source: doc/papers/concurrency/Paper.tex @ 287da46

aaron-thesisarm-ehcleanup-dtorsdeferred_resndemanglerjacob/cs343-translationjenkins-sandboxnew-astnew-ast-unique-exprno_listpersistent-indexer
Last change on this file since 287da46 was 287da46, checked in by Peter A. Buhr <pabuhr@…>, 4 years ago

more updates

  • Property mode set to 100644
File size: 122.3 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 for concurrency.
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 show comparable 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 (thread pool)~\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 access to shared data and safe communication.
265\newterm{Locking}, and by extension \newterm{locks}, are defined as a mechanism to prevent progress of threads to provide safety.
266\newterm{Parallelism} is running multiple threads simultaneously.
267Parallelism implies \emph{actual} simultaneous execution, where concurrency only requires \emph{apparent} simultaneous execution.
268As such, parallelism only affects performance, which is observed through differences in space and/or time at runtime.
269Hence, there are two problems to be solved: concurrency and parallelism.
270While these two concepts are often combined, they are distinct, requiring different tools~\cite[\S~2]{Buhr05a}.
271Concurrency tools handle mutual exclusion and synchronization, while parallelism tools handle performance, cost, and resource utilization.
272
273The proposed concurrency API is implemented in a dialect of C, called \CFA.
274The paper discusses how the language features are added to the \CFA translator with respect to parsing, semantics, and type checking, and the corresponding high-performance runtime-library to implement the concurrent features.
275
276
277\section{\CFA Overview}
278
279The following is a quick introduction to the \CFA language, specifically tailored to the features needed to support concurrency.
280Extended versions and explanation of the following code examples are available at the \CFA website~\cite{Cforall} or in Moss~\etal~\cite{Moss18}.
281
282\CFA is a non-object-oriented extension of ISO-C, and hence, supports all C paradigms.
283%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.
284Like C, the building blocks of \CFA are structures and routines.
285Virtually all of the code generated by the \CFA translator respects C memory layouts and calling conventions.
286While \CFA is not object oriented, 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}.
287While some \CFA features are common in object-oriented programming, they are independent capabilities allowing \CFA to adopt them while maintaining a procedural paradigm.
288
289
290\subsection{References}
291
292\CFA provides multi-level rebindable references, as an alternative to pointers, which significantly reduces syntactic noise.
293\begin{cfa}
294int x = 1, y = 2, z = 3;
295int * p1 = &x, ** p2 = &p1,  *** p3 = &p2,      $\C{// pointers to x}$
296    `&` r1 = x,   `&&` r2 = r1,   `&&&` r3 = r2;        $\C{// references to x}$
297int * p4 = &z, `&` r4 = z;
298
299*p1 = 3; **p2 = 3; ***p3 = 3;       // change x
300r1 =  3;     r2 = 3;      r3 = 3;        // change x: implicit dereferences *r1, **r2, ***r3
301**p3 = &y; *p3 = &p4;                // change p1, p2
302`&`r3 = &y; `&&`r3 = &`&`r4;             // change r1, r2: cancel implicit dereferences (&*)**r3, (&(&*)*)*r3, &(&*)r4
303\end{cfa}
304A reference is a handle to an object, like a pointer, but is automatically dereferenced the specified number of levels.
305Referencing (address-of @&@) a reference variable cancels one of the implicit dereferences, until there are no more implicit references, after which normal expression behaviour applies.
306
307
308\subsection{\texorpdfstring{\protect\lstinline{with} Statement}{with Statement}}
309\label{s:WithStatement}
310
311Heterogeneous data is aggregated into a structure/union.
312To 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.
313\begin{cquote}
314\vspace*{-\baselineskip}%???
315\lstDeleteShortInline@%
316\begin{cfa}
317struct S { char c; int i; double d; };
318struct T { double m, n; };
319// multiple aggregate parameters
320\end{cfa}
321\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
322\begin{cfa}
323void f( S & s, T & t ) {
324        `s.`c; `s.`i; `s.`d;
325        `t.`m; `t.`n;
326}
327\end{cfa}
328&
329\begin{cfa}
330void f( S & s, T & t ) `with ( s, t )` {
331        c; i; d;                // no qualification
332        m; n;
333}
334\end{cfa}
335\end{tabular}
336\lstMakeShortInline@%
337\end{cquote}
338Object-oriented programming languages only provide implicit qualification for the receiver.
339
340In detail, the @with@ statement has the form:
341\begin{cfa}
342$\emph{with-statement}$:
343        'with' '(' $\emph{expression-list}$ ')' $\emph{compound-statement}$
344\end{cfa}
345and may appear as the body of a routine or nested within a routine body.
346Each expression in the expression-list provides a type and object.
347The type must be an aggregate type.
348(Enumerations are already opened.)
349The object is the implicit qualifier for the open structure-fields.
350All 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.
351
352
353\subsection{Overloading}
354
355\CFA maximizes the ability to reuse names via overloading to aggressively address the naming problem.
356Both variables and routines may be overloaded, where selection is based on types, and number of returns (as in Ada~\cite{Ada}) and arguments.
357\begin{cquote}
358\vspace*{-\baselineskip}%???
359\lstDeleteShortInline@%
360\begin{cfa}
361// selection based on type
362\end{cfa}
363\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
364\begin{cfa}
365const short int `MIN` = -32768;
366const int `MIN` = -2147483648;
367const long int `MIN` = -9223372036854775808L;
368\end{cfa}
369&
370\begin{cfa}
371short int si = `MIN`;
372int i = `MIN`;
373long int li = `MIN`;
374\end{cfa}
375\end{tabular}
376\begin{cfa}
377// selection based on type and number of parameters
378\end{cfa}
379\begin{tabular}{@{}l@{\hspace{2.7\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
380\begin{cfa}
381void `f`( void );
382void `f`( char );
383void `f`( int, double );
384\end{cfa}
385&
386\begin{cfa}
387`f`();
388`f`( 'a' );
389`f`( 3, 5.2 );
390\end{cfa}
391\end{tabular}
392\begin{cfa}
393// selection based on type and number of returns
394\end{cfa}
395\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
396\begin{cfa}
397char `f`( int );
398double `f`( int );
399[char, double] `f`( int );
400\end{cfa}
401&
402\begin{cfa}
403char c = `f`( 3 );
404double d = `f`( 3 );
405[d, c] = `f`( 3 );
406\end{cfa}
407\end{tabular}
408\lstMakeShortInline@%
409\end{cquote}
410Overloading is important for \CFA concurrency since the runtime system relies on creating different types to represent concurrency objects.
411Therefore, overloading eliminates long prefixes and other naming conventions to prevent name clashes.
412As seen in Section~\ref{s:Concurrency}, routine @main@ is heavily overloaded.
413For example, variable overloading is useful in the parallel semantics of the @with@ statement for fields with the same name:
414\begin{cfa}
415struct S { int `i`; int j; double m; } s;
416struct T { int `i`; int k; int m; } t;
417with ( s, t ) {
418        j + k;                                                                  $\C{// unambiguous, s.j + t.k}$
419        m = 5.0;                                                                $\C{// unambiguous, s.m = 5.0}$
420        m = 1;                                                                  $\C{// unambiguous, t.m = 1}$
421        int a = m;                                                              $\C{// unambiguous, a = t.m }$
422        double b = m;                                                   $\C{// unambiguous, b = s.m}$
423        int c = `s.i` + `t.i`;                                  $\C{// unambiguous, qualification}$
424        (double)m;                                                              $\C{// unambiguous, cast s.m}$
425}
426\end{cfa}
427For parallel semantics, both @s.i@ and @t.i@ are visible with the same type, so only @i@ is ambiguous without qualification.
428
429
430\subsection{Operators}
431
432Overloading also extends to operators.
433Operator-overloading syntax creates a routine name with an operator symbol and question marks for the operands:
434\begin{cquote}
435\lstDeleteShortInline@%
436\begin{tabular}{@{}ll@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
437\begin{cfa}
438int ++?(int op);
439int ?++(int op);
440int `?+?`(int op1, int op2);
441int ?<=?(int op1, int op2);
442int ?=? (int & op1, int op2);
443int ?+=?(int & op1, int op2);
444\end{cfa}
445&
446\begin{cfa}
447// unary prefix increment
448// unary postfix increment
449// binary plus
450// binary less than
451// binary assignment
452// binary plus-assignment
453\end{cfa}
454&
455\begin{cfa}
456struct S { int i, j; };
457S `?+?`( S op1, S op2) { // add two structures
458        return (S){op1.i + op2.i, op1.j + op2.j};
459}
460S s1 = {1, 2}, s2 = {2, 3}, s3;
461s3 = s1 `+` s2;         // compute sum: s3 == {2, 5}
462\end{cfa}
463\end{tabular}
464\lstMakeShortInline@%
465\end{cquote}
466While concurrency does not use operator overloading directly, it provides an introduction for the syntax of constructors.
467
468
469\subsection{Constructors / Destructors}
470
471Object lifetime is a challenge in non-managed programming languages.
472\CFA responds with \CC-like constructors and destructors:
473\begin{cfa}
474struct VLA { int len, * data; };                        $\C{// variable length array of integers}$
475void ?{}( VLA & vla ) with ( vla ) { len = 10;  data = alloc( len ); }  $\C{// default constructor}$
476void ?{}( VLA & vla, int size, char fill ) with ( vla ) { len = size;  data = alloc( len, fill ); } // initialization
477void ?{}( VLA & vla, VLA other ) { vla.len = other.len;  vla.data = other.data; } $\C{// copy, shallow}$
478void ^?{}( VLA & vla ) with ( vla ) { free( data ); } $\C{// destructor}$
479{
480        VLA  x,            y = { 20, 0x01 },     z = y; $\C{// z points to y}$
481        // $\LstCommentStyle{\color{red}\ \ \ x\{\};\ \ \ \ \ \ \ \ \ y\{ 20, 0x01 \};\ \ \ \ \ \ \ \ \ \ z\{ z, y \};\ \ \ \ \ \ \ implicit calls}$
482        ^x{};                                                                   $\C{// deallocate x}$
483        x{};                                                                    $\C{// reallocate x}$
484        z{ 5, 0xff };                                                   $\C{// reallocate z, not pointing to y}$
485        ^y{};                                                                   $\C{// deallocate y}$
486        y{ x };                                                                 $\C{// reallocate y, points to x}$
487        x{};                                                                    $\C{// reallocate x, not pointing to y}$
488}       //  $\LstCommentStyle{\color{red}\^{}z\{\};\ \ \^{}y\{\};\ \ \^{}x\{\};\ \ \ implicit calls}$
489\end{cfa}
490Like \CC, construction is implicit on allocation (stack/heap) and destruction is implicit on deallocation.
491The object and all their fields are constructed/destructed.
492\CFA also provides @new@ and @delete@, which behave like @malloc@ and @free@, in addition to constructing and destructing objects:
493\begin{cfa}
494{
495        ... struct S s = {10}; ...                              $\C{// allocation, call constructor}$
496}                                                                                       $\C{// deallocation, call destructor}$
497struct S * s = new();                                           $\C{// allocation, call constructor}$
498...
499delete( s );                                                            $\C{// deallocation, call destructor}$
500\end{cfa}
501\CFA concurrency uses object lifetime as a means of mutual exclusion and/or synchronization.
502
503
504\subsection{Parametric Polymorphism}
505\label{s:ParametricPolymorphism}
506
507The 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.
508For example, the following sum routine works for any type that supports construction from 0 and addition:
509\begin{cfa}
510forall( otype T | { void `?{}`( T *, zero_t ); T `?+?`( T, T ); } ) // constraint type, 0 and +
511T sum( T a[$\,$], size_t size ) {
512        `T` total = { `0` };                                    $\C{// initialize by 0 constructor}$
513        for ( size_t i = 0; i < size; i += 1 )
514                total = total `+` a[i];                         $\C{// select appropriate +}$
515        return total;
516}
517S sa[5];
518int i = sum( sa, 5 );                                           $\C{// use S's 0 construction and +}$
519\end{cfa}
520The 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.
521
522\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:
523\begin{cfa}
524trait `sumable`( otype T ) {
525        void `?{}`( T &, zero_t );                              $\C{// 0 literal constructor}$
526        T `?+?`( T, T );                                                $\C{// assortment of additions}$
527        T ?+=?( T &, T );
528        T ++?( T & );
529        T ?++( T & );
530};
531forall( otype T `| sumable( T )` )                      $\C{// use trait}$
532T sum( T a[$\,$], size_t size );
533\end{cfa}
534
535Assertions can be @otype@ or @dtype@.
536@otype@ refers to a \emph{complete} object, \ie an object has a size, alignment, default constructor, copy constructor, destructor and an assignment operator.
537@dtype@ refers to an \emph{incomplete} object, \ie, an object only has a size and alignment.
538
539Using the return type for overload discrimination, it is possible to write a type-safe @alloc@ based on the C @malloc@:
540\begin{cfa}
541forall( dtype T | sized(T) ) T * alloc( void ) { return (T *)malloc( sizeof(T) ); }
542int * ip = alloc();                                                     $\C{// select type and size from left-hand side}$
543double * dp = alloc();
544struct S {...} * sp = alloc();
545\end{cfa}
546where the return type supplies the type/size of the allocation, which is impossible in most type systems.
547
548
549\section{Concurrency}
550\label{s:Concurrency}
551
552At its core, concurrency is based on multiple call-stacks and scheduling threads executing on these stacks.
553Multiple 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}.
554In coroutining, the single thread is self-scheduling across the stacks, so execution is deterministic, \ie the execution path from input to output is fixed and predictable.
555A \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;
556a \newterm{stackfull} coroutine executes on its own stack, allowing full generality.
557Only stackfull coroutines are a stepping stone to concurrency.
558
559The transition to concurrency, even for execution with a single thread and multiple stacks, occurs when coroutines also context switch to a \newterm{scheduling oracle}, introducing non-determinism from the coroutine perspective~\cite[\S~3]{Buhr05a}.
560Therefore, a minimal concurrency system is possible using coroutines (see Section \ref{coroutine}) in conjunction with a scheduler to decide where to context switch next.
561The resulting execution system now follows a cooperative threading-model, called \newterm{non-preemptive scheduling}.
562
563Because the scheduler is special, it can either be a stackless or stackfull coroutine.
564For 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.
565For 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.
566A stackfull scheduler is often used for simplicity and security, even through there is a slightly higher runtime-cost.
567
568Regardless of the approach used, a subset of concurrency related challenges start to appear.
569For 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}.
570While a scheduler introduces uncertainty in the order of execution, preemption introduces uncertainty where context switches occur.
571Interestingly, uncertainty is necessary for the runtime (operating) system to give the illusion of parallelism on a single processor and increase performance on multiple processors.
572The reason is that only the runtime has complete knowledge about resources and how to best utilized them.
573However, the introduction of unrestricted non-determinism results in the need for \newterm{mutual exclusion} and \newterm{synchronization} to restrict non-determinism for correctness;
574otherwise, it is impossible to write meaningful programs.
575Optimal performance in concurrent applications is often obtained by having as much non-determinism as correctness allows.
576
577
578\subsection{\protect\CFA's Thread Building Blocks}
579
580An 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.
581As such, library support for threading is far from widespread.
582At the time of writing the paper, neither \protect\lstinline@gcc@ nor \protect\lstinline@clang@ support \protect\lstinline@threads.h@ in their standard libraries.}.
583In 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.
584As 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.
585Furthermore, because C is a system-level language, programmers expect to choose precisely which features they need and which cost they are willing to pay.
586Hence, concurrent programs should be written using high-level mechanisms, and only step down to lower-level mechanisms when performance bottlenecks are encountered.
587
588
589\subsection{Coroutines: A Stepping Stone}\label{coroutine}
590
591While the focus of this discussion is concurrency and parallelism, it is important to address coroutines, which are a significant building block of a concurrency system (but not concurrent among themselves).
592Coroutines are generalized routines allowing execution to be temporarily suspend and later resumed.
593Hence, 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.
594This capability is accomplish via the coroutine's stack, where suspend/resume context switch among stacks.
595Because 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.
596Therefore, the core \CFA coroutine-API for has two fundamental features: independent call-stacks and @suspend@/@resume@ operations.
597
598For example, a problem made easier with coroutines is unbounded generators, \eg generating an infinite sequence of Fibonacci numbers
599\begin{displaymath}
600\mathsf{fib}(n) = \left \{
601\begin{array}{ll}
6020                                       & n = 0         \\
6031                                       & n = 1         \\
604\mathsf{fib}(n-1) + \mathsf{fib}(n-2)   & n \ge 2       \\
605\end{array}
606\right.
607\end{displaymath}
608where Figure~\ref{f:C-fibonacci} shows conventional approaches for writing a Fibonacci generator in C.
609Figure~\ref{f:GlobalVariables} illustrates the following problems: unique unencapsulated global variables necessary to retain state between calls, only one Fibonacci generator, and execution state must be explicitly retained via explicit state variables.
610Figure~\ref{f:ExternalState} addresses these issues: unencapsulated program global variables become encapsulated structure variables, unique global variables are replaced by multiple Fibonacci objects, and explicit execution state is removed by precomputing the first two Fibonacci numbers and returning $\mathsf{fib}(n-2)$.
611
612\begin{figure}
613\centering
614\newbox\myboxA
615\begin{lrbox}{\myboxA}
616\begin{cfa}[aboveskip=0pt,belowskip=0pt]
617`int f1, f2, state = 1;`   // single global variables
618int fib() {
619        int fn;
620        `switch ( state )` {  // explicit execution state
621          case 1: fn = 0;  f1 = fn;  state = 2;  break;
622          case 2: fn = 1;  f2 = f1;  f1 = fn;  state = 3;  break;
623          case 3: fn = f1 + f2;  f2 = f1;  f1 = fn;  break;
624        }
625        return fn;
626}
627int main() {
628
629        for ( int i = 0; i < 10; i += 1 ) {
630                printf( "%d\n", fib() );
631        }
632}
633\end{cfa}
634\end{lrbox}
635
636\newbox\myboxB
637\begin{lrbox}{\myboxB}
638\begin{cfa}[aboveskip=0pt,belowskip=0pt]
639#define FIB_INIT `{ 0, 1 }`
640typedef struct { int f2, f1; } Fib;
641int fib( Fib * f ) {
642
643        int ret = f->f2;
644        int fn = f->f1 + f->f2;
645        f->f2 = f->f1; f->f1 = fn;
646
647        return ret;
648}
649int main() {
650        Fib f1 = FIB_INIT, f2 = FIB_INIT;
651        for ( int i = 0; i < 10; i += 1 ) {
652                printf( "%d %d\n", fib( &f1 ), fib( &f2 ) );
653        }
654}
655\end{cfa}
656\end{lrbox}
657
658\subfloat[3 States: global variables]{\label{f:GlobalVariables}\usebox\myboxA}
659\qquad
660\subfloat[1 State: external variables]{\label{f:ExternalState}\usebox\myboxB}
661\caption{C Fibonacci Implementations}
662\label{f:C-fibonacci}
663
664\bigskip
665
666\newbox\myboxA
667\begin{lrbox}{\myboxA}
668\begin{cfa}[aboveskip=0pt,belowskip=0pt]
669`coroutine` Fib { int fn; };
670void main( Fib & fib ) with( fib ) {
671        int f1, f2;
672        fn = 0;  f1 = fn;  `suspend()`;
673        fn = 1;  f2 = f1;  f1 = fn;  `suspend()`;
674        for ( ;; ) {
675                fn = f1 + f2;  f2 = f1;  f1 = fn;  `suspend()`;
676        }
677}
678int next( Fib & fib ) with( fib ) {
679        `resume( fib );`
680        return fn;
681}
682int main() {
683        Fib f1, f2;
684        for ( int i = 1; i <= 10; i += 1 ) {
685                sout | next( f1 ) | next( f2 ) | endl;
686        }
687}
688\end{cfa}
689\end{lrbox}
690\newbox\myboxB
691\begin{lrbox}{\myboxB}
692\begin{cfa}[aboveskip=0pt,belowskip=0pt]
693`coroutine` Fib { int ret; };
694void main( Fib & f ) with( fib ) {
695        int fn, f1 = 1, f2 = 0;
696        for ( ;; ) {
697                ret = f2;
698
699                fn = f1 + f2;  f2 = f1;  f1 = fn; `suspend();`
700        }
701}
702int next( Fib & fib ) with( fib ) {
703        `resume( fib );`
704        return ret;
705}
706
707
708
709
710
711
712\end{cfa}
713\end{lrbox}
714\subfloat[3 States, internal variables]{\label{f:Coroutine3States}\usebox\myboxA}
715\qquad\qquad
716\subfloat[1 State, internal variables]{\label{f:Coroutine1State}\usebox\myboxB}
717\caption{\CFA Coroutine Fibonacci Implementations}
718\label{f:cfa-fibonacci}
719\end{figure}
720
721Using a coroutine, it is possible to express the Fibonacci formula directly without any of the C problems.
722Figure~\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@.
723Like 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.
724The 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@.
725The interface routine @next@, takes a Fibonacci instance and context switches to it using @resume@;
726on restart, the Fibonacci field, @fn@, contains the next value in the sequence, which is returned.
727The first @resume@ is special because it allocates the coroutine stack and cocalls its coroutine main on that stack;
728when the coroutine main returns, its stack is deallocated.
729Hence, @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.
730Figure~\ref{f:Coroutine1State} shows the coroutine version of the C version in Figure~\ref{f:ExternalState}.
731Coroutine generators are called \newterm{output coroutines} because values are only returned.
732
733Figure~\ref{f:CFAFmt} shows an \newterm{input coroutine}, @Format@, for restructuring text into groups of characters of fixed-size blocks.
734For example, the input of the left is reformatted into the output on the right.
735\begin{quote}
736\tt
737\begin{tabular}{@{}l|l@{}}
738\multicolumn{1}{c|}{\textbf{\textrm{input}}} & \multicolumn{1}{c}{\textbf{\textrm{output}}} \\
739abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz
740&
741\begin{tabular}[t]{@{}lllll@{}}
742abcd    & efgh  & ijkl  & mnop  & qrst  \\
743uvwx    & yzab  & cdef  & ghij  & klmn  \\
744opqr    & stuv  & wxyz  &               &
745\end{tabular}
746\end{tabular}
747\end{quote}
748The 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.
749The destruction provides a newline, if formatted text ends with a full line.
750Figure~\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.
751(Linearized code is the bane of device drivers.)
752
753\begin{figure}
754\centering
755\newbox\myboxA
756\begin{lrbox}{\myboxA}
757\begin{cfa}[aboveskip=0pt,belowskip=0pt]
758`coroutine` Format {
759        char ch;   // used for communication
760        int g, b;  // global because used in destructor
761};
762void main( Format & fmt ) with( fmt ) {
763        for ( ;; ) {   
764                for ( g = 0; g < 5; g += 1 ) {      // group
765                        for ( b = 0; b < 4; b += 1 ) { // block
766                                `suspend();`
767                                sout | ch;              // separator
768                        }
769                        sout | "  ";               // separator
770                }
771                sout | endl;
772        }
773}
774void ?{}( Format & fmt ) { `resume( fmt );` }
775void ^?{}( Format & fmt ) with( fmt ) {
776        if ( g != 0 || b != 0 ) sout | endl;
777}
778void format( Format & fmt ) {
779        `resume( fmt );`
780}
781int main() {
782        Format fmt;
783        eof: for ( ;; ) {
784                sin | fmt.ch;
785          if ( eof( sin ) ) break eof;
786                format( fmt );
787        }
788}
789\end{cfa}
790\end{lrbox}
791
792\newbox\myboxB
793\begin{lrbox}{\myboxB}
794\begin{cfa}[aboveskip=0pt,belowskip=0pt]
795struct Format {
796        char ch;
797        int g, b;
798};
799void format( struct Format * fmt ) {
800        if ( fmt->ch != -1 ) {      // not EOF ?
801                printf( "%c", fmt->ch );
802                fmt->b += 1;
803                if ( fmt->b == 4 ) {  // block
804                        printf( "  " );      // separator
805                        fmt->b = 0;
806                        fmt->g += 1;
807                }
808                if ( fmt->g == 5 ) {  // group
809                        printf( "\n" );     // separator
810                        fmt->g = 0;
811                }
812        } else {
813                if ( fmt->g != 0 || fmt->b != 0 ) printf( "\n" );
814        }
815}
816int main() {
817        struct Format fmt = { 0, 0, 0 };
818        for ( ;; ) {
819                scanf( "%c", &fmt.ch );
820          if ( feof( stdin ) ) break;
821                format( &fmt );
822        }
823        fmt.ch = -1;
824        format( &fmt );
825}
826\end{cfa}
827\end{lrbox}
828\subfloat[\CFA Coroutine]{\label{f:CFAFmt}\usebox\myboxA}
829\qquad
830\subfloat[C Linearized]{\label{f:CFmt}\usebox\myboxB}
831\caption{Formatting text into lines of 5 blocks of 4 characters.}
832\label{f:fmt-line}
833\end{figure}
834
835The 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.
836However, @resume@ and @suspend@ context switch among existing stack-frames, rather than create new ones so there is no stack growth.
837\newterm{Symmetric (full) coroutine}s have a coroutine call to a resuming routine for another coroutine, and its coroutine main has a call to another resuming routine, which eventually forms a resuming-call cycle.
838(The trivial cycle is a coroutine resuming itself.)
839This control flow is similar to recursion for normal routines, but again there is no stack growth from the context switch.
840
841\begin{figure}
842\centering
843\lstset{language=CFA,escapechar={},moredelim=**[is][\protect\color{red}]{`}{`}}% allow $
844\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
845\begin{cfa}
846`coroutine` Prod {
847        Cons & c;
848        int N, money, receipt;
849};
850void main( Prod & prod ) with( prod ) {
851        // 1st resume starts here
852        for ( int i = 0; i < N; i += 1 ) {
853                int p1 = random( 100 ), p2 = random( 100 );
854                sout | p1 | " " | p2 | endl;
855                int status = delivery( c, p1, p2 );
856                sout | " $" | money | endl | status | endl;
857                receipt += 1;
858        }
859        stop( c );
860        sout | "prod stops" | endl;
861}
862int payment( Prod & prod, int money ) {
863        prod.money = money;
864        `resume( prod );`
865        return prod.receipt;
866}
867void start( Prod & prod, int N, Cons &c ) {
868        &prod.c = &c;
869        prod.[N, receipt] = [N, 0];
870        `resume( prod );`
871}
872int main() {
873        Prod prod;
874        Cons cons = { prod };
875        start( prod, 5, cons );
876}
877\end{cfa}
878&
879\begin{cfa}
880`coroutine` Cons {
881        Prod & p;
882        int p1, p2, status;
883        _Bool done;
884};
885void ?{}( Cons & cons, Prod & p ) {
886        &cons.p = &p;
887        cons.[status, done ] = [0, false];
888}
889void ^?{}( Cons & cons ) {}
890void main( Cons & cons ) with( cons ) {
891        // 1st resume starts here
892        int money = 1, receipt;
893        for ( ; ! done; ) {
894                sout | p1 | " " | p2 | endl | " $" | money | endl;
895                status += 1;
896                receipt = payment( p, money );
897                sout | " #" | receipt | endl;
898                money += 1;
899        }
900        sout | "cons stops" | endl;
901}
902int delivery( Cons & cons, int p1, int p2 ) {
903        cons.[p1, p2] = [p1, p2];
904        `resume( cons );`
905        return cons.status;
906}
907void stop( Cons & cons ) {
908        cons.done = true;
909        `resume( cons );`
910}
911\end{cfa}
912\end{tabular}
913\caption{Producer / consumer: resume-resume cycle, bi-directional communication}
914\label{f:ProdCons}
915\end{figure}
916
917Figure~\ref{f:ProdCons} shows a producer/consumer symmetric-coroutine performing bi-directional communication.
918Since 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@.
919The @start@ routine communicates both the number of elements to be produced and the consumer into the producer's coroutine-structure.
920Then the @resume@ to @prod@ creates @prod@'s stack with a frame for @prod@'s coroutine main at the top, and context switches to it.
921@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.
922
923The producer call to @delivery@ transfers values into the consumer's communication variables, resumes the consumer, and returns the consumer status.
924For the first resume, @cons@'s stack is initialized, creating local variables retained between subsequent activations of the coroutine.
925The 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).
926The call from the consumer to @payment@ introduces the cycle between producer and consumer.
927When @payment@ is called, the consumer copies values into the producer's communication variable and a resume is executed.
928The context switch restarts the producer at the point where it was last context switched, so it continues in @delivery@ after the resume.
929
930@delivery@ returns the status value in @prod@'s coroutine main, where the status is printed.
931The loop then repeats calling @delivery@, where each call resumes the consumer coroutine.
932The context switch to the consumer continues in @payment@.
933The consumer increments and returns the receipt to the call in @cons@'s coroutine main.
934The loop then repeats calling @payment@, where each call resumes the producer coroutine.
935
936After iterating $N$ times, the producer calls @stop@.
937The @done@ flag is set to stop the consumer's execution and a resume is executed.
938The context switch restarts @cons@ in @payment@ and it returns with the last receipt.
939The 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@.
940@stop@ returns and @prod@'s coroutine main terminates.
941The program main restarts after the resume in @start@.
942@start@ returns and the program main terminates.
943
944
945\subsection{Coroutine Implementation}
946
947A 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.
948There are several solutions to this problem and the chosen option forced the \CFA coroutine design.
949
950Object-oriented inheritance provides extra fields and code in a restricted context, but it requires programmers to explicitly perform the inheritance:
951\begin{cfa}[morekeywords={class,inherits}]
952class mycoroutine inherits baseCoroutine { ... }
953\end{cfa}
954and the programming language (and possibly its tool set, \eg debugger) may need to understand @baseCoroutine@ because of the stack.
955Furthermore, the execution of constructs/destructors is in the wrong order for certain operations.
956For 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.
957
958An alternatively is composition:
959\begin{cfa}
960struct mycoroutine {
961        ... // declarations
962        baseCoroutine dummy; // composition, last declaration
963}
964\end{cfa}
965which also requires an explicit declaration that must be the last one to ensure correct initialization order.
966However, there is nothing preventing wrong placement or multiple declarations.
967
968For coroutines as for threads, many implementations are based on routine pointers or routine objects~\cite{Butenhof97, C++14, MS:VisualC++, BoostCoroutines15}.
969For example, Boost implements coroutines in terms of four functor object-types:
970\begin{cfa}
971asymmetric_coroutine<>::pull_type
972asymmetric_coroutine<>::push_type
973symmetric_coroutine<>::call_type
974symmetric_coroutine<>::yield_type
975\end{cfa}
976Similarly, the canonical threading paradigm is often based on routine pointers, \eg @pthreads@~\cite{Butenhof97}, \Csharp~\cite{Csharp}, Go~\cite{Go}, and Scala~\cite{Scala}.
977However, the generic thread-handle (identifier) is limited (few operations), unless it is wrapped in a custom type.
978\begin{cfa}
979void mycor( coroutine_t cid, void * arg ) {
980        int * value = (int *)arg;                               $\C{// type unsafe, pointer-size only}$
981        // Coroutine body
982}
983int main() {
984        int input = 0, output;
985        coroutine_t cid = coroutine_create( &mycor, (void *)&input ); $\C{// type unsafe, pointer-size only}$
986        coroutine_resume( cid, (void *)input, (void **)&output ); $\C{// type unsafe, pointer-size only}$
987}
988\end{cfa}
989Since the custom type is simple to write in \CFA and solves several issues, added support for routine/lambda-based coroutines adds very little.
990
991Note, the type @coroutine_t@ must be an abstract handle to the coroutine, because the coroutine descriptor and its stack are non-copyable.
992Copying the coroutine descriptor results in copies being out of date with the current state of the stack.
993Correspondingly, 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.
994(There is no mechanism in C to find all stack-specific pointers and update them as part of a copy.)
995
996The selected approach is to use language support by introducing a new kind of aggregate (structure):
997\begin{cfa}
998coroutine Fibonacci {
999        int fn; // communication variables
1000};
1001\end{cfa}
1002The @coroutine@ keyword means the compiler (and tool set) can find and inject code where needed.
1003The downside of this approach is that it makes coroutine a special case in the language.
1004Users wanting to extend coroutines or build their own for various reasons can only do so in ways offered by the language.
1005Furthermore, implementing coroutines without language supports also displays the power of a programming language.
1006While this is ultimately the option used for idiomatic \CFA code, coroutines and threads can still be constructed without using the language support.
1007The reserved keyword simply eases use for the common cases.
1008
1009Part 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:
1010\begin{cfa}
1011trait is_coroutine( `dtype` T ) {
1012        void main( T & );
1013        coroutine_desc * get_coroutine( T & );
1014};
1015forall( `dtype` T | is_coroutine(T) ) void suspend( T & );
1016forall( `dtype` T | is_coroutine(T) ) void resume( T & );
1017\end{cfa}
1018The @dtype@ property of the trait ensures the coroutine descriptor is non-copyable, so all coroutines must be passed by reference (pointer).
1019The 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.
1020The @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.
1021The 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.
1022The 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@.
1023The \CFA keyword @coroutine@ implicitly implements the getter and forward declarations required for implementing the coroutine main:
1024\begin{cquote}
1025\begin{tabular}{@{}ccc@{}}
1026\begin{cfa}
1027coroutine MyCor {
1028        int value;
1029
1030};
1031\end{cfa}
1032&
1033{\Large $\Rightarrow$}
1034&
1035\begin{tabular}{@{}ccc@{}}
1036\begin{cfa}
1037struct MyCor {
1038        int value;
1039        coroutine_desc cor;
1040};
1041\end{cfa}
1042&
1043\begin{cfa}
1044static inline coroutine_desc *
1045get_coroutine( MyCor & this ) {
1046        return &this.cor;
1047}
1048\end{cfa}
1049&
1050\begin{cfa}
1051void main( MyCor * this );
1052
1053
1054
1055\end{cfa}
1056\end{tabular}
1057\end{tabular}
1058\end{cquote}
1059The 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.
1060
1061
1062\subsection{Thread Interface}
1063\label{threads}
1064
1065Both user and kernel threads are supported, where user threads provide concurrency and kernel threads provide parallelism.
1066Like 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:
1067\begin{cquote}
1068\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1069\begin{cfa}
1070thread myThread {
1071        // communication variables
1072};
1073
1074
1075\end{cfa}
1076&
1077\begin{cfa}
1078trait is_thread( `dtype` T ) {
1079      void main( T & );
1080      thread_desc * get_thread( T & );
1081      void ^?{}( T & `mutex` );
1082};
1083\end{cfa}
1084\end{tabular}
1085\end{cquote}
1086(The qualifier @mutex@ for the destructor parameter is discussed in Section~\ref{s:Monitor}.)
1087Like a coroutine, the statically-typed @main@ routine is the starting point (first stack frame) of a user thread.
1088The difference is that a coroutine borrows a thread from its caller, so the first thread resuming a coroutine creates an instance of @main@;
1089whereas, 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{
1090The \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.}
1091No 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.
1092
1093\begin{comment} % put in appendix with coroutine version ???
1094As such the @main@ routine of a thread can be defined as
1095\begin{cfa}
1096thread foo {};
1097
1098void main(foo & this) {
1099        sout | "Hello World!" | endl;
1100}
1101\end{cfa}
1102
1103In 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.
1104With 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.
1105\begin{cfa}
1106typedef void (*voidRtn)(int);
1107
1108thread RtnRunner {
1109        voidRtn func;
1110        int arg;
1111};
1112
1113void ?{}(RtnRunner & this, voidRtn inRtn, int arg) {
1114        this.func = inRtn;
1115        this.arg  = arg;
1116}
1117
1118void main(RtnRunner & this) {
1119        // thread starts here and runs the routine
1120        this.func( this.arg );
1121}
1122
1123void hello(/*unused*/ int) {
1124        sout | "Hello World!" | endl;
1125}
1126
1127int main() {
1128        RtnRunner f = {hello, 42};
1129        return 0?
1130}
1131\end{cfa}
1132A 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}.
1133\end{comment}
1134
1135For user threads to be useful, it must be possible to start and stop the underlying thread, and wait for it to complete execution.
1136While using an API such as @fork@ and @join@ is relatively common, such an interface is awkward and unnecessary.
1137A simple approach is to use allocation/deallocation principles, and have threads implicitly @fork@ after construction and @join@ before destruction.
1138\begin{cfa}
1139thread World {};
1140void main( World & this ) {
1141        sout | "World!" | endl;
1142}
1143int main() {
1144        World w`[10]`;                                                  $\C{// implicit forks after creation}$
1145        sout | "Hello " | endl;                                 $\C{// "Hello " and 10 "World!" printed concurrently}$
1146}                                                                                       $\C{// implicit joins before destruction}$
1147\end{cfa}
1148This semantics ensures a thread is started and stopped exactly once, eliminating some programming error, and scales to multiple threads for basic (termination) synchronization.
1149This 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.
1150\begin{cfa}
1151int main() {
1152        MyThread * heapLived;
1153        {
1154                MyThread blockLived;                            $\C{// fork block-based thread}$
1155                heapLived = `new`( MyThread );          $\C{// fork heap-based thread}$
1156                ...
1157        }                                                                               $\C{// join block-based thread}$
1158        ...
1159        `delete`( heapLived );                                  $\C{// join heap-based thread}$
1160}
1161\end{cfa}
1162The heap-based approach allows arbitrary thread-creation topologies, with respect to fork/join-style concurrency.
1163
1164Figure~\ref{s:ConcurrentMatrixSummation} shows concurrently adding the rows of a matrix and then totalling the subtotals sequential, after all the row threads have terminated.
1165The program uses heap-based threads because each thread needs different constructor values.
1166(Python provides a simple iteration mechanism to initialize array elements to different values allowing stack allocation.)
1167The allocation/deallocation pattern appears unusual because allocated objects are immediately deallocated without any intervening code.
1168However, for threads, the deletion provides implicit synchronization, which is the intervening code.
1169While 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.
1170
1171\begin{figure}
1172\begin{cfa}
1173thread Adder {
1174    int * row, cols, & subtotal;                        $\C{// communication}$
1175};
1176void ?{}( Adder & adder, int row[], int cols, int & subtotal ) {
1177    adder.[ row, cols, &subtotal ] = [ row, cols, &subtotal ];
1178}
1179void main( Adder & adder ) with( adder ) {
1180    subtotal = 0;
1181    for ( int c = 0; c < cols; c += 1 ) { subtotal += row[c]; }
1182}
1183int main() {
1184    const int rows = 10, cols = 1000;
1185    int matrix[rows][cols], subtotals[rows], total = 0;
1186    // read matrix
1187    Adder * adders[rows];
1188    for ( int r = 0; r < rows; r += 1 ) {       $\C{// start threads to sum rows}$
1189                adders[r] = new( matrix[r], cols, &subtotals[r] );
1190    }
1191    for ( int r = 0; r < rows; r += 1 ) {       $\C{// wait for threads to finish}$
1192                delete( adders[r] );                            $\C{// termination join}$
1193                total += subtotals[r];                          $\C{// total subtotal}$
1194    }
1195    sout | total | endl;
1196}
1197\end{cfa}
1198\caption{Concurrent Matrix Summation}
1199\label{s:ConcurrentMatrixSummation}
1200\end{figure}
1201
1202
1203\section{Mutual Exclusion / Synchronization}
1204
1205Uncontrolled non-deterministic execution is meaningless.
1206To 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}.
1207Since 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).
1208In 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}.
1209However, 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.
1210Hence, a programmer must learn and manipulate two sets of design patterns.
1211While this distinction can be hidden away in library code, effective use of the library still has to take both paradigms into account.
1212In contrast, approaches based on statefull models more closely resemble the standard call/return programming-model, resulting in a single programming paradigm.
1213
1214At the lowest level, concurrent control is implemented by atomic operations, upon which different kinds of locking mechanisms are constructed, \eg semaphores~\cite{Dijkstra68b}, barriers, and path expressions~\cite{Campbell74}.
1215However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
1216A newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
1217While 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.
1218
1219One of the most natural, elegant, and efficient mechanisms for mutual exclusion and synchronization for shared-memory systems is the \emph{monitor}.
1220First 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}.
1221In 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.
1222For these reasons, \CFA selected monitors as the core high-level concurrency-construct, upon which higher-level approaches can be easily constructed.
1223
1224
1225\subsection{Mutual Exclusion}
1226
1227A group of instructions manipulating a specific instance of shared data that must be performed atomically is called an (individual) \newterm{critical-section}~\cite{Dijkstra65}.
1228The 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.
1229The 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.
1230\newterm{Mutual exclusion} enforces that the correct kind and number of threads are using a critical section.
1231
1232However, many solutions exist for mutual exclusion, which vary in terms of performance, flexibility and ease of use.
1233Methods 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.
1234Ease 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.
1235For 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.
1236However, a significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
1237Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
1238
1239
1240\subsection{Synchronization}
1241
1242Synchronization enforces relative ordering of execution, and synchronization tools provide numerous mechanisms to establish these timing relationships.
1243Low-level synchronization primitives offer good performance and flexibility at the cost of ease of use;
1244higher-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.
1245Often 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.
1246If a writer thread is scheduled for next access, but another reader thread acquires the critical section first, that reader has \newterm{barged}.
1247Barging 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).
1248Preventing or detecting barging is an involved challenge with low-level locks, which can be made much easier by higher-level constructs.
1249This challenge is often split into two different approaches: barging avoidance and barging prevention.
1250Algorithms that allow a barger, but divert it until later using current synchronization state (flags), are avoiding the barger;
1251algorithms that preclude a barger from entering during synchronization in the critical section prevent barging completely.
1252Techniques like baton-pass locks~\cite{Andrews89} between threads instead of unconditionally releasing locks is an example of barging prevention.
1253
1254
1255\section{Monitor}
1256\label{s:Monitor}
1257
1258A \textbf{monitor} is a set of routines that ensure mutual exclusion when accessing shared state.
1259More 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).
1260The strong association with the call/return paradigm eases programmability, readability and maintainability, at a slight cost in flexibility and efficiency.
1261
1262Note, 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.
1263Copying 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.
1264Copying 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.
1265As for coroutines/tasks, a non-copyable (@dtype@) trait is used to capture this requirement, so all locks/monitors must be passed by reference (pointer).
1266\begin{cfa}
1267trait is_monitor( `dtype` T ) {
1268        monitor_desc * get_monitor( T & );
1269        void ^?{}( T & mutex );
1270};
1271\end{cfa}
1272
1273
1274\subsection{Mutex Acquisition}
1275\label{s:MutexAcquisition}
1276
1277While correctness implicitly implies a monitor's mutual exclusion is acquired and released, there are implementation options about when and where the locking/unlocking occurs.
1278(Much of this discussion also applies to basic locks.)
1279For example, a monitor may need to be passed through multiple helper routines before it becomes necessary to acquire the monitor mutual-exclusion.
1280\begin{cfa}[morekeywords=nomutex]
1281monitor Aint { int cnt; };                                      $\C{// atomic integer counter}$
1282void ?{}( Aint & `nomutex` this ) with( this ) { cnt = 0; } $\C{// constructor}$
1283int ?=?( Aint & `mutex`$\(_{opt}\)$ lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions}$
1284void ?{}( int & this, Aint & `mutex`$\(_{opt}\)$ v ) { this = v.cnt; }
1285int ?=?( int & lhs, Aint & `mutex`$\(_{opt}\)$ rhs ) with( rhs ) { lhs = cnt; }
1286int ++?( Aint & `mutex`$\(_{opt}\)$ this ) with( this ) { return ++cnt; } $\C{// increment}$
1287\end{cfa}
1288The @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.
1289(While a constructor may publish its address into a global variable, doing so generates a race-condition.)
1290The conversion operators for initializing and assigning with a normal integer only need @mutex@, if reading/writing the implementation type is not atomic.
1291Finally, 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.
1292
1293The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics, which is similar to the \CC template @std::atomic@.
1294\begin{cfa}
1295Aint x, y, z;
1296++x; ++y; ++z;                                                          $\C{// safe increment by multiple threads}$
1297x = 2; y = 2; z = 2;                                            $\C{// conversions}$
1298int i = x, j = y, k = z;
1299i = x; j = y; k = z;
1300\end{cfa}
1301
1302For maximum usability, monitors have \newterm{multi-acquire} semantics allowing a thread to acquire it multiple times without deadlock.
1303For example, atomically printing the contents of a binary tree:
1304\begin{cfa}
1305monitor Tree {
1306        Tree * left, right;
1307        // value
1308};
1309void print( Tree & mutex tree ) {                       $\C{// prefix traversal}$
1310        // write value
1311        print( tree->left );                                    $\C{// multiply acquire monitor lock on each recursion}$
1312        print( tree->right );
1313}
1314\end{cfa}
1315
1316Mandatory monitor qualifiers have the benefit of being self-documenting, but requiring both @mutex@ and \lstinline[morekeywords=nomutex]@nomutex@ for all monitor parameter is redundant.
1317Instead, one of qualifier semantics can be the default, and the other required.
1318For example, assume the safe @mutex@ qualifier for all monitor parameters because assuming \lstinline[morekeywords=nomutex]@nomutex@ may cause subtle errors.
1319On the other hand, assuming \lstinline[morekeywords=nomutex]@nomutex@ is the \emph{normal} parameter behaviour, stating explicitly \emph{this parameter is not special}.
1320Providing a default qualifier implies knowing whether a parameter is a monitor.
1321Since \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.
1322For 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@.
1323
1324The next semantic decision is establishing which parameter \emph{types} may be qualified with @mutex@.
1325Given:
1326\begin{cfa}
1327monitor M { ... }
1328int f1( M & mutex m );
1329int f2( M * mutex m );
1330int f3( M * mutex m[] );
1331int f4( stack( M * ) & mutex m );
1332\end{cfa}
1333the issue is that some of these parameter types are composed of multiple objects.
1334For @f1@, there is only a single parameter object.
1335Adding indirection in @f2@ still identifies a single object.
1336However, the matrix in @f3@ introduces multiple objects.
1337While shown shortly, multiple acquisition is possible;
1338however array lengths are often unknown in C.
1339This issue is exacerbated in @f4@, where the data structure must be safely traversed to acquire all of its elements.
1340
1341To make the issue tractable, \CFA only acquires one monitor per parameter with at most one level of indirection.
1342However, the C type-system has an ambiguity with respects to arrays.
1343Is the argument for @f2@ a single object or an array of objects?
1344If it is an array, only the first element of the array is acquired, which seems unsafe;
1345hence, @mutex@ is disallowed for array parameters.
1346\begin{cfa}
1347int f1( M & mutex m );                                          $\C{// allowed: recommended case}$
1348int f2( M * mutex m );                                          $\C{// disallowed: could be an array}$
1349int f3( M mutex m[$\,$] );                                      $\C{// disallowed: array length unknown}$
1350int f4( M ** mutex m );                                         $\C{// disallowed: could be an array}$
1351int f5( M * mutex m[$\,$] );                            $\C{// disallowed: array length unknown}$
1352\end{cfa}
1353% Note, not all array routines have distinct types: @f2@ and @f3@ have the same type, as do @f4@ and @f5@.
1354% However, even if the code generation could tell the difference, the extra information is still not sufficient to extend meaningfully the monitor call semantic.
1355
1356For object-oriented monitors, calling a mutex member \emph{implicitly} acquires mutual exclusion of the receiver object, @`rec`.foo(...)@.
1357\CFA has no receiver, and hence, must use an explicit mechanism to specify which object has mutual exclusion acquired.
1358A positive consequence of this design decision is the ability to support multi-monitor routines.
1359\begin{cfa}
1360int f( M & mutex x, M & mutex y );              $\C{// multiple monitor parameter of any type}$
1361M m1, m2;
1362f( m1, m2 );
1363\end{cfa}
1364(While object-oriented monitors can be extended with a mutex qualifier for multiple-monitor members, no prior example of this feature could be found.)
1365In practice, writing multi-locking routines that do not deadlock is tricky.
1366Having language support for such a feature is therefore a significant asset for \CFA.
1367
1368The capability to acquire multiple locks before entering a critical section is called \newterm{bulk acquire}.
1369In the previous example, \CFA guarantees the order of acquisition is consistent across calls to different routines using the same monitors as arguments.
1370This consistent ordering means acquiring multiple monitors is safe from deadlock.
1371However, users can force the acquiring order.
1372For example, notice the use of @mutex@/\lstinline[morekeywords=nomutex]@nomutex@ and how this affects the acquiring order:
1373\begin{cfa}
1374void foo( M & mutex m1, M & mutex m2 );         $\C{// acquire m1 and m2}$
1375void bar( M & mutex m1, M & /* nomutex */ m2 ) { $\C{// acquire m1}$
1376        ... foo( m1, m2 ); ...                                  $\C{// acquire m2}$
1377}
1378void baz( M & /* nomutex */ m1, M & mutex m2 ) { $\C{// acquire m2}$
1379        ... foo( m1, m2 ); ...                                  $\C{// acquire m1}$
1380}
1381\end{cfa}
1382The multi-acquire semantics allows @bar@ or @baz@ to acquire a monitor lock and reacquire it in @foo@.
1383In the calls to @bar@ and @baz@, the monitors are acquired in opposite order.
1384
1385However, such use leads to lock acquiring order problems resulting in deadlock~\cite{Lister77}, where detecting it requires dynamic tracking of monitor calls, and dealing with it requires rollback semantics~\cite{Dice10}.
1386In \CFA, safety is guaranteed by using bulk acquire of all monitors to shared objects, whereas other monitor systems provide no aid.
1387While \CFA provides only a partial solution, the \CFA partial solution handles many useful cases.
1388\begin{cfa}
1389monitor Bank { ... };
1390void deposit( Bank & `mutex` b, int deposit );
1391void transfer( Bank & `mutex` mybank, Bank & `mutex` yourbank, int me2you) {
1392        deposit( mybank, `-`me2you );                   $\C{// debit}$
1393        deposit( yourbank, me2you );                    $\C{// credit}$
1394}
1395\end{cfa}
1396This example shows a trivial solution to the bank-account transfer problem.
1397Without multi- and bulk acquire, the solution to this problem requires careful engineering.
1398
1399
1400\subsection{\protect\lstinline|mutex| statement} \label{mutex-stmt}
1401
1402The monitor call-semantics associate all locking semantics to routines.
1403Like Java, \CFA offers an alternative @mutex@ statement to reduce refactoring and naming.
1404\begin{cquote}
1405\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1406\begin{cfa}
1407monitor M {};
1408void foo( M & mutex m1, M & mutex m2 ) {
1409        // critical section
1410}
1411void bar( M & m1, M & m2 ) {
1412        foo( m1, m2 );
1413}
1414\end{cfa}
1415&
1416\begin{cfa}
1417
1418void bar( M & m1, M & m2 ) {
1419        mutex( m1, m2 ) {       // remove refactoring and naming
1420                // critical section
1421        }
1422}
1423
1424\end{cfa}
1425\\
1426\multicolumn{1}{c}{\textbf{routine call}} & \multicolumn{1}{c}{\lstinline@mutex@ \textbf{statement}}
1427\end{tabular}
1428\end{cquote}
1429
1430
1431\section{Scheduling}
1432\label{s:Scheduling}
1433
1434While monitor mutual-exclusion provides safe access to shared data, the monitor data may indicate that a thread accessing it cannot proceed.
1435For example, Figure~\ref{f:GenericBoundedBuffer} shows a bounded buffer that may be full/empty so produce/consumer threads must block.
1436Leaving the monitor and trying again (busy waiting) is impractical for high-level programming.
1437Monitors 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.
1438Synchronization 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.
1439\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.
1440
1441Figure~\ref{f:BBInt} shows a \CFA generic bounded-buffer with internal scheduling, where producers/consumers enter the monitor, see the buffer is full/empty, and block on an appropriate condition lock, @full@/@empty@.
1442The @wait@ routine atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the routine's parameter list.
1443The appropriate condition lock is signalled to unblock an opposite kind of thread after an element is inserted/removed from the buffer.
1444Signalling is unconditional, because signalling an empty condition lock does nothing.
1445
1446Signalling semantics cannot have the signaller and signalled thread in the monitor simultaneously, which means:
1447\begin{enumerate}
1448\item
1449The signalling thread returns immediately, and the signalled thread continues.
1450\item
1451The signalling thread continues and the signalled thread is marked for urgent unblocking at the next scheduling point (exit/wait).
1452\item
1453The signalling thread blocks but is marked for urgrent unblocking at the next scheduling point and the signalled thread continues.
1454\end{enumerate}
1455The first approach is too restrictive, as it precludes solving a reasonable class of problems, \eg dating service (see Figure~\ref{f:DatingService}).
1456\CFA supports the next two semantics as both are useful.
1457Finally, while it is common to store a @condition@ as a field of the monitor, in \CFA, a @condition@ variable can be created/stored independently.
1458Furthermore, 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.
1459
1460\begin{figure}
1461\centering
1462\newbox\myboxA
1463\begin{lrbox}{\myboxA}
1464\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1465forall( otype T ) { // distribute forall
1466        monitor Buffer {
1467                `condition` full, empty;
1468                int front, back, count;
1469                T elements[10];
1470        };
1471        void ?{}( Buffer(T) & buffer ) with(buffer) {
1472                [front, back, count] = 0;
1473        }
1474
1475        void insert( Buffer(T) & mutex buffer, T elem )
1476                                with(buffer) {
1477                if ( count == 10 ) `wait( empty )`;
1478                // insert elem into buffer
1479                `signal( full )`;
1480        }
1481        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1482                if ( count == 0 ) `wait( full )`;
1483                // remove elem from buffer
1484                `signal( empty )`;
1485                return elem;
1486        }
1487}
1488\end{cfa}
1489\end{lrbox}
1490
1491\newbox\myboxB
1492\begin{lrbox}{\myboxB}
1493\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1494forall( otype T ) { // distribute forall
1495        monitor Buffer {
1496
1497                int front, back, count;
1498                T elements[10];
1499        };
1500        void ?{}( Buffer(T) & buffer ) with(buffer) {
1501                [front, back, count] = 0;
1502        }
1503        T remove( Buffer(T) & mutex buffer ); // forward
1504        void insert( Buffer(T) & mutex buffer, T elem )
1505                                with(buffer) {
1506                if ( count == 10 ) `waitfor( remove, buffer )`;
1507                // insert elem into buffer
1508
1509        }
1510        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1511                if ( count == 0 ) `waitfor( insert, buffer )`;
1512                // remove elem from buffer
1513
1514                return elem;
1515        }
1516}
1517\end{cfa}
1518\end{lrbox}
1519
1520\subfloat[Internal Scheduling]{\label{f:BBInt}\usebox\myboxA}
1521%\qquad
1522\subfloat[External Scheduling]{\label{f:BBExt}\usebox\myboxB}
1523\caption{Generic Bounded-Buffer}
1524\label{f:GenericBoundedBuffer}
1525\end{figure}
1526
1527Figure~\ref{f:BBExt} shows a \CFA generic bounded-buffer with external scheduling, where producers/consumers detecting a full/empty buffer block and prevent more producers/consumers from entering the monitor until the buffer has a free/empty slot.
1528External 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.
1529If 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.
1530Threads 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.
1531% External scheduling is more constrained and explicit, which helps programmers reduce the non-deterministic nature of concurrency.
1532External scheduling allows users to wait for events from other threads without concern of unrelated events occurring.
1533The mechnaism can be done in terms of control flow, \eg Ada @accept@ or \uC @_Accept@, or in terms of data, \eg Go channels.
1534While both mechanisms have strengths and weaknesses, this project uses a control-flow mechanism to stay consistent with other language semantics.
1535Two 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}).
1536
1537For internal scheduling, non-blocking signalling (as in the producer/consumer example) is used when the signaller is providing the cooperation for a waiting thread;
1538the 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.
1539The waiter unblocks next from the urgent queue, uses/takes the state, and exits the monitor.
1540Blocking signalling is the reverse, where the waiter is providing the cooperation for the signalling thread;
1541the 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.
1542The waiter changes state and exits the monitor, and the signaller unblocks next from the urgent queue to use/take the state.
1543
1544Figure~\ref{f:DatingService} shows a dating service demonstrating non-blocking and blocking signalling.
1545The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
1546A thread blocks until an appropriate partner arrives.
1547The complexity is exchanging phone numbers in the monitor because of the mutual-exclusion property.
1548For signal 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.
1549For signal-block scheduling, the implicit urgent-queue replaces the explict @exchange@-condition and @signal_block@ puts the finding thread on the urgent condition and unblocks the matcher.
1550
1551The 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;
1552as well, an arriving thread may not find a partner and must wait, which requires a condition variable, and condition variables imply internal scheduling.
1553
1554\begin{figure}
1555\centering
1556\newbox\myboxA
1557\begin{lrbox}{\myboxA}
1558\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1559enum { CCodes = 20 };
1560monitor DS {
1561        int GirlPhNo, BoyPhNo;
1562        condition Girls[CCodes], Boys[CCodes];
1563        condition exchange;
1564};
1565int girl( DS & mutex ds, int phNo, int ccode ) {
1566        if ( is_empty( Boys[ccode] ) ) {
1567                wait( Girls[ccode] );
1568                GirlPhNo = phNo;
1569                `exchange.signal();`
1570        } else {
1571                GirlPhNo = phNo;
1572                `signal( Boys[ccode] );`
1573                `exchange.wait();`
1574        } // if
1575        return BoyPhNo;
1576}
1577int boy( DS & mutex ds, int phNo, int ccode ) {
1578        // as above with boy/girl interchanged
1579}
1580\end{cfa}
1581\end{lrbox}
1582
1583\newbox\myboxB
1584\begin{lrbox}{\myboxB}
1585\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1586
1587monitor DS {
1588        int GirlPhNo, BoyPhNo;
1589        condition Girls[CCodes], Boys[CCodes];
1590
1591};
1592int girl( DS & mutex ds, int phNo, int ccode ) {
1593        if ( is_empty( Boys[ccode] ) ) { // no compatible
1594                wait( Girls[ccode] ); // wait for boy
1595                GirlPhNo = phNo; // make phone number available
1596
1597        } else {
1598                GirlPhNo = phNo; // make phone number available
1599                `signal_block( Boys[ccode] );` // restart boy
1600
1601        } // if
1602        return BoyPhNo;
1603}
1604int boy( DS & mutex ds, int phNo, int ccode ) {
1605        // as above with boy/girl interchanged
1606}
1607\end{cfa}
1608\end{lrbox}
1609
1610\subfloat[\lstinline@signal@]{\label{f:DatingSignal}\usebox\myboxA}
1611\qquad
1612\subfloat[\lstinline@signal_block@]{\label{f:DatingSignalBlock}\usebox\myboxB}
1613\caption{Dating service. }
1614\label{f:DatingService}
1615\end{figure}
1616
1617Both internal and external scheduling extend to multiple monitors in a natural way.
1618\begin{cquote}
1619\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1620\begin{cfa}
1621monitor M { `condition e`; ... };
1622void foo( M & mutex m1, M & mutex m2 ) {
1623        ... wait( `e` ); ...   // wait( e, m1, m2 )
1624        ... wait( `e, m1` ); ...
1625        ... wait( `e, m2` ); ...
1626}
1627\end{cfa}
1628&
1629\begin{cfa}
1630void rtn$\(_1\)$( M & mutex m1, M & mutex m2 );
1631void rtn$\(_2\)$( M & mutex m1 );
1632void bar( M & mutex m1, M & mutex m2 ) {
1633        ... waitfor( `rtn` ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\), m1, m2 )}$
1634        ... waitfor( `rtn, m1` ); ... // $\LstCommentStyle{waitfor( rtn\(_2\), m1 )}$
1635}
1636\end{cfa}
1637\end{tabular}
1638\end{cquote}
1639For @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 )@.
1640To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @wait( e, m1 )@.
1641Wait statically verifies the released monitors are the acquired mutex-parameters so unconditional release is safe.
1642Finally, a signaller,
1643\begin{cfa}
1644void baz( M & mutex m1, M & mutex m2 ) {
1645        ... signal( e ); ...
1646}
1647\end{cfa}
1648must have acquired monitor locks that are greater than or equal to the number of locks for the waiting thread signalled from the condition queue.
1649
1650Similarly, 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 )@.
1651To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn, m1 )@.
1652@waitfor@ statically verifies the released monitors are the same as the acquired mutex-parameters of the given routine or routine pointer.
1653To statically verify the released monitors match with the accepted routine's mutex parameters, the routine (pointer) prototype must be accessible.
1654
1655When an overloaded routine appears in an @waitfor@ statement, calls to any routine with that name are accepted.
1656The rationale is that members with the same name should perform a similar function, and therefore, all should be eligible to accept a call.
1657As always, overloaded routines can be disambiguated using a cast:
1658\begin{cfa}
1659void rtn( M & mutex m );
1660`int` rtn( M & mutex m );
1661waitfor( (`int` (*)( M & mutex ))rtn, m1, m2 );
1662\end{cfa}
1663
1664Given the ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock.
1665\begin{cfa}
1666void foo( M & mutex m1, M & mutex m2 ) {
1667        ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired )}$
1668void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2 )}$
1669        ... signal( `e` ); ...
1670\end{cfa}
1671The @wait@ only releases @m1@ so the signalling thread cannot acquire both @m1@ and @m2@ to  enter @bar@ to get to the @signal@.
1672While deadlock issues can occur with multiple/nesting acquisition, this issue results from the fact that locks, and by extension monitors, are not perfectly composable.
1673
1674Finally, an important aspect of monitor implementation is barging, \ie can calling threads barge ahead of signalled threads?
1675If 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).
1676\begin{quote}
1677However, 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.
1678It 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}
1679\end{quote}
1680\CFA scheduling \emph{precludes} barging, which simplifies synchronization among threads in the monitor and increases correctness.
1681For example, there are no loops in either bounded buffer solution in Figure~\ref{f:GenericBoundedBuffer}.
1682Supporting 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.
1683Furthermore, \CFA concurrency has no superious wakeup~\cite[\S~9]{Buhr05a}, which eliminates an implict form of barging.
1684
1685
1686\subsection{Barging Prevention}
1687
1688Figure~\ref{f:BargingPrevention} shows \CFA code where bulk acquire adds complexity to the internal-signalling semantics.
1689The complexity begins at the end of the inner @mutex@ statement, where the semantics of internal scheduling need to be extended for multiple monitors.
1690The problem is that bulk acquire is used in the inner @mutex@ statement where one of the monitors is already acquired.
1691When 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.
1692However, both the signalling and waiting thread W1 still need monitor @m1@.
1693
1694\begin{figure}
1695\newbox\myboxA
1696\begin{lrbox}{\myboxA}
1697\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1698monitor M m1, m2;
1699condition c;
1700mutex( m1 ) { // $\LstCommentStyle{\color{red}outer}$
1701        ...
1702        mutex( m1, m2 ) { // $\LstCommentStyle{\color{red}inner}$
1703                ... `signal( c )`; ...
1704                // m1, m2 acquired
1705        } // $\LstCommentStyle{\color{red}release m2}$
1706        // m1 acquired
1707} // release m1
1708\end{cfa}
1709\end{lrbox}
1710
1711\newbox\myboxB
1712\begin{lrbox}{\myboxB}
1713\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1714
1715
1716mutex( m1 ) {
1717        ...
1718        mutex( m1, m2 ) {
1719                ... `wait( c )`; // block and release m1, m2
1720                // m1, m2 acquired
1721        } // $\LstCommentStyle{\color{red}release m2}$
1722        // m1 acquired
1723} // release m1
1724\end{cfa}
1725\end{lrbox}
1726
1727\newbox\myboxC
1728\begin{lrbox}{\myboxC}
1729\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1730
1731
1732mutex( m2 ) {
1733        ... `wait( c )`; ...
1734        // m2 acquired
1735} // $\LstCommentStyle{\color{red}release m2}$
1736
1737
1738
1739
1740\end{cfa}
1741\end{lrbox}
1742
1743\begin{cquote}
1744\subfloat[Signalling Thread]{\label{f:SignallingThread}\usebox\myboxA}
1745\hspace{2\parindentlnth}
1746\subfloat[Waiting Thread (W1)]{\label{f:WaitingThread}\usebox\myboxB}
1747\hspace{2\parindentlnth}
1748\subfloat[Waiting Thread (W2)]{\label{f:OtherWaitingThread}\usebox\myboxC}
1749\end{cquote}
1750\caption{Barging Prevention}
1751\label{f:BargingPrevention}
1752\end{figure}
1753
1754One 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.
1755However, Figure~\ref{f:OtherWaitingThread} shows this solution is complex depending on other waiters, resulting is choices when the signaller finishes the inner mutex-statement.
1756The 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@.
1757In 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.
1758Furthermore, there is an execution sequence where the signaller always finds waiter W2, and hence, waiter W1 starves.
1759
1760While a number of approaches were examined~\cite[\S~4.3]{Delisle18}, the solution chosen for \CFA is a novel techique called \newterm{partial signalling}.
1761Signalled threads are moved to the urgent queue and the waiter at the front defines the set of monitors necessary for it to unblock.
1762Partial signalling transfers ownership of monitors to the front waiter.
1763When the signaller thread exits or waits in the monitor the front waiter is unblocked if all its monitors are released.
1764This 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.
1765
1766
1767\subsection{Loose Object Definitions}
1768\label{s:LooseObjectDefinitions}
1769
1770In an object-oriented programming-language, a class includes an exhaustive list of operations.
1771However, new members can be added via static inheritance or dynaic members, \eg JavaScript~\cite{JavaScript}.
1772Similarly, monitor routines can be added at any time in \CFA, making it less clear for programmers and more difficult to implement.
1773\begin{cfa}
1774monitor M {};
1775void `f`( M & mutex m );
1776void g( M & mutex m ) { waitfor( `f` ); }       $\C{// clear which f}$
1777void `f`( M & mutex m, int );                           $\C{// different f}$
1778void h( M & mutex m ) { waitfor( `f` ); }       $\C{// unclear which f}$
1779\end{cfa}
1780Hence, the cfa-code for the entering a monitor looks like:
1781\begin{cfa}
1782if ( $\textrm{\textit{monitor is free}}$ ) $\LstCommentStyle{// \color{red}enter}$
1783else if ( $\textrm{\textit{already own monitor}}$ ) $\LstCommentStyle{// \color{red}continue}$
1784else if ( $\textrm{\textit{monitor accepts me}}$ ) $\LstCommentStyle{// \color{red}enter}$
1785else $\LstCommentStyle{// \color{red}block}$
1786\end{cfa}
1787For the first two conditions, it is easy to implement a check that can evaluate the condition in a few instructions.
1788However, a fast check for \emph{monitor accepts me} is much harder to implement depending on the constraints put on the monitors.
1789Figure~\ref{fig:ClassicalMonitor} shows monitors are often expressed as an entry (calling) queue, some acceptor queues, and an urgent stack/queue.
1790
1791\begin{figure}
1792\centering
1793\subfloat[Classical monitor] {
1794\label{fig:ClassicalMonitor}
1795{\resizebox{0.45\textwidth}{!}{\input{monitor.pstex_t}}}
1796}% subfloat
1797\quad
1798\subfloat[Bulk acquire monitor] {
1799\label{fig:BulkMonitor}
1800{\resizebox{0.45\textwidth}{!}{\input{ext_monitor.pstex_t}}}
1801}% subfloat
1802\caption{Monitor Implementation}
1803\label{f:MonitorImplementation}
1804\end{figure}
1805
1806For 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.
1807This approach requires a unique dense ordering of routines with a small upper-bound and the ordering must be consistent across translation units.
1808For 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.
1809
1810Figure~\ref{fig:BulkMonitor} shows the \CFA monitor implementation.
1811The mutex routine called is associated with each thread on the entry queue, while a list of acceptable routines is kept separately.
1812The accepted list is a variable-sized array of accepted routine pointers, so the single instruction bitmask comparison is replaced by dereferencing a pointer followed by a (usually short) linear search.
1813
1814
1815\subsection{Multi-Monitor Scheduling}
1816\label{s:Multi-MonitorScheduling}
1817
1818External scheduling, like internal scheduling, becomes significantly more complex for multi-monitor semantics.
1819Even in the simplest case, new semantics needs to be established.
1820\begin{cfa}
1821monitor M {};
1822void f( M & mutex m1 );
1823void g( M & mutex m1, M & mutex m2 ) {
1824        waitfor( f );                                                   $\C{\color{red}// pass m1 or m2 to f?}$
1825}
1826\end{cfa}
1827The solution is for the programmer to disambiguate:
1828\begin{cfa}
1829        waitfor( f, m2 );                                               $\C{\color{red}// wait for call to f with argument m2}$
1830\end{cfa}
1831Routine @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@.
1832This behaviour can be extended to the multi-monitor @waitfor@ statement.
1833\begin{cfa}
1834monitor M {};
1835void f( M & mutex m1, M & mutex m2 );
1836void g( M & mutex m1, M & mutex m2 ) {
1837        waitfor( f, m1, m2 );                                   $\C{\color{red}// wait for call to f with arguments m1 and m2}$
1838}
1839\end{cfa}
1840Again, the set of monitors passed to the @waitfor@ statement must be entirely contained in the set of monitors already acquired by the accepting routine.
1841
1842Note, for internal and external scheduling with multiple monitors, a signalling or accepting thread must match exactly, \ie partial matching results in waiting.
1843\begin{cquote}
1844\lstDeleteShortInline@%
1845\begin{tabular}{@{}l@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
1846\begin{cfa}
1847monitor M1 {} m11, m12;
1848monitor M2 {} m2;
1849condition c;
1850void f( M1 & mutex m1, M2 & mutex m2 ) {
1851        signal( c );
1852}
1853void g( M1 & mutex m1, M2 & mutex m2 ) {
1854        wait( c );
1855}
1856g( `m11`, m2 ); // block on accept
1857f( `m12`, m2 ); // cannot fulfil
1858\end{cfa}
1859&
1860\begin{cfa}
1861monitor M1 {} m11, m12;
1862monitor M2 {} m2;
1863
1864void f( M1 & mutex m1, M2 & mutex m2 ) {
1865
1866}
1867void g( M1 & mutex m1, M2 & mutex m2 ) {
1868        waitfor( f, m1, m2 );
1869}
1870g( `m11`, m2 ); // block on accept
1871f( `m12`, m2 ); // cannot fulfil
1872\end{cfa}
1873\end{tabular}
1874\lstMakeShortInline@%
1875\end{cquote}
1876For both internal and external scheduling, the set of monitors is disjoint so unblocking is impossible.
1877
1878
1879\subsection{Extended \protect\lstinline@waitfor@}
1880
1881The 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.
1882\begin{cfa}
1883`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1884        waitfor( $\emph{mutex-member-name}$ )
1885                $\emph{statement}$                                      $\C{// action after call}$
1886`or` `when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
1887        waitfor( $\emph{mutex-member-name}$ )
1888                $\emph{statement}$                                      $\C{// action after call}$
1889`or`    ...                                                                     $\C{// list of waitfor clauses}$
1890`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1891        `timeout`                                                               $\C{// optional terminating timeout clause}$
1892                $\emph{statement}$                                      $\C{// action after timeout}$
1893`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1894        `else`                                                                  $\C{// optional terminating clause}$
1895                $\emph{statement}$                                      $\C{// action when no immediate calls}$
1896\end{cfa}
1897For a @waitfor@ clause to be executed, its @when@ must be true and an outstanding call to its corresponding member(s) must exist.
1898The \emph{conditional-expression} of a @when@ may call a routine, but the routine must not block or context switch.
1899If there are several mutex calls that can be accepted, selection occurs top-to-bottom in the @waitfor@ clauses versus non-deterministically.
1900If 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.
1901If 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.
1902Hence, the terminating @else@ clause allows a conditional attempt to accept a call without blocking.
1903If 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.
1904In all cases, the statement following is executed \emph{after} a clause is executed to know which of the clauses executed.
1905
1906Note, a group of conditional @waitfor@ clauses is \emph{not} the same as a group of @if@ statements, e.g.:
1907\begin{cfa}
1908if ( C1 ) waitfor( mem1 );                       when ( C1 ) waitfor( mem1 );
1909else if ( C2 ) waitfor( mem2 );         or when ( C2 ) waitfor( mem2 );
1910\end{cfa}
1911The left example accepts only @mem1@ if @C1@ is true or only @mem2@ if @C2@ is true.
1912The right example accepts either @mem1@ or @mem2@ if @C1@ and @C2@ are true.
1913
1914An interesting use of @waitfor@ is accepting the @mutex@ destructor to know when an object is deallocated.
1915\begin{cfa}
1916void insert( Buffer(T) & mutex buffer, T elem ) with( buffer ) {
1917        if ( count == BufferSize )
1918                waitfor( remove, buffer ) {
1919                        elements[back] = elem;
1920                        back = ( back + 1 ) % BufferSize;
1921                        count += 1;
1922                } or `waitfor( ^?{}, buffer )` throw insertFail;
1923}
1924\end{cfa}
1925When the buffer is deallocated, the current waiter is unblocked and informed, so it can perform an appropriate action.
1926However, the basic @waitfor@ semantics do not support this functionality, since using an object after its destructor is called is undefined.
1927Therefore, to make this useful capability work, the semantics for accepting the destructor is the same as @signal@, \ie the call to the destructor is placed on the urgent queue and the acceptor continues execution, which throws an exception to the acceptor and then the caller is unbloced from the urgent queue to deallocate the object.
1928Accepting the destructor is an idiomatic way to terminate a thread in \CFA.
1929
1930
1931\subsection{\protect\lstinline@mutex@ Threads}
1932
1933Threads in \CFA are monitors to allow direct communication among threads, \ie threads can have mutex routines that are called by other threads.
1934Hence, all monitor features are available when using threads.
1935Figure~\ref{f:pingpong} shows an example of two threads directly calling each other and accepting calls from each other in a cycle.
1936Note, both ping/pong threads are globally declared, @pi@/@po@, and hence, start (and possibly complete) before the program main starts.
1937
1938\begin{figure}
1939\lstDeleteShortInline@%
1940\begin{cquote}
1941\begin{cfa}
1942thread Ping {} pi;
1943thread Pong {} po;
1944void ping( Ping & mutex ) {}
1945void pong( Pong & mutex ) {}
1946int main() {}
1947\end{cfa}
1948\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1949\begin{cfa}
1950void main( Ping & pi ) {
1951        for ( int i = 0; i < 10; i += 1 ) {
1952                `waitfor( ping, pi );`
1953                `pong( po );`
1954        }
1955}
1956\end{cfa}
1957&
1958\begin{cfa}
1959void main( Pong & po ) {
1960        for ( int i = 0; i < 10; i += 1 ) {
1961                `ping( pi );`
1962                `waitfor( pong, po );`
1963        }
1964}
1965\end{cfa}
1966\end{tabular}
1967\lstMakeShortInline@%
1968\end{cquote}
1969\caption{Threads ping/pong using external scheduling}
1970\label{f:pingpong}
1971\end{figure}
1972
1973\section{Parallelism}
1974
1975Historically, computer performance was about processor speeds.
1976However, with heat dissipation being a direct consequence of speed increase, parallelism has become the new source for increased performance~\cite{Sutter05, Sutter05b}.
1977Now, high-performance applications must care about parallelism, which requires concurrency.
1978The lowest-level approach of parallelism is to use \newterm{kernel threads} in combination with semantics like @fork@, @join@, \etc.
1979However, kernel threads are better as an implementation tool because of complexity and higher cost.
1980Therefore, different abstractions are often layered onto kernel threads to simplify them, \eg pthreads.
1981
1982
1983\subsection{User Threads with Preemption}
1984
1985A direct improvement on kernel threads is user threads, \eg Erlang~\cite{Erlang} and \uC~\cite{uC++book}.
1986This approach provides an interface that matches the language paradigms, more control over concurrency by the language runtime, and an abstract (and portable) interface to the underlying kernel threads across operating systems.
1987In many cases, user threads can be used on a much larger scale (100,000 threads).
1988Like kernel threads, user threads support preemption, which maximizes nondeterminism, but introduces the same concurrency errors: race, livelock, starvation, and deadlock.
1989\CFA adopts user-threads as they represent the truest realization of concurrency and can build any other concurrency approach, \eg thread pools and actors~\cite{Actors}.
1990
1991
1992\subsection{User Threads without Preemption (Fiber)}
1993\label{s:fibers}
1994
1995A variant of user thread is \newterm{fibers}, which removes preemption, \eg Go~\cite{Go}.
1996Like 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.
1997However, preemption is necessary for concurrency that relies on spinning, so there are a class of problems that cannot be programmed without preemption.
1998
1999
2000\subsection{Thread Pools}
2001
2002In contrast to direct threading is indirect \newterm{thread pools}, where small jobs (work units) are insert into a work pool for execution.
2003If the jobs are dependent, \ie interact, there is an implicit/explicit dependency graph that ties them together.
2004While removing direct concurrency, and hence the amount of context switching, thread pools significantly limit the interaction that can occur among jobs.
2005Indeed, jobs should not block because that also block the underlying thread, which effectively means the CPU utilization, and therefore throughput, suffers.
2006While 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.
2007As well, concurrency errors return, which threads pools are suppose to mitigate.
2008Intel's TBB library~\cite{TBB} is the gold standard for thread pools.
2009
2010
2011\section{\protect\CFA Runtime Structure}
2012
2013Figure~\ref{f:RunTimeStructure} illustrates the runtime structure of a \CFA program.
2014In addition to the new kinds of objects introduced by \CFA, there are two more runtime entities used to control parallel execution: cluster and (virtual) processor.
2015An executing thread is illustrated by its containment in a processor.
2016
2017\begin{figure}
2018\centering
2019\input{RunTimeStructure}
2020\caption{\CFA Runtime Structure}
2021\label{f:RunTimeStructure}
2022\end{figure}
2023
2024
2025\subsection{Cluster}
2026\label{s:RuntimeStructureCluster}
2027
2028A \newterm{cluster} is a collection of threads and virtual processors (abstract kernel-thread) that execute the threads (like a virtual machine).
2029The purpose of a cluster is to control the amount of parallelism that is possible among threads, plus scheduling and other execution defaults.
2030The default cluster-scheduler is single-queue multi-server, which provides automatic load-balancing of threads on processors.
2031However, the scheduler is pluggable, supporting alternative schedulers.
2032If several clusters exist, both threads and virtual processors, can be explicitly migrated from one cluster to another.
2033No automatic load balancing among clusters is performed by \CFA.
2034
2035When a \CFA program begins execution, it creates a user cluster with a single processor and a special processor to handle preemption that does not execute user threads.
2036The user cluster is created to contain the application user-threads.
2037Having all threads execute on the one cluster often maximizes utilization of processors, which minimizes runtime.
2038However, because of limitations of the underlying operating system, heterogeneous hardware, or scheduling requirements (real-time), it is sometimes necessary to have multiple clusters.
2039
2040
2041\subsection{Virtual Processor}
2042\label{s:RuntimeStructureProcessor}
2043
2044A 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.
2045Programs may use more virtual processors than hardware processors.
2046On a multiprocessor, kernel threads are distributed across the hardware processors resulting in virtual processors executing in parallel.
2047(It is possible to use affinity to lock a virtual processor onto a particular hardware processor~\cite{affinityLinux, affinityWindows, affinityFreebsd, affinityNetbsd, affinityMacosx}, which is used when caching issues occur or for heterogeneous hardware processors.)
2048The \CFA runtime attempts to block unused processors and unblock processors as the system load increases;
2049balancing the workload with processors is difficult.
2050Preemption occurs on virtual processors rather than user threads, via operating-system interrupts.
2051Thus virtual processors execute user threads, where preemption frequency applies to a virtual processor, so preemption occurs randomly across the executed user threads.
2052Turning off preemption transforms user threads into fibers.
2053
2054
2055\subsection{Debug Kernel}
2056
2057There are two versions of the \CFA runtime kernel: debug and non-debug.
2058The 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.
2059After a program is debugged, the non-debugging version can be used to decrease space and increase performance.
2060
2061
2062\section{Implementation}
2063
2064Currently, \CFA has fixed-sized stacks, where the stack size can be set at coroutine/thread creation but with no subsequent growth.
2065Schemes exist for dynamic stack-growth, such as stack copying and chained stacks.
2066However, stack copying requires pointer adjustment to items on the stack, which is impossible without some form of garage collection.
2067As well, chained stacks require all modules be recompiled to use this feature, which breaks backward compatibility with existing C libraries.
2068In the long term, it is likely C libraries will migrate to stack chaining to support concurrency, at only a minimal cost to sequential programs.
2069Nevertheless, experience teaching \uC~\cite{CS343} shows fixed-sized stacks are rarely an issue in most concurrent programs.
2070
2071A primary implementation challenge is avoiding contention from dynamically allocating memory because of bulk acquire, \eg the internal-scheduling design is (almost) free of allocations.
2072All 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.
2073Furthermore, several bulk-acquire operations need a variable amount of memory.
2074This 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.
2075
2076In \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.
2077When a mutex call is made, pointers to the concerned monitors are aggregated into a variable-length array and sorted.
2078This array persists for the entire duration of the mutual-exclusion and its ordering reused extensively.
2079
2080To 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;
2081the corresponding registers are then restored for the other context.
2082Note, the instruction pointer is untouched since the context switch is always inside the same routine.
2083Unlike coroutines, threads do not context switch among each other;
2084they context switch to the cluster scheduler.
2085This method is a 2-step context-switch and provides a clear distinction between user and kernel code, where scheduling and other system operations happen.
2086The 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.
2087Experimental results (not presented) show the performance difference between these two approaches is virtually equivalent, because the 1-step performance is dominated by a lock instruction to prevent a race condition.
2088
2089All kernel threads (@pthreads@) created a stack.
2090Each \CFA virtual processor is implemented as a coroutine and these coroutines run directly on the kernel-thread stack, effectively stealing this stack.
2091The exception to this rule is the program main, \ie the initial kernel thread that is given to any program.
2092In 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.
2093
2094Finally, 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.
2095Because preemption frequency is usually long, 1 millisecond, performance cost is negligible.
2096Preemption is normally handled by setting a count-down timer on each virtual processor.
2097When 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.
2098Multiple signal handlers may be pending.
2099When 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.
2100The 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;
2101therefore, all virtual processors in a cluster need to have the same signal mask.
2102
2103However, on current UNIX systems:
2104\begin{quote}
2105A process-directed signal may be delivered to any one of the threads that does not currently have the signal blocked.
2106If more than one of the threads has the signal unblocked, then the kernel chooses an arbitrary thread to which to deliver the signal.
2107SIGNAL(7) - Linux Programmer's Manual
2108\end{quote}
2109Hence, 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).
2110To ensure each virtual processor receives its own preemption signals, a discrete-event simulation is run on a special virtual processor, and only it sets and receives timer events.
2111Virtual processors register an expiration time with the discrete-event simulator, which is inserted in sorted order.
2112The 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.
2113Processing a preemption event sends an \emph{internal} @SIGUSR1@ signal to the registered virtual processor, which is always delivered to that processor.
2114
2115
2116\section{Performance}
2117\label{results}
2118
2119To verify the implementation of the \CFA runtime, a series of microbenchmarks are performed comparing \CFA with other widely used programming languages with concurrency.
2120Table~\ref{t:machine} shows the specifications of the computer used to run the benchmarks, and the versions of the software used in the comparison.
2121
2122\begin{table}
2123\centering
2124\caption{Experiment environment}
2125\label{t:machine}
2126
2127\begin{tabular}{|l|r||l|r|}
2128\hline
2129Architecture            & x86\_64                               & NUMA node(s)  & 8 \\
2130\hline
2131CPU op-mode(s)          & 32-bit, 64-bit                & Model name    & AMD Opteron\texttrademark\ Processor 6380 \\
2132\hline
2133Byte Order                      & Little Endian                 & CPU Freq              & 2.5 GHz \\
2134\hline
2135CPU(s)                          & 64                                    & L1d cache     & 16 KiB \\
2136\hline
2137Thread(s) per core      & 2                                     & L1i cache     & 64 KiB \\
2138\hline
2139Core(s) per socket      & 8                                     & L2 cache              & 2048 KiB \\
2140\hline
2141Socket(s)                       & 4                                     & L3 cache              & 6144 KiB \\
2142\hline
2143\hline
2144Operating system        & Ubuntu 16.04.3 LTS    & Kernel                & Linux 4.4-97-generic \\
2145\hline
2146gcc                                     & 6.3                                   & \CFA                  & 1.0.0 \\
2147\hline
2148Java                            & OpenJDK-9                     & Go                    & 1.9.2 \\
2149\hline
2150\end{tabular}
2151\end{table}
2152
2153All benchmarks are run using the following harness:
2154\begin{cfa}
2155unsigned int N = 10_000_000;
2156#define BENCH( run ) Time before = getTimeNsec(); run; Duration result = (getTimeNsec() - before) / N;
2157\end{cfa}
2158The method used to get time is @clock_gettime( CLOCK_REALTIME )@.
2159Each 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.
2160All omitted tests for other languages are functionally identical to the shown \CFA test.
2161
2162
2163\paragraph{Context-Switching}
2164
2165In procedural programming, the cost of a routine call is important as modularization (refactoring) increases.
2166(In many cases, a compiler inlines routine calls to eliminate this cost.)
2167Similarly, when modularization extends to coroutines/tasks, the time for a context switch becomes a relevant factor.
2168The coroutine context-switch is 2-step using resume/suspend, \ie from resumer to suspender and from suspender to resumer.
2169The thread context switch is 2-step using yield, \ie enter and return from the runtime kernel.
2170Figure~\ref{f:ctx-switch} shows the code for coroutines/threads with all results in Table~\ref{tab:ctx-switch}.
2171The difference in performance between coroutine and thread context-switch is the cost of scheduling for threads, whereas coroutines are self-scheduling.
2172
2173\begin{figure}
2174\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2175
2176\newbox\myboxA
2177\begin{lrbox}{\myboxA}
2178\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2179coroutine C {} c;
2180void main( C & ) { for ( ;; ) { @suspend();@ } }
2181int main() {
2182        BENCH(
2183                for ( size_t i = 0; i < N; i += 1 ) { @resume( c );@ } )
2184        sout | result`ns | endl;
2185}
2186\end{cfa}
2187\end{lrbox}
2188
2189\newbox\myboxB
2190\begin{lrbox}{\myboxB}
2191\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2192
2193
2194int main() {
2195        BENCH(
2196                for ( size_t i = 0; i < N; i += 1 ) { @yield();@ } )
2197        sout | result`ns | endl;
2198}
2199\end{cfa}
2200\end{lrbox}
2201
2202\subfloat[Coroutine]{\usebox\myboxA}
2203\quad
2204\subfloat[Thread]{\usebox\myboxB}
2205\captionof{figure}{\CFA context-switch benchmark}
2206\label{f:ctx-switch}
2207
2208\centering
2209
2210\captionof{table}{Context switch comparison (nanoseconds)}
2211\label{tab:ctx-switch}
2212\bigskip
2213\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2214\cline{2-4}
2215\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2216\hline
2217Kernel Thread   & 241.5         & 243.86        & 5.08 \\
2218\CFA Coroutine  & 38            & 38            & 0    \\
2219\CFA Thread             & 103           & 102.96        & 2.96 \\
2220\uC Coroutine   & 46            & 45.86         & 0.35 \\
2221\uC Thread              & 98            & 99.11         & 1.42 \\
2222Goroutine               & 150           & 149.96        & 3.16 \\
2223Java Thread             & 289           & 290.68        & 8.72 \\
2224\hline
2225\end{tabular}
2226
2227\bigskip
2228\bigskip
2229
2230\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2231\begin{cfa}
2232monitor M {} m1/*, m2, m3, m4*/;
2233void __attribute__((noinline)) do_call( M & mutex m/*, m2, m3, m4*/ ) {}
2234int main() {
2235        BENCH( for( size_t i = 0; i < N; i += 1 ) { @do_call( m1/*, m2, m3, m4*/ );@ } )
2236        sout | result`ns | endl;
2237}
2238\end{cfa}
2239\captionof{figure}{\CFA acquire/release mutex benchmark}
2240\label{f:mutex}
2241
2242\centering
2243
2244\captionof{table}{Mutex comparison (nanoseconds)}
2245\label{tab:mutex}
2246\bigskip
2247
2248\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2249\cline{2-4}
2250\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2251\hline
2252C routine                                                       & 2                     & 2                     & 0    \\
2253FetchAdd + FetchSub                                     & 26            & 26            & 0    \\
2254Pthreads Mutex Lock                                     & 31            & 31.86         & 0.99 \\
2255\uC @monitor@ member routine            & 30            & 30            & 0    \\
2256\CFA @mutex@ routine, 1 argument        & 41            & 41.57         & 0.9  \\
2257\CFA @mutex@ routine, 2 argument        & 76            & 76.96         & 1.57 \\
2258\CFA @mutex@ routine, 4 argument        & 145           & 146.68        & 3.85 \\
2259Java synchronized routine                       & 27            & 28.57         & 2.6  \\
2260\hline
2261\end{tabular}
2262\end{figure}
2263
2264
2265\paragraph{Mutual-Exclusion}
2266
2267Mutual exclusion is measured by entering/leaving a critical section.
2268For monitors, entering and leaving a monitor routine is measured.
2269Figure~\ref{f:mutex} shows the code for \CFA with all results in Table~\ref{tab:mutex}.
2270To 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.
2271Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2272
2273
2274\paragraph{Internal Scheduling}
2275
2276Internal scheduling is measured by waiting on and signalling a condition variable.
2277Figure~\ref{f:int-sched} shows the code for \CFA, with results in Table~\ref{tab:int-sched}.
2278Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2279
2280\begin{figure}
2281\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2282\begin{cfa}
2283volatile int go = 0;
2284condition c;
2285monitor M {} m;
2286void __attribute__((noinline)) do_call( M & mutex a1 ) { signal( c ); }
2287thread T {};
2288void main( T & this ) {
2289        while ( go == 0 ) { yield(); }  // wait for other thread to start
2290        while ( go == 1 ) { @do_call( m );@ }
2291}
2292int  __attribute__((noinline)) do_wait( M & mutex m ) {
2293        go = 1; // continue other thread
2294        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @wait( c );@ } );
2295        go = 0; // stop other thread
2296        sout | result`ns | endl;
2297}
2298int main() {
2299        T t;
2300        do_wait( m );
2301}
2302\end{cfa}
2303\captionof{figure}{\CFA Internal scheduling benchmark}
2304\label{f:int-sched}
2305
2306\centering
2307\captionof{table}{Internal scheduling comparison (nanoseconds)}
2308\label{tab:int-sched}
2309\bigskip
2310
2311\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2312\cline{2-4}
2313\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2314\hline
2315Pthreads Condition Variable             & 5902.5        & 6093.29       & 714.78 \\
2316\uC @signal@                                    & 322           & 323           & 3.36   \\
2317\CFA @signal@, 1 @monitor@              & 352.5         & 353.11        & 3.66   \\
2318\CFA @signal@, 2 @monitor@              & 430           & 430.29        & 8.97   \\
2319\CFA @signal@, 4 @monitor@              & 594.5         & 606.57        & 18.33  \\
2320Java @notify@                                   & 13831.5       & 15698.21      & 4782.3 \\
2321\hline
2322\end{tabular}
2323\end{figure}
2324
2325
2326\paragraph{External Scheduling}
2327
2328External scheduling is measured by accepting a call using the @waitfor@ statement (@_Accept@ in \uC).
2329Figure~\ref{f:ext-sched} shows the code for \CFA, with results in Table~\ref{tab:ext-sched}.
2330Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2331
2332\begin{figure}
2333\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2334\begin{cfa}
2335volatile int go = 0;
2336monitor M {} m;
2337thread T {};
2338void __attribute__((noinline)) do_call( M & mutex ) {}
2339void main( T & ) {
2340        while ( go == 0 ) { yield(); }  // wait for other thread to start
2341        while ( go == 1 ) { @do_call( m );@ }
2342}
2343int __attribute__((noinline)) do_wait( M & mutex m ) {
2344        go = 1; // continue other thread
2345        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @waitfor( do_call, m );@ } )
2346        go = 0; // stop other thread
2347        sout | result`ns | endl;
2348}
2349int main() {
2350        T t;
2351        do_wait( m );
2352}
2353\end{cfa}
2354\captionof{figure}{\CFA external scheduling benchmark}
2355\label{f:ext-sched}
2356
2357\centering
2358
2359\captionof{table}{External scheduling comparison (nanoseconds)}
2360\label{tab:ext-sched}
2361\bigskip
2362\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2363\cline{2-4}
2364\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2365\hline
2366\uC @_Accept@                           & 350           & 350.61        & 3.11  \\
2367\CFA @waitfor@, 1 @monitor@     & 358.5         & 358.36        & 3.82  \\
2368\CFA @waitfor@, 2 @monitor@     & 422           & 426.79        & 7.95  \\
2369\CFA @waitfor@, 4 @monitor@     & 579.5         & 585.46        & 11.25 \\
2370\hline
2371\end{tabular}
2372
2373\bigskip
2374\medskip
2375
2376\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2377\begin{cfa}
2378thread MyThread {};
2379void main( MyThread & ) {}
2380int main() {
2381        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @MyThread m;@ } )
2382        sout | result`ns | endl;
2383}
2384\end{cfa}
2385\captionof{figure}{\CFA object creation benchmark}
2386\label{f:creation}
2387
2388\centering
2389
2390\captionof{table}{Creation comparison (nanoseconds)}
2391\label{tab:creation}
2392\bigskip
2393
2394\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2395\cline{2-4}
2396\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} & \multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2397\hline
2398Pthreads                                & 26996         & 26984.71      & 156.6  \\
2399\CFA Coroutine Lazy             & 6                     & 5.71          & 0.45   \\
2400\CFA Coroutine Eager    & 708           & 706.68        & 4.82   \\
2401\CFA Thread                             & 1173.5        & 1176.18       & 15.18  \\
2402\uC Coroutine                   & 109           & 107.46        & 1.74   \\
2403\uC Thread                              & 526           & 530.89        & 9.73   \\
2404Goroutine                               & 2520.5        & 2530.93       & 61.56  \\
2405Java Thread                             & 91114.5       & 92272.79      & 961.58 \\
2406\hline
2407\end{tabular}
2408\end{figure}
2409
2410
2411\paragraph{Object Creation}
2412
2413Object creation is measured by creating/deleting the specific kind of concurrent object.
2414Figure~\ref{f:creation} shows the code for \CFA, with results in Table~\ref{tab:creation}.
2415The 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.
2416
2417
2418\section{Conclusion}
2419
2420This paper demonstrate a concurrency API that is simple, efficient, and able to build higher-level concurrency features.
2421The 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.
2422The M:N model is judged to be efficient and provide greater flexibility than a 1:1 threading model.
2423High-level objects (monitor/task) are the core mechanism for mutual exclusion and synchronization.
2424A novel aspect is allowing multiple mutex-objects to be accessed simultaneously reducing the potential for deadlock for this complex scenario.
2425These concepts and the entire \CFA runtime-system are written in the \CFA language, demonstrating the expressiveness of the \CFA language.
2426Performance 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.
2427C 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.
2428
2429
2430\section{Future Work}
2431
2432While concurrency in \CFA has a strong start, development is still underway and there are missing features.
2433
2434\paragraph{Flexible Scheduling}
2435\label{futur:sched}
2436
2437An important part of concurrency is scheduling.
2438Different scheduling algorithms can affect performance (both in terms of average and variation).
2439However, no single scheduler is optimal for all workloads and therefore there is value in being able to change the scheduler for given programs.
2440One solution is to offer various tweaking options, allowing the scheduler to be adjusted to the requirements of the workload.
2441However, to be truly flexible, it is necessary to have a pluggable scheduler.
2442Currently, 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.
2443
2444\paragraph{Non-Blocking I/O}
2445\label{futur:nbio}
2446
2447Many modern workloads are not bound by computation but IO operations, a common case being web servers and XaaS~\cite{XaaS} (anything as a service).
2448These types of workloads require significant engineering to amortizing costs of blocking IO-operations.
2449At 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.
2450Current 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.
2451However, these solutions lead to code that is hard create, read and maintain.
2452A 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.
2453A non-blocking I/O library is currently under development for \CFA.
2454
2455\paragraph{Other Concurrency Tools}
2456\label{futur:tools}
2457
2458While monitors offer a flexible and powerful concurrent for \CFA, other concurrency tools are also necessary for a complete multi-paradigm concurrency package.
2459Examples of such tools can include futures and promises~\cite{promises}, executors and actors.
2460These additional features are useful when monitors offer a level of abstraction that is inadequate for certain tasks.
2461
2462\paragraph{Implicit Threading}
2463\label{futur:implcit}
2464
2465Basic 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.
2466This type of concurrency can be achieved both at the language level and at the library level.
2467The canonical example of implicit concurrency is concurrent nested @for@ loops, which are amenable to divide and conquer algorithms~\cite{uC++book}.
2468The \CFA language features should make it possible to develop a reasonable number of implicit concurrency mechanism to solve basic HPC data-concurrency problems.
2469However, implicit concurrency is a restrictive solution and has its limitations, so it can never replace explicit concurrent programming.
2470
2471
2472\section{Acknowledgements}
2473
2474The authors would like to recognize the design assistance of Aaron Moss, Rob Schluntz and Andrew Beach on the features described in this paper.
2475Funding 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.
2476
2477{%
2478\fontsize{9bp}{12bp}\selectfont%
2479\bibliography{pl,local}
2480}%
2481
2482\end{document}
2483
2484% Local Variables: %
2485% tab-width: 4 %
2486% fill-column: 120 %
2487% compile-command: "make" %
2488% End: %
Note: See TracBrowser for help on using the repository browser.