source: doc/papers/concurrency/Paper.tex @ 45af7e1

aaron-thesisarm-ehcleanup-dtorsenumforall-pointer-decayjacob/cs343-translationjenkins-sandboxnew-astnew-ast-unique-exprpersistent-indexer
Last change on this file since 45af7e1 was 45af7e1, checked in by Peter A. Buhr <pabuhr@…>, 3 years ago

start rewrite of concurrency paper for SPE

  • Property mode set to 100644
File size: 128.0 KB
Line 
1\documentclass[AMA,STIX1COL]{WileyNJD-v2}
2
3\articletype{RESEARCH ARTICLE}%
4
5\received{XXXXX}
6\revised{XXXXX}
7\accepted{XXXXX}
8
9\raggedbottom
10
11%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12
13% Latex packages used in the document.
14
15\usepackage{epic,eepic}
16\usepackage{xspace}
17\usepackage{comment}
18\usepackage{upquote}                                            % switch curled `'" to straight
19\usepackage{listings}                                           % format program code
20\usepackage[labelformat=simple,aboveskip=0pt,farskip=0pt]{subfig}
21\renewcommand{\thesubfigure}{(\Alph{subfigure})}
22\captionsetup{justification=raggedright,singlelinecheck=false}
23\usepackage{dcolumn}                                            % align decimal points in tables
24\usepackage{capt-of}
25
26\hypersetup{breaklinks=true}
27\definecolor{OliveGreen}{cmyk}{0.64 0 0.95 0.40}
28\definecolor{Mahogany}{cmyk}{0 0.85 0.87 0.35}
29\definecolor{Plum}{cmyk}{0.50 1 0 0}
30
31\usepackage[pagewise]{lineno}
32\renewcommand{\linenumberfont}{\scriptsize\sffamily}
33
34\renewcommand{\topfraction}{0.8}                        % float must be greater than X of the page before it is forced onto its own page
35\renewcommand{\bottomfraction}{0.8}                     % float must be greater than X of the page before it is forced onto its own page
36\renewcommand{\floatpagefraction}{0.8}          % float must be greater than X of the page before it is forced onto its own page
37\renewcommand{\textfraction}{0.0}                       % the entire page maybe devoted to floats with no text on the page at all
38
39\lefthyphenmin=3                                                        % hyphen only after 4 characters
40\righthyphenmin=3
41
42%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
43
44% Names used in the document.
45
46\newcommand{\CFAIcon}{\textsf{C}\raisebox{\depth}{\rotatebox{180}{\textsf{A}}}\xspace} % Cforall symbolic name
47\newcommand{\CFA}{\protect\CFAIcon}             % safe for section/caption
48\newcommand{\CFL}{\textrm{Cforall}\xspace}      % Cforall symbolic name
49\newcommand{\Celeven}{\textrm{C11}\xspace}      % C11 symbolic name
50\newcommand{\CC}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}\xspace} % C++ symbolic name
51\newcommand{\CCeleven}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}11\xspace} % C++11 symbolic name
52\newcommand{\CCfourteen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}14\xspace} % C++14 symbolic name
53\newcommand{\CCseventeen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}17\xspace} % C++17 symbolic name
54\newcommand{\CCtwenty}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}20\xspace} % C++20 symbolic name
55\newcommand{\Csharp}{C\raisebox{-0.7ex}{\Large$^\sharp$}\xspace} % C# symbolic name
56
57%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
58
59\newcommand{\Textbf}[2][red]{{\color{#1}{\textbf{#2}}}}
60\newcommand{\Emph}[2][red]{{\color{#1}\textbf{\emph{#2}}}}
61\newcommand{\uC}{$\mu$\CC}
62\newcommand{\TODO}[1]{{\Textbf{#1}}}
63
64%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
65
66% Default underscore is too low and wide. Cannot use lstlisting "literate" as replacing underscore
67% removes it as a variable-name character so keywords in variables are highlighted. MUST APPEAR
68% AFTER HYPERREF.
69%\DeclareTextCommandDefault{\textunderscore}{\leavevmode\makebox[1.2ex][c]{\rule{1ex}{0.1ex}}}
70\renewcommand{\textunderscore}{\leavevmode\makebox[1.2ex][c]{\rule{1ex}{0.075ex}}}
71
72\renewcommand*{\thefootnote}{\Alph{footnote}} % hack because fnsymbol does not work
73%\renewcommand*{\thefootnote}{\fnsymbol{footnote}}
74
75\makeatletter
76% parindent is relative, i.e., toggled on/off in environments like itemize, so store the value for
77% use rather than use \parident directly.
78\newlength{\parindentlnth}
79\setlength{\parindentlnth}{\parindent}
80
81\newcommand{\LstBasicStyle}[1]{{\lst@basicstyle{\lst@basicstyle{#1}}}}
82\newcommand{\LstKeywordStyle}[1]{{\lst@basicstyle{\lst@keywordstyle{#1}}}}
83\newcommand{\LstCommentStyle}[1]{{\lst@basicstyle{\lst@commentstyle{#1}}}}
84
85\newlength{\gcolumnposn}                                        % temporary hack because lstlisting does not handle tabs correctly
86\newlength{\columnposn}
87\setlength{\gcolumnposn}{3.5in}
88\setlength{\columnposn}{\gcolumnposn}
89
90\newcommand{\C}[2][\@empty]{\ifx#1\@empty\else\global\setlength{\columnposn}{#1}\global\columnposn=\columnposn\fi\hfill\makebox[\textwidth-\columnposn][l]{\lst@basicstyle{\LstCommentStyle{#2}}}}
91\newcommand{\CRT}{\global\columnposn=\gcolumnposn}
92
93% Denote newterms in particular font and index them without particular font and in lowercase, e.g., \newterm{abc}.
94% The option parameter provides an index term different from the new term, e.g., \newterm[\texttt{abc}]{abc}
95% The star version does not lowercase the index information, e.g., \newterm*{IBM}.
96\newcommand{\newtermFontInline}{\emph}
97\newcommand{\newterm}{\@ifstar\@snewterm\@newterm}
98\newcommand{\@newterm}[2][\@empty]{\lowercase{\def\temp{#2}}{\newtermFontInline{#2}}\ifx#1\@empty\index{\temp}\else\index{#1@{\protect#2}}\fi}
99\newcommand{\@snewterm}[2][\@empty]{{\newtermFontInline{#2}}\ifx#1\@empty\index{#2}\else\index{#1@{\protect#2}}\fi}
100
101% Latin abbreviation
102\newcommand{\abbrevFont}{\textit}                       % set empty for no italics
103\@ifundefined{eg}{
104\newcommand{\EG}{\abbrevFont{e}\abbrevFont{g}}
105\newcommand*{\eg}{%
106        \@ifnextchar{,}{\EG}%
107                {\@ifnextchar{:}{\EG}%
108                        {\EG,\xspace}}%
109}}{}%
110\@ifundefined{ie}{
111\newcommand{\IE}{\abbrevFont{i}\abbrevFont{e}}
112\newcommand*{\ie}{%
113        \@ifnextchar{,}{\IE}%
114                {\@ifnextchar{:}{\IE}%
115                        {\IE,\xspace}}%
116}}{}%
117\@ifundefined{etc}{
118\newcommand{\ETC}{\abbrevFont{etc}}
119\newcommand*{\etc}{%
120        \@ifnextchar{.}{\ETC}%
121        {\ETC.\xspace}%
122}}{}%
123\@ifundefined{etal}{
124\newcommand{\ETAL}{\abbrevFont{et}~\abbrevFont{al}}
125\newcommand*{\etal}{%
126        \@ifnextchar{.}{\protect\ETAL}%
127                {\protect\ETAL.\xspace}%
128}}{}%
129\@ifundefined{viz}{
130\newcommand{\VIZ}{\abbrevFont{viz}}
131\newcommand*{\viz}{%
132        \@ifnextchar{.}{\VIZ}%
133                {\VIZ.\xspace}%
134}}{}%
135\makeatother
136
137\newenvironment{cquote}
138               {\list{}{\lstset{resetmargins=true,aboveskip=0pt,belowskip=0pt}\topsep=3pt\parsep=0pt\leftmargin=\parindentlnth\rightmargin\leftmargin}%
139                \item\relax}
140               {\endlist}
141
142%\newenvironment{cquote}{%
143%\list{}{\lstset{resetmargins=true,aboveskip=0pt,belowskip=0pt}\topsep=3pt\parsep=0pt\leftmargin=\parindentlnth\rightmargin\leftmargin}%
144%\item\relax%
145%}{%
146%\endlist%
147%}% cquote
148
149% CFA programming language, based on ANSI C (with some gcc additions)
150\lstdefinelanguage{CFA}[ANSI]{C}{
151        morekeywords={
152                _Alignas, _Alignof, __alignof, __alignof__, asm, __asm, __asm__, __attribute, __attribute__,
153                auto, _Bool, catch, catchResume, choose, _Complex, __complex, __complex__, __const, __const__,
154                coroutine, disable, dtype, enable, exception, __extension__, fallthrough, fallthru, finally,
155                __float80, float80, __float128, float128, forall, ftype, _Generic, _Imaginary, __imag, __imag__,
156                inline, __inline, __inline__, __int128, int128, __label__, monitor, mutex, _Noreturn, one_t, or,
157                otype, restrict, __restrict, __restrict__, __signed, __signed__, _Static_assert, thread,
158                _Thread_local, throw, throwResume, timeout, trait, try, ttype, typeof, __typeof, __typeof__,
159                virtual, __volatile, __volatile__, waitfor, when, with, zero_t},
160        moredirectives={defined,include_next}%
161}
162
163\lstset{
164language=CFA,
165columns=fullflexible,
166basicstyle=\linespread{0.9}\sf,                                                 % reduce line spacing and use sanserif font
167stringstyle=\tt,                                                                                % use typewriter font
168tabsize=5,                                                                                              % N space tabbing
169xleftmargin=\parindentlnth,                                                             % indent code to paragraph indentation
170%mathescape=true,                                                                               % LaTeX math escape in CFA code $...$
171escapechar=\$,                                                                                  % LaTeX escape in CFA code
172keepspaces=true,                                                                                %
173showstringspaces=false,                                                                 % do not show spaces with cup
174showlines=true,                                                                                 % show blank lines at end of code
175aboveskip=4pt,                                                                                  % spacing above/below code block
176belowskip=3pt,
177% replace/adjust listing characters that look bad in sanserif
178literate={-}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.1ex}}}}1 {^}{\raisebox{0.6ex}{$\scriptstyle\land\,$}}1
179        {~}{\raisebox{0.3ex}{$\scriptstyle\sim\,$}}1 % {`}{\ttfamily\upshape\hspace*{-0.1ex}`}1
180        {<}{\textrm{\textless}}1 {>}{\textrm{\textgreater}}1
181        {<-}{$\leftarrow$}2 {=>}{$\Rightarrow$}2 {->}{\makebox[1ex][c]{\raisebox{0.5ex}{\rule{0.8ex}{0.075ex}}}\kern-0.2ex{\textrm{\textgreater}}}2,
182moredelim=**[is][\color{red}]{`}{`},
183}% lstset
184
185% uC++ programming language, based on ANSI C++
186\lstdefinelanguage{uC++}[ANSI]{C++}{
187        morekeywords={
188                _Accept, _AcceptReturn, _AcceptWait, _Actor, _At, _CatchResume, _Cormonitor, _Coroutine, _Disable,
189                _Else, _Enable, _Event, _Finally, _Monitor, _Mutex, _Nomutex, _PeriodicTask, _RealTimeTask,
190                _Resume, _Select, _SporadicTask, _Task, _Timeout, _When, _With, _Throw},
191}
192\lstdefinelanguage{Golang}{
193        morekeywords=[1]{package,import,func,type,struct,return,defer,panic,recover,select,var,const,iota,},
194        morekeywords=[2]{string,uint,uint8,uint16,uint32,uint64,int,int8,int16,int32,int64,
195                bool,float32,float64,complex64,complex128,byte,rune,uintptr, error,interface},
196        morekeywords=[3]{map,slice,make,new,nil,len,cap,copy,close,true,false,delete,append,real,imag,complex,chan,},
197        morekeywords=[4]{for,break,continue,range,goto,switch,case,fallthrough,if,else,default,},
198        morekeywords=[5]{Println,Printf,Error,},
199        sensitive=true,
200        morecomment=[l]{//},
201        morecomment=[s]{/*}{*/},
202        morestring=[b]',
203        morestring=[b]",
204        morestring=[s]{`}{`},
205}
206
207\lstnewenvironment{cfa}[1][]
208{\lstset{#1}}
209{}
210\lstnewenvironment{C++}[1][]                            % use C++ style
211{\lstset{language=C++,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
212{}
213\lstnewenvironment{uC++}[1][]
214{\lstset{#1}}
215{}
216\lstnewenvironment{Go}[1][]
217{\lstset{#1}}
218{}
219
220% inline code @...@
221\lstMakeShortInline@%
222
223\let\OLDthebibliography\thebibliography
224\renewcommand\thebibliography[1]{
225  \OLDthebibliography{#1}
226  \setlength{\parskip}{0pt}
227  \setlength{\itemsep}{4pt plus 0.3ex}
228}
229
230\title{\texorpdfstring{Concurrency in \protect\CFA}{Concurrency in Cforall}}
231
232\author[1]{Thierry Delisle}
233\author[1]{Peter A. Buhr*}
234\authormark{DELISLE \textsc{et al.}}
235
236\address[1]{\orgdiv{Cheriton School of Computer Science}, \orgname{University of Waterloo}, \orgaddress{\state{Waterloo, ON}, \country{Canada}}}
237
238\corres{*Peter A. Buhr, Cheriton School of Computer Science, University of Waterloo, 200 University Avenue West, Waterloo, ON, N2L 3G1, Canada. \email{pabuhr{\char`\@}uwaterloo.ca}}
239
240\fundingInfo{Natural Sciences and Engineering Research Council of Canada}
241
242\abstract[Summary]{
243\CFA is a modern, polymorphic, \emph{non-object-oriented}, backwards-compatible extension of the C programming language.
244This paper discusses the concurrency and parallelism features in \CFA, and its runtime system.
245These features are created from scratch as ISO C's concurrency is low-level and unimplemented, so C programmers continue to rely on the C pthreads library.
246\CFA provides high-level control-flow mechanisms, like coroutines and lightweight (user) threads, and monitors for mutual exclusion and synchronization.
247A unique contribution of this work is allowing multiple monitors to be safely acquired \emph{simultaneously}.
248All features respect the expectations of C programmers, while being fully integrate with the \CFA polymorphic type-system and other language features.
249Experimental results show comparable performance of the new features with similar mechanisms in other concurrent programming-languages.
250}%
251
252\keywords{concurrency, parallelism, coroutines, threads, monitors, runtime, C, \CFA (Cforall)}
253
254
255\begin{document}
256\linenumbers                                            % comment out to turn off line numbering
257
258\maketitle
259
260
261\section{Introduction}
262
263This paper discusses the design of the high-level concurrency and parallelism features in \CFA, and its runtime.
264\CFA is a modern, polymorphic, \emph{non-object-oriented}, backwards-compatible extension of the C programming language~\cite{Moss18}.
265Within the \CFA framework, new concurrency features were created from scratch.
266While ISO \Celeven defines concurrency~\cite[\S~7.26]{C11}, it is largely wrappers for a subset of the pthreads library~\cite{Butenhof97,Pthreads}.
267Furthermore, \Celeven and pthreads concurrency is simple: create/join threads in a function and a few locks, which is low-level and error prone;
268no high-level language concurrency features exist.
269Interestingly, 8 years since publication of the \Celeven standard, neither gcc-8 nor clang-8 (most recent versions) support \Celeven @threads.h@, indicating little interest in the C concurrency approach.
270Finally, while the \Celeven standard does not state a concurrent threading-model, the strong association with pthreads suggests the threading model is kernel-level threading (1:1)~\cite{ThreadModel}.
271
272There has been a re-interest during the past decade in user-level (M:N, green) threading in new and old programming languages, and providing high-level constructs like coroutines, monitors, tasks, and actors for presenting advanced control-flow.
273As multi-core hardware became available in the 1980/90s, both user and kernel threading were examined.
274Kernel threading was chosen, largely because of its simplicity and fit with the simpler operating systems and hardware architectures at the time, which gave it a performance advantage~\cite{Drepper03}.
275Libraries like pthreads were developed for C and the Solaris operating-system switched from user (JDK 1.1~\cite{JDK1.1}) to kernel threads.
276As a result, languages like Java, Scala~\cite{Scala}, Objective-C~\cite{obj-c-book}, \CCeleven~\cite{C11}, and C\#~\cite{Csharp} adopted the 1:1 kernel-threading model, with a variety of presentation mechanisms.
277From 2000 onwards, languages like Go~\cite{Go}, Erlang~\cite{Erlang}, Haskell~\cite{Haskell}, D~\cite{D}, and \uC~\cite{uC++,uC++book} have championed the M:N user-threading model, and many user-threading libraries have appeared~\cite{Qthreads,MPC,BoostThreads}, including putting green threads back into Java~\cite{Quasar}.
278Because advanced control-flow (including exception handling) is pervasive in a programming language and its runtime, these features must be understood by the language (i.e., not added via a library) to prevent invalidation by sequential optimizations~\cite{Buhr95a,Boehm05}.
279
280The main argument for user-level threading is matching the concurrency model with the programming-language style, versus adapting language concurrency to one general approach.
281For example, it is possible to provide coroutines, monitors, and tasks as specialized types in an object-oriented language, integrating these constructs to allow leveraging the type-system (static type-checking) and all other object-oriented capabilities~\cite{uC++}.
282The user-threading approach facilitates a simpler concurrency construction using thread objects and leveraging sequential patterns versus call-backs and events~\cite{vonBehren03}.
283As well, user-level threads are lighter weight than kernel threads, so there is less restriction on programming styles that encourage large numbers of threads performing smaller work-units to facilitate load balancing by the runtime~\cite{Verch12}.
284User threading is also able to layer multiple concurrency models into a single language (locks, monitors, tasks, actors, futures), so programmers can chose the model that best fits an application.
285Finally, it is possible to discretely fold locking and non-blocking I/O multiplexing into the language's I/O libraries, so threading implicitly dovetails with the I/O subsystem.
286Performant user-threading implementations (both time and space) are appearing that are competitive with direct kernel-threading implementations, while achieving the programming advantages of high concurrency levels and safety.
287
288Adding advanced control-flow to \CFA is similar to current and future extensions in \CCeleven through to \CCtwenty.
289However, we contend the \CFA extensions are demonstrably better than those proposed for \CC.
290For example, a unique contribution of this work is allowing multiple monitors to be safely acquired \emph{simultaneously} (deadlock free), while integrating this capability with all monitor synchronization mechanisms.
291As well, all control-flow features respect the expectations of C programmers, with statically type-safe interfaces that integrate with the \CFA polymorphic type-system and other language features.
292Experimental results show comparable performance of the new features with similar mechanisms in other concurrent programming-languages.
293
294\begin{comment}
295This paper provides a minimal concurrency \newterm{Application Program Interface} (API) that is simple, efficient and can be used to build other concurrency features.
296While the simplest concurrency system is a thread and a lock, this low-level approach is hard to master.
297An easier approach for programmers is to support higher-level constructs as the basis of concurrency.
298Indeed, for highly-productive concurrent-programming, high-level approaches are much more popular~\cite{Hochstein05}.
299Examples 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}.
300
301The following terminology is used.
302A \newterm{thread} is a fundamental unit of execution that runs a sequence of code and requires a stack to maintain state.
303Multiple simultaneous threads give rise to \newterm{concurrency}, which requires locking to ensure access to shared data and safe communication.
304\newterm{Locking}, and by extension \newterm{locks}, are defined as a mechanism to prevent progress of threads to provide safety.
305\newterm{Parallelism} is running multiple threads simultaneously.
306Parallelism implies \emph{actual} simultaneous execution, where concurrency only requires \emph{apparent} simultaneous execution.
307As such, parallelism only affects performance, which is observed through differences in space and/or time at runtime.
308Hence, there are two problems to be solved: concurrency and parallelism.
309While these two concepts are often combined, they are distinct, requiring different tools~\cite[\S~2]{Buhr05a}.
310Concurrency tools handle mutual exclusion and synchronization, while parallelism tools handle performance, cost, and resource utilization.
311
312The proposed concurrency API is implemented in a dialect of C, called \CFA (pronounced C-for-all).
313The 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.
314\end{comment}
315
316
317\begin{comment}
318\section{\CFA Overview}
319
320The following is a quick introduction to the \CFA language, specifically tailored to the features needed to support concurrency.
321Extended versions and explanation of the following code examples are available at the \CFA website~\cite{Cforall} or in Moss~\etal~\cite{Moss18}.
322
323\CFA is a non-object-oriented extension of ISO-C, and hence, supports all C paradigms.
324Like C, the building blocks of \CFA are structures and routines.
325Virtually all of the code generated by the \CFA translator respects C memory layouts and calling conventions.
326While \CFA is not object oriented, lacking the concept of a receiver (\eg @this@) and nominal inheritance-relationships, C has a notion of objects: ``region of data storage in the execution environment, the contents of which can represent values''~\cite[3.15]{C11}.
327While some object-oriented features appear in \CFA, they are independent capabilities, allowing \CFA to adopt them while maintaining a procedural paradigm.
328
329
330\subsection{References}
331
332\CFA provides multi-level rebindable references, as an alternative to pointers, which significantly reduces syntactic noise.
333\begin{cfa}
334int x = 1, y = 2, z = 3;
335int * p1 = &x, ** p2 = &p1,  *** p3 = &p2,      $\C{// pointers to x}$
336    `&` r1 = x,   `&&` r2 = r1,   `&&&` r3 = r2;        $\C{// references to x}$
337int * p4 = &z, `&` r4 = z;
338
339*p1 = 3; **p2 = 3; ***p3 = 3;       // change x
340r1 =  3;     r2 = 3;      r3 = 3;        // change x: implicit dereferences *r1, **r2, ***r3
341**p3 = &y; *p3 = &p4;                // change p1, p2
342`&`r3 = &y; `&&`r3 = &`&`r4;             // change r1, r2: cancel implicit dereferences (&*)**r3, (&(&*)*)*r3, &(&*)r4
343\end{cfa}
344A reference is a handle to an object, like a pointer, but is automatically dereferenced the specified number of levels.
345Referencing (address-of @&@) a reference variable cancels one of the implicit dereferences, until there are no more implicit references, after which normal expression behaviour applies.
346
347
348\subsection{\texorpdfstring{\protect\lstinline{with} Statement}{with Statement}}
349\label{s:WithStatement}
350
351Heterogeneous data is aggregated into a structure/union.
352To 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.
353\begin{cquote}
354\vspace*{-\baselineskip}%???
355\lstDeleteShortInline@%
356\begin{cfa}
357struct S { char c; int i; double d; };
358struct T { double m, n; };
359// multiple aggregate parameters
360\end{cfa}
361\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
362\begin{cfa}
363void f( S & s, T & t ) {
364        `s.`c; `s.`i; `s.`d;
365        `t.`m; `t.`n;
366}
367\end{cfa}
368&
369\begin{cfa}
370void f( S & s, T & t ) `with ( s, t )` {
371        c; i; d;                // no qualification
372        m; n;
373}
374\end{cfa}
375\end{tabular}
376\lstMakeShortInline@%
377\end{cquote}
378Object-oriented programming languages only provide implicit qualification for the receiver.
379
380In detail, the @with@-statement syntax is:
381\begin{cfa}
382$\emph{with-statement}$:
383        'with' '(' $\emph{expression-list}$ ')' $\emph{compound-statement}$
384\end{cfa}
385and may appear as the body of a routine or nested within a routine body.
386Each expression in the expression-list provides a type and object.
387The type must be an aggregate type.
388(Enumerations are already opened.)
389The object is the implicit qualifier for the open structure-fields.
390All expressions in the expression list are opened in parallel within the compound statement, which is different from Pascal, which nests the openings from left to right.
391
392
393\subsection{Overloading}
394
395\CFA maximizes the ability to reuse names via overloading to aggressively address the naming problem.
396Both variables and routines may be overloaded, where selection is based on number and types of returns and arguments (as in Ada~\cite{Ada}).
397\newpage
398\vspace*{-2\baselineskip}%???
399\begin{cquote}
400\begin{cfa}
401// selection based on type
402\end{cfa}
403\lstDeleteShortInline@%
404\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
405\begin{cfa}
406const short int `MIN` = -32768;
407const int `MIN` = -2147483648;
408const long int `MIN` = -9223372036854775808L;
409\end{cfa}
410&
411\begin{cfa}
412short int si = `MIN`;
413int i = `MIN`;
414long int li = `MIN`;
415\end{cfa}
416\end{tabular}
417\begin{cfa}
418// selection based on type and number of parameters
419\end{cfa}
420\begin{tabular}{@{}l@{\hspace{2.7\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
421\begin{cfa}
422void `f`( void );
423void `f`( char );
424void `f`( int, double );
425\end{cfa}
426&
427\begin{cfa}
428`f`();
429`f`( 'a' );
430`f`( 3, 5.2 );
431\end{cfa}
432\end{tabular}
433\begin{cfa}
434// selection based on type and number of returns
435\end{cfa}
436\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
437\begin{cfa}
438char `f`( int );
439double `f`( int );
440[char, double] `f`( int );
441\end{cfa}
442&
443\begin{cfa}
444char c = `f`( 3 );
445double d = `f`( 3 );
446[d, c] = `f`( 3 );
447\end{cfa}
448\end{tabular}
449\lstMakeShortInline@%
450\end{cquote}
451Overloading is important for \CFA concurrency since the runtime system relies on creating different types to represent concurrency objects.
452Therefore, overloading eliminates long prefixes and other naming conventions to prevent name clashes.
453As seen in Section~\ref{s:Concurrency}, routine @main@ is heavily overloaded.
454As another example, variable overloading is useful in the parallel semantics of the @with@ statement for fields with the same name:
455\begin{cfa}
456struct S { int `i`; int j; double m; } s;
457struct T { int `i`; int k; int m; } t;
458with ( s, t ) {
459        j + k;                                                                  $\C{// unambiguous, s.j + t.k}$
460        m = 5.0;                                                                $\C{// unambiguous, s.m = 5.0}$
461        m = 1;                                                                  $\C{// unambiguous, t.m = 1}$
462        int a = m;                                                              $\C{// unambiguous, a = t.m }$
463        double b = m;                                                   $\C{// unambiguous, b = s.m}$
464        int c = `s.i` + `t.i`;                                  $\C{// unambiguous, qualification}$
465        (double)m;                                                              $\C{// unambiguous, cast s.m}$
466}
467\end{cfa}
468For parallel semantics, both @s.i@ and @t.i@ are visible with the same type, so only @i@ is ambiguous without qualification.
469
470
471\subsection{Operators}
472
473Overloading also extends to operators.
474Operator-overloading syntax creates a routine name with an operator symbol and question marks for the operands:
475\begin{cquote}
476\lstDeleteShortInline@%
477\begin{tabular}{@{}ll@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
478\begin{cfa}
479int ++?(int op);
480int ?++(int op);
481int `?+?`(int op1, int op2);
482int ?<=?(int op1, int op2);
483int ?=? (int & op1, int op2);
484int ?+=?(int & op1, int op2);
485\end{cfa}
486&
487\begin{cfa}
488// unary prefix increment
489// unary postfix increment
490// binary plus
491// binary less than
492// binary assignment
493// binary plus-assignment
494\end{cfa}
495&
496\begin{cfa}
497struct S { int i, j; };
498S `?+?`( S op1, S op2) { // add two structures
499        return (S){op1.i + op2.i, op1.j + op2.j};
500}
501S s1 = {1, 2}, s2 = {2, 3}, s3;
502s3 = s1 `+` s2;         // compute sum: s3 == {2, 5}
503\end{cfa}
504\end{tabular}
505\lstMakeShortInline@%
506\end{cquote}
507
508
509\subsection{Constructors / Destructors}
510
511Object lifetime is a challenge in non-managed programming languages.
512\CFA responds with \CC-like constructors and destructors, using a different operator-overloading syntax.
513\begin{cfa}
514struct VLA { int len, * data; };                        $\C{// variable length array of integers}$
515void ?{}( VLA & vla ) with ( vla ) { len = 10;  data = alloc( len ); }  $\C{// default constructor}$
516void ?{}( VLA & vla, int size, char fill ) with ( vla ) { len = size;  data = alloc( len, fill ); } // initialization
517void ?{}( VLA & vla, VLA other ) { vla.len = other.len;  vla.data = other.data; } $\C{// copy, shallow}$
518void ^?{}( VLA & vla ) with ( vla ) { free( data ); } $\C{// destructor}$
519{
520        VLA  x,            y = { 20, 0x01 },     z = y; $\C{// z points to y}$
521        // $\LstCommentStyle{\color{red}\ \ \ x\{\};\ \ \ \ \ \ \ \ \ y\{ 20, 0x01 \};\ \ \ \ \ \ \ \ \ \ z\{ z, y \};\ \ \ \ \ \ \ implicit calls}$
522        ^x{};                                                                   $\C{// deallocate x}$
523        x{};                                                                    $\C{// reallocate x}$
524        z{ 5, 0xff };                                                   $\C{// reallocate z, not pointing to y}$
525        ^y{};                                                                   $\C{// deallocate y}$
526        y{ x };                                                                 $\C{// reallocate y, points to x}$
527        x{};                                                                    $\C{// reallocate x, not pointing to y}$
528}       //  $\LstCommentStyle{\color{red}\^{}z\{\};\ \ \^{}y\{\};\ \ \^{}x\{\};\ \ \ implicit calls}$
529\end{cfa}
530Like \CC, construction is implicit on allocation (stack/heap) and destruction is implicit on deallocation.
531The object and all their fields are constructed/destructed.
532\CFA also provides @new@ and @delete@ as library routines, which behave like @malloc@ and @free@, in addition to constructing and destructing objects:
533\begin{cfa}
534{
535        ... struct S s = {10}; ...                              $\C{// allocation, call constructor}$
536}                                                                                       $\C{// deallocation, call destructor}$
537struct S * s = new();                                           $\C{// allocation, call constructor}$
538...
539delete( s );                                                            $\C{// deallocation, call destructor}$
540\end{cfa}
541\CFA concurrency uses object lifetime as a means of mutual exclusion and/or synchronization.
542
543
544\subsection{Parametric Polymorphism}
545\label{s:ParametricPolymorphism}
546
547The 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.
548For example, the following sum routine works for any type that supports construction from 0 and addition:
549\begin{cfa}
550forall( otype T | { void `?{}`( T *, zero_t ); T `?+?`( T, T ); } ) // constraint type, 0 and +
551T sum( T a[$\,$], size_t size ) {
552        `T` total = { `0` };                                    $\C{// initialize by 0 constructor}$
553        for ( size_t i = 0; i < size; i += 1 )
554                total = total `+` a[i];                         $\C{// select appropriate +}$
555        return total;
556}
557S sa[5];
558int i = sum( sa, 5 );                                           $\C{// use S's 0 construction and +}$
559\end{cfa}
560Type variables can be @otype@ or @dtype@.
561@otype@ refers to a \emph{complete type}, \ie, a type with size, alignment, default constructor, copy constructor, destructor, and assignment operator.
562@dtype@ refers to an \emph{incomplete type}, \eg, void or a forward-declared type.
563The builtin types @zero_t@ and @one_t@ overload constant 0 and 1 for a new types, where both 0 and 1 have special meaning in C.
564
565\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:
566\begin{cfa}
567trait `sumable`( otype T ) {
568        void `?{}`( T &, zero_t );                              $\C{// 0 literal constructor}$
569        T `?+?`( T, T );                                                $\C{// assortment of additions}$
570        T ?+=?( T &, T );
571        T ++?( T & );
572        T ?++( T & );
573};
574forall( otype T `| sumable( T )` )                      $\C{// use trait}$
575T sum( T a[$\,$], size_t size );
576\end{cfa}
577
578Using the return type for overload discrimination, it is possible to write a type-safe @alloc@ based on the C @malloc@:
579\begin{cfa}
580forall( dtype T | sized(T) ) T * alloc( void ) { return (T *)malloc( sizeof(T) ); }
581int * ip = alloc();                                                     $\C{// select type and size from left-hand side}$
582double * dp = alloc();
583struct S {...} * sp = alloc();
584\end{cfa}
585where the return type supplies the type/size of the allocation, which is impossible in most type systems.
586\end{comment}
587
588
589\section{Concurrency}
590\label{s:Concurrency}
591
592At its core, concurrency is based on multiple call-stacks and scheduling threads executing on these stacks.
593Multiple 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}.
594In 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.
595A \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;
596a \newterm{stackful} coroutine executes on its own stack, allowing full generality.
597Only stackful coroutines are a stepping stone to concurrency.
598
599The 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}.
600Therefore, a minimal concurrency system is possible using coroutines (see Section \ref{coroutine}) in conjunction with a scheduler to decide where to context switch next.
601The resulting execution system now follows a cooperative threading-model, called \newterm{non-preemptive scheduling}.
602
603Because the scheduler is special, it can either be a stackless or stackful coroutine.
604For 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.
605For stackful, the current coroutine switches to the scheduler, which performs scheduling, and it then switches to the next coroutine, so there are two context switches.
606A stackful scheduler is often used for simplicity and security.
607
608Regardless of the approach used, a subset of concurrency related challenges start to appear.
609For 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}.
610While a scheduler introduces uncertainty in the order of execution, preemption introduces uncertainty about where context switches occur.
611Interestingly, uncertainty is necessary for the runtime (operating) system to give the illusion of parallelism on a single processor and increase performance on multiple processors.
612The reason is that only the runtime has complete knowledge about resources and how to best utilized them.
613However, the introduction of unrestricted non-determinism results in the need for \newterm{mutual exclusion} and \newterm{synchronization} to restrict non-determinism for correctness;
614otherwise, it is impossible to write meaningful programs.
615Optimal performance in concurrent applications is often obtained by having as much non-determinism as correctness allows.
616
617An 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.
618As such, library support for threading is far from widespread.
619At the time of writing the paper, neither \protect\lstinline@gcc@ nor \protect\lstinline@clang@ support \protect\lstinline@threads.h@ in their standard libraries.}.
620In 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.
621As 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.
622Furthermore, because C is a system-level language, programmers expect to choose precisely which features they need and which cost they are willing to pay.
623Hence, concurrent programs should be written using high-level mechanisms, and only step down to lower-level mechanisms when performance bottlenecks are encountered.
624
625
626\subsection{Coroutines: A Stepping Stone}\label{coroutine}
627
628While 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).
629Coroutines are generalized routines allowing execution to be temporarily suspended and later resumed.
630Hence, 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.
631This capability is accomplished via the coroutine's stack, where suspend/resume context switch among stacks.
632Because 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.
633Therefore, the two fundamental features of the core \CFA coroutine-API are independent call-stacks and @suspend@/@resume@ operations.
634
635For example, a problem made easier with coroutines is unbounded generators, \eg generating an infinite sequence of Fibonacci numbers
636\begin{displaymath}
637\mathsf{fib}(n) = \left \{
638\begin{array}{ll}
6390                                       & n = 0         \\
6401                                       & n = 1         \\
641\mathsf{fib}(n-1) + \mathsf{fib}(n-2)   & n \ge 2       \\
642\end{array}
643\right.
644\end{displaymath}
645where Figure~\ref{f:C-fibonacci} shows conventional approaches for writing a Fibonacci generator in C.
646Figure~\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.
647Figure~\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 $\mathsf{fib}(n-2)$.
648
649\begin{figure}
650\centering
651\newbox\myboxA
652\begin{lrbox}{\myboxA}
653\begin{cfa}[aboveskip=0pt,belowskip=0pt]
654`int f1, f2, state = 1;`   // single global variables
655int fib() {
656        int fn;
657        `switch ( state )` {  // explicit execution state
658          case 1: fn = 0;  f1 = fn;  state = 2;  break;
659          case 2: fn = 1;  f2 = f1;  f1 = fn;  state = 3;  break;
660          case 3: fn = f1 + f2;  f2 = f1;  f1 = fn;  break;
661        }
662        return fn;
663}
664int main() {
665
666        for ( int i = 0; i < 10; i += 1 ) {
667                printf( "%d\n", fib() );
668        }
669}
670\end{cfa}
671\end{lrbox}
672
673\newbox\myboxB
674\begin{lrbox}{\myboxB}
675\begin{cfa}[aboveskip=0pt,belowskip=0pt]
676#define FIB_INIT `{ 0, 1 }`
677typedef struct { int f2, f1; } Fib;
678int fib( Fib * f ) {
679
680        int ret = f->f2;
681        int fn = f->f1 + f->f2;
682        f->f2 = f->f1; f->f1 = fn;
683
684        return ret;
685}
686int main() {
687        Fib f1 = FIB_INIT, f2 = FIB_INIT;
688        for ( int i = 0; i < 10; i += 1 ) {
689                printf( "%d %d\n", fib( &f1 ), fib( &f2 ) );
690        }
691}
692\end{cfa}
693\end{lrbox}
694
695\subfloat[3 States: global variables]{\label{f:GlobalVariables}\usebox\myboxA}
696\qquad
697\subfloat[1 State: external variables]{\label{f:ExternalState}\usebox\myboxB}
698\caption{C Fibonacci Implementations}
699\label{f:C-fibonacci}
700
701\bigskip
702
703\newbox\myboxA
704\begin{lrbox}{\myboxA}
705\begin{cfa}[aboveskip=0pt,belowskip=0pt]
706`coroutine` Fib { int fn; };
707void main( Fib & fib ) with( fib ) {
708        int f1, f2;
709        fn = 0;  f1 = fn;  `suspend()`;
710        fn = 1;  f2 = f1;  f1 = fn;  `suspend()`;
711        for ( ;; ) {
712                fn = f1 + f2;  f2 = f1;  f1 = fn;  `suspend()`;
713        }
714}
715int next( Fib & fib ) with( fib ) {
716        `resume( fib );`
717        return fn;
718}
719int main() {
720        Fib f1, f2;
721        for ( int i = 1; i <= 10; i += 1 ) {
722                sout | next( f1 ) | next( f2 );
723        }
724}
725\end{cfa}
726\end{lrbox}
727\newbox\myboxB
728\begin{lrbox}{\myboxB}
729\begin{cfa}[aboveskip=0pt,belowskip=0pt]
730`coroutine` Fib { int ret; };
731void main( Fib & f ) with( fib ) {
732        int fn, f1 = 1, f2 = 0;
733        for ( ;; ) {
734                ret = f2;
735
736                fn = f1 + f2;  f2 = f1;  f1 = fn; `suspend();`
737        }
738}
739int next( Fib & fib ) with( fib ) {
740        `resume( fib );`
741        return ret;
742}
743
744
745
746
747
748
749\end{cfa}
750\end{lrbox}
751\subfloat[3 States, internal variables]{\label{f:Coroutine3States}\usebox\myboxA}
752\qquad\qquad
753\subfloat[1 State, internal variables]{\label{f:Coroutine1State}\usebox\myboxB}
754\caption{\CFA Coroutine Fibonacci Implementations}
755\label{f:cfa-fibonacci}
756\end{figure}
757
758Using a coroutine, it is possible to express the Fibonacci formula directly without any of the C problems.
759Figure~\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@.
760Like 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.
761The coroutine main's stack holds the state for the next generation, @f1@ and @f2@, and the code represents the three states in the Fibonacci formula via the three suspend points, to context switch back to the caller's @resume@.
762The interface routine @next@, takes a Fibonacci instance and context switches to it using @resume@;
763on restart, the Fibonacci field, @fn@, contains the next value in the sequence, which is returned.
764The first @resume@ is special because it allocates the coroutine stack and cocalls its coroutine main on that stack;
765when the coroutine main returns, its stack is deallocated.
766Hence, @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.
767Figure~\ref{f:Coroutine1State} shows the coroutine version of the C version in Figure~\ref{f:ExternalState}.
768Coroutine generators are called \newterm{output coroutines} because values are only returned.
769
770Figure~\ref{f:CFAFmt} shows an \newterm{input coroutine}, @Format@, for restructuring text into groups of characters of fixed-size blocks.
771For example, the input of the left is reformatted into the output on the right.
772\begin{quote}
773\tt
774\begin{tabular}{@{}l|l@{}}
775\multicolumn{1}{c|}{\textbf{\textrm{input}}} & \multicolumn{1}{c}{\textbf{\textrm{output}}} \\
776abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz
777&
778\begin{tabular}[t]{@{}lllll@{}}
779abcd    & efgh  & ijkl  & mnop  & qrst  \\
780uvwx    & yzab  & cdef  & ghij  & klmn  \\
781opqr    & stuv  & wxyz  &               &
782\end{tabular}
783\end{tabular}
784\end{quote}
785The 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.
786The destructor provides a newline, if formatted text ends with a full line.
787Figure~\ref{f:CFmt} shows the C equivalent formatter, where the loops of the coroutine are flattened (linearized) and rechecked on each call because execution location is not retained between calls.
788(Linearized code is the bane of device drivers.)
789
790\begin{figure}
791\centering
792\newbox\myboxA
793\begin{lrbox}{\myboxA}
794\begin{cfa}[aboveskip=0pt,belowskip=0pt]
795`coroutine` Format {
796        char ch;   // used for communication
797        int g, b;  // global because used in destructor
798};
799void main( Format & fmt ) with( fmt ) {
800        for ( ;; ) {
801                for ( g = 0; g < 5; g += 1 ) {      // group
802                        for ( b = 0; b < 4; b += 1 ) { // block
803                                `suspend();`
804                                sout | ch;              // separator
805                        }
806                        sout | "  ";               // separator
807                }
808                sout | nl;
809        }
810}
811void ?{}( Format & fmt ) { `resume( fmt );` }
812void ^?{}( Format & fmt ) with( fmt ) {
813        if ( g != 0 || b != 0 ) sout | nl;
814}
815void format( Format & fmt ) {
816        `resume( fmt );`
817}
818int main() {
819        Format fmt;
820        eof: for ( ;; ) {
821                sin | fmt.ch;
822          if ( eof( sin ) ) break eof;
823                format( fmt );
824        }
825}
826\end{cfa}
827\end{lrbox}
828
829\newbox\myboxB
830\begin{lrbox}{\myboxB}
831\begin{cfa}[aboveskip=0pt,belowskip=0pt]
832struct Format {
833        char ch;
834        int g, b;
835};
836void format( struct Format * fmt ) {
837        if ( fmt->ch != -1 ) {      // not EOF ?
838                printf( "%c", fmt->ch );
839                fmt->b += 1;
840                if ( fmt->b == 4 ) {  // block
841                        printf( "  " );      // separator
842                        fmt->b = 0;
843                        fmt->g += 1;
844                }
845                if ( fmt->g == 5 ) {  // group
846                        printf( "\n" );     // separator
847                        fmt->g = 0;
848                }
849        } else {
850                if ( fmt->g != 0 || fmt->b != 0 ) printf( "\n" );
851        }
852}
853int main() {
854        struct Format fmt = { 0, 0, 0 };
855        for ( ;; ) {
856                scanf( "%c", &fmt.ch );
857          if ( feof( stdin ) ) break;
858                format( &fmt );
859        }
860        fmt.ch = -1;
861        format( &fmt );
862}
863\end{cfa}
864\end{lrbox}
865\subfloat[\CFA Coroutine]{\label{f:CFAFmt}\usebox\myboxA}
866\qquad
867\subfloat[C Linearized]{\label{f:CFmt}\usebox\myboxB}
868\caption{Formatting text into lines of 5 blocks of 4 characters.}
869\label{f:fmt-line}
870\end{figure}
871
872The 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.
873However, @resume@ and @suspend@ context switch among existing stack-frames, rather than create new ones so there is no stack growth.
874\newterm{Symmetric (full) coroutine}s have a coroutine call to a resuming routine for another coroutine, and its coroutine main calls another resuming routine, which eventually forms a resuming-call cycle.
875(The trivial cycle is a coroutine resuming itself.)
876This control flow is similar to recursion for normal routines, but again there is no stack growth from the context switch.
877
878\begin{figure}
879\centering
880\lstset{language=CFA,escapechar={},moredelim=**[is][\protect\color{red}]{`}{`}}% allow $
881\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
882\begin{cfa}
883`coroutine` Prod {
884        Cons & c;
885        int N, money, receipt;
886};
887void main( Prod & prod ) with( prod ) {
888        // 1st resume starts here
889        for ( int i = 0; i < N; i += 1 ) {
890                int p1 = random( 100 ), p2 = random( 100 );
891                sout | p1 | " " | p2;
892                int status = delivery( c, p1, p2 );
893                sout | " $" | money | nl | status;
894                receipt += 1;
895        }
896        stop( c );
897        sout | "prod stops";
898}
899int payment( Prod & prod, int money ) {
900        prod.money = money;
901        `resume( prod );`
902        return prod.receipt;
903}
904void start( Prod & prod, int N, Cons &c ) {
905        &prod.c = &c;
906        prod.[N, receipt] = [N, 0];
907        `resume( prod );`
908}
909int main() {
910        Prod prod;
911        Cons cons = { prod };
912        start( prod, 5, cons );
913}
914\end{cfa}
915&
916\begin{cfa}
917`coroutine` Cons {
918        Prod & p;
919        int p1, p2, status;
920        _Bool done;
921};
922void ?{}( Cons & cons, Prod & p ) {
923        &cons.p = &p;
924        cons.[status, done ] = [0, false];
925}
926void ^?{}( Cons & cons ) {}
927void main( Cons & cons ) with( cons ) {
928        // 1st resume starts here
929        int money = 1, receipt;
930        for ( ; ! done; ) {
931                sout | p1 | " " | p2 | nl | " $" | money;
932                status += 1;
933                receipt = payment( p, money );
934                sout | " #" | receipt;
935                money += 1;
936        }
937        sout | "cons stops";
938}
939int delivery( Cons & cons, int p1, int p2 ) {
940        cons.[p1, p2] = [p1, p2];
941        `resume( cons );`
942        return cons.status;
943}
944void stop( Cons & cons ) {
945        cons.done = true;
946        `resume( cons );`
947}
948\end{cfa}
949\end{tabular}
950\caption{Producer / consumer: resume-resume cycle, bi-directional communication}
951\label{f:ProdCons}
952\end{figure}
953
954Figure~\ref{f:ProdCons} shows a producer/consumer symmetric-coroutine performing bi-directional communication.
955Since 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@.
956The @start@ routine communicates both the number of elements to be produced and the consumer into the producer's coroutine-structure.
957Then the @resume@ to @prod@ creates @prod@'s stack with a frame for @prod@'s coroutine main at the top, and context switches to it.
958@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.
959
960The producer call to @delivery@ transfers values into the consumer's communication variables, resumes the consumer, and returns the consumer status.
961For the first resume, @cons@'s stack is initialized, creating local variables retained between subsequent activations of the coroutine.
962The 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).
963The call from the consumer to @payment@ introduces the cycle between producer and consumer.
964When @payment@ is called, the consumer copies values into the producer's communication variable and a resume is executed.
965The context switch restarts the producer at the point where it last context switched, so it continues in @delivery@ after the resume.
966
967@delivery@ returns the status value in @prod@'s coroutine main, where the status is printed.
968The loop then repeats calling @delivery@, where each call resumes the consumer coroutine.
969The context switch to the consumer continues in @payment@.
970The consumer increments and returns the receipt to the call in @cons@'s coroutine main.
971The loop then repeats calling @payment@, where each call resumes the producer coroutine.
972
973After iterating $N$ times, the producer calls @stop@.
974The @done@ flag is set to stop the consumer's execution and a resume is executed.
975The context switch restarts @cons@ in @payment@ and it returns with the last receipt.
976The 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@.
977@stop@ returns and @prod@'s coroutine main terminates.
978The program main restarts after the resume in @start@.
979@start@ returns and the program main terminates.
980
981
982\subsection{Coroutine Implementation}
983
984A 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.
985There are several solutions to this problem and the chosen option forced the \CFA coroutine design.
986
987Object-oriented inheritance provides extra fields and code in a restricted context, but it requires programmers to explicitly perform the inheritance:
988\begin{cfa}[morekeywords={class,inherits}]
989class mycoroutine inherits baseCoroutine { ... }
990\end{cfa}
991and the programming language (and possibly its tool set, \eg debugger) may need to understand @baseCoroutine@ because of the stack.
992Furthermore, the execution of constructors/destructors is in the wrong order for certain operations.
993For 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.
994
995An alternative is composition:
996\begin{cfa}
997struct mycoroutine {
998        ... // declarations
999        baseCoroutine dummy; // composition, last declaration
1000}
1001\end{cfa}
1002which also requires an explicit declaration that must be the last one to ensure correct initialization order.
1003However, there is nothing preventing wrong placement or multiple declarations.
1004
1005For coroutines as for threads, many implementations are based on routine pointers or routine objects~\cite{Butenhof97, C++14, MS:VisualC++, BoostCoroutines15}.
1006For example, Boost implements coroutines in terms of four functor object-types:
1007\begin{cfa}
1008asymmetric_coroutine<>::pull_type
1009asymmetric_coroutine<>::push_type
1010symmetric_coroutine<>::call_type
1011symmetric_coroutine<>::yield_type
1012\end{cfa}
1013Similarly, the canonical threading paradigm is often based on routine pointers, \eg @pthreads@~\cite{Butenhof97}, \Csharp~\cite{Csharp}, Go~\cite{Go}, and Scala~\cite{Scala}.
1014However, the generic thread-handle (identifier) is limited (few operations), unless it is wrapped in a custom type.
1015\begin{cfa}
1016void mycor( coroutine_t cid, void * arg ) {
1017        int * value = (int *)arg;                               $\C{// type unsafe, pointer-size only}$
1018        // Coroutine body
1019}
1020int main() {
1021        int input = 0, output;
1022        coroutine_t cid = coroutine_create( &mycor, (void *)&input ); $\C{// type unsafe, pointer-size only}$
1023        coroutine_resume( cid, (void *)input, (void **)&output ); $\C{// type unsafe, pointer-size only}$
1024}
1025\end{cfa}
1026Since the custom type is simple to write in \CFA and solves several issues, added support for routine/lambda-based coroutines adds very little.
1027
1028Note, the type @coroutine_t@ must be an abstract handle to the coroutine, because the coroutine descriptor and its stack are non-copyable.
1029Copying the coroutine descriptor results in copies being out of date with the current state of the stack.
1030Correspondingly, 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.
1031(There is no mechanism in C to find all stack-specific pointers and update them as part of a copy.)
1032
1033The selected approach is to use language support by introducing a new kind of aggregate (structure):
1034\begin{cfa}
1035coroutine Fibonacci {
1036        int fn; // communication variables
1037};
1038\end{cfa}
1039The @coroutine@ keyword means the compiler (and tool set) can find and inject code where needed.
1040The downside of this approach is that it makes coroutine a special case in the language.
1041Users wanting to extend coroutines or build their own for various reasons can only do so in ways offered by the language.
1042Furthermore, implementing coroutines without language supports also displays the power of a programming language.
1043While this is ultimately the option used for idiomatic \CFA code, coroutines and threads can still be constructed without language support.
1044The reserved keyword simply eases use for the common case.
1045
1046Part of the mechanism to generalize coroutines is using a \CFA trait, which defines a coroutine as anything satisfying the trait @is_coroutine@, and this trait restricts the available set of coroutine-manipulation routines:
1047\begin{cfa}
1048trait is_coroutine( `dtype` T ) {
1049        void main( T & );
1050        coroutine_desc * get_coroutine( T & );
1051};
1052forall( `dtype` T | is_coroutine(T) ) void suspend( T & );
1053forall( `dtype` T | is_coroutine(T) ) void resume( T & );
1054\end{cfa}
1055The @dtype@ property provides no implicit copying operations and the @is_coroutine@ trait provides no explicit copying operations, so all coroutines must be passed by reference (pointer).
1056The 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.
1057The @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.
1058The 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@.
1059The \CFA keyword @coroutine@ implicitly implements the getter and forward declarations required for implementing the coroutine main:
1060\begin{cquote}
1061\begin{tabular}{@{}ccc@{}}
1062\begin{cfa}
1063coroutine MyCor {
1064        int value;
1065
1066};
1067\end{cfa}
1068&
1069{\Large $\Rightarrow$}
1070&
1071\begin{tabular}{@{}ccc@{}}
1072\begin{cfa}
1073struct MyCor {
1074        int value;
1075        coroutine_desc cor;
1076};
1077\end{cfa}
1078&
1079\begin{cfa}
1080static inline coroutine_desc *
1081get_coroutine( MyCor & this ) {
1082        return &this.cor;
1083}
1084\end{cfa}
1085&
1086\begin{cfa}
1087void main( MyCor * this );
1088
1089
1090
1091\end{cfa}
1092\end{tabular}
1093\end{tabular}
1094\end{cquote}
1095The 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.
1096
1097
1098\subsection{Thread Interface}
1099\label{threads}
1100
1101Both user and kernel threads are supported, where user threads provide concurrency and kernel threads provide parallelism.
1102Like 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:
1103\begin{cquote}
1104\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1105\begin{cfa}
1106thread myThread {
1107        // communication variables
1108};
1109
1110
1111\end{cfa}
1112&
1113\begin{cfa}
1114trait is_thread( `dtype` T ) {
1115      void main( T & );
1116      thread_desc * get_thread( T & );
1117      void ^?{}( T & `mutex` );
1118};
1119\end{cfa}
1120\end{tabular}
1121\end{cquote}
1122(The qualifier @mutex@ for the destructor parameter is discussed in Section~\ref{s:Monitor}.)
1123Like a coroutine, the statically-typed @main@ routine is the starting point (first stack frame) of a user thread.
1124The difference is that a coroutine borrows a thread from its caller, so the first thread resuming a coroutine creates an instance of @main@;
1125whereas, 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{
1126The \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.}
1127No 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.
1128
1129\begin{comment} % put in appendix with coroutine version ???
1130As such the @main@ routine of a thread can be defined as
1131\begin{cfa}
1132thread foo {};
1133
1134void main(foo & this) {
1135        sout | "Hello World!";
1136}
1137\end{cfa}
1138
1139In 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.
1140With 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.
1141\begin{cfa}
1142typedef void (*voidRtn)(int);
1143
1144thread RtnRunner {
1145        voidRtn func;
1146        int arg;
1147};
1148
1149void ?{}(RtnRunner & this, voidRtn inRtn, int arg) {
1150        this.func = inRtn;
1151        this.arg  = arg;
1152}
1153
1154void main(RtnRunner & this) {
1155        // thread starts here and runs the routine
1156        this.func( this.arg );
1157}
1158
1159void hello(/*unused*/ int) {
1160        sout | "Hello World!";
1161}
1162
1163int main() {
1164        RtnRunner f = {hello, 42};
1165        return 0?
1166}
1167\end{cfa}
1168A 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}.
1169\end{comment}
1170
1171For user threads to be useful, it must be possible to start and stop the underlying thread, and wait for it to complete execution.
1172While using an API such as @fork@ and @join@ is relatively common, such an interface is awkward and unnecessary.
1173A simple approach is to use allocation/deallocation principles, and have threads implicitly @fork@ after construction and @join@ before destruction.
1174\begin{cfa}
1175thread World {};
1176void main( World & this ) {
1177        sout | "World!";
1178}
1179int main() {
1180        World w`[10]`;                                                  $\C{// implicit forks after creation}$
1181        sout | "Hello ";                                        $\C{// "Hello " and 10 "World!" printed concurrently}$
1182}                                                                                       $\C{// implicit joins before destruction}$
1183\end{cfa}
1184This semantics ensures a thread is started and stopped exactly once, eliminating some programming error, and scales to multiple threads for basic (termination) synchronization.
1185This 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.
1186\begin{cfa}
1187int main() {
1188        MyThread * heapLive;
1189        {
1190                MyThread blockLive;                                     $\C{// fork block-based thread}$
1191                heapLive = `new`( MyThread );           $\C{// fork heap-based thread}$
1192                ...
1193        }                                                                               $\C{// join block-based thread}$
1194        ...
1195        `delete`( heapLive );                                   $\C{// join heap-based thread}$
1196}
1197\end{cfa}
1198The heap-based approach allows arbitrary thread-creation topologies, with respect to fork/join-style concurrency.
1199
1200Figure~\ref{s:ConcurrentMatrixSummation} shows concurrently adding the rows of a matrix and then totalling the subtotals sequentially, after all the row threads have terminated.
1201The program uses heap-based threads because each thread needs different constructor values.
1202(Python provides a simple iteration mechanism to initialize array elements to different values allowing stack allocation.)
1203The allocation/deallocation pattern appears unusual because allocated objects are immediately deallocated without any intervening code.
1204However, for threads, the deletion provides implicit synchronization, which is the intervening code.
1205While the subtotals are added in linear order rather than completion order, which slightly inhibits concurrency, the computation is restricted by the critical-path thread (\ie the thread that takes the longest), and so any inhibited concurrency is very small as totalling the subtotals is trivial.
1206
1207\begin{figure}
1208\begin{cfa}
1209`thread` Adder { int * row, cols, & subtotal; } $\C{// communication variables}$
1210void ?{}( Adder & adder, int row[], int cols, int & subtotal ) {
1211    adder.[ row, cols, &subtotal ] = [ row, cols, &subtotal ];
1212}
1213void main( Adder & adder ) with( adder ) {
1214    subtotal = 0;
1215    for ( int c = 0; c < cols; c += 1 ) { subtotal += row[c]; }
1216}
1217int main() {
1218    const int rows = 10, cols = 1000;
1219    int matrix[rows][cols], subtotals[rows], total = 0;
1220    // read matrix
1221    Adder * adders[rows];
1222    for ( int r = 0; r < rows; r += 1 ) {       $\C{// start threads to sum rows}$
1223                adders[r] = `new( matrix[r], cols, &subtotals[r] );`
1224    }
1225    for ( int r = 0; r < rows; r += 1 ) {       $\C{// wait for threads to finish}$
1226                `delete( adders[r] );`                          $\C{// termination join}$
1227                total += subtotals[r];                          $\C{// total subtotal}$
1228    }
1229    sout | total;
1230}
1231\end{cfa}
1232\caption{Concurrent Matrix Summation}
1233\label{s:ConcurrentMatrixSummation}
1234\end{figure}
1235
1236
1237\section{Mutual Exclusion / Synchronization}
1238
1239Uncontrolled non-deterministic execution is meaningless.
1240To 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}.
1241Since 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).
1242In these paradigms, interaction among concurrent objects is performed by stateless message-passing~\cite{Thoth,Harmony,V-Kernel} or other paradigms closely related to networking concepts, \eg channels~\cite{CSP,Go}.
1243However, 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.
1244Hence, a programmer must learn and manipulate two sets of design patterns.
1245While this distinction can be hidden away in library code, effective use of the library still has to take both paradigms into account.
1246In contrast, approaches based on stateful models more closely resemble the standard call/return programming-model, resulting in a single programming paradigm.
1247
1248At 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}.
1249However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
1250A newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
1251While this approach is pursued in hardware~\cite{Nakaike15} and system languages, like \CC~\cite{Cpp-Transactions}, the performance and feature set is still too restrictive to be the main concurrency paradigm for system languages, which is why it is rejected as the core paradigm for concurrency in \CFA.
1252
1253One of the most natural, elegant, and efficient mechanisms for mutual exclusion and synchronization for shared-memory systems is the \emph{monitor}.
1254First 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}.
1255In 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.
1256For these reasons, \CFA selected monitors as the core high-level concurrency-construct, upon which higher-level approaches can be easily constructed.
1257
1258
1259\subsection{Mutual Exclusion}
1260
1261A group of instructions manipulating a specific instance of shared data that must be performed atomically is called an (individual) \newterm{critical-section}~\cite{Dijkstra65}.
1262The 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.
1263The 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.
1264\newterm{Mutual exclusion} enforces that the correct kind and number of threads are using a critical section.
1265
1266However, many solutions exist for mutual exclusion, which vary in terms of performance, flexibility and ease of use.
1267Methods 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.
1268Ease 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.
1269For 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.
1270However, a significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
1271Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
1272
1273
1274\subsection{Synchronization}
1275
1276Synchronization enforces relative ordering of execution, and synchronization tools provide numerous mechanisms to establish these timing relationships.
1277Low-level synchronization primitives offer good performance and flexibility at the cost of ease of use;
1278higher-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.
1279Often 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.
1280If a writer thread is scheduled for next access, but another reader thread acquires the critical section first, that reader \newterm{barged}.
1281Barging 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).
1282Preventing or detecting barging is an involved challenge with low-level locks, which can be made much easier by higher-level constructs.
1283This challenge is often split into two different approaches: barging avoidance and barging prevention.
1284Algorithms that allow a barger, but divert it until later using current synchronization state (flags), are avoiding the barger;
1285algorithms that preclude a barger from entering during synchronization in the critical section prevent barging completely.
1286Techniques like baton-passing locks~\cite{Andrews89} between threads instead of unconditionally releasing locks is an example of barging prevention.
1287
1288
1289\section{Monitor}
1290\label{s:Monitor}
1291
1292A \textbf{monitor} is a set of routines that ensure mutual exclusion when accessing shared state.
1293More 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).
1294The strong association with the call/return paradigm eases programmability, readability and maintainability, at a slight cost in flexibility and efficiency.
1295
1296Note, 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.
1297Copying 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.
1298Copying 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.
1299As for coroutines/tasks, the @dtype@ property provides no implicit copying operations and the @is_monitor@ trait provides no explicit copying operations, so all locks/monitors must be passed by reference (pointer).
1300\begin{cfa}
1301trait is_monitor( `dtype` T ) {
1302        monitor_desc * get_monitor( T & );
1303        void ^?{}( T & mutex );
1304};
1305\end{cfa}
1306
1307
1308\subsection{Mutex Acquisition}
1309\label{s:MutexAcquisition}
1310
1311While correctness implies a monitor's mutual exclusion is acquired and released, there are implementation options about when and where the locking/unlocking occurs.
1312(Much of this discussion also applies to basic locks.)
1313For example, a monitor may need to be passed through multiple helper routines before it becomes necessary to acquire the monitor mutual-exclusion.
1314\begin{cfa}[morekeywords=nomutex]
1315monitor Aint { int cnt; };                                      $\C{// atomic integer counter}$
1316void ?{}( Aint & `nomutex` this ) with( this ) { cnt = 0; } $\C{// constructor}$
1317int ?=?( Aint & `mutex`$\(_{opt}\)$ lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions}$
1318void ?{}( int & this, Aint & `mutex`$\(_{opt}\)$ v ) { this = v.cnt; }
1319int ?=?( int & lhs, Aint & `mutex`$\(_{opt}\)$ rhs ) with( rhs ) { lhs = cnt; }
1320int ++?( Aint & `mutex`$\(_{opt}\)$ this ) with( this ) { return ++cnt; } $\C{// increment}$
1321\end{cfa}
1322The @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.
1323(While a constructor may publish its address into a global variable, doing so generates a race-condition.)
1324The conversion operators for initializing and assigning with a normal integer only need @mutex@, if reading/writing the implementation type is not atomic.
1325Finally, 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.
1326
1327The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics, which is similar to the \CC template @std::atomic@.
1328\begin{cfa}
1329Aint x, y, z;
1330++x; ++y; ++z;                                                          $\C{// safe increment by multiple threads}$
1331x = 2; y = 2; z = 2;                                            $\C{// conversions}$
1332int i = x, j = y, k = z;
1333i = x; j = y; k = z;
1334\end{cfa}
1335
1336For maximum usability, monitors have \newterm{multi-acquire} semantics allowing a thread to acquire it multiple times without deadlock.
1337\begin{cfa}
1338monitor M { ... } m;
1339void foo( M & mutex m ) { ... }                         $\C{// acquire mutual exclusion}$
1340void bar( M & mutex m ) {                                       $\C{// acquire mutual exclusion}$
1341        ... `foo( m );` ...                                             $\C{// reacquire mutual exclusion}$
1342}
1343`bar( m );`                                                                     $\C{// nested monitor call}$
1344\end{cfa}
1345
1346The benefit of mandatory monitor qualifiers is self-documentation, but requiring both @mutex@ and \lstinline[morekeywords=nomutex]@nomutex@ for all monitor parameters is redundant.
1347Instead, the semantics have one qualifier as the default, and the other required.
1348For example, make the safe @mutex@ qualifier the default because assuming \lstinline[morekeywords=nomutex]@nomutex@ may cause subtle errors.
1349Alternatively, make the unsafe \lstinline[morekeywords=nomutex]@nomutex@ qualifier the default because it is the \emph{normal} parameter semantics while @mutex@ parameters are rare.
1350Providing a default qualifier implies knowing whether a parameter is a monitor.
1351Since \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.
1352For 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@.
1353
1354The next semantic decision is establishing which parameter \emph{types} may be qualified with @mutex@.
1355Given:
1356\begin{cfa}
1357monitor M { ... }
1358int f1( M & mutex m );
1359int f2( M * mutex m );
1360int f3( M * mutex m[] );
1361int f4( stack( M * ) & mutex m );
1362\end{cfa}
1363the issue is that some of these parameter types are composed of multiple objects.
1364For @f1@, there is only a single parameter object.
1365Adding indirection in @f2@ still identifies a single object.
1366However, the matrix in @f3@ introduces multiple objects.
1367While shown shortly, multiple acquisition is possible;
1368however array lengths are often unknown in C.
1369This issue is exacerbated in @f4@, where the data structure must be safely traversed to acquire all of its elements.
1370
1371To make the issue tractable, \CFA only acquires one monitor per parameter with at most one level of indirection.
1372However, there is an ambiguity in the C type-system with respects to arrays.
1373Is the argument for @f2@ a single object or an array of objects?
1374If it is an array, only the first element of the array is acquired, which seems unsafe;
1375hence, @mutex@ is disallowed for array parameters.
1376\begin{cfa}
1377int f1( M & mutex m );                                          $\C{// allowed: recommended case}$
1378int f2( M * mutex m );                                          $\C{// disallowed: could be an array}$
1379int f3( M mutex m[$\,$] );                                      $\C{// disallowed: array length unknown}$
1380int f4( M ** mutex m );                                         $\C{// disallowed: could be an array}$
1381int f5( M * mutex m[$\,$] );                            $\C{// disallowed: array length unknown}$
1382\end{cfa}
1383% Note, not all array routines have distinct types: @f2@ and @f3@ have the same type, as do @f4@ and @f5@.
1384% However, even if the code generation could tell the difference, the extra information is still not sufficient to extend meaningfully the monitor call semantic.
1385
1386For object-oriented monitors, calling a mutex member \emph{implicitly} acquires mutual exclusion of the receiver object, @`rec`.foo(...)@.
1387\CFA has no receiver, and hence, must use an explicit mechanism to specify which object acquires mutual exclusion.
1388A positive consequence of this design decision is the ability to support multi-monitor routines.
1389\begin{cfa}
1390int f( M & mutex x, M & mutex y );              $\C{// multiple monitor parameter of any type}$
1391M m1, m2;
1392f( m1, m2 );
1393\end{cfa}
1394(While object-oriented monitors can be extended with a mutex qualifier for multiple-monitor members, no prior example of this feature could be found.)
1395In practice, writing multi-locking routines that do not deadlock is tricky.
1396Having language support for such a feature is therefore a significant asset for \CFA.
1397
1398The capability to acquire multiple locks before entering a critical section is called \newterm{bulk acquire} (see Section~\ref{s:Implementation} for implementation details).
1399In the previous example, \CFA guarantees the order of acquisition is consistent across calls to different routines using the same monitors as arguments.
1400This consistent ordering means acquiring multiple monitors is safe from deadlock.
1401However, users can force the acquiring order.
1402For example, notice the use of @mutex@/\lstinline[morekeywords=nomutex]@nomutex@ and how this affects the acquiring order:
1403\begin{cfa}
1404void foo( M & mutex m1, M & mutex m2 );         $\C{// acquire m1 and m2}$
1405void bar( M & mutex m1, M & /* nomutex */ m2 ) { $\C{// acquire m1}$
1406        ... foo( m1, m2 ); ...                                  $\C{// acquire m2}$
1407}
1408void baz( M & /* nomutex */ m1, M & mutex m2 ) { $\C{// acquire m2}$
1409        ... foo( m1, m2 ); ...                                  $\C{// acquire m1}$
1410}
1411\end{cfa}
1412The multi-acquire semantics allows @bar@ or @baz@ to acquire a monitor lock and reacquire it in @foo@.
1413In the calls to @bar@ and @baz@, the monitors are acquired in opposite order.
1414
1415However, 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}.
1416In \CFA, a safety aid is provided by using bulk acquire of all monitors to shared objects, whereas other monitor systems provide no aid.
1417While \CFA provides only a partial solution, it handles many useful cases, \eg:
1418\begin{cfa}
1419monitor BankAccount { ... };
1420void deposit( BankAccount & `mutex` b, int deposit );
1421void transfer( BankAccount & `mutex` my, BankAccount & `mutex` your, int me2you ) {
1422        deposit( my, `-`me2you );                               $\C{// debit}$
1423        deposit( your, me2you );                                $\C{// credit}$
1424}
1425\end{cfa}
1426This example shows a trivial solution to the bank-account transfer problem.
1427Without multi- and bulk acquire, the solution to this problem requires careful engineering.
1428
1429
1430\subsection{\protect\lstinline@mutex@ statement}
1431\label{mutex-stmt}
1432
1433The monitor call-semantics associate all locking semantics to routines.
1434Like Java, \CFA offers an alternative @mutex@ statement to reduce refactoring and naming.
1435\begin{cquote}
1436\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1437\begin{cfa}
1438monitor M { ... };
1439void foo( M & mutex m1, M & mutex m2 ) {
1440        // critical section
1441}
1442void bar( M & m1, M & m2 ) {
1443        foo( m1, m2 );
1444}
1445\end{cfa}
1446&
1447\begin{cfa}
1448
1449void bar( M & m1, M & m2 ) {
1450        mutex( m1, m2 ) {       // remove refactoring and naming
1451                // critical section
1452        }
1453}
1454
1455\end{cfa}
1456\\
1457\multicolumn{1}{c}{\textbf{routine call}} & \multicolumn{1}{c}{\lstinline@mutex@ \textbf{statement}}
1458\end{tabular}
1459\end{cquote}
1460
1461
1462\section{Scheduling}
1463\label{s:Scheduling}
1464
1465While monitor mutual-exclusion provides safe access to shared data, the monitor data may indicate that a thread accessing it cannot proceed.
1466For example, Figure~\ref{f:GenericBoundedBuffer} shows a bounded buffer that may be full/empty so produce/consumer threads must block.
1467Leaving the monitor and trying again (busy waiting) is impractical for high-level programming.
1468Monitors eliminate busy waiting by providing synchronization to schedule threads needing access to the shared data, where threads block versus spinning.
1469Synchronization 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.
1470\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.
1471
1472Figure~\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@.
1473The @wait@ routine atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the routine's parameter list.
1474The appropriate condition lock is signalled to unblock an opposite kind of thread after an element is inserted/removed from the buffer.
1475Signalling is unconditional, because signalling an empty condition lock does nothing.
1476
1477Signalling semantics cannot have the signaller and signalled thread in the monitor simultaneously, which means:
1478\begin{enumerate}
1479\item
1480The signalling thread returns immediately, and the signalled thread continues.
1481\item
1482The signalling thread continues and the signalled thread is marked for urgent unblocking at the next scheduling point (exit/wait).
1483\item
1484The signalling thread blocks but is marked for urgrent unblocking at the next scheduling point and the signalled thread continues.
1485\end{enumerate}
1486The first approach is too restrictive, as it precludes solving a reasonable class of problems, \eg dating service (see Figure~\ref{f:DatingService}).
1487\CFA supports the next two semantics as both are useful.
1488Finally, while it is common to store a @condition@ as a field of the monitor, in \CFA, a @condition@ variable can be created/stored independently.
1489Furthermore, 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.
1490
1491\begin{figure}
1492\centering
1493\newbox\myboxA
1494\begin{lrbox}{\myboxA}
1495\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1496forall( otype T ) { // distribute forall
1497        monitor Buffer {
1498                `condition` full, empty;
1499                int front, back, count;
1500                T elements[10];
1501        };
1502        void ?{}( Buffer(T) & buffer ) with(buffer) {
1503                [front, back, count] = 0;
1504        }
1505
1506        void insert( Buffer(T) & mutex buffer, T elem )
1507                                with(buffer) {
1508                if ( count == 10 ) `wait( empty )`;
1509                // insert elem into buffer
1510                `signal( full )`;
1511        }
1512        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1513                if ( count == 0 ) `wait( full )`;
1514                // remove elem from buffer
1515                `signal( empty )`;
1516                return elem;
1517        }
1518}
1519\end{cfa}
1520\end{lrbox}
1521
1522\newbox\myboxB
1523\begin{lrbox}{\myboxB}
1524\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1525forall( otype T ) { // distribute forall
1526        monitor Buffer {
1527
1528                int front, back, count;
1529                T elements[10];
1530        };
1531        void ?{}( Buffer(T) & buffer ) with(buffer) {
1532                [front, back, count] = 0;
1533        }
1534        T remove( Buffer(T) & mutex buffer ); // forward
1535        void insert( Buffer(T) & mutex buffer, T elem )
1536                                with(buffer) {
1537                if ( count == 10 ) `waitfor( remove, buffer )`;
1538                // insert elem into buffer
1539
1540        }
1541        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1542                if ( count == 0 ) `waitfor( insert, buffer )`;
1543                // remove elem from buffer
1544
1545                return elem;
1546        }
1547}
1548\end{cfa}
1549\end{lrbox}
1550
1551\subfloat[Internal Scheduling]{\label{f:BBInt}\usebox\myboxA}
1552%\qquad
1553\subfloat[External Scheduling]{\label{f:BBExt}\usebox\myboxB}
1554\caption{Generic Bounded-Buffer}
1555\label{f:GenericBoundedBuffer}
1556\end{figure}
1557
1558Figure~\ref{f:BBExt} shows a \CFA generic bounded-buffer with external scheduling, where producers/consumers detecting a full/empty buffer block and prevent more producers/consumers from entering the monitor until there is a free/empty slot in the buffer.
1559External 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.
1560If 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.
1561Threads making calls to routines that are currently excluded, block outside of (external to) the monitor on a calling queue, versus blocking on condition queues inside of (internal to) the monitor.
1562External scheduling allows users to wait for events from other threads without concern of unrelated events occurring.
1563The mechnaism can be done in terms of control flow, \eg Ada @accept@ or \uC @_Accept@, or in terms of data, \eg Go channels.
1564While both mechanisms have strengths and weaknesses, this project uses a control-flow mechanism to stay consistent with other language semantics.
1565Two 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}).
1566
1567For internal scheduling, non-blocking signalling (as in the producer/consumer example) is used when the signaller is providing the cooperation for a waiting thread;
1568the 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.
1569The waiter unblocks next from the urgent queue, uses/takes the state, and exits the monitor.
1570Blocking signalling is the reverse, where the waiter is providing the cooperation for the signalling thread;
1571the 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.
1572The waiter changes state and exits the monitor, and the signaller unblocks next from the urgent queue to use/take the state.
1573
1574Figure~\ref{f:DatingService} shows a dating service demonstrating non-blocking and blocking signalling.
1575The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
1576A thread blocks until an appropriate partner arrives.
1577The complexity is exchanging phone numbers in the monitor because of the mutual-exclusion property.
1578For signal scheduling, the @exchange@ condition is necessary to block the thread finding the match, while the matcher unblocks to take the opposite number, post its phone number, and unblock the partner.
1579For 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.
1580
1581The 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;
1582as well, an arriving thread may not find a partner and must wait, which requires a condition variable, and condition variables imply internal scheduling.
1583
1584\begin{figure}
1585\centering
1586\newbox\myboxA
1587\begin{lrbox}{\myboxA}
1588\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1589enum { CCodes = 20 };
1590monitor DS {
1591        int GirlPhNo, BoyPhNo;
1592        condition Girls[CCodes], Boys[CCodes];
1593        condition exchange;
1594};
1595int girl( DS & mutex ds, int phNo, int ccode ) {
1596        if ( is_empty( Boys[ccode] ) ) {
1597                wait( Girls[ccode] );
1598                GirlPhNo = phNo;
1599                `signal( exchange );`
1600        } else {
1601                GirlPhNo = phNo;
1602                `signal( Boys[ccode] );`
1603                `wait( exchange );`
1604        } // if
1605        return BoyPhNo;
1606}
1607int boy( DS & mutex ds, int phNo, int ccode ) {
1608        // as above with boy/girl interchanged
1609}
1610\end{cfa}
1611\end{lrbox}
1612
1613\newbox\myboxB
1614\begin{lrbox}{\myboxB}
1615\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1616
1617monitor DS {
1618        int GirlPhNo, BoyPhNo;
1619        condition Girls[CCodes], Boys[CCodes];
1620
1621};
1622int girl( DS & mutex ds, int phNo, int ccode ) {
1623        if ( is_empty( Boys[ccode] ) ) { // no compatible
1624                wait( Girls[ccode] ); // wait for boy
1625                GirlPhNo = phNo; // make phone number available
1626
1627        } else {
1628                GirlPhNo = phNo; // make phone number available
1629                `signal_block( Boys[ccode] );` // restart boy
1630
1631        } // if
1632        return BoyPhNo;
1633}
1634int boy( DS & mutex ds, int phNo, int ccode ) {
1635        // as above with boy/girl interchanged
1636}
1637\end{cfa}
1638\end{lrbox}
1639
1640\subfloat[\lstinline@signal@]{\label{f:DatingSignal}\usebox\myboxA}
1641\qquad
1642\subfloat[\lstinline@signal_block@]{\label{f:DatingSignalBlock}\usebox\myboxB}
1643\caption{Dating service. }
1644\label{f:DatingService}
1645\end{figure}
1646
1647Both internal and external scheduling extend to multiple monitors in a natural way.
1648\begin{cquote}
1649\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1650\begin{cfa}
1651monitor M { `condition e`; ... };
1652void foo( M & mutex m1, M & mutex m2 ) {
1653        ... wait( `e` ); ...   // wait( e, m1, m2 )
1654        ... wait( `e, m1` ); ...
1655        ... wait( `e, m2` ); ...
1656}
1657\end{cfa}
1658&
1659\begin{cfa}
1660void rtn$\(_1\)$( M & mutex m1, M & mutex m2 );
1661void rtn$\(_2\)$( M & mutex m1 );
1662void bar( M & mutex m1, M & mutex m2 ) {
1663        ... waitfor( `rtn` ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\), m1, m2 )}$
1664        ... waitfor( `rtn, m1` ); ... // $\LstCommentStyle{waitfor( rtn\(_2\), m1 )}$
1665}
1666\end{cfa}
1667\end{tabular}
1668\end{cquote}
1669For @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 )@.
1670To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @wait( e, m1 )@.
1671Wait statically verifies the released monitors are the acquired mutex-parameters so unconditional release is safe.
1672Finally, a signaller,
1673\begin{cfa}
1674void baz( M & mutex m1, M & mutex m2 ) {
1675        ... signal( e ); ...
1676}
1677\end{cfa}
1678must have acquired at least the same locks as the waiting thread signalled from the condition queue.
1679
1680Similarly, 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 )@.
1681To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn, m1 )@.
1682@waitfor@ statically verifies the released monitors are the same as the acquired mutex-parameters of the given routine or routine pointer.
1683To statically verify the released monitors match with the accepted routine's mutex parameters, the routine (pointer) prototype must be accessible.
1684% When an overloaded routine appears in an @waitfor@ statement, calls to any routine with that name are accepted.
1685% The rationale is that members with the same name should perform a similar function, and therefore, all should be eligible to accept a call.
1686Overloaded routines can be disambiguated using a cast:
1687\begin{cfa}
1688void rtn( M & mutex m );
1689`int` rtn( M & mutex m );
1690waitfor( (`int` (*)( M & mutex ))rtn, m );
1691\end{cfa}
1692
1693The ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock.
1694\begin{cfa}
1695void foo( M & mutex m1, M & mutex m2 ) {
1696        ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired )}$
1697void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2 )}$
1698        ... signal( `e` ); ...
1699\end{cfa}
1700The @wait@ only releases @m1@ so the signalling thread cannot acquire both @m1@ and @m2@ to  enter @bar@ to get to the @signal@.
1701While deadlock issues can occur with multiple/nesting acquisition, this issue results from the fact that locks, and by extension monitors, are not perfectly composable.
1702
1703Finally, an important aspect of monitor implementation is barging, \ie can calling threads barge ahead of signalled threads?
1704If barging is allowed, synchronization between a signaller and signallee is difficult, often requiring multiple unblock/block cycles (looping around a wait rechecking if a condition is met).
1705In fact, signals-as-hints is completely opposite from that proposed by Hoare in the seminal paper on monitors:
1706\begin{quote}
1707However, 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.
1708It 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}
1709\end{quote}
1710\CFA scheduling \emph{precludes} barging, which simplifies synchronization among threads in the monitor and increases correctness.
1711Furthermore, \CFA concurrency has no spurious wakeup~\cite[\S~9]{Buhr05a}, which eliminates an implict form of barging.
1712For example, there are no loops in either bounded buffer solution in Figure~\ref{f:GenericBoundedBuffer}.
1713Supporting 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.
1714
1715
1716\subsection{Barging Prevention}
1717
1718Figure~\ref{f:BargingPrevention} shows \CFA code where bulk acquire adds complexity to the internal-signalling semantics.
1719The complexity begins at the end of the inner @mutex@ statement, where the semantics of internal scheduling need to be extended for multiple monitors.
1720The problem is that bulk acquire is used in the inner @mutex@ statement where one of the monitors is already acquired.
1721When 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.
1722However, both the signalling and waiting thread W1 still need monitor @m1@.
1723
1724\begin{figure}
1725\newbox\myboxA
1726\begin{lrbox}{\myboxA}
1727\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1728monitor M m1, m2;
1729condition c;
1730mutex( m1 ) { // $\LstCommentStyle{\color{red}outer}$
1731        ...
1732        mutex( m1, m2 ) { // $\LstCommentStyle{\color{red}inner}$
1733                ... `signal( c )`; ...
1734                // m1, m2 acquired
1735        } // $\LstCommentStyle{\color{red}release m2}$
1736        // m1 acquired
1737} // release m1
1738\end{cfa}
1739\end{lrbox}
1740
1741\newbox\myboxB
1742\begin{lrbox}{\myboxB}
1743\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1744
1745
1746mutex( m1 ) {
1747        ...
1748        mutex( m1, m2 ) {
1749                ... `wait( c )`; // block and release m1, m2
1750                // m1, m2 acquired
1751        } // $\LstCommentStyle{\color{red}release m2}$
1752        // m1 acquired
1753} // release m1
1754\end{cfa}
1755\end{lrbox}
1756
1757\newbox\myboxC
1758\begin{lrbox}{\myboxC}
1759\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1760
1761
1762mutex( m2 ) {
1763        ... `wait( c )`; ...
1764        // m2 acquired
1765} // $\LstCommentStyle{\color{red}release m2}$
1766
1767
1768
1769
1770\end{cfa}
1771\end{lrbox}
1772
1773\begin{cquote}
1774\subfloat[Signalling Thread]{\label{f:SignallingThread}\usebox\myboxA}
1775\hspace{2\parindentlnth}
1776\subfloat[Waiting Thread (W1)]{\label{f:WaitingThread}\usebox\myboxB}
1777\hspace{2\parindentlnth}
1778\subfloat[Waiting Thread (W2)]{\label{f:OtherWaitingThread}\usebox\myboxC}
1779\end{cquote}
1780\caption{Barging Prevention}
1781\label{f:BargingPrevention}
1782\end{figure}
1783
1784One 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.
1785However, Figure~\ref{f:OtherWaitingThread} shows this solution is complex depending on other waiters, resulting in options when the signaller finishes the inner mutex-statement.
1786The signaller can retain @m2@ until completion of the outer mutex statement and pass the locks to waiter W1, or it can pass @m2@ to waiter W2 after completing the inner mutex-statement, while continuing to hold @m1@.
1787In 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.
1788Furthermore, there is an execution sequence where the signaller always finds waiter W2, and hence, waiter W1 starves.
1789
1790While a number of approaches were examined~\cite[\S~4.3]{Delisle18}, the solution chosen for \CFA is a novel techique called \newterm{partial signalling}.
1791Signalled threads are moved to the urgent queue and the waiter at the front defines the set of monitors necessary for it to unblock.
1792Partial signalling transfers ownership of monitors to the front waiter.
1793When the signaller thread exits or waits in the monitor, the front waiter is unblocked if all its monitors are released.
1794The benefit is encapsulating complexity into only two actions: passing monitors to the next owner when they should be released and conditionally waking threads if all conditions are met.
1795
1796
1797\subsection{Loose Object Definitions}
1798\label{s:LooseObjectDefinitions}
1799
1800In an object-oriented programming-language, a class includes an exhaustive list of operations.
1801However, new members can be added via static inheritance or dynamic members, \eg JavaScript~\cite{JavaScript}.
1802Similarly, monitor routines can be added at any time in \CFA, making it less clear for programmers and more difficult to implement.
1803\begin{cfa}
1804monitor M { ... };
1805void `f`( M & mutex m );
1806void g( M & mutex m ) { waitfor( `f` ); }       $\C{// clear which f}$
1807void `f`( M & mutex m, int );                           $\C{// different f}$
1808void h( M & mutex m ) { waitfor( `f` ); }       $\C{// unclear which f}$
1809\end{cfa}
1810Hence, the cfa-code for entering a monitor looks like:
1811\begin{cfa}
1812if ( $\textrm{\textit{monitor is free}}$ ) $\LstCommentStyle{// \color{red}enter}$
1813else if ( $\textrm{\textit{already own monitor}}$ ) $\LstCommentStyle{// \color{red}continue}$
1814else if ( $\textrm{\textit{monitor accepts me}}$ ) $\LstCommentStyle{// \color{red}enter}$
1815else $\LstCommentStyle{// \color{red}block}$
1816\end{cfa}
1817For the first two conditions, it is easy to implement a check that can evaluate the condition in a few instructions.
1818However, a fast check for \emph{monitor accepts me} is much harder to implement depending on the constraints put on the monitors.
1819Figure~\ref{fig:ClassicalMonitor} shows monitors are often expressed as an entry (calling) queue, some acceptor queues, and an urgent stack/queue.
1820
1821\begin{figure}
1822\centering
1823\subfloat[Classical monitor] {
1824\label{fig:ClassicalMonitor}
1825{\resizebox{0.45\textwidth}{!}{\input{monitor.pstex_t}}}
1826}% subfloat
1827\quad
1828\subfloat[Bulk acquire monitor] {
1829\label{fig:BulkMonitor}
1830{\resizebox{0.45\textwidth}{!}{\input{ext_monitor.pstex_t}}}
1831}% subfloat
1832\caption{Monitor Implementation}
1833\label{f:MonitorImplementation}
1834\end{figure}
1835
1836For 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.
1837This approach requires a unique dense ordering of routines with a small upper-bound and the ordering must be consistent across translation units.
1838For 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.
1839
1840Figure~\ref{fig:BulkMonitor} shows the \CFA monitor implementation.
1841The mutex routine called is associated with each thread on the entry queue, while a list of acceptable routines is kept separately.
1842The 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.
1843
1844
1845\subsection{Multi-Monitor Scheduling}
1846\label{s:Multi-MonitorScheduling}
1847
1848External scheduling, like internal scheduling, becomes significantly more complex for multi-monitor semantics.
1849Even in the simplest case, new semantics needs to be established.
1850\newpage
1851\begin{cfa}
1852monitor M { ... };
1853void f( M & mutex m1 );
1854void g( M & mutex m1, M & mutex m2 ) {
1855        waitfor( f );                                                   $\C{\color{red}// pass m1 or m2 to f?}$
1856}
1857\end{cfa}
1858The solution is for the programmer to disambiguate:
1859\begin{cfa}
1860        waitfor( f, m2 );                                               $\C{\color{red}// wait for call to f with argument m2}$
1861\end{cfa}
1862Both locks are acquired by routine @g@, so when routine @f@ is called, the lock for monitor @m2@ is passed from @g@ to @f@, while @g@ still holds lock @m1@.
1863This behaviour can be extended to the multi-monitor @waitfor@ statement.
1864\begin{cfa}
1865monitor M { ... };
1866void f( M & mutex m1, M & mutex m2 );
1867void g( M & mutex m1, M & mutex m2 ) {
1868        waitfor( f, m1, m2 );                                   $\C{\color{red}// wait for call to f with arguments m1 and m2}$
1869}
1870\end{cfa}
1871Again, the set of monitors passed to the @waitfor@ statement must be entirely contained in the set of monitors already acquired by the accepting routine.
1872Also, the order of the monitors in a @waitfor@ statement is unimportant.
1873
1874Figure~\ref{f:UnmatchedMutexSets} shows an example where, for internal and external scheduling with multiple monitors, a signalling or accepting thread must match exactly, \ie partial matching results in waiting.
1875For both examples, the set of monitors is disjoint so unblocking is impossible.
1876
1877\begin{figure}
1878\lstDeleteShortInline@%
1879\begin{tabular}{@{}l@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
1880\begin{cfa}
1881monitor M1 {} m11, m12;
1882monitor M2 {} m2;
1883condition c;
1884void f( M1 & mutex m1, M2 & mutex m2 ) {
1885        signal( c );
1886}
1887void g( M1 & mutex m1, M2 & mutex m2 ) {
1888        wait( c );
1889}
1890g( `m11`, m2 ); // block on wait
1891f( `m12`, m2 ); // cannot fulfil
1892\end{cfa}
1893&
1894\begin{cfa}
1895monitor M1 {} m11, m12;
1896monitor M2 {} m2;
1897
1898void f( M1 & mutex m1, M2 & mutex m2 ) {
1899
1900}
1901void g( M1 & mutex m1, M2 & mutex m2 ) {
1902        waitfor( f, m1, m2 );
1903}
1904g( `m11`, m2 ); // block on accept
1905f( `m12`, m2 ); // cannot fulfil
1906\end{cfa}
1907\end{tabular}
1908\lstMakeShortInline@%
1909\caption{Unmatched \protect\lstinline@mutex@ sets}
1910\label{f:UnmatchedMutexSets}
1911\end{figure}
1912
1913
1914\subsection{Extended \protect\lstinline@waitfor@}
1915
1916Figure~\ref{f:ExtendedWaitfor} show the extended form of the @waitfor@ statement to conditionally accept one of a group of mutex routines, with a specific action to be performed \emph{after} the mutex routine finishes.
1917For a @waitfor@ clause to be executed, its @when@ must be true and an outstanding call to its corresponding member(s) must exist.
1918The \emph{conditional-expression} of a @when@ may call a routine, but the routine must not block or context switch.
1919If there are multiple acceptable mutex calls, selection occurs top-to-bottom (prioritized) in the @waitfor@ clauses, whereas some programming languages with similar mechanisms accept non-deterministically for this case, \eg Go \lstinline[morekeywords=select]@select@.
1920If 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.
1921If 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.
1922Hence, the terminating @else@ clause allows a conditional attempt to accept a call without blocking.
1923If there is a @timeout@ clause, it provides an upper bound on waiting.
1924If both a @timeout@ clause and an @else@ clause are present, the @else@ must be conditional, or the @timeout@ is never triggered.
1925In all cases, the statement following is executed \emph{after} a clause is executed to know which of the clauses executed.
1926
1927\begin{figure}
1928\begin{cfa}
1929`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1930        waitfor( $\emph{mutex-member-name}$ )
1931                $\emph{statement}$                                      $\C{// action after call}$
1932`or` `when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
1933        waitfor( $\emph{mutex-member-name}$ )
1934                $\emph{statement}$                                      $\C{// action after call}$
1935`or`    ...                                                                     $\C{// list of waitfor clauses}$
1936`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1937        `timeout`                                                               $\C{// optional terminating timeout clause}$
1938                $\emph{statement}$                                      $\C{// action after timeout}$
1939`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
1940        `else`                                                                  $\C{// optional terminating clause}$
1941                $\emph{statement}$                                      $\C{// action when no immediate calls}$
1942\end{cfa}
1943\caption{Extended \protect\lstinline@waitfor@}
1944\label{f:ExtendedWaitfor}
1945\end{figure}
1946
1947Note, a group of conditional @waitfor@ clauses is \emph{not} the same as a group of @if@ statements, e.g.:
1948\begin{cfa}
1949if ( C1 ) waitfor( mem1 );                       when ( C1 ) waitfor( mem1 );
1950else if ( C2 ) waitfor( mem2 );         or when ( C2 ) waitfor( mem2 );
1951\end{cfa}
1952The left example accepts only @mem1@ if @C1@ is true or only @mem2@ if @C2@ is true.
1953The right example accepts either @mem1@ or @mem2@ if @C1@ and @C2@ are true.
1954
1955An interesting use of @waitfor@ is accepting the @mutex@ destructor to know when an object is deallocated.
1956\begin{cfa}
1957void insert( Buffer(T) & mutex buffer, T elem ) with( buffer ) {
1958        if ( count == 10 )
1959                waitfor( remove, buffer ) {
1960                        // insert elem into buffer
1961                } or `waitfor( ^?{}, buffer )` throw insertFail;
1962}
1963\end{cfa}
1964When the buffer is deallocated, the current waiter is unblocked and informed, so it can perform an appropriate action.
1965However, the basic @waitfor@ semantics do not support this functionality, since using an object after its destructor is called is undefined.
1966Therefore, to make this useful capability work, the semantics for accepting the destructor is the same as @signal@, \ie the call to the destructor is placed on the urgent queue and the acceptor continues execution, which throws an exception to the acceptor and then the caller is unblocked from the urgent queue to deallocate the object.
1967Accepting the destructor is an idiomatic way to terminate a thread in \CFA.
1968
1969
1970\subsection{\protect\lstinline@mutex@ Threads}
1971
1972Threads in \CFA are monitors to allow direct communication among threads, \ie threads can have mutex routines that are called by other threads.
1973Hence, all monitor features are available when using threads.
1974The following shows an example of two threads directly calling each other and accepting calls from each other in a cycle.
1975\begin{cfa}
1976thread Ping {} pi;
1977thread Pong {} po;
1978void ping( Ping & mutex ) {}
1979void pong( Pong & mutex ) {}
1980int main() {}
1981\end{cfa}
1982\vspace{-0.8\baselineskip}
1983\begin{cquote}
1984\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1985\begin{cfa}
1986void main( Ping & pi ) {
1987        for ( int i = 0; i < 10; i += 1 ) {
1988                `waitfor( ping, pi );`
1989                `pong( po );`
1990        }
1991}
1992\end{cfa}
1993&
1994\begin{cfa}
1995void main( Pong & po ) {
1996        for ( int i = 0; i < 10; i += 1 ) {
1997                `ping( pi );`
1998                `waitfor( pong, po );`
1999        }
2000}
2001\end{cfa}
2002\end{tabular}
2003\end{cquote}
2004% \lstMakeShortInline@%
2005% \caption{Threads ping/pong using external scheduling}
2006% \label{f:pingpong}
2007% \end{figure}
2008Note, the ping/pong threads are globally declared, @pi@/@po@, and hence, start (and possibly complete) before the program main starts.
2009
2010
2011\subsection{Low-level Locks}
2012
2013For completeness and efficiency, \CFA provides a standard set of low-level locks: recursive mutex, condition, semaphore, barrier, \etc, and atomic instructions: @fetchAssign@, @fetchAdd@, @testSet@, @compareSet@, \etc.
2014
2015
2016\section{Parallelism}
2017
2018Historically, computer performance was about processor speeds.
2019However, with heat dissipation being a direct consequence of speed increase, parallelism is the new source for increased performance~\cite{Sutter05, Sutter05b}.
2020Now, high-performance applications must care about parallelism, which requires concurrency.
2021The lowest-level approach of parallelism is to use \newterm{kernel threads} in combination with semantics like @fork@, @join@, \etc.
2022However, kernel threads are better as an implementation tool because of complexity and higher cost.
2023Therefore, different abstractions are often layered onto kernel threads to simplify them, \eg pthreads.
2024
2025
2026\subsection{User Threads with Preemption}
2027
2028A direct improvement on kernel threads is user threads, \eg Erlang~\cite{Erlang} and \uC~\cite{uC++book}.
2029This 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.
2030In many cases, user threads can be used on a much larger scale (100,000 threads).
2031Like kernel threads, user threads support preemption, which maximizes nondeterminism, but introduces the same concurrency errors: race, livelock, starvation, and deadlock.
2032\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}.
2033
2034
2035\subsection{User Threads without Preemption (Fiber)}
2036\label{s:fibers}
2037
2038A variant of user thread is \newterm{fibers}, which removes preemption, \eg Go~\cite{Go} @goroutine@s.
2039Like functional programming, which removes mutation and its associated problems, removing preemption from concurrency reduces nondeterminism, making race and deadlock errors more difficult to generate.
2040However, preemption is necessary for concurrency that relies on spinning, so there are a class of problems that cannot be programmed without preemption.
2041
2042
2043\subsection{Thread Pools}
2044
2045In contrast to direct threading is indirect \newterm{thread pools}, where small jobs (work units) are inserted into a work pool for execution.
2046If the jobs are dependent, \ie interact, there is an implicit/explicit dependency graph that ties them together.
2047While removing direct concurrency, and hence the amount of context switching, thread pools significantly limit the interaction that can occur among jobs.
2048Indeed, jobs should not block because that also blocks the underlying thread, which effectively means the CPU utilization, and therefore throughput, suffers.
2049While 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.
2050As well, concurrency errors return, which threads pools are suppose to mitigate.
2051
2052
2053\section{\protect\CFA Runtime Structure}
2054
2055Figure~\ref{f:RunTimeStructure} illustrates the runtime structure of a \CFA program.
2056In 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.
2057An executing thread is illustrated by its containment in a processor.
2058
2059\begin{figure}
2060\centering
2061\input{RunTimeStructure}
2062\caption{\CFA Runtime Structure}
2063\label{f:RunTimeStructure}
2064\end{figure}
2065
2066
2067\subsection{Cluster}
2068\label{s:RuntimeStructureCluster}
2069
2070A \newterm{cluster} is a collection of threads and virtual processors (abstract kernel-thread) that execute the threads (like a virtual machine).
2071The purpose of a cluster is to control the amount of parallelism that is possible among threads, plus scheduling and other execution defaults.
2072The default cluster-scheduler is single-queue multi-server, which provides automatic load-balancing of threads on processors.
2073However, the scheduler is pluggable, supporting alternative schedulers.
2074If several clusters exist, both threads and virtual processors, can be explicitly migrated from one cluster to another.
2075No automatic load balancing among clusters is performed by \CFA.
2076
2077When 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.
2078The user cluster is created to contain the application user-threads.
2079Having all threads execute on the one cluster often maximizes utilization of processors, which minimizes runtime.
2080However, because of limitations of the underlying operating system, heterogeneous hardware, or scheduling requirements (real-time), multiple clusters are sometimes necessary.
2081
2082
2083\subsection{Virtual Processor}
2084\label{s:RuntimeStructureProcessor}
2085
2086A 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.
2087Programs may use more virtual processors than hardware processors.
2088On a multiprocessor, kernel threads are distributed across the hardware processors resulting in virtual processors executing in parallel.
2089(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.)
2090The \CFA runtime attempts to block unused processors and unblock processors as the system load increases;
2091balancing the workload with processors is difficult.
2092Preemption occurs on virtual processors rather than user threads, via operating-system interrupts.
2093Thus virtual processors execute user threads, where preemption frequency applies to a virtual processor, so preemption occurs randomly across the executed user threads.
2094Turning off preemption transforms user threads into fibers.
2095
2096
2097\subsection{Debug Kernel}
2098
2099There are two versions of the \CFA runtime kernel: debug and non-debug.
2100The 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.
2101After a program is debugged, the non-debugging version can be used to decrease space and increase performance.
2102
2103
2104\section{Implementation}
2105\label{s:Implementation}
2106
2107Currently, \CFA has fixed-sized stacks, where the stack size can be set at coroutine/thread creation but with no subsequent growth.
2108Schemes exist for dynamic stack-growth, such as stack copying and chained stacks.
2109However, stack copying requires pointer adjustment to items on the stack, which is impossible without some form of garbage collection.
2110As well, chained stacks require all modules be recompiled to use this feature, which breaks backward compatibility with existing C libraries.
2111In the long term, it is likely C libraries will migrate to stack chaining to support concurrency, at only a minimal cost to sequential programs.
2112Nevertheless, experience teaching \uC~\cite{CS343} shows fixed-sized stacks are rarely an issue in most concurrent programs.
2113
2114A primary implementation challenge is avoiding contention from dynamically allocating memory because of bulk acquire, \eg the internal-scheduling design is (almost) free of allocations.
2115All 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.
2116Furthermore, several bulk-acquire operations need a variable amount of memory.
2117This 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.
2118
2119In \CFA, ordering of monitor acquisition relies on memory ordering to prevent deadlock~\cite{Havender68}, because all objects have distinct non-overlapping memory layouts, and mutual-exclusion for a monitor is only defined for its lifetime.
2120When a mutex call is made, pointers to the concerned monitors are aggregated into a variable-length array and sorted.
2121This array persists for the entire duration of the mutual exclusion and is used extensively for synchronization operations.
2122
2123To improve performance and simplicity, context switching occurs inside a routine call, so only callee-saved registers are copied onto the stack and then the stack register is switched;
2124the corresponding registers are then restored for the other context.
2125Note, the instruction pointer is untouched since the context switch is always inside the same routine.
2126Unlike coroutines, threads do not context switch among each other;
2127they context switch to the cluster scheduler.
2128This method is a 2-step context-switch and provides a clear distinction between user and kernel code, where scheduling and other system operations happen.
2129The 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.
2130Experimental results (not presented) show the performance of these two approaches is virtually equivalent, because both approaches are dominated by locking to prevent a race condition.
2131
2132All kernel threads (@pthreads@) created a stack.
2133Each \CFA virtual processor is implemented as a coroutine and these coroutines run directly on the kernel-thread stack, effectively stealing this stack.
2134The exception to this rule is the program main, \ie the initial kernel thread that is given to any program.
2135In 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.
2136
2137Finally, 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.
2138Because preemption frequency is usually long (1 millisecond) performance cost is negligible.
2139Preemption is normally handled by setting a count-down timer on each virtual processor.
2140When 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.
2141Multiple signal handlers may be pending.
2142When control eventually switches back to the signal handler, it returns normally, and execution continues in the interrupted user thread, even though the return from the signal handler may be on a different kernel thread than the one where the signal is delivered.
2143The 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;
2144therefore, the same signal mask is required for all virtual processors in a cluster.
2145
2146However, on current UNIX systems:
2147\begin{quote}
2148A process-directed signal may be delivered to any one of the threads that does not currently have the signal blocked.
2149If more than one of the threads has the signal unblocked, then the kernel chooses an arbitrary thread to which to deliver the signal.
2150SIGNAL(7) - Linux Programmer's Manual
2151\end{quote}
2152Hence, 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).
2153To 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.
2154Virtual processors register an expiration time with the discrete-event simulator, which is inserted in sorted order.
2155The 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.
2156Processing a preemption event sends an \emph{internal} @SIGUSR1@ signal to the registered virtual processor, which is always delivered to that processor.
2157
2158
2159\section{Performance}
2160\label{results}
2161
2162To verify the implementation of the \CFA runtime, a series of microbenchmarks are performed comparing \CFA with other widely used programming languages with concurrency.
2163Table~\ref{t:machine} shows the specifications of the computer used to run the benchmarks, and the versions of the software used in the comparison.
2164
2165\begin{table}
2166\centering
2167\caption{Experiment environment}
2168\label{t:machine}
2169
2170\begin{tabular}{|l|r||l|r|}
2171\hline
2172Architecture            & x86\_64                               & NUMA node(s)  & 8 \\
2173\hline
2174CPU op-mode(s)          & 32-bit, 64-bit                & Model name    & AMD Opteron\texttrademark\ Processor 6380 \\
2175\hline
2176Byte Order                      & Little Endian                 & CPU Freq              & 2.5 GHz \\
2177\hline
2178CPU(s)                          & 64                                    & L1d cache     & 16 KiB \\
2179\hline
2180Thread(s) per core      & 2                                     & L1i cache     & 64 KiB \\
2181\hline
2182Core(s) per socket      & 8                                     & L2 cache              & 2048 KiB \\
2183\hline
2184Socket(s)                       & 4                                     & L3 cache              & 6144 KiB \\
2185\hline
2186\hline
2187Operating system        & Ubuntu 16.04.3 LTS    & Kernel                & Linux 4.4-97-generic \\
2188\hline
2189gcc                                     & 6.3                                   & \CFA                  & 1.0.0 \\
2190\hline
2191Java                            & OpenJDK-9                     & Go                    & 1.9.2 \\
2192\hline
2193\end{tabular}
2194\end{table}
2195
2196All benchmarks are run using the following harness:
2197\begin{cfa}
2198unsigned int N = 10_000_000;
2199#define BENCH( run ) Time before = getTimeNsec(); run; Duration result = (getTimeNsec() - before) / N;
2200\end{cfa}
2201The method used to get time is @clock_gettime( CLOCK_REALTIME )@.
2202Each benchmark is performed @N@ times, where @N@ varies depending on the benchmark;
2203the total time is divided by @N@ to obtain the average time for a benchmark.
2204All omitted tests for other languages are functionally identical to the shown \CFA test.
2205
2206
2207\paragraph{Context-Switching}
2208
2209In procedural programming, the cost of a routine call is important as modularization (refactoring) increases.
2210(In many cases, a compiler inlines routine calls to eliminate this cost.)
2211Similarly, when modularization extends to coroutines/tasks, the time for a context switch becomes a relevant factor.
2212The coroutine context-switch is 2-step using resume/suspend, \ie from resumer to suspender and from suspender to resumer.
2213The thread context switch is 2-step using yield, \ie enter and return from the runtime kernel.
2214Figure~\ref{f:ctx-switch} shows the code for coroutines/threads with all results in Table~\ref{tab:ctx-switch}.
2215The difference in performance between coroutine and thread context-switch is the cost of scheduling for threads, whereas coroutines are self-scheduling.
2216
2217\begin{figure}
2218\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2219
2220\newbox\myboxA
2221\begin{lrbox}{\myboxA}
2222\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2223coroutine C {} c;
2224void main( C & ) { for ( ;; ) { @suspend();@ } }
2225int main() {
2226        BENCH(
2227                for ( size_t i = 0; i < N; i += 1 ) { @resume( c );@ } )
2228        sout | result`ns;
2229}
2230\end{cfa}
2231\end{lrbox}
2232
2233\newbox\myboxB
2234\begin{lrbox}{\myboxB}
2235\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2236
2237
2238int main() {
2239        BENCH(
2240                for ( size_t i = 0; i < N; i += 1 ) { @yield();@ } )
2241        sout | result`ns;
2242}
2243\end{cfa}
2244\end{lrbox}
2245
2246\subfloat[Coroutine]{\usebox\myboxA}
2247\quad
2248\subfloat[Thread]{\usebox\myboxB}
2249\captionof{figure}{\CFA context-switch benchmark}
2250\label{f:ctx-switch}
2251
2252\centering
2253
2254\captionof{table}{Context switch comparison (nanoseconds)}
2255\label{tab:ctx-switch}
2256\bigskip
2257\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2258\cline{2-4}
2259\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2260\hline
2261Kernel Thread   & 333.5 & 332.96        & 4.1 \\
2262\CFA Coroutine  & 49            & 48.68 & 0.47    \\
2263\CFA Thread             & 105           & 105.57        & 1.37 \\
2264\uC Coroutine   & 44            & 44            & 0 \\
2265\uC Thread              & 100           & 99.29 & 0.96 \\
2266Goroutine               & 145           & 147.25        & 4.15 \\
2267Java Thread             & 373.5 & 375.14        & 8.72 \\
2268\hline
2269\end{tabular}
2270
2271\bigskip
2272\bigskip
2273
2274\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2275\begin{cfa}
2276monitor M { ... } m1/*, m2, m3, m4*/;
2277void __attribute__((noinline)) do_call( M & mutex m/*, m2, m3, m4*/ ) {}
2278int main() {
2279        BENCH( for( size_t i = 0; i < N; i += 1 ) { @do_call( m1/*, m2, m3, m4*/ );@ } )
2280        sout | result`ns;
2281}
2282\end{cfa}
2283\captionof{figure}{\CFA acquire/release mutex benchmark}
2284\label{f:mutex}
2285
2286\centering
2287
2288\captionof{table}{Mutex comparison (nanoseconds)}
2289\label{tab:mutex}
2290\bigskip
2291
2292\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2293\cline{2-4}
2294\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2295\hline
2296C routine                                       & 2             & 2             & 0    \\
2297FetchAdd + FetchSub                     & 26            & 26            & 0    \\
2298Pthreads Mutex Lock                     & 31            & 31.71 & 0.97 \\
2299\uC @monitor@ member routine            & 31            & 31            & 0    \\
2300\CFA @mutex@ routine, 1 argument        & 46            & 46.68 & 0.93  \\
2301\CFA @mutex@ routine, 2 argument        & 84            & 85.36 & 1.99 \\
2302\CFA @mutex@ routine, 4 argument        & 158           & 161           & 4.22 \\
2303Java synchronized routine               & 27.5  & 29.79 & 2.93  \\
2304\hline
2305\end{tabular}
2306\end{figure}
2307
2308
2309\paragraph{Mutual-Exclusion}
2310
2311Mutual exclusion is measured by entering/leaving a critical section.
2312For monitors, entering and leaving a monitor routine is measured.
2313Figure~\ref{f:mutex} shows the code for \CFA with all results in Table~\ref{tab:mutex}.
2314To 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.
2315Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2316
2317
2318\paragraph{Internal Scheduling}
2319
2320Internal scheduling is measured by waiting on and signalling a condition variable.
2321Figure~\ref{f:int-sched} shows the code for \CFA, with results in Table~\ref{tab:int-sched}.
2322Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2323Java scheduling is significantly greater because the benchmark explicitly creates multiple thread in order to prevent the JIT from making the program sequential, \ie removing all locking.
2324
2325\begin{figure}
2326\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2327\begin{cfa}
2328volatile int go = 0;
2329condition c;
2330monitor M { ... } m;
2331void __attribute__((noinline)) do_call( M & mutex a1 ) { signal( c ); }
2332thread T {};
2333void main( T & this ) {
2334        while ( go == 0 ) { yield(); }  // wait for other thread to start
2335        while ( go == 1 ) { @do_call( m );@ }
2336}
2337int  __attribute__((noinline)) do_wait( M & mutex m ) {
2338        go = 1; // continue other thread
2339        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @wait( c );@ } );
2340        go = 0; // stop other thread
2341        sout | result`ns;
2342}
2343int main() {
2344        T t;
2345        do_wait( m );
2346}
2347\end{cfa}
2348\captionof{figure}{\CFA Internal-scheduling benchmark}
2349\label{f:int-sched}
2350
2351\centering
2352\captionof{table}{Internal-scheduling comparison (nanoseconds)}
2353\label{tab:int-sched}
2354\bigskip
2355
2356\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2357\cline{2-4}
2358\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2359\hline
2360Pthreads Condition Variable             & 6005  & 5681.43       & 835.45 \\
2361\uC @signal@                                    & 324           & 325.54        & 3,02   \\
2362\CFA @signal@, 1 @monitor@              & 368.5         & 370.61        & 4.77   \\
2363\CFA @signal@, 2 @monitor@              & 467           & 470.5 & 6.79   \\
2364\CFA @signal@, 4 @monitor@              & 700.5         & 702.46        & 7.23  \\
2365Java @notify@                                   & 15471 & 172511        & 5689 \\
2366\hline
2367\end{tabular}
2368\end{figure}
2369
2370
2371\paragraph{External Scheduling}
2372
2373External scheduling is measured by accepting a call using the @waitfor@ statement (@_Accept@ in \uC).
2374Figure~\ref{f:ext-sched} shows the code for \CFA, with results in Table~\ref{tab:ext-sched}.
2375Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2376
2377\begin{figure}
2378\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2379\begin{cfa}
2380volatile int go = 0;
2381monitor M { ... } m;
2382thread T {};
2383void __attribute__((noinline)) do_call( M & mutex ) {}
2384void main( T & ) {
2385        while ( go == 0 ) { yield(); }  // wait for other thread to start
2386        while ( go == 1 ) { @do_call( m );@ }
2387}
2388int __attribute__((noinline)) do_wait( M & mutex m ) {
2389        go = 1; // continue other thread
2390        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @waitfor( do_call, m );@ } )
2391        go = 0; // stop other thread
2392        sout | result`ns;
2393}
2394int main() {
2395        T t;
2396        do_wait( m );
2397}
2398\end{cfa}
2399\captionof{figure}{\CFA external-scheduling benchmark}
2400\label{f:ext-sched}
2401
2402\centering
2403
2404\captionof{table}{External-scheduling comparison (nanoseconds)}
2405\label{tab:ext-sched}
2406\bigskip
2407\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2408\cline{2-4}
2409\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2410\hline
2411\uC @_Accept@                           & 358           & 359.11        & 2.53  \\
2412\CFA @waitfor@, 1 @monitor@     & 359           & 360.93        & 4.07  \\
2413\CFA @waitfor@, 2 @monitor@     & 450           & 449.39        & 6.62  \\
2414\CFA @waitfor@, 4 @monitor@     & 652           & 655.64        & 7.73 \\
2415\hline
2416\end{tabular}
2417
2418\bigskip
2419\medskip
2420
2421\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2422\begin{cfa}
2423thread MyThread {};
2424void main( MyThread & ) {}
2425int main() {
2426        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @MyThread m;@ } )
2427        sout | result`ns;
2428}
2429\end{cfa}
2430\captionof{figure}{\CFA object-creation benchmark}
2431\label{f:creation}
2432
2433\centering
2434
2435\captionof{table}{Creation comparison (nanoseconds)}
2436\label{tab:creation}
2437\bigskip
2438
2439\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2440\cline{2-4}
2441\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} & \multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2442\hline
2443Pthreads                                & 28091         & 28073.39      & 163.1  \\
2444\CFA Coroutine Lazy             & 6                     & 6.07          & 0.26   \\
2445\CFA Coroutine Eager    & 520           & 520.61        & 2.04   \\
2446\CFA Thread                             & 2032  & 2016.29       & 112.07  \\
2447\uC Coroutine                   & 106           & 107.36        & 1.47   \\
2448\uC Thread                              & 536.5 & 537.07        & 4.64   \\
2449Goroutine                               & 3103  & 3086.29       & 90.25  \\
2450Java Thread                             & 103416.5      & 103732.29     & 1137 \\
2451\hline
2452\end{tabular}
2453\end{figure}
2454
2455
2456\paragraph{Object Creation}
2457
2458Object creation is measured by creating/deleting the specific kind of concurrent object.
2459Figure~\ref{f:creation} shows the code for \CFA, with results in Table~\ref{tab:creation}.
2460The 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.
2461
2462
2463\section{Conclusion}
2464
2465This paper demonstrates a concurrency API that is simple, efficient, and able to build higher-level concurrency features.
2466The 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.
2467The M:N model is judged to be efficient and provide greater flexibility than a 1:1 threading model.
2468High-level objects (monitor/task) are the core mechanism for mutual exclusion and synchronization.
2469A novel aspect is allowing multiple mutex-objects to be accessed simultaneously reducing the potential for deadlock for this complex scenario.
2470These concepts and the entire \CFA runtime-system are written in the \CFA language, demonstrating the expressiveness of the \CFA language.
2471Performance 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.
2472C 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.
2473
2474
2475\section{Future Work}
2476
2477While concurrency in \CFA has a strong start, development is still underway and there are missing features.
2478
2479\paragraph{Flexible Scheduling}
2480\label{futur:sched}
2481
2482An important part of concurrency is scheduling.
2483Different scheduling algorithms can affect performance (both in terms of average and variation).
2484However, no single scheduler is optimal for all workloads and therefore there is value in being able to change the scheduler for given programs.
2485One solution is to offer various tweaking options, allowing the scheduler to be adjusted to the requirements of the workload.
2486However, to be truly flexible, a pluggable scheduler is necessary.
2487Currently, 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.
2488
2489\paragraph{Non-Blocking I/O}
2490\label{futur:nbio}
2491
2492Many modern workloads are not bound by computation but IO operations, a common case being web servers and XaaS~\cite{XaaS} (anything as a service).
2493These types of workloads require significant engineering to amortizing costs of blocking IO-operations.
2494At 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.
2495Current 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.
2496However, these solutions lead to code that is hard to create, read and maintain.
2497A 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.
2498A non-blocking I/O library is currently under development for \CFA.
2499
2500\paragraph{Other Concurrency Tools}
2501\label{futur:tools}
2502
2503While monitors offer flexible and powerful concurrency for \CFA, other concurrency tools are also necessary for a complete multi-paradigm concurrency package.
2504Examples of such tools can include futures and promises~\cite{promises}, executors and actors.
2505These additional features are useful when monitors offer a level of abstraction that is inadequate for certain tasks.
2506As well, new \CFA extensions should make it possible to create a uniform interface for virtually all mutual exclusion, including monitors and low-level locks.
2507
2508\paragraph{Implicit Threading}
2509\label{futur:implcit}
2510
2511Basic 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.
2512This type of concurrency can be achieved both at the language level and at the library level.
2513The canonical example of implicit concurrency is concurrent nested @for@ loops, which are amenable to divide and conquer algorithms~\cite{uC++book}.
2514The \CFA language features should make it possible to develop a reasonable number of implicit concurrency mechanism to solve basic HPC data-concurrency problems.
2515However, implicit concurrency is a restrictive solution with significant limitations, so it can never replace explicit concurrent programming.
2516
2517
2518\section{Acknowledgements}
2519
2520The authors would like to recognize the design assistance of Aaron Moss, Rob Schluntz and Andrew Beach on the features described in this paper.
2521Funding 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.
2522
2523{%
2524\fontsize{9bp}{12bp}\selectfont%
2525\bibliography{pl,local}
2526}%
2527
2528\end{document}
2529
2530% Local Variables: %
2531% tab-width: 4 %
2532% fill-column: 120 %
2533% compile-command: "make" %
2534% End: %
Note: See TracBrowser for help on using the repository browser.