source: doc/papers/concurrency/Paper.tex @ eb28d7e

ADTarm-ehast-experimentalcleanup-dtorsenumforall-pointer-decayjacob/cs343-translationjenkins-sandboxnew-astnew-ast-unique-exprpthread-emulationqualifiedEnum
Last change on this file since eb28d7e was eb28d7e, checked in by Peter A. Buhr <pabuhr@…>, 5 years ago

rewrite abstract and introduction

  • Property mode set to 100644
File size: 135.9 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{language=go,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
218{}
219\lstnewenvironment{python}[1][]
220{\lstset{language=python,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
221{}
222
223% inline code @...@
224\lstMakeShortInline@%
225
226\let\OLDthebibliography\thebibliography
227\renewcommand\thebibliography[1]{
228  \OLDthebibliography{#1}
229  \setlength{\parskip}{0pt}
230  \setlength{\itemsep}{4pt plus 0.3ex}
231}
232
233\title{\texorpdfstring{Advanced Control-flow and Concurrency in \protect\CFA}{Advanced Control-flow in Cforall}}
234
235\author[1]{Thierry Delisle}
236\author[1]{Peter A. Buhr*}
237\authormark{DELISLE \textsc{et al.}}
238
239\address[1]{\orgdiv{Cheriton School of Computer Science}, \orgname{University of Waterloo}, \orgaddress{\state{Waterloo, ON}, \country{Canada}}}
240
241\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}}
242
243% \fundingInfo{Natural Sciences and Engineering Research Council of Canada}
244
245\abstract[Summary]{
246\CFA is a polymorphic, non-object-oriented, concurrent, backwards-compatible extension of the C programming language.
247This paper discusses the design philosophy and implementation of its advanced control-flow and concurrency/parallelism features, along with the supporting runtime.
248These features are created from scratch as ISO C has only low-level and/or unimplemented concurrency, so C programmers continue to rely on library features, like C pthreads.
249\CFA introduces modern language-level control-flow mechanisms, like coroutines, user-level threading, and monitors for mutual exclusion and synchronization.
250The design contributions provide significant programmer simplification and safety by eliminating spurious wakeup and barging in monitors.
251As well, multiple monitors can be safely acquired \emph{simultaneously} (deadlock free), which is fully integrated with the monitor synchronization mechanisms.
252These features also integrate with the \CFA polymorphic type-system and exception handling, while respecting the expectations and style of C programmers.
253Experimental results show comparable performance of the new features with similar (weaker) mechanisms in other concurrent programming-languages.
254}%
255
256\keywords{coroutines, concurrency, parallelism, threads, monitors, runtime, C, \CFA (Cforall)}
257
258
259\begin{document}
260\linenumbers                                            % comment out to turn off line numbering
261
262\maketitle
263
264
265\section{Introduction}
266
267This paper discusses the design philosophy and implementation of advanced language-level control-flow and concurrency/parallelism extensions in \CFA and its runtime.
268\CFA is a modern, polymorphic, non-object-oriented\footnote{
269\CFA has features often associated with object-oriented programming languages, such as constructors, destructors, virtuals and simple inheritance.
270However, functions \emph{cannot} be nested in structures, so there is no lexical binding between a structure and set of functions (member/method) implemented by an implicit \lstinline@this@ (receiver) parameter.},
271backwards-compatible extension of the C programming language~\cite{Moss18}.
272Within the \CFA framework, new control-flow features are created from scratch.
273ISO \Celeven defines only a subset of the \CFA extensions, where the overlapping features are concurrency~\cite[\S~7.26]{C11}.
274However, \Celeven concurrency is largely wrappers for a subset of the pthreads library~\cite{Butenhof97,Pthreads}.
275Furthermore, \Celeven and pthreads concurrency is simple, based on thread fork/join in a function and a few locks, which is low-level and error prone;
276no high-level language concurrency features are defined.
277Interestingly, almost a decade after publication of the \Celeven standard, neither gcc-8, clang-8 nor msvc-19 (most recent versions) support the \Celeven include @threads.h@, indicating little interest in the C11 concurrency approach.
278Finally, while the \Celeven standard does not state a threading model, the historical association with pthreads suggests implementations would adopt kernel-level threading (1:1)~\cite{ThreadModel}.
279
280In contrast, there has been a renewed interest during the past decade in user-level (M:N, green) threading in old and new programming languages.
281As multi-core hardware became available in the 1980/90s, both user and kernel threading were examined.
282Kernel 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}.
283Libraries like pthreads were developed for C, and the Solaris operating-system switched from user (JDK 1.1~\cite{JDK1.1}) to kernel threads.
284As a result, languages like Java, Scala~\cite{Scala}, Objective-C~\cite{obj-c-book}, \CCeleven~\cite{C11}, and C\#~\cite{Csharp} adopt the 1:1 kernel-threading model, with a variety of presentation mechanisms.
285From 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}.
286The main argument for user-level threading is that they are lighter weight than kernel threads (locking and context switching do not cross the kernel boundary), so there is less restriction on programming styles that encourage large numbers of threads performing smaller work-units to facilitate load balancing by the runtime~\cite{Verch12}.
287As well, user-threading facilitates a simpler concurrency approach using thread objects that leverage sequential patterns versus events with call-backs~\cite{vonBehren03}.
288Finally, performant user-threading implementations (both time and space) are largely competitive with direct kernel-threading implementations, while achieving the programming advantages of high concurrency levels and safety.
289
290A further effort over the past two decades is the development of language memory-models to deal with the conflict between language features and compiler/hardware optimizations, i.e., some language features are unsafe in the presence of aggressive sequential optimizations~\cite{Buhr95a,Boehm05}.
291The consequence is that a language must provide sufficient tools to program around safety issues, as inline and library code is all sequential to the compiler.
292One solution is low-level qualifiers and functions (e.g., @volatile@ and atomics) allowing \emph{programmers} to explicitly write safe (race-free~\cite{Boehm12}) programs.
293A safer solution is high-level language constructs so the \emph{compiler} knows the optimization boundaries, and hence, provides implicit safety.
294This problem is best know with respect to concurrency, but applies to other complex control-flow, like exceptions\footnote{
295\CFA exception handling will be presented in a separate paper.
296The key feature that dovetails with this paper is non-local exceptions allowing exceptions to be raised across stacks, with synchronous exceptions raised among coroutines and asynchronous exceptions raised among threads, similar to that in \uC~\cite[\S~5]{uC++}
297} and coroutines.
298Finally, solutions in the language allows matching constructs with language paradigm, i.e., imperative and functional languages have different presentations of the same concept.
299
300Finally, it is important for a language to provide safety over performance \emph{as the default}, allowing careful reduction of safety for performance when necessary.
301Two concurrency violations of this philosophy are \emph{spurious wakeup} and \emph{barging}, i.e., random wakeup~\cite[\S~8]{Buhr05a} and signalling-as-hints~\cite[\S~8]{Buhr05a}, where one begats the other.
302If you believe spurious wakeup is a foundational concurrency property, than unblocking (signalling) a thread is always a hint.
303If you \emph{do not} believe spurious wakeup is foundational, than signalling-as-hints is a performance decision.
304Most importantly, removing spurious wakeup and signals-as-hints makes concurrent programming significantly safer because it removes local non-determinism.
305Clawing back performance where the local non-determinism is unimportant, should be an option not the default.
306
307\begin{comment}
308For example, it is possible to provide exceptions, coroutines, monitors, and tasks as specialized types in an object-oriented language, integrating these constructs to allow leveraging the type-system (static type-checking) and all other object-oriented capabilities~\cite{uC++}.
309It is also possible to leverage call/return for blocking communication via new control structures, versus switching to alternative communication paradigms, like channels or message passing.
310As well, user threading is often a complementary feature, allowing light-weight threading to match with low-cost objects, while hiding the application/kernel boundary.
311User threading also allows layering of implicit concurrency models (no explicit thread creation), such executors, data-flow, actors, into a single language, so programmers can chose the model that best fits an algorithm.\footnote{
312All implicit concurrency models have explicit threading in their implementation, and hence, can be build from explicit threading;
313however, the reverse is seldom true, i.e., given implicit concurrency, e.g., actors, it is virtually impossible to create explicit concurrency, e.g., blocking thread objects.}
314Finally, with extended language features and user-level threading it is possible to discretely fold locking and non-blocking I/O multiplexing into the language's I/O libraries, so threading implicitly dovetails with the I/O subsystem.
315\CFA embraces language extensions and user-level threading to provide advanced control-flow (exception handling\footnote{
316\CFA exception handling will be presented in a separate paper.
317The key feature that dovetails with this paper is non-local exceptions allowing exceptions to be raised across stacks, with synchronous exceptions raised among coroutines and asynchronous exceptions raised among threads, similar to that in \uC~\cite[\S~5]{uC++}
318} and coroutines) and concurrency.
319
320Most augmented traditional (Fortran 18~\cite{Fortran18}, Cobol 14~\cite{Cobol14}, Ada 12~\cite{Ada12}, Java 11~\cite{Java11}) and new languages (Go~\cite{Go}, Rust~\cite{Rust}, and D~\cite{D}), except \CC, diverge from C with different syntax and semantics, only interoperate indirectly with C, and are not systems languages, for those with managed memory.
321As a result, there is a significant learning curve to move to these languages, and C legacy-code must be rewritten.
322While \CC, like \CFA, takes an evolutionary approach to extend C, \CC's constantly growing complex and interdependent features-set (e.g., objects, inheritance, templates, etc.) mean idiomatic \CC code is difficult to use from C, and C programmers must expend significant effort learning \CC.
323Hence, rewriting and retraining costs for these languages, even \CC, are prohibitive for companies with a large C software-base.
324\CFA with its orthogonal feature-set, its high-performance runtime, and direct access to all existing C libraries circumvents these problems.
325\end{comment}
326
327\CFA embraces user-level threading, language extensions for advanced control-flow, and safety as the default.
328We present comparative examples so the reader can judge if the \CFA control-flow extensions are better and safer than those in or proposed for \Celeven, \CC and other concurrent, imperative programming languages, and perform experiments to show the \CFA runtime is competitive with other similar mechanisms.
329The main contributions of this work are:
330\begin{itemize}
331\item
332expressive language-level coroutines and user-level threading, which respect the expectations of C programmers.
333\item
334monitor synchronization without barging.
335\item
336safely acquiring multiple monitors \emph{simultaneously} (deadlock free), while seamlessly integrating this capability with all monitor synchronization mechanisms.
337\item
338providing statically type-safe interfaces that integrate with the \CFA polymorphic type-system and other language features.
339\item
340a runtime system with no spurious wakeup.
341\item
342experimental results showing comparable performance of the new features with similar mechanisms in other concurrent programming-languages.
343\end{itemize}
344
345\begin{comment}
346This paper provides a minimal concurrency \newterm{Application Program Interface} (API) that is simple, efficient and can be used to build other concurrency features.
347While the simplest concurrency system is a thread and a lock, this low-level approach is hard to master.
348An easier approach for programmers is to support higher-level constructs as the basis of concurrency.
349Indeed, for highly-productive concurrent-programming, high-level approaches are much more popular~\cite{Hochstein05}.
350Examples 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}.
351
352The following terminology is used.
353A \newterm{thread} is a fundamental unit of execution that runs a sequence of code and requires a stack to maintain state.
354Multiple simultaneous threads give rise to \newterm{concurrency}, which requires locking to ensure access to shared data and safe communication.
355\newterm{Locking}, and by extension \newterm{locks}, are defined as a mechanism to prevent progress of threads to provide safety.
356\newterm{Parallelism} is running multiple threads simultaneously.
357Parallelism implies \emph{actual} simultaneous execution, where concurrency only requires \emph{apparent} simultaneous execution.
358As such, parallelism only affects performance, which is observed through differences in space and/or time at runtime.
359Hence, there are two problems to be solved: concurrency and parallelism.
360While these two concepts are often combined, they are distinct, requiring different tools~\cite[\S~2]{Buhr05a}.
361Concurrency tools handle mutual exclusion and synchronization, while parallelism tools handle performance, cost, and resource utilization.
362
363The proposed concurrency API is implemented in a dialect of C, called \CFA (pronounced C-for-all).
364The 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.
365\end{comment}
366
367
368\begin{comment}
369\section{\CFA Overview}
370
371The following is a quick introduction to the \CFA language, specifically tailored to the features needed to support concurrency.
372Extended versions and explanation of the following code examples are available at the \CFA website~\cite{Cforall} or in Moss~\etal~\cite{Moss18}.
373
374\CFA is a non-object-oriented extension of ISO-C, and hence, supports all C paradigms.
375Like C, the building blocks of \CFA are structures and routines.
376Virtually all of the code generated by the \CFA translator respects C memory layouts and calling conventions.
377While \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}.
378While some object-oriented features appear in \CFA, they are independent capabilities, allowing \CFA to adopt them while maintaining a procedural paradigm.
379
380
381\subsection{References}
382
383\CFA provides multi-level rebindable references, as an alternative to pointers, which significantly reduces syntactic noise.
384\begin{cfa}
385int x = 1, y = 2, z = 3;
386int * p1 = &x, ** p2 = &p1,  *** p3 = &p2,      $\C{// pointers to x}$
387    `&` r1 = x,   `&&` r2 = r1,   `&&&` r3 = r2;        $\C{// references to x}$
388int * p4 = &z, `&` r4 = z;
389
390*p1 = 3; **p2 = 3; ***p3 = 3;       // change x
391r1 =  3;     r2 = 3;      r3 = 3;        // change x: implicit dereferences *r1, **r2, ***r3
392**p3 = &y; *p3 = &p4;                // change p1, p2
393`&`r3 = &y; `&&`r3 = &`&`r4;             // change r1, r2: cancel implicit dereferences (&*)**r3, (&(&*)*)*r3, &(&*)r4
394\end{cfa}
395A reference is a handle to an object, like a pointer, but is automatically dereferenced the specified number of levels.
396Referencing (address-of @&@) a reference variable cancels one of the implicit dereferences, until there are no more implicit references, after which normal expression behaviour applies.
397
398
399\subsection{\texorpdfstring{\protect\lstinline{with} Statement}{with Statement}}
400\label{s:WithStatement}
401
402Heterogeneous data is aggregated into a structure/union.
403To 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.
404\begin{cquote}
405\vspace*{-\baselineskip}%???
406\lstDeleteShortInline@%
407\begin{cfa}
408struct S { char c; int i; double d; };
409struct T { double m, n; };
410// multiple aggregate parameters
411\end{cfa}
412\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
413\begin{cfa}
414void f( S & s, T & t ) {
415        `s.`c; `s.`i; `s.`d;
416        `t.`m; `t.`n;
417}
418\end{cfa}
419&
420\begin{cfa}
421void f( S & s, T & t ) `with ( s, t )` {
422        c; i; d;                // no qualification
423        m; n;
424}
425\end{cfa}
426\end{tabular}
427\lstMakeShortInline@%
428\end{cquote}
429Object-oriented programming languages only provide implicit qualification for the receiver.
430
431In detail, the @with@-statement syntax is:
432\begin{cfa}
433$\emph{with-statement}$:
434        'with' '(' $\emph{expression-list}$ ')' $\emph{compound-statement}$
435\end{cfa}
436and may appear as the body of a routine or nested within a routine body.
437Each expression in the expression-list provides a type and object.
438The type must be an aggregate type.
439(Enumerations are already opened.)
440The object is the implicit qualifier for the open structure-fields.
441All 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.
442
443
444\subsection{Overloading}
445
446\CFA maximizes the ability to reuse names via overloading to aggressively address the naming problem.
447Both variables and routines may be overloaded, where selection is based on number and types of returns and arguments (as in Ada~\cite{Ada}).
448\newpage
449\vspace*{-2\baselineskip}%???
450\begin{cquote}
451\begin{cfa}
452// selection based on type
453\end{cfa}
454\lstDeleteShortInline@%
455\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
456\begin{cfa}
457const short int `MIN` = -32768;
458const int `MIN` = -2147483648;
459const long int `MIN` = -9223372036854775808L;
460\end{cfa}
461&
462\begin{cfa}
463short int si = `MIN`;
464int i = `MIN`;
465long int li = `MIN`;
466\end{cfa}
467\end{tabular}
468\begin{cfa}
469// selection based on type and number of parameters
470\end{cfa}
471\begin{tabular}{@{}l@{\hspace{2.7\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
472\begin{cfa}
473void `f`( void );
474void `f`( char );
475void `f`( int, double );
476\end{cfa}
477&
478\begin{cfa}
479`f`();
480`f`( 'a' );
481`f`( 3, 5.2 );
482\end{cfa}
483\end{tabular}
484\begin{cfa}
485// selection based on type and number of returns
486\end{cfa}
487\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}|@{\hspace{2\parindentlnth}}l@{}}
488\begin{cfa}
489char `f`( int );
490double `f`( int );
491[char, double] `f`( int );
492\end{cfa}
493&
494\begin{cfa}
495char c = `f`( 3 );
496double d = `f`( 3 );
497[d, c] = `f`( 3 );
498\end{cfa}
499\end{tabular}
500\lstMakeShortInline@%
501\end{cquote}
502Overloading is important for \CFA concurrency since the runtime system relies on creating different types to represent concurrency objects.
503Therefore, overloading eliminates long prefixes and other naming conventions to prevent name clashes.
504As seen in Section~\ref{s:Concurrency}, routine @main@ is heavily overloaded.
505As another example, variable overloading is useful in the parallel semantics of the @with@ statement for fields with the same name:
506\begin{cfa}
507struct S { int `i`; int j; double m; } s;
508struct T { int `i`; int k; int m; } t;
509with ( s, t ) {
510        j + k;                                                                  $\C{// unambiguous, s.j + t.k}$
511        m = 5.0;                                                                $\C{// unambiguous, s.m = 5.0}$
512        m = 1;                                                                  $\C{// unambiguous, t.m = 1}$
513        int a = m;                                                              $\C{// unambiguous, a = t.m }$
514        double b = m;                                                   $\C{// unambiguous, b = s.m}$
515        int c = `s.i` + `t.i`;                                  $\C{// unambiguous, qualification}$
516        (double)m;                                                              $\C{// unambiguous, cast s.m}$
517}
518\end{cfa}
519For parallel semantics, both @s.i@ and @t.i@ are visible with the same type, so only @i@ is ambiguous without qualification.
520
521
522\subsection{Operators}
523
524Overloading also extends to operators.
525Operator-overloading syntax creates a routine name with an operator symbol and question marks for the operands:
526\begin{cquote}
527\lstDeleteShortInline@%
528\begin{tabular}{@{}ll@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
529\begin{cfa}
530int ++?(int op);
531int ?++(int op);
532int `?+?`(int op1, int op2);
533int ?<=?(int op1, int op2);
534int ?=? (int & op1, int op2);
535int ?+=?(int & op1, int op2);
536\end{cfa}
537&
538\begin{cfa}
539// unary prefix increment
540// unary postfix increment
541// binary plus
542// binary less than
543// binary assignment
544// binary plus-assignment
545\end{cfa}
546&
547\begin{cfa}
548struct S { int i, j; };
549S `?+?`( S op1, S op2) { // add two structures
550        return (S){op1.i + op2.i, op1.j + op2.j};
551}
552S s1 = {1, 2}, s2 = {2, 3}, s3;
553s3 = s1 `+` s2;         // compute sum: s3 == {2, 5}
554\end{cfa}
555\end{tabular}
556\lstMakeShortInline@%
557\end{cquote}
558
559
560\subsection{Constructors / Destructors}
561
562Object lifetime is a challenge in non-managed programming languages.
563\CFA responds with \CC-like constructors and destructors, using a different operator-overloading syntax.
564\begin{cfa}
565struct VLA { int len, * data; };                        $\C{// variable length array of integers}$
566void ?{}( VLA & vla ) with ( vla ) { len = 10;  data = alloc( len ); }  $\C{// default constructor}$
567void ?{}( VLA & vla, int size, char fill ) with ( vla ) { len = size;  data = alloc( len, fill ); } // initialization
568void ?{}( VLA & vla, VLA other ) { vla.len = other.len;  vla.data = other.data; } $\C{// copy, shallow}$
569void ^?{}( VLA & vla ) with ( vla ) { free( data ); } $\C{// destructor}$
570{
571        VLA  x,            y = { 20, 0x01 },     z = y; $\C{// z points to y}$
572        // $\LstCommentStyle{\color{red}\ \ \ x\{\};\ \ \ \ \ \ \ \ \ y\{ 20, 0x01 \};\ \ \ \ \ \ \ \ \ \ z\{ z, y \};\ \ \ \ \ \ \ implicit calls}$
573        ^x{};                                                                   $\C{// deallocate x}$
574        x{};                                                                    $\C{// reallocate x}$
575        z{ 5, 0xff };                                                   $\C{// reallocate z, not pointing to y}$
576        ^y{};                                                                   $\C{// deallocate y}$
577        y{ x };                                                                 $\C{// reallocate y, points to x}$
578        x{};                                                                    $\C{// reallocate x, not pointing to y}$
579}       //  $\LstCommentStyle{\color{red}\^{}z\{\};\ \ \^{}y\{\};\ \ \^{}x\{\};\ \ \ implicit calls}$
580\end{cfa}
581Like \CC, construction is implicit on allocation (stack/heap) and destruction is implicit on deallocation.
582The object and all their fields are constructed/destructed.
583\CFA also provides @new@ and @delete@ as library routines, which behave like @malloc@ and @free@, in addition to constructing and destructing objects:
584\begin{cfa}
585{
586        ... struct S s = {10}; ...                              $\C{// allocation, call constructor}$
587}                                                                                       $\C{// deallocation, call destructor}$
588struct S * s = new();                                           $\C{// allocation, call constructor}$
589...
590delete( s );                                                            $\C{// deallocation, call destructor}$
591\end{cfa}
592\CFA concurrency uses object lifetime as a means of mutual exclusion and/or synchronization.
593
594
595\subsection{Parametric Polymorphism}
596\label{s:ParametricPolymorphism}
597
598The 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.
599For example, the following sum routine works for any type that supports construction from 0 and addition:
600\begin{cfa}
601forall( otype T | { void `?{}`( T *, zero_t ); T `?+?`( T, T ); } ) // constraint type, 0 and +
602T sum( T a[$\,$], size_t size ) {
603        `T` total = { `0` };                                    $\C{// initialize by 0 constructor}$
604        for ( size_t i = 0; i < size; i += 1 )
605                total = total `+` a[i];                         $\C{// select appropriate +}$
606        return total;
607}
608S sa[5];
609int i = sum( sa, 5 );                                           $\C{// use S's 0 construction and +}$
610\end{cfa}
611Type variables can be @otype@ or @dtype@.
612@otype@ refers to a \emph{complete type}, \ie, a type with size, alignment, default constructor, copy constructor, destructor, and assignment operator.
613@dtype@ refers to an \emph{incomplete type}, \eg, void or a forward-declared type.
614The 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.
615
616\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:
617\begin{cfa}
618trait `sumable`( otype T ) {
619        void `?{}`( T &, zero_t );                              $\C{// 0 literal constructor}$
620        T `?+?`( T, T );                                                $\C{// assortment of additions}$
621        T ?+=?( T &, T );
622        T ++?( T & );
623        T ?++( T & );
624};
625forall( otype T `| sumable( T )` )                      $\C{// use trait}$
626T sum( T a[$\,$], size_t size );
627\end{cfa}
628
629Using the return type for overload discrimination, it is possible to write a type-safe @alloc@ based on the C @malloc@:
630\begin{cfa}
631forall( dtype T | sized(T) ) T * alloc( void ) { return (T *)malloc( sizeof(T) ); }
632int * ip = alloc();                                                     $\C{// select type and size from left-hand side}$
633double * dp = alloc();
634struct S {...} * sp = alloc();
635\end{cfa}
636where the return type supplies the type/size of the allocation, which is impossible in most type systems.
637\end{comment}
638
639
640\section{Coroutines: Stepping Stone}
641\label{coroutine}
642
643Coroutines are generalized routines allowing execution to be temporarily suspended and later resumed.
644Hence, 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.
645This capability is accomplished via the coroutine's stack, where suspend/resume context switch among stacks.
646Because 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.
647Therefore, the two fundamental features of the core \CFA coroutine-API are independent call-stacks and @suspend@/@resume@ operations.
648
649For example, a problem made easier with coroutines is unbounded generators, \eg generating an infinite sequence of Fibonacci numbers
650\begin{displaymath}
651\mathsf{fib}(n) = \left \{
652\begin{array}{ll}
6530                                       & n = 0         \\
6541                                       & n = 1         \\
655\mathsf{fib}(n-1) + \mathsf{fib}(n-2)   & n \ge 2       \\
656\end{array}
657\right.
658\end{displaymath}
659where Figure~\ref{f:C-fibonacci} shows conventional approaches for writing a Fibonacci generator in C.
660Figure~\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.
661Figure~\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 $\mathsf{fib}(n-2)$.
662
663\begin{figure}
664\centering
665\newbox\myboxA
666% \begin{lrbox}{\myboxA}
667% \begin{cfa}[aboveskip=0pt,belowskip=0pt]
668% `int fn1, fn2, state = 1;`   // single global variables
669% int fib() {
670%       int fn;
671%       `switch ( state )` {  // explicit execution state
672%         case 1: fn = 0;  fn1 = fn;  state = 2;  break;
673%         case 2: fn = 1;  fn2 = fn1;  fn1 = fn;  state = 3;  break;
674%         case 3: fn = fn1 + fn2;  fn2 = fn1;  fn1 = fn;  break;
675%       }
676%       return fn;
677% }
678% int main() {
679%
680%       for ( int i = 0; i < 10; i += 1 ) {
681%               printf( "%d\n", fib() );
682%       }
683% }
684% \end{cfa}
685% \end{lrbox}
686\begin{lrbox}{\myboxA}
687\begin{cfa}[aboveskip=0pt,belowskip=0pt]
688#define FIB_INIT { 0, 1 }
689typedef struct { int fn1, fn; } Fib;
690int fib( Fib * f ) {
691
692        int ret = f->fn1;
693        f->fn1 = f->fn;
694        f->fn = ret + f->fn;
695        return ret;
696}
697
698
699
700int main() {
701        Fib f1 = FIB_INIT, f2 = FIB_INIT;
702        for ( int i = 0; i < 10; i += 1 ) {
703                printf( "%d %d\n",
704                                fib( &f1 ), fib( &f2 ) );
705        }
706}
707\end{cfa}
708\end{lrbox}
709
710\newbox\myboxB
711\begin{lrbox}{\myboxB}
712\begin{cfa}[aboveskip=0pt,belowskip=0pt]
713`coroutine` Fib { int fn1; };
714void main( Fib & fib ) with( fib ) {
715        int fn;
716        [fn1, fn] = [0, 1];
717        for () {
718                `suspend();`
719                [fn1, fn] = [fn, fn1 + fn];
720        }
721}
722int ?()( Fib & fib ) with( fib ) {
723        `resume( fib );`  return fn1;
724}
725int main() {
726        Fib f1, f2;
727        for ( 10 ) {
728                sout | f1() | f2();
729}
730
731
732\end{cfa}
733\end{lrbox}
734
735\newbox\myboxC
736\begin{lrbox}{\myboxC}
737\begin{python}[aboveskip=0pt,belowskip=0pt]
738
739def Fib():
740
741    fn1, fn = 0, 1
742    while True:
743        `yield fn1`
744        fn1, fn = fn, fn1 + fn
745
746
747// next prewritten
748
749
750f1 = Fib()
751f2 = Fib()
752for i in range( 10 ):
753        print( next( f1 ), next( f2 ) )
754
755
756
757\end{python}
758\end{lrbox}
759
760\subfloat[C]{\label{f:GlobalVariables}\usebox\myboxA}
761\hspace{3pt}
762\vrule
763\hspace{3pt}
764\subfloat[\CFA]{\label{f:ExternalState}\usebox\myboxB}
765\hspace{3pt}
766\vrule
767\hspace{3pt}
768\subfloat[Python]{\label{f:ExternalState}\usebox\myboxC}
769\caption{Fibonacci Generator}
770\label{f:C-fibonacci}
771
772% \bigskip
773%
774% \newbox\myboxA
775% \begin{lrbox}{\myboxA}
776% \begin{cfa}[aboveskip=0pt,belowskip=0pt]
777% `coroutine` Fib { int fn; };
778% void main( Fib & fib ) with( fib ) {
779%       fn = 0;  int fn1 = fn; `suspend()`;
780%       fn = 1;  int fn2 = fn1;  fn1 = fn; `suspend()`;
781%       for () {
782%               fn = fn1 + fn2; fn2 = fn1; fn1 = fn; `suspend()`; }
783% }
784% int next( Fib & fib ) with( fib ) { `resume( fib );` return fn; }
785% int main() {
786%       Fib f1, f2;
787%       for ( 10 )
788%               sout | next( f1 ) | next( f2 );
789% }
790% \end{cfa}
791% \end{lrbox}
792% \newbox\myboxB
793% \begin{lrbox}{\myboxB}
794% \begin{python}[aboveskip=0pt,belowskip=0pt]
795%
796% def Fibonacci():
797%       fn = 0; fn1 = fn; `yield fn`  # suspend
798%       fn = 1; fn2 = fn1; fn1 = fn; `yield fn`
799%       while True:
800%               fn = fn1 + fn2; fn2 = fn1; fn1 = fn; `yield fn`
801%
802%
803% f1 = Fibonacci()
804% f2 = Fibonacci()
805% for i in range( 10 ):
806%       print( `next( f1 )`, `next( f2 )` ) # resume
807%
808% \end{python}
809% \end{lrbox}
810% \subfloat[\CFA]{\label{f:Coroutine3States}\usebox\myboxA}
811% \qquad
812% \subfloat[Python]{\label{f:Coroutine1State}\usebox\myboxB}
813% \caption{Fibonacci input coroutine, 3 states, internal variables}
814% \label{f:cfa-fibonacci}
815\end{figure}
816
817Using a coroutine, it is possible to express the Fibonacci formula directly without any of the C problems.
818Figure~\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@.
819Like 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.
820The 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@.
821The interface routine @next@, takes a Fibonacci instance and context switches to it using @resume@;
822on restart, the Fibonacci field, @fn@, contains the next value in the sequence, which is returned.
823The first @resume@ is special because it allocates the coroutine stack and cocalls its coroutine main on that stack;
824when the coroutine main returns, its stack is deallocated.
825Hence, @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.
826Figure~\ref{f:Coroutine1State} shows the coroutine version of the C version in Figure~\ref{f:ExternalState}.
827Coroutine generators are called \newterm{output coroutines} because values are only returned.
828
829Figure~\ref{f:CFAFmt} shows an \newterm{input coroutine}, @Format@, for restructuring text into groups of characters of fixed-size blocks.
830For example, the input of the left is reformatted into the output on the right.
831\begin{quote}
832\tt
833\begin{tabular}{@{}l|l@{}}
834\multicolumn{1}{c|}{\textbf{\textrm{input}}} & \multicolumn{1}{c}{\textbf{\textrm{output}}} \\
835abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz
836&
837\begin{tabular}[t]{@{}lllll@{}}
838abcd    & efgh  & ijkl  & mnop  & qrst  \\
839uvwx    & yzab  & cdef  & ghij  & klmn  \\
840opqr    & stuv  & wxyz  &               &
841\end{tabular}
842\end{tabular}
843\end{quote}
844The 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.
845The destructor provides a newline, if formatted text ends with a full line.
846Figure~\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.
847(Linearized code is the bane of device drivers.)
848
849\begin{figure}
850\centering
851\newbox\myboxA
852\begin{lrbox}{\myboxA}
853\begin{cfa}[aboveskip=0pt,belowskip=0pt]
854`coroutine` Fmt {
855        char ch;   // communication variables
856        int g, b;   // needed in destructor
857};
858void main( Fmt & fmt ) with( fmt ) {
859        for () {
860                for ( g = 0; g < 5; g += 1 ) { // groups
861                        for ( b = 0; b < 4; b += 1 ) { // blocks
862                                `suspend();`
863                                sout | ch; } // print character
864                        sout | "  "; } // block separator
865                sout | nl; }  // group separator
866}
867void ?{}( Fmt & fmt ) { `resume( fmt );` } // prime
868void ^?{}( Fmt & fmt ) with( fmt ) { // destructor
869        if ( g != 0 || b != 0 ) // special case
870                sout | nl; }
871void send( Fmt & fmt, char c ) { fmt.ch = c; `resume( fmt )`; }
872int main() {
873        Fmt fmt;
874        sout | nlOff;   // turn off auto newline
875        for ( 41 )
876                send( fmt, 'a' );
877}
878\end{cfa}
879\end{lrbox}
880
881\newbox\myboxB
882\begin{lrbox}{\myboxB}
883\begin{python}[aboveskip=0pt,belowskip=0pt]
884
885
886
887def Fmt():
888        try:
889                while True:
890                        for g in range( 5 ):
891                                for b in range( 4 ):
892
893                                        print( `(yield)`, end='' )
894                                print( '  ', end='' )
895                        print()
896
897
898        except GeneratorExit:
899                if g != 0 | b != 0:
900                        print()
901
902
903fmt = Fmt()
904`next( fmt )`                    # prime
905for i in range( 41 ):
906        `fmt.send( 'a' );`      # send to yield
907
908\end{python}
909\end{lrbox}
910\subfloat[\CFA]{\label{f:CFAFmt}\usebox\myboxA}
911\qquad
912\subfloat[Python]{\label{f:CFmt}\usebox\myboxB}
913\caption{Output formatting text}
914\label{f:fmt-line}
915\end{figure}
916
917The 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.
918However, @resume@ and @suspend@ context switch among existing stack-frames, rather than create new ones so there is no stack growth.
919\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.
920(The trivial cycle is a coroutine resuming itself.)
921This control flow is similar to recursion for normal routines, but again there is no stack growth from the context switch.
922
923\begin{figure}
924\centering
925\lstset{language=CFA,escapechar={},moredelim=**[is][\protect\color{red}]{`}{`}}% allow $
926\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
927\begin{cfa}
928`coroutine` Prod {
929        Cons & c;
930        int N, money, receipt;
931};
932void main( Prod & prod ) with( prod ) {
933        // 1st resume starts here
934        for ( i; N ) {
935                int p1 = random( 100 ), p2 = random( 100 );
936                sout | p1 | " " | p2;
937                int status = delivery( c, p1, p2 );
938                sout | " $" | money | nl | status;
939                receipt += 1;
940        }
941        stop( c );
942        sout | "prod stops";
943}
944int payment( Prod & prod, int money ) {
945        prod.money = money;
946        `resume( prod );`
947        return prod.receipt;
948}
949void start( Prod & prod, int N, Cons &c ) {
950        &prod.c = &c; // reassignable reference
951        prod.[N, receipt] = [N, 0];
952        `resume( prod );`
953}
954int main() {
955        Prod prod;
956        Cons cons = { prod };
957        start( prod, 5, cons );
958}
959\end{cfa}
960&
961\begin{cfa}
962`coroutine` Cons {
963        Prod & p;
964        int p1, p2, status;
965        bool done;
966};
967void ?{}( Cons & cons, Prod & p ) {
968        &cons.p = &p; // reassignable reference
969        cons.[status, done ] = [0, false];
970}
971void ^?{}( Cons & cons ) {}
972void main( Cons & cons ) with( cons ) {
973        // 1st resume starts here
974        int money = 1, receipt;
975        for ( ; ! done; ) {
976                sout | p1 | " " | p2 | nl | " $" | money;
977                status += 1;
978                receipt = payment( p, money );
979                sout | " #" | receipt;
980                money += 1;
981        }
982        sout | "cons stops";
983}
984int delivery( Cons & cons, int p1, int p2 ) {
985        cons.[p1, p2] = [p1, p2];
986        `resume( cons );`
987        return cons.status;
988}
989void stop( Cons & cons ) {
990        cons.done = true;
991        `resume( cons );`
992}
993\end{cfa}
994\end{tabular}
995\caption{Producer / consumer: resume-resume cycle, bi-directional communication}
996\label{f:ProdCons}
997\end{figure}
998
999Figure~\ref{f:ProdCons} shows a producer/consumer symmetric-coroutine performing bi-directional communication.
1000Since 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@.
1001The @start@ routine communicates both the number of elements to be produced and the consumer into the producer's coroutine-structure.
1002Then the @resume@ to @prod@ creates @prod@'s stack with a frame for @prod@'s coroutine main at the top, and context switches to it.
1003@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.
1004
1005The producer call to @delivery@ transfers values into the consumer's communication variables, resumes the consumer, and returns the consumer status.
1006For the first resume, @cons@'s stack is initialized, creating local variables retained between subsequent activations of the coroutine.
1007The 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).
1008The call from the consumer to @payment@ introduces the cycle between producer and consumer.
1009When @payment@ is called, the consumer copies values into the producer's communication variable and a resume is executed.
1010The context switch restarts the producer at the point where it last context switched, so it continues in @delivery@ after the resume.
1011
1012@delivery@ returns the status value in @prod@'s coroutine main, where the status is printed.
1013The loop then repeats calling @delivery@, where each call resumes the consumer coroutine.
1014The context switch to the consumer continues in @payment@.
1015The consumer increments and returns the receipt to the call in @cons@'s coroutine main.
1016The loop then repeats calling @payment@, where each call resumes the producer coroutine.
1017
1018After iterating $N$ times, the producer calls @stop@.
1019The @done@ flag is set to stop the consumer's execution and a resume is executed.
1020The context switch restarts @cons@ in @payment@ and it returns with the last receipt.
1021The 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@.
1022@stop@ returns and @prod@'s coroutine main terminates.
1023The program main restarts after the resume in @start@.
1024@start@ returns and the program main terminates.
1025
1026One \emph{killer} application for a coroutine is device drivers, which at one time caused 70\%-85\% of failures in Windows/Linux~\cite{Swift05}.
1027Many device drivers are a finite state-machine parsing a protocol, e.g.:
1028\begin{tabbing}
1029\ldots STX \= \ldots message \ldots \= ESC \= ETX \= \ldots message \ldots  \= ETX \= 2-byte crc \= \ldots      \kill
1030\ldots STX \> \ldots message \ldots \> ESC \> ETX \> \ldots message \ldots  \> ETX \> 2-byte crc \> \ldots
1031\end{tabbing}
1032where a network message begins with the control character STX and ends with an ETX, followed by a 2-byte cyclic-redundancy check.
1033Control characters may appear in a message if preceded by an ESC.
1034Because FSMs can be complex and occur frequently in important domains, direct support of the coroutine is crucial in a systems programminglanguage.
1035
1036\begin{figure}
1037\begin{cfa}
1038enum Status { CONT, MSG, ESTX, ELNTH, ECRC };
1039`coroutine` Driver {
1040        Status status;
1041        char * msg, byte;
1042};
1043void ?{}( Driver & d, char * m ) { d.msg = m; }         $\C[3.0in]{// constructor}$
1044Status next( Driver & d, char b ) with( d ) {           $\C{// 'with' opens scope}$
1045        byte = b; `resume( d );` return status;
1046}
1047void main( Driver & d ) with( d ) {
1048        enum { STX = '\002', ESC = '\033', ETX = '\003', MaxMsg = 64 };
1049        unsigned short int crc;                                                 $\C{// error checking}$
1050  msg: for () {                                                                         $\C{// parse message}$
1051                status = CONT;
1052                unsigned int lnth = 0, sum = 0;
1053                while ( byte != STX ) `suspend();`
1054          emsg: for () {
1055                        `suspend();`                                                    $\C{// process byte}$
1056                        choose ( byte ) {                                               $\C{// switch with default break}$
1057                          case STX:
1058                                status = ESTX; `suspend();` continue msg;
1059                          case ETX:
1060                                break emsg;
1061                          case ESC:
1062                                suspend();
1063                        } // choose
1064                        if ( lnth >= MaxMsg ) {                                 $\C{// buffer full ?}$
1065                                status = ELNTH; `suspend();` continue msg; }
1066                        msg[lnth++] = byte;
1067                        sum += byte;
1068                } // for
1069                msg[lnth] = '\0';                                                       $\C{// terminate string}\CRT$
1070                `suspend();`
1071                crc = (unsigned char)byte << 8; // prevent sign extension for signed char
1072                `suspend();`
1073                status = (crc | (unsigned char)byte) == sum ? MSG : ECRC;
1074                `suspend();`
1075        } // for
1076}
1077\end{cfa}
1078\caption{Device driver for simple communication protocol}
1079\end{figure}
1080
1081
1082\subsection{Coroutine Implementation}
1083
1084A 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.
1085There are several solutions to this problem and the chosen option forced the \CFA coroutine design.
1086
1087Object-oriented inheritance provides extra fields and code in a restricted context, but it requires programmers to explicitly perform the inheritance:
1088\begin{cfa}[morekeywords={class,inherits}]
1089class mycoroutine inherits baseCoroutine { ... }
1090\end{cfa}
1091and the programming language (and possibly its tool set, \eg debugger) may need to understand @baseCoroutine@ because of the stack.
1092Furthermore, the execution of constructors/destructors is in the wrong order for certain operations.
1093For 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.
1094
1095An alternative is composition:
1096\begin{cfa}
1097struct mycoroutine {
1098        ... // declarations
1099        baseCoroutine dummy; // composition, last declaration
1100}
1101\end{cfa}
1102which also requires an explicit declaration that must be the last one to ensure correct initialization order.
1103However, there is nothing preventing wrong placement or multiple declarations.
1104
1105For coroutines as for threads, many implementations are based on routine pointers or routine objects~\cite{Butenhof97, C++14, MS:VisualC++, BoostCoroutines15}.
1106For example, Boost implements coroutines in terms of four functor object-types:
1107\begin{cfa}
1108asymmetric_coroutine<>::pull_type
1109asymmetric_coroutine<>::push_type
1110symmetric_coroutine<>::call_type
1111symmetric_coroutine<>::yield_type
1112\end{cfa}
1113Similarly, the canonical threading paradigm is often based on routine pointers, \eg @pthreads@~\cite{Butenhof97}, \Csharp~\cite{Csharp}, Go~\cite{Go}, and Scala~\cite{Scala}.
1114However, the generic thread-handle (identifier) is limited (few operations), unless it is wrapped in a custom type.
1115\begin{cfa}
1116void mycor( coroutine_t cid, void * arg ) {
1117        int * value = (int *)arg;                               $\C{// type unsafe, pointer-size only}$
1118        // Coroutine body
1119}
1120int main() {
1121        int input = 0, output;
1122        coroutine_t cid = coroutine_create( &mycor, (void *)&input ); $\C{// type unsafe, pointer-size only}$
1123        coroutine_resume( cid, (void *)input, (void **)&output ); $\C{// type unsafe, pointer-size only}$
1124}
1125\end{cfa}
1126Since the custom type is simple to write in \CFA and solves several issues, added support for routine/lambda-based coroutines adds very little.
1127
1128Note, the type @coroutine_t@ must be an abstract handle to the coroutine, because the coroutine descriptor and its stack are non-copyable.
1129Copying the coroutine descriptor results in copies being out of date with the current state of the stack.
1130Correspondingly, 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.
1131(There is no mechanism in C to find all stack-specific pointers and update them as part of a copy.)
1132
1133The selected approach is to use language support by introducing a new kind of aggregate (structure):
1134\begin{cfa}
1135coroutine Fibonacci {
1136        int fn; // communication variables
1137};
1138\end{cfa}
1139The @coroutine@ keyword means the compiler (and tool set) can find and inject code where needed.
1140The downside of this approach is that it makes coroutine a special case in the language.
1141Users wanting to extend coroutines or build their own for various reasons can only do so in ways offered by the language.
1142Furthermore, implementing coroutines without language supports also displays the power of a programming language.
1143While this is ultimately the option used for idiomatic \CFA code, coroutines and threads can still be constructed without language support.
1144The reserved keyword simply eases use for the common case.
1145
1146Part 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:
1147\begin{cfa}
1148trait is_coroutine( `dtype` T ) {
1149        void main( T & );
1150        coroutine_desc * get_coroutine( T & );
1151};
1152forall( `dtype` T | is_coroutine(T) ) void suspend( T & );
1153forall( `dtype` T | is_coroutine(T) ) void resume( T & );
1154\end{cfa}
1155The @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).
1156The 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.
1157The @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.
1158The 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@.
1159The \CFA keyword @coroutine@ implicitly implements the getter and forward declarations required for implementing the coroutine main:
1160\begin{cquote}
1161\begin{tabular}{@{}ccc@{}}
1162\begin{cfa}
1163coroutine MyCor {
1164        int value;
1165
1166};
1167\end{cfa}
1168&
1169{\Large $\Rightarrow$}
1170&
1171\begin{tabular}{@{}ccc@{}}
1172\begin{cfa}
1173struct MyCor {
1174        int value;
1175        coroutine_desc cor;
1176};
1177\end{cfa}
1178&
1179\begin{cfa}
1180static inline coroutine_desc *
1181get_coroutine( MyCor & this ) {
1182        return &this.cor;
1183}
1184\end{cfa}
1185&
1186\begin{cfa}
1187void main( MyCor * this );
1188
1189
1190
1191\end{cfa}
1192\end{tabular}
1193\end{tabular}
1194\end{cquote}
1195The 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.
1196
1197
1198\section{Concurrency}
1199\label{s:Concurrency}
1200
1201At its core, concurrency is based on multiple call-stacks and scheduling threads executing on these stacks.
1202Multiple 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}.
1203In 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.
1204A \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;
1205a \newterm{stackful} coroutine executes on its own stack, allowing full generality.
1206Only stackful coroutines are a stepping stone to concurrency.
1207
1208The 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}.
1209Therefore, a minimal concurrency system is possible using coroutines (see Section \ref{coroutine}) in conjunction with a scheduler to decide where to context switch next.
1210The resulting execution system now follows a cooperative threading-model, called \newterm{non-preemptive scheduling}.
1211
1212Because the scheduler is special, it can either be a stackless or stackful coroutine.
1213For 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.
1214For 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.
1215A stackful scheduler is often used for simplicity and security.
1216
1217Regardless of the approach used, a subset of concurrency related challenges start to appear.
1218For 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}.
1219While a scheduler introduces uncertainty in the order of execution, preemption introduces uncertainty about where context switches occur.
1220Interestingly, uncertainty is necessary for the runtime (operating) system to give the illusion of parallelism on a single processor and increase performance on multiple processors.
1221The reason is that only the runtime has complete knowledge about resources and how to best utilized them.
1222However, the introduction of unrestricted non-determinism results in the need for \newterm{mutual exclusion} and \newterm{synchronization} to restrict non-determinism for correctness;
1223otherwise, it is impossible to write meaningful programs.
1224Optimal performance in concurrent applications is often obtained by having as much non-determinism as correctness allows.
1225
1226An 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.
1227As such, library support for threading is far from widespread.
1228At the time of writing the paper, neither \protect\lstinline@gcc@ nor \protect\lstinline@clang@ support \protect\lstinline@threads.h@ in their standard libraries.}.
1229In 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.
1230As 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.
1231Furthermore, because C is a system-level language, programmers expect to choose precisely which features they need and which cost they are willing to pay.
1232Hence, concurrent programs should be written using high-level mechanisms, and only step down to lower-level mechanisms when performance bottlenecks are encountered.
1233
1234
1235\subsection{Thread Interface}
1236\label{threads}
1237
1238Both user and kernel threads are supported, where user threads provide concurrency and kernel threads provide parallelism.
1239Like 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:
1240\begin{cquote}
1241\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1242\begin{cfa}
1243thread myThread {
1244        // communication variables
1245};
1246
1247
1248\end{cfa}
1249&
1250\begin{cfa}
1251trait is_thread( `dtype` T ) {
1252      void main( T & );
1253      thread_desc * get_thread( T & );
1254      void ^?{}( T & `mutex` );
1255};
1256\end{cfa}
1257\end{tabular}
1258\end{cquote}
1259(The qualifier @mutex@ for the destructor parameter is discussed in Section~\ref{s:Monitor}.)
1260Like a coroutine, the statically-typed @main@ routine is the starting point (first stack frame) of a user thread.
1261The difference is that a coroutine borrows a thread from its caller, so the first thread resuming a coroutine creates an instance of @main@;
1262whereas, 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{
1263The \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.}
1264No 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.
1265
1266\begin{comment} % put in appendix with coroutine version ???
1267As such the @main@ routine of a thread can be defined as
1268\begin{cfa}
1269thread foo {};
1270
1271void main(foo & this) {
1272        sout | "Hello World!";
1273}
1274\end{cfa}
1275
1276In 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.
1277With 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.
1278\begin{cfa}
1279typedef void (*voidRtn)(int);
1280
1281thread RtnRunner {
1282        voidRtn func;
1283        int arg;
1284};
1285
1286void ?{}(RtnRunner & this, voidRtn inRtn, int arg) {
1287        this.func = inRtn;
1288        this.arg  = arg;
1289}
1290
1291void main(RtnRunner & this) {
1292        // thread starts here and runs the routine
1293        this.func( this.arg );
1294}
1295
1296void hello(/*unused*/ int) {
1297        sout | "Hello World!";
1298}
1299
1300int main() {
1301        RtnRunner f = {hello, 42};
1302        return 0?
1303}
1304\end{cfa}
1305A 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}.
1306\end{comment}
1307
1308For user threads to be useful, it must be possible to start and stop the underlying thread, and wait for it to complete execution.
1309While using an API such as @fork@ and @join@ is relatively common, such an interface is awkward and unnecessary.
1310A simple approach is to use allocation/deallocation principles, and have threads implicitly @fork@ after construction and @join@ before destruction.
1311\begin{cfa}
1312thread World {};
1313void main( World & this ) {
1314        sout | "World!";
1315}
1316int main() {
1317        World w`[10]`;                                                  $\C{// implicit forks after creation}$
1318        sout | "Hello ";                                        $\C{// "Hello " and 10 "World!" printed concurrently}$
1319}                                                                                       $\C{// implicit joins before destruction}$
1320\end{cfa}
1321This semantics ensures a thread is started and stopped exactly once, eliminating some programming error, and scales to multiple threads for basic (termination) synchronization.
1322This 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.
1323\begin{cfa}
1324int main() {
1325        MyThread * heapLive;
1326        {
1327                MyThread blockLive;                                     $\C{// fork block-based thread}$
1328                heapLive = `new`( MyThread );           $\C{// fork heap-based thread}$
1329                ...
1330        }                                                                               $\C{// join block-based thread}$
1331        ...
1332        `delete`( heapLive );                                   $\C{// join heap-based thread}$
1333}
1334\end{cfa}
1335The heap-based approach allows arbitrary thread-creation topologies, with respect to fork/join-style concurrency.
1336
1337Figure~\ref{s:ConcurrentMatrixSummation} shows concurrently adding the rows of a matrix and then totalling the subtotals sequentially, after all the row threads have terminated.
1338The program uses heap-based threads because each thread needs different constructor values.
1339(Python provides a simple iteration mechanism to initialize array elements to different values allowing stack allocation.)
1340The allocation/deallocation pattern appears unusual because allocated objects are immediately deallocated without any intervening code.
1341However, for threads, the deletion provides implicit synchronization, which is the intervening code.
1342While 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.
1343
1344\begin{figure}
1345\begin{cfa}
1346`thread` Adder { int * row, cols, & subtotal; } $\C{// communication variables}$
1347void ?{}( Adder & adder, int row[], int cols, int & subtotal ) {
1348    adder.[ row, cols, &subtotal ] = [ row, cols, &subtotal ];
1349}
1350void main( Adder & adder ) with( adder ) {
1351    subtotal = 0;
1352    for ( int c = 0; c < cols; c += 1 ) { subtotal += row[c]; }
1353}
1354int main() {
1355    const int rows = 10, cols = 1000;
1356    int matrix[rows][cols], subtotals[rows], total = 0;
1357    // read matrix
1358    Adder * adders[rows];
1359    for ( int r = 0; r < rows; r += 1 ) {       $\C{// start threads to sum rows}$
1360                adders[r] = `new( matrix[r], cols, &subtotals[r] );`
1361    }
1362    for ( int r = 0; r < rows; r += 1 ) {       $\C{// wait for threads to finish}$
1363                `delete( adders[r] );`                          $\C{// termination join}$
1364                total += subtotals[r];                          $\C{// total subtotal}$
1365    }
1366    sout | total;
1367}
1368\end{cfa}
1369\caption{Concurrent Matrix Summation}
1370\label{s:ConcurrentMatrixSummation}
1371\end{figure}
1372
1373
1374\section{Mutual Exclusion / Synchronization}
1375
1376Uncontrolled non-deterministic execution is meaningless.
1377To 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}.
1378Since 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).
1379In 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}.
1380However, 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.
1381Hence, a programmer must learn and manipulate two sets of design patterns.
1382While this distinction can be hidden away in library code, effective use of the library still has to take both paradigms into account.
1383In contrast, approaches based on stateful models more closely resemble the standard call/return programming-model, resulting in a single programming paradigm.
1384
1385At 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}.
1386However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
1387A newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
1388While 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.
1389
1390One of the most natural, elegant, and efficient mechanisms for mutual exclusion and synchronization for shared-memory systems is the \emph{monitor}.
1391First 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}.
1392In 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.
1393For these reasons, \CFA selected monitors as the core high-level concurrency-construct, upon which higher-level approaches can be easily constructed.
1394
1395
1396\subsection{Mutual Exclusion}
1397
1398A group of instructions manipulating a specific instance of shared data that must be performed atomically is called an (individual) \newterm{critical-section}~\cite{Dijkstra65}.
1399The 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.
1400The 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.
1401\newterm{Mutual exclusion} enforces that the correct kind and number of threads are using a critical section.
1402
1403However, many solutions exist for mutual exclusion, which vary in terms of performance, flexibility and ease of use.
1404Methods 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.
1405Ease 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.
1406For 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.
1407However, a significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
1408Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
1409
1410
1411\subsection{Synchronization}
1412
1413Synchronization enforces relative ordering of execution, and synchronization tools provide numerous mechanisms to establish these timing relationships.
1414Low-level synchronization primitives offer good performance and flexibility at the cost of ease of use;
1415higher-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.
1416Often 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.
1417If a writer thread is scheduled for next access, but another reader thread acquires the critical section first, that reader \newterm{barged}.
1418Barging 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).
1419Preventing or detecting barging is an involved challenge with low-level locks, which can be made much easier by higher-level constructs.
1420This challenge is often split into two different approaches: barging avoidance and barging prevention.
1421Algorithms that allow a barger, but divert it until later using current synchronization state (flags), are avoiding the barger;
1422algorithms that preclude a barger from entering during synchronization in the critical section prevent barging completely.
1423Techniques like baton-passing locks~\cite{Andrews89} between threads instead of unconditionally releasing locks is an example of barging prevention.
1424
1425
1426\section{Monitor}
1427\label{s:Monitor}
1428
1429A \textbf{monitor} is a set of routines that ensure mutual exclusion when accessing shared state.
1430More 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).
1431The strong association with the call/return paradigm eases programmability, readability and maintainability, at a slight cost in flexibility and efficiency.
1432
1433Note, 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.
1434Copying 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.
1435Copying 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.
1436As 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).
1437\begin{cfa}
1438trait is_monitor( `dtype` T ) {
1439        monitor_desc * get_monitor( T & );
1440        void ^?{}( T & mutex );
1441};
1442\end{cfa}
1443
1444
1445\subsection{Mutex Acquisition}
1446\label{s:MutexAcquisition}
1447
1448While correctness implies a monitor's mutual exclusion is acquired and released, there are implementation options about when and where the locking/unlocking occurs.
1449(Much of this discussion also applies to basic locks.)
1450For example, a monitor may need to be passed through multiple helper routines before it becomes necessary to acquire the monitor mutual-exclusion.
1451\begin{cfa}[morekeywords=nomutex]
1452monitor Aint { int cnt; };                                      $\C{// atomic integer counter}$
1453void ?{}( Aint & `nomutex` this ) with( this ) { cnt = 0; } $\C{// constructor}$
1454int ?=?( Aint & `mutex`$\(_{opt}\)$ lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions}$
1455void ?{}( int & this, Aint & `mutex`$\(_{opt}\)$ v ) { this = v.cnt; }
1456int ?=?( int & lhs, Aint & `mutex`$\(_{opt}\)$ rhs ) with( rhs ) { lhs = cnt; }
1457int ++?( Aint & `mutex`$\(_{opt}\)$ this ) with( this ) { return ++cnt; } $\C{// increment}$
1458\end{cfa}
1459The @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.
1460(While a constructor may publish its address into a global variable, doing so generates a race-condition.)
1461The conversion operators for initializing and assigning with a normal integer only need @mutex@, if reading/writing the implementation type is not atomic.
1462Finally, 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.
1463
1464The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics, which is similar to the \CC template @std::atomic@.
1465\begin{cfa}
1466Aint x, y, z;
1467++x; ++y; ++z;                                                          $\C{// safe increment by multiple threads}$
1468x = 2; y = 2; z = 2;                                            $\C{// conversions}$
1469int i = x, j = y, k = z;
1470i = x; j = y; k = z;
1471\end{cfa}
1472
1473For maximum usability, monitors have \newterm{multi-acquire} semantics allowing a thread to acquire it multiple times without deadlock.
1474\begin{cfa}
1475monitor M { ... } m;
1476void foo( M & mutex m ) { ... }                         $\C{// acquire mutual exclusion}$
1477void bar( M & mutex m ) {                                       $\C{// acquire mutual exclusion}$
1478        ... `foo( m );` ...                                             $\C{// reacquire mutual exclusion}$
1479}
1480`bar( m );`                                                                     $\C{// nested monitor call}$
1481\end{cfa}
1482
1483The benefit of mandatory monitor qualifiers is self-documentation, but requiring both @mutex@ and \lstinline[morekeywords=nomutex]@nomutex@ for all monitor parameters is redundant.
1484Instead, the semantics have one qualifier as the default, and the other required.
1485For example, make the safe @mutex@ qualifier the default because assuming \lstinline[morekeywords=nomutex]@nomutex@ may cause subtle errors.
1486Alternatively, make the unsafe \lstinline[morekeywords=nomutex]@nomutex@ qualifier the default because it is the \emph{normal} parameter semantics while @mutex@ parameters are rare.
1487Providing a default qualifier implies knowing whether a parameter is a monitor.
1488Since \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.
1489For 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@.
1490
1491The next semantic decision is establishing which parameter \emph{types} may be qualified with @mutex@.
1492Given:
1493\begin{cfa}
1494monitor M { ... }
1495int f1( M & mutex m );
1496int f2( M * mutex m );
1497int f3( M * mutex m[] );
1498int f4( stack( M * ) & mutex m );
1499\end{cfa}
1500the issue is that some of these parameter types are composed of multiple objects.
1501For @f1@, there is only a single parameter object.
1502Adding indirection in @f2@ still identifies a single object.
1503However, the matrix in @f3@ introduces multiple objects.
1504While shown shortly, multiple acquisition is possible;
1505however array lengths are often unknown in C.
1506This issue is exacerbated in @f4@, where the data structure must be safely traversed to acquire all of its elements.
1507
1508To make the issue tractable, \CFA only acquires one monitor per parameter with at most one level of indirection.
1509However, there is an ambiguity in the C type-system with respects to arrays.
1510Is the argument for @f2@ a single object or an array of objects?
1511If it is an array, only the first element of the array is acquired, which seems unsafe;
1512hence, @mutex@ is disallowed for array parameters.
1513\begin{cfa}
1514int f1( M & mutex m );                                          $\C{// allowed: recommended case}$
1515int f2( M * mutex m );                                          $\C{// disallowed: could be an array}$
1516int f3( M mutex m[$\,$] );                                      $\C{// disallowed: array length unknown}$
1517int f4( M ** mutex m );                                         $\C{// disallowed: could be an array}$
1518int f5( M * mutex m[$\,$] );                            $\C{// disallowed: array length unknown}$
1519\end{cfa}
1520% Note, not all array routines have distinct types: @f2@ and @f3@ have the same type, as do @f4@ and @f5@.
1521% However, even if the code generation could tell the difference, the extra information is still not sufficient to extend meaningfully the monitor call semantic.
1522
1523For object-oriented monitors, calling a mutex member \emph{implicitly} acquires mutual exclusion of the receiver object, @`rec`.foo(...)@.
1524\CFA has no receiver, and hence, must use an explicit mechanism to specify which object acquires mutual exclusion.
1525A positive consequence of this design decision is the ability to support multi-monitor routines.
1526\begin{cfa}
1527int f( M & mutex x, M & mutex y );              $\C{// multiple monitor parameter of any type}$
1528M m1, m2;
1529f( m1, m2 );
1530\end{cfa}
1531(While object-oriented monitors can be extended with a mutex qualifier for multiple-monitor members, no prior example of this feature could be found.)
1532In practice, writing multi-locking routines that do not deadlock is tricky.
1533Having language support for such a feature is therefore a significant asset for \CFA.
1534
1535The capability to acquire multiple locks before entering a critical section is called \newterm{bulk acquire} (see Section~\ref{s:Implementation} for implementation details).
1536In the previous example, \CFA guarantees the order of acquisition is consistent across calls to different routines using the same monitors as arguments.
1537This consistent ordering means acquiring multiple monitors is safe from deadlock.
1538However, users can force the acquiring order.
1539For example, notice the use of @mutex@/\lstinline[morekeywords=nomutex]@nomutex@ and how this affects the acquiring order:
1540\begin{cfa}
1541void foo( M & mutex m1, M & mutex m2 );         $\C{// acquire m1 and m2}$
1542void bar( M & mutex m1, M & /* nomutex */ m2 ) { $\C{// acquire m1}$
1543        ... foo( m1, m2 ); ...                                  $\C{// acquire m2}$
1544}
1545void baz( M & /* nomutex */ m1, M & mutex m2 ) { $\C{// acquire m2}$
1546        ... foo( m1, m2 ); ...                                  $\C{// acquire m1}$
1547}
1548\end{cfa}
1549The multi-acquire semantics allows @bar@ or @baz@ to acquire a monitor lock and reacquire it in @foo@.
1550In the calls to @bar@ and @baz@, the monitors are acquired in opposite order.
1551
1552However, 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}.
1553In \CFA, a safety aid is provided by using bulk acquire of all monitors to shared objects, whereas other monitor systems provide no aid.
1554While \CFA provides only a partial solution, it handles many useful cases, \eg:
1555\begin{cfa}
1556monitor BankAccount { ... };
1557void deposit( BankAccount & `mutex` b, int deposit );
1558void transfer( BankAccount & `mutex` my, BankAccount & `mutex` your, int me2you ) {
1559        deposit( my, `-`me2you );                               $\C{// debit}$
1560        deposit( your, me2you );                                $\C{// credit}$
1561}
1562\end{cfa}
1563This example shows a trivial solution to the bank-account transfer problem.
1564Without multi- and bulk acquire, the solution to this problem requires careful engineering.
1565
1566
1567\subsection{\protect\lstinline@mutex@ statement}
1568\label{mutex-stmt}
1569
1570The monitor call-semantics associate all locking semantics to routines.
1571Like Java, \CFA offers an alternative @mutex@ statement to reduce refactoring and naming.
1572\begin{cquote}
1573\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1574\begin{cfa}
1575monitor M { ... };
1576void foo( M & mutex m1, M & mutex m2 ) {
1577        // critical section
1578}
1579void bar( M & m1, M & m2 ) {
1580        foo( m1, m2 );
1581}
1582\end{cfa}
1583&
1584\begin{cfa}
1585
1586void bar( M & m1, M & m2 ) {
1587        mutex( m1, m2 ) {       // remove refactoring and naming
1588                // critical section
1589        }
1590}
1591
1592\end{cfa}
1593\\
1594\multicolumn{1}{c}{\textbf{routine call}} & \multicolumn{1}{c}{\lstinline@mutex@ \textbf{statement}}
1595\end{tabular}
1596\end{cquote}
1597
1598
1599\section{Scheduling}
1600\label{s:Scheduling}
1601
1602While monitor mutual-exclusion provides safe access to shared data, the monitor data may indicate that a thread accessing it cannot proceed.
1603For example, Figure~\ref{f:GenericBoundedBuffer} shows a bounded buffer that may be full/empty so produce/consumer threads must block.
1604Leaving the monitor and trying again (busy waiting) is impractical for high-level programming.
1605Monitors eliminate busy waiting by providing synchronization to schedule threads needing access to the shared data, where threads block versus spinning.
1606Synchronization 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.
1607\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.
1608
1609Figure~\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@.
1610The @wait@ routine atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the routine's parameter list.
1611The appropriate condition lock is signalled to unblock an opposite kind of thread after an element is inserted/removed from the buffer.
1612Signalling is unconditional, because signalling an empty condition lock does nothing.
1613
1614Signalling semantics cannot have the signaller and signalled thread in the monitor simultaneously, which means:
1615\begin{enumerate}
1616\item
1617The signalling thread returns immediately, and the signalled thread continues.
1618\item
1619The signalling thread continues and the signalled thread is marked for urgent unblocking at the next scheduling point (exit/wait).
1620\item
1621The signalling thread blocks but is marked for urgrent unblocking at the next scheduling point and the signalled thread continues.
1622\end{enumerate}
1623The first approach is too restrictive, as it precludes solving a reasonable class of problems, \eg dating service (see Figure~\ref{f:DatingService}).
1624\CFA supports the next two semantics as both are useful.
1625Finally, while it is common to store a @condition@ as a field of the monitor, in \CFA, a @condition@ variable can be created/stored independently.
1626Furthermore, 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.
1627
1628\begin{figure}
1629\centering
1630\newbox\myboxA
1631\begin{lrbox}{\myboxA}
1632\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1633forall( otype T ) { // distribute forall
1634        monitor Buffer {
1635                `condition` full, empty;
1636                int front, back, count;
1637                T elements[10];
1638        };
1639        void ?{}( Buffer(T) & buffer ) with(buffer) {
1640                [front, back, count] = 0;
1641        }
1642
1643        void insert( Buffer(T) & mutex buffer, T elem )
1644                                with(buffer) {
1645                if ( count == 10 ) `wait( empty )`;
1646                // insert elem into buffer
1647                `signal( full )`;
1648        }
1649        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1650                if ( count == 0 ) `wait( full )`;
1651                // remove elem from buffer
1652                `signal( empty )`;
1653                return elem;
1654        }
1655}
1656\end{cfa}
1657\end{lrbox}
1658
1659\newbox\myboxB
1660\begin{lrbox}{\myboxB}
1661\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1662forall( otype T ) { // distribute forall
1663        monitor Buffer {
1664
1665                int front, back, count;
1666                T elements[10];
1667        };
1668        void ?{}( Buffer(T) & buffer ) with(buffer) {
1669                [front, back, count] = 0;
1670        }
1671        T remove( Buffer(T) & mutex buffer ); // forward
1672        void insert( Buffer(T) & mutex buffer, T elem )
1673                                with(buffer) {
1674                if ( count == 10 ) `waitfor( remove, buffer )`;
1675                // insert elem into buffer
1676
1677        }
1678        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1679                if ( count == 0 ) `waitfor( insert, buffer )`;
1680                // remove elem from buffer
1681
1682                return elem;
1683        }
1684}
1685\end{cfa}
1686\end{lrbox}
1687
1688\subfloat[Internal Scheduling]{\label{f:BBInt}\usebox\myboxA}
1689%\qquad
1690\subfloat[External Scheduling]{\label{f:BBExt}\usebox\myboxB}
1691\caption{Generic Bounded-Buffer}
1692\label{f:GenericBoundedBuffer}
1693\end{figure}
1694
1695Figure~\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.
1696External 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.
1697If 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.
1698Threads 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.
1699External scheduling allows users to wait for events from other threads without concern of unrelated events occurring.
1700The mechnaism can be done in terms of control flow, \eg Ada @accept@ or \uC @_Accept@, or in terms of data, \eg Go channels.
1701While both mechanisms have strengths and weaknesses, this project uses a control-flow mechanism to stay consistent with other language semantics.
1702Two 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}).
1703
1704For internal scheduling, non-blocking signalling (as in the producer/consumer example) is used when the signaller is providing the cooperation for a waiting thread;
1705the 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.
1706The waiter unblocks next from the urgent queue, uses/takes the state, and exits the monitor.
1707Blocking signalling is the reverse, where the waiter is providing the cooperation for the signalling thread;
1708the 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.
1709The waiter changes state and exits the monitor, and the signaller unblocks next from the urgent queue to use/take the state.
1710
1711Figure~\ref{f:DatingService} shows a dating service demonstrating non-blocking and blocking signalling.
1712The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
1713A thread blocks until an appropriate partner arrives.
1714The complexity is exchanging phone numbers in the monitor because of the mutual-exclusion property.
1715For 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.
1716For 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.
1717
1718The 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;
1719as well, an arriving thread may not find a partner and must wait, which requires a condition variable, and condition variables imply internal scheduling.
1720
1721\begin{figure}
1722\centering
1723\newbox\myboxA
1724\begin{lrbox}{\myboxA}
1725\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1726enum { CCodes = 20 };
1727monitor DS {
1728        int GirlPhNo, BoyPhNo;
1729        condition Girls[CCodes], Boys[CCodes];
1730        condition exchange;
1731};
1732int girl( DS & mutex ds, int phNo, int ccode ) {
1733        if ( is_empty( Boys[ccode] ) ) {
1734                wait( Girls[ccode] );
1735                GirlPhNo = phNo;
1736                `signal( exchange );`
1737        } else {
1738                GirlPhNo = phNo;
1739                `signal( Boys[ccode] );`
1740                `wait( exchange );`
1741        } // if
1742        return BoyPhNo;
1743}
1744int boy( DS & mutex ds, int phNo, int ccode ) {
1745        // as above with boy/girl interchanged
1746}
1747\end{cfa}
1748\end{lrbox}
1749
1750\newbox\myboxB
1751\begin{lrbox}{\myboxB}
1752\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1753
1754monitor DS {
1755        int GirlPhNo, BoyPhNo;
1756        condition Girls[CCodes], Boys[CCodes];
1757
1758};
1759int girl( DS & mutex ds, int phNo, int ccode ) {
1760        if ( is_empty( Boys[ccode] ) ) { // no compatible
1761                wait( Girls[ccode] ); // wait for boy
1762                GirlPhNo = phNo; // make phone number available
1763
1764        } else {
1765                GirlPhNo = phNo; // make phone number available
1766                `signal_block( Boys[ccode] );` // restart boy
1767
1768        } // if
1769        return BoyPhNo;
1770}
1771int boy( DS & mutex ds, int phNo, int ccode ) {
1772        // as above with boy/girl interchanged
1773}
1774\end{cfa}
1775\end{lrbox}
1776
1777\subfloat[\lstinline@signal@]{\label{f:DatingSignal}\usebox\myboxA}
1778\qquad
1779\subfloat[\lstinline@signal_block@]{\label{f:DatingSignalBlock}\usebox\myboxB}
1780\caption{Dating service. }
1781\label{f:DatingService}
1782\end{figure}
1783
1784Both internal and external scheduling extend to multiple monitors in a natural way.
1785\begin{cquote}
1786\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1787\begin{cfa}
1788monitor M { `condition e`; ... };
1789void foo( M & mutex m1, M & mutex m2 ) {
1790        ... wait( `e` ); ...   // wait( e, m1, m2 )
1791        ... wait( `e, m1` ); ...
1792        ... wait( `e, m2` ); ...
1793}
1794\end{cfa}
1795&
1796\begin{cfa}
1797void rtn$\(_1\)$( M & mutex m1, M & mutex m2 );
1798void rtn$\(_2\)$( M & mutex m1 );
1799void bar( M & mutex m1, M & mutex m2 ) {
1800        ... waitfor( `rtn` ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\), m1, m2 )}$
1801        ... waitfor( `rtn, m1` ); ... // $\LstCommentStyle{waitfor( rtn\(_2\), m1 )}$
1802}
1803\end{cfa}
1804\end{tabular}
1805\end{cquote}
1806For @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 )@.
1807To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @wait( e, m1 )@.
1808Wait statically verifies the released monitors are the acquired mutex-parameters so unconditional release is safe.
1809Finally, a signaller,
1810\begin{cfa}
1811void baz( M & mutex m1, M & mutex m2 ) {
1812        ... signal( e ); ...
1813}
1814\end{cfa}
1815must have acquired at least the same locks as the waiting thread signalled from the condition queue.
1816
1817Similarly, 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 )@.
1818To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn, m1 )@.
1819@waitfor@ statically verifies the released monitors are the same as the acquired mutex-parameters of the given routine or routine pointer.
1820To statically verify the released monitors match with the accepted routine's mutex parameters, the routine (pointer) prototype must be accessible.
1821% When an overloaded routine appears in an @waitfor@ statement, calls to any routine with that name are accepted.
1822% The rationale is that members with the same name should perform a similar function, and therefore, all should be eligible to accept a call.
1823Overloaded routines can be disambiguated using a cast:
1824\begin{cfa}
1825void rtn( M & mutex m );
1826`int` rtn( M & mutex m );
1827waitfor( (`int` (*)( M & mutex ))rtn, m );
1828\end{cfa}
1829
1830The ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock.
1831\begin{cfa}
1832void foo( M & mutex m1, M & mutex m2 ) {
1833        ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired )}$
1834void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2 )}$
1835        ... signal( `e` ); ...
1836\end{cfa}
1837The @wait@ only releases @m1@ so the signalling thread cannot acquire both @m1@ and @m2@ to  enter @bar@ to get to the @signal@.
1838While deadlock issues can occur with multiple/nesting acquisition, this issue results from the fact that locks, and by extension monitors, are not perfectly composable.
1839
1840Finally, an important aspect of monitor implementation is barging, \ie can calling threads barge ahead of signalled threads?
1841If 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).
1842In fact, signals-as-hints is completely opposite from that proposed by Hoare in the seminal paper on monitors:
1843\begin{quote}
1844However, 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.
1845It 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}
1846\end{quote}
1847\CFA scheduling \emph{precludes} barging, which simplifies synchronization among threads in the monitor and increases correctness.
1848Furthermore, \CFA concurrency has no spurious wakeup~\cite[\S~9]{Buhr05a}, which eliminates an implict form of barging.
1849For example, there are no loops in either bounded buffer solution in Figure~\ref{f:GenericBoundedBuffer}.
1850Supporting 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.
1851
1852
1853\subsection{Barging Prevention}
1854
1855Figure~\ref{f:BargingPrevention} shows \CFA code where bulk acquire adds complexity to the internal-signalling semantics.
1856The complexity begins at the end of the inner @mutex@ statement, where the semantics of internal scheduling need to be extended for multiple monitors.
1857The problem is that bulk acquire is used in the inner @mutex@ statement where one of the monitors is already acquired.
1858When 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.
1859However, both the signalling and waiting thread W1 still need monitor @m1@.
1860
1861\begin{figure}
1862\newbox\myboxA
1863\begin{lrbox}{\myboxA}
1864\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1865monitor M m1, m2;
1866condition c;
1867mutex( m1 ) { // $\LstCommentStyle{\color{red}outer}$
1868        ...
1869        mutex( m1, m2 ) { // $\LstCommentStyle{\color{red}inner}$
1870                ... `signal( c )`; ...
1871                // m1, m2 acquired
1872        } // $\LstCommentStyle{\color{red}release m2}$
1873        // m1 acquired
1874} // release m1
1875\end{cfa}
1876\end{lrbox}
1877
1878\newbox\myboxB
1879\begin{lrbox}{\myboxB}
1880\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1881
1882
1883mutex( m1 ) {
1884        ...
1885        mutex( m1, m2 ) {
1886                ... `wait( c )`; // block and release m1, m2
1887                // m1, m2 acquired
1888        } // $\LstCommentStyle{\color{red}release m2}$
1889        // m1 acquired
1890} // release m1
1891\end{cfa}
1892\end{lrbox}
1893
1894\newbox\myboxC
1895\begin{lrbox}{\myboxC}
1896\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1897
1898
1899mutex( m2 ) {
1900        ... `wait( c )`; ...
1901        // m2 acquired
1902} // $\LstCommentStyle{\color{red}release m2}$
1903
1904
1905
1906
1907\end{cfa}
1908\end{lrbox}
1909
1910\begin{cquote}
1911\subfloat[Signalling Thread]{\label{f:SignallingThread}\usebox\myboxA}
1912\hspace{2\parindentlnth}
1913\subfloat[Waiting Thread (W1)]{\label{f:WaitingThread}\usebox\myboxB}
1914\hspace{2\parindentlnth}
1915\subfloat[Waiting Thread (W2)]{\label{f:OtherWaitingThread}\usebox\myboxC}
1916\end{cquote}
1917\caption{Barging Prevention}
1918\label{f:BargingPrevention}
1919\end{figure}
1920
1921One 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.
1922However, Figure~\ref{f:OtherWaitingThread} shows this solution is complex depending on other waiters, resulting in options when the signaller finishes the inner mutex-statement.
1923The 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@.
1924In 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.
1925Furthermore, there is an execution sequence where the signaller always finds waiter W2, and hence, waiter W1 starves.
1926
1927While a number of approaches were examined~\cite[\S~4.3]{Delisle18}, the solution chosen for \CFA is a novel techique called \newterm{partial signalling}.
1928Signalled threads are moved to the urgent queue and the waiter at the front defines the set of monitors necessary for it to unblock.
1929Partial signalling transfers ownership of monitors to the front waiter.
1930When the signaller thread exits or waits in the monitor, the front waiter is unblocked if all its monitors are released.
1931The 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.
1932
1933
1934\subsection{Loose Object Definitions}
1935\label{s:LooseObjectDefinitions}
1936
1937In an object-oriented programming-language, a class includes an exhaustive list of operations.
1938However, new members can be added via static inheritance or dynamic members, \eg JavaScript~\cite{JavaScript}.
1939Similarly, monitor routines can be added at any time in \CFA, making it less clear for programmers and more difficult to implement.
1940\begin{cfa}
1941monitor M { ... };
1942void `f`( M & mutex m );
1943void g( M & mutex m ) { waitfor( `f` ); }       $\C{// clear which f}$
1944void `f`( M & mutex m, int );                           $\C{// different f}$
1945void h( M & mutex m ) { waitfor( `f` ); }       $\C{// unclear which f}$
1946\end{cfa}
1947Hence, the cfa-code for entering a monitor looks like:
1948\begin{cfa}
1949if ( $\textrm{\textit{monitor is free}}$ ) $\LstCommentStyle{// \color{red}enter}$
1950else if ( $\textrm{\textit{already own monitor}}$ ) $\LstCommentStyle{// \color{red}continue}$
1951else if ( $\textrm{\textit{monitor accepts me}}$ ) $\LstCommentStyle{// \color{red}enter}$
1952else $\LstCommentStyle{// \color{red}block}$
1953\end{cfa}
1954For the first two conditions, it is easy to implement a check that can evaluate the condition in a few instructions.
1955However, a fast check for \emph{monitor accepts me} is much harder to implement depending on the constraints put on the monitors.
1956Figure~\ref{fig:ClassicalMonitor} shows monitors are often expressed as an entry (calling) queue, some acceptor queues, and an urgent stack/queue.
1957
1958\begin{figure}
1959\centering
1960\subfloat[Classical monitor] {
1961\label{fig:ClassicalMonitor}
1962{\resizebox{0.45\textwidth}{!}{\input{monitor.pstex_t}}}
1963}% subfloat
1964\quad
1965\subfloat[Bulk acquire monitor] {
1966\label{fig:BulkMonitor}
1967{\resizebox{0.45\textwidth}{!}{\input{ext_monitor.pstex_t}}}
1968}% subfloat
1969\caption{Monitor Implementation}
1970\label{f:MonitorImplementation}
1971\end{figure}
1972
1973For 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.
1974This approach requires a unique dense ordering of routines with a small upper-bound and the ordering must be consistent across translation units.
1975For 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.
1976
1977Figure~\ref{fig:BulkMonitor} shows the \CFA monitor implementation.
1978The mutex routine called is associated with each thread on the entry queue, while a list of acceptable routines is kept separately.
1979The 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.
1980
1981
1982\subsection{Multi-Monitor Scheduling}
1983\label{s:Multi-MonitorScheduling}
1984
1985External scheduling, like internal scheduling, becomes significantly more complex for multi-monitor semantics.
1986Even in the simplest case, new semantics needs to be established.
1987\newpage
1988\begin{cfa}
1989monitor M { ... };
1990void f( M & mutex m1 );
1991void g( M & mutex m1, M & mutex m2 ) {
1992        waitfor( f );                                                   $\C{\color{red}// pass m1 or m2 to f?}$
1993}
1994\end{cfa}
1995The solution is for the programmer to disambiguate:
1996\begin{cfa}
1997        waitfor( f, m2 );                                               $\C{\color{red}// wait for call to f with argument m2}$
1998\end{cfa}
1999Both 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@.
2000This behaviour can be extended to the multi-monitor @waitfor@ statement.
2001\begin{cfa}
2002monitor M { ... };
2003void f( M & mutex m1, M & mutex m2 );
2004void g( M & mutex m1, M & mutex m2 ) {
2005        waitfor( f, m1, m2 );                                   $\C{\color{red}// wait for call to f with arguments m1 and m2}$
2006}
2007\end{cfa}
2008Again, the set of monitors passed to the @waitfor@ statement must be entirely contained in the set of monitors already acquired by the accepting routine.
2009Also, the order of the monitors in a @waitfor@ statement is unimportant.
2010
2011Figure~\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.
2012For both examples, the set of monitors is disjoint so unblocking is impossible.
2013
2014\begin{figure}
2015\lstDeleteShortInline@%
2016\begin{tabular}{@{}l@{\hspace{\parindentlnth}}|@{\hspace{\parindentlnth}}l@{}}
2017\begin{cfa}
2018monitor M1 {} m11, m12;
2019monitor M2 {} m2;
2020condition c;
2021void f( M1 & mutex m1, M2 & mutex m2 ) {
2022        signal( c );
2023}
2024void g( M1 & mutex m1, M2 & mutex m2 ) {
2025        wait( c );
2026}
2027g( `m11`, m2 ); // block on wait
2028f( `m12`, m2 ); // cannot fulfil
2029\end{cfa}
2030&
2031\begin{cfa}
2032monitor M1 {} m11, m12;
2033monitor M2 {} m2;
2034
2035void f( M1 & mutex m1, M2 & mutex m2 ) {
2036
2037}
2038void g( M1 & mutex m1, M2 & mutex m2 ) {
2039        waitfor( f, m1, m2 );
2040}
2041g( `m11`, m2 ); // block on accept
2042f( `m12`, m2 ); // cannot fulfil
2043\end{cfa}
2044\end{tabular}
2045\lstMakeShortInline@%
2046\caption{Unmatched \protect\lstinline@mutex@ sets}
2047\label{f:UnmatchedMutexSets}
2048\end{figure}
2049
2050
2051\subsection{Extended \protect\lstinline@waitfor@}
2052
2053Figure~\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.
2054For a @waitfor@ clause to be executed, its @when@ must be true and an outstanding call to its corresponding member(s) must exist.
2055The \emph{conditional-expression} of a @when@ may call a routine, but the routine must not block or context switch.
2056If 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@.
2057If 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.
2058If 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.
2059Hence, the terminating @else@ clause allows a conditional attempt to accept a call without blocking.
2060If there is a @timeout@ clause, it provides an upper bound on waiting.
2061If both a @timeout@ clause and an @else@ clause are present, the @else@ must be conditional, or the @timeout@ is never triggered.
2062In all cases, the statement following is executed \emph{after} a clause is executed to know which of the clauses executed.
2063
2064\begin{figure}
2065\begin{cfa}
2066`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
2067        waitfor( $\emph{mutex-member-name}$ )
2068                $\emph{statement}$                                      $\C{// action after call}$
2069`or` `when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
2070        waitfor( $\emph{mutex-member-name}$ )
2071                $\emph{statement}$                                      $\C{// action after call}$
2072`or`    ...                                                                     $\C{// list of waitfor clauses}$
2073`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
2074        `timeout`                                                               $\C{// optional terminating timeout clause}$
2075                $\emph{statement}$                                      $\C{// action after timeout}$
2076`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
2077        `else`                                                                  $\C{// optional terminating clause}$
2078                $\emph{statement}$                                      $\C{// action when no immediate calls}$
2079\end{cfa}
2080\caption{Extended \protect\lstinline@waitfor@}
2081\label{f:ExtendedWaitfor}
2082\end{figure}
2083
2084Note, a group of conditional @waitfor@ clauses is \emph{not} the same as a group of @if@ statements, e.g.:
2085\begin{cfa}
2086if ( C1 ) waitfor( mem1 );                       when ( C1 ) waitfor( mem1 );
2087else if ( C2 ) waitfor( mem2 );         or when ( C2 ) waitfor( mem2 );
2088\end{cfa}
2089The left example accepts only @mem1@ if @C1@ is true or only @mem2@ if @C2@ is true.
2090The right example accepts either @mem1@ or @mem2@ if @C1@ and @C2@ are true.
2091
2092An interesting use of @waitfor@ is accepting the @mutex@ destructor to know when an object is deallocated.
2093\begin{cfa}
2094void insert( Buffer(T) & mutex buffer, T elem ) with( buffer ) {
2095        if ( count == 10 )
2096                waitfor( remove, buffer ) {
2097                        // insert elem into buffer
2098                } or `waitfor( ^?{}, buffer )` throw insertFail;
2099}
2100\end{cfa}
2101When the buffer is deallocated, the current waiter is unblocked and informed, so it can perform an appropriate action.
2102However, the basic @waitfor@ semantics do not support this functionality, since using an object after its destructor is called is undefined.
2103Therefore, 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.
2104Accepting the destructor is an idiomatic way to terminate a thread in \CFA.
2105
2106
2107\subsection{\protect\lstinline@mutex@ Threads}
2108
2109Threads in \CFA are monitors to allow direct communication among threads, \ie threads can have mutex routines that are called by other threads.
2110Hence, all monitor features are available when using threads.
2111The following shows an example of two threads directly calling each other and accepting calls from each other in a cycle.
2112\begin{cfa}
2113thread Ping {} pi;
2114thread Pong {} po;
2115void ping( Ping & mutex ) {}
2116void pong( Pong & mutex ) {}
2117int main() {}
2118\end{cfa}
2119\vspace{-0.8\baselineskip}
2120\begin{cquote}
2121\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
2122\begin{cfa}
2123void main( Ping & pi ) {
2124        for ( int i = 0; i < 10; i += 1 ) {
2125                `waitfor( ping, pi );`
2126                `pong( po );`
2127        }
2128}
2129\end{cfa}
2130&
2131\begin{cfa}
2132void main( Pong & po ) {
2133        for ( int i = 0; i < 10; i += 1 ) {
2134                `ping( pi );`
2135                `waitfor( pong, po );`
2136        }
2137}
2138\end{cfa}
2139\end{tabular}
2140\end{cquote}
2141% \lstMakeShortInline@%
2142% \caption{Threads ping/pong using external scheduling}
2143% \label{f:pingpong}
2144% \end{figure}
2145Note, the ping/pong threads are globally declared, @pi@/@po@, and hence, start (and possibly complete) before the program main starts.
2146
2147
2148\subsection{Low-level Locks}
2149
2150For 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.
2151
2152
2153\section{Parallelism}
2154
2155Historically, computer performance was about processor speeds.
2156However, with heat dissipation being a direct consequence of speed increase, parallelism is the new source for increased performance~\cite{Sutter05, Sutter05b}.
2157Now, high-performance applications must care about parallelism, which requires concurrency.
2158The lowest-level approach of parallelism is to use \newterm{kernel threads} in combination with semantics like @fork@, @join@, \etc.
2159However, kernel threads are better as an implementation tool because of complexity and higher cost.
2160Therefore, different abstractions are often layered onto kernel threads to simplify them, \eg pthreads.
2161
2162
2163\subsection{User Threads with Preemption}
2164
2165A direct improvement on kernel threads is user threads, \eg Erlang~\cite{Erlang} and \uC~\cite{uC++book}.
2166This 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.
2167In many cases, user threads can be used on a much larger scale (100,000 threads).
2168Like kernel threads, user threads support preemption, which maximizes nondeterminism, but introduces the same concurrency errors: race, livelock, starvation, and deadlock.
2169\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}.
2170
2171
2172\subsection{User Threads without Preemption (Fiber)}
2173\label{s:fibers}
2174
2175A variant of user thread is \newterm{fibers}, which removes preemption, \eg Go~\cite{Go} @goroutine@s.
2176Like functional programming, which removes mutation and its associated problems, removing preemption from concurrency reduces nondeterminism, making race and deadlock errors more difficult to generate.
2177However, preemption is necessary for concurrency that relies on spinning, so there are a class of problems that cannot be programmed without preemption.
2178
2179
2180\subsection{Thread Pools}
2181
2182In contrast to direct threading is indirect \newterm{thread pools}, where small jobs (work units) are inserted into a work pool for execution.
2183If the jobs are dependent, \ie interact, there is an implicit/explicit dependency graph that ties them together.
2184While removing direct concurrency, and hence the amount of context switching, thread pools significantly limit the interaction that can occur among jobs.
2185Indeed, jobs should not block because that also blocks the underlying thread, which effectively means the CPU utilization, and therefore throughput, suffers.
2186While 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.
2187As well, concurrency errors return, which threads pools are suppose to mitigate.
2188
2189
2190\section{\protect\CFA Runtime Structure}
2191
2192Figure~\ref{f:RunTimeStructure} illustrates the runtime structure of a \CFA program.
2193In 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.
2194An executing thread is illustrated by its containment in a processor.
2195
2196\begin{figure}
2197\centering
2198\input{RunTimeStructure}
2199\caption{\CFA Runtime Structure}
2200\label{f:RunTimeStructure}
2201\end{figure}
2202
2203
2204\subsection{Cluster}
2205\label{s:RuntimeStructureCluster}
2206
2207A \newterm{cluster} is a collection of threads and virtual processors (abstract kernel-thread) that execute the threads (like a virtual machine).
2208The purpose of a cluster is to control the amount of parallelism that is possible among threads, plus scheduling and other execution defaults.
2209The default cluster-scheduler is single-queue multi-server, which provides automatic load-balancing of threads on processors.
2210However, the scheduler is pluggable, supporting alternative schedulers.
2211If several clusters exist, both threads and virtual processors, can be explicitly migrated from one cluster to another.
2212No automatic load balancing among clusters is performed by \CFA.
2213
2214When 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.
2215The user cluster is created to contain the application user-threads.
2216Having all threads execute on the one cluster often maximizes utilization of processors, which minimizes runtime.
2217However, because of limitations of the underlying operating system, heterogeneous hardware, or scheduling requirements (real-time), multiple clusters are sometimes necessary.
2218
2219
2220\subsection{Virtual Processor}
2221\label{s:RuntimeStructureProcessor}
2222
2223A 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.
2224Programs may use more virtual processors than hardware processors.
2225On a multiprocessor, kernel threads are distributed across the hardware processors resulting in virtual processors executing in parallel.
2226(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.)
2227The \CFA runtime attempts to block unused processors and unblock processors as the system load increases;
2228balancing the workload with processors is difficult.
2229Preemption occurs on virtual processors rather than user threads, via operating-system interrupts.
2230Thus virtual processors execute user threads, where preemption frequency applies to a virtual processor, so preemption occurs randomly across the executed user threads.
2231Turning off preemption transforms user threads into fibers.
2232
2233
2234\subsection{Debug Kernel}
2235
2236There are two versions of the \CFA runtime kernel: debug and non-debug.
2237The 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.
2238After a program is debugged, the non-debugging version can be used to decrease space and increase performance.
2239
2240
2241\section{Implementation}
2242\label{s:Implementation}
2243
2244Currently, \CFA has fixed-sized stacks, where the stack size can be set at coroutine/thread creation but with no subsequent growth.
2245Schemes exist for dynamic stack-growth, such as stack copying and chained stacks.
2246However, stack copying requires pointer adjustment to items on the stack, which is impossible without some form of garbage collection.
2247As well, chained stacks require all modules be recompiled to use this feature, which breaks backward compatibility with existing C libraries.
2248In the long term, it is likely C libraries will migrate to stack chaining to support concurrency, at only a minimal cost to sequential programs.
2249Nevertheless, experience teaching \uC~\cite{CS343} shows fixed-sized stacks are rarely an issue in most concurrent programs.
2250
2251A primary implementation challenge is avoiding contention from dynamically allocating memory because of bulk acquire, \eg the internal-scheduling design is (almost) free of allocations.
2252All 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.
2253Furthermore, several bulk-acquire operations need a variable amount of memory.
2254This 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.
2255
2256In \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.
2257When a mutex call is made, pointers to the concerned monitors are aggregated into a variable-length array and sorted.
2258This array persists for the entire duration of the mutual exclusion and is used extensively for synchronization operations.
2259
2260To 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;
2261the corresponding registers are then restored for the other context.
2262Note, the instruction pointer is untouched since the context switch is always inside the same routine.
2263Unlike coroutines, threads do not context switch among each other;
2264they context switch to the cluster scheduler.
2265This method is a 2-step context-switch and provides a clear distinction between user and kernel code, where scheduling and other system operations happen.
2266The 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.
2267Experimental 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.
2268
2269All kernel threads (@pthreads@) created a stack.
2270Each \CFA virtual processor is implemented as a coroutine and these coroutines run directly on the kernel-thread stack, effectively stealing this stack.
2271The exception to this rule is the program main, \ie the initial kernel thread that is given to any program.
2272In 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.
2273
2274Finally, 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.
2275Because preemption frequency is usually long (1 millisecond) performance cost is negligible.
2276Preemption is normally handled by setting a count-down timer on each virtual processor.
2277When 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.
2278Multiple signal handlers may be pending.
2279When 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.
2280The 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;
2281therefore, the same signal mask is required for all virtual processors in a cluster.
2282
2283However, on current UNIX systems:
2284\begin{quote}
2285A process-directed signal may be delivered to any one of the threads that does not currently have the signal blocked.
2286If more than one of the threads has the signal unblocked, then the kernel chooses an arbitrary thread to which to deliver the signal.
2287SIGNAL(7) - Linux Programmer's Manual
2288\end{quote}
2289Hence, 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).
2290To 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.
2291Virtual processors register an expiration time with the discrete-event simulator, which is inserted in sorted order.
2292The 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.
2293Processing a preemption event sends an \emph{internal} @SIGUSR1@ signal to the registered virtual processor, which is always delivered to that processor.
2294
2295
2296\section{Performance}
2297\label{results}
2298
2299To verify the implementation of the \CFA runtime, a series of microbenchmarks are performed comparing \CFA with other widely used programming languages with concurrency.
2300Table~\ref{t:machine} shows the specifications of the computer used to run the benchmarks, and the versions of the software used in the comparison.
2301
2302\begin{table}
2303\centering
2304\caption{Experiment environment}
2305\label{t:machine}
2306
2307\begin{tabular}{|l|r||l|r|}
2308\hline
2309Architecture            & x86\_64                               & NUMA node(s)  & 8 \\
2310\hline
2311CPU op-mode(s)          & 32-bit, 64-bit                & Model name    & AMD Opteron\texttrademark\ Processor 6380 \\
2312\hline
2313Byte Order                      & Little Endian                 & CPU Freq              & 2.5 GHz \\
2314\hline
2315CPU(s)                          & 64                                    & L1d cache     & 16 KiB \\
2316\hline
2317Thread(s) per core      & 2                                     & L1i cache     & 64 KiB \\
2318\hline
2319Core(s) per socket      & 8                                     & L2 cache              & 2048 KiB \\
2320\hline
2321Socket(s)                       & 4                                     & L3 cache              & 6144 KiB \\
2322\hline
2323\hline
2324Operating system        & Ubuntu 16.04.3 LTS    & Kernel                & Linux 4.4-97-generic \\
2325\hline
2326gcc                                     & 6.3                                   & \CFA                  & 1.0.0 \\
2327\hline
2328Java                            & OpenJDK-9                     & Go                    & 1.9.2 \\
2329\hline
2330\end{tabular}
2331\end{table}
2332
2333All benchmarks are run using the following harness:
2334\begin{cfa}
2335unsigned int N = 10_000_000;
2336#define BENCH( run ) Time before = getTimeNsec(); run; Duration result = (getTimeNsec() - before) / N;
2337\end{cfa}
2338The method used to get time is @clock_gettime( CLOCK_REALTIME )@.
2339Each benchmark is performed @N@ times, where @N@ varies depending on the benchmark;
2340the total time is divided by @N@ to obtain the average time for a benchmark.
2341All omitted tests for other languages are functionally identical to the shown \CFA test.
2342
2343
2344\paragraph{Context-Switching}
2345
2346In procedural programming, the cost of a routine call is important as modularization (refactoring) increases.
2347(In many cases, a compiler inlines routine calls to eliminate this cost.)
2348Similarly, when modularization extends to coroutines/tasks, the time for a context switch becomes a relevant factor.
2349The coroutine context-switch is 2-step using resume/suspend, \ie from resumer to suspender and from suspender to resumer.
2350The thread context switch is 2-step using yield, \ie enter and return from the runtime kernel.
2351Figure~\ref{f:ctx-switch} shows the code for coroutines/threads with all results in Table~\ref{tab:ctx-switch}.
2352The difference in performance between coroutine and thread context-switch is the cost of scheduling for threads, whereas coroutines are self-scheduling.
2353
2354\begin{figure}
2355\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2356
2357\newbox\myboxA
2358\begin{lrbox}{\myboxA}
2359\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2360coroutine C {} c;
2361void main( C & ) { for ( ;; ) { @suspend();@ } }
2362int main() {
2363        BENCH(
2364                for ( size_t i = 0; i < N; i += 1 ) { @resume( c );@ } )
2365        sout | result`ns;
2366}
2367\end{cfa}
2368\end{lrbox}
2369
2370\newbox\myboxB
2371\begin{lrbox}{\myboxB}
2372\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2373
2374
2375int main() {
2376        BENCH(
2377                for ( size_t i = 0; i < N; i += 1 ) { @yield();@ } )
2378        sout | result`ns;
2379}
2380\end{cfa}
2381\end{lrbox}
2382
2383\subfloat[Coroutine]{\usebox\myboxA}
2384\quad
2385\subfloat[Thread]{\usebox\myboxB}
2386\captionof{figure}{\CFA context-switch benchmark}
2387\label{f:ctx-switch}
2388
2389\centering
2390
2391\captionof{table}{Context switch comparison (nanoseconds)}
2392\label{tab:ctx-switch}
2393\bigskip
2394\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2395\cline{2-4}
2396\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2397\hline
2398Kernel Thread   & 333.5 & 332.96        & 4.1 \\
2399\CFA Coroutine  & 49            & 48.68 & 0.47    \\
2400\CFA Thread             & 105           & 105.57        & 1.37 \\
2401\uC Coroutine   & 44            & 44            & 0 \\
2402\uC Thread              & 100           & 99.29 & 0.96 \\
2403Goroutine               & 145           & 147.25        & 4.15 \\
2404Java Thread             & 373.5 & 375.14        & 8.72 \\
2405\hline
2406\end{tabular}
2407
2408\bigskip
2409\bigskip
2410
2411\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2412\begin{cfa}
2413monitor M { ... } m1/*, m2, m3, m4*/;
2414void __attribute__((noinline)) do_call( M & mutex m/*, m2, m3, m4*/ ) {}
2415int main() {
2416        BENCH( for( size_t i = 0; i < N; i += 1 ) { @do_call( m1/*, m2, m3, m4*/ );@ } )
2417        sout | result`ns;
2418}
2419\end{cfa}
2420\captionof{figure}{\CFA acquire/release mutex benchmark}
2421\label{f:mutex}
2422
2423\centering
2424
2425\captionof{table}{Mutex comparison (nanoseconds)}
2426\label{tab:mutex}
2427\bigskip
2428
2429\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2430\cline{2-4}
2431\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2432\hline
2433C routine                                       & 2             & 2             & 0    \\
2434FetchAdd + FetchSub                     & 26            & 26            & 0    \\
2435Pthreads Mutex Lock                     & 31            & 31.71 & 0.97 \\
2436\uC @monitor@ member routine            & 31            & 31            & 0    \\
2437\CFA @mutex@ routine, 1 argument        & 46            & 46.68 & 0.93  \\
2438\CFA @mutex@ routine, 2 argument        & 84            & 85.36 & 1.99 \\
2439\CFA @mutex@ routine, 4 argument        & 158           & 161           & 4.22 \\
2440Java synchronized routine               & 27.5  & 29.79 & 2.93  \\
2441\hline
2442\end{tabular}
2443\end{figure}
2444
2445
2446\paragraph{Mutual-Exclusion}
2447
2448Mutual exclusion is measured by entering/leaving a critical section.
2449For monitors, entering and leaving a monitor routine is measured.
2450Figure~\ref{f:mutex} shows the code for \CFA with all results in Table~\ref{tab:mutex}.
2451To 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.
2452Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2453
2454
2455\paragraph{Internal Scheduling}
2456
2457Internal scheduling is measured by waiting on and signalling a condition variable.
2458Figure~\ref{f:int-sched} shows the code for \CFA, with results in Table~\ref{tab:int-sched}.
2459Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2460Java 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.
2461
2462\begin{figure}
2463\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2464\begin{cfa}
2465volatile int go = 0;
2466condition c;
2467monitor M { ... } m;
2468void __attribute__((noinline)) do_call( M & mutex a1 ) { signal( c ); }
2469thread T {};
2470void main( T & this ) {
2471        while ( go == 0 ) { yield(); }  // wait for other thread to start
2472        while ( go == 1 ) { @do_call( m );@ }
2473}
2474int  __attribute__((noinline)) do_wait( M & mutex m ) {
2475        go = 1; // continue other thread
2476        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @wait( c );@ } );
2477        go = 0; // stop other thread
2478        sout | result`ns;
2479}
2480int main() {
2481        T t;
2482        do_wait( m );
2483}
2484\end{cfa}
2485\captionof{figure}{\CFA Internal-scheduling benchmark}
2486\label{f:int-sched}
2487
2488\centering
2489\captionof{table}{Internal-scheduling comparison (nanoseconds)}
2490\label{tab:int-sched}
2491\bigskip
2492
2493\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2494\cline{2-4}
2495\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2496\hline
2497Pthreads Condition Variable             & 6005  & 5681.43       & 835.45 \\
2498\uC @signal@                                    & 324           & 325.54        & 3,02   \\
2499\CFA @signal@, 1 @monitor@              & 368.5         & 370.61        & 4.77   \\
2500\CFA @signal@, 2 @monitor@              & 467           & 470.5 & 6.79   \\
2501\CFA @signal@, 4 @monitor@              & 700.5         & 702.46        & 7.23  \\
2502Java @notify@                                   & 15471 & 172511        & 5689 \\
2503\hline
2504\end{tabular}
2505\end{figure}
2506
2507
2508\paragraph{External Scheduling}
2509
2510External scheduling is measured by accepting a call using the @waitfor@ statement (@_Accept@ in \uC).
2511Figure~\ref{f:ext-sched} shows the code for \CFA, with results in Table~\ref{tab:ext-sched}.
2512Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2513
2514\begin{figure}
2515\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2516\begin{cfa}
2517volatile int go = 0;
2518monitor M { ... } m;
2519thread T {};
2520void __attribute__((noinline)) do_call( M & mutex ) {}
2521void main( T & ) {
2522        while ( go == 0 ) { yield(); }  // wait for other thread to start
2523        while ( go == 1 ) { @do_call( m );@ }
2524}
2525int __attribute__((noinline)) do_wait( M & mutex m ) {
2526        go = 1; // continue other thread
2527        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @waitfor( do_call, m );@ } )
2528        go = 0; // stop other thread
2529        sout | result`ns;
2530}
2531int main() {
2532        T t;
2533        do_wait( m );
2534}
2535\end{cfa}
2536\captionof{figure}{\CFA external-scheduling benchmark}
2537\label{f:ext-sched}
2538
2539\centering
2540
2541\captionof{table}{External-scheduling comparison (nanoseconds)}
2542\label{tab:ext-sched}
2543\bigskip
2544\begin{tabular}{|r|*{3}{D{.}{.}{3.2}|}}
2545\cline{2-4}
2546\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} &\multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2547\hline
2548\uC @_Accept@                           & 358           & 359.11        & 2.53  \\
2549\CFA @waitfor@, 1 @monitor@     & 359           & 360.93        & 4.07  \\
2550\CFA @waitfor@, 2 @monitor@     & 450           & 449.39        & 6.62  \\
2551\CFA @waitfor@, 4 @monitor@     & 652           & 655.64        & 7.73 \\
2552\hline
2553\end{tabular}
2554
2555\bigskip
2556\medskip
2557
2558\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2559\begin{cfa}
2560thread MyThread {};
2561void main( MyThread & ) {}
2562int main() {
2563        BENCH( for ( size_t i = 0; i < N; i += 1 ) { @MyThread m;@ } )
2564        sout | result`ns;
2565}
2566\end{cfa}
2567\captionof{figure}{\CFA object-creation benchmark}
2568\label{f:creation}
2569
2570\centering
2571
2572\captionof{table}{Creation comparison (nanoseconds)}
2573\label{tab:creation}
2574\bigskip
2575
2576\begin{tabular}{|r|*{3}{D{.}{.}{5.2}|}}
2577\cline{2-4}
2578\multicolumn{1}{c|}{} & \multicolumn{1}{c|}{Median} & \multicolumn{1}{c|}{Average} & \multicolumn{1}{c|}{Std Dev} \\
2579\hline
2580Pthreads                                & 28091         & 28073.39      & 163.1  \\
2581\CFA Coroutine Lazy             & 6                     & 6.07          & 0.26   \\
2582\CFA Coroutine Eager    & 520           & 520.61        & 2.04   \\
2583\CFA Thread                             & 2032  & 2016.29       & 112.07  \\
2584\uC Coroutine                   & 106           & 107.36        & 1.47   \\
2585\uC Thread                              & 536.5 & 537.07        & 4.64   \\
2586Goroutine                               & 3103  & 3086.29       & 90.25  \\
2587Java Thread                             & 103416.5      & 103732.29     & 1137 \\
2588\hline
2589\end{tabular}
2590\end{figure}
2591
2592
2593\paragraph{Object Creation}
2594
2595Object creation is measured by creating/deleting the specific kind of concurrent object.
2596Figure~\ref{f:creation} shows the code for \CFA, with results in Table~\ref{tab:creation}.
2597The 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.
2598
2599
2600\section{Conclusion}
2601
2602This paper demonstrates a concurrency API that is simple, efficient, and able to build higher-level concurrency features.
2603The 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.
2604The M:N model is judged to be efficient and provide greater flexibility than a 1:1 threading model.
2605High-level objects (monitor/task) are the core mechanism for mutual exclusion and synchronization.
2606A novel aspect is allowing multiple mutex-objects to be accessed simultaneously reducing the potential for deadlock for this complex scenario.
2607These concepts and the entire \CFA runtime-system are written in the \CFA language, demonstrating the expressiveness of the \CFA language.
2608Performance 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.
2609C 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.
2610
2611
2612\section{Future Work}
2613
2614While concurrency in \CFA has a strong start, development is still underway and there are missing features.
2615
2616\paragraph{Flexible Scheduling}
2617\label{futur:sched}
2618
2619An important part of concurrency is scheduling.
2620Different scheduling algorithms can affect performance (both in terms of average and variation).
2621However, no single scheduler is optimal for all workloads and therefore there is value in being able to change the scheduler for given programs.
2622One solution is to offer various tweaking options, allowing the scheduler to be adjusted to the requirements of the workload.
2623However, to be truly flexible, a pluggable scheduler is necessary.
2624Currently, 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.
2625
2626\paragraph{Non-Blocking I/O}
2627\label{futur:nbio}
2628
2629Many modern workloads are not bound by computation but IO operations, a common case being web servers and XaaS~\cite{XaaS} (anything as a service).
2630These types of workloads require significant engineering to amortizing costs of blocking IO-operations.
2631At 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.
2632Current 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.
2633However, these solutions lead to code that is hard to create, read and maintain.
2634A 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.
2635A non-blocking I/O library is currently under development for \CFA.
2636
2637\paragraph{Other Concurrency Tools}
2638\label{futur:tools}
2639
2640While monitors offer flexible and powerful concurrency for \CFA, other concurrency tools are also necessary for a complete multi-paradigm concurrency package.
2641Examples of such tools can include futures and promises~\cite{promises}, executors and actors.
2642These additional features are useful when monitors offer a level of abstraction that is inadequate for certain tasks.
2643As well, new \CFA extensions should make it possible to create a uniform interface for virtually all mutual exclusion, including monitors and low-level locks.
2644
2645\paragraph{Implicit Threading}
2646\label{futur:implcit}
2647
2648Basic 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.
2649This type of concurrency can be achieved both at the language level and at the library level.
2650The canonical example of implicit concurrency is concurrent nested @for@ loops, which are amenable to divide and conquer algorithms~\cite{uC++book}.
2651The \CFA language features should make it possible to develop a reasonable number of implicit concurrency mechanism to solve basic HPC data-concurrency problems.
2652However, implicit concurrency is a restrictive solution with significant limitations, so it can never replace explicit concurrent programming.
2653
2654
2655\section{Acknowledgements}
2656
2657The authors would like to recognize the design assistance of Aaron Moss, Rob Schluntz and Andrew Beach on the features described in this paper.
2658Funding 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.
2659
2660{%
2661\fontsize{9bp}{12bp}\selectfont%
2662\bibliography{pl,local}
2663}%
2664
2665\end{document}
2666
2667% Local Variables: %
2668% tab-width: 4 %
2669% fill-column: 120 %
2670% compile-command: "make" %
2671% End: %
Note: See TracBrowser for help on using the repository browser.