowest priority monitor, every monitor needs to have the correct queues even though it is probable that half the multi-monitor queues will go unused for the entire duration of the program.
833
834% ======================================================================
835% ======================================================================
836\section{Other concurrency tools}
837% ======================================================================
838% ======================================================================
839% \TODO
Note: See TracBrowser for help on using the repository browser.