source: doc/papers/concurrency/Paper.tex @ 64188cc

aaron-thesisarm-ehcleanup-dtorsdeferred_resndemanglerjacob/cs343-translationjenkins-sandboxnew-astnew-ast-unique-exprno_listpersistent-indexer
Last change on this file since 64188cc was 64188cc, checked in by Peter A. Buhr <pabuhr@…>, 4 years ago

more updates

  • Property mode set to 100644
File size: 121.6 KB
Line 
1\documentclass[AMA,STIX1COL]{WileyNJD-v2}
2
3\articletype{RESEARCH ARTICLE}%
4
5\received{26 April 2016}
6\revised{6 June 2016}
7\accepted{6 June 2016}
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{\textfraction}{0.0}                       % the entire page maybe devoted to floats with no text on the page at all
35
36\lefthyphenmin=3                                                        % hyphen only after 4 characters
37\righthyphenmin=3
38
39%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
40
41% Names used in the document.
42
43\newcommand{\CFAIcon}{\textsf{C}\raisebox{\depth}{\rotatebox{180}{\textsf{A}}}\xspace} % Cforall symbolic name
44\newcommand{\CFA}{\protect\CFAIcon}             % safe for section/caption
45\newcommand{\CFL}{\textrm{Cforall}\xspace}      % Cforall symbolic name
46\newcommand{\Celeven}{\textrm{C11}\xspace}      % C11 symbolic name
47\newcommand{\CC}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}\xspace} % C++ symbolic name
48\newcommand{\CCeleven}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}11\xspace} % C++11 symbolic name
49\newcommand{\CCfourteen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}14\xspace} % C++14 symbolic name
50\newcommand{\CCseventeen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}17\xspace} % C++17 symbolic name
51\newcommand{\CCtwenty}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}20\xspace} % C++20 symbolic name
52\newcommand{\Csharp}{C\raisebox{-0.7ex}{\Large$^\sharp$}\xspace} % C# symbolic name
53
54%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
55
56\newcommand{\Textbf}[2][red]{{\color{#1}{\textbf{#2}}}}
57\newcommand{\Emph}[2][red]{{\color{#1}\textbf{\emph{#2}}}}
58\newcommand{\uC}{$\mu$\CC}
59\newcommand{\TODO}[1]{{\Textbf{#1}}}
60
61%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
62
63% Default underscore is too low and wide. Cannot use lstlisting "literate" as replacing underscore
64% removes it as a variable-name character so keywords in variables are highlighted. MUST APPEAR
65% AFTER HYPERREF.
66%\DeclareTextCommandDefault{\textunderscore}{\leavevmode\makebox[1.2ex][c]{\rule{1ex}{0.1ex}}}
67\renewcommand{\textunderscore}{\leavevmode\makebox[1.2ex][c]{\rule{1ex}{0.075ex}}}
68
69\renewcommand*{\thefootnote}{\Alph{footnote}} % hack because fnsymbol does not work
70%\renewcommand*{\thefootnote}{\fnsymbol{footnote}}
71
72\makeatletter
73% parindent is relative, i.e., toggled on/off in environments like itemize, so store the value for
74% use rather than use \parident directly.
75\newlength{\parindentlnth}
76\setlength{\parindentlnth}{\parindent}
77
78\newcommand{\LstBasicStyle}[1]{{\lst@basicstyle{\lst@basicstyle{#1}}}}
79\newcommand{\LstKeywordStyle}[1]{{\lst@basicstyle{\lst@keywordstyle{#1}}}}
80\newcommand{\LstCommentStyle}[1]{{\lst@basicstyle{\lst@commentstyle{#1}}}}
81
82\newlength{\gcolumnposn}                                        % temporary hack because lstlisting does not handle tabs correctly
83\newlength{\columnposn}
84\setlength{\gcolumnposn}{3.5in}
85\setlength{\columnposn}{\gcolumnposn}
86
87\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}}}}
88\newcommand{\CRT}{\global\columnposn=\gcolumnposn}
89
90% Denote newterms in particular font and index them without particular font and in lowercase, e.g., \newterm{abc}.
91% The option parameter provides an index term different from the new term, e.g., \newterm[\texttt{abc}]{abc}
92% The star version does not lowercase the index information, e.g., \newterm*{IBM}.
93\newcommand{\newtermFontInline}{\emph}
94\newcommand{\newterm}{\@ifstar\@snewterm\@newterm}
95\newcommand{\@newterm}[2][\@empty]{\lowercase{\def\temp{#2}}{\newtermFontInline{#2}}\ifx#1\@empty\index{\temp}\else\index{#1@{\protect#2}}\fi}
96\newcommand{\@snewterm}[2][\@empty]{{\newtermFontInline{#2}}\ifx#1\@empty\index{#2}\else\index{#1@{\protect#2}}\fi}
97
98% Latin abbreviation
99\newcommand{\abbrevFont}{\textit}                       % set empty for no italics
100\@ifundefined{eg}{
101\newcommand{\EG}{\abbrevFont{e}\abbrevFont{g}}
102\newcommand*{\eg}{%
103        \@ifnextchar{,}{\EG}%
104                {\@ifnextchar{:}{\EG}%
105                        {\EG,\xspace}}%
106}}{}%
107\@ifundefined{ie}{
108\newcommand{\IE}{\abbrevFont{i}\abbrevFont{e}}
109\newcommand*{\ie}{%
110        \@ifnextchar{,}{\IE}%
111                {\@ifnextchar{:}{\IE}%
112                        {\IE,\xspace}}%
113}}{}%
114\@ifundefined{etc}{
115\newcommand{\ETC}{\abbrevFont{etc}}
116\newcommand*{\etc}{%
117        \@ifnextchar{.}{\ETC}%
118        {\ETC.\xspace}%
119}}{}%
120\@ifundefined{etal}{
121\newcommand{\ETAL}{\abbrevFont{et}~\abbrevFont{al}}
122\newcommand*{\etal}{%
123        \@ifnextchar{.}{\protect\ETAL}%
124                {\protect\ETAL.\xspace}%
125}}{}%
126\@ifundefined{viz}{
127\newcommand{\VIZ}{\abbrevFont{viz}}
128\newcommand*{\viz}{%
129        \@ifnextchar{.}{\VIZ}%
130                {\VIZ.\xspace}%
131}}{}%
132\makeatother
133
134\newenvironment{cquote}{%
135        \list{}{\lstset{resetmargins=true,aboveskip=0pt,belowskip=0pt}\topsep=3pt\parsep=0pt\leftmargin=\parindentlnth\rightmargin\leftmargin}%
136        \item\relax
137}{%
138        \endlist
139}% cquote
140
141% CFA programming language, based on ANSI C (with some gcc additions)
142\lstdefinelanguage{CFA}[ANSI]{C}{
143        morekeywords={
144                _Alignas, _Alignof, __alignof, __alignof__, asm, __asm, __asm__, __attribute, __attribute__,
145                auto, _Bool, catch, catchResume, choose, _Complex, __complex, __complex__, __const, __const__,
146                coroutine, disable, dtype, enable, exception, __extension__, fallthrough, fallthru, finally,
147                __float80, float80, __float128, float128, forall, ftype, _Generic, _Imaginary, __imag, __imag__,
148                inline, __inline, __inline__, __int128, int128, __label__, monitor, mutex, _Noreturn, one_t, or,
149                otype, restrict, __restrict, __restrict__, __signed, __signed__, _Static_assert, thread,
150                _Thread_local, throw, throwResume, timeout, trait, try, ttype, typeof, __typeof, __typeof__,
151                virtual, __volatile, __volatile__, waitfor, when, with, zero_t},
152        moredirectives={defined,include_next}%
153}
154
155\lstset{
156language=CFA,
157columns=fullflexible,
158basicstyle=\linespread{0.9}\sf,                                                 % reduce line spacing and use sanserif font
159stringstyle=\tt,                                                                                % use typewriter font
160tabsize=5,                                                                                              % N space tabbing
161xleftmargin=\parindentlnth,                                                             % indent code to paragraph indentation
162%mathescape=true,                                                                               % LaTeX math escape in CFA code $...$
163escapechar=\$,                                                                                  % LaTeX escape in CFA code
164keepspaces=true,                                                                                %
165showstringspaces=false,                                                                 % do not show spaces with cup
166showlines=true,                                                                                 % show blank lines at end of code
167aboveskip=4pt,                                                                                  % spacing above/below code block
168belowskip=3pt,
169% replace/adjust listing characters that look bad in sanserif
170literate={-}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.1ex}}}}1 {^}{\raisebox{0.6ex}{$\scriptstyle\land\,$}}1
171        {~}{\raisebox{0.3ex}{$\scriptstyle\sim\,$}}1 % {`}{\ttfamily\upshape\hspace*{-0.1ex}`}1
172        {<}{\textrm{\textless}}1 {>}{\textrm{\textgreater}}1
173        {<-}{$\leftarrow$}2 {=>}{$\Rightarrow$}2 {->}{\makebox[1ex][c]{\raisebox{0.5ex}{\rule{0.8ex}{0.075ex}}}\kern-0.2ex{\textrm{\textgreater}}}2,
174moredelim=**[is][\color{red}]{`}{`},
175}% lstset
176
177% uC++ programming language, based on ANSI C++
178\lstdefinelanguage{uC++}[ANSI]{C++}{
179        morekeywords={
180                _Accept, _AcceptReturn, _AcceptWait, _Actor, _At, _CatchResume, _Cormonitor, _Coroutine, _Disable,
181                _Else, _Enable, _Event, _Finally, _Monitor, _Mutex, _Nomutex, _PeriodicTask, _RealTimeTask,
182                _Resume, _Select, _SporadicTask, _Task, _Timeout, _When, _With, _Throw},
183}
184\lstdefinelanguage{Golang}{
185        morekeywords=[1]{package,import,func,type,struct,return,defer,panic,recover,select,var,const,iota,},
186        morekeywords=[2]{string,uint,uint8,uint16,uint32,uint64,int,int8,int16,int32,int64,
187                bool,float32,float64,complex64,complex128,byte,rune,uintptr, error,interface},
188        morekeywords=[3]{map,slice,make,new,nil,len,cap,copy,close,true,false,delete,append,real,imag,complex,chan,},
189        morekeywords=[4]{for,break,continue,range,goto,switch,case,fallthrough,if,else,default,},
190        morekeywords=[5]{Println,Printf,Error,},
191        sensitive=true,
192        morecomment=[l]{//},
193        morecomment=[s]{/*}{*/},
194        morestring=[b]',
195        morestring=[b]",
196        morestring=[s]{`}{`},
197}
198
199\lstnewenvironment{cfa}[1][]
200{\lstset{#1}}
201{}
202\lstnewenvironment{C++}[1][]                            % use C++ style
203{\lstset{language=C++,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
204{}
205\lstnewenvironment{uC++}[1][]
206{\lstset{#1}}
207{}
208\lstnewenvironment{Go}[1][]
209{\lstset{#1}}
210{}
211
212% inline code @...@
213\lstMakeShortInline@%
214
215\let\OLDthebibliography\thebibliography
216\renewcommand\thebibliography[1]{
217  \OLDthebibliography{#1}
218  \setlength{\parskip}{0pt}
219  \setlength{\itemsep}{4pt plus 0.3ex}
220}
221
222\title{\texorpdfstring{Concurrency in \protect\CFA}{Concurrency in Cforall}}
223
224\author[1]{Thierry Delisle}
225\author[1]{Peter A. Buhr*}
226\authormark{DELISLE \textsc{et al.}}
227
228\address[1]{\orgdiv{Cheriton School of Computer Science}, \orgname{University of Waterloo}, \orgaddress{\state{Waterloo, ON}, \country{Canada}}}
229
230\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}}
231
232\fundingInfo{Natural Sciences and Engineering Research Council of Canada}
233
234\abstract[Summary]{
235\CFA is a modern, polymorphic, \emph{non-object-oriented} extension of the C programming language.
236This paper discusses the design of the concurrency and parallelism features in \CFA, and the concurrent runtime-system.
237These features are created from scratch as ISO C lacks concurrency, relying largely on the pthreads library.
238Coroutines and lightweight (user) threads are introduced into the language.
239In addition, monitors are added as a high-level mechanism for mutual exclusion and synchronization.
240A unique contribution is allowing multiple monitors to be safely acquired simultaneously.
241All features respect the expectations of C programmers, while being fully integrate with the \CFA polymorphic type-system and other language features.
242Finally, experimental results are presented to compare the performance of the new features with similar mechanisms in other concurrent programming-languages.
243}%
244
245\keywords{concurrency, parallelism, coroutines, threads, monitors, runtime, C, Cforall}
246
247
248\begin{document}
249\linenumbers                                            % comment out to turn off line numbering
250
251\maketitle
252
253
254\section{Introduction}
255
256This paper provides a minimal concurrency \newterm{Application Program Interface} (API) that is simple, efficient and can be used to build other concurrency features.
257While the simplest concurrency system is a thread and a lock, this low-level approach is hard to master.
258An easier approach for programmers is to support higher-level constructs as the basis of concurrency.
259Indeed, for highly productive concurrent programming, high-level approaches are much more popular~\cite{Hochstein05}.
260Examples of high-level approaches are jobs (tasks) based~\cite{TBB}, implicit threading~\cite{OpenMP}, monitors~\cite{Java}, channels~\cite{CSP,Go}, and message passing~\cite{Erlang,MPI}.
261
262The following terminology is used.
263A \newterm{thread} is a fundamental unit of execution that runs a sequence of code and requires a stack to maintain state.
264Multiple simultaneous threads give rise to \newterm{concurrency}, which requires locking to ensure safe communication and access to shared data.
265% Correspondingly, concurrency is defined as the concepts and challenges that occur when multiple independent (sharing memory, timing dependencies, \etc) concurrent threads are introduced.
266\newterm{Locking}, and by extension \newterm{locks}, are defined as a mechanism to prevent progress of threads to provide safety.
267\newterm{Parallelism} is running multiple threads simultaneously.
268Parallelism implies \emph{actual} simultaneous execution, where concurrency only requires \emph{apparent} simultaneous execution.
269As such, parallelism only affects performance, which is observed through differences in space and/or time at runtime.
270
271Hence, there are two problems to be solved: concurrency and parallelism.
272While these two concepts are often combined, they are distinct, requiring different tools~\cite[\S~2]{Buhr05a}.
273Concurrency tools handle mutual exclusion and synchronization, while parallelism tools handle performance, cost, and resource utilization.
274
275The proposed concurrency API is implemented in a dialect of C, called \CFA.
276The paper discusses how the language features are added to the \CFA translator with respect to parsing, semantic, and type checking, and the corresponding high-performance runtime-library to implement the concurrency features.
277
278
279\section{\CFA Overview}
280
281The following is a quick introduction to the \CFA language, specifically tailored to the features needed to support concurrency.
282Extended versions and explanation of the following code examples are available at the \CFA website~\cite{Cforall} or in Moss~\etal~\cite{Moss18}.
283
284\CFA is a non-object-oriented extension of ISO-C, and hence, supports all C paradigms.
285%It is a non-object-oriented system-language, meaning most of the major abstractions have either no runtime overhead or can be opted out easily.
286Like C, the building blocks of \CFA are structures and routines.
287Virtually all of the code generated by the \CFA translator respects C memory layouts and calling conventions.
288While \CFA is not an object-oriented language, lacking the concept of a receiver (\eg @this@) and nominal inheritance-relationships, C does have a notion of objects: ``region of data storage in the execution environment, the contents of which can represent values''~\cite[3.15]{C11}.
289While some \CFA features are common in object-oriented programming-languages, they are an independent capability allowing \CFA to adopt them while retaining a procedural paradigm.
290
291
292\subsection{References}
293
294\CFA provides multi-level rebindable references, as an alternative to pointers, which significantly reduces syntactic noise.
295\begin{cfa}
296int x = 1, y = 2, z = 3;
297int * p1 = &x, ** p2 = &p1,  *** p3 = &p2,      $\C{// pointers to x}$
298    `&` r1 = x,   `&&` r2 = r1,   `&&&` r3 = r2;        $\C{// references to x}$
299int * p4 = &z, `&` r4 = z;
300
301*p1 = 3; **p2 = 3; ***p3 = 3;       // change x
302r1 =  3;     r2 = 3;      r3 = 3;        // change x: implicit dereferences *r1, **r2, ***r3
303**p3 = &y; *p3 = &p4;                // change p1, p2
304`&`r3 = &y; `&&`r3 = &`&`r4;             // change r1, r2: cancel implicit dereferences (&*)**r3, (&(&*)*)*r3, &(&*)r4
305\end{cfa}
306A reference is a handle to an object, like a pointer, but is automatically dereferenced the specified number of levels.
307Referencing (address-of @&@) a reference variable cancels one of the implicit dereferences, until there are no more implicit references, after which normal expression behaviour applies.
308
309
310\subsection{\texorpdfstring{\protect\lstinline{with} Statement}{with Statement}}
311\label{s:WithStatement}
312
313Heterogeneous data is aggregated into a structure/union.
314To 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.
315\begin{cquote}
316\vspace*{-\baselineskip}%???
317\lstDeleteShortInline@%
318\begin{cfa}
319struct S { char c; int i; double d; };
320struct T { double m, n; };
321// multiple aggregate parameters
322\end{cfa}
323\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
324\begin{cfa}
325void f( S & s, T & t ) {
326        `s.`c; `s.`i; `s.`d;
327        `t.`m; `t.`n;
328}
329\end{cfa}
330&
331\begin{cfa}
332void f( S & s, T & t ) `with ( s, t )` {
333        c; i; d;                // no qualification
334        m; n;
335}
336\end{cfa}
337\end{tabular}
338\lstMakeShortInline@%
339\end{cquote}
340Object-oriented programming languages only provide implicit qualification for the receiver.
341
342In detail, the @with@ statement has the form:
343\begin{cfa}
344$\emph{with-statement}$:
345        'with' '(' $\emph{expression-list}$ ')' $\emph{compound-statement}$
346\end{cfa}
347and may appear as the body of a routine or nested within a routine body.
348Each expression in the expression-list provides a type and object.
349The type must be an aggregate type.
350(Enumerations are already opened.)
351The object is the implicit qualifier for the open structure-fields.
352All expressions in the expression list are open in parallel within the compound statement, which is different from Pascal, which nests the openings from left to right.
353
354
355\subsection{Overloading}
356
357\CFA maximizes the ability to reuse names via overloading to aggressively address the naming problem.
358Both variables and routines may be overloaded, where selection is based on types, and number of returns (as in Ada~\cite{Ada}) and arguments.
359\begin{cquote}
360\vspace*{-\baselineskip}%???
361\lstDeleteShortInline@%
362\begin{cfa}
363// selection based on type
364\end{cfa}
365\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
366\begin{cfa}
367const short int `MIN` = -32768;
368const int `MIN` = -2147483648;
369const long int `MIN` = -9223372036854775808L;
370\end{cfa}
371&
372\begin{cfa}
373short int si = `MIN`;
374int i = `MIN`;
375long int li = `MIN`;
376\end{cfa}
377\end{tabular}
378\begin{cfa}
379// selection based on type and number of parameters
380\end{cfa}
381\begin{tabular}{@{}l@{\hspace{2.7\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
382\begin{cfa}
383void `f`( void );
384void `f`( char );
385void `f`( int, double );
386\end{cfa}
387&
388\begin{cfa}
389`f`();
390`f`( 'a' );
391`f`( 3, 5.2 );
392\end{cfa}
393\end{tabular}
394\begin{cfa}
395// selection based on type and number of returns
396\end{cfa}
397\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
398\begin{cfa}
399char `f`( int );
400double `f`( int );
401[char, double] `f`( int );
402\end{cfa}
403&
404\begin{cfa}
405char c = `f`( 3 );
406double d = `f`( 3 );
407[d, c] = `f`( 3 );
408\end{cfa}
409\end{tabular}
410\lstMakeShortInline@%
411\end{cquote}
412Overloading is important for \CFA concurrency since the runtime system relies on creating different types to represent concurrency objects.
413Therefore, overloading eliminates long prefixes and other naming conventions to prevent name clashes.
414As seen in Section~\ref{basics}, routine @main@ is heavily overloaded.
415For example, variable overloading is useful in the parallel semantics of the @with@ statement for fields with the same name:
416\begin{cfa}
417struct S { int `i`; int j; double m; } s;
418struct T { int `i`; int k; int m; } t;
419with ( s, t ) {
420        j + k;                                                                  $\C{// unambiguous, s.j + t.k}$
421        m = 5.0;                                                                $\C{// unambiguous, s.m = 5.0}$
422        m = 1;                                                                  $\C{// unambiguous, t.m = 1}$
423        int a = m;                                                              $\C{// unambiguous, a = t.m }$
424        double b = m;                                                   $\C{// unambiguous, b = s.m}$
425        int c = `s.i` + `t.i`;                                  $\C{// unambiguous, qualification}$
426        (double)m;                                                              $\C{// unambiguous, cast s.m}$
427}
428\end{cfa}
429For parallel semantics, both @s.i@ and @t.i@ are visible with the same type, so only @i@ is ambiguous without qualification.
430
431
432\subsection{Operators}
433
434Overloading also extends to operators.
435Operator-overloading syntax creates a routine name with an operator symbol and question marks for the operands:
436\begin{cquote}
437\lstDeleteShortInline@%
438\begin{tabular}{@{}ll@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
439\begin{cfa}
440int ++?(int op);
441int ?++(int op);
442int `?+?`(int op1, int op2);
443int ?<=?(int op1, int op2);
444int ?=? (int & op1, int op2);
445int ?+=?(int & op1, int op2);
446\end{cfa}
447&
448\begin{cfa}
449// unary prefix increment
450// unary postfix increment
451// binary plus
452// binary less than
453// binary assignment
454// binary plus-assignment
455\end{cfa}
456&
457\begin{cfa}
458struct S { int i, j; };
459S `?+?`( S op1, S op2) { // add two structures
460        return (S){op1.i + op2.i, op1.j + op2.j};
461}
462S s1 = {1, 2}, s2 = {2, 3}, s3;
463s3 = s1 `+` s2;         // compute sum: s3 == {2, 5}
464\end{cfa}
465\end{tabular}
466\lstMakeShortInline@%
467\end{cquote}
468While concurrency does not use operator overloading directly, it provides an introduction for the syntax of constructors.
469
470
471\subsection{Constructors / Destructors}
472
473Object lifetime is a challenge in non-managed programming languages.
474\CFA responds with \CC-like constructors and destructors:
475\begin{cfa}
476struct VLA { int len, * data; };                        $\C{// variable length array of integers}$
477void ?{}( VLA & vla ) with ( vla ) { len = 10;  data = alloc( len ); }  $\C{// default constructor}$
478void ?{}( VLA & vla, int size, char fill ) with ( vla ) { len = size;  data = alloc( len, fill ); } // initialization
479void ?{}( VLA & vla, VLA other ) { vla.len = other.len;  vla.data = other.data; } $\C{// copy, shallow}$
480void ^?{}( VLA & vla ) with ( vla ) { free( data ); } $\C{// destructor}$
481{
482        VLA  x,            y = { 20, 0x01 },     z = y; $\C{// z points to y}$
483        //    x{};         y{ 20, 0x01 };          z{ z, y };
484        ^x{};                                                                   $\C{// deallocate x}$
485        x{};                                                                    $\C{// reallocate x}$
486        z{ 5, 0xff };                                                   $\C{// reallocate z, not pointing to y}$
487        ^y{};                                                                   $\C{// deallocate y}$
488        y{ x };                                                                 $\C{// reallocate y, points to x}$
489        x{};                                                                    $\C{// reallocate x, not pointing to y}$
490        //  ^z{}^y{}^x{};
491}
492\end{cfa}
493Like \CC, construction is implicit on allocation (stack/heap) and destruction is implicit on deallocation.
494The object and all their fields are constructed/destructed.
495\CFA also provides @new@ and @delete@, which behave like @malloc@ and @free@, in addition to constructing and destructing objects:
496\begin{cfa}
497{
498        ... struct S s = {10}; ...                              $\C{// allocation, call constructor}$
499}                                                                                       $\C{// deallocation, call destructor}$
500struct S * s = new();                                           $\C{// allocation, call constructor}$
501...
502delete( s );                                                            $\C{// deallocation, call destructor}$
503\end{cfa}
504\CFA concurrency uses object lifetime as a means of mutual exclusion and/or synchronization.
505
506
507\subsection{Parametric Polymorphism}
508\label{s:ParametricPolymorphism}
509
510The 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.
511For example, the following sum routine works for any type that supports construction from 0 and addition:
512\begin{cfa}
513forall( otype T | { void `?{}`( T *, zero_t ); T `?+?`( T, T ); } ) // constraint type, 0 and +
514T sum( T a[$\,$], size_t size ) {
515        `T` total = { `0` };                                    $\C{// initialize by 0 constructor}$
516        for ( size_t i = 0; i < size; i += 1 )
517                total = total `+` a[i];                         $\C{// select appropriate +}$
518        return total;
519}
520S sa[5];
521int i = sum( sa, 5 );                                           $\C{// use S's 0 construction and +}$
522\end{cfa}
523The builtin type @zero_t@ (and @one_t@) overload constant 0 (and 1) for a new types, where both 0 and 1 have special meaning in C.
524
525\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:
526\begin{cfa}
527trait `sumable`( otype T ) {
528        void `?{}`( T &, zero_t );                              $\C{// 0 literal constructor}$
529        T `?+?`( T, T );                                                $\C{// assortment of additions}$
530        T ?+=?( T &, T );
531        T ++?( T & );
532        T ?++( T & );
533};
534forall( otype T `| sumable( T )` )                      $\C{// use trait}$
535T sum( T a[$\,$], size_t size );
536\end{cfa}
537
538Assertions can be @otype@ or @dtype@.
539@otype@ refers to a ``complete'' object, \ie an object has a size, default constructor, copy constructor, destructor and an assignment operator.
540@dtype@ only guarantees an object has a size and alignment.
541
542Using the return type for discrimination, it is possible to write a type-safe @alloc@ based on the C @malloc@:
543\begin{cfa}
544forall( dtype T | sized(T) ) T * alloc( void ) { return (T *)malloc( sizeof(T) ); }
545int * ip = alloc();                                                     $\C{// select type and size from left-hand side}$
546double * dp = alloc();
547struct S {...} * sp = alloc();
548\end{cfa}
549where the return type supplies the type/size of the allocation, which is impossible in most type systems.
550
551
552\section{Concurrency Basics}\label{basics}
553
554At its core, concurrency is based on multiple call-stacks and scheduling threads executing on these stacks.
555Multiple 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}.
556In coroutining, the single thread is self-scheduling across the stacks, so execution is deterministic, \ie given fixed inputs, the execution path to the outputs is fixed and predictable.
557A \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;
558a \newterm{stackfull} coroutine executes on its own stack, allowing full generality.
559Only stackfull coroutines are a stepping-stone to concurrency.
560
561The transition to concurrency, even for execution with a single thread and multiple stacks, occurs when coroutines also context switch to a scheduling oracle, introducing non-determinism from the coroutine perspective~\cite[\S~3]{Buhr05a}.
562Therefore, a minimal concurrency system is possible using coroutines (see Section \ref{coroutine}) in conjunction with a scheduler to decide where to context switch next.
563The resulting execution system now follows a cooperative threading-model, called \newterm{non-preemptive scheduling}.
564
565Because the scheduler is special, it can either be a stackless or stackfull coroutine.
566For 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.
567For stackfull, the current coroutine switches to the scheduler, which performs scheduling, and it then switches to the next coroutine, so there are two context switches.
568A stackfull scheduler is often used for simplicity and security, even through there is a slightly higher runtime-cost.
569
570Regardless of the approach used, a subset of concurrency related challenges start to appear.
571For 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}.
572While a scheduler introduces uncertainty in the order of execution, preemption introduces uncertainty where context switches occur.
573Interestingly, uncertainty is necessary for the runtime (operating) system to give the illusion of parallelism on a single processor and increase performance on multiple processors.
574The reason is that only the runtime has complete knowledge about resources and how to best utilized them.
575However, the introduction of unrestricted non-determinism results in the need for \newterm{mutual exclusion} and \newterm{synchronization} to restrict non-determinism for correctness;
576otherwise, it is impossible to write meaningful programs.
577Optimal performance in concurrent applications is often obtained by having as much non-determinism as correctness allows.
578
579
580\subsection{\protect\CFA's Thread Building Blocks}
581
582An 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.
583As such, library support for threading is far from widespread.
584At the time of writing the paper, neither \protect\lstinline@gcc@ nor \protect\lstinline@clang@ support \protect\lstinline@threads.h@ in their standard libraries.}.
585In 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.
586As 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.
587Furthermore, because C is a system-level language, programmers expect to choose precisely which features they need and which cost they are willing to pay.
588Hence, concurrent programs should be written using high-level mechanisms, and only step down to lower-level mechanisms when performance bottlenecks are encountered.
589
590
591\subsection{Coroutines: A Stepping Stone}\label{coroutine}
592
593While 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.
594Coroutines are generalized routines allowing execution to be temporarily suspend and later resumed.
595Hence, 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.
596This capability is accomplish via the coroutine's stack, where suspend/resume context switch among stacks.
597Because 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.
598Therefore, the core \CFA coroutine-API for has two fundamental features: independent call-stacks and @suspend@/@resume@ operations.
599
600For example, a problem made easier with coroutines is unbounded generators, \eg generating an infinite sequence of Fibonacci numbers, where Figure~\ref{f:C-fibonacci} shows conventional approaches for writing a Fibonacci generator in C.
601\begin{displaymath}
602\mathsf{fib}(n) = \left \{
603\begin{array}{ll}
6040                                       & n = 0         \\
6051                                       & n = 1         \\
606\mathsf{fib}(n-1) + \mathsf{fib}(n-2)   & n \ge 2       \\
607\end{array}
608\right.
609\end{displaymath}
610Figure~\ref{f:GlobalVariables} illustrates the following problems:
611unique unencapsulated global variables necessary to retain state between calls;
612only one Fibonacci generator;
613execution state must be explicitly retained via explicit state variables.
614Figure~\ref{f:ExternalState} addresses these issues:
615unencapsulated program global variables become encapsulated structure variables;
616unique global variables are replaced by multiple Fibonacci objects;
617explicit execution state is removed by precomputing the first two Fibonacci numbers and returning $\mathsf{fib}(n-2)$.
618
619\begin{figure}
620\centering
621\newbox\myboxA
622\begin{lrbox}{\myboxA}
623\begin{cfa}[aboveskip=0pt,belowskip=0pt]
624`int f1, f2, state = 1;`   // single global variables
625int fib() {
626        int fn;
627        `switch ( state )` {  // explicit execution state
628          case 1: fn = 0;  f1 = fn;  state = 2;  break;
629          case 2: fn = 1;  f2 = f1;  f1 = fn;  state = 3;  break;
630          case 3: fn = f1 + f2;  f2 = f1;  f1 = fn;  break;
631        }
632        return fn;
633}
634int main() {
635
636        for ( int i = 0; i < 10; i += 1 ) {
637                printf( "%d\n", fib() );
638        }
639}
640\end{cfa}
641\end{lrbox}
642
643\newbox\myboxB
644\begin{lrbox}{\myboxB}
645\begin{cfa}[aboveskip=0pt,belowskip=0pt]
646#define FIB_INIT `{ 0, 1 }`
647typedef struct { int f2, f1; } Fib;
648int fib( Fib * f ) {
649
650        int ret = f->f2;
651        int fn = f->f1 + f->f2;
652        f->f2 = f->f1; f->f1 = fn;
653
654        return ret;
655}
656int main() {
657        Fib f1 = FIB_INIT, f2 = FIB_INIT;
658        for ( int i = 0; i < 10; i += 1 ) {
659                printf( "%d %d\n", fib( &f1 ), fib( &f2 ) );
660        }
661}
662\end{cfa}
663\end{lrbox}
664
665\subfloat[3 States: global variables]{\usebox\myboxA}
666\qquad
667\subfloat[1 State: external variables]{\usebox\myboxB}
668\caption{C Fibonacci Implementations}
669\label{f:C-fibonacci}
670
671\bigskip
672
673\newbox\myboxA
674\begin{lrbox}{\myboxA}
675\begin{cfa}[aboveskip=0pt,belowskip=0pt]
676`coroutine` Fib { int fn; };
677void main( Fib & fib ) with( fib ) {
678        int f1, f2;
679        fn = 0;  f1 = fn;  `suspend()`;
680        fn = 1;  f2 = f1;  f1 = fn;  `suspend()`;
681        for ( ;; ) {
682                fn = f1 + f2;  f2 = f1;  f1 = fn;  `suspend()`;
683        }
684}
685int next( Fib & fib ) with( fib ) {
686        `resume( fib );`
687        return fn;
688}
689int main() {
690        Fib f1, f2;
691        for ( int i = 1; i <= 10; i += 1 ) {
692                sout | next( f1 ) | next( f2 ) | endl;
693        }
694}
695\end{cfa}
696\end{lrbox}
697\newbox\myboxB
698\begin{lrbox}{\myboxB}
699\begin{cfa}[aboveskip=0pt,belowskip=0pt]
700`coroutine` Fib { int ret; };
701void main( Fib & f ) with( fib ) {
702        int fn, f1 = 1, f2 = 0;
703        for ( ;; ) {
704                ret = f2;
705
706                fn = f1 + f2;  f2 = f1;  f1 = fn; `suspend();`
707        }
708}
709int next( Fib & fib ) with( fib ) {
710        `resume( fib );`
711        return ret;
712}
713
714
715
716
717
718
719\end{cfa}
720\end{lrbox}
721\subfloat[3 States, internal variables]{\label{f:Coroutine3States}\usebox\myboxA}
722\qquad\qquad
723\subfloat[1 State, internal variables]{\label{f:Coroutine1State}\usebox\myboxB}
724\caption{\CFA Coroutine Fibonacci Implementations}
725\label{f:fibonacci-cfa}
726\end{figure}
727
728Using a coroutine, it is possible to express the Fibonacci formula directly without any of the C problems.
729Figure~\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@.
730Like 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.
731The coroutine main's stack holds the state for the next generation, @f1@ and @f2@, and the code has the three suspend points, representing the three states in the Fibonacci formula, to context switch back to the caller's @resume@.
732The interface routine @next@, takes a Fibonacci instance and context switches to it using @resume@;
733on restart, the Fibonacci field, @fn@, contains the next value in the sequence, which is returned.
734The first @resume@ is special because it cocalls the coroutine at its coroutine main and allocates the stack;
735when the coroutine main returns, its stack is deallocated.
736Hence, @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.
737Figure~\ref{f:Coroutine1State} shows the coroutine version of the C version in Figure~\ref{f:ExternalState}.
738Coroutine generators are called \newterm{output coroutines} because values are only returned.
739
740Figure~\ref{f:CFAFmt} shows an \newterm{input coroutine}, @Format@, for restructuring text into groups of characters of fixed-size blocks.
741For example, the input of the left is reformatted into the output on the right.
742\begin{quote}
743\tt
744\begin{tabular}{@{}l|l@{}}
745\multicolumn{1}{c|}{\textbf{\textrm{input}}} & \multicolumn{1}{c}{\textbf{\textrm{output}}} \\
746abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz
747&
748\begin{tabular}[t]{@{}lllll@{}}
749abcd    & efgh  & ijkl  & mnop  & qrst  \\
750uvwx    & yzab  & cdef  & ghij  & klmn  \\
751opqr    & stuv  & wxyz  &               &
752\end{tabular}
753\end{tabular}
754\end{quote}
755The 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.
756The destruction provides a newline if formatted text ends with a full line.
757Figure~\ref{f:CFmt} shows the C equivalent formatter, where the loops of the coroutine are flatten (linearized) and rechecked on each call because execution location is not retained between calls.
758
759\begin{figure}
760\centering
761\newbox\myboxA
762\begin{lrbox}{\myboxA}
763\begin{cfa}[aboveskip=0pt,belowskip=0pt]
764`coroutine` Format {
765        char ch;   // used for communication
766        int g, b;  // global because used in destructor
767};
768void main( Format & fmt ) with( fmt ) {
769        for ( ;; ) {   
770                for ( g = 0; g < 5; g += 1 ) {      // group
771                        for ( b = 0; b < 4; b += 1 ) { // block
772                                `suspend();`
773                                sout | ch;              // separator
774                        }
775                        sout | "  ";               // separator
776                }
777                sout | endl;
778        }
779}
780void ?{}( Format & fmt ) { `resume( fmt );` }
781void ^?{}( Format & fmt ) with( fmt ) {
782        if ( g != 0 || b != 0 ) sout | endl;
783}
784void format( Format & fmt ) {
785        `resume( fmt );`
786}
787int main() {
788        Format fmt;
789        eof: for ( ;; ) {
790                sin | fmt.ch;
791          if ( eof( sin ) ) break eof;
792                format( fmt );
793        }
794}
795\end{cfa}
796\end{lrbox}
797
798\newbox\myboxB
799\begin{lrbox}{\myboxB}
800\begin{cfa}[aboveskip=0pt,belowskip=0pt]
801struct Format {
802        char ch;
803        int g, b;
804};
805void format( struct Format * fmt ) {
806        if ( fmt->ch != -1 ) {      // not EOF ?
807                printf( "%c", fmt->ch );
808                fmt->b += 1;
809                if ( fmt->b == 4 ) {  // block
810                        printf( "  " );      // separator
811                        fmt->b = 0;
812                        fmt->g += 1;
813                }
814                if ( fmt->g == 5 ) {  // group
815                        printf( "\n" );     // separator
816                        fmt->g = 0;
817                }
818        } else {
819                if ( fmt->g != 0 || fmt->b != 0 ) printf( "\n" );
820        }
821}
822int main() {
823        struct Format fmt = { 0, 0, 0 };
824        for ( ;; ) {
825                scanf( "%c", &fmt.ch );
826          if ( feof( stdin ) ) break;
827                format( &fmt );
828        }
829        fmt.ch = -1;
830        format( &fmt );
831}
832\end{cfa}
833\end{lrbox}
834\subfloat[\CFA Coroutine]{\label{f:CFAFmt}\usebox\myboxA}
835\qquad
836\subfloat[C Linearized]{\label{f:CFmt}\usebox\myboxB}
837\caption{Formatting text into lines of 5 blocks of 4 characters.}
838\label{f:fmt-line}
839\end{figure}
840
841The 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.
842However,@resume@/@suspend@ context switch to existing stack-frames rather than create new ones so there is no stack growth.
843\newterm{Symmetric (full) coroutine}s have a coroutine call a resuming routine for another coroutine, which eventually forms a resuming-call cycle.
844(The trivial cycle is a coroutine resuming itself.)
845This control flow is similar to recursion for normal routines, but again there is no stack growth from the context switch.
846
847\begin{figure}
848\centering
849\lstset{language=CFA,escapechar={},moredelim=**[is][\protect\color{red}]{`}{`}}% allow $
850\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
851\begin{cfa}
852`coroutine` Prod {
853        Cons & c;
854        int N, money, receipt;
855};
856void main( Prod & prod ) with( prod ) {
857        // 1st resume starts here
858        for ( int i = 0; i < N; i += 1 ) {
859                int p1 = random( 100 ), p2 = random( 100 );
860                sout | p1 | " " | p2 | endl;
861                int status = delivery( c, p1, p2 );
862                sout | " $" | money | endl | status | endl;
863                receipt += 1;
864        }
865        stop( c );
866        sout | "prod stops" | endl;
867}
868int payment( Prod & prod, int money ) {
869        prod.money = money;
870        `resume( prod );`
871        return prod.receipt;
872}
873void start( Prod & prod, int N, Cons &c ) {
874        &prod.c = &c;
875        prod.[N, receipt] = [N, 0];
876        `resume( prod );`
877}
878int main() {
879        Prod prod;
880        Cons cons = { prod };
881        start( prod, 5, cons );
882}
883\end{cfa}
884&
885\begin{cfa}
886`coroutine` Cons {
887        Prod & p;
888        int p1, p2, status;
889        _Bool done;
890};
891void ?{}( Cons & cons, Prod & p ) {
892        &cons.p = &p;
893        cons.[status, done ] = [0, false];
894}
895void ^?{}( Cons & cons ) {}
896void main( Cons & cons ) with( cons ) {
897        // 1st resume starts here
898        int money = 1, receipt;
899        for ( ; ! done; ) {
900                sout | p1 | " " | p2 | endl | " $" | money | endl;
901                status += 1;
902                receipt = payment( p, money );
903                sout | " #" | receipt | endl;
904                money += 1;
905        }
906        sout | "cons stops" | endl;
907}
908int delivery( Cons & cons, int p1, int p2 ) {
909        cons.[p1, p2] = [p1, p2];
910        `resume( cons );`
911        return cons.status;
912}
913void stop( Cons & cons ) {
914        cons.done = true;
915        `resume( cons );`
916}
917\end{cfa}
918\end{tabular}
919\caption{Producer / consumer: resume-resume cycle, bi-directional communication}
920\label{f:ProdCons}
921\end{figure}
922
923Figure~\ref{f:ProdCons} shows a producer/consumer symmetric-coroutine performing bi-directional communication.
924Since 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@.
925The @start@ routine communicates both the number of elements to be produced and the consumer into the producer's coroutine structure.
926Then the @resume@ to @prod@ creates @prod@'s stack with a frame for @prod@'s coroutine main at the top, and context switches to it.
927@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.
928
929The producer call to @delivery@ transfers values into the consumer's communication variables, resumes the consumer, and returns the consumer status.
930For the first resume, @cons@'s stack is initialized, creating local variables retained between subsequent activations of the coroutine.
931The 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).
932The call from the consumer to the @payment@ introduces the cycle between producer and consumer.
933When @payment@ is called, the consumer copies values into the producer's communication variable and a resume is executed.
934The context switch restarts the producer at the point where it was last context switched, so it continues in @delivery@ after the resume.
935
936@delivery@ returns the status value in @prod@'s coroutine main, where the status is printed.
937The loop then repeats calling @delivery@, where each call resumes the consumer coroutine.
938The context switch to the consumer continues in @payment@.
939The consumer increments and returns the receipt to the call in @cons@'s coroutine main.
940The loop then repeats calling @payment@, where each call resumes the producer coroutine.
941
942After iterating $N$ times, the producer calls @stop@.
943The @done@ flag is set to stop the consumer's execution and a resume is executed.
944The context switch restarts @cons@ in @payment@ and it returns with the last receipt.
945The 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@.
946@stop@ returns and @prod@'s coroutine main terminates.
947The program main restarts after the resume in @start@.
948@start@ returns and the program main terminates.
949
950
951\subsection{Coroutine Implementation}
952
953A 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.
954There are several solutions to this problem and the chosen option forced the \CFA coroutine design.
955
956Object-oriented inheritance provides extra fields and code in a restricted context, but it requires programmers to explicitly perform the inheritance:
957\begin{cfa}
958struct mycoroutine $\textbf{\textsf{inherits}}$ baseCoroutine { ... }
959\end{cfa}
960and the programming language (and possibly its tool set, \eg debugger) may need to understand @baseCoroutine@ because of the stack.
961Furthermore, the execution of constructs/destructors is in the wrong order for certain operations.
962For 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.
963
964An alternatively is composition:
965\begin{cfa}
966struct mycoroutine {
967        ... // declarations
968        baseCoroutine dummy; // composition, last declaration
969}
970\end{cfa}
971which also requires an explicit declaration that must be the last one to ensure correct initialization order.
972However, there is nothing preventing wrong placement or multiple declarations.
973
974For coroutines as for threads, many implementations are based on routine pointers or routine objects~\cite{Butenhof97, C++14, MS:VisualC++, BoostCoroutines15}.
975For example, Boost implements coroutines in terms of four functor object-types:
976\begin{cfa}
977asymmetric_coroutine<>::pull_type
978asymmetric_coroutine<>::push_type
979symmetric_coroutine<>::call_type
980symmetric_coroutine<>::yield_type
981\end{cfa}
982Similarly, the canonical threading paradigm is often based on routine pointers, \eg @pthreads@~\cite{Butenhof97}, \Csharp~\cite{Csharp}, Go~\cite{Go}, and Scala~\cite{Scala}.
983However, the generic thread-handle (identifier) is limited (few operations), unless it is wrapped in a custom type.
984\begin{cfa}
985void mycor( coroutine_t cid, void * arg ) {
986        int * value = (int *)arg;                               $\C{// type unsafe, pointer-size only}$
987        // Coroutine body
988}
989int main() {
990        int input = 0, output;
991        coroutine_t cid = coroutine_create( &mycor, (void *)&input ); $\C{// type unsafe, pointer-size only}$
992        coroutine_resume( cid, (void *)input, (void **)&output ); $\C{// type unsafe, pointer-size only}$
993}
994\end{cfa}
995Since the custom type is simple to write in \CFA and solves several issues, added support for routine/lambda-based coroutines adds very little.
996
997Note, the type @coroutine_t@ must be an abstract handle to the coroutine, because the coroutine descriptor and its stack are non-copyable.
998Copying the coroutine descriptor results in copies being out of date with the current state of the stack.
999Correspondingly, 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.
1000(There is no mechanism in C to find all stack-specific pointers and update them as part of a copy.)
1001
1002The selected approach is to use language support by introducing a new kind of aggregate (structure):
1003\begin{cfa}
1004coroutine Fibonacci {
1005        int fn; // communication variables
1006};
1007\end{cfa}
1008The @coroutine@ keyword means the compiler (and tool set) can find and inject code where needed.
1009The downside of this approach is that it makes coroutine a special case in the language.
1010Users wanting to extend coroutines or build their own for various reasons can only do so in ways offered by the language.
1011Furthermore, implementing coroutines without language supports also displays the power of a programming language.
1012While this is ultimately the option used for idiomatic \CFA code, coroutines and threads can still be constructed without using the language support.
1013The reserved keyword simply eases use for the common cases.
1014
1015Part 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 is used to restrict coroutine-manipulation routines:
1016\begin{cfa}
1017trait is_coroutine( `dtype` T ) {
1018        void main( T & );
1019        coroutine_desc * get_coroutine( T & );
1020};
1021forall( `dtype` T | is_coroutine(T) ) void suspend( T & );
1022forall( `dtype` T | is_coroutine(T) ) void resume( T & );
1023\end{cfa}
1024The @dtype@ property of the trait ensures the coroutine descriptor is non-copyable, so all coroutines must be passed by reference (pointer).
1025The 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.
1026The @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.
1027The generic routines @suspend@ and @resume@ can be redefined, but any object passed to them is a coroutine since it must satisfy the @is_coroutine@ trait to compile.
1028The 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@.
1029The \CFA keyword @coroutine@ implicitly implements the getter and forward declarations required for implementing the coroutine main:
1030\begin{cquote}
1031\begin{tabular}{@{}ccc@{}}
1032\begin{cfa}
1033coroutine MyCor {
1034        int value;
1035
1036};
1037\end{cfa}
1038&
1039{\Large $\Rightarrow$}
1040&
1041\begin{tabular}{@{}ccc@{}}
1042\begin{cfa}
1043struct MyCor {
1044        int value;
1045        coroutine_desc cor;
1046};
1047\end{cfa}
1048&
1049\begin{cfa}
1050static inline coroutine_desc *
1051get_coroutine( MyCor & this ) {
1052        return &this.cor;
1053}
1054\end{cfa}
1055&
1056\begin{cfa}
1057void main( MyCor * this );
1058
1059
1060
1061\end{cfa}
1062\end{tabular}
1063\end{tabular}
1064\end{cquote}
1065The 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.
1066
1067
1068\subsection{Thread Interface}
1069\label{threads}
1070
1071Both user and kernel threads are supported, where user threads provide concurrency and kernel threads provide parallelism.
1072Like 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:
1073\begin{cquote}
1074\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1075\begin{cfa}
1076thread myThread {
1077        // communication variables
1078};
1079
1080
1081\end{cfa}
1082&
1083\begin{cfa}
1084trait is_thread( `dtype` T ) {
1085      void main( T & );
1086      thread_desc * get_thread( T & );
1087      void ^?{}( T & `mutex` );
1088};
1089\end{cfa}
1090\end{tabular}
1091\end{cquote}
1092(The qualifier @mutex@ for the destructor parameter is discussed in Section~\ref{s:Monitors}.)
1093Like a coroutine, the statically-typed @main@ routine is the starting point (first stack frame) of a user thread.
1094The difference is that a coroutine borrows a thread from its caller, so the first thread resuming a coroutine creates an instance of @main@;
1095whereas, 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{
1096The \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.}
1097No 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.
1098
1099\begin{comment} % put in appendix with coroutine version ???
1100As such the @main@ routine of a thread can be defined as
1101\begin{cfa}
1102thread foo {};
1103
1104void main(foo & this) {
1105        sout | "Hello World!" | endl;
1106}
1107\end{cfa}
1108
1109In 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.
1110With 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.
1111\begin{cfa}
1112typedef void (*voidRtn)(int);
1113
1114thread RtnRunner {
1115        voidRtn func;
1116        int arg;
1117};
1118
1119void ?{}(RtnRunner & this, voidRtn inRtn, int arg) {
1120        this.func = inRtn;
1121        this.arg  = arg;
1122}
1123
1124void main(RtnRunner & this) {
1125        // thread starts here and runs the routine
1126        this.func( this.arg );
1127}
1128
1129void hello(/*unused*/ int) {
1130        sout | "Hello World!" | endl;
1131}
1132
1133int main() {
1134        RtnRunner f = {hello, 42};
1135        return 0?
1136}
1137\end{cfa}
1138A 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}.
1139\end{comment}
1140
1141For user threads to be useful, it must be possible to start and stop the underlying thread, and wait for it to complete execution.
1142While using an API such as @fork@ and @join@ is relatively common, such an interface is awkward and unnecessary.
1143A simple approach is to use allocation/deallocation principles, and have threads implicitly @fork@ after construction and @join@ before destruction.
1144\begin{cfa}
1145thread World {};
1146void main( World & this ) {
1147        sout | "World!" | endl;
1148}
1149int main() {
1150        World w`[10]`;                                                  $\C{// implicit forks after creation}$
1151        sout | "Hello " | endl;                                 $\C{// "Hello " and 10 "World!" printed concurrently}$
1152}                                                                                       $\C{// implicit joins before destruction}$
1153\end{cfa}
1154This semantics ensures a thread is started and stopped exactly once, eliminating some programming error, and scales to multiple threads for basic (termination) synchronization.
1155This 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.
1156\begin{cfa}
1157int main() {
1158        MyThread * heapLived;
1159        {
1160                MyThread blockLived;                            $\C{// fork block-based thread}$
1161                heapLived = `new`( MyThread );          $\C{// fork heap-based thread}$
1162                ...
1163        }                                                                               $\C{// join block-based thread}$
1164        ...
1165        `delete`( heapLived );                                  $\C{// join heap-based thread}$
1166}
1167\end{cfa}
1168The heap-based approach allows arbitrary thread-creation topologies, with respect to fork/join-style concurrency.
1169
1170Figure~\ref{s:ConcurrentMatrixSummation} shows concurrently adding the rows of a matrix and then totalling the subtotals sequential, after all the row threads have terminated.
1171The program uses heap-based threads because each thread needs different constructor values.
1172(Python provides a simple iteration mechanism to initialize array elements to different values allowing stack allocation.)
1173The allocation/deallocation pattern appears unusual because allocated objects are immediately deleted without any intervening code.
1174However, for threads, the deletion provides implicit synchronization, which is the intervening code.
1175While the subtotals are added in linear order rather than completion order, which slight 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.
1176
1177\begin{figure}
1178\begin{cfa}
1179thread Adder {
1180    int * row, cols, & subtotal;                        $\C{// communication}$
1181};
1182void ?{}( Adder & adder, int row[], int cols, int & subtotal ) {
1183    adder.[ row, cols, &subtotal ] = [ row, cols, &subtotal ];
1184}
1185void main( Adder & adder ) with( adder ) {
1186    subtotal = 0;
1187    for ( int c = 0; c < cols; c += 1 ) { subtotal += row[c]; }
1188}
1189int main() {
1190    const int rows = 10, cols = 1000;
1191    int matrix[rows][cols], subtotals[rows], total = 0;
1192    // read matrix
1193    Adder * adders[rows];
1194    for ( int r = 0; r < rows; r += 1 ) {       $\C{// start threads to sum rows}$
1195                adders[r] = new( matrix[r], cols, &subtotals[r] );
1196    }
1197    for ( int r = 0; r < rows; r += 1 ) {       $\C{// wait for threads to finish}$
1198                delete( adders[r] );                            $\C{// termination join}$
1199                total += subtotals[r];                          $\C{// total subtotal}$
1200    }
1201    sout | total | endl;
1202}
1203\end{cfa}
1204\caption{Concurrent Matrix Summation}
1205\label{s:ConcurrentMatrixSummation}
1206\end{figure}
1207
1208
1209\section{Mutual Exclusion / Synchronization}
1210
1211Uncontrolled non-deterministic execution is meaningless.
1212To 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}.
1213Since 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).
1214In these paradigms, interaction among concurrent objects is performed by stateless message-passing~\cite{Thoth,Harmony,V-Kernel} or other paradigms closely relate to networking concepts, \eg channels~\cite{CSP,Go}.
1215However, 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.
1216Hence, a programmer must learn and manipulate two sets of design patterns.
1217While this distinction can be hidden away in library code, effective use of the library still has to take both paradigms into account.
1218In contrast, approaches based on statefull models more closely resemble the standard call/return programming-model, resulting in a single programming paradigm.
1219
1220At the lowest level, concurrent control is implemented by atomic operations, upon which different kinds of locks mechanism are constructed, \eg semaphores~\cite{Dijkstra68b}, barriers, and path expressions~\cite{Campbell74}.
1221However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
1222A newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
1223While 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 was rejected as the core paradigm for concurrency in \CFA.
1224
1225One of the most natural, elegant, and efficient mechanisms for mutual exclusion and synchronization for shared-memory systems is the \emph{monitor}.
1226First 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}.
1227In 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.
1228For these reasons, \CFA selected monitors as the core high-level concurrency-construct, upon which higher-level approaches can be easily constructed.
1229
1230
1231\subsection{Mutual Exclusion}
1232
1233A group of instructions manipulating a specific instance of shared data that must be performed atomically is called an (individual) \newterm{critical-section}~\cite{Dijkstra65}.
1234The 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.
1235The 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.
1236\newterm{Mutual exclusion} enforces that the correct kind and number of threads are using a critical section.
1237
1238However, many solutions exist for mutual exclusion, which vary in terms of performance, flexibility and ease of use.
1239Methods 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.
1240Ease 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.
1241For 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.
1242However, a significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
1243Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
1244
1245
1246\subsection{Synchronization}
1247
1248Synchronization enforces relative ordering of execution, and synchronization tools provide numerous mechanisms to establish these timing relationships.
1249Low-level synchronization primitives offer good performance and flexibility at the cost of ease of use;
1250higher-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.
1251Often 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.
1252If a writer thread is scheduled for next access, but another reader thread acquires the critical section first, that reader has \newterm{barged}.
1253Barging 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).
1254Preventing or detecting barging is an involved challenge with low-level locks, which can be made much easier by higher-level constructs.
1255This challenge is often split into two different approaches: barging avoidance and barging prevention.
1256Algorithms that allow a barger, but divert it until later using current synchronization state (flags), are avoiding the barger;
1257algorithms that preclude a barger from entering during synchronization in the critical section prevent barging completely.
1258Techniques like baton-pass locks~\cite{Andrews89} between threads instead of unconditionally releasing locks is an example of barging prevention.
1259
1260
1261\section{Monitors}
1262\label{s:Monitors}
1263
1264A \textbf{monitor} is a set of routines that ensure mutual exclusion when accessing shared state.
1265More 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).
1266The strong association with the call/return paradigm eases programmability, readability and maintainability, at a slight cost in flexibility and efficiency.
1267
1268Note, 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.
1269Copying 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.
1270Copying 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.
1271As for coroutines/tasks, a non-copyable (@dtype@) trait is used to capture this requirement, so all locks/monitors must be passed by reference (pointer).
1272\begin{cfa}
1273trait is_monitor( `dtype` T ) {
1274        monitor_desc * get_monitor( T & );
1275        void ^?{}( T & mutex );
1276};
1277\end{cfa}
1278
1279
1280\subsection{Mutex Acquisition}
1281\label{s:MutexAcquisition}
1282
1283While correctness implicitly implies a monitor's mutual exclusion is acquired and released, there are implementation options about when and where the locking/unlocking occurs.
1284(Much of this discussion also applies to basic locks.)
1285For example, a monitor may need to be passed through multiple helper routines before it becomes necessary to acquire the monitor mutual-exclusion.
1286\begin{cfa}[morekeywords=nomutex]
1287monitor Aint { int cnt; };                                      $\C{// atomic integer counter}$
1288void ?{}( Aint & `nomutex` this ) with( this ) { cnt = 0; } $\C{// constructor}$
1289int ?=?( Aint & `mutex`$\(_{opt}\)$ lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions}$
1290void ?{}( int & this, Aint & `mutex`$\(_{opt}\)$ v ) { this = v.cnt; }
1291int ?=?( int & lhs, Aint & `mutex`$\(_{opt}\)$ rhs ) with( rhs ) { lhs = cnt; }
1292int ++?( Aint & `mutex`$\(_{opt}\)$ this ) with( this ) { return ++cnt; } $\C{// increment}$
1293\end{cfa}
1294The @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.
1295(While a constructor may publish its address into a global variable, doing so generates a race-condition.)
1296The conversion operators for initializing and assigning with a normal integer only need @mutex@, if reading/writing the implementation type is not atomic.
1297Finally, 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.
1298
1299The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics, which is similar to the \CC template @std::atomic@.
1300\begin{cfa}
1301Aint x, y, z;
1302++x; ++y; ++z;                                                          $\C{// safe increment by multiple threads}$
1303x = 2; y = 2; z = 2;                                            $\C{// conversions}$
1304int i = x, j = y, k = z;
1305i = x; j = y; k = z;
1306\end{cfa}
1307
1308For maximum usability, monitors have \newterm{multi-acquire} semantics allowing a thread to acquire it multiple times without deadlock.
1309For example, atomically printing the contents of a binary tree:
1310\begin{cfa}
1311monitor Tree {
1312        Tree * left, right;
1313        // value
1314};
1315void print( Tree & mutex tree ) {                       $\C{// prefix traversal}$
1316        // write value
1317        print( tree->left );                                    $\C{// multiply acquire monitor lock on each recursion}$
1318        print( tree->right );
1319}
1320\end{cfa}
1321
1322Mandatory monitor qualifiers have the benefit of being self-documented, but requiring both @mutex@ and \lstinline[morekeywords=nomutex]@nomutex@ for all monitor parameter is redundant.
1323Instead, one of qualifier semantics can be the default, and the other required.
1324For example, assume the safe @mutex@ option for a monitor parameter because assuming \lstinline[morekeywords=nomutex]@nomutex@ may cause subtle errors.
1325On the other hand, assuming \lstinline[morekeywords=nomutex]@nomutex@ is the \emph{normal} parameter behaviour, stating explicitly ``this parameter is not special''.
1326Providing a default qualifier implies knowing whether a parameter is a monitor.
1327Since \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.
1328For 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@.
1329
1330The next semantic decision is establishing which parameter \emph{types} may be qualified with @mutex@.
1331Given:
1332\begin{cfa}
1333monitor M { ... }
1334int f1( M & mutex m );
1335int f2( M * mutex m );
1336int f3( M * mutex m[] );
1337int f4( stack( M * ) & mutex m );
1338\end{cfa}
1339the issue is that some of these parameter types are composed of multiple objects.
1340For @f1@, there is only a single parameter object.
1341Adding indirection in @f2@ still identifies a single object.
1342However, the matrix in @f3@ introduces multiple objects.
1343While shown shortly, multiple acquisition is possible;
1344however array lengths are often unknown in C.
1345This issue is exacerbated in @f4@, where the data structure must be safely traversed to acquire all of its elements.
1346
1347To make the issue tractable, \CFA only acquires one monitor per parameter with at most one level of indirection.
1348However, the C type-system has an ambiguity with respects to arrays.
1349Is the argument for @f2@ a single object or an array of objects?
1350If it is an array, only the first element of the array is acquired, which seems unsafe;
1351hence, @mutex@ is disallowed for array parameters.
1352\begin{cfa}
1353int f1( M & mutex m );                                          $\C{// allowed: recommended case}$
1354int f2( M * mutex m );                                          $\C{// disallowed: could be an array}$
1355int f3( M mutex m[$\,$] );                                      $\C{// disallowed: array length unknown}$
1356int f4( M ** mutex m );                                         $\C{// disallowed: could be an array}$
1357int f5( M * mutex m[$\,$] );                            $\C{// disallowed: array length unknown}$
1358\end{cfa}
1359% Note, not all array routines have distinct types: @f2@ and @f3@ have the same type, as do @f4@ and @f5@.
1360% However, even if the code generation could tell the difference, the extra information is still not sufficient to extend meaningfully the monitor call semantic.
1361
1362For object-oriented monitors, calling a mutex member \emph{implicitly} acquires mutual exclusion of the receiver object, @`rec`.foo(...)@.
1363\CFA has no receiver, and hence, must use an explicit mechanism to specify which object has mutual exclusion acquired.
1364A positive consequence of this design decision is the ability to support multi-monitor routines.
1365\begin{cfa}
1366int f( M & mutex x, M & mutex y );              $\C{// multiple monitor parameter of any type}$
1367M m1, m2;
1368f( m1, m2 );
1369\end{cfa}
1370(While object-oriented monitors can be extended with a mutex qualifier for multiple-monitor members, no prior example of this feature could be found.)
1371In practice, writing multi-locking routines that do not deadlock is tricky.
1372Having language support for such a feature is therefore a significant asset for \CFA.
1373
1374The capability to acquire multiple locks before entering a critical section is called \newterm{bulk acquire}.
1375In the previous example, \CFA guarantees the order of acquisition is consistent across calls to different routines using the same monitors as arguments.
1376This consistent ordering means acquiring multiple monitors is safe from deadlock.
1377However, users can force the acquiring order.
1378For example, notice the use of @mutex@/\lstinline[morekeywords=nomutex]@nomutex@ and how this affects the acquiring order:
1379\begin{cfa}
1380void foo( M & mutex m1, M & mutex m2 );         $\C{// acquire m1 and m2}$
1381void bar( M & mutex m1, M & /* nomutex */ m2 ) { $\C{// acquire m1}$
1382        ... foo( m1, m2 ); ...                                  $\C{// acquire m2}$
1383}
1384void baz( M & /* nomutex */ m1, M & mutex m2 ) { $\C{// acquire m2}$
1385        ... foo( m1, m2 ); ...                                  $\C{// acquire m1}$
1386}
1387\end{cfa}
1388The multi-acquire semantics allows @bar@ or @baz@ to acquire a monitor lock and reacquire it in @foo@.
1389In the calls to @bar@ and @baz@, the monitors are acquired in opposite order.
1390
1391However, such use leads to lock acquiring order problems resulting in deadlock~\cite{Lister77}, where detecting it requires dynamically tracking of monitor calls, and dealing with it requires rollback semantics~\cite{Dice10}.
1392In \CFA, safety is guaranteed by using bulk acquire of all monitors to shared objects, whereas other monitor systems provide no aid.
1393While \CFA provides only a partial solution, the \CFA partial solution handles many useful cases.
1394\begin{cfa}
1395monitor Bank { ... };
1396void deposit( Bank & `mutex` b, int deposit );
1397void transfer( Bank & `mutex` mybank, Bank & `mutex` yourbank, int me2you) {
1398        deposit( mybank, `-`me2you );                   $\C{// debit}$
1399        deposit( yourbank, me2you );                    $\C{// credit}$
1400}
1401\end{cfa}
1402This example shows a trivial solution to the bank-account transfer problem.
1403Without multi- and bulk acquire, the solution to this problem requires careful engineering.
1404
1405
1406\subsection{\protect\lstinline|mutex| statement} \label{mutex-stmt}
1407
1408The monitor call-semantics associate all locking semantics to routines.
1409Like Java, \CFA offers an alternative @mutex@ statement to reduce refactoring and naming.
1410\begin{cquote}
1411\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1412\begin{cfa}
1413monitor M {};
1414void foo( M & mutex m1, M & mutex m2 ) {
1415        // critical section
1416}
1417void bar( M & m1, M & m2 ) {
1418        foo( m1, m2 );
1419}
1420\end{cfa}
1421&
1422\begin{cfa}
1423
1424void bar( M & m1, M & m2 ) {
1425        mutex( m1, m2 ) {       // remove refactoring and naming
1426                // critical section
1427        }
1428}
1429
1430\end{cfa}
1431\\
1432\multicolumn{1}{c}{\textbf{routine call}} & \multicolumn{1}{c}{\lstinline@mutex@ \textbf{statement}}
1433\end{tabular}
1434\end{cquote}
1435
1436
1437\section{Scheduling}
1438\label{s:Scheduling}
1439
1440While monitor mutual-exclusion provides safe access to shared data, the monitor data may indicate that a thread accessing it cannot proceed.
1441For example, Figure~\ref{f:GenericBoundedBuffer} shows a bounded buffer that may be full/empty so produce/consumer threads must block.
1442Leaving the monitor and trying again (busy waiting) is impractical for high-level programming.
1443Monitors eliminate busy waiting by providing internal synchronization to schedule threads needing access to the shared data, where the synchronization is blocking (threads are parked) versus spinning.
1444Synchronization 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.
1445\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.
1446
1447Figure~\ref{f:BBInt} shows a \CFA 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@.
1448The @wait@ routine atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the routine's parameter list.
1449The appropriate condition lock is signalled to unblock an opposite kind of thread after an element is inserted/removed from the buffer.
1450Signalling is unconditional, because signalling an empty condition lock does nothing.
1451
1452Signalling semantics cannot have the signaller and signalled thread in the monitor simultaneously, which means:
1453\begin{enumerate}
1454\item
1455The signalling thread returns immediately, and the signalled thread continues.
1456\item
1457The signalling thread continues and the signalled thread is marked for urgent unblocking at the next scheduling point (exit/wait).
1458\item
1459The signalling thread blocks but is marked for urgrent unblocking at the next scheduling point and the signalled thread continues.
1460\end{enumerate}
1461The first approach is too restrictive, as it precludes solving a reasonable class of problems, \eg dating service.
1462\CFA supports the next two semantics as both are useful.
1463Finally, while it is common to store a @condition@ as a field of the monitor, in \CFA, a @condition@ variable can be created/stored independently.
1464Furthermore, 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.
1465
1466\begin{figure}
1467\centering
1468\newbox\myboxA
1469\begin{lrbox}{\myboxA}
1470\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1471forall( otype T ) { // distribute forall
1472        monitor Buffer {
1473                `condition` full, empty;
1474                int front, back, count;
1475                T elements[10];
1476        };
1477        void ?{}( Buffer(T) & buffer ) with(buffer) {
1478                [front, back, count] = 0;
1479        }
1480
1481        void insert( Buffer(T) & mutex buffer, T elem )
1482                                with(buffer) {
1483                if ( count == 10 ) `wait( empty )`;
1484                // insert elem into buffer
1485                `signal( full )`;
1486        }
1487        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1488                if ( count == 0 ) `wait( full )`;
1489                // remove elem from buffer
1490                `signal( empty )`;
1491                return elem;
1492        }
1493}
1494\end{cfa}
1495\end{lrbox}
1496
1497\newbox\myboxB
1498\begin{lrbox}{\myboxB}
1499\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1500forall( otype T ) { // distribute forall
1501        monitor Buffer {
1502
1503                int front, back, count;
1504                T elements[10];
1505        };
1506        void ?{}( Buffer(T) & buffer ) with(buffer) {
1507                [front, back, count] = 0;
1508        }
1509        T remove( Buffer(T) & mutex buffer ); // forward
1510        void insert( Buffer(T) & mutex buffer, T elem )
1511                                with(buffer) {
1512                if ( count == 10 ) `waitfor( remove, buffer )`;
1513                // insert elem into buffer
1514
1515        }
1516        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1517                if ( count == 0 ) `waitfor( insert, buffer )`;
1518                // remove elem from buffer
1519
1520                return elem;
1521        }
1522}
1523\end{cfa}
1524\end{lrbox}
1525
1526\subfloat[Internal Scheduling]{\label{f:BBInt}\usebox\myboxA}
1527%\qquad
1528\subfloat[External Scheduling]{\label{f:BBExt}\usebox\myboxB}
1529\caption{Generic Bounded-Buffer}
1530\label{f:GenericBoundedBuffer}
1531\end{figure}
1532
1533Figure~\ref{f:BBExt} shows a \CFA bounded-buffer with external scheduling, where producers/consumers detecting a full/empty buffer block and prevent more producers/consumers from entering the monitor until the buffer has a free/empty slot.
1534External 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.
1535If 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.
1536Threads making calls to routines that are currently excluded block outside (external) of the monitor on a calling queue, versus blocking on condition queues inside (internal) of the monitor.
1537% External scheduling is more constrained and explicit, which helps programmers reduce the non-deterministic nature of concurrency.
1538External scheduling allows users to wait for events from other threads without concern of unrelated events occurring.
1539The mechnaism can be done in terms of control flow, \eg Ada @accept@ or \uC @_Accept@, or in terms of data, \eg Go channels.
1540While both mechanisms have strengths and weaknesses, this project uses a control-flow mechanism to stay consistent with other language semantics.
1541Two 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}).
1542
1543For internal scheduling, non-blocking signalling (as in the producer/consumer example) is used when the signaller is providing the cooperation for a waiting thread;
1544the 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.
1545The waiter unblocks next, uses/takes the state, and exits the monitor.
1546Blocking signalling is the reverse, where the waiter is providing the cooperation for the signalling thread;
1547the 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.
1548The waiter changes state and exits the monitor, and the signaller unblocks next from the urgent queue to use/take the state.
1549
1550Figure~\ref{f:DatingService} shows a dating service demonstrating the two forms of signalling: non-blocking and blocking.
1551The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
1552A thread blocks until an appropriate partner arrives.
1553The complexity is exchanging phone number in the monitor because the monitor mutual-exclusion property prevents exchanging numbers.
1554For internal scheduling, the @exchange@ condition is necessary to block the thread finding the match, while the matcher unblocks to take the oppose number, post its phone number, and unblock the partner.
1555For external scheduling, the implicit urgent-condition replaces the explict @exchange@-condition and @signal_block@ puts the finding thread on the urgent condition and unblocks the matcher..
1556
1557The 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;
1558as well, an arriving thread may not find a partner and must wait, which requires a condition variable, and condition variables imply internal scheduling.
1559
1560\begin{figure}
1561\centering
1562\newbox\myboxA
1563\begin{lrbox}{\myboxA}
1564\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1565enum { CCodes = 20 };
1566monitor DS {
1567        int GirlPhNo, BoyPhNo;
1568        condition Girls[CCodes], Boys[CCodes];
1569        condition exchange;
1570};
1571int girl( DS & mutex ds, int phNo, int ccode ) {
1572        if ( is_empty( Boys[ccode] ) ) {
1573                wait( Girls[ccode] );
1574                GirlPhNo = phNo;
1575                `exchange.signal();`
1576        } else {
1577                GirlPhNo = phNo;
1578                `signal( Boys[ccode] );`
1579                `exchange.wait();`
1580        } // if
1581        return BoyPhNo;
1582}
1583int boy( DS & mutex ds, int phNo, int ccode ) {
1584        // as above with boy/girl interchanged
1585}
1586\end{cfa}
1587\end{lrbox}
1588
1589\newbox\myboxB
1590\begin{lrbox}{\myboxB}
1591\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1592
1593monitor DS {
1594        int GirlPhNo, BoyPhNo;
1595        condition Girls[CCodes], Boys[CCodes];
1596
1597};
1598int girl( DS & mutex ds, int phNo, int ccode ) {
1599        if ( is_empty( Boys[ccode] ) ) { // no compatible
1600                wait( Girls[ccode] ); // wait for boy
1601                GirlPhNo = phNo; // make phone number available
1602
1603        } else {
1604                GirlPhNo = phNo; // make phone number available
1605                `signal_block( Boys[ccode] );` // restart boy
1606
1607        } // if
1608        return BoyPhNo;
1609}
1610int boy( DS & mutex ds, int phNo, int ccode ) {
1611        // as above with boy/girl interchanged
1612}
1613\end{cfa}
1614\end{lrbox}
1615
1616\subfloat[\lstinline@signal@]{\label{f:DatingSignal}\usebox\myboxA}
1617\qquad
1618\subfloat[\lstinline@signal_block@]{\label{f:DatingSignalBlock}\usebox\myboxB}
1619\caption{Dating service. }
1620\label{f:DatingService}
1621\end{figure}
1622
1623Both internal and external scheduling extend to multiple monitors in a natural way.
1624\begin{cquote}
1625\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1626\begin{cfa}
1627monitor M { `condition e`; ... };
1628void foo( M & mutex m1, M & mutex m2 ) {
1629        ... wait( `e` ); ...   // wait( e, m1, m2 )
1630        ... wait( `e, m1` ); ...
1631        ... wait( `e, m2` ); ...
1632}
1633\end{cfa}
1634&
1635\begin{cfa}
1636void rtn$\(_1\)$( M & mutex m1, M & mutex m2 );
1637void rtn$\(_2\)$( M & mutex m1 );
1638void bar( M & mutex m1, M & mutex m2 ) {
1639        ... waitfor( `rtn` ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\), m1, m2 )}$
1640        ... waitfor( `rtn, m1` ); ... // $\LstCommentStyle{waitfor( rtn\(_2\), m1 )}$
1641}
1642\end{cfa}
1643\end{tabular}
1644\end{cquote}
1645For @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 )@.
1646To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @wait( e, m1 )@.
1647Wait statically verifies the released monitors are the acquired mutex-parameters so unconditional release is safe.
1648Finally, a signaller,
1649\begin{cfa}
1650void baz( M & mutex m1, M & mutex m2 ) {
1651        ... signal( e ); ...
1652}
1653\end{cfa}
1654must have acquired monitor locks that are greater than or equal to the number of locks for the waiting thread signalled from the condition queue.
1655
1656Similarly, 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 )@.
1657To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn, m1 )@.
1658Waitfor statically verifies the released monitors are the same as the acquired mutex-parameters of the given routine or routine pointer.
1659To statically verify the released monitors match with the accepted routine's mutex parameters, the routine (pointer) prototype must be accessible.
1660
1661When an overloaded routine appears in an @waitfor@ statement, calls to any routine with that name are accepted.
1662The rationale is that members with the same name should perform a similar function, and therefore, all should be eligible to accept a call.
1663As always, overloaded routines can be disambiguated using a cast:
1664\begin{cfa}
1665void rtn( M & mutex m );
1666`int` rtn( M & mutex m );
1667waitfor( (`int` (*)( M & mutex ))rtn, m1, m2 );
1668\end{cfa}
1669
1670Given the ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock.
1671\begin{cfa}
1672void foo( M & mutex m1, M & mutex m2 ) {
1673        ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired )}$
1674void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2 )}$
1675        ... signal( `e` ); ...
1676\end{cfa}
1677The @wait@ only releases @m1@ so the signalling thread cannot acquire both @m1@ and @m2@ to  enter @bar@ to get to the @signal@.
1678While deadlock issues can occur with multiple/nesting acquisition, this issue results from the fact that locks, and by extension monitors, are not perfectly composable.
1679
1680Finally, an important aspect of monitor implementation is barging, \ie can calling threads barge ahead of signalled threads?
1681If barging is allowed, synchronization between a singller and signallee is difficult, often requiring multiple unblock/block cycles (looping around a wait rechecking if a condition is met).
1682\begin{quote}
1683However, 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.
1684It 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}
1685\end{quote}
1686\CFA scheduling \emph{precludes} barging, which simplifies synchronization among threads in the monitor and increases correctness.
1687For example, there are no loops in either bounded buffer solution in Figure~\ref{f:GenericBoundedBuffer}.
1688Supporting 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.
1689
1690
1691\subsection{Barging Prevention}
1692
1693Figure~\ref{f:BargingPrevention} shows \CFA code where bulk acquire adds complexity to the internal-signalling semantics.
1694The complexity begins at the end of the inner @mutex@ statement, where the semantics of internal scheduling need to be extended for multiple monitors.
1695The problem is that bulk acquire is used in the inner @mutex@ statement where one of the monitors is already acquired.
1696When 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.
1697However, both the signalling and waiting thread W1 still need monitor @m1@.
1698
1699\begin{figure}
1700\newbox\myboxA
1701\begin{lrbox}{\myboxA}
1702\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1703monitor M m1, m2;
1704condition c;
1705mutex( m1 ) { // $\LstCommentStyle{\color{red}outer}$
1706        ...
1707        mutex( m1, m2 ) { // $\LstCommentStyle{\color{red}inner}$
1708                ... `signal( c )`; ...
1709                // m1, m2 acquired
1710        } // $\LstCommentStyle{\color{red}release m2}$
1711        // m1 acquired
1712} // release m1
1713\end{cfa}
1714\end{lrbox}
1715
1716\newbox\myboxB
1717\begin{lrbox}{\myboxB}
1718\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1719
1720
1721mutex( m1 ) {
1722        ...
1723        mutex( m1, m2 ) {
1724                ... `wait( c )`; // block and release m1, m2
1725                // m1, m2 acquired
1726        } // $\LstCommentStyle{\color{red}release m2}$
1727        // m1 acquired
1728} // release m1
1729\end{cfa}
1730\end{lrbox}
1731
1732\newbox\myboxC
1733\begin{lrbox}{\myboxC}
1734\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1735
1736
1737mutex( m2 ) {
1738        ... `wait( c )`; ...
1739        // m2 acquired
1740} // $\LstCommentStyle{\color{red}release m2}$
1741
1742
1743
1744
1745\end{cfa}
1746\end{lrbox}
1747
1748\begin{cquote}
1749\subfloat[Signalling Thread]{\label{f:SignallingThread}\usebox\myboxA}
1750\hspace{2\parindentlnth}
1751\subfloat[Waiting Thread (W1)]{\label{f:WaitingThread}\usebox\myboxB}
1752\hspace{2\parindentlnth}
1753\subfloat[Waiting Thread (W2)]{\label{f:OtherWaitingThread}\usebox\myboxC}
1754\end{cquote}
1755\caption{Barging Prevention}
1756\label{f:BargingPrevention}
1757\end{figure}
1758
1759One 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.
1760However, Figure~\ref{f:OtherWaitingThread} shows this solution is complex depending on other waiters, resulting is choices when the signaller finishes the inner mutex-statement.
1761The singaller 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@.
1762In 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.
1763Furthermore, there is an execution sequence where the signaller always finds waiter W2, and hence, waiter W1 starves.
1764
1765While a number of approaches were examined~\cite[\S~4.3]{Delisle18}, the solution chosen for \CFA is a novel techique called \newterm{partial signalling}.
1766Signalled threads are moved to an urgent queue and the waiter at the front defines the set of monitors necessary for it to unblock.
1767Partial signalling transfers ownership of monitors to the front waiter.
1768When the signaller thread exits or waits in the monitor the front waiter is unblocked if all its monitors are released.
1769This solution has the benefit that complexity is encapsulated into only two actions: passing monitors to the next owner when they should be released and conditionally waking threads if all conditions are met.
1770
1771
1772\subsection{Loose Object Definitions}
1773\label{s:LooseObjectDefinitions}
1774
1775In an object-oriented programming-language, a class includes an exhaustive list of operations.
1776However, new members can be added via static inheritance or dynaic members, \eg JavaScript~\cite{JavaScript}.
1777Similarly, monitor routines can be added at any time in \CFA, making it less clear for programmers and more difficult to implement.
1778\begin{cfa}
1779monitor M {};
1780void `f`( M & mutex m );
1781void g( M & mutex m ) { waitfor( `f` ); }       $\C{// clear which f}$
1782void `f`( M & mutex m, int );                           $\C{// different f}$
1783void h( M & mutex m ) { waitfor( `f` ); }       $\C{// unclear which f}$
1784\end{cfa}
1785Hence, the cfa-code for the entering a monitor looks like:
1786\begin{cfa}
1787if ( $\textrm{\textit{monitor is free}}$ ) $\LstCommentStyle{// \color{red}enter}$
1788else if ( $\textrm{\textit{already own monitor}}$ ) $\LstCommentStyle{// \color{red}continue}$
1789else if ( $\textrm{\textit{monitor accepts me}}$ ) $\LstCommentStyle{// \color{red}enter}$
1790else $\LstCommentStyle{// \color{red}block}$
1791\end{cfa}
1792For the first two conditions, it is easy to implement a check that can evaluate the condition in a few instructions.
1793However, a fast check for \emph{monitor accepts me} is much harder to implement depending on the constraints put on the monitors.
1794Figure~\ref{fig:ClassicalMonitor} shows monitors are often expressed as an entry (calling) queue, some acceptor queues, and an urgent stack/queue.
1795
1796\begin{figure}
1797\centering
1798\subfloat[Classical monitor] {
1799\label{fig:ClassicalMonitor}
1800{\resizebox{0.45\textwidth}{!}{\input{monitor.pstex_t}}}
1801}% subfloat
1802\quad
1803\subfloat[Bulk acquire monitor] {
1804\label{fig:BulkMonitor}
1805{\resizebox{0.45\textwidth}{!}{\input{ext_monitor.pstex_t}}}
1806}% subfloat
1807\caption{Monitor Implementation}
1808\label{f:MonitorImplementation}
1809\end{figure}
1810
1811For 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.
1812This approach requires a unique dense ordering of routines with a small upper-bound and the ordering must be consistent across translation units.
1813For 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.
1814
1815Figure~\ref{fig:BulkMonitor} shows the \CFA monitor implementation.
1816The mutex routine called is associated with each thread on the entry queue, while a list of acceptable routines is kept separately.
1817The 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 linear search.
1818
1819
1820\subsection{Multi-Monitor Scheduling}
1821\label{s:Multi-MonitorScheduling}
1822
1823External scheduling, like internal scheduling, becomes significantly more complex when introducing multi-monitor syntax.
1824Even in the simplest possible case, new semantics needs to be established:
1825\begin{cfa}
1826monitor M {};
1827void f( M & mutex m1 );
1828void g( M & mutex m1, M & mutex m2 ) {
1829        waitfor( f );                                                   $\C{\color{red}// pass m1 or m2 to f?}$
1830}
1831\end{cfa}
1832The solution is for the programmer to disambiguate:
1833\begin{cfa}
1834        waitfor( f, m2 );                                               $\C{\color{red}// wait for call to f with argument m2}$
1835\end{cfa}
1836Routine @g@ has acquired both locks, so when routine @f@ is called, the lock for monitor @m2@ is passed from @g@ to @f@, while @g@ still holds lock @m1@.
1837This behaviour can be extended to the multi-monitor @waitfor@ statement.
1838\begin{cfa}
1839monitor M {};
1840void f( M & mutex m1, M & mutex m2 );
1841void g( M & mutex m1, M & mutex m2 ) {
1842        waitfor( f, m1, m2 );                                   $\C{\color{red}// wait for call to f with arguments m1 and m2}$
1843}
1844\end{cfa}
1845Again, the set of monitors passed to the @waitfor@ statement must be entirely contained in the set of monitors already acquired by accepting routine.
1846
1847Note, for internal and external scheduling with multiple monitors, a signalling or accepting thread must match exactly, \ie partial matching results in waiting.
1848\begin{cquote}
1849\lstDeleteShortInline@%
1850\begin{tabular}{@{}l@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
1851\begin{cfa}
1852monitor M1 {} m11, m12;
1853monitor M2 {} m2;
1854condition c;
1855void f( M1 & mutex m1, M2 & mutex m2 ) {
1856        signal( c );
1857}
1858void g( M1 & mutex m1, M2 & mutex m2 ) {
1859        wait( c );
1860}
1861g( `m11`, m2 ); // block on accept
1862f( `m12`, m2 ); // cannot fulfil
1863\end{cfa}
1864&
1865\begin{cfa}
1866monitor M1 {} m11, m12;
1867monitor M2 {} m2;
1868
1869void f( M1 & mutex m1, M2 & mutex m2 ) {
1870
1871}
1872void g( M1 & mutex m1, M2 & mutex m2 ) {
1873        waitfor( f, m1, m2 );
1874}
1875g( `m11`, m2 ); // block on accept
1876f( `m12`, m2 ); // cannot fulfil
1877\end{cfa}
1878\end{tabular}
1879\lstMakeShortInline@%
1880\end{cquote}
1881
1882
1883\subsection{Extended \protect\lstinline@waitfor@}
1884
1885The extended form of the @waitfor@ statement conditionally accepts one of a group of mutex routines and allows a specific action to be performed \emph{after} the mutex routine finishes.
1886\begin{cfa}
1887`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1888        waitfor( $\emph{mutex-member-name}$ )
1889                $\emph{statement}$                                      $\C{// action after call}$
1890`or` `when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
1891        waitfor( $\emph{mutex-member-name}$ )
1892                $\emph{statement}$                                      $\C{// action after call}$
1893`or`    ...                                                                     $\C{// list of waitfor clauses}$
1894`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1895        `timeout`                                                               $\C{// optional terminating timeout clause}$
1896                $\emph{statement}$                                      $\C{// action after timeout}$
1897`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1898        `else`                                                                  $\C{// optional terminating clause}$
1899                $\emph{statement}$                                      $\C{// action when no immediate calls}$
1900\end{cfa}
1901For a @waitfor@ clause to be executed, its @when@ must be true and an outstanding call to its corresponding member(s) must exist.
1902The \emph{conditional-expression} of a @when@ may call a routine, but the routine must not block or context switch.
1903If there are several mutex calls that can be accepted, selection occurs top-to-bottom in the @waitfor@ clauses versus non-deterministically.
1904If 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.
1905If 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.
1906Hence, the terminating @else@ clause allows a conditional attempt to accept a call without blocking.
1907If there is a @timeout@ clause, it provides an upper bound on waiting, and can only appear with a conditional @else@, otherwise the timeout cannot be triggered.
1908In all cases, the statement following is executed \emph{after} a clause is executed to know which of the clauses executed.
1909
1910A group of conditional @waitfor@ clauses is \emph{not} the same as a group of @if@ statements, e.g.:
1911\begin{cfa}
1912if ( C1 ) waitfor( mem1 );                       when ( C1 ) waitfor( mem1 );
1913else if ( C2 ) waitfor( mem2 );         or when ( C2 ) waitfor( mem2 );
1914\end{cfa}
1915The left example accepts only @mem1@ if @C1@ is true or only @mem2@ if @C2@ is true.
1916The right example accepts either @mem1@ or @mem2@ if @C1@ and @C2@ are true.
1917
1918An interesting use of @waitfor@ is accepting the @mutex@ destructor to know when an object deallocated.
1919\begin{cfa}
1920void insert( Buffer(T) & mutex buffer, T elem ) with( buffer ) {
1921        if ( count == BufferSize )
1922                waitfor( remove, buffer ) {
1923                        elements[back] = elem;
1924                        back = ( back + 1 ) % BufferSize;
1925                        count += 1;
1926                } or `waitfor( ^?{}, buffer )` throw insertFail;
1927}
1928\end{cfa}
1929However, the @waitfor@ semantics do not work, since using an object after its destructor is called is undefined.
1930Therefore, 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 deallocates the object.
1931Accepting the destructor is an idiomatic way to terminate a thread in \CFA.
1932
1933
1934\subsection{\protect\lstinline@mutex@ Threads}
1935
1936Threads in \CFA are monitors, so all monitor features are available when using threads.
1937Figure~\ref{f:pingpong} shows an example of two threads calling and accepting calls from each other in a cycle.
1938Note, both ping/pong threads are globally declared, @pi@/@po@, and hence, start (and possibly complete) before the program starts.
1939
1940\begin{figure}
1941\lstDeleteShortInline@%
1942\begin{cquote}
1943\begin{cfa}
1944thread Ping {} pi;
1945thread Pong {} po;
1946void ping( Ping & mutex ) {}
1947void pong( Pong & mutex ) {}
1948int main() {}
1949\end{cfa}
1950\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1951\begin{cfa}
1952void main( Ping & pi ) {
1953        for ( int i = 0; i < 10; i += 1 ) {
1954                `waitfor( ping, pi );`
1955                `pong( po );`
1956        }
1957}
1958\end{cfa}
1959&
1960\begin{cfa}
1961void main( Pong & po ) {
1962        for ( int i = 0; i < 10; i += 1 ) {
1963                `ping( pi );`
1964                `waitfor( pong, po );`
1965        }
1966}
1967\end{cfa}
1968\end{tabular}
1969\lstMakeShortInline@%
1970\end{cquote}
1971\caption{Threads ping/pong using external scheduling}
1972\label{f:pingpong}
1973\end{figure}
1974
1975\section{Parallelism}
1976
1977Historically, computer performance was about processor speeds.
1978However, with heat dissipation being a direct consequence of speed increase, parallelism has become the new source for increased performance~\cite{Sutter05, Sutter05b}.
1979Now, high-performance applications must care about parallelism, which requires concurrency.
1980The lowest-level approach of parallelism is to use \newterm{kernel threads} in combination with semantics like @fork@, @join@, \etc.
1981However, kernel threads are better as an implementation tool because of complexity and high cost.
1982Therefore, different abstractions are layered onto kernel threads to simplify them.
1983
1984
1985\subsection{User Threads with Preemption}
1986
1987A direct improvement on kernel threads is user threads, \eg Erlang~\cite{Erlang} and \uC~\cite{uC++book}.
1988This approach provides an interface that matches the language paradigms, more control over concurrency in the language runtime, and an abstract (and portable) interface to the underlying kernel threads across operating systems.
1989In many cases, user threads can be used on a much larger scale (100,000 threads).
1990Like kernel threads, user threads support preemption, which maximizes nondeterminism, but introduces concurrency errors: race, livelock, starvation, and deadlock.
1991\CFA adopts user-threads as they represent the truest realization of concurrency and can build the following approaches and more, \eg actors~\cite{Actors}.
1992
1993
1994\subsection{User Threads without Preemption (Fiber)}
1995\label{s:fibers}
1996
1997A variant of user thread is \newterm{fibers}, which removes preemption, \eg Go~\cite{Go}.
1998Like functional programming, which removes mutation and its associated problems, removing preemption from concurrency reduces nondeterminism, hence race and deadlock errors are more difficult to generate.
1999However, preemption is necessary for concurrency that relies on spinning, so there are a class of problems that cannot be programmed without preemption.
2000
2001
2002\subsection{Thread Pools}
2003
2004In contrast to direct threading is indirect \newterm{thread pools}, where small jobs (work units) are insert into a work pool for execution.
2005If the jobs are dependent, \ie interact, there is an implicit/explicit dependency graph that ties them together.
2006While removing direct concurrency, and hence the amount of context switching, thread pools significantly limit the interaction that can occur among jobs.
2007Indeed, jobs should not block because that also block the underlying thread, which effectively means the CPU utilization, and therefore throughput, suffers.
2008While 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.
2009As well, concurrency errors return, which threads pools are suppose to mitigate.
2010The gold standard for thread pool is Intel's TBB library~\cite{TBB}.
2011
2012
2013\section{\protect\CFA Runtime Structure}
2014
2015Figure~\ref{f:RunTimeStructure} illustrates the runtime structure of a \CFA program.
2016In addition to the new kinds of objects introduced by \CFA, there are two more runtime entities used to control parallel execution.
2017An executing thread is illustrated by its containment in a processor.
2018
2019\begin{figure}
2020\centering
2021\input{RunTimeStructure}
2022\caption{\CFA Runtime Structure}
2023\label{f:RunTimeStructure}
2024\end{figure}
2025
2026
2027\subsection{Cluster}
2028\label{s:RuntimeStructureCluster}
2029
2030A \newterm{cluster} is a collection of threads and virtual processors (abstraction a kernel thread) that execute the threads (like a virtual machine).
2031The purpose of a cluster is to control the amount of parallelism that is possible among threads, plus scheduling and other execution defaults.
2032The default cluster-scheduler is single-queue multi-server, which provides automatic load-balancing of threads on processors.
2033However, the scheduler is pluggable, supporting alternative schedulers.
2034If several clusters exist, both threads and virtual processors, can be explicitly migrated from one cluster to another.
2035No automatic load balancing among clusters is performed by \CFA.
2036
2037When a \CFA program begins execution, it creates two clusters: system and user.
2038The system cluster contains a processor that does not execute user threads.
2039Instead, the system cluster handles system-related operations, such as catching errors that occur on the user clusters, printing appropriate error information, and shutting down \CFA.
2040A user cluster is created to contain the user threads.
2041Having all threads execute on the one cluster often maximizes utilization of processors, which minimizes runtime.
2042However, because of limitations of the underlying operating system, special hardware, or scheduling requirements (real-time), it is sometimes necessary to have multiple clusters.
2043
2044
2045\subsection{Virtual Processor}
2046\label{s:RuntimeStructureProcessor}
2047
2048A 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.
2049Programs may use more virtual processors than hardware processors.
2050On a multiprocessor, kernel threads are distributed across the hardware processors resulting in virtual processors executing in parallel.
2051(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 processor.)
2052The \CFA runtime attempts to block unused processors and unblock processors as the system load increases;
2053balancing the workload with processors is difficult.
2054Preemption occurs on virtual processors rather than user threads, via operating-system interrupts.
2055Thus virtual processors execute user threads, where preemption frequency applies to a virtual processor, so preemption occurs randomly across the executed user threads.
2056Turning off preemption transforms user threads into fibers.
2057
2058
2059\subsection{Debug Kernel}
2060
2061There are two versions of the \CFA runtime kernel: debug and non-debug.
2062The 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.
2063After a program is debugged, the non-debugging version can be used to decrease space and increase performance.
2064
2065
2066\section{Implementation}
2067
2068Currently, \CFA has fixed-sized stacks, where the stack size can be set at coroutine/thread creation but with no subsequent growth.
2069Schemes exist for dynamic stack-growth, such as stack copying and chained stacks.
2070However, stack copying requires pointer adjustment to items on the stack, which is impossible without some form of garage collection.
2071As well, chained stacks require all modules be recompiled to use this feature, which breaks backward compatibility with existing C libraries.
2072In the long term, it is likely C libraries will migrate to stack chaining to support concurrency, at only a minimal cost to sequential programs.
2073Nevertheless, experience teaching \uC~\cite{CS343} shows fixed-sized stacks are rarely an issue in the most concurrent programs.
2074
2075A primary implementation challenge is avoiding contention from dynamically allocating memory because of bulk acquire, \eg the internal-scheduling design is (almost) free of allocations.
2076All 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.
2077Furthermore, several bulk-acquire operations need a variable amount of memory.
2078This 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.
2079
2080In \CFA, ordering of monitor acquisition relies on memory ordering to prevent deadlock~\cite{Havender68}, because all objects are guaranteed to have distinct non-overlapping memory layouts, and mutual-exclusion for a monitor is only defined for its lifetime.
2081When a mutex call is made, pointers to the concerned monitors are aggregated into a variable-length array and sorted.
2082This array persists for the entire duration of the mutual-exclusion and its ordering reused extensively.
2083
2084To improve performance and simplicity, context switching occur inside a routine call, so only callee-saved registers are copied onto the stack and then the stack register is switched;
2085the corresponding registers are then restored for the other context.
2086Note, the instruction pointer is untouched since the context switch is always inside the same routine.
2087Unlike coroutines, threads do not context switch among each other;
2088they context switch to the cluster scheduler.
2089This method is a 2-step context-switch and provides a clear distinction between user and kernel code, where scheduling and other system operations happen.
2090The 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.
2091Experimental results (not shown) show the performance difference between these two approaches is virtually equivalent, because the 1-step performance is dominated by locking instructions to prevent a race condition.
2092
2093All kernel threads (@pthreads@) created a stack.
2094Each \CFA virtual processor is implemented as a coroutine and these coroutines run directly on the kernel-thread stack, effectively stealing this stack.
2095The exception to this rule is the program main, \ie the initial kernel thread that is given to any program.
2096In 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.
2097
2098Finally, 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.
2099Because preemption frequency is usually long, 1 millisecond, performance cost is negligible.
2100
2101Preemption is normally handled by setting a count-down timer on each virtual processor.
2102When 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.
2103Multiple signal handlers may be pending.
2104When 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 was delivered.
2105The 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;
2106therefore, all virtual processors in a cluster need to have the same signal mask.
2107
2108However, on UNIX systems:
2109\begin{quote}
2110A process-directed signal may be delivered to any one of the threads that does not currently have the signal blocked.
2111If more than one of the threads has the signal unblocked, then the kernel chooses an arbitrary thread to which to deliver the signal.
2112SIGNAL(7) - Linux Programmer's Manual
2113\end{quote}
2114Hence, 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).
2115To ensure each virtual processor receives its own preemption signals, a discrete-event simulation is run on one virtual processor, and only it sets timer events.
2116Virtual processors register an expiration time with the discrete-event simulator, which is inserted in sorted order.
2117The 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.
2118Processing a preemption event sends an \emph{internal} @SIGUSR1@ signal to the registered virtual processor, which is always delivered to that processor.
2119
2120
2121\section{Performance}
2122\label{results}
2123
2124To verify the implementation of the \CFA runtime, a series of microbenchmarks are performed comparing \CFA with other widely used programming languages with concurrency.
2125Table~\ref{t:machine} shows the specifications of the computer used to run the benchmarks, and the versions of the software used in the comparison.
2126
2127\begin{table}[h]
2128\centering
2129\caption{Experiment environment}
2130\label{t:machine}
2131
2132\begin{tabular}{|l|r||l|r|}
2133\hline
2134Architecture            & x86\_64                               & NUMA node(s)  & 8 \\
2135\hline
2136CPU op-mode(s)          & 32-bit, 64-bit                & Model name    & AMD Opteron\texttrademark\ Processor 6380 \\
2137\hline
2138Byte Order                      & Little Endian                 & CPU Freq              & 2.5 GHz \\
2139\hline
2140CPU(s)                          & 64                                    & L1d cache     & 16 KiB \\
2141\hline
2142Thread(s) per core      & 2                                     & L1i cache     & 64 KiB \\
2143\hline
2144Core(s) per socket      & 8                                     & L2 cache              & 2048 KiB \\
2145\hline
2146Socket(s)                       & 4                                     & L3 cache              & 6144 KiB \\
2147\hline
2148\hline
2149Operating system        & Ubuntu 16.04.3 LTS    & Kernel                & Linux 4.4-97-generic \\
2150\hline
2151gcc                                     & 6.3                                   & \CFA                  & 1.0.0 \\
2152\hline
2153Java                            & OpenJDK-9                     & Go                    & 1.9.2 \\
2154\hline
2155\end{tabular}
2156\end{table}
2157
2158All benchmarks are run using the following harness:
2159\begin{cfa}
2160unsigned int N = 10_000_000;
2161#define BENCH( run ) Time before = getTimeNsec(); run; Duration result = (getTimeNsec() - before) / N;
2162\end{cfa}
2163The method used to get time is @clock_gettime( CLOCK_REALTIME )@.
2164Each benchmark is performed @N@ times, where @N@ varies depending on the benchmark, the total time is divided by @N@ to obtain the average time for a benchmark.
2165
2166
2167\paragraph{Context-Switching}
2168
2169In procedural programming, the cost of a routine call is important as modularization (refactoring) increases.
2170(In many cases, a compiler inlines routine calls to eliminate this cost.)
2171Similarly, when modularization extends to coroutines/tasks, the time for a context switch becomes a relevant factor.
2172The coroutine context-switch is 2-step using resume/suspend, \ie from resumer to suspender and from suspender to resumer.
2173The thread context switch is 2-step using yield, \ie enter and return from the runtime kernel.
2174Figure~\ref{f:ctx-switch} shows the code for coroutines/threads with all results in Table~\ref{tab:ctx-switch}.
2175All omitted tests for other languages are functionally identical to this test (as for all other tests).
2176The difference in performance between coroutine and thread context-switch is the cost of scheduling for threads, whereas coroutines are self-scheduling.
2177
2178\begin{figure}
2179\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2180
2181\newbox\myboxA
2182\begin{lrbox}{\myboxA}
2183\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2184coroutine C {} c;
2185void main( C & ) { for ( ;; ) { @suspend();@ } }
2186int main() {
2187        BENCH(
2188                for ( size_t i = 0; i < N; i += 1 ) { @resume( c );@ } )
2189        sout | result`ns | endl;
2190}
2191\end{cfa}
2192\end{lrbox}
2193
2194\newbox\myboxB
2195\begin{lrbox}{\myboxB}
2196\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2197
2198
2199int main() {
2200        BENCH(
2201                for ( size_t i = 0; i < N; i += 1 ) { @yield();@ } )
2202        sout | result`ns | endl;
2203}
2204\end{cfa}
2205\end{lrbox}
2206
2207\subfloat[Coroutine]{\label{f:GlobalVariables}\usebox\myboxA}
2208\quad
2209\subfloat[Thread]{\label{f:ExternalState}\usebox\myboxB}
2210\captionof{figure}{\CFA context-switch benchmark}
2211\label{f:ctx-switch}
2212
2213\centering
2214
2215\captionof{table}{Context switch comparison (nanoseconds)}
2216\label{tab:ctx-switch}
2217\bigskip
2218\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2219\cline{2-4}
2220\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2221\hline
2222Kernel Thread   & 241.5         & 243.86        & 5.08 \\
2223\CFA Coroutine  & 38            & 38            & 0    \\
2224\CFA Thread             & 103           & 102.96        & 2.96 \\
2225\uC Coroutine   & 46            & 45.86         & 0.35 \\
2226\uC Thread              & 98            & 99.11         & 1.42 \\
2227Goroutine               & 150           & 149.96        & 3.16 \\
2228Java Thread             & 289           & 290.68        & 8.72 \\
2229\hline
2230\end{tabular}
2231
2232\bigskip
2233\bigskip
2234
2235\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2236\begin{cfa}
2237monitor M {} m1/*, m2, m3, m4*/;
2238void __attribute__((noinline)) do_call( M & mutex m/*, m2, m3, m4*/ ) {}
2239int main() {
2240        BENCH( for( size_t i = 0; i < N; i += 1 ) { @do_call( m1/*, m2, m3, m4*/ );@ } )
2241        sout | result`ns | endl;
2242}
2243\end{cfa}
2244\captionof{figure}{\CFA acquire/release mutex benchmark}
2245\label{f:mutex}
2246
2247\centering
2248
2249\captionof{table}{Mutex comparison (nanoseconds)}
2250\label{tab:mutex}
2251\bigskip
2252
2253\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2254\cline{2-4}
2255\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2256\hline
2257C routine                                                       & 2                     & 2                     & 0    \\
2258FetchAdd + FetchSub                                     & 26            & 26            & 0    \\
2259Pthreads Mutex Lock                                     & 31            & 31.86         & 0.99 \\
2260\uC @monitor@ member routine            & 30            & 30            & 0    \\
2261\CFA @mutex@ routine, 1 argument        & 41            & 41.57         & 0.9  \\
2262\CFA @mutex@ routine, 2 argument        & 76            & 76.96         & 1.57 \\
2263\CFA @mutex@ routine, 4 argument        & 145           & 146.68        & 3.85 \\
2264Java synchronized routine                       & 27            & 28.57         & 2.6  \\
2265\hline
2266\end{tabular}
2267\end{figure}
2268
2269
2270\paragraph{Mutual-Exclusion}
2271
2272Mutual exclusion is measured by entering/leaving a critical section.
2273For monitors, entering and leaving a monitor routine is measured.
2274Figure~\ref{f:mutex} shows the code for \CFA with all results in Table~\ref{tab:mutex}.
2275To 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.
2276Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2277
2278
2279\paragraph{Internal Scheduling}
2280
2281Internal scheduling is measured by waiting on and signalling a condition variable.
2282Figure~\ref{f:int-sched} shows the code for \CFA, with results in Table~\ref{tab:int-sched}.
2283Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2284
2285\begin{figure}
2286\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2287\begin{cfa}
2288volatile int go = 0;
2289condition c;
2290monitor M {} m;
2291void __attribute__((noinline)) do_call( M & mutex a1 ) { signal( c ); }
2292thread T {};
2293void main( T & this ) {
2294        while ( go == 0 ) { yield(); }  // wait for other thread to start
2295        while ( go == 1 ) { @do_call( m );@ }
2296}
2297int  __attribute__((noinline)) do_wait( M & mutex m ) {
2298        go = 1; // continue other thread
2299        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @wait( c );@ } );
2300        go = 0; // stop other thread
2301        sout | result`ns | endl;
2302}
2303int main() {
2304        T t;
2305        do_wait( m );
2306}
2307\end{cfa}
2308\captionof{figure}{\CFA Internal scheduling benchmark}
2309\label{f:int-sched}
2310
2311\centering
2312\captionof{table}{Internal scheduling comparison (nanoseconds)}
2313\label{tab:int-sched}
2314\bigskip
2315
2316\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2317\cline{2-4}
2318\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2319\hline
2320Pthreads Condition Variable             & 5902.5        & 6093.29       & 714.78 \\
2321\uC @signal@                                    & 322           & 323           & 3.36   \\
2322\CFA @signal@, 1 @monitor@              & 352.5         & 353.11        & 3.66   \\
2323\CFA @signal@, 2 @monitor@              & 430           & 430.29        & 8.97   \\
2324\CFA @signal@, 4 @monitor@              & 594.5         & 606.57        & 18.33  \\
2325Java @notify@                                   & 13831.5       & 15698.21      & 4782.3 \\
2326\hline
2327\end{tabular}
2328\end{figure}
2329
2330
2331\paragraph{External Scheduling}
2332
2333External scheduling is measured by accepting a call using the @waitfor@ statement (@_Accept@ in \uC).
2334Figure~\ref{f:ext-sched} shows the code for \CFA, with results in Table~\ref{tab:ext-sched}.
2335Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2336
2337\begin{figure}
2338\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2339\begin{cfa}
2340volatile int go = 0;
2341monitor M {} m;
2342thread T {};
2343void __attribute__((noinline)) do_call( M & mutex ) {}
2344void main( T & ) {
2345        while ( go == 0 ) { yield(); }  // wait for other thread to start
2346        while ( go == 1 ) { @do_call( m );@ }
2347}
2348int __attribute__((noinline)) do_wait( M & mutex m ) {
2349        go = 1; // continue other thread
2350        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @waitfor( do_call, m );@ } )
2351        go = 0; // stop other thread
2352        sout | result`ns | endl;
2353}
2354int main() {
2355        T t;
2356        do_wait( m );
2357}
2358\end{cfa}
2359\captionof{figure}{\CFA external scheduling benchmark}
2360\label{f:ext-sched}
2361
2362\centering
2363
2364\captionof{table}{External scheduling comparison (nanoseconds)}
2365\label{tab:ext-sched}
2366\bigskip
2367\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2368\cline{2-4}
2369\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2370\hline
2371\uC @_Accept@                           & 350           & 350.61        & 3.11  \\
2372\CFA @waitfor@, 1 @monitor@     & 358.5         & 358.36        & 3.82  \\
2373\CFA @waitfor@, 2 @monitor@     & 422           & 426.79        & 7.95  \\
2374\CFA @waitfor@, 4 @monitor@     & 579.5         & 585.46        & 11.25 \\
2375\hline
2376\end{tabular}
2377
2378\bigskip
2379\medskip
2380
2381\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2382\begin{cfa}
2383thread MyThread {};
2384void main( MyThread & ) {}
2385int main() {
2386        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @MyThread m;@ } )
2387        sout | result`ns | endl;
2388}
2389\end{cfa}
2390\captionof{figure}{\CFA object creation benchmark}
2391\label{f:creation}
2392
2393\centering
2394
2395\captionof{table}{Creation comparison (nanoseconds)}
2396\label{tab:creation}
2397\bigskip
2398
2399\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2400\cline{2-4}
2401\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} & \multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2402\hline
2403Pthreads                                & 26996         & 26984.71      & 156.6  \\
2404\CFA Coroutine Lazy             & 6                     & 5.71          & 0.45   \\
2405\CFA Coroutine Eager    & 708           & 706.68        & 4.82   \\
2406\CFA Thread                             & 1173.5        & 1176.18       & 15.18  \\
2407\uC Coroutine                   & 109           & 107.46        & 1.74   \\
2408\uC Thread                              & 526           & 530.89        & 9.73   \\
2409Goroutine                               & 2520.5        & 2530.93       & 61.56  \\
2410Java Thread                             & 91114.5       & 92272.79      & 961.58 \\
2411\hline
2412\end{tabular}
2413\end{figure}
2414
2415
2416\paragraph{Object Creation}
2417
2418Object creation is measured by creating/deleting the specific kind of concurrent object.
2419Figure~\ref{f:creation} shows the code for \CFA, with results in Table~\ref{tab:creation}.
2420The 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.
2421
2422
2423\section{Conclusion}
2424
2425This paper demonstrate a concurrency API that is simple, efficient, and able to build higher-level concurrency features.
2426The 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.
2427The M:N model is judged to be efficient and provide greater flexibility than a 1:1 threading model.
2428High-level objects (monitor/task) are the core mechanism for mutual exclusion and synchronization.
2429A novel aspect is allowing multiple mutex-objects to be accessed simultaneously reducing the potential for deadlock for this complex scenario.
2430These concepts and the entire \CFA runtime-system are written in the \CFA language, demonstrating the expressiveness of the \CFA language.
2431Performance 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.
2432C 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.
2433
2434
2435\section{Future Work}
2436
2437While concurrency in \CFA has a strong start, development is still underway and there are missing features.
2438
2439\paragraph{Flexible Scheduling}
2440\label{futur:sched}
2441
2442An important part of concurrency is scheduling.
2443Different scheduling algorithms can affect performance (both in terms of average and variation).
2444However, no single scheduler is optimal for all workloads and therefore there is value in being able to change the scheduler for given programs.
2445One solution is to offer various tweaking options, allowing the scheduler to be adjusted to the requirements of the workload.
2446However, to be truly flexible, it is necessary to have a pluggable scheduler.
2447Currently, 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.
2448
2449\paragraph{Non-Blocking I/O}
2450\label{futur:nbio}
2451
2452Many modern workloads are not bound by computation but IO operations, a common case being web servers and XaaS~\cite{XaaS} (anything as a service).
2453These types of workloads require significant engineering to amortizing costs of blocking IO-operations.
2454At 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.
2455Current 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.
2456However, these solutions lead to code that is hard create, read and maintain.
2457A 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.
2458A non-blocking I/O library is currently under development for \CFA.
2459
2460\paragraph{Other Concurrency Tools}
2461\label{futur:tools}
2462
2463While monitors offer a flexible and powerful concurrent for \CFA, other concurrency tools are also necessary for a complete multi-paradigm concurrency package.
2464Examples of such tools can include futures and promises~\cite{promises}, executors and actors.
2465These additional features are useful when monitors offer a level of abstraction that is inadequate for certain tasks.
2466
2467\paragraph{Implicit Threading}
2468\label{futur:implcit}
2469
2470Basic 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.
2471This type of concurrency can be achieved both at the language level and at the library level.
2472The canonical example of implicit concurrency is concurrent nested @for@ loops, which are amenable to divide and conquer algorithms~\cite{uC++book}.
2473The \CFA language features should make it possible to develop a reasonable number of implicit concurrency mechanism to solve basic HPC data-concurrency problems.
2474However, implicit concurrency is a restrictive solution and has its limitations, so it can never replace explicit concurrent programming.
2475
2476
2477\section{Acknowledgements}
2478
2479The authors would like to recognize the design assistance of Aaron Moss, Rob Schluntz and Andrew Beach on the features described in this paper.
2480Funding 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.
2481
2482{%
2483\fontsize{9bp}{12bp}\selectfont%
2484\bibliography{pl,local}
2485}%
2486
2487\end{document}
2488
2489% Local Variables: %
2490% tab-width: 4 %
2491% fill-column: 120 %
2492% compile-command: "make" %
2493% End: %
Note: See TracBrowser for help on using the repository browser.