source: doc/papers/concurrency/Paper.tex @ ca0f061f

aaron-thesisarm-ehcleanup-dtorsjacob/cs343-translationjenkins-sandboxnew-astnew-ast-unique-exprpersistent-indexer
Last change on this file since ca0f061f was ca0f061f, checked in by Peter A. Buhr <pabuhr@…>, 3 years ago

second introduction update

  • Property mode set to 100644
File size: 131.3 KB
Line 
1\documentclass[AMA,STIX1COL]{WileyNJD-v2}
2
3\articletype{RESEARCH ARTICLE}%
4
5\received{XXXXX}
6\revised{XXXXX}
7\accepted{XXXXX}
8
9\raggedbottom
10
11%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12
13% Latex packages used in the document.
14
15\usepackage{epic,eepic}
16\usepackage{xspace}
17\usepackage{comment}
18\usepackage{upquote}                                            % switch curled `'" to straight
19\usepackage{listings}                                           % format program code
20\usepackage[labelformat=simple,aboveskip=0pt,farskip=0pt]{subfig}
21\renewcommand{\thesubfigure}{(\Alph{subfigure})}
22\captionsetup{justification=raggedright,singlelinecheck=false}
23\usepackage{dcolumn}                                            % align decimal points in tables
24\usepackage{capt-of}
25
26\hypersetup{breaklinks=true}
27\definecolor{OliveGreen}{cmyk}{0.64 0 0.95 0.40}
28\definecolor{Mahogany}{cmyk}{0 0.85 0.87 0.35}
29\definecolor{Plum}{cmyk}{0.50 1 0 0}
30
31\usepackage[pagewise]{lineno}
32\renewcommand{\linenumberfont}{\scriptsize\sffamily}
33
34\renewcommand{\topfraction}{0.8}                        % float must be greater than X of the page before it is forced onto its own page
35\renewcommand{\bottomfraction}{0.8}                     % float must be greater than X of the page before it is forced onto its own page
36\renewcommand{\floatpagefraction}{0.8}          % float must be greater than X of the page before it is forced onto its own page
37\renewcommand{\textfraction}{0.0}                       % the entire page maybe devoted to floats with no text on the page at all
38
39\lefthyphenmin=3                                                        % hyphen only after 4 characters
40\righthyphenmin=3
41
42%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
43
44% Names used in the document.
45
46\newcommand{\CFAIcon}{\textsf{C}\raisebox{\depth}{\rotatebox{180}{\textsf{A}}}\xspace} % Cforall symbolic name
47\newcommand{\CFA}{\protect\CFAIcon}             % safe for section/caption
48\newcommand{\CFL}{\textrm{Cforall}\xspace}      % Cforall symbolic name
49\newcommand{\Celeven}{\textrm{C11}\xspace}      % C11 symbolic name
50\newcommand{\CC}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}\xspace} % C++ symbolic name
51\newcommand{\CCeleven}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}11\xspace} % C++11 symbolic name
52\newcommand{\CCfourteen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}14\xspace} % C++14 symbolic name
53\newcommand{\CCseventeen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}17\xspace} % C++17 symbolic name
54\newcommand{\CCtwenty}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}20\xspace} % C++20 symbolic name
55\newcommand{\Csharp}{C\raisebox{-0.7ex}{\Large$^\sharp$}\xspace} % C# symbolic name
56
57%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
58
59\newcommand{\Textbf}[2][red]{{\color{#1}{\textbf{#2}}}}
60\newcommand{\Emph}[2][red]{{\color{#1}\textbf{\emph{#2}}}}
61\newcommand{\uC}{$\mu$\CC}
62\newcommand{\TODO}[1]{{\Textbf{#1}}}
63
64%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
65
66% Default underscore is too low and wide. Cannot use lstlisting "literate" as replacing underscore
67% removes it as a variable-name character so keywords in variables are highlighted. MUST APPEAR
68% AFTER HYPERREF.
69%\DeclareTextCommandDefault{\textunderscore}{\leavevmode\makebox[1.2ex][c]{\rule{1ex}{0.1ex}}}
70\renewcommand{\textunderscore}{\leavevmode\makebox[1.2ex][c]{\rule{1ex}{0.075ex}}}
71
72\renewcommand*{\thefootnote}{\Alph{footnote}} % hack because fnsymbol does not work
73%\renewcommand*{\thefootnote}{\fnsymbol{footnote}}
74
75\makeatletter
76% parindent is relative, i.e., toggled on/off in environments like itemize, so store the value for
77% use rather than use \parident directly.
78\newlength{\parindentlnth}
79\setlength{\parindentlnth}{\parindent}
80
81\newcommand{\LstBasicStyle}[1]{{\lst@basicstyle{\lst@basicstyle{#1}}}}
82\newcommand{\LstKeywordStyle}[1]{{\lst@basicstyle{\lst@keywordstyle{#1}}}}
83\newcommand{\LstCommentStyle}[1]{{\lst@basicstyle{\lst@commentstyle{#1}}}}
84
85\newlength{\gcolumnposn}                                        % temporary hack because lstlisting does not handle tabs correctly
86\newlength{\columnposn}
87\setlength{\gcolumnposn}{3.5in}
88\setlength{\columnposn}{\gcolumnposn}
89
90\newcommand{\C}[2][\@empty]{\ifx#1\@empty\else\global\setlength{\columnposn}{#1}\global\columnposn=\columnposn\fi\hfill\makebox[\textwidth-\columnposn][l]{\lst@basicstyle{\LstCommentStyle{#2}}}}
91\newcommand{\CRT}{\global\columnposn=\gcolumnposn}
92
93% Denote newterms in particular font and index them without particular font and in lowercase, e.g., \newterm{abc}.
94% The option parameter provides an index term different from the new term, e.g., \newterm[\texttt{abc}]{abc}
95% The star version does not lowercase the index information, e.g., \newterm*{IBM}.
96\newcommand{\newtermFontInline}{\emph}
97\newcommand{\newterm}{\@ifstar\@snewterm\@newterm}
98\newcommand{\@newterm}[2][\@empty]{\lowercase{\def\temp{#2}}{\newtermFontInline{#2}}\ifx#1\@empty\index{\temp}\else\index{#1@{\protect#2}}\fi}
99\newcommand{\@snewterm}[2][\@empty]{{\newtermFontInline{#2}}\ifx#1\@empty\index{#2}\else\index{#1@{\protect#2}}\fi}
100
101% Latin abbreviation
102\newcommand{\abbrevFont}{\textit}                       % set empty for no italics
103\@ifundefined{eg}{
104\newcommand{\EG}{\abbrevFont{e}\abbrevFont{g}}
105\newcommand*{\eg}{%
106        \@ifnextchar{,}{\EG}%
107                {\@ifnextchar{:}{\EG}%
108                        {\EG,\xspace}}%
109}}{}%
110\@ifundefined{ie}{
111\newcommand{\IE}{\abbrevFont{i}\abbrevFont{e}}
112\newcommand*{\ie}{%
113        \@ifnextchar{,}{\IE}%
114                {\@ifnextchar{:}{\IE}%
115                        {\IE,\xspace}}%
116}}{}%
117\@ifundefined{etc}{
118\newcommand{\ETC}{\abbrevFont{etc}}
119\newcommand*{\etc}{%
120        \@ifnextchar{.}{\ETC}%
121        {\ETC.\xspace}%
122}}{}%
123\@ifundefined{etal}{
124\newcommand{\ETAL}{\abbrevFont{et}~\abbrevFont{al}}
125\newcommand*{\etal}{%
126        \@ifnextchar{.}{\protect\ETAL}%
127                {\protect\ETAL.\xspace}%
128}}{}%
129\@ifundefined{viz}{
130\newcommand{\VIZ}{\abbrevFont{viz}}
131\newcommand*{\viz}{%
132        \@ifnextchar{.}{\VIZ}%
133                {\VIZ.\xspace}%
134}}{}%
135\makeatother
136
137\newenvironment{cquote}
138               {\list{}{\lstset{resetmargins=true,aboveskip=0pt,belowskip=0pt}\topsep=3pt\parsep=0pt\leftmargin=\parindentlnth\rightmargin\leftmargin}%
139                \item\relax}
140               {\endlist}
141
142%\newenvironment{cquote}{%
143%\list{}{\lstset{resetmargins=true,aboveskip=0pt,belowskip=0pt}\topsep=3pt\parsep=0pt\leftmargin=\parindentlnth\rightmargin\leftmargin}%
144%\item\relax%
145%}{%
146%\endlist%
147%}% cquote
148
149% CFA programming language, based on ANSI C (with some gcc additions)
150\lstdefinelanguage{CFA}[ANSI]{C}{
151        morekeywords={
152                _Alignas, _Alignof, __alignof, __alignof__, asm, __asm, __asm__, __attribute, __attribute__,
153                auto, _Bool, catch, catchResume, choose, _Complex, __complex, __complex__, __const, __const__,
154                coroutine, disable, dtype, enable, exception, __extension__, fallthrough, fallthru, finally,
155                __float80, float80, __float128, float128, forall, ftype, _Generic, _Imaginary, __imag, __imag__,
156                inline, __inline, __inline__, __int128, int128, __label__, monitor, mutex, _Noreturn, one_t, or,
157                otype, restrict, __restrict, __restrict__, __signed, __signed__, _Static_assert, thread,
158                _Thread_local, throw, throwResume, timeout, trait, try, ttype, typeof, __typeof, __typeof__,
159                virtual, __volatile, __volatile__, waitfor, when, with, zero_t},
160        moredirectives={defined,include_next}%
161}
162
163\lstset{
164language=CFA,
165columns=fullflexible,
166basicstyle=\linespread{0.9}\sf,                                                 % reduce line spacing and use sanserif font
167stringstyle=\tt,                                                                                % use typewriter font
168tabsize=5,                                                                                              % N space tabbing
169xleftmargin=\parindentlnth,                                                             % indent code to paragraph indentation
170%mathescape=true,                                                                               % LaTeX math escape in CFA code $...$
171escapechar=\$,                                                                                  % LaTeX escape in CFA code
172keepspaces=true,                                                                                %
173showstringspaces=false,                                                                 % do not show spaces with cup
174showlines=true,                                                                                 % show blank lines at end of code
175aboveskip=4pt,                                                                                  % spacing above/below code block
176belowskip=3pt,
177% replace/adjust listing characters that look bad in sanserif
178literate={-}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.1ex}}}}1 {^}{\raisebox{0.6ex}{$\scriptstyle\land\,$}}1
179        {~}{\raisebox{0.3ex}{$\scriptstyle\sim\,$}}1 % {`}{\ttfamily\upshape\hspace*{-0.1ex}`}1
180        {<}{\textrm{\textless}}1 {>}{\textrm{\textgreater}}1
181        {<-}{$\leftarrow$}2 {=>}{$\Rightarrow$}2 {->}{\makebox[1ex][c]{\raisebox{0.5ex}{\rule{0.8ex}{0.075ex}}}\kern-0.2ex{\textrm{\textgreater}}}2,
182moredelim=**[is][\color{red}]{`}{`},
183}% lstset
184
185% uC++ programming language, based on ANSI C++
186\lstdefinelanguage{uC++}[ANSI]{C++}{
187        morekeywords={
188                _Accept, _AcceptReturn, _AcceptWait, _Actor, _At, _CatchResume, _Cormonitor, _Coroutine, _Disable,
189                _Else, _Enable, _Event, _Finally, _Monitor, _Mutex, _Nomutex, _PeriodicTask, _RealTimeTask,
190                _Resume, _Select, _SporadicTask, _Task, _Timeout, _When, _With, _Throw},
191}
192\lstdefinelanguage{Golang}{
193        morekeywords=[1]{package,import,func,type,struct,return,defer,panic,recover,select,var,const,iota,},
194        morekeywords=[2]{string,uint,uint8,uint16,uint32,uint64,int,int8,int16,int32,int64,
195                bool,float32,float64,complex64,complex128,byte,rune,uintptr, error,interface},
196        morekeywords=[3]{map,slice,make,new,nil,len,cap,copy,close,true,false,delete,append,real,imag,complex,chan,},
197        morekeywords=[4]{for,break,continue,range,goto,switch,case,fallthrough,if,else,default,},
198        morekeywords=[5]{Println,Printf,Error,},
199        sensitive=true,
200        morecomment=[l]{//},
201        morecomment=[s]{/*}{*/},
202        morestring=[b]',
203        morestring=[b]",
204        morestring=[s]{`}{`},
205}
206
207\lstnewenvironment{cfa}[1][]
208{\lstset{#1}}
209{}
210\lstnewenvironment{C++}[1][]                            % use C++ style
211{\lstset{language=C++,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
212{}
213\lstnewenvironment{uC++}[1][]
214{\lstset{#1}}
215{}
216\lstnewenvironment{Go}[1][]
217{\lstset{#1}}
218{}
219
220% inline code @...@
221\lstMakeShortInline@%
222
223\let\OLDthebibliography\thebibliography
224\renewcommand\thebibliography[1]{
225  \OLDthebibliography{#1}
226  \setlength{\parskip}{0pt}
227  \setlength{\itemsep}{4pt plus 0.3ex}
228}
229
230\title{\texorpdfstring{Advanced Control-flow in \protect\CFA}{Advanced Control-flow in Cforall}}
231
232\author[1]{Thierry Delisle}
233\author[1]{Peter A. Buhr*}
234\authormark{DELISLE \textsc{et al.}}
235
236\address[1]{\orgdiv{Cheriton School of Computer Science}, \orgname{University of Waterloo}, \orgaddress{\state{Waterloo, ON}, \country{Canada}}}
237
238\corres{*Peter A. Buhr, Cheriton School of Computer Science, University of Waterloo, 200 University Avenue West, Waterloo, ON, N2L 3G1, Canada. \email{pabuhr{\char`\@}uwaterloo.ca}}
239
240\fundingInfo{Natural Sciences and Engineering Research Council of Canada}
241
242\abstract[Summary]{
243\CFA is a modern, polymorphic, non-object-oriented, backwards-compatible extension of the C programming language.
244This paper discusses the advanced control-flow features in \CFA, which include concurrency and parallelism, and its supporting runtime system.
245These features are created from scratch as ISO C's concurrency is low-level and unimplemented, so C programmers continue to rely on the C pthreads library.
246\CFA provides high-level control-flow mechanisms, like coroutines and user-level threads, and monitors for mutual exclusion and synchronization.
247A unique contribution of this work is allowing multiple monitors to be safely acquired \emph{simultaneously} (deadlock free), while integrating this capability with all monitor synchronization mechanisms.
248All features respect the expectations of C programmers, while being fully integrate with the \CFA polymorphic type-system and other language features.
249Experimental results show comparable performance of the new features with similar mechanisms in other concurrent programming-languages.
250}%
251
252\keywords{coroutines, concurrency, parallelism, threads, monitors, runtime, C, \CFA (Cforall)}
253
254
255\begin{document}
256\linenumbers                                            % comment out to turn off line numbering
257
258\maketitle
259
260
261\section{Introduction}
262
263This paper discusses the design of advanced, high-level control-flow extensions (especially concurrency and parallelism) in \CFA and its runtime.
264\CFA is a modern, polymorphic, non-object-oriented\footnote{
265\CFA has features often associated with object-oriented programming languages, such as constructors, destructors, virtuals and simple inheritance.
266However, functions \emph{cannot} be nested in structures, so there is no lexical binding between a structure and set of functions (member/method) implemented by an implicit \lstinline@this@ (receiver) parameter.},
267backwards-compatible extension of the C programming language~\cite{Moss18}.
268Within the \CFA framework, new control-flow features were created from scratch.
269ISO \Celeven defines only a subset of the \CFA extensions, and with respect to concurrency~\cite[\S~7.26]{C11}, the features are largely wrappers for a subset of the pthreads library~\cite{Butenhof97,Pthreads}.
270Furthermore, \Celeven and pthreads concurrency is basic, based on thread fork/join in a function and a few locks, which is low-level and error prone;
271no high-level language concurrency features exist.
272Interestingly, almost a decade after publication of the \Celeven standard, neither gcc-8, clang-8 nor msvc-19 (most recent versions) support the \Celeven include @threads.h@, indicating little interest in the C concurrency approach.
273Finally, while the \Celeven standard does not state a concurrent threading-model, the historical association with pthreads suggests the threading model is kernel-level threading (1:1)~\cite{ThreadModel}.
274
275In contrast, there has been a renewed interest during the past decade in user-level (M:N, green) threading in old and new programming languages.
276As multi-core hardware became available in the 1980/90s, both user and kernel threading were examined.
277Kernel threading was chosen, largely because of its simplicity and fit with the simpler operating systems and hardware architectures at the time, which gave it a performance advantage~\cite{Drepper03}.
278Libraries like pthreads were developed for C, and the Solaris operating-system switched from user (JDK 1.1~\cite{JDK1.1}) to kernel threads.
279As a result, languages like Java, Scala~\cite{Scala}, Objective-C~\cite{obj-c-book}, \CCeleven~\cite{C11}, and C\#~\cite{Csharp} adopted the 1:1 kernel-threading model, with a variety of presentation mechanisms.
280From 2000 onwards, languages like Go~\cite{Go}, Erlang~\cite{Erlang}, Haskell~\cite{Haskell}, D~\cite{D}, and \uC~\cite{uC++,uC++book} have championed the M:N user-threading model, and many user-threading libraries have appeared~\cite{Qthreads,MPC,BoostThreads}, including putting green threads back into Java~\cite{Quasar}.
281The main argument for user-level threading is that they are lighter weight than kernel threads (locking and context switching do not cross the kernel boundary), so there is less restriction on programming styles that encourage large numbers of threads performing smaller work-units to facilitate load balancing by the runtime~\cite{Verch12}.
282As well, user-threading facilitates a simpler concurrency approach using thread objects that leverage sequential patterns versus events with call-backs~\cite{vonBehren03}.
283Finally, performant user-threading implementations (both time and space) are largely competitive with direct kernel-threading implementations, while achieving the programming advantages of high concurrency levels and safety.
284
285A further effort over the past decade is the development of language memory-models to deal with the conflict between certain language features and compiler/hardware optimizations.
286This issue can be rephrased as some features are pervasive (language and runtime) and cannot be safely added via a library to prevent invalidation by sequential optimizations~\cite{Buhr95a,Boehm05}.
287The consequence is that a language must be cognizant of these features and provide sufficient tools to program around any safety issues.
288For example, C created the @volatile@ qualifier to provide correct execution for @setjmp@/@logjmp@ (concurrency came later).
289The simplest solution is to provide a handful of complex qualifiers and functions (e.g., @volatile@ and atomics) allowing programmers to write consistent/race-free programs, often in the sequentially-consistent memory-model~\cite{Boehm12}.
290
291While having a sufficient memory-model allows sound libraries to be constructed, writing these libraries can quickly become awkward and error prone, and using these low-level libraries has the same issues.
292Essentially, using low-level explicit locks is the concurrent equivalent of assembler programming.
293Just as most assembler programming is replaced with programming in a high-level language, explicit locks can be replaced with high-level concurrency constructs in a programming language.
294The goal is to get the compiler to check for correct usage and follow any complex coding conventions implicitly.
295The drawback is that language constructs may preclude certain specialized techniques, therefore introducing inefficiency or inhibiting concurrency.
296For most concurrent programs, these drawbacks are insignificant in comparison to the speed of composition, and subsequent reliability and maintainability of the high-level concurrent program.
297(The same is true for high-level programming versus assembler programming.)
298Only very rarely should it be necessary to drop down to races and/or explicit locks to apply a specialized technique to achieve maximum speed or concurrency.
299As stated, this observation applies to non-concurrent forms of complex control-flow, like exception handling and coroutines.
300
301Adapting the programming language allows matching the control-flow model with the programming-language style, versus adapting to one general (sound) library/paradigm.
302For example, it is possible to provide exceptions, coroutines, monitors, and tasks as specialized types in an object-oriented language, integrating these constructs to allow leveraging the type-system (static type-checking) and all other object-oriented capabilities~\cite{uC++}.
303It is also possible to leverage call/return for blocking communication via new control structures, versus switching to alternative communication paradigms, like channels or message passing.
304As well, user threading is often a complementary feature, allowing light-weight threading to match with low-cost objects, while hiding the application/kernel boundary.
305User threading also allows layering of implicit concurrency models (no explicit thread creation), such executors, data-flow, actors, into a single language, so programmers can chose the model that best fits an algorithm.\footnote{
306All implicit concurrency models have explicit threading in their implementation, and hence, can be build from explicit threading;
307however, the reverse is seldom true, i.e., given implicit concurrency, e.g., actors, it is virtually impossible to create explicit concurrency, e.g., blocking thread objects.}
308Finally, with extended language features and user-level threading it is possible to discretely fold locking and non-blocking I/O multiplexing into the language's I/O libraries, so threading implicitly dovetails with the I/O subsystem.
309
310\CFA embraces language extensions and user-level threading to provide advanced control-flow and concurrency.
311We attempt to show the \CFA extensions and runtime are demonstrably better than those proposed for \CC and other concurrent, imperative programming languages.
312The contributions of this work are:
313\begin{itemize}
314\item
315allowing multiple monitors to be safely acquired \emph{simultaneously} (deadlock free), while seamlessly integrating this capability with all monitor synchronization mechanisms.
316\item
317all control-flow features respect the expectations of C programmers, with statically type-safe interfaces that integrate with the \CFA polymorphic type-system and other language features.
318\item
319experimental results show comparable performance of the new features with similar mechanisms in other concurrent programming-languages.
320\end{itemize}
321
322\begin{comment}
323This paper provides a minimal concurrency \newterm{Application Program Interface} (API) that is simple, efficient and can be used to build other concurrency features.
324While the simplest concurrency system is a thread and a lock, this low-level approach is hard to master.
325An easier approach for programmers is to support higher-level constructs as the basis of concurrency.
326Indeed, for highly-productive concurrent-programming, high-level approaches are much more popular~\cite{Hochstein05}.
327Examples of high-level approaches are jobs (thread pool)~\cite{TBB}, implicit threading~\cite{OpenMP}, monitors~\cite{Java}, channels~\cite{CSP,Go}, and message passing~\cite{Erlang,MPI}.
328
329The following terminology is used.
330A \newterm{thread} is a fundamental unit of execution that runs a sequence of code and requires a stack to maintain state.
331Multiple simultaneous threads give rise to \newterm{concurrency}, which requires locking to ensure access to shared data and safe communication.
332\newterm{Locking}, and by extension \newterm{locks}, are defined as a mechanism to prevent progress of threads to provide safety.
333\newterm{Parallelism} is running multiple threads simultaneously.
334Parallelism implies \emph{actual} simultaneous execution, where concurrency only requires \emph{apparent} simultaneous execution.
335As such, parallelism only affects performance, which is observed through differences in space and/or time at runtime.
336Hence, there are two problems to be solved: concurrency and parallelism.
337While these two concepts are often combined, they are distinct, requiring different tools~\cite[\S~2]{Buhr05a}.
338Concurrency tools handle mutual exclusion and synchronization, while parallelism tools handle performance, cost, and resource utilization.
339
340The proposed concurrency API is implemented in a dialect of C, called \CFA (pronounced C-for-all).
341The paper discusses how the language features are added to the \CFA translator with respect to parsing, semantics, and type checking, and the corresponding high-performance runtime-library to implement the concurrent features.
342\end{comment}
343
344
345\begin{comment}
346\section{\CFA Overview}
347
348The following is a quick introduction to the \CFA language, specifically tailored to the features needed to support concurrency.
349Extended versions and explanation of the following code examples are available at the \CFA website~\cite{Cforall} or in Moss~\etal~\cite{Moss18}.
350
351\CFA is a non-object-oriented extension of ISO-C, and hence, supports all C paradigms.
352Like C, the building blocks of \CFA are structures and routines.
353Virtually all of the code generated by the \CFA translator respects C memory layouts and calling conventions.
354While \CFA is not object oriented, lacking the concept of a receiver (\eg @this@) and nominal inheritance-relationships, C has a notion of objects: ``region of data storage in the execution environment, the contents of which can represent values''~\cite[3.15]{C11}.
355While some object-oriented features appear in \CFA, they are independent capabilities, allowing \CFA to adopt them while maintaining a procedural paradigm.
356
357
358\subsection{References}
359
360\CFA provides multi-level rebindable references, as an alternative to pointers, which significantly reduces syntactic noise.
361\begin{cfa}
362int x = 1, y = 2, z = 3;
363int * p1 = &x, ** p2 = &p1,  *** p3 = &p2,      $\C{// pointers to x}$
364    `&` r1 = x,   `&&` r2 = r1,   `&&&` r3 = r2;        $\C{// references to x}$
365int * p4 = &z, `&` r4 = z;
366
367*p1 = 3; **p2 = 3; ***p3 = 3;       // change x
368r1 =  3;     r2 = 3;      r3 = 3;        // change x: implicit dereferences *r1, **r2, ***r3
369**p3 = &y; *p3 = &p4;                // change p1, p2
370`&`r3 = &y; `&&`r3 = &`&`r4;             // change r1, r2: cancel implicit dereferences (&*)**r3, (&(&*)*)*r3, &(&*)r4
371\end{cfa}
372A reference is a handle to an object, like a pointer, but is automatically dereferenced the specified number of levels.
373Referencing (address-of @&@) a reference variable cancels one of the implicit dereferences, until there are no more implicit references, after which normal expression behaviour applies.
374
375
376\subsection{\texorpdfstring{\protect\lstinline{with} Statement}{with Statement}}
377\label{s:WithStatement}
378
379Heterogeneous data is aggregated into a structure/union.
380To reduce syntactic noise, \CFA provides a @with@ statement (see Pascal~\cite[\S~4.F]{Pascal}) to elide aggregate field-qualification by opening a scope containing the field identifiers.
381\begin{cquote}
382\vspace*{-\baselineskip}%???
383\lstDeleteShortInline@%
384\begin{cfa}
385struct S { char c; int i; double d; };
386struct T { double m, n; };
387// multiple aggregate parameters
388\end{cfa}
389\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
390\begin{cfa}
391void f( S & s, T & t ) {
392        `s.`c; `s.`i; `s.`d;
393        `t.`m; `t.`n;
394}
395\end{cfa}
396&
397\begin{cfa}
398void f( S & s, T & t ) `with ( s, t )` {
399        c; i; d;                // no qualification
400        m; n;
401}
402\end{cfa}
403\end{tabular}
404\lstMakeShortInline@%
405\end{cquote}
406Object-oriented programming languages only provide implicit qualification for the receiver.
407
408In detail, the @with@-statement syntax is:
409\begin{cfa}
410$\emph{with-statement}$:
411        'with' '(' $\emph{expression-list}$ ')' $\emph{compound-statement}$
412\end{cfa}
413and may appear as the body of a routine or nested within a routine body.
414Each expression in the expression-list provides a type and object.
415The type must be an aggregate type.
416(Enumerations are already opened.)
417The object is the implicit qualifier for the open structure-fields.
418All expressions in the expression list are opened in parallel within the compound statement, which is different from Pascal, which nests the openings from left to right.
419
420
421\subsection{Overloading}
422
423\CFA maximizes the ability to reuse names via overloading to aggressively address the naming problem.
424Both variables and routines may be overloaded, where selection is based on number and types of returns and arguments (as in Ada~\cite{Ada}).
425\newpage
426\vspace*{-2\baselineskip}%???
427\begin{cquote}
428\begin{cfa}
429// selection based on type
430\end{cfa}
431\lstDeleteShortInline@%
432\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
433\begin{cfa}
434const short int `MIN` = -32768;
435const int `MIN` = -2147483648;
436const long int `MIN` = -9223372036854775808L;
437\end{cfa}
438&
439\begin{cfa}
440short int si = `MIN`;
441int i = `MIN`;
442long int li = `MIN`;
443\end{cfa}
444\end{tabular}
445\begin{cfa}
446// selection based on type and number of parameters
447\end{cfa}
448\begin{tabular}{@{}l@{\hspace{2.7\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
449\begin{cfa}
450void `f`( void );
451void `f`( char );
452void `f`( int, double );
453\end{cfa}
454&
455\begin{cfa}
456`f`();
457`f`( 'a' );
458`f`( 3, 5.2 );
459\end{cfa}
460\end{tabular}
461\begin{cfa}
462// selection based on type and number of returns
463\end{cfa}
464\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
465\begin{cfa}
466char `f`( int );
467double `f`( int );
468[char, double] `f`( int );
469\end{cfa}
470&
471\begin{cfa}
472char c = `f`( 3 );
473double d = `f`( 3 );
474[d, c] = `f`( 3 );
475\end{cfa}
476\end{tabular}
477\lstMakeShortInline@%
478\end{cquote}
479Overloading is important for \CFA concurrency since the runtime system relies on creating different types to represent concurrency objects.
480Therefore, overloading eliminates long prefixes and other naming conventions to prevent name clashes.
481As seen in Section~\ref{s:Concurrency}, routine @main@ is heavily overloaded.
482As another example, variable overloading is useful in the parallel semantics of the @with@ statement for fields with the same name:
483\begin{cfa}
484struct S { int `i`; int j; double m; } s;
485struct T { int `i`; int k; int m; } t;
486with ( s, t ) {
487        j + k;                                                                  $\C{// unambiguous, s.j + t.k}$
488        m = 5.0;                                                                $\C{// unambiguous, s.m = 5.0}$
489        m = 1;                                                                  $\C{// unambiguous, t.m = 1}$
490        int a = m;                                                              $\C{// unambiguous, a = t.m }$
491        double b = m;                                                   $\C{// unambiguous, b = s.m}$
492        int c = `s.i` + `t.i`;                                  $\C{// unambiguous, qualification}$
493        (double)m;                                                              $\C{// unambiguous, cast s.m}$
494}
495\end{cfa}
496For parallel semantics, both @s.i@ and @t.i@ are visible with the same type, so only @i@ is ambiguous without qualification.
497
498
499\subsection{Operators}
500
501Overloading also extends to operators.
502Operator-overloading syntax creates a routine name with an operator symbol and question marks for the operands:
503\begin{cquote}
504\lstDeleteShortInline@%
505\begin{tabular}{@{}ll@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
506\begin{cfa}
507int ++?(int op);
508int ?++(int op);
509int `?+?`(int op1, int op2);
510int ?<=?(int op1, int op2);
511int ?=? (int & op1, int op2);
512int ?+=?(int & op1, int op2);
513\end{cfa}
514&
515\begin{cfa}
516// unary prefix increment
517// unary postfix increment
518// binary plus
519// binary less than
520// binary assignment
521// binary plus-assignment
522\end{cfa}
523&
524\begin{cfa}
525struct S { int i, j; };
526S `?+?`( S op1, S op2) { // add two structures
527        return (S){op1.i + op2.i, op1.j + op2.j};
528}
529S s1 = {1, 2}, s2 = {2, 3}, s3;
530s3 = s1 `+` s2;         // compute sum: s3 == {2, 5}
531\end{cfa}
532\end{tabular}
533\lstMakeShortInline@%
534\end{cquote}
535
536
537\subsection{Constructors / Destructors}
538
539Object lifetime is a challenge in non-managed programming languages.
540\CFA responds with \CC-like constructors and destructors, using a different operator-overloading syntax.
541\begin{cfa}
542struct VLA { int len, * data; };                        $\C{// variable length array of integers}$
543void ?{}( VLA & vla ) with ( vla ) { len = 10;  data = alloc( len ); }  $\C{// default constructor}$
544void ?{}( VLA & vla, int size, char fill ) with ( vla ) { len = size;  data = alloc( len, fill ); } // initialization
545void ?{}( VLA & vla, VLA other ) { vla.len = other.len;  vla.data = other.data; } $\C{// copy, shallow}$
546void ^?{}( VLA & vla ) with ( vla ) { free( data ); } $\C{// destructor}$
547{
548        VLA  x,            y = { 20, 0x01 },     z = y; $\C{// z points to y}$
549        // $\LstCommentStyle{\color{red}\ \ \ x\{\};\ \ \ \ \ \ \ \ \ y\{ 20, 0x01 \};\ \ \ \ \ \ \ \ \ \ z\{ z, y \};\ \ \ \ \ \ \ implicit calls}$
550        ^x{};                                                                   $\C{// deallocate x}$
551        x{};                                                                    $\C{// reallocate x}$
552        z{ 5, 0xff };                                                   $\C{// reallocate z, not pointing to y}$
553        ^y{};                                                                   $\C{// deallocate y}$
554        y{ x };                                                                 $\C{// reallocate y, points to x}$
555        x{};                                                                    $\C{// reallocate x, not pointing to y}$
556}       //  $\LstCommentStyle{\color{red}\^{}z\{\};\ \ \^{}y\{\};\ \ \^{}x\{\};\ \ \ implicit calls}$
557\end{cfa}
558Like \CC, construction is implicit on allocation (stack/heap) and destruction is implicit on deallocation.
559The object and all their fields are constructed/destructed.
560\CFA also provides @new@ and @delete@ as library routines, which behave like @malloc@ and @free@, in addition to constructing and destructing objects:
561\begin{cfa}
562{
563        ... struct S s = {10}; ...                              $\C{// allocation, call constructor}$
564}                                                                                       $\C{// deallocation, call destructor}$
565struct S * s = new();                                           $\C{// allocation, call constructor}$
566...
567delete( s );                                                            $\C{// deallocation, call destructor}$
568\end{cfa}
569\CFA concurrency uses object lifetime as a means of mutual exclusion and/or synchronization.
570
571
572\subsection{Parametric Polymorphism}
573\label{s:ParametricPolymorphism}
574
575The signature feature of \CFA is parametric-polymorphic routines~\cite{Cforall} with routines generalized using a @forall@ clause (giving the language its name), which allow separately compiled routines to support generic usage over multiple types.
576For example, the following sum routine works for any type that supports construction from 0 and addition:
577\begin{cfa}
578forall( otype T | { void `?{}`( T *, zero_t ); T `?+?`( T, T ); } ) // constraint type, 0 and +
579T sum( T a[$\,$], size_t size ) {
580        `T` total = { `0` };                                    $\C{// initialize by 0 constructor}$
581        for ( size_t i = 0; i < size; i += 1 )
582                total = total `+` a[i];                         $\C{// select appropriate +}$
583        return total;
584}
585S sa[5];
586int i = sum( sa, 5 );                                           $\C{// use S's 0 construction and +}$
587\end{cfa}
588Type variables can be @otype@ or @dtype@.
589@otype@ refers to a \emph{complete type}, \ie, a type with size, alignment, default constructor, copy constructor, destructor, and assignment operator.
590@dtype@ refers to an \emph{incomplete type}, \eg, void or a forward-declared type.
591The builtin types @zero_t@ and @one_t@ overload constant 0 and 1 for a new types, where both 0 and 1 have special meaning in C.
592
593\CFA provides \newterm{traits} to name a group of type assertions, where the trait name allows specifying the same set of assertions in multiple locations, preventing repetition mistakes at each routine declaration:
594\begin{cfa}
595trait `sumable`( otype T ) {
596        void `?{}`( T &, zero_t );                              $\C{// 0 literal constructor}$
597        T `?+?`( T, T );                                                $\C{// assortment of additions}$
598        T ?+=?( T &, T );
599        T ++?( T & );
600        T ?++( T & );
601};
602forall( otype T `| sumable( T )` )                      $\C{// use trait}$
603T sum( T a[$\,$], size_t size );
604\end{cfa}
605
606Using the return type for overload discrimination, it is possible to write a type-safe @alloc@ based on the C @malloc@:
607\begin{cfa}
608forall( dtype T | sized(T) ) T * alloc( void ) { return (T *)malloc( sizeof(T) ); }
609int * ip = alloc();                                                     $\C{// select type and size from left-hand side}$
610double * dp = alloc();
611struct S {...} * sp = alloc();
612\end{cfa}
613where the return type supplies the type/size of the allocation, which is impossible in most type systems.
614\end{comment}
615
616
617\section{Coroutines: A Stepping Stone}\label{coroutine}
618
619Advanced controlWhile the focus of this discussion is concurrency and parallelism, it is important to address coroutines, which are a significant building block of a concurrency system (but not concurrent among themselves).
620Coroutines are generalized routines allowing execution to be temporarily suspended and later resumed.
621Hence, unlike a normal routine, a coroutine may not terminate when it returns to its caller, allowing it to be restarted with the values and execution location present at the point of suspension.
622This capability is accomplished via the coroutine's stack, where suspend/resume context switch among stacks.
623Because threading design-challenges are present in coroutines, their design effort is relevant, and this effort can be easily exposed to programmers giving them a useful new programming paradigm because a coroutine handles the class of problems that need to retain state between calls, \eg plugins, device drivers, and finite-state machines.
624Therefore, the two fundamental features of the core \CFA coroutine-API are independent call-stacks and @suspend@/@resume@ operations.
625
626For example, a problem made easier with coroutines is unbounded generators, \eg generating an infinite sequence of Fibonacci numbers
627\begin{displaymath}
628\mathsf{fib}(n) = \left \{
629\begin{array}{ll}
6300                                       & n = 0         \\
6311                                       & n = 1         \\
632\mathsf{fib}(n-1) + \mathsf{fib}(n-2)   & n \ge 2       \\
633\end{array}
634\right.
635\end{displaymath}
636where Figure~\ref{f:C-fibonacci} shows conventional approaches for writing a Fibonacci generator in C.
637Figure~\ref{f:GlobalVariables} illustrates the following problems: unique unencapsulated global variables necessary to retain state between calls, only one Fibonacci generator, and execution state must be explicitly retained via explicit state variables.
638Figure~\ref{f:ExternalState} addresses these issues: unencapsulated program global variables become encapsulated structure variables, unique global variables are replaced by multiple Fibonacci objects, and explicit execution state is removed by precomputing the first two Fibonacci numbers and returning $\mathsf{fib}(n-2)$.
639
640\begin{figure}
641\centering
642\newbox\myboxA
643\begin{lrbox}{\myboxA}
644\begin{cfa}[aboveskip=0pt,belowskip=0pt]
645`int f1, f2, state = 1;`   // single global variables
646int fib() {
647        int fn;
648        `switch ( state )` {  // explicit execution state
649          case 1: fn = 0;  f1 = fn;  state = 2;  break;
650          case 2: fn = 1;  f2 = f1;  f1 = fn;  state = 3;  break;
651          case 3: fn = f1 + f2;  f2 = f1;  f1 = fn;  break;
652        }
653        return fn;
654}
655int main() {
656
657        for ( int i = 0; i < 10; i += 1 ) {
658                printf( "%d\n", fib() );
659        }
660}
661\end{cfa}
662\end{lrbox}
663
664\newbox\myboxB
665\begin{lrbox}{\myboxB}
666\begin{cfa}[aboveskip=0pt,belowskip=0pt]
667#define FIB_INIT `{ 0, 1 }`
668typedef struct { int f2, f1; } Fib;
669int fib( Fib * f ) {
670
671        int ret = f->f2;
672        int fn = f->f1 + f->f2;
673        f->f2 = f->f1; f->f1 = fn;
674
675        return ret;
676}
677int main() {
678        Fib f1 = FIB_INIT, f2 = FIB_INIT;
679        for ( int i = 0; i < 10; i += 1 ) {
680                printf( "%d %d\n", fib( &f1 ), fib( &f2 ) );
681        }
682}
683\end{cfa}
684\end{lrbox}
685
686\subfloat[3 States: global variables]{\label{f:GlobalVariables}\usebox\myboxA}
687\qquad
688\subfloat[1 State: external variables]{\label{f:ExternalState}\usebox\myboxB}
689\caption{C Fibonacci Implementations}
690\label{f:C-fibonacci}
691
692\bigskip
693
694\newbox\myboxA
695\begin{lrbox}{\myboxA}
696\begin{cfa}[aboveskip=0pt,belowskip=0pt]
697`coroutine` Fib { int fn; };
698void main( Fib & fib ) with( fib ) {
699        int f1, f2;
700        fn = 0;  f1 = fn;  `suspend()`;
701        fn = 1;  f2 = f1;  f1 = fn;  `suspend()`;
702        for ( ;; ) {
703                fn = f1 + f2;  f2 = f1;  f1 = fn;  `suspend()`;
704        }
705}
706int next( Fib & fib ) with( fib ) {
707        `resume( fib );`
708        return fn;
709}
710int main() {
711        Fib f1, f2;
712        for ( int i = 1; i <= 10; i += 1 ) {
713                sout | next( f1 ) | next( f2 );
714        }
715}
716\end{cfa}
717\end{lrbox}
718\newbox\myboxB
719\begin{lrbox}{\myboxB}
720\begin{cfa}[aboveskip=0pt,belowskip=0pt]
721`coroutine` Fib { int ret; };
722void main( Fib & f ) with( fib ) {
723        int fn, f1 = 1, f2 = 0;
724        for ( ;; ) {
725                ret = f2;
726
727                fn = f1 + f2;  f2 = f1;  f1 = fn; `suspend();`
728        }
729}
730int next( Fib & fib ) with( fib ) {
731        `resume( fib );`
732        return ret;
733}
734
735
736
737
738
739
740\end{cfa}
741\end{lrbox}
742\subfloat[3 States, internal variables]{\label{f:Coroutine3States}\usebox\myboxA}
743\qquad\qquad
744\subfloat[1 State, internal variables]{\label{f:Coroutine1State}\usebox\myboxB}
745\caption{\CFA Coroutine Fibonacci Implementations}
746\label{f:cfa-fibonacci}
747\end{figure}
748
749Using a coroutine, it is possible to express the Fibonacci formula directly without any of the C problems.
750Figure~\ref{f:Coroutine3States} creates a @coroutine@ type, @`coroutine` Fib { int fn; }@, which provides communication, @fn@, for the \newterm{coroutine main}, @main@, which runs on the coroutine stack, and possibly multiple interface routines, \eg @next@.
751Like the structure in Figure~\ref{f:ExternalState}, the coroutine type allows multiple instances, where instances of this type are passed to the (overloaded) coroutine main.
752The coroutine main's stack holds the state for the next generation, @f1@ and @f2@, and the code represents the three states in the Fibonacci formula via the three suspend points, to context switch back to the caller's @resume@.
753The interface routine @next@, takes a Fibonacci instance and context switches to it using @resume@;
754on restart, the Fibonacci field, @fn@, contains the next value in the sequence, which is returned.
755The first @resume@ is special because it allocates the coroutine stack and cocalls its coroutine main on that stack;
756when the coroutine main returns, its stack is deallocated.
757Hence, @Fib@ is an object at creation, transitions to a coroutine on its first resume, and transitions back to an object when the coroutine main finishes.
758Figure~\ref{f:Coroutine1State} shows the coroutine version of the C version in Figure~\ref{f:ExternalState}.
759Coroutine generators are called \newterm{output coroutines} because values are only returned.
760
761Figure~\ref{f:CFAFmt} shows an \newterm{input coroutine}, @Format@, for restructuring text into groups of characters of fixed-size blocks.
762For example, the input of the left is reformatted into the output on the right.
763\begin{quote}
764\tt
765\begin{tabular}{@{}l|l@{}}
766\multicolumn{1}{c|}{\textbf{\textrm{input}}} & \multicolumn{1}{c}{\textbf{\textrm{output}}} \\
767abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz
768&
769\begin{tabular}[t]{@{}lllll@{}}
770abcd    & efgh  & ijkl  & mnop  & qrst  \\
771uvwx    & yzab  & cdef  & ghij  & klmn  \\
772opqr    & stuv  & wxyz  &               &
773\end{tabular}
774\end{tabular}
775\end{quote}
776The example takes advantage of resuming a coroutine in the constructor to prime the loops so the first character sent for formatting appears inside the nested loops.
777The destructor provides a newline, if formatted text ends with a full line.
778Figure~\ref{f:CFmt} shows the C equivalent formatter, where the loops of the coroutine are flattened (linearized) and rechecked on each call because execution location is not retained between calls.
779(Linearized code is the bane of device drivers.)
780
781\begin{figure}
782\centering
783\newbox\myboxA
784\begin{lrbox}{\myboxA}
785\begin{cfa}[aboveskip=0pt,belowskip=0pt]
786`coroutine` Format {
787        char ch;   // used for communication
788        int g, b;  // global because used in destructor
789};
790void main( Format & fmt ) with( fmt ) {
791        for ( ;; ) {
792                for ( g = 0; g < 5; g += 1 ) {      // group
793                        for ( b = 0; b < 4; b += 1 ) { // block
794                                `suspend();`
795                                sout | ch;              // separator
796                        }
797                        sout | "  ";               // separator
798                }
799                sout | nl;
800        }
801}
802void ?{}( Format & fmt ) { `resume( fmt );` }
803void ^?{}( Format & fmt ) with( fmt ) {
804        if ( g != 0 || b != 0 ) sout | nl;
805}
806void format( Format & fmt ) {
807        `resume( fmt );`
808}
809int main() {
810        Format fmt;
811        eof: for ( ;; ) {
812                sin | fmt.ch;
813          if ( eof( sin ) ) break eof;
814                format( fmt );
815        }
816}
817\end{cfa}
818\end{lrbox}
819
820\newbox\myboxB
821\begin{lrbox}{\myboxB}
822\begin{cfa}[aboveskip=0pt,belowskip=0pt]
823struct Format {
824        char ch;
825        int g, b;
826};
827void format( struct Format * fmt ) {
828        if ( fmt->ch != -1 ) {      // not EOF ?
829                printf( "%c", fmt->ch );
830                fmt->b += 1;
831                if ( fmt->b == 4 ) {  // block
832                        printf( "  " );      // separator
833                        fmt->b = 0;
834                        fmt->g += 1;
835                }
836                if ( fmt->g == 5 ) {  // group
837                        printf( "\n" );     // separator
838                        fmt->g = 0;
839                }
840        } else {
841                if ( fmt->g != 0 || fmt->b != 0 ) printf( "\n" );
842        }
843}
844int main() {
845        struct Format fmt = { 0, 0, 0 };
846        for ( ;; ) {
847                scanf( "%c", &fmt.ch );
848          if ( feof( stdin ) ) break;
849                format( &fmt );
850        }
851        fmt.ch = -1;
852        format( &fmt );
853}
854\end{cfa}
855\end{lrbox}
856\subfloat[\CFA Coroutine]{\label{f:CFAFmt}\usebox\myboxA}
857\qquad
858\subfloat[C Linearized]{\label{f:CFmt}\usebox\myboxB}
859\caption{Formatting text into lines of 5 blocks of 4 characters.}
860\label{f:fmt-line}
861\end{figure}
862
863The previous examples are \newterm{asymmetric (semi) coroutine}s because one coroutine always calls a resuming routine for another coroutine, and the resumed coroutine always suspends back to its last resumer, similar to call/return for normal routines.
864However, @resume@ and @suspend@ context switch among existing stack-frames, rather than create new ones so there is no stack growth.
865\newterm{Symmetric (full) coroutine}s have a coroutine call to a resuming routine for another coroutine, and its coroutine main calls another resuming routine, which eventually forms a resuming-call cycle.
866(The trivial cycle is a coroutine resuming itself.)
867This control flow is similar to recursion for normal routines, but again there is no stack growth from the context switch.
868
869\begin{figure}
870\centering
871\lstset{language=CFA,escapechar={},moredelim=**[is][\protect\color{red}]{`}{`}}% allow $
872\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
873\begin{cfa}
874`coroutine` Prod {
875        Cons & c;
876        int N, money, receipt;
877};
878void main( Prod & prod ) with( prod ) {
879        // 1st resume starts here
880        for ( int i = 0; i < N; i += 1 ) {
881                int p1 = random( 100 ), p2 = random( 100 );
882                sout | p1 | " " | p2;
883                int status = delivery( c, p1, p2 );
884                sout | " $" | money | nl | status;
885                receipt += 1;
886        }
887        stop( c );
888        sout | "prod stops";
889}
890int payment( Prod & prod, int money ) {
891        prod.money = money;
892        `resume( prod );`
893        return prod.receipt;
894}
895void start( Prod & prod, int N, Cons &c ) {
896        &prod.c = &c;
897        prod.[N, receipt] = [N, 0];
898        `resume( prod );`
899}
900int main() {
901        Prod prod;
902        Cons cons = { prod };
903        start( prod, 5, cons );
904}
905\end{cfa}
906&
907\begin{cfa}
908`coroutine` Cons {
909        Prod & p;
910        int p1, p2, status;
911        _Bool done;
912};
913void ?{}( Cons & cons, Prod & p ) {
914        &cons.p = &p;
915        cons.[status, done ] = [0, false];
916}
917void ^?{}( Cons & cons ) {}
918void main( Cons & cons ) with( cons ) {
919        // 1st resume starts here
920        int money = 1, receipt;
921        for ( ; ! done; ) {
922                sout | p1 | " " | p2 | nl | " $" | money;
923                status += 1;
924                receipt = payment( p, money );
925                sout | " #" | receipt;
926                money += 1;
927        }
928        sout | "cons stops";
929}
930int delivery( Cons & cons, int p1, int p2 ) {
931        cons.[p1, p2] = [p1, p2];
932        `resume( cons );`
933        return cons.status;
934}
935void stop( Cons & cons ) {
936        cons.done = true;
937        `resume( cons );`
938}
939\end{cfa}
940\end{tabular}
941\caption{Producer / consumer: resume-resume cycle, bi-directional communication}
942\label{f:ProdCons}
943\end{figure}
944
945Figure~\ref{f:ProdCons} shows a producer/consumer symmetric-coroutine performing bi-directional communication.
946Since the solution involves a full-coroutining cycle, the program main creates one coroutine in isolation, passes this coroutine to its partner, and closes the cycle at the call to @start@.
947The @start@ routine communicates both the number of elements to be produced and the consumer into the producer's coroutine-structure.
948Then the @resume@ to @prod@ creates @prod@'s stack with a frame for @prod@'s coroutine main at the top, and context switches to it.
949@prod@'s coroutine main starts, creates local variables that are retained between coroutine activations, and executes $N$ iterations, each generating two random values, calling the consumer to deliver the values, and printing the status returned from the consumer.
950
951The producer call to @delivery@ transfers values into the consumer's communication variables, resumes the consumer, and returns the consumer status.
952For the first resume, @cons@'s stack is initialized, creating local variables retained between subsequent activations of the coroutine.
953The consumer iterates until the @done@ flag is set, prints the values delivered by the producer, increments status, and calls back to the producer via @payment@, and on return from @payment@, prints the receipt from the producer and increments @money@ (inflation).
954The call from the consumer to @payment@ introduces the cycle between producer and consumer.
955When @payment@ is called, the consumer copies values into the producer's communication variable and a resume is executed.
956The context switch restarts the producer at the point where it last context switched, so it continues in @delivery@ after the resume.
957
958@delivery@ returns the status value in @prod@'s coroutine main, where the status is printed.
959The loop then repeats calling @delivery@, where each call resumes the consumer coroutine.
960The context switch to the consumer continues in @payment@.
961The consumer increments and returns the receipt to the call in @cons@'s coroutine main.
962The loop then repeats calling @payment@, where each call resumes the producer coroutine.
963
964After iterating $N$ times, the producer calls @stop@.
965The @done@ flag is set to stop the consumer's execution and a resume is executed.
966The context switch restarts @cons@ in @payment@ and it returns with the last receipt.
967The consumer terminates its loops because @done@ is true, its @main@ terminates, so @cons@ transitions from a coroutine back to an object, and @prod@ reactivates after the resume in @stop@.
968@stop@ returns and @prod@'s coroutine main terminates.
969The program main restarts after the resume in @start@.
970@start@ returns and the program main terminates.
971
972
973\subsection{Coroutine Implementation}
974
975A significant implementation challenge for coroutines (and threads, see Section~\ref{threads}) is adding extra fields and executing code after/before the coroutine constructor/destructor and coroutine main to create/initialize/de-initialize/destroy extra fields and the stack.
976There are several solutions to this problem and the chosen option forced the \CFA coroutine design.
977
978Object-oriented inheritance provides extra fields and code in a restricted context, but it requires programmers to explicitly perform the inheritance:
979\begin{cfa}[morekeywords={class,inherits}]
980class mycoroutine inherits baseCoroutine { ... }
981\end{cfa}
982and the programming language (and possibly its tool set, \eg debugger) may need to understand @baseCoroutine@ because of the stack.
983Furthermore, the execution of constructors/destructors is in the wrong order for certain operations.
984For example, for threads if the thread is implicitly started, it must start \emph{after} all constructors, because the thread relies on a completely initialized object, but the inherited constructor runs \emph{before} the derived.
985
986An alternative is composition:
987\begin{cfa}
988struct mycoroutine {
989        ... // declarations
990        baseCoroutine dummy; // composition, last declaration
991}
992\end{cfa}
993which also requires an explicit declaration that must be the last one to ensure correct initialization order.
994However, there is nothing preventing wrong placement or multiple declarations.
995
996For coroutines as for threads, many implementations are based on routine pointers or routine objects~\cite{Butenhof97, C++14, MS:VisualC++, BoostCoroutines15}.
997For example, Boost implements coroutines in terms of four functor object-types:
998\begin{cfa}
999asymmetric_coroutine<>::pull_type
1000asymmetric_coroutine<>::push_type
1001symmetric_coroutine<>::call_type
1002symmetric_coroutine<>::yield_type
1003\end{cfa}
1004Similarly, the canonical threading paradigm is often based on routine pointers, \eg @pthreads@~\cite{Butenhof97}, \Csharp~\cite{Csharp}, Go~\cite{Go}, and Scala~\cite{Scala}.
1005However, the generic thread-handle (identifier) is limited (few operations), unless it is wrapped in a custom type.
1006\begin{cfa}
1007void mycor( coroutine_t cid, void * arg ) {
1008        int * value = (int *)arg;                               $\C{// type unsafe, pointer-size only}$
1009        // Coroutine body
1010}
1011int main() {
1012        int input = 0, output;
1013        coroutine_t cid = coroutine_create( &mycor, (void *)&input ); $\C{// type unsafe, pointer-size only}$
1014        coroutine_resume( cid, (void *)input, (void **)&output ); $\C{// type unsafe, pointer-size only}$
1015}
1016\end{cfa}
1017Since the custom type is simple to write in \CFA and solves several issues, added support for routine/lambda-based coroutines adds very little.
1018
1019Note, the type @coroutine_t@ must be an abstract handle to the coroutine, because the coroutine descriptor and its stack are non-copyable.
1020Copying the coroutine descriptor results in copies being out of date with the current state of the stack.
1021Correspondingly, copying the stack results is copies being out of date with the coroutine descriptor, and pointers in the stack being out of date to data on the stack.
1022(There is no mechanism in C to find all stack-specific pointers and update them as part of a copy.)
1023
1024The selected approach is to use language support by introducing a new kind of aggregate (structure):
1025\begin{cfa}
1026coroutine Fibonacci {
1027        int fn; // communication variables
1028};
1029\end{cfa}
1030The @coroutine@ keyword means the compiler (and tool set) can find and inject code where needed.
1031The downside of this approach is that it makes coroutine a special case in the language.
1032Users wanting to extend coroutines or build their own for various reasons can only do so in ways offered by the language.
1033Furthermore, implementing coroutines without language supports also displays the power of a programming language.
1034While this is ultimately the option used for idiomatic \CFA code, coroutines and threads can still be constructed without language support.
1035The reserved keyword simply eases use for the common case.
1036
1037Part of the mechanism to generalize coroutines is using a \CFA trait, which defines a coroutine as anything satisfying the trait @is_coroutine@, and this trait restricts the available set of coroutine-manipulation routines:
1038\begin{cfa}
1039trait is_coroutine( `dtype` T ) {
1040        void main( T & );
1041        coroutine_desc * get_coroutine( T & );
1042};
1043forall( `dtype` T | is_coroutine(T) ) void suspend( T & );
1044forall( `dtype` T | is_coroutine(T) ) void resume( T & );
1045\end{cfa}
1046The @dtype@ property provides no implicit copying operations and the @is_coroutine@ trait provides no explicit copying operations, so all coroutines must be passed by reference (pointer).
1047The routine definitions ensures there is a statically-typed @main@ routine that is the starting point (first stack frame) of a coroutine, and a mechanism to get (read) the currently executing coroutine handle.
1048The @main@ routine has no return value or additional parameters because the coroutine type allows an arbitrary number of interface routines with corresponding arbitrary typed input/output values versus fixed ones.
1049The advantage of this approach is that users can easily create different types of coroutines, \eg changing the memory layout of a coroutine is trivial when implementing the @get_coroutine@ routine, and possibly redefining @suspend@ and @resume@.
1050The \CFA keyword @coroutine@ implicitly implements the getter and forward declarations required for implementing the coroutine main:
1051\begin{cquote}
1052\begin{tabular}{@{}ccc@{}}
1053\begin{cfa}
1054coroutine MyCor {
1055        int value;
1056
1057};
1058\end{cfa}
1059&
1060{\Large $\Rightarrow$}
1061&
1062\begin{tabular}{@{}ccc@{}}
1063\begin{cfa}
1064struct MyCor {
1065        int value;
1066        coroutine_desc cor;
1067};
1068\end{cfa}
1069&
1070\begin{cfa}
1071static inline coroutine_desc *
1072get_coroutine( MyCor & this ) {
1073        return &this.cor;
1074}
1075\end{cfa}
1076&
1077\begin{cfa}
1078void main( MyCor * this );
1079
1080
1081
1082\end{cfa}
1083\end{tabular}
1084\end{tabular}
1085\end{cquote}
1086The combination of these two approaches allows an easy and concise specification to coroutining (and concurrency) for normal users, while more advanced users have tighter control on memory layout and initialization.
1087
1088
1089\section{Concurrency}
1090\label{s:Concurrency}
1091
1092At its core, concurrency is based on multiple call-stacks and scheduling threads executing on these stacks.
1093Multiple call stacks (or contexts) and a single thread of execution, called \newterm{coroutining}~\cite{Conway63,Marlin80}, does \emph{not} imply concurrency~\cite[\S~2]{Buhr05a}.
1094In coroutining, the single thread is self-scheduling across the stacks, so execution is deterministic, \ie the execution path from input to output is fixed and predictable.
1095A \newterm{stackless} coroutine executes on the caller's stack~\cite{Python} but this approach is restrictive, \eg preventing modularization and supporting only iterator/generator-style programming;
1096a \newterm{stackful} coroutine executes on its own stack, allowing full generality.
1097Only stackful coroutines are a stepping stone to concurrency.
1098
1099The transition to concurrency, even for execution with a single thread and multiple stacks, occurs when coroutines also context switch to a \newterm{scheduling oracle}, introducing non-determinism from the coroutine perspective~\cite[\S~3]{Buhr05a}.
1100Therefore, a minimal concurrency system is possible using coroutines (see Section \ref{coroutine}) in conjunction with a scheduler to decide where to context switch next.
1101The resulting execution system now follows a cooperative threading-model, called \newterm{non-preemptive scheduling}.
1102
1103Because the scheduler is special, it can either be a stackless or stackful coroutine.
1104For stackless, the scheduler performs scheduling on the stack of the current coroutine and switches directly to the next coroutine, so there is one context switch.
1105For stackful, the current coroutine switches to the scheduler, which performs scheduling, and it then switches to the next coroutine, so there are two context switches.
1106A stackful scheduler is often used for simplicity and security.
1107
1108Regardless of the approach used, a subset of concurrency related challenges start to appear.
1109For the complete set of concurrency challenges to occur, the missing feature is \newterm{preemption}, where context switching occurs randomly between any two instructions, often based on a timer interrupt, called \newterm{preemptive scheduling}.
1110While a scheduler introduces uncertainty in the order of execution, preemption introduces uncertainty about where context switches occur.
1111Interestingly, uncertainty is necessary for the runtime (operating) system to give the illusion of parallelism on a single processor and increase performance on multiple processors.
1112The reason is that only the runtime has complete knowledge about resources and how to best utilized them.
1113However, the introduction of unrestricted non-determinism results in the need for \newterm{mutual exclusion} and \newterm{synchronization} to restrict non-determinism for correctness;
1114otherwise, it is impossible to write meaningful programs.
1115Optimal performance in concurrent applications is often obtained by having as much non-determinism as correctness allows.
1116
1117An important missing feature in C is threading\footnote{While the C11 standard defines a \protect\lstinline@threads.h@ header, it is minimal and defined as optional.
1118As such, library support for threading is far from widespread.
1119At the time of writing the paper, neither \protect\lstinline@gcc@ nor \protect\lstinline@clang@ support \protect\lstinline@threads.h@ in their standard libraries.}.
1120In modern programming languages, a lack of threading is unacceptable~\cite{Sutter05, Sutter05b}, and therefore existing and new programming languages must have tools for writing efficient concurrent programs to take advantage of parallelism.
1121As 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.
1122Furthermore, because C is a system-level language, programmers expect to choose precisely which features they need and which cost they are willing to pay.
1123Hence, concurrent programs should be written using high-level mechanisms, and only step down to lower-level mechanisms when performance bottlenecks are encountered.
1124
1125
1126\subsection{Thread Interface}
1127\label{threads}
1128
1129Both user and kernel threads are supported, where user threads provide concurrency and kernel threads provide parallelism.
1130Like coroutines and for the same design reasons, the selected approach for user threads is to use language support by introducing a new kind of aggregate (structure) and a \CFA trait:
1131\begin{cquote}
1132\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1133\begin{cfa}
1134thread myThread {
1135        // communication variables
1136};
1137
1138
1139\end{cfa}
1140&
1141\begin{cfa}
1142trait is_thread( `dtype` T ) {
1143      void main( T & );
1144      thread_desc * get_thread( T & );
1145      void ^?{}( T & `mutex` );
1146};
1147\end{cfa}
1148\end{tabular}
1149\end{cquote}
1150(The qualifier @mutex@ for the destructor parameter is discussed in Section~\ref{s:Monitor}.)
1151Like a coroutine, the statically-typed @main@ routine is the starting point (first stack frame) of a user thread.
1152The difference is that a coroutine borrows a thread from its caller, so the first thread resuming a coroutine creates an instance of @main@;
1153whereas, a user thread receives its own thread from the runtime system, which starts in @main@ as some point after the thread constructor is run.\footnote{
1154The \lstinline@main@ routine is already a special routine in C, \ie where the program's initial thread begins, so it is a natural extension of this semantics to use overloading to declare \lstinline@main@s for user coroutines and threads.}
1155No return value or additional parameters are necessary for this routine because the task type allows an arbitrary number of interface routines with corresponding arbitrary typed input/output values.
1156
1157\begin{comment} % put in appendix with coroutine version ???
1158As such the @main@ routine of a thread can be defined as
1159\begin{cfa}
1160thread foo {};
1161
1162void main(foo & this) {
1163        sout | "Hello World!";
1164}
1165\end{cfa}
1166
1167In this example, threads of type @foo@ start execution in the @void main(foo &)@ routine, which prints @"Hello World!".@ While this paper encourages this approach to enforce strongly typed programming, users may prefer to use the routine-based thread semantics for the sake of simplicity.
1168With the static semantics it is trivial to write a thread type that takes a routine pointer as a parameter and executes it on its stack asynchronously.
1169\begin{cfa}
1170typedef void (*voidRtn)(int);
1171
1172thread RtnRunner {
1173        voidRtn func;
1174        int arg;
1175};
1176
1177void ?{}(RtnRunner & this, voidRtn inRtn, int arg) {
1178        this.func = inRtn;
1179        this.arg  = arg;
1180}
1181
1182void main(RtnRunner & this) {
1183        // thread starts here and runs the routine
1184        this.func( this.arg );
1185}
1186
1187void hello(/*unused*/ int) {
1188        sout | "Hello World!";
1189}
1190
1191int main() {
1192        RtnRunner f = {hello, 42};
1193        return 0?
1194}
1195\end{cfa}
1196A 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 \textbf{API}.
1197\end{comment}
1198
1199For user threads to be useful, it must be possible to start and stop the underlying thread, and wait for it to complete execution.
1200While using an API such as @fork@ and @join@ is relatively common, such an interface is awkward and unnecessary.
1201A simple approach is to use allocation/deallocation principles, and have threads implicitly @fork@ after construction and @join@ before destruction.
1202\begin{cfa}
1203thread World {};
1204void main( World & this ) {
1205        sout | "World!";
1206}
1207int main() {
1208        World w`[10]`;                                                  $\C{// implicit forks after creation}$
1209        sout | "Hello ";                                        $\C{// "Hello " and 10 "World!" printed concurrently}$
1210}                                                                                       $\C{// implicit joins before destruction}$
1211\end{cfa}
1212This semantics ensures a thread is started and stopped exactly once, eliminating some programming error, and scales to multiple threads for basic (termination) synchronization.
1213This tree-structure (lattice) create/delete from C block-structure is generalized 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.
1214\begin{cfa}
1215int main() {
1216        MyThread * heapLive;
1217        {
1218                MyThread blockLive;                                     $\C{// fork block-based thread}$
1219                heapLive = `new`( MyThread );           $\C{// fork heap-based thread}$
1220                ...
1221        }                                                                               $\C{// join block-based thread}$
1222        ...
1223        `delete`( heapLive );                                   $\C{// join heap-based thread}$
1224}
1225\end{cfa}
1226The heap-based approach allows arbitrary thread-creation topologies, with respect to fork/join-style concurrency.
1227
1228Figure~\ref{s:ConcurrentMatrixSummation} shows concurrently adding the rows of a matrix and then totalling the subtotals sequentially, after all the row threads have terminated.
1229The program uses heap-based threads because each thread needs different constructor values.
1230(Python provides a simple iteration mechanism to initialize array elements to different values allowing stack allocation.)
1231The allocation/deallocation pattern appears unusual because allocated objects are immediately deallocated without any intervening code.
1232However, for threads, the deletion provides implicit synchronization, which is the intervening code.
1233While the subtotals are added in linear order rather than completion order, which slightly inhibits concurrency, the computation is restricted by the critical-path thread (\ie the thread that takes the longest), and so any inhibited concurrency is very small as totalling the subtotals is trivial.
1234
1235\begin{figure}
1236\begin{cfa}
1237`thread` Adder { int * row, cols, & subtotal; } $\C{// communication variables}$
1238void ?{}( Adder & adder, int row[], int cols, int & subtotal ) {
1239    adder.[ row, cols, &subtotal ] = [ row, cols, &subtotal ];
1240}
1241void main( Adder & adder ) with( adder ) {
1242    subtotal = 0;
1243    for ( int c = 0; c < cols; c += 1 ) { subtotal += row[c]; }
1244}
1245int main() {
1246    const int rows = 10, cols = 1000;
1247    int matrix[rows][cols], subtotals[rows], total = 0;
1248    // read matrix
1249    Adder * adders[rows];
1250    for ( int r = 0; r < rows; r += 1 ) {       $\C{// start threads to sum rows}$
1251                adders[r] = `new( matrix[r], cols, &subtotals[r] );`
1252    }
1253    for ( int r = 0; r < rows; r += 1 ) {       $\C{// wait for threads to finish}$
1254                `delete( adders[r] );`                          $\C{// termination join}$
1255                total += subtotals[r];                          $\C{// total subtotal}$
1256    }
1257    sout | total;
1258}
1259\end{cfa}
1260\caption{Concurrent Matrix Summation}
1261\label{s:ConcurrentMatrixSummation}
1262\end{figure}
1263
1264
1265\section{Mutual Exclusion / Synchronization}
1266
1267Uncontrolled non-deterministic execution is meaningless.
1268To reestablish meaningful execution requires mechanisms to reintroduce determinism, \ie restrict non-determinism, called mutual exclusion and synchronization, where mutual exclusion is an access-control mechanism on data shared by threads, and synchronization is a timing relationship among threads~\cite[\S~4]{Buhr05a}.
1269Since many deterministic challenges appear with the use of mutable shared state, some languages/libraries disallow it, \eg Erlang~\cite{Erlang}, Haskell~\cite{Haskell}, Akka~\cite{Akka} (Scala).
1270In these paradigms, interaction among concurrent objects is performed by stateless message-passing~\cite{Thoth,Harmony,V-Kernel} or other paradigms closely related to networking concepts, \eg channels~\cite{CSP,Go}.
1271However, in call/return-based languages, these approaches force a clear distinction, \ie introduce a new programming paradigm between regular and concurrent computation, \eg routine call versus message passing.
1272Hence, a programmer must learn and manipulate two sets of design patterns.
1273While this distinction can be hidden away in library code, effective use of the library still has to take both paradigms into account.
1274In contrast, approaches based on stateful models more closely resemble the standard call/return programming-model, resulting in a single programming paradigm.
1275
1276At the lowest level, concurrent control is implemented by atomic operations, upon which different kinds of locking mechanisms are constructed, \eg semaphores~\cite{Dijkstra68b}, barriers, and path expressions~\cite{Campbell74}.
1277However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
1278A newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
1279While this approach is pursued in hardware~\cite{Nakaike15} and system languages, like \CC~\cite{Cpp-Transactions}, the performance and feature set is still too restrictive to be the main concurrency paradigm for system languages, which is why it is rejected as the core paradigm for concurrency in \CFA.
1280
1281One of the most natural, elegant, and efficient mechanisms for mutual exclusion and synchronization for shared-memory systems is the \emph{monitor}.
1282First proposed by Brinch Hansen~\cite{Hansen73} and later described and extended by C.A.R.~Hoare~\cite{Hoare74}, many concurrent programming-languages provide monitors as an explicit language construct: \eg 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}.
1283In addition, operating-system kernels and device drivers have a monitor-like structure, although they often use lower-level primitives such as mutex locks or semaphores to simulate monitors.
1284For these reasons, \CFA selected monitors as the core high-level concurrency-construct, upon which higher-level approaches can be easily constructed.
1285
1286
1287\subsection{Mutual Exclusion}
1288
1289A group of instructions manipulating a specific instance of shared data that must be performed atomically is called an (individual) \newterm{critical-section}~\cite{Dijkstra65}.
1290The generalization is called a \newterm{group critical-section}~\cite{Joung00}, where multiple tasks with the same session may use the resource simultaneously, but different sessions may not use the resource simultaneously.
1291The readers/writer problem~\cite{Courtois71} is an instance of a group critical-section, where readers have the same session and all writers have a unique session.
1292\newterm{Mutual exclusion} enforces that the correct kind and number of threads are using a critical section.
1293
1294However, many solutions exist for mutual exclusion, which vary in terms of performance, flexibility and ease of use.
1295Methods range from low-level locks, which are fast and flexible but require significant attention for correctness, to higher-level concurrency techniques, which sacrifice some performance to improve ease of use.
1296Ease of use comes by either guaranteeing some problems cannot occur, \eg deadlock free, or by offering a more explicit coupling between shared data and critical section.
1297For example, the \CC @std::atomic<T>@ offers an easy way to express mutual-exclusion on a restricted set of operations, \eg reading/writing, for numerical types.
1298However, a significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
1299Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
1300
1301
1302\subsection{Synchronization}
1303
1304Synchronization enforces relative ordering of execution, and synchronization tools provide numerous mechanisms to establish these timing relationships.
1305Low-level synchronization primitives offer good performance and flexibility at the cost of ease of use;
1306higher-level mechanisms often simplify usage by adding better coupling between synchronization and data, \eg message passing, or offering a simpler solution to otherwise involved challenges, \eg barrier lock.
1307Often synchronization is used to order access to a critical section, \eg ensuring a reader thread is the next kind of thread to enter a critical section.
1308If a writer thread is scheduled for next access, but another reader thread acquires the critical section first, that reader \newterm{barged}.
1309Barging can result in staleness/freshness problems, where a reader barges ahead of a writer and reads temporally stale data, or a writer barges ahead of another writer overwriting data with a fresh value preventing the previous value from ever being read (lost computation).
1310Preventing or detecting barging is an involved challenge with low-level locks, which can be made much easier by higher-level constructs.
1311This challenge is often split into two different approaches: barging avoidance and barging prevention.
1312Algorithms that allow a barger, but divert it until later using current synchronization state (flags), are avoiding the barger;
1313algorithms that preclude a barger from entering during synchronization in the critical section prevent barging completely.
1314Techniques like baton-passing locks~\cite{Andrews89} between threads instead of unconditionally releasing locks is an example of barging prevention.
1315
1316
1317\section{Monitor}
1318\label{s:Monitor}
1319
1320A \textbf{monitor} is a set of routines that ensure mutual exclusion when accessing shared state.
1321More precisely, a monitor is a programming technique that binds mutual exclusion to routine scope, as opposed to locks, where mutual-exclusion is defined by acquire/release calls, independent of lexical context (analogous to block and heap storage allocation).
1322The strong association with the call/return paradigm eases programmability, readability and maintainability, at a slight cost in flexibility and efficiency.
1323
1324Note, like coroutines/threads, both locks and monitors require an abstract handle to reference them, because at their core, both mechanisms are manipulating non-copyable shared-state.
1325Copying a lock is insecure because it is possible to copy an open lock and then use the open copy when the original lock is closed to simultaneously access the shared data.
1326Copying a monitor is secure because both the lock and shared data are copies, but copying the shared data is meaningless because it no longer represents a unique entity.
1327As for coroutines/tasks, the @dtype@ property provides no implicit copying operations and the @is_monitor@ trait provides no explicit copying operations, so all locks/monitors must be passed by reference (pointer).
1328\begin{cfa}
1329trait is_monitor( `dtype` T ) {
1330        monitor_desc * get_monitor( T & );
1331        void ^?{}( T & mutex );
1332};
1333\end{cfa}
1334
1335
1336\subsection{Mutex Acquisition}
1337\label{s:MutexAcquisition}
1338
1339While correctness implies a monitor's mutual exclusion is acquired and released, there are implementation options about when and where the locking/unlocking occurs.
1340(Much of this discussion also applies to basic locks.)
1341For example, a monitor may need to be passed through multiple helper routines before it becomes necessary to acquire the monitor mutual-exclusion.
1342\begin{cfa}[morekeywords=nomutex]
1343monitor Aint { int cnt; };                                      $\C{// atomic integer counter}$
1344void ?{}( Aint & `nomutex` this ) with( this ) { cnt = 0; } $\C{// constructor}$
1345int ?=?( Aint & `mutex`$\(_{opt}\)$ lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions}$
1346void ?{}( int & this, Aint & `mutex`$\(_{opt}\)$ v ) { this = v.cnt; }
1347int ?=?( int & lhs, Aint & `mutex`$\(_{opt}\)$ rhs ) with( rhs ) { lhs = cnt; }
1348int ++?( Aint & `mutex`$\(_{opt}\)$ this ) with( this ) { return ++cnt; } $\C{// increment}$
1349\end{cfa}
1350The @Aint@ constructor, @?{}@, uses the \lstinline[morekeywords=nomutex]@nomutex@ qualifier indicating mutual exclusion is unnecessary during construction because an object is inaccessible (private) until after it is initialized.
1351(While a constructor may publish its address into a global variable, doing so generates a race-condition.)
1352The conversion operators for initializing and assigning with a normal integer only need @mutex@, if reading/writing the implementation type is not atomic.
1353Finally, the prefix increment operato, @++?@, is normally @mutex@ to protect the incrementing from race conditions, unless there is an atomic increment instruction for the implementation type.
1354
1355The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics, which is similar to the \CC template @std::atomic@.
1356\begin{cfa}
1357Aint x, y, z;
1358++x; ++y; ++z;                                                          $\C{// safe increment by multiple threads}$
1359x = 2; y = 2; z = 2;                                            $\C{// conversions}$
1360int i = x, j = y, k = z;
1361i = x; j = y; k = z;
1362\end{cfa}
1363
1364For maximum usability, monitors have \newterm{multi-acquire} semantics allowing a thread to acquire it multiple times without deadlock.
1365\begin{cfa}
1366monitor M { ... } m;
1367void foo( M & mutex m ) { ... }                         $\C{// acquire mutual exclusion}$
1368void bar( M & mutex m ) {                                       $\C{// acquire mutual exclusion}$
1369        ... `foo( m );` ...                                             $\C{// reacquire mutual exclusion}$
1370}
1371`bar( m );`                                                                     $\C{// nested monitor call}$
1372\end{cfa}
1373
1374The benefit of mandatory monitor qualifiers is self-documentation, but requiring both @mutex@ and \lstinline[morekeywords=nomutex]@nomutex@ for all monitor parameters is redundant.
1375Instead, the semantics have one qualifier as the default, and the other required.
1376For example, make the safe @mutex@ qualifier the default because assuming \lstinline[morekeywords=nomutex]@nomutex@ may cause subtle errors.
1377Alternatively, make the unsafe \lstinline[morekeywords=nomutex]@nomutex@ qualifier the default because it is the \emph{normal} parameter semantics while @mutex@ parameters are rare.
1378Providing a default qualifier implies knowing whether a parameter is a monitor.
1379Since \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.
1380For this reason, \CFA requires programmers to identify the kind of parameter with the @mutex@ keyword and uses no keyword to mean \lstinline[morekeywords=nomutex]@nomutex@.
1381
1382The next semantic decision is establishing which parameter \emph{types} may be qualified with @mutex@.
1383Given:
1384\begin{cfa}
1385monitor M { ... }
1386int f1( M & mutex m );
1387int f2( M * mutex m );
1388int f3( M * mutex m[] );
1389int f4( stack( M * ) & mutex m );
1390\end{cfa}
1391the issue is that some of these parameter types are composed of multiple objects.
1392For @f1@, there is only a single parameter object.
1393Adding indirection in @f2@ still identifies a single object.
1394However, the matrix in @f3@ introduces multiple objects.
1395While shown shortly, multiple acquisition is possible;
1396however array lengths are often unknown in C.
1397This issue is exacerbated in @f4@, where the data structure must be safely traversed to acquire all of its elements.
1398
1399To make the issue tractable, \CFA only acquires one monitor per parameter with at most one level of indirection.
1400However, there is an ambiguity in the C type-system with respects to arrays.
1401Is the argument for @f2@ a single object or an array of objects?
1402If it is an array, only the first element of the array is acquired, which seems unsafe;
1403hence, @mutex@ is disallowed for array parameters.
1404\begin{cfa}
1405int f1( M & mutex m );                                          $\C{// allowed: recommended case}$
1406int f2( M * mutex m );                                          $\C{// disallowed: could be an array}$
1407int f3( M mutex m[$\,$] );                                      $\C{// disallowed: array length unknown}$
1408int f4( M ** mutex m );                                         $\C{// disallowed: could be an array}$
1409int f5( M * mutex m[$\,$] );                            $\C{// disallowed: array length unknown}$
1410\end{cfa}
1411% Note, not all array routines have distinct types: @f2@ and @f3@ have the same type, as do @f4@ and @f5@.
1412% However, even if the code generation could tell the difference, the extra information is still not sufficient to extend meaningfully the monitor call semantic.
1413
1414For object-oriented monitors, calling a mutex member \emph{implicitly} acquires mutual exclusion of the receiver object, @`rec`.foo(...)@.
1415\CFA has no receiver, and hence, must use an explicit mechanism to specify which object acquires mutual exclusion.
1416A positive consequence of this design decision is the ability to support multi-monitor routines.
1417\begin{cfa}
1418int f( M & mutex x, M & mutex y );              $\C{// multiple monitor parameter of any type}$
1419M m1, m2;
1420f( m1, m2 );
1421\end{cfa}
1422(While object-oriented monitors can be extended with a mutex qualifier for multiple-monitor members, no prior example of this feature could be found.)
1423In practice, writing multi-locking routines that do not deadlock is tricky.
1424Having language support for such a feature is therefore a significant asset for \CFA.
1425
1426The capability to acquire multiple locks before entering a critical section is called \newterm{bulk acquire} (see Section~\ref{s:Implementation} for implementation details).
1427In the previous example, \CFA guarantees the order of acquisition is consistent across calls to different routines using the same monitors as arguments.
1428This consistent ordering means acquiring multiple monitors is safe from deadlock.
1429However, users can force the acquiring order.
1430For example, notice the use of @mutex@/\lstinline[morekeywords=nomutex]@nomutex@ and how this affects the acquiring order:
1431\begin{cfa}
1432void foo( M & mutex m1, M & mutex m2 );         $\C{// acquire m1 and m2}$
1433void bar( M & mutex m1, M & /* nomutex */ m2 ) { $\C{// acquire m1}$
1434        ... foo( m1, m2 ); ...                                  $\C{// acquire m2}$
1435}
1436void baz( M & /* nomutex */ m1, M & mutex m2 ) { $\C{// acquire m2}$
1437        ... foo( m1, m2 ); ...                                  $\C{// acquire m1}$
1438}
1439\end{cfa}
1440The multi-acquire semantics allows @bar@ or @baz@ to acquire a monitor lock and reacquire it in @foo@.
1441In the calls to @bar@ and @baz@, the monitors are acquired in opposite order.
1442
1443However, such use leads to lock acquiring order problems resulting in deadlock~\cite{Lister77}, where detecting it requires dynamic tracking of monitor calls, and dealing with it requires rollback semantics~\cite{Dice10}.
1444In \CFA, a safety aid is provided by using bulk acquire of all monitors to shared objects, whereas other monitor systems provide no aid.
1445While \CFA provides only a partial solution, it handles many useful cases, \eg:
1446\begin{cfa}
1447monitor BankAccount { ... };
1448void deposit( BankAccount & `mutex` b, int deposit );
1449void transfer( BankAccount & `mutex` my, BankAccount & `mutex` your, int me2you ) {
1450        deposit( my, `-`me2you );                               $\C{// debit}$
1451        deposit( your, me2you );                                $\C{// credit}$
1452}
1453\end{cfa}
1454This example shows a trivial solution to the bank-account transfer problem.
1455Without multi- and bulk acquire, the solution to this problem requires careful engineering.
1456
1457
1458\subsection{\protect\lstinline@mutex@ statement}
1459\label{mutex-stmt}
1460
1461The monitor call-semantics associate all locking semantics to routines.
1462Like Java, \CFA offers an alternative @mutex@ statement to reduce refactoring and naming.
1463\begin{cquote}
1464\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1465\begin{cfa}
1466monitor M { ... };
1467void foo( M & mutex m1, M & mutex m2 ) {
1468        // critical section
1469}
1470void bar( M & m1, M & m2 ) {
1471        foo( m1, m2 );
1472}
1473\end{cfa}
1474&
1475\begin{cfa}
1476
1477void bar( M & m1, M & m2 ) {
1478        mutex( m1, m2 ) {       // remove refactoring and naming
1479                // critical section
1480        }
1481}
1482
1483\end{cfa}
1484\\
1485\multicolumn{1}{c}{\textbf{routine call}} & \multicolumn{1}{c}{\lstinline@mutex@ \textbf{statement}}
1486\end{tabular}
1487\end{cquote}
1488
1489
1490\section{Scheduling}
1491\label{s:Scheduling}
1492
1493While monitor mutual-exclusion provides safe access to shared data, the monitor data may indicate that a thread accessing it cannot proceed.
1494For example, Figure~\ref{f:GenericBoundedBuffer} shows a bounded buffer that may be full/empty so produce/consumer threads must block.
1495Leaving the monitor and trying again (busy waiting) is impractical for high-level programming.
1496Monitors eliminate busy waiting by providing synchronization to schedule threads needing access to the shared data, where threads block versus spinning.
1497Synchronization is generally achieved with internal~\cite{Hoare74} or external~\cite[\S~2.9.2]{uC++} scheduling, where \newterm{scheduling} defines which thread acquires the critical section next.
1498\newterm{Internal scheduling} is characterized by each thread entering the monitor and making an individual decision about proceeding or blocking, while \newterm{external scheduling} is characterized by an entering thread making a decision about proceeding for itself and on behalf of other threads attempting entry.
1499
1500Figure~\ref{f:BBInt} shows a \CFA generic bounded-buffer with internal scheduling, where producers/consumers enter the monitor, see the buffer is full/empty, and block on an appropriate condition lock, @full@/@empty@.
1501The @wait@ routine atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the routine's parameter list.
1502The appropriate condition lock is signalled to unblock an opposite kind of thread after an element is inserted/removed from the buffer.
1503Signalling is unconditional, because signalling an empty condition lock does nothing.
1504
1505Signalling semantics cannot have the signaller and signalled thread in the monitor simultaneously, which means:
1506\begin{enumerate}
1507\item
1508The signalling thread returns immediately, and the signalled thread continues.
1509\item
1510The signalling thread continues and the signalled thread is marked for urgent unblocking at the next scheduling point (exit/wait).
1511\item
1512The signalling thread blocks but is marked for urgrent unblocking at the next scheduling point and the signalled thread continues.
1513\end{enumerate}
1514The first approach is too restrictive, as it precludes solving a reasonable class of problems, \eg dating service (see Figure~\ref{f:DatingService}).
1515\CFA supports the next two semantics as both are useful.
1516Finally, while it is common to store a @condition@ as a field of the monitor, in \CFA, a @condition@ variable can be created/stored independently.
1517Furthermore, a condition variable is tied to a \emph{group} of monitors on first use, called \newterm{branding}, which means that using internal scheduling with distinct sets of monitors requires one condition variable per set of monitors.
1518
1519\begin{figure}
1520\centering
1521\newbox\myboxA
1522\begin{lrbox}{\myboxA}
1523\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1524forall( otype T ) { // distribute forall
1525        monitor Buffer {
1526                `condition` full, empty;
1527                int front, back, count;
1528                T elements[10];
1529        };
1530        void ?{}( Buffer(T) & buffer ) with(buffer) {
1531                [front, back, count] = 0;
1532        }
1533
1534        void insert( Buffer(T) & mutex buffer, T elem )
1535                                with(buffer) {
1536                if ( count == 10 ) `wait( empty )`;
1537                // insert elem into buffer
1538                `signal( full )`;
1539        }
1540        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1541                if ( count == 0 ) `wait( full )`;
1542                // remove elem from buffer
1543                `signal( empty )`;
1544                return elem;
1545        }
1546}
1547\end{cfa}
1548\end{lrbox}
1549
1550\newbox\myboxB
1551\begin{lrbox}{\myboxB}
1552\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1553forall( otype T ) { // distribute forall
1554        monitor Buffer {
1555
1556                int front, back, count;
1557                T elements[10];
1558        };
1559        void ?{}( Buffer(T) & buffer ) with(buffer) {
1560                [front, back, count] = 0;
1561        }
1562        T remove( Buffer(T) & mutex buffer ); // forward
1563        void insert( Buffer(T) & mutex buffer, T elem )
1564                                with(buffer) {
1565                if ( count == 10 ) `waitfor( remove, buffer )`;
1566                // insert elem into buffer
1567
1568        }
1569        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1570                if ( count == 0 ) `waitfor( insert, buffer )`;
1571                // remove elem from buffer
1572
1573                return elem;
1574        }
1575}
1576\end{cfa}
1577\end{lrbox}
1578
1579\subfloat[Internal Scheduling]{\label{f:BBInt}\usebox\myboxA}
1580%\qquad
1581\subfloat[External Scheduling]{\label{f:BBExt}\usebox\myboxB}
1582\caption{Generic Bounded-Buffer}
1583\label{f:GenericBoundedBuffer}
1584\end{figure}
1585
1586Figure~\ref{f:BBExt} shows a \CFA generic bounded-buffer with external scheduling, where producers/consumers detecting a full/empty buffer block and prevent more producers/consumers from entering the monitor until there is a free/empty slot in the buffer.
1587External scheduling is controlled by the @waitfor@ statement, which atomically blocks the calling thread, releases the monitor lock, and restricts the routine calls that can next acquire mutual exclusion.
1588If the buffer is full, only calls to @remove@ can acquire the buffer, and if the buffer is empty, only calls to @insert@ can acquire the buffer.
1589Threads making calls to routines that are currently excluded, block outside of (external to) the monitor on a calling queue, versus blocking on condition queues inside of (internal to) the monitor.
1590External scheduling allows users to wait for events from other threads without concern of unrelated events occurring.
1591The mechnaism can be done in terms of control flow, \eg Ada @accept@ or \uC @_Accept@, or in terms of data, \eg Go channels.
1592While both mechanisms have strengths and weaknesses, this project uses a control-flow mechanism to stay consistent with other language semantics.
1593Two challenges specific to \CFA for external scheduling are loose object-definitions (see Section~\ref{s:LooseObjectDefinitions}) and multiple-monitor routines (see Section~\ref{s:Multi-MonitorScheduling}).
1594
1595For internal scheduling, non-blocking signalling (as in the producer/consumer example) is used when the signaller is providing the cooperation for a waiting thread;
1596the signaller enters the monitor and changes state, detects a waiting threads that can use the state, performs a non-blocking signal on the condition queue for the waiting thread, and exits the monitor to run concurrently.
1597The waiter unblocks next from the urgent queue, uses/takes the state, and exits the monitor.
1598Blocking signalling is the reverse, where the waiter is providing the cooperation for the signalling thread;
1599the signaller enters the monitor, detects a waiting thread providing the necessary state, performs a blocking signal to place it on the urgent queue and unblock the waiter.
1600The waiter changes state and exits the monitor, and the signaller unblocks next from the urgent queue to use/take the state.
1601
1602Figure~\ref{f:DatingService} shows a dating service demonstrating non-blocking and blocking signalling.
1603The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
1604A thread blocks until an appropriate partner arrives.
1605The complexity is exchanging phone numbers in the monitor because of the mutual-exclusion property.
1606For signal scheduling, the @exchange@ condition is necessary to block the thread finding the match, while the matcher unblocks to take the opposite number, post its phone number, and unblock the partner.
1607For signal-block scheduling, the implicit urgent-queue replaces the explict @exchange@-condition and @signal_block@ puts the finding thread on the urgent condition and unblocks the matcher.
1608
1609The dating service is an example of a monitor that cannot be written using external scheduling because it requires knowledge of calling parameters to make scheduling decisions, and parameters of waiting threads are unavailable;
1610as well, an arriving thread may not find a partner and must wait, which requires a condition variable, and condition variables imply internal scheduling.
1611
1612\begin{figure}
1613\centering
1614\newbox\myboxA
1615\begin{lrbox}{\myboxA}
1616\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1617enum { CCodes = 20 };
1618monitor DS {
1619        int GirlPhNo, BoyPhNo;
1620        condition Girls[CCodes], Boys[CCodes];
1621        condition exchange;
1622};
1623int girl( DS & mutex ds, int phNo, int ccode ) {
1624        if ( is_empty( Boys[ccode] ) ) {
1625                wait( Girls[ccode] );
1626                GirlPhNo = phNo;
1627                `signal( exchange );`
1628        } else {
1629                GirlPhNo = phNo;
1630                `signal( Boys[ccode] );`
1631                `wait( exchange );`
1632        } // if
1633        return BoyPhNo;
1634}
1635int boy( DS & mutex ds, int phNo, int ccode ) {
1636        // as above with boy/girl interchanged
1637}
1638\end{cfa}
1639\end{lrbox}
1640
1641\newbox\myboxB
1642\begin{lrbox}{\myboxB}
1643\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1644
1645monitor DS {
1646        int GirlPhNo, BoyPhNo;
1647        condition Girls[CCodes], Boys[CCodes];
1648
1649};
1650int girl( DS & mutex ds, int phNo, int ccode ) {
1651        if ( is_empty( Boys[ccode] ) ) { // no compatible
1652                wait( Girls[ccode] ); // wait for boy
1653                GirlPhNo = phNo; // make phone number available
1654
1655        } else {
1656                GirlPhNo = phNo; // make phone number available
1657                `signal_block( Boys[ccode] );` // restart boy
1658
1659        } // if
1660        return BoyPhNo;
1661}
1662int boy( DS & mutex ds, int phNo, int ccode ) {
1663        // as above with boy/girl interchanged
1664}
1665\end{cfa}
1666\end{lrbox}
1667
1668\subfloat[\lstinline@signal@]{\label{f:DatingSignal}\usebox\myboxA}
1669\qquad
1670\subfloat[\lstinline@signal_block@]{\label{f:DatingSignalBlock}\usebox\myboxB}
1671\caption{Dating service. }
1672\label{f:DatingService}
1673\end{figure}
1674
1675Both internal and external scheduling extend to multiple monitors in a natural way.
1676\begin{cquote}
1677\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1678\begin{cfa}
1679monitor M { `condition e`; ... };
1680void foo( M & mutex m1, M & mutex m2 ) {
1681        ... wait( `e` ); ...   // wait( e, m1, m2 )
1682        ... wait( `e, m1` ); ...
1683        ... wait( `e, m2` ); ...
1684}
1685\end{cfa}
1686&
1687\begin{cfa}
1688void rtn$\(_1\)$( M & mutex m1, M & mutex m2 );
1689void rtn$\(_2\)$( M & mutex m1 );
1690void bar( M & mutex m1, M & mutex m2 ) {
1691        ... waitfor( `rtn` ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\), m1, m2 )}$
1692        ... waitfor( `rtn, m1` ); ... // $\LstCommentStyle{waitfor( rtn\(_2\), m1 )}$
1693}
1694\end{cfa}
1695\end{tabular}
1696\end{cquote}
1697For @wait( e )@, the default semantics is to atomically block the signaller and release all acquired mutex types in the parameter list, \ie @wait( e, m1, m2 )@.
1698To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @wait( e, m1 )@.
1699Wait statically verifies the released monitors are the acquired mutex-parameters so unconditional release is safe.
1700Finally, a signaller,
1701\begin{cfa}
1702void baz( M & mutex m1, M & mutex m2 ) {
1703        ... signal( e ); ...
1704}
1705\end{cfa}
1706must have acquired at least the same locks as the waiting thread signalled from the condition queue.
1707
1708Similarly, for @waitfor( rtn )@, the default semantics is to atomically block the acceptor and release all acquired mutex types in the parameter list, \ie @waitfor( rtn, m1, m2 )@.
1709To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn, m1 )@.
1710@waitfor@ statically verifies the released monitors are the same as the acquired mutex-parameters of the given routine or routine pointer.
1711To statically verify the released monitors match with the accepted routine's mutex parameters, the routine (pointer) prototype must be accessible.
1712% When an overloaded routine appears in an @waitfor@ statement, calls to any routine with that name are accepted.
1713% The rationale is that members with the same name should perform a similar function, and therefore, all should be eligible to accept a call.
1714Overloaded routines can be disambiguated using a cast:
1715\begin{cfa}
1716void rtn( M & mutex m );
1717`int` rtn( M & mutex m );
1718waitfor( (`int` (*)( M & mutex ))rtn, m );
1719\end{cfa}
1720
1721The ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock.
1722\begin{cfa}
1723void foo( M & mutex m1, M & mutex m2 ) {
1724        ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired )}$
1725void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2 )}$
1726        ... signal( `e` ); ...
1727\end{cfa}
1728The @wait@ only releases @m1@ so the signalling thread cannot acquire both @m1@ and @m2@ to  enter @bar@ to get to the @signal@.
1729While deadlock issues can occur with multiple/nesting acquisition, this issue results from the fact that locks, and by extension monitors, are not perfectly composable.
1730
1731Finally, an important aspect of monitor implementation is barging, \ie can calling threads barge ahead of signalled threads?
1732If barging is allowed, synchronization between a signaller and signallee is difficult, often requiring multiple unblock/block cycles (looping around a wait rechecking if a condition is met).
1733In fact, signals-as-hints is completely opposite from that proposed by Hoare in the seminal paper on monitors:
1734\begin{quote}
1735However, we decree that a signal operation be followed immediately by resumption of a waiting program, without possibility of an intervening procedure call from yet a third program.
1736It is only in this way that a waiting program has an absolute guarantee that it can acquire the resource just released by the signalling program without any danger that a third program will interpose a monitor entry and seize the resource instead.~\cite[p.~550]{Hoare74}
1737\end{quote}
1738\CFA scheduling \emph{precludes} barging, which simplifies synchronization among threads in the monitor and increases correctness.
1739Furthermore, \CFA concurrency has no spurious wakeup~\cite[\S~9]{Buhr05a}, which eliminates an implict form of barging.
1740For example, there are no loops in either bounded buffer solution in Figure~\ref{f:GenericBoundedBuffer}.
1741Supporting barging prevention as well as extending internal scheduling to multiple monitors is the main source of complexity in the design and implementation of \CFA concurrency.
1742
1743
1744\subsection{Barging Prevention}
1745
1746Figure~\ref{f:BargingPrevention} shows \CFA code where bulk acquire adds complexity to the internal-signalling semantics.
1747The complexity begins at the end of the inner @mutex@ statement, where the semantics of internal scheduling need to be extended for multiple monitors.
1748The problem is that bulk acquire is used in the inner @mutex@ statement where one of the monitors is already acquired.
1749When the signalling thread reaches the end of the inner @mutex@ statement, it should transfer ownership of @m1@ and @m2@ to the waiting threads to prevent barging into the outer @mutex@ statement by another thread.
1750However, both the signalling and waiting thread W1 still need monitor @m1@.
1751
1752\begin{figure}
1753\newbox\myboxA
1754\begin{lrbox}{\myboxA}
1755\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1756monitor M m1, m2;
1757condition c;
1758mutex( m1 ) { // $\LstCommentStyle{\color{red}outer}$
1759        ...
1760        mutex( m1, m2 ) { // $\LstCommentStyle{\color{red}inner}$
1761                ... `signal( c )`; ...
1762                // m1, m2 acquired
1763        } // $\LstCommentStyle{\color{red}release m2}$
1764        // m1 acquired
1765} // release m1
1766\end{cfa}
1767\end{lrbox}
1768
1769\newbox\myboxB
1770\begin{lrbox}{\myboxB}
1771\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1772
1773
1774mutex( m1 ) {
1775        ...
1776        mutex( m1, m2 ) {
1777                ... `wait( c )`; // block and release m1, m2
1778                // m1, m2 acquired
1779        } // $\LstCommentStyle{\color{red}release m2}$
1780        // m1 acquired
1781} // release m1
1782\end{cfa}
1783\end{lrbox}
1784
1785\newbox\myboxC
1786\begin{lrbox}{\myboxC}
1787\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1788
1789
1790mutex( m2 ) {
1791        ... `wait( c )`; ...
1792        // m2 acquired
1793} // $\LstCommentStyle{\color{red}release m2}$
1794
1795
1796
1797
1798\end{cfa}
1799\end{lrbox}
1800
1801\begin{cquote}
1802\subfloat[Signalling Thread]{\label{f:SignallingThread}\usebox\myboxA}
1803\hspace{2\parindentlnth}
1804\subfloat[Waiting Thread (W1)]{\label{f:WaitingThread}\usebox\myboxB}
1805\hspace{2\parindentlnth}
1806\subfloat[Waiting Thread (W2)]{\label{f:OtherWaitingThread}\usebox\myboxC}
1807\end{cquote}
1808\caption{Barging Prevention}
1809\label{f:BargingPrevention}
1810\end{figure}
1811
1812One scheduling solution is for the signaller to keep ownership of all locks until the last lock is ready to be transferred, because this semantics fits most closely to the behaviour of single-monitor scheduling.
1813However, Figure~\ref{f:OtherWaitingThread} shows this solution is complex depending on other waiters, resulting in options when the signaller finishes the inner mutex-statement.
1814The signaller can retain @m2@ until completion of the outer mutex statement and pass the locks to waiter W1, or it can pass @m2@ to waiter W2 after completing the inner mutex-statement, while continuing to hold @m1@.
1815In the latter case, waiter W2 must eventually pass @m2@ to waiter W1, which is complex because W1 may have waited before W2, so W2 is unaware of it.
1816Furthermore, there is an execution sequence where the signaller always finds waiter W2, and hence, waiter W1 starves.
1817
1818While a number of approaches were examined~\cite[\S~4.3]{Delisle18}, the solution chosen for \CFA is a novel techique called \newterm{partial signalling}.
1819Signalled threads are moved to the urgent queue and the waiter at the front defines the set of monitors necessary for it to unblock.
1820Partial signalling transfers ownership of monitors to the front waiter.
1821When the signaller thread exits or waits in the monitor, the front waiter is unblocked if all its monitors are released.
1822The benefit is encapsulating complexity into only two actions: passing monitors to the next owner when they should be released and conditionally waking threads if all conditions are met.
1823
1824
1825\subsection{Loose Object Definitions}
1826\label{s:LooseObjectDefinitions}
1827
1828In an object-oriented programming-language, a class includes an exhaustive list of operations.
1829However, new members can be added via static inheritance or dynamic members, \eg JavaScript~\cite{JavaScript}.
1830Similarly, monitor routines can be added at any time in \CFA, making it less clear for programmers and more difficult to implement.
1831\begin{cfa}
1832monitor M { ... };
1833void `f`( M & mutex m );
1834void g( M & mutex m ) { waitfor( `f` ); }       $\C{// clear which f}$
1835void `f`( M & mutex m, int );                           $\C{// different f}$
1836void h( M & mutex m ) { waitfor( `f` ); }       $\C{// unclear which f}$
1837\end{cfa}
1838Hence, the cfa-code for entering a monitor looks like:
1839\begin{cfa}
1840if ( $\textrm{\textit{monitor is free}}$ ) $\LstCommentStyle{// \color{red}enter}$
1841else if ( $\textrm{\textit{already own monitor}}$ ) $\LstCommentStyle{// \color{red}continue}$
1842else if ( $\textrm{\textit{monitor accepts me}}$ ) $\LstCommentStyle{// \color{red}enter}$
1843else $\LstCommentStyle{// \color{red}block}$
1844\end{cfa}
1845For the first two conditions, it is easy to implement a check that can evaluate the condition in a few instructions.
1846However, a fast check for \emph{monitor accepts me} is much harder to implement depending on the constraints put on the monitors.
1847Figure~\ref{fig:ClassicalMonitor} shows monitors are often expressed as an entry (calling) queue, some acceptor queues, and an urgent stack/queue.
1848
1849\begin{figure}
1850\centering
1851\subfloat[Classical monitor] {
1852\label{fig:ClassicalMonitor}
1853{\resizebox{0.45\textwidth}{!}{\input{monitor.pstex_t}}}
1854}% subfloat
1855\quad
1856\subfloat[Bulk acquire monitor] {
1857\label{fig:BulkMonitor}
1858{\resizebox{0.45\textwidth}{!}{\input{ext_monitor.pstex_t}}}
1859}% subfloat
1860\caption{Monitor Implementation}
1861\label{f:MonitorImplementation}
1862\end{figure}
1863
1864For a fixed (small) number of mutex routines (\eg 128), the accept check reduces to a bitmask of allowed callers, which can be checked with a single instruction.
1865This approach requires a unique dense ordering of routines with a small upper-bound and the ordering must be consistent across translation units.
1866For object-oriented languages these constraints are common, but \CFA mutex routines can be added in any scope and are only visible in certain translation unit, precluding program-wide dense-ordering among mutex routines.
1867
1868Figure~\ref{fig:BulkMonitor} shows the \CFA monitor implementation.
1869The mutex routine called is associated with each thread on the entry queue, while a list of acceptable routines is kept separately.
1870The accepted list is a variable-sized array of accepted routine pointers, so the single instruction bitmask comparison is replaced by dereferencing a pointer followed by a (usually short) linear search.
1871
1872
1873\subsection{Multi-Monitor Scheduling}
1874\label{s:Multi-MonitorScheduling}
1875
1876External scheduling, like internal scheduling, becomes significantly more complex for multi-monitor semantics.
1877Even in the simplest case, new semantics needs to be established.
1878\newpage
1879\begin{cfa}
1880monitor M { ... };
1881void f( M & mutex m1 );
1882void g( M & mutex m1, M & mutex m2 ) {
1883        waitfor( f );                                                   $\C{\color{red}// pass m1 or m2 to f?}$
1884}
1885\end{cfa}
1886The solution is for the programmer to disambiguate:
1887\begin{cfa}
1888        waitfor( f, m2 );                                               $\C{\color{red}// wait for call to f with argument m2}$
1889\end{cfa}
1890Both locks are acquired by routine @g@, so when routine @f@ is called, the lock for monitor @m2@ is passed from @g@ to @f@, while @g@ still holds lock @m1@.
1891This behaviour can be extended to the multi-monitor @waitfor@ statement.
1892\begin{cfa}
1893monitor M { ... };
1894void f( M & mutex m1, M & mutex m2 );
1895void g( M & mutex m1, M & mutex m2 ) {
1896        waitfor( f, m1, m2 );                                   $\C{\color{red}// wait for call to f with arguments m1 and m2}$
1897}
1898\end{cfa}
1899Again, the set of monitors passed to the @waitfor@ statement must be entirely contained in the set of monitors already acquired by the accepting routine.
1900Also, the order of the monitors in a @waitfor@ statement is unimportant.
1901
1902Figure~\ref{f:UnmatchedMutexSets} shows an example where, for internal and external scheduling with multiple monitors, a signalling or accepting thread must match exactly, \ie partial matching results in waiting.
1903For both examples, the set of monitors is disjoint so unblocking is impossible.
1904
1905\begin{figure}
1906\lstDeleteShortInline@%
1907\begin{tabular}{@{}l@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
1908\begin{cfa}
1909monitor M1 {} m11, m12;
1910monitor M2 {} m2;
1911condition c;
1912void f( M1 & mutex m1, M2 & mutex m2 ) {
1913        signal( c );
1914}
1915void g( M1 & mutex m1, M2 & mutex m2 ) {
1916        wait( c );
1917}
1918g( `m11`, m2 ); // block on wait
1919f( `m12`, m2 ); // cannot fulfil
1920\end{cfa}
1921&
1922\begin{cfa}
1923monitor M1 {} m11, m12;
1924monitor M2 {} m2;
1925
1926void f( M1 & mutex m1, M2 & mutex m2 ) {
1927
1928}
1929void g( M1 & mutex m1, M2 & mutex m2 ) {
1930        waitfor( f, m1, m2 );
1931}
1932g( `m11`, m2 ); // block on accept
1933f( `m12`, m2 ); // cannot fulfil
1934\end{cfa}
1935\end{tabular}
1936\lstMakeShortInline@%
1937\caption{Unmatched \protect\lstinline@mutex@ sets}
1938\label{f:UnmatchedMutexSets}
1939\end{figure}
1940
1941
1942\subsection{Extended \protect\lstinline@waitfor@}
1943
1944Figure~\ref{f:ExtendedWaitfor} show the extended form of the @waitfor@ statement to conditionally accept one of a group of mutex routines, with a specific action to be performed \emph{after} the mutex routine finishes.
1945For a @waitfor@ clause to be executed, its @when@ must be true and an outstanding call to its corresponding member(s) must exist.
1946The \emph{conditional-expression} of a @when@ may call a routine, but the routine must not block or context switch.
1947If there are multiple acceptable mutex calls, selection occurs top-to-bottom (prioritized) in the @waitfor@ clauses, whereas some programming languages with similar mechanisms accept non-deterministically for this case, \eg Go \lstinline[morekeywords=select]@select@.
1948If some accept guards are true and there are no outstanding calls to these members, the acceptor is accept-blocked until a call to one of these members is made.
1949If all the accept guards are false, the statement does nothing, unless there is a terminating @else@ clause with a true guard, which is executed instead.
1950Hence, the terminating @else@ clause allows a conditional attempt to accept a call without blocking.
1951If there is a @timeout@ clause, it provides an upper bound on waiting.
1952If both a @timeout@ clause and an @else@ clause are present, the @else@ must be conditional, or the @timeout@ is never triggered.
1953In all cases, the statement following is executed \emph{after} a clause is executed to know which of the clauses executed.
1954
1955\begin{figure}
1956\begin{cfa}
1957`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1958        waitfor( $\emph{mutex-member-name}$ )
1959                $\emph{statement}$                                      $\C{// action after call}$
1960`or` `when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
1961        waitfor( $\emph{mutex-member-name}$ )
1962                $\emph{statement}$                                      $\C{// action after call}$
1963`or`    ...                                                                     $\C{// list of waitfor clauses}$
1964`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1965        `timeout`                                                               $\C{// optional terminating timeout clause}$
1966                $\emph{statement}$                                      $\C{// action after timeout}$
1967`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1968        `else`                                                                  $\C{// optional terminating clause}$
1969                $\emph{statement}$                                      $\C{// action when no immediate calls}$
1970\end{cfa}
1971\caption{Extended \protect\lstinline@waitfor@}
1972\label{f:ExtendedWaitfor}
1973\end{figure}
1974
1975Note, a group of conditional @waitfor@ clauses is \emph{not} the same as a group of @if@ statements, e.g.:
1976\begin{cfa}
1977if ( C1 ) waitfor( mem1 );                       when ( C1 ) waitfor( mem1 );
1978else if ( C2 ) waitfor( mem2 );         or when ( C2 ) waitfor( mem2 );
1979\end{cfa}
1980The left example accepts only @mem1@ if @C1@ is true or only @mem2@ if @C2@ is true.
1981The right example accepts either @mem1@ or @mem2@ if @C1@ and @C2@ are true.
1982
1983An interesting use of @waitfor@ is accepting the @mutex@ destructor to know when an object is deallocated.
1984\begin{cfa}
1985void insert( Buffer(T) & mutex buffer, T elem ) with( buffer ) {
1986        if ( count == 10 )
1987                waitfor( remove, buffer ) {
1988                        // insert elem into buffer
1989                } or `waitfor( ^?{}, buffer )` throw insertFail;
1990}
1991\end{cfa}
1992When the buffer is deallocated, the current waiter is unblocked and informed, so it can perform an appropriate action.
1993However, the basic @waitfor@ semantics do not support this functionality, since using an object after its destructor is called is undefined.
1994Therefore, to make this useful capability work, the semantics for accepting the destructor is the same as @signal@, \ie the call to the destructor is placed on the urgent queue and the acceptor continues execution, which throws an exception to the acceptor and then the caller is unblocked from the urgent queue to deallocate the object.
1995Accepting the destructor is an idiomatic way to terminate a thread in \CFA.
1996
1997
1998\subsection{\protect\lstinline@mutex@ Threads}
1999
2000Threads in \CFA are monitors to allow direct communication among threads, \ie threads can have mutex routines that are called by other threads.
2001Hence, all monitor features are available when using threads.
2002The following shows an example of two threads directly calling each other and accepting calls from each other in a cycle.
2003\begin{cfa}
2004thread Ping {} pi;
2005thread Pong {} po;
2006void ping( Ping & mutex ) {}
2007void pong( Pong & mutex ) {}
2008int main() {}
2009\end{cfa}
2010\vspace{-0.8\baselineskip}
2011\begin{cquote}
2012\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
2013\begin{cfa}
2014void main( Ping & pi ) {
2015        for ( int i = 0; i < 10; i += 1 ) {
2016                `waitfor( ping, pi );`
2017                `pong( po );`
2018        }
2019}
2020\end{cfa}
2021&
2022\begin{cfa}
2023void main( Pong & po ) {
2024        for ( int i = 0; i < 10; i += 1 ) {
2025                `ping( pi );`
2026                `waitfor( pong, po );`
2027        }
2028}
2029\end{cfa}
2030\end{tabular}
2031\end{cquote}
2032% \lstMakeShortInline@%
2033% \caption{Threads ping/pong using external scheduling}
2034% \label{f:pingpong}
2035% \end{figure}
2036Note, the ping/pong threads are globally declared, @pi@/@po@, and hence, start (and possibly complete) before the program main starts.
2037
2038
2039\subsection{Low-level Locks}
2040
2041For completeness and efficiency, \CFA provides a standard set of low-level locks: recursive mutex, condition, semaphore, barrier, \etc, and atomic instructions: @fetchAssign@, @fetchAdd@, @testSet@, @compareSet@, \etc.
2042
2043
2044\section{Parallelism}
2045
2046Historically, computer performance was about processor speeds.
2047However, with heat dissipation being a direct consequence of speed increase, parallelism is the new source for increased performance~\cite{Sutter05, Sutter05b}.
2048Now, high-performance applications must care about parallelism, which requires concurrency.
2049The lowest-level approach of parallelism is to use \newterm{kernel threads} in combination with semantics like @fork@, @join@, \etc.
2050However, kernel threads are better as an implementation tool because of complexity and higher cost.
2051Therefore, different abstractions are often layered onto kernel threads to simplify them, \eg pthreads.
2052
2053
2054\subsection{User Threads with Preemption}
2055
2056A direct improvement on kernel threads is user threads, \eg Erlang~\cite{Erlang} and \uC~\cite{uC++book}.
2057This approach provides an interface that matches the language paradigms, more control over concurrency by the language runtime, and an abstract (and portable) interface to the underlying kernel threads across operating systems.
2058In many cases, user threads can be used on a much larger scale (100,000 threads).
2059Like kernel threads, user threads support preemption, which maximizes nondeterminism, but introduces the same concurrency errors: race, livelock, starvation, and deadlock.
2060\CFA adopts user-threads as they represent the truest realization of concurrency and can build any other concurrency approach, \eg thread pools and actors~\cite{Actors}.
2061
2062
2063\subsection{User Threads without Preemption (Fiber)}
2064\label{s:fibers}
2065
2066A variant of user thread is \newterm{fibers}, which removes preemption, \eg Go~\cite{Go} @goroutine@s.
2067Like functional programming, which removes mutation and its associated problems, removing preemption from concurrency reduces nondeterminism, making race and deadlock errors more difficult to generate.
2068However, preemption is necessary for concurrency that relies on spinning, so there are a class of problems that cannot be programmed without preemption.
2069
2070
2071\subsection{Thread Pools}
2072
2073In contrast to direct threading is indirect \newterm{thread pools}, where small jobs (work units) are inserted into a work pool for execution.
2074If the jobs are dependent, \ie interact, there is an implicit/explicit dependency graph that ties them together.
2075While removing direct concurrency, and hence the amount of context switching, thread pools significantly limit the interaction that can occur among jobs.
2076Indeed, jobs should not block because that also blocks the underlying thread, which effectively means the CPU utilization, and therefore throughput, suffers.
2077While it is possible to tune the thread pool with sufficient threads, it becomes difficult to obtain high throughput and good core utilization as job interaction increases.
2078As well, concurrency errors return, which threads pools are suppose to mitigate.
2079
2080
2081\section{\protect\CFA Runtime Structure}
2082
2083Figure~\ref{f:RunTimeStructure} illustrates the runtime structure of a \CFA program.
2084In addition to the new kinds of objects introduced by \CFA, there are two more runtime entities used to control parallel execution: cluster and (virtual) processor.
2085An executing thread is illustrated by its containment in a processor.
2086
2087\begin{figure}
2088\centering
2089\input{RunTimeStructure}
2090\caption{\CFA Runtime Structure}
2091\label{f:RunTimeStructure}
2092\end{figure}
2093
2094
2095\subsection{Cluster}
2096\label{s:RuntimeStructureCluster}
2097
2098A \newterm{cluster} is a collection of threads and virtual processors (abstract kernel-thread) that execute the threads (like a virtual machine).
2099The purpose of a cluster is to control the amount of parallelism that is possible among threads, plus scheduling and other execution defaults.
2100The default cluster-scheduler is single-queue multi-server, which provides automatic load-balancing of threads on processors.
2101However, the scheduler is pluggable, supporting alternative schedulers.
2102If several clusters exist, both threads and virtual processors, can be explicitly migrated from one cluster to another.
2103No automatic load balancing among clusters is performed by \CFA.
2104
2105When a \CFA program begins execution, it creates a user cluster with a single processor and a special processor to handle preemption that does not execute user threads.
2106The user cluster is created to contain the application user-threads.
2107Having all threads execute on the one cluster often maximizes utilization of processors, which minimizes runtime.
2108However, because of limitations of the underlying operating system, heterogeneous hardware, or scheduling requirements (real-time), multiple clusters are sometimes necessary.
2109
2110
2111\subsection{Virtual Processor}
2112\label{s:RuntimeStructureProcessor}
2113
2114A virtual processor is implemented by a kernel thread (\eg UNIX process), which is subsequently scheduled for execution on a hardware processor by the underlying operating system.
2115Programs may use more virtual processors than hardware processors.
2116On a multiprocessor, kernel threads are distributed across the hardware processors resulting in virtual processors executing in parallel.
2117(It is possible to use affinity to lock a virtual processor onto a particular hardware processor~\cite{affinityLinux, affinityWindows, affinityFreebsd, affinityNetbsd, affinityMacosx}, which is used when caching issues occur or for heterogeneous hardware processors.)
2118The \CFA runtime attempts to block unused processors and unblock processors as the system load increases;
2119balancing the workload with processors is difficult.
2120Preemption occurs on virtual processors rather than user threads, via operating-system interrupts.
2121Thus virtual processors execute user threads, where preemption frequency applies to a virtual processor, so preemption occurs randomly across the executed user threads.
2122Turning off preemption transforms user threads into fibers.
2123
2124
2125\subsection{Debug Kernel}
2126
2127There are two versions of the \CFA runtime kernel: debug and non-debug.
2128The debugging version has many runtime checks and internal assertions, \eg stack (non-writable) guard page, and checks for stack overflow whenever context switches occur among coroutines and threads, which catches most stack overflows.
2129After a program is debugged, the non-debugging version can be used to decrease space and increase performance.
2130
2131
2132\section{Implementation}
2133\label{s:Implementation}
2134
2135Currently, \CFA has fixed-sized stacks, where the stack size can be set at coroutine/thread creation but with no subsequent growth.
2136Schemes exist for dynamic stack-growth, such as stack copying and chained stacks.
2137However, stack copying requires pointer adjustment to items on the stack, which is impossible without some form of garbage collection.
2138As well, chained stacks require all modules be recompiled to use this feature, which breaks backward compatibility with existing C libraries.
2139In the long term, it is likely C libraries will migrate to stack chaining to support concurrency, at only a minimal cost to sequential programs.
2140Nevertheless, experience teaching \uC~\cite{CS343} shows fixed-sized stacks are rarely an issue in most concurrent programs.
2141
2142A primary implementation challenge is avoiding contention from dynamically allocating memory because of bulk acquire, \eg the internal-scheduling design is (almost) free of allocations.
2143All blocking operations are made by parking threads onto queues, therefore all queues are designed with intrusive nodes, where each node has preallocated link fields for chaining.
2144Furthermore, several bulk-acquire operations need a variable amount of memory.
2145This storage is allocated at the base of a thread's stack before blocking, which means programmers must add a small amount of extra space for stacks.
2146
2147In \CFA, ordering of monitor acquisition relies on memory ordering to prevent deadlock~\cite{Havender68}, because all objects have distinct non-overlapping memory layouts, and mutual-exclusion for a monitor is only defined for its lifetime.
2148When a mutex call is made, pointers to the concerned monitors are aggregated into a variable-length array and sorted.
2149This array persists for the entire duration of the mutual exclusion and is used extensively for synchronization operations.
2150
2151To improve performance and simplicity, context switching occurs inside a routine call, so only callee-saved registers are copied onto the stack and then the stack register is switched;
2152the corresponding registers are then restored for the other context.
2153Note, the instruction pointer is untouched since the context switch is always inside the same routine.
2154Unlike coroutines, threads do not context switch among each other;
2155they context switch to the cluster scheduler.
2156This method is a 2-step context-switch and provides a clear distinction between user and kernel code, where scheduling and other system operations happen.
2157The alternative 1-step context-switch uses the \emph{from} thread's stack to schedule and then context-switches directly to the \emph{to} thread's stack.
2158Experimental results (not presented) show the performance of these two approaches is virtually equivalent, because both approaches are dominated by locking to prevent a race condition.
2159
2160All kernel threads (@pthreads@) created a stack.
2161Each \CFA virtual processor is implemented as a coroutine and these coroutines run directly on the kernel-thread stack, effectively stealing this stack.
2162The exception to this rule is the program main, \ie the initial kernel thread that is given to any program.
2163In order to respect C expectations, the stack of the initial kernel thread is used by program main rather than the main processor, allowing it to grow dynamically as in a normal C program.
2164
2165Finally, an important aspect for a complete threading system is preemption, which introduces extra non-determinism via transparent interleaving, rather than cooperation among threads for proper scheduling and processor fairness from long-running threads.
2166Because preemption frequency is usually long (1 millisecond) performance cost is negligible.
2167Preemption is normally handled by setting a count-down timer on each virtual processor.
2168When the timer expires, an interrupt is delivered, and the interrupt handler resets the count-down timer, and if the virtual processor is executing in user code, the signal handler performs a user-level context-switch, or if executing in the language runtime-kernel, the preemption is ignored or rolled forward to the point where the runtime kernel context switches back to user code.
2169Multiple signal handlers may be pending.
2170When control eventually switches back to the signal handler, it returns normally, and execution continues in the interrupted user thread, even though the return from the signal handler may be on a different kernel thread than the one where the signal is delivered.
2171The only issue with this approach is that signal masks from one kernel thread may be restored on another as part of returning from the signal handler;
2172therefore, the same signal mask is required for all virtual processors in a cluster.
2173
2174However, on current UNIX systems:
2175\begin{quote}
2176A process-directed signal may be delivered to any one of the threads that does not currently have the signal blocked.
2177If more than one of the threads has the signal unblocked, then the kernel chooses an arbitrary thread to which to deliver the signal.
2178SIGNAL(7) - Linux Programmer's Manual
2179\end{quote}
2180Hence, the timer-expiry signal, which is generated \emph{externally} by the UNIX kernel to the UNIX process, is delivered to any of its UNIX subprocesses (kernel threads).
2181To ensure each virtual processor receives its own preemption signals, a discrete-event simulation is run on a special virtual processor, and only it sets and receives timer events.
2182Virtual processors register an expiration time with the discrete-event simulator, which is inserted in sorted order.
2183The simulation sets the count-down timer to the value at the head of the event list, and when the timer expires, all events less than or equal to the current time are processed.
2184Processing a preemption event sends an \emph{internal} @SIGUSR1@ signal to the registered virtual processor, which is always delivered to that processor.
2185
2186
2187\section{Performance}
2188\label{results}
2189
2190To verify the implementation of the \CFA runtime, a series of microbenchmarks are performed comparing \CFA with other widely used programming languages with concurrency.
2191Table~\ref{t:machine} shows the specifications of the computer used to run the benchmarks, and the versions of the software used in the comparison.
2192
2193\begin{table}
2194\centering
2195\caption{Experiment environment}
2196\label{t:machine}
2197
2198\begin{tabular}{|l|r||l|r|}
2199\hline
2200Architecture            & x86\_64                               & NUMA node(s)  & 8 \\
2201\hline
2202CPU op-mode(s)          & 32-bit, 64-bit                & Model name    & AMD Opteron\texttrademark\ Processor 6380 \\
2203\hline
2204Byte Order                      & Little Endian                 & CPU Freq              & 2.5 GHz \\
2205\hline
2206CPU(s)                          & 64                                    & L1d cache     & 16 KiB \\
2207\hline
2208Thread(s) per core      & 2                                     & L1i cache     & 64 KiB \\
2209\hline
2210Core(s) per socket      & 8                                     & L2 cache              & 2048 KiB \\
2211\hline
2212Socket(s)                       & 4                                     & L3 cache              & 6144 KiB \\
2213\hline
2214\hline
2215Operating system        & Ubuntu 16.04.3 LTS    & Kernel                & Linux 4.4-97-generic \\
2216\hline
2217gcc                                     & 6.3                                   & \CFA                  & 1.0.0 \\
2218\hline
2219Java                            & OpenJDK-9                     & Go                    & 1.9.2 \\
2220\hline
2221\end{tabular}
2222\end{table}
2223
2224All benchmarks are run using the following harness:
2225\begin{cfa}
2226unsigned int N = 10_000_000;
2227#define BENCH( run ) Time before = getTimeNsec(); run; Duration result = (getTimeNsec() - before) / N;
2228\end{cfa}
2229The method used to get time is @clock_gettime( CLOCK_REALTIME )@.
2230Each benchmark is performed @N@ times, where @N@ varies depending on the benchmark;
2231the total time is divided by @N@ to obtain the average time for a benchmark.
2232All omitted tests for other languages are functionally identical to the shown \CFA test.
2233
2234
2235\paragraph{Context-Switching}
2236
2237In procedural programming, the cost of a routine call is important as modularization (refactoring) increases.
2238(In many cases, a compiler inlines routine calls to eliminate this cost.)
2239Similarly, when modularization extends to coroutines/tasks, the time for a context switch becomes a relevant factor.
2240The coroutine context-switch is 2-step using resume/suspend, \ie from resumer to suspender and from suspender to resumer.
2241The thread context switch is 2-step using yield, \ie enter and return from the runtime kernel.
2242Figure~\ref{f:ctx-switch} shows the code for coroutines/threads with all results in Table~\ref{tab:ctx-switch}.
2243The difference in performance between coroutine and thread context-switch is the cost of scheduling for threads, whereas coroutines are self-scheduling.
2244
2245\begin{figure}
2246\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2247
2248\newbox\myboxA
2249\begin{lrbox}{\myboxA}
2250\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2251coroutine C {} c;
2252void main( C & ) { for ( ;; ) { @suspend();@ } }
2253int main() {
2254        BENCH(
2255                for ( size_t i = 0; i < N; i += 1 ) { @resume( c );@ } )
2256        sout | result`ns;
2257}
2258\end{cfa}
2259\end{lrbox}
2260
2261\newbox\myboxB
2262\begin{lrbox}{\myboxB}
2263\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2264
2265
2266int main() {
2267        BENCH(
2268                for ( size_t i = 0; i < N; i += 1 ) { @yield();@ } )
2269        sout | result`ns;
2270}
2271\end{cfa}
2272\end{lrbox}
2273
2274\subfloat[Coroutine]{\usebox\myboxA}
2275\quad
2276\subfloat[Thread]{\usebox\myboxB}
2277\captionof{figure}{\CFA context-switch benchmark}
2278\label{f:ctx-switch}
2279
2280\centering
2281
2282\captionof{table}{Context switch comparison (nanoseconds)}
2283\label{tab:ctx-switch}
2284\bigskip
2285\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2286\cline{2-4}
2287\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2288\hline
2289Kernel Thread   & 333.5 & 332.96        & 4.1 \\
2290\CFA Coroutine  & 49            & 48.68 & 0.47    \\
2291\CFA Thread             & 105           & 105.57        & 1.37 \\
2292\uC Coroutine   & 44            & 44            & 0 \\
2293\uC Thread              & 100           & 99.29 & 0.96 \\
2294Goroutine               & 145           & 147.25        & 4.15 \\
2295Java Thread             & 373.5 & 375.14        & 8.72 \\
2296\hline
2297\end{tabular}
2298
2299\bigskip
2300\bigskip
2301
2302\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2303\begin{cfa}
2304monitor M { ... } m1/*, m2, m3, m4*/;
2305void __attribute__((noinline)) do_call( M & mutex m/*, m2, m3, m4*/ ) {}
2306int main() {
2307        BENCH( for( size_t i = 0; i < N; i += 1 ) { @do_call( m1/*, m2, m3, m4*/ );@ } )
2308        sout | result`ns;
2309}
2310\end{cfa}
2311\captionof{figure}{\CFA acquire/release mutex benchmark}
2312\label{f:mutex}
2313
2314\centering
2315
2316\captionof{table}{Mutex comparison (nanoseconds)}
2317\label{tab:mutex}
2318\bigskip
2319
2320\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2321\cline{2-4}
2322\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2323\hline
2324C routine                                       & 2             & 2             & 0    \\
2325FetchAdd + FetchSub                     & 26            & 26            & 0    \\
2326Pthreads Mutex Lock                     & 31            & 31.71 & 0.97 \\
2327\uC @monitor@ member routine            & 31            & 31            & 0    \\
2328\CFA @mutex@ routine, 1 argument        & 46            & 46.68 & 0.93  \\
2329\CFA @mutex@ routine, 2 argument        & 84            & 85.36 & 1.99 \\
2330\CFA @mutex@ routine, 4 argument        & 158           & 161           & 4.22 \\
2331Java synchronized routine               & 27.5  & 29.79 & 2.93  \\
2332\hline
2333\end{tabular}
2334\end{figure}
2335
2336
2337\paragraph{Mutual-Exclusion}
2338
2339Mutual exclusion is measured by entering/leaving a critical section.
2340For monitors, entering and leaving a monitor routine is measured.
2341Figure~\ref{f:mutex} shows the code for \CFA with all results in Table~\ref{tab:mutex}.
2342To put the results in context, the cost of entering a non-inline routine and the cost of acquiring and releasing a @pthread_mutex@ lock is also measured.
2343Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2344
2345
2346\paragraph{Internal Scheduling}
2347
2348Internal scheduling is measured by waiting on and signalling a condition variable.
2349Figure~\ref{f:int-sched} shows the code for \CFA, with results in Table~\ref{tab:int-sched}.
2350Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2351Java scheduling is significantly greater because the benchmark explicitly creates multiple thread in order to prevent the JIT from making the program sequential, \ie removing all locking.
2352
2353\begin{figure}
2354\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2355\begin{cfa}
2356volatile int go = 0;
2357condition c;
2358monitor M { ... } m;
2359void __attribute__((noinline)) do_call( M & mutex a1 ) { signal( c ); }
2360thread T {};
2361void main( T & this ) {
2362        while ( go == 0 ) { yield(); }  // wait for other thread to start
2363        while ( go == 1 ) { @do_call( m );@ }
2364}
2365int  __attribute__((noinline)) do_wait( M & mutex m ) {
2366        go = 1; // continue other thread
2367        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @wait( c );@ } );
2368        go = 0; // stop other thread
2369        sout | result`ns;
2370}
2371int main() {
2372        T t;
2373        do_wait( m );
2374}
2375\end{cfa}
2376\captionof{figure}{\CFA Internal-scheduling benchmark}
2377\label{f:int-sched}
2378
2379\centering
2380\captionof{table}{Internal-scheduling comparison (nanoseconds)}
2381\label{tab:int-sched}
2382\bigskip
2383
2384\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2385\cline{2-4}
2386\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2387\hline
2388Pthreads Condition Variable             & 6005  & 5681.43       & 835.45 \\
2389\uC @signal@                                    & 324           & 325.54        & 3,02   \\
2390\CFA @signal@, 1 @monitor@              & 368.5         & 370.61        & 4.77   \\
2391\CFA @signal@, 2 @monitor@              & 467           & 470.5 & 6.79   \\
2392\CFA @signal@, 4 @monitor@              & 700.5         & 702.46        & 7.23  \\
2393Java @notify@                                   & 15471 & 172511        & 5689 \\
2394\hline
2395\end{tabular}
2396\end{figure}
2397
2398
2399\paragraph{External Scheduling}
2400
2401External scheduling is measured by accepting a call using the @waitfor@ statement (@_Accept@ in \uC).
2402Figure~\ref{f:ext-sched} shows the code for \CFA, with results in Table~\ref{tab:ext-sched}.
2403Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2404
2405\begin{figure}
2406\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2407\begin{cfa}
2408volatile int go = 0;
2409monitor M { ... } m;
2410thread T {};
2411void __attribute__((noinline)) do_call( M & mutex ) {}
2412void main( T & ) {
2413        while ( go == 0 ) { yield(); }  // wait for other thread to start
2414        while ( go == 1 ) { @do_call( m );@ }
2415}
2416int __attribute__((noinline)) do_wait( M & mutex m ) {
2417        go = 1; // continue other thread
2418        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @waitfor( do_call, m );@ } )
2419        go = 0; // stop other thread
2420        sout | result`ns;
2421}
2422int main() {
2423        T t;
2424        do_wait( m );
2425}
2426\end{cfa}
2427\captionof{figure}{\CFA external-scheduling benchmark}
2428\label{f:ext-sched}
2429
2430\centering
2431
2432\captionof{table}{External-scheduling comparison (nanoseconds)}
2433\label{tab:ext-sched}
2434\bigskip
2435\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2436\cline{2-4}
2437\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2438\hline
2439\uC @_Accept@                           & 358           & 359.11        & 2.53  \\
2440\CFA @waitfor@, 1 @monitor@     & 359           & 360.93        & 4.07  \\
2441\CFA @waitfor@, 2 @monitor@     & 450           & 449.39        & 6.62  \\
2442\CFA @waitfor@, 4 @monitor@     & 652           & 655.64        & 7.73 \\
2443\hline
2444\end{tabular}
2445
2446\bigskip
2447\medskip
2448
2449\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2450\begin{cfa}
2451thread MyThread {};
2452void main( MyThread & ) {}
2453int main() {
2454        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @MyThread m;@ } )
2455        sout | result`ns;
2456}
2457\end{cfa}
2458\captionof{figure}{\CFA object-creation benchmark}
2459\label{f:creation}
2460
2461\centering
2462
2463\captionof{table}{Creation comparison (nanoseconds)}
2464\label{tab:creation}
2465\bigskip
2466
2467\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2468\cline{2-4}
2469\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} & \multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2470\hline
2471Pthreads                                & 28091         & 28073.39      & 163.1  \\
2472\CFA Coroutine Lazy             & 6                     & 6.07          & 0.26   \\
2473\CFA Coroutine Eager    & 520           & 520.61        & 2.04   \\
2474\CFA Thread                             & 2032  & 2016.29       & 112.07  \\
2475\uC Coroutine                   & 106           & 107.36        & 1.47   \\
2476\uC Thread                              & 536.5 & 537.07        & 4.64   \\
2477Goroutine                               & 3103  & 3086.29       & 90.25  \\
2478Java Thread                             & 103416.5      & 103732.29     & 1137 \\
2479\hline
2480\end{tabular}
2481\end{figure}
2482
2483
2484\paragraph{Object Creation}
2485
2486Object creation is measured by creating/deleting the specific kind of concurrent object.
2487Figure~\ref{f:creation} shows the code for \CFA, with results in Table~\ref{tab:creation}.
2488The only note here is that the call stacks of \CFA coroutines are lazily created, therefore without priming the coroutine to force stack creation, the creation cost is artificially low.
2489
2490
2491\section{Conclusion}
2492
2493This paper demonstrates a concurrency API that is simple, efficient, and able to build higher-level concurrency features.
2494The approach provides concurrency based on a preemptive M:N user-level threading-system, executing in clusters, which encapsulate scheduling of work on multiple kernel threads providing parallelism.
2495The M:N model is judged to be efficient and provide greater flexibility than a 1:1 threading model.
2496High-level objects (monitor/task) are the core mechanism for mutual exclusion and synchronization.
2497A novel aspect is allowing multiple mutex-objects to be accessed simultaneously reducing the potential for deadlock for this complex scenario.
2498These concepts and the entire \CFA runtime-system are written in the \CFA language, demonstrating the expressiveness of the \CFA language.
2499Performance comparisons with other concurrent systems/languages show the \CFA approach is competitive across all low-level operations, which translates directly into good performance in well-written concurrent applications.
2500C programmers should feel comfortable using these mechanisms for developing concurrent applications, with the ability to obtain maximum available performance by mechanisms at the appropriate level.
2501
2502
2503\section{Future Work}
2504
2505While concurrency in \CFA has a strong start, development is still underway and there are missing features.
2506
2507\paragraph{Flexible Scheduling}
2508\label{futur:sched}
2509
2510An important part of concurrency is scheduling.
2511Different scheduling algorithms can affect performance (both in terms of average and variation).
2512However, no single scheduler is optimal for all workloads and therefore there is value in being able to change the scheduler for given programs.
2513One solution is to offer various tweaking options, allowing the scheduler to be adjusted to the requirements of the workload.
2514However, to be truly flexible, a pluggable scheduler is necessary.
2515Currently, the \CFA pluggable scheduler is too simple to handle complex scheduling, \eg quality of service and real-time, where the scheduler must interact with mutex objects to deal with issues like priority inversion.
2516
2517\paragraph{Non-Blocking I/O}
2518\label{futur:nbio}
2519
2520Many modern workloads are not bound by computation but IO operations, a common case being web servers and XaaS~\cite{XaaS} (anything as a service).
2521These types of workloads require significant engineering to amortizing costs of blocking IO-operations.
2522At its core, non-blocking I/O is an operating-system level feature queuing IO operations, \eg network operations, and registering for notifications instead of waiting for requests to complete.
2523Current trends use asynchronous programming like callbacks, futures, and/or promises, \eg Node.js~\cite{NodeJs} for JavaScript, Spring MVC~\cite{SpringMVC} for Java, and Django~\cite{Django} for Python.
2524However, these solutions lead to code that is hard to create, read and maintain.
2525A better approach is to tie non-blocking I/O into the concurrency system to provide ease of use with low overhead, \eg thread-per-connection web-services.
2526A non-blocking I/O library is currently under development for \CFA.
2527
2528\paragraph{Other Concurrency Tools}
2529\label{futur:tools}
2530
2531While monitors offer flexible and powerful concurrency for \CFA, other concurrency tools are also necessary for a complete multi-paradigm concurrency package.
2532Examples of such tools can include futures and promises~\cite{promises}, executors and actors.
2533These additional features are useful when monitors offer a level of abstraction that is inadequate for certain tasks.
2534As well, new \CFA extensions should make it possible to create a uniform interface for virtually all mutual exclusion, including monitors and low-level locks.
2535
2536\paragraph{Implicit Threading}
2537\label{futur:implcit}
2538
2539Basic concurrent (embarrassingly parallel) applications can benefit greatly from implicit concurrency, where sequential programs are converted to concurrent, possibly with some help from pragmas to guide the conversion.
2540This type of concurrency can be achieved both at the language level and at the library level.
2541The canonical example of implicit concurrency is concurrent nested @for@ loops, which are amenable to divide and conquer algorithms~\cite{uC++book}.
2542The \CFA language features should make it possible to develop a reasonable number of implicit concurrency mechanism to solve basic HPC data-concurrency problems.
2543However, implicit concurrency is a restrictive solution with significant limitations, so it can never replace explicit concurrent programming.
2544
2545
2546\section{Acknowledgements}
2547
2548The authors would like to recognize the design assistance of Aaron Moss, Rob Schluntz and Andrew Beach on the features described in this paper.
2549Funding for this project has been provided by Huawei Ltd.\ (\url{http://www.huawei.com}), and Peter Buhr is partially funded by the Natural Sciences and Engineering Research Council of Canada.
2550
2551{%
2552\fontsize{9bp}{12bp}\selectfont%
2553\bibliography{pl,local}
2554}%
2555
2556\end{document}
2557
2558% Local Variables: %
2559% tab-width: 4 %
2560% fill-column: 120 %
2561% compile-command: "make" %
2562% End: %
Note: See TracBrowser for help on using the repository browser.