source: doc/papers/concurrency/Paper.tex @ 5453237

arm-ehjacob/cs343-translationjenkins-sandboxnew-astnew-ast-unique-expr
Last change on this file since 5453237 was 5453237, checked in by Peter A. Buhr <pabuhr@…>, 2 years ago

make Dave Dice changes to concurrency paper

  • Property mode set to 100644
File size: 158.1 KB
Line 
1\documentclass[AMA,STIX1COL]{WileyNJD-v2}
2
3\articletype{RESEARCH ARTICLE}%
4
5% Referees
6% Doug Lea, dl@cs.oswego.edu, SUNY Oswego
7% Herb Sutter, hsutter@microsoft.com, Microsoft Corp
8% Gor Nishanov, gorn@microsoft.com, Microsoft Corp
9% James Noble, kjx@ecs.vuw.ac.nz, Victoria University of Wellington, School of Engineering and Computer Science
10
11\received{XXXXX}
12\revised{XXXXX}
13\accepted{XXXXX}
14
15\raggedbottom
16
17%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
18
19% Latex packages used in the document.
20
21\usepackage{epic,eepic}
22\usepackage{xspace}
23\usepackage{enumitem}
24\usepackage{comment}
25\usepackage{upquote}                                            % switch curled `'" to straight
26\usepackage{listings}                                           % format program code
27\usepackage[labelformat=simple,aboveskip=0pt,farskip=0pt]{subfig}
28\renewcommand{\thesubfigure}{(\Alph{subfigure})}
29\captionsetup{justification=raggedright,singlelinecheck=false}
30\usepackage{dcolumn}                                            % align decimal points in tables
31\usepackage{capt-of}
32\setlength{\multicolsep}{6.0pt plus 2.0pt minus 1.5pt}
33
34\hypersetup{breaklinks=true}
35\definecolor{OliveGreen}{cmyk}{0.64 0 0.95 0.40}
36\definecolor{Mahogany}{cmyk}{0 0.85 0.87 0.35}
37\definecolor{Plum}{cmyk}{0.50 1 0 0}
38
39\usepackage[pagewise]{lineno}
40\renewcommand{\linenumberfont}{\scriptsize\sffamily}
41
42\renewcommand{\topfraction}{0.8}                        % float must be greater than X of the page before it is forced onto its own page
43\renewcommand{\bottomfraction}{0.8}                     % float must be greater than X of the page before it is forced onto its own page
44\renewcommand{\floatpagefraction}{0.8}          % float must be greater than X of the page before it is forced onto its own page
45\renewcommand{\textfraction}{0.0}                       % the entire page maybe devoted to floats with no text on the page at all
46
47\lefthyphenmin=3                                                        % hyphen only after 4 characters
48\righthyphenmin=3
49
50%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
51
52% Names used in the document.
53
54\newcommand{\CFAIcon}{\textsf{C}\raisebox{\depth}{\rotatebox{180}{\textsf{A}}}\xspace} % Cforall symbolic name
55\newcommand{\CFA}{\protect\CFAIcon}             % safe for section/caption
56\newcommand{\CFL}{\textrm{Cforall}\xspace}      % Cforall symbolic name
57\newcommand{\Celeven}{\textrm{C11}\xspace}      % C11 symbolic name
58\newcommand{\CC}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}\xspace} % C++ symbolic name
59\newcommand{\CCeleven}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}11\xspace} % C++11 symbolic name
60\newcommand{\CCfourteen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}14\xspace} % C++14 symbolic name
61\newcommand{\CCseventeen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}17\xspace} % C++17 symbolic name
62\newcommand{\CCtwenty}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}20\xspace} % C++20 symbolic name
63\newcommand{\Csharp}{C\raisebox{-0.7ex}{\Large$^\sharp$}\xspace} % C# symbolic name
64
65%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
66
67\newcommand{\Textbf}[2][red]{{\color{#1}{\textbf{#2}}}}
68\newcommand{\Emph}[2][red]{{\color{#1}\textbf{\emph{#2}}}}
69\newcommand{\uC}{$\mu$\CC}
70\newcommand{\TODO}[1]{{\Textbf{#1}}}
71
72%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
73
74% Default underscore is too low and wide. Cannot use lstlisting "literate" as replacing underscore
75% removes it as a variable-name character so keywords in variables are highlighted. MUST APPEAR
76% AFTER HYPERREF.
77%\DeclareTextCommandDefault{\textunderscore}{\leavevmode\makebox[1.2ex][c]{\rule{1ex}{0.1ex}}}
78\renewcommand{\textunderscore}{\leavevmode\makebox[1.2ex][c]{\rule{1ex}{0.075ex}}}
79
80\renewcommand*{\thefootnote}{\Alph{footnote}} % hack because fnsymbol does not work
81%\renewcommand*{\thefootnote}{\fnsymbol{footnote}}
82
83\makeatletter
84% parindent is relative, i.e., toggled on/off in environments like itemize, so store the value for
85% use rather than use \parident directly.
86\newlength{\parindentlnth}
87\setlength{\parindentlnth}{\parindent}
88
89\newcommand{\LstBasicStyle}[1]{{\lst@basicstyle{\lst@basicstyle{#1}}}}
90\newcommand{\LstKeywordStyle}[1]{{\lst@basicstyle{\lst@keywordstyle{#1}}}}
91\newcommand{\LstCommentStyle}[1]{{\lst@basicstyle{\lst@commentstyle{#1}}}}
92
93\newlength{\gcolumnposn}                                        % temporary hack because lstlisting does not handle tabs correctly
94\newlength{\columnposn}
95\setlength{\gcolumnposn}{3.5in}
96\setlength{\columnposn}{\gcolumnposn}
97
98\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}}}}
99\newcommand{\CRT}{\global\columnposn=\gcolumnposn}
100
101% Denote newterms in particular font and index them without particular font and in lowercase, e.g., \newterm{abc}.
102% The option parameter provides an index term different from the new term, e.g., \newterm[\texttt{abc}]{abc}
103% The star version does not lowercase the index information, e.g., \newterm*{IBM}.
104\newcommand{\newtermFontInline}{\emph}
105\newcommand{\newterm}{\@ifstar\@snewterm\@newterm}
106\newcommand{\@newterm}[2][\@empty]{\lowercase{\def\temp{#2}}{\newtermFontInline{#2}}\ifx#1\@empty\index{\temp}\else\index{#1@{\protect#2}}\fi}
107\newcommand{\@snewterm}[2][\@empty]{{\newtermFontInline{#2}}\ifx#1\@empty\index{#2}\else\index{#1@{\protect#2}}\fi}
108
109% Latin abbreviation
110\newcommand{\abbrevFont}{\textit}                       % set empty for no italics
111\@ifundefined{eg}{
112\newcommand{\EG}{\abbrevFont{e}\abbrevFont{g}}
113\newcommand*{\eg}{%
114        \@ifnextchar{,}{\EG}%
115                {\@ifnextchar{:}{\EG}%
116                        {\EG,\xspace}}%
117}}{}%
118\@ifundefined{ie}{
119\newcommand{\IE}{\abbrevFont{i}\abbrevFont{e}}
120\newcommand*{\ie}{%
121        \@ifnextchar{,}{\IE}%
122                {\@ifnextchar{:}{\IE}%
123                        {\IE,\xspace}}%
124}}{}%
125\@ifundefined{etc}{
126\newcommand{\ETC}{\abbrevFont{etc}}
127\newcommand*{\etc}{%
128        \@ifnextchar{.}{\ETC}%
129        {\ETC.\xspace}%
130}}{}%
131\@ifundefined{etal}{
132\newcommand{\ETAL}{\abbrevFont{et}~\abbrevFont{al}}
133\newcommand*{\etal}{%
134        \@ifnextchar{.}{\protect\ETAL}%
135                {\protect\ETAL.\xspace}%
136}}{}%
137\@ifundefined{viz}{
138\newcommand{\VIZ}{\abbrevFont{viz}}
139\newcommand*{\viz}{%
140        \@ifnextchar{.}{\VIZ}%
141                {\VIZ.\xspace}%
142}}{}%
143\makeatother
144
145\newenvironment{cquote}
146               {\list{}{\lstset{resetmargins=true,aboveskip=0pt,belowskip=0pt}\topsep=3pt\parsep=0pt\leftmargin=\parindentlnth\rightmargin\leftmargin}%
147                \item\relax}
148               {\endlist}
149
150%\newenvironment{cquote}{%
151%\list{}{\lstset{resetmargins=true,aboveskip=0pt,belowskip=0pt}\topsep=3pt\parsep=0pt\leftmargin=\parindentlnth\rightmargin\leftmargin}%
152%\item\relax%
153%}{%
154%\endlist%
155%}% cquote
156
157% CFA programming language, based on ANSI C (with some gcc additions)
158\lstdefinelanguage{CFA}[ANSI]{C}{
159        morekeywords={
160                _Alignas, _Alignof, __alignof, __alignof__, asm, __asm, __asm__, __attribute, __attribute__,
161                auto, _Bool, catch, catchResume, choose, _Complex, __complex, __complex__, __const, __const__,
162                coroutine, disable, dtype, enable, exception, __extension__, fallthrough, fallthru, finally,
163                __float80, float80, __float128, float128, forall, ftype, generator, _Generic, _Imaginary, __imag, __imag__,
164                inline, __inline, __inline__, __int128, int128, __label__, monitor, mutex, _Noreturn, one_t, or,
165                otype, restrict, __restrict, __restrict__, __signed, __signed__, _Static_assert, thread,
166                _Thread_local, throw, throwResume, timeout, trait, try, ttype, typeof, __typeof, __typeof__,
167                virtual, __volatile, __volatile__, waitfor, when, with, zero_t},
168        moredirectives={defined,include_next},
169        % replace/adjust listing characters that look bad in sanserif
170        literate={-}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.1ex}}}}1 {^}{\raisebox{0.6ex}{$\scriptstyle\land\,$}}1
171                {~}{\raisebox{0.3ex}{$\scriptstyle\sim\,$}}1 % {`}{\ttfamily\upshape\hspace*{-0.1ex}`}1
172                {<}{\textrm{\textless}}1 {>}{\textrm{\textgreater}}1
173                {<-}{$\leftarrow$}2 {=>}{$\Rightarrow$}2 {->}{\makebox[1ex][c]{\raisebox{0.5ex}{\rule{0.8ex}{0.075ex}}}\kern-0.2ex{\textrm{\textgreater}}}2,
174}
175
176\lstset{
177language=CFA,
178columns=fullflexible,
179basicstyle=\linespread{0.9}\sf,                                                 % reduce line spacing and use sanserif font
180stringstyle=\tt,                                                                                % use typewriter font
181tabsize=5,                                                                                              % N space tabbing
182xleftmargin=\parindentlnth,                                                             % indent code to paragraph indentation
183%mathescape=true,                                                                               % LaTeX math escape in CFA code $...$
184escapechar=\$,                                                                                  % LaTeX escape in CFA code
185keepspaces=true,                                                                                %
186showstringspaces=false,                                                                 % do not show spaces with cup
187showlines=true,                                                                                 % show blank lines at end of code
188aboveskip=4pt,                                                                                  % spacing above/below code block
189belowskip=3pt,
190moredelim=**[is][\color{red}]{`}{`},
191}% lstset
192
193% uC++ programming language, based on ANSI C++
194\lstdefinelanguage{uC++}[ANSI]{C++}{
195        morekeywords={
196                _Accept, _AcceptReturn, _AcceptWait, _Actor, _At, _CatchResume, _Cormonitor, _Coroutine, _Disable,
197                _Else, _Enable, _Event, _Finally, _Monitor, _Mutex, _Nomutex, _PeriodicTask, _RealTimeTask,
198                _Resume, _Select, _SporadicTask, _Task, _Timeout, _When, _With, _Throw},
199}
200\lstdefinelanguage{Golang}{
201        morekeywords=[1]{package,import,func,type,struct,return,defer,panic,recover,select,var,const,iota,},
202        morekeywords=[2]{string,uint,uint8,uint16,uint32,uint64,int,int8,int16,int32,int64,
203                bool,float32,float64,complex64,complex128,byte,rune,uintptr, error,interface},
204        morekeywords=[3]{map,slice,make,new,nil,len,cap,copy,close,true,false,delete,append,real,imag,complex,chan,},
205        morekeywords=[4]{for,break,continue,range,goto,switch,case,fallthrough,if,else,default,},
206        morekeywords=[5]{Println,Printf,Error,},
207        sensitive=true,
208        morecomment=[l]{//},
209        morecomment=[s]{/*}{*/},
210        morestring=[b]',
211        morestring=[b]",
212        morestring=[s]{`}{`},
213}
214
215% Go programming language: https://github.com/julienc91/listings-golang/blob/master/listings-golang.sty
216\lstdefinelanguage{Golang}{
217        morekeywords=[1]{package,import,func,type,struct,return,defer,panic,recover,select,var,const,iota,},
218        morekeywords=[2]{string,uint,uint8,uint16,uint32,uint64,int,int8,int16,int32,int64,
219                bool,float32,float64,complex64,complex128,byte,rune,uintptr, error,interface},
220        morekeywords=[3]{map,slice,make,new,nil,len,cap,copy,close,true,false,delete,append,real,imag,complex,chan,},
221        morekeywords=[4]{for,break,continue,range,goto,switch,case,fallthrough,if,else,default,},
222        morekeywords=[5]{Println,Printf,Error,},
223        sensitive=true,
224        morecomment=[l]{//},
225        morecomment=[s]{/*}{*/},
226        morestring=[b]',
227        morestring=[b]",
228        morestring=[s]{`}{`},
229        % replace/adjust listing characters that look bad in sanserif
230        literate={-}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.1ex}}}}1 {^}{\raisebox{0.6ex}{$\scriptstyle\land\,$}}1
231                {~}{\raisebox{0.3ex}{$\scriptstyle\sim\,$}}1 % {`}{\ttfamily\upshape\hspace*{-0.1ex}`}1
232                {<}{\textrm{\textless}}1 {>}{\textrm{\textgreater}}1
233                {<-}{\makebox[2ex][c]{\textrm{\textless}\raisebox{0.5ex}{\rule{0.8ex}{0.075ex}}}}2,
234}
235
236\lstnewenvironment{cfa}[1][]
237{\lstset{#1}}
238{}
239\lstnewenvironment{C++}[1][]                            % use C++ style
240{\lstset{language=C++,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
241{}
242\lstnewenvironment{uC++}[1][]
243{\lstset{#1}}
244{}
245\lstnewenvironment{Go}[1][]
246{\lstset{language=Golang,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
247{}
248\lstnewenvironment{python}[1][]
249{\lstset{language=python,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
250{}
251
252% inline code @...@
253\lstMakeShortInline@%
254
255\let\OLDthebibliography\thebibliography
256\renewcommand\thebibliography[1]{
257  \OLDthebibliography{#1}
258  \setlength{\parskip}{0pt}
259  \setlength{\itemsep}{4pt plus 0.3ex}
260}
261
262\newbox\myboxA
263\newbox\myboxB
264\newbox\myboxC
265\newbox\myboxD
266
267\title{\texorpdfstring{Advanced Control-flow and Concurrency in \protect\CFA}{Advanced Control-flow in Cforall}}
268
269\author[1]{Thierry Delisle}
270\author[1]{Peter A. Buhr*}
271\authormark{DELISLE \textsc{et al.}}
272
273\address[1]{\orgdiv{Cheriton School of Computer Science}, \orgname{University of Waterloo}, \orgaddress{\state{Waterloo, ON}, \country{Canada}}}
274
275\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}}
276
277% \fundingInfo{Natural Sciences and Engineering Research Council of Canada}
278
279\abstract[Summary]{
280\CFA is a polymorphic, non-object-oriented, concurrent, backwards-compatible extension of the C programming language.
281This paper discusses the design philosophy and implementation of its advanced control-flow and concurrent/parallel features, along with the supporting runtime written in \CFA.
282These 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 pthreads.
283\CFA introduces modern language-level control-flow mechanisms, like generators, coroutines, user-level threading, and monitors for mutual exclusion and synchronization.
284% Library extension for executors, futures, and actors are built on these basic mechanisms.
285The runtime provides significant programmer simplification and safety by eliminating spurious wakeup and monitor barging.
286The runtime also ensures multiple monitors can be safely acquired \emph{simultaneously} (deadlock free), and this feature is fully integrated with all monitor synchronization mechanisms.
287All control-flow features integrate with the \CFA polymorphic type-system and exception handling, while respecting the expectations and style of C programmers.
288Experimental results show comparable performance of the new features with similar mechanisms in other concurrent programming languages.
289}%
290
291\keywords{generator, coroutine, concurrency, parallelism, thread, monitor, runtime, C, \CFA (Cforall)}
292
293
294\begin{document}
295\linenumbers                                            % comment out to turn off line numbering
296
297\maketitle
298
299
300\section{Introduction}
301
302This paper discusses the design philosophy and implementation of advanced language-level control-flow and concurrent/parallel features in \CFA~\cite{Moss18,Cforall} and its runtime, which is written entirely in \CFA.
303\CFA is a modern, polymorphic, non-object-oriented\footnote{
304\CFA has features often associated with object-oriented programming languages, such as constructors, destructors, virtuals and simple inheritance.
305However, 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.},
306backwards-compatible extension of the C programming language.
307In many ways, \CFA is to C as Scala~\cite{Scala} is to Java, providing a \emph{research vehicle} for new typing and control-flow capabilities on top of a highly popular programming language allowing immediate dissemination.
308Within the \CFA framework, new control-flow features are created from scratch because ISO \Celeven defines only a subset of the \CFA extensions, where the overlapping features are concurrency~\cite[\S~7.26]{C11}.
309However, \Celeven concurrency is largely wrappers for a subset of the pthreads library~\cite{Butenhof97,Pthreads}, and \Celeven and pthreads concurrency is simple, based on thread fork/join in a function and mutex/condition locks, which is low-level and error-prone;
310no high-level language concurrency features are defined.
311Interestingly, almost a decade after publication of the \Celeven standard, neither gcc-8, clang-9 nor msvc-19 (most recent versions) support the \Celeven include @threads.h@, indicating little interest in the C11 concurrency approach (possibly because the effort to add concurrency to \CC).
312Finally, 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}.
313
314In contrast, there has been a renewed interest during the past decade in user-level (M:N, green) threading in old and new programming languages.
315As multi-core hardware became available in the 1980/90s, both user and kernel threading were examined.
316Kernel 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}.
317Libraries like pthreads were developed for C, and the Solaris operating-system switched from user (JDK 1.1~\cite{JDK1.1}) to kernel threads.
318As a result, languages like Java, 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.
319From 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,Marcel}, including putting green threads back into Java~\cite{Quasar}.
320The main argument for user-level threading is that it is lighter weight than kernel threading (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 medium work units to facilitate load balancing by the runtime~\cite{Verch12}.
321As well, user-threading facilitates a simpler concurrency approach using thread objects that leverage sequential patterns versus events with call-backs~\cite{Adya02,vonBehren03}.
322Finally, performant user-threading implementations (both time and space) meet or exceed direct kernel-threading implementations, while achieving the programming advantages of high concurrency levels and safety.
323
324A 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, \ie some language features are unsafe in the presence of aggressive sequential optimizations~\cite{Buhr95a,Boehm05}.
325The 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.
326One solution is low-level qualifiers and functions (\eg @volatile@ and atomics) allowing \emph{programmers} to explicitly write safe (race-free~\cite{Boehm12}) programs.
327A safer solution is high-level language constructs so the \emph{compiler} knows the optimization boundaries, and hence, provides implicit safety.
328This problem is best known with respect to concurrency, but applies to other complex control-flow, like exceptions\footnote{
329\CFA exception handling will be presented in a separate paper.
330The key feature that dovetails with this paper is nonlocal 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++}
331} and coroutines.
332Finally, language solutions allow matching constructs with language paradigm, \ie imperative and functional languages often have different presentations of the same concept to fit their programming model.
333
334Finally, it is important for a language to provide safety over performance \emph{as the default}, allowing careful reduction of safety for performance when necessary.
335Two concurrency violations of this philosophy are \emph{spurious wakeup} (random wakeup~\cite[\S~8]{Buhr05a}) and \emph{barging}\footnote{
336The notion of competitive succession instead of direct handoff, \ie a lock owner releases the lock and an arriving thread acquires it ahead of preexisting waiter threads.
337} (signals-as-hints~\cite[\S~8]{Buhr05a}), where one is a consequence of the other, \ie once there is spurious wakeup, signals-as-hints follow.
338However, spurious wakeup is \emph{not} a foundational concurrency property~\cite[\S~8]{Buhr05a}, it is a performance design choice.
339Similarly, signals-as-hints are often a performance decision.
340We argue removing spurious wakeup and signals-as-hints make concurrent programming significantly safer because it removes local non-determinism and matches with programmer expectation.
341(Author experience teaching concurrency is that students are highly confused by these semantics.)
342Clawing back performance, when local non-determinism is unimportant, should be an option not the default.
343
344\begin{comment}
345Most 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.
346As a result, there is a significant learning curve to move to these languages, and C legacy-code must be rewritten.
347While \CC, like \CFA, takes an evolutionary approach to extend C, \CC's constantly growing complex and interdependent features-set (\eg objects, inheritance, templates, etc.) mean idiomatic \CC code is difficult to use from C, and C programmers must expend significant effort learning \CC.
348Hence, rewriting and retraining costs for these languages, even \CC, are prohibitive for companies with a large C software-base.
349\CFA with its orthogonal feature-set, its high-performance runtime, and direct access to all existing C libraries circumvents these problems.
350\end{comment}
351
352\CFA embraces user-level threading, language extensions for advanced control-flow, and safety as the default.
353We present comparative examples so the reader can judge if the \CFA control-flow extensions are better and safer than those in other concurrent, imperative programming languages, and perform experiments to show the \CFA runtime is competitive with other similar mechanisms.
354The main contributions of this work are:
355\begin{itemize}[topsep=3pt,itemsep=1pt]
356\item
357language-level generators, coroutines and user-level threading, which respect the expectations of C programmers.
358\item
359monitor synchronization without barging, and the ability to safely acquiring multiple monitors \emph{simultaneously} (deadlock free), while seamlessly integrating these capabilities with all monitor synchronization mechanisms.
360\item
361providing statically type-safe interfaces that integrate with the \CFA polymorphic type-system and other language features.
362% \item
363% library extensions for executors, futures, and actors built on the basic mechanisms.
364\item
365a runtime system with no spurious wakeup.
366\item
367a dynamic partitioning mechanism to segregate the execution environment for specialized requirements.
368% \item
369% a non-blocking I/O library
370\item
371experimental results showing comparable performance of the new features with similar mechanisms in other programming languages.
372\end{itemize}
373
374Section~\ref{s:StatefulFunction} begins advanced control by introducing sequential functions that retain data and execution state between calls, which produces constructs @generator@ and @coroutine@.
375Section~\ref{s:Concurrency} begins concurrency, or how to create (fork) and destroy (join) a thread, which produces the @thread@ construct.
376Section~\ref{s:MutualExclusionSynchronization} discusses the two mechanisms to restricted nondeterminism when controlling shared access to resources (mutual exclusion) and timing relationships among threads (synchronization).
377Section~\ref{s:Monitor} shows how both mutual exclusion and synchronization are safely embedded in the @monitor@ and @thread@ constructs.
378Section~\ref{s:CFARuntimeStructure} describes the large-scale mechanism to structure (cluster) threads and virtual processors (kernel threads).
379Section~\ref{s:Performance} uses a series of microbenchmarks to compare \CFA threading with pthreads, Java OpenJDK-9, Go 1.12.6 and \uC 7.0.0.
380
381
382\section{Stateful Function}
383\label{s:StatefulFunction}
384
385The stateful function is an old idea~\cite{Conway63,Marlin80} that is new again~\cite{C++20Coroutine19}, where execution is temporarily suspended and later resumed, \eg plugin, device driver, finite-state machine.
386Hence, a stateful function may not end when it returns to its caller, allowing it to be restarted with the data and execution location present at the point of suspension.
387This capability is accomplished by retaining a data/execution \emph{closure} between invocations.
388If the closure is fixed size, we call it a \emph{generator} (or \emph{stackless}), and its control flow is restricted, \eg suspending outside the generator is prohibited.
389If the closure is variable size, we call it a \emph{coroutine} (or \emph{stackful}), and as the names implies, often implemented with a separate stack with no programming restrictions.
390Hence, refactoring a stackless coroutine may require changing it to stackful.
391A foundational property of all \emph{stateful functions} is that resume/suspend \emph{do not} cause incremental stack growth, \ie resume/suspend operations are remembered through the closure not the stack.
392As well, activating a stateful function is \emph{asymmetric} or \emph{symmetric}, identified by resume/suspend (no cycles) and resume/resume (cycles).
393A fixed closure activated by modified call/return is faster than a variable closure activated by context switching.
394Additionally, any storage management for the closure (especially in unmanaged languages, \ie no garbage collection) must also be factored into design and performance.
395Therefore, selecting between stackless and stackful semantics is a tradeoff between programming requirements and performance, where stackless is faster and stackful is more general.
396Note, creation cost is amortized across usage, so activation cost is usually the dominant factor.
397
398\begin{figure}
399\centering
400\begin{lrbox}{\myboxA}
401\begin{cfa}[aboveskip=0pt,belowskip=0pt]
402typedef struct {
403        int fn1, fn;
404} Fib;
405#define FibCtor { 1, 0 }
406int fib( Fib * f ) {
407
408
409
410        int fn = f->fn; f->fn = f->fn1;
411                f->fn1 = f->fn + fn;
412        return fn;
413
414}
415int main() {
416        Fib f1 = FibCtor, f2 = FibCtor;
417        for ( int i = 0; i < 10; i += 1 )
418                printf( "%d %d\n",
419                           fib( &f1 ), fib( &f2 ) );
420}
421\end{cfa}
422\end{lrbox}
423
424\begin{lrbox}{\myboxB}
425\begin{cfa}[aboveskip=0pt,belowskip=0pt]
426`generator` Fib {
427        int fn1, fn;
428};
429
430void `main(Fib & fib)` with(fib) {
431
432        [fn1, fn] = [1, 0];
433        for () {
434                `suspend;`
435                [fn1, fn] = [fn, fn + fn1];
436
437        }
438}
439int main() {
440        Fib f1, f2;
441        for ( 10 )
442                sout | `resume( f1 )`.fn
443                         | `resume( f2 )`.fn;
444}
445\end{cfa}
446\end{lrbox}
447
448\begin{lrbox}{\myboxC}
449\begin{cfa}[aboveskip=0pt,belowskip=0pt]
450typedef struct {
451        int fn1, fn;  void * `next`;
452} Fib;
453#define FibCtor { 1, 0, NULL }
454Fib * comain( Fib * f ) {
455        if ( f->next ) goto *f->next;
456        f->next = &&s1;
457        for ( ;; ) {
458                return f;
459          s1:; int fn = f->fn + f->fn1;
460                        f->fn1 = f->fn; f->fn = fn;
461        }
462}
463int main() {
464        Fib f1 = FibCtor, f2 = FibCtor;
465        for ( int i = 0; i < 10; i += 1 )
466                printf("%d %d\n",comain(&f1)->fn,
467                                 comain(&f2)->fn);
468}
469\end{cfa}
470\end{lrbox}
471
472\subfloat[C asymmetric generator]{\label{f:CFibonacci}\usebox\myboxA}
473\hspace{3pt}
474\vrule
475\hspace{3pt}
476\subfloat[\CFA asymmetric generator]{\label{f:CFAFibonacciGen}\usebox\myboxB}
477\hspace{3pt}
478\vrule
479\hspace{3pt}
480\subfloat[C generator implementation]{\label{f:CFibonacciSim}\usebox\myboxC}
481\caption{Fibonacci (output) asymmetric generator}
482\label{f:FibonacciAsymmetricGenerator}
483
484\bigskip
485
486\begin{lrbox}{\myboxA}
487\begin{cfa}[aboveskip=0pt,belowskip=0pt]
488`generator Fmt` {
489        char ch;
490        int g, b;
491};
492void ?{}( Fmt & fmt ) { `resume(fmt);` } // constructor
493void ^?{}( Fmt & f ) with(f) { $\C[1.75in]{// destructor}$
494        if ( g != 0 || b != 0 ) sout | nl; }
495void `main( Fmt & f )` with(f) {
496        for () { $\C{// until destructor call}$
497                for ( ; g < 5; g += 1 ) { $\C{// groups}$
498                        for ( ; b < 4; b += 1 ) { $\C{// blocks}$
499                                `suspend;` $\C{// wait for character}$
500                                while ( ch == '\n' ) `suspend;` // ignore
501                                sout | ch;                                              // newline
502                        } sout | " ";  // block spacer
503                } sout | nl; // group newline
504        }
505}
506int main() {
507        Fmt fmt; $\C{// fmt constructor called}$
508        for () {
509                sin | fmt.ch; $\C{// read into generator}$
510          if ( eof( sin ) ) break;
511                `resume( fmt );`
512        }
513
514} $\C{// fmt destructor called}\CRT$
515\end{cfa}
516\end{lrbox}
517
518\begin{lrbox}{\myboxB}
519\begin{cfa}[aboveskip=0pt,belowskip=0pt]
520typedef struct {
521        void * next;
522        char ch;
523        int g, b;
524} Fmt;
525void comain( Fmt * f ) {
526        if ( f->next ) goto *f->next;
527        f->next = &&s1;
528        for ( ;; ) {
529                for ( f->g = 0; f->g < 5; f->g += 1 ) {
530                        for ( f->b = 0; f->b < 4; f->b += 1 ) {
531                                return;
532                          s1:;  while ( f->ch == '\n' ) return;
533                                printf( "%c", f->ch );
534                        } printf( " " );
535                } printf( "\n" );
536        }
537}
538int main() {
539        Fmt fmt = { NULL };  comain( &fmt ); // prime
540        for ( ;; ) {
541                scanf( "%c", &fmt.ch );
542          if ( feof( stdin ) ) break;
543                comain( &fmt );
544        }
545        if ( fmt.g != 0 || fmt.b != 0 ) printf( "\n" );
546}
547\end{cfa}
548\end{lrbox}
549
550\subfloat[\CFA asymmetric generator]{\label{f:CFAFormatGen}\usebox\myboxA}
551\hspace{3pt}
552\vrule
553\hspace{3pt}
554\subfloat[C generator simulation]{\label{f:CFormatSim}\usebox\myboxB}
555\hspace{3pt}
556\caption{Formatter (input) asymmetric generator}
557\label{f:FormatterAsymmetricGenerator}
558\end{figure}
559
560Stateful functions appear as generators, coroutines, and threads, where presentations are based on function objects or pointers~\cite{Butenhof97, C++14, MS:VisualC++, BoostCoroutines15}.
561For example, Python presents generators as a function object:
562\begin{python}
563def Gen():
564        ... `yield val` ...
565gen = Gen()
566for i in range( 10 ):
567        print( next( gen ) )
568\end{python}
569Boost presents coroutines in terms of four functor object-types:
570\begin{cfa}
571asymmetric_coroutine<>::pull_type
572asymmetric_coroutine<>::push_type
573symmetric_coroutine<>::call_type
574symmetric_coroutine<>::yield_type
575\end{cfa}
576and many languages present threading using function pointers, @pthreads@~\cite{Butenhof97}, \Csharp~\cite{Csharp}, Go~\cite{Go}, and Scala~\cite{Scala}, \eg pthreads:
577\begin{cfa}
578void * rtn( void * arg ) { ... }
579int i = 3, rc;
580pthread_t t; $\C{// thread id}$
581`rc = pthread_create( &t, rtn, (void *)i );` $\C{// create and initialized task, type-unsafe input parameter}$
582\end{cfa}
583% void mycor( pthread_t cid, void * arg ) {
584%       int * value = (int *)arg;                               $\C{// type unsafe, pointer-size only}$
585%       // thread body
586% }
587% int main() {
588%       int input = 0, output;
589%       coroutine_t cid = coroutine_create( &mycor, (void *)&input ); $\C{// type unsafe, pointer-size only}$
590%       coroutine_resume( cid, (void *)input, (void **)&output ); $\C{// type unsafe, pointer-size only}$
591% }
592\CFA's preferred presentation model for generators/coroutines/threads is a hybrid of objects and functions, with an object-oriented flavour.
593Essentially, the generator/coroutine/thread function is semantically coupled with a generator/coroutine/thread custom type.
594The custom type solves several issues, while accessing the underlying mechanisms used by the custom types is still allowed.
595
596
597\subsection{Generator}
598
599Stackless generators have the potential to be very small and fast, \ie as small and fast as function call/return for both creation and execution.
600The \CFA goal is to achieve this performance target, possibly at the cost of some semantic complexity.
601A series of different kinds of generators and their implementation demonstrate how this goal is accomplished.
602
603Figure~\ref{f:FibonacciAsymmetricGenerator} shows an unbounded asymmetric generator for an infinite sequence of Fibonacci numbers written in C and \CFA, with a simple C implementation for the \CFA version.
604This generator is an \emph{output generator}, producing a new result on each resumption.
605To compute Fibonacci, the previous two values in the sequence are retained to generate the next value, \ie @fn1@ and @fn@, plus the execution location where control restarts when the generator is resumed, \ie top or middle.
606An additional requirement is the ability to create an arbitrary number of generators (of any kind), \ie retaining one state in global variables is insufficient;
607hence, state is retained in a closure between calls.
608Figure~\ref{f:CFibonacci} shows the C approach of manually creating the closure in structure @Fib@, and multiple instances of this closure provide multiple Fibonacci generators.
609The C version only has the middle execution state because the top execution state is declaration initialization.
610Figure~\ref{f:CFAFibonacciGen} shows the \CFA approach, which also has a manual closure, but replaces the structure with a custom \CFA @generator@ type.
611This generator type is then connected to a function that \emph{must be named \lstinline|main|},\footnote{
612The name \lstinline|main| has special meaning in C, specifically the function where a program starts execution.
613Hence, overloading this name for other starting points (generator/coroutine/thread) is a logical extension.}
614called a \emph{generator main},which takes as its only parameter a reference to the generator type.
615The generator main contains @suspend@ statements that suspend execution without ending the generator versus @return@.
616For the Fibonacci generator-main,\footnote{
617The \CFA \lstinline|with| opens an aggregate scope making its fields directly accessible, like Pascal \lstinline|with|, but using parallel semantics.
618Multiple aggregates may be opened.}
619the top initialization state appears at the start and the middle execution state is denoted by statement @suspend@.
620Any local variables in @main@ \emph{are not retained} between calls;
621hence local variables are only for temporary computations \emph{between} suspends.
622All retained state \emph{must} appear in the generator's type.
623As well, generator code containing a @suspend@ cannot be refactored into a helper function called by the generator, because @suspend@ is implemented via @return@, so a return from the helper function goes back to the current generator not the resumer.
624The generator is started by calling function @resume@ with a generator instance, which begins execution at the top of the generator main, and subsequent @resume@ calls restart the generator at its point of last suspension.
625Resuming an ended (returned) generator is undefined.
626Function @resume@ returns its argument generator so it can be cascaded in an expression, in this case to print the next Fibonacci value @fn@ computed in the generator instance.
627Figure~\ref{f:CFibonacciSim} shows the C implementation of the \CFA generator only needs one additional field, @next@, to handle retention of execution state.
628The computed @goto@ at the start of the generator main, which branches after the previous suspend, adds very little cost to the resume call.
629Finally, an explicit generator type provides both design and performance benefits, such as multiple type-safe interface functions taking and returning arbitrary types.\footnote{
630The \CFA operator syntax uses \lstinline|?| to denote operands, which allows precise definitions for pre, post, and infix operators, \eg \lstinline|++?|, \lstinline|?++|, and \lstinline|?+?|, in addition \lstinline|?\{\}| denotes a constructor, as in \lstinline|foo `f` = `\{`...`\}`|, \lstinline|^?\{\}| denotes a destructor, and \lstinline|?()| is \CC function call \lstinline|operator()|.
631}%
632\begin{cfa}
633int ?()( Fib & fib ) { return `resume( fib )`.fn; } $\C[3.9in]{// function-call interface}$
634int ?()( Fib & fib, int N ) { for ( N - 1 ) `fib()`; return `fib()`; } $\C{// use function-call interface to skip N values}$
635double ?()( Fib & fib ) { return (int)`fib()` / 3.14159; } $\C{// different return type, cast prevents recursive call}\CRT$
636sout | (int)f1() | (double)f1() | f2( 2 ); // alternative interface, cast selects call based on return type, step 2 values
637\end{cfa}
638Now, the generator can be a separately compiled opaque-type only accessed through its interface functions.
639For contrast, Figure~\ref{f:PythonFibonacci} shows the equivalent Python Fibonacci generator, which does not use a generator type, and hence only has a single interface, but an implicit closure.
640
641Having to manually create the generator closure by moving local-state variables into the generator type is an additional programmer burden.
642(This restriction is removed by the coroutine in Section~\ref{s:Coroutine}.)
643This requirement follows from the generality of variable-size local-state, \eg local state with a variable-length array requires dynamic allocation because the array size is unknown at compile time.
644However, dynamic allocation significantly increases the cost of generator creation/destruction and is a showstopper for embedded real-time programming.
645But more importantly, the size of the generator type is tied to the local state in the generator main, which precludes separate compilation of the generator main, \ie a generator must be inlined or local state must be dynamically allocated.
646With respect to safety, we believe static analysis can discriminate local state from temporary variables in a generator, \ie variable usage spanning @suspend@, and generate a compile-time error.
647Finally, our current experience is that most generator problems have simple data state, including local state, but complex execution state, so the burden of creating the generator type is small.
648As well, C programmers are not afraid of this kind of semantic programming requirement, if it results in very small, fast generators.
649
650Figure~\ref{f:CFAFormatGen} shows an asymmetric \newterm{input generator}, @Fmt@, for restructuring text into groups of characters of fixed-size blocks, \ie the input on the left is reformatted into the output on the right, where newlines are ignored.
651\begin{center}
652\tt
653\begin{tabular}{@{}l|l@{}}
654\multicolumn{1}{c|}{\textbf{\textrm{input}}} & \multicolumn{1}{c}{\textbf{\textrm{output}}} \\
655\begin{tabular}[t]{@{}ll@{}}
656abcdefghijklmnopqrstuvwxyz \\
657abcdefghijklmnopqrstuvwxyz
658\end{tabular}
659&
660\begin{tabular}[t]{@{}lllll@{}}
661abcd    & efgh  & ijkl  & mnop  & qrst  \\
662uvwx    & yzab  & cdef  & ghij  & klmn  \\
663opqr    & stuv  & wxyz  &               &
664\end{tabular}
665\end{tabular}
666\end{center}
667The example takes advantage of resuming a generator in the constructor to prime the loops so the first character sent for formatting appears inside the nested loops.
668The destructor provides a newline, if formatted text ends with a full line.
669Figure~\ref{f:CFormatSim} shows the C implementation of the \CFA input generator with one additional field and the computed @goto@.
670For contrast, Figure~\ref{f:PythonFormatter} shows the equivalent Python format generator with the same properties as the Fibonacci generator.
671
672Figure~\ref{f:DeviceDriverGen} shows a \emph{killer} asymmetric generator, a device-driver, because device drivers caused 70\%-85\% of failures in Windows/Linux~\cite{Swift05}.
673Device drives follow the pattern of simple data state but complex execution state, \ie finite state-machine (FSM) parsing a protocol.
674For example, the following protocol:
675\begin{center}
676\ldots\, STX \ldots\, message \ldots\, ESC ETX \ldots\, message \ldots\, ETX 2-byte crc \ldots
677\end{center}
678is a network message beginning with the control character STX, ending with an ETX, and followed by a 2-byte cyclic-redundancy check.
679Control characters may appear in a message if preceded by an ESC.
680When a message byte arrives, it triggers an interrupt, and the operating system services the interrupt by calling the device driver with the byte read from a hardware register.
681The device driver returns a status code of its current state, and when a complete message is obtained, the operating system knows the message is in the message buffer.
682Hence, the device driver is an input/output generator.
683
684Note, the cost of creating and resuming the device-driver generator, @Driver@, is virtually identical to call/return, so performance in an operating-system kernel is excellent.
685As well, the data state is small, where variables @byte@ and @msg@ are communication variables for passing in message bytes and returning the message, and variables @lnth@, @crc@, and @sum@ are local variable that must be retained between calls and are manually hoisted into the generator type.
686% Manually, detecting and hoisting local-state variables is easy when the number is small.
687In contrast, the execution state is large, with one @resume@ and seven @suspend@s.
688Hence, the key benefits of the generator are correctness, safety, and maintenance because the execution states are transcribed directly into the programming language rather than using a table-driven approach.
689Because FSMs can be complex and frequently occur in important domains, direct generator support is important in a system programming language.
690
691\begin{figure}
692\centering
693\newbox\myboxA
694\begin{lrbox}{\myboxA}
695\begin{python}[aboveskip=0pt,belowskip=0pt]
696def Fib():
697        fn1, fn = 0, 1
698        while True:
699                `yield fn1`
700                fn1, fn = fn, fn1 + fn
701f1 = Fib()
702f2 = Fib()
703for i in range( 10 ):
704        print( next( f1 ), next( f2 ) )
705
706
707
708
709
710
711\end{python}
712\end{lrbox}
713
714\newbox\myboxB
715\begin{lrbox}{\myboxB}
716\begin{python}[aboveskip=0pt,belowskip=0pt]
717def Fmt():
718        try:
719                while True:
720                        for g in range( 5 ):
721                                for b in range( 4 ):
722                                        print( `(yield)`, end='' )
723                                print( '  ', end='' )
724                        print()
725        except GeneratorExit:
726                if g != 0 | b != 0:
727                        print()
728fmt = Fmt()
729`next( fmt )`                    # prime, next prewritten
730for i in range( 41 ):
731        `fmt.send( 'a' );`      # send to yield
732\end{python}
733\end{lrbox}
734\subfloat[Fibonacci]{\label{f:PythonFibonacci}\usebox\myboxA}
735\hspace{3pt}
736\vrule
737\hspace{3pt}
738\subfloat[Formatter]{\label{f:PythonFormatter}\usebox\myboxB}
739\caption{Python generator}
740\label{f:PythonGenerator}
741
742\bigskip
743
744\begin{tabular}{@{}l|l@{}}
745\begin{cfa}[aboveskip=0pt,belowskip=0pt]
746enum Status { CONT, MSG, ESTX,
747                                ELNTH, ECRC };
748`generator` Driver {
749        Status status;
750        unsigned char byte, * msg; // communication
751        unsigned int lnth, sum;      // local state
752        unsigned short int crc;
753};
754void ?{}( Driver & d, char * m ) { d.msg = m; }
755Status next( Driver & d, char b ) with( d ) {
756        byte = b; `resume( d );` return status;
757}
758void main( Driver & d ) with( d ) {
759        enum { STX = '\002', ESC = '\033',
760                        ETX = '\003', MaxMsg = 64 };
761  msg: for () { // parse message
762                status = CONT;
763                lnth = 0; sum = 0;
764                while ( byte != STX ) `suspend;`
765          emsg: for () {
766                        `suspend;` // process byte
767\end{cfa}
768&
769\begin{cfa}[aboveskip=0pt,belowskip=0pt]
770                        choose ( byte ) { // switch with implicit break
771                          case STX:
772                                status = ESTX; `suspend;` continue msg;
773                          case ETX:
774                                break emsg;
775                          case ESC:
776                                `suspend;`
777                        }
778                        if ( lnth >= MaxMsg ) { // buffer full ?
779                                status = ELNTH; `suspend;` continue msg; }
780                        msg[lnth++] = byte;
781                        sum += byte;
782                }
783                msg[lnth] = '\0'; // terminate string
784                `suspend;`
785                crc = byte << 8;
786                `suspend;`
787                status = (crc | byte) == sum ? MSG : ECRC;
788                `suspend;`
789        }
790}
791\end{cfa}
792\end{tabular}
793\caption{Device-driver generator for communication protocol}
794\label{f:DeviceDriverGen}
795\end{figure}
796
797Figure~\ref{f:CFAPingPongGen} shows a symmetric generator, where the generator resumes another generator, forming a resume/resume cycle.
798(The trivial cycle is a generator resuming itself.)
799This control flow is similar to recursion for functions but without stack growth.
800The steps for symmetric control-flow are creating, executing, and terminating the cycle.
801Constructing the cycle must deal with definition-before-use to close the cycle, \ie, the first generator must know about the last generator, which is not within scope.
802(This issue occurs for any cyclic data structure.)
803% The example creates all the generators and then assigns the partners that form the cycle.
804% Alternatively, the constructor can assign the partners as they are declared, except the first, and the first-generator partner is set after the last generator declaration to close the cycle.
805Once the cycle is formed, the program main resumes one of the generators, and the generators can then traverse an arbitrary cycle using @resume@ to activate partner generator(s).
806Terminating the cycle is accomplished by @suspend@ or @return@, both of which go back to the stack frame that started the cycle (program main in the example).
807The starting stack-frame is below the last active generator because the resume/resume cycle does not grow the stack.
808Also, since local variables are not retained in the generator function, it does not contain any objects with destructors that must be called, so the  cost is the same as a function return.
809Destructor cost occurs when the generator instance is deallocated, which is easily controlled by the programmer.
810
811Figure~\ref{f:CPingPongSim} shows the implementation of the symmetric generator, where the complexity is the @resume@, which needs an extension to the calling convention to perform a forward rather than backward jump.
812This jump-starts at the top of the next generator main to re-execute the normal calling convention to make space on the stack for its local variables.
813However, before the jump, the caller must reset its stack (and any registers) equivalent to a @return@, but subsequently jump forward.
814This semantics is basically a tail-call optimization, which compilers already perform.
815The example shows the assembly code to undo the generator's entry code before the direct jump.
816This assembly code depends on what entry code is generated, specifically if there are local variables and the level of optimization.
817To provide this new calling convention requires a mechanism built into the compiler, which is beyond the scope of \CFA at this time.
818Nevertheless, it is possible to hand generate any symmetric generators for proof of concept and performance testing.
819A compiler could also eliminate other artifacts in the generator simulation to further increase performance, \eg LLVM has various coroutine support~\cite{CoroutineTS}, and \CFA can leverage this support should it fork @clang@.
820
821\begin{figure}
822\centering
823\begin{lrbox}{\myboxA}
824\begin{cfa}[aboveskip=0pt,belowskip=0pt]
825`generator PingPong` {
826        const char * name;
827        int N;
828        int i;                          // local state
829        PingPong & partner; // rebindable reference
830};
831
832void `main( PingPong & pp )` with(pp) {
833        for ( ; i < N; i += 1 ) {
834                sout | name | i;
835                `resume( partner );`
836        }
837}
838int main() {
839        enum { N = 5 };
840        PingPong ping = {"ping",N,0}, pong = {"pong",N,0};
841        &ping.partner = &pong;  &pong.partner = &ping;
842        `resume( ping );`
843}
844\end{cfa}
845\end{lrbox}
846
847\begin{lrbox}{\myboxB}
848\begin{cfa}[escapechar={},aboveskip=0pt,belowskip=0pt]
849typedef struct PingPong {
850        const char * name;
851        int N, i;
852        struct PingPong * partner;
853        void * next;
854} PingPong;
855#define PPCtor(name, N) {name,N,0,NULL,NULL}
856void comain( PingPong * pp ) {
857        if ( pp->next ) goto *pp->next;
858        pp->next = &&cycle;
859        for ( ; pp->i < pp->N; pp->i += 1 ) {
860                printf( "%s %d\n", pp->name, pp->i );
861                asm( "mov  %0,%%rdi" : "=m" (pp->partner) );
862                asm( "mov  %rdi,%rax" );
863                asm( "popq %rbx" );
864                asm( "jmp  comain" );
865          cycle: ;
866        }
867}
868\end{cfa}
869\end{lrbox}
870
871\subfloat[\CFA symmetric generator]{\label{f:CFAPingPongGen}\usebox\myboxA}
872\hspace{3pt}
873\vrule
874\hspace{3pt}
875\subfloat[C generator simulation]{\label{f:CPingPongSim}\usebox\myboxB}
876\hspace{3pt}
877\caption{Ping-Pong symmetric generator}
878\label{f:PingPongSymmetricGenerator}
879\end{figure}
880
881Finally, part of this generator work was inspired by the recent \CCtwenty generator proposal~\cite{C++20Coroutine19} (which they call coroutines).
882Our work provides the same high-performance asymmetric generators as \CCtwenty, and extends their work with symmetric generators.
883An additional \CCtwenty generator feature allows @suspend@ and @resume@ to be followed by a restricted compound statement that is executed after the current generator has reset its stack but before calling the next generator, specified with \CFA syntax:
884\begin{cfa}
885... suspend`{ ... }`;
886... resume( C )`{ ... }` ...
887\end{cfa}
888Since the current generator's stack is released before calling the compound statement, the compound statement can only reference variables in the generator's type.
889This feature is useful when a generator is used in a concurrent context to ensure it is stopped before releasing a lock in the compound statement, which might immediately allow another thread to resume the generator.
890Hence, this mechanism provides a general and safe handoff of the generator among competing threads.
891
892
893\subsection{Coroutine}
894\label{s:Coroutine}
895
896Stackful coroutines extend generator semantics, \ie there is an implicit closure and @suspend@ may appear in a helper function called from the coroutine main.
897A coroutine is specified by replacing @generator@ with @coroutine@ for the type.
898Coroutine generality results in higher cost for creation, due to dynamic stack allocation, execution, due to context switching among stacks, and terminating, due to possible stack unwinding and dynamic stack deallocation.
899A series of different kinds of coroutines and their implementations demonstrate how coroutines extend generators.
900
901First, the previous generator examples are converted to their coroutine counterparts, allowing local-state variables to be moved from the generator type into the coroutine main.
902\begin{description}
903\item[Fibonacci]
904Move the declaration of @fn1@ to the start of coroutine main.
905\begin{cfa}[xleftmargin=0pt]
906void main( Fib & fib ) with(fib) {
907        `int fn1;`
908\end{cfa}
909\item[Formatter]
910Move the declaration of @g@ and @b@ to the for loops in the coroutine main.
911\begin{cfa}[xleftmargin=0pt]
912for ( `g`; 5 ) {
913        for ( `b`; 4 ) {
914\end{cfa}
915\item[Device Driver]
916Move the declaration of @lnth@ and @sum@ to their points of initialization.
917\begin{cfa}[xleftmargin=0pt]
918        status = CONT;
919        `unsigned int lnth = 0, sum = 0;`
920        ...
921        `unsigned short int crc = byte << 8;`
922\end{cfa}
923\item[PingPong]
924Move the declaration of @i@ to the for loop in the coroutine main.
925\begin{cfa}[xleftmargin=0pt]
926void main( PingPong & pp ) with(pp) {
927        for ( `i`; N ) {
928\end{cfa}
929\end{description}
930It is also possible to refactor code containing local-state and @suspend@ statements into a helper function, like the computation of the CRC for the device driver.
931\begin{cfa}
932unsigned int Crc() {
933        `suspend;`
934        unsigned short int crc = byte << 8;
935        `suspend;`
936        status = (crc | byte) == sum ? MSG : ECRC;
937        return crc;
938}
939\end{cfa}
940A call to this function is placed at the end of the driver's coroutine-main.
941For complex finite-state machines, refactoring is part of normal program abstraction, especially when code is used in multiple places.
942Again, this complexity is usually associated with execution state rather than data state.
943
944\begin{comment}
945Figure~\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 functions, \eg @next@.
946Like 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.
947The 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@.
948The interface function @next@, takes a Fibonacci instance and context switches to it using @resume@;
949on restart, the Fibonacci field, @fn@, contains the next value in the sequence, which is returned.
950The first @resume@ is special because it allocates the coroutine stack and cocalls its coroutine main on that stack;
951when the coroutine main returns, its stack is deallocated.
952Hence, @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.
953Figure~\ref{f:Coroutine1State} shows the coroutine version of the C version in Figure~\ref{f:ExternalState}.
954Coroutine generators are called \newterm{output coroutines} because values are only returned.
955
956\begin{figure}
957\centering
958\newbox\myboxA
959% \begin{lrbox}{\myboxA}
960% \begin{cfa}[aboveskip=0pt,belowskip=0pt]
961% `int fn1, fn2, state = 1;`   // single global variables
962% int fib() {
963%       int fn;
964%       `switch ( state )` {  // explicit execution state
965%         case 1: fn = 0;  fn1 = fn;  state = 2;  break;
966%         case 2: fn = 1;  fn2 = fn1;  fn1 = fn;  state = 3;  break;
967%         case 3: fn = fn1 + fn2;  fn2 = fn1;  fn1 = fn;  break;
968%       }
969%       return fn;
970% }
971% int main() {
972%
973%       for ( int i = 0; i < 10; i += 1 ) {
974%               printf( "%d\n", fib() );
975%       }
976% }
977% \end{cfa}
978% \end{lrbox}
979\begin{lrbox}{\myboxA}
980\begin{cfa}[aboveskip=0pt,belowskip=0pt]
981#define FibCtor { 0, 1 }
982typedef struct { int fn1, fn; } Fib;
983int fib( Fib * f ) {
984
985        int ret = f->fn1;
986        f->fn1 = f->fn;
987        f->fn = ret + f->fn;
988        return ret;
989}
990
991
992
993int main() {
994        Fib f1 = FibCtor, f2 = FibCtor;
995        for ( int i = 0; i < 10; i += 1 ) {
996                printf( "%d %d\n",
997                                fib( &f1 ), fib( &f2 ) );
998        }
999}
1000\end{cfa}
1001\end{lrbox}
1002
1003\newbox\myboxB
1004\begin{lrbox}{\myboxB}
1005\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1006`coroutine` Fib { int fn1; };
1007void main( Fib & fib ) with( fib ) {
1008        int fn;
1009        [fn1, fn] = [0, 1];
1010        for () {
1011                `suspend;`
1012                [fn1, fn] = [fn, fn1 + fn];
1013        }
1014}
1015int ?()( Fib & fib ) with( fib ) {
1016        return `resume( fib )`.fn1;
1017}
1018int main() {
1019        Fib f1, f2;
1020        for ( 10 ) {
1021                sout | f1() | f2();
1022}
1023
1024
1025\end{cfa}
1026\end{lrbox}
1027
1028\newbox\myboxC
1029\begin{lrbox}{\myboxC}
1030\begin{python}[aboveskip=0pt,belowskip=0pt]
1031
1032def Fib():
1033
1034        fn1, fn = 0, 1
1035        while True:
1036                `yield fn1`
1037                fn1, fn = fn, fn1 + fn
1038
1039
1040// next prewritten
1041
1042
1043f1 = Fib()
1044f2 = Fib()
1045for i in range( 10 ):
1046        print( next( f1 ), next( f2 ) )
1047
1048
1049
1050\end{python}
1051\end{lrbox}
1052
1053\subfloat[C]{\label{f:GlobalVariables}\usebox\myboxA}
1054\hspace{3pt}
1055\vrule
1056\hspace{3pt}
1057\subfloat[\CFA]{\label{f:ExternalState}\usebox\myboxB}
1058\hspace{3pt}
1059\vrule
1060\hspace{3pt}
1061\subfloat[Python]{\label{f:ExternalState}\usebox\myboxC}
1062\caption{Fibonacci generator}
1063\label{f:C-fibonacci}
1064\end{figure}
1065
1066\bigskip
1067
1068\newbox\myboxA
1069\begin{lrbox}{\myboxA}
1070\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1071`coroutine` Fib { int fn; };
1072void main( Fib & fib ) with( fib ) {
1073        fn = 0;  int fn1 = fn; `suspend`;
1074        fn = 1;  int fn2 = fn1;  fn1 = fn; `suspend`;
1075        for () {
1076                fn = fn1 + fn2; fn2 = fn1; fn1 = fn; `suspend`; }
1077}
1078int next( Fib & fib ) with( fib ) { `resume( fib );` return fn; }
1079int main() {
1080        Fib f1, f2;
1081        for ( 10 )
1082                sout | next( f1 ) | next( f2 );
1083}
1084\end{cfa}
1085\end{lrbox}
1086\newbox\myboxB
1087\begin{lrbox}{\myboxB}
1088\begin{python}[aboveskip=0pt,belowskip=0pt]
1089
1090def Fibonacci():
1091        fn = 0; fn1 = fn; `yield fn`  # suspend
1092        fn = 1; fn2 = fn1; fn1 = fn; `yield fn`
1093        while True:
1094                fn = fn1 + fn2; fn2 = fn1; fn1 = fn; `yield fn`
1095
1096
1097f1 = Fibonacci()
1098f2 = Fibonacci()
1099for i in range( 10 ):
1100        print( `next( f1 )`, `next( f2 )` ) # resume
1101
1102\end{python}
1103\end{lrbox}
1104\subfloat[\CFA]{\label{f:Coroutine3States}\usebox\myboxA}
1105\qquad
1106\subfloat[Python]{\label{f:Coroutine1State}\usebox\myboxB}
1107\caption{Fibonacci input coroutine, 3 states, internal variables}
1108\label{f:cfa-fibonacci}
1109\end{figure}
1110\end{comment}
1111
1112\begin{figure}
1113\centering
1114\lstset{language=CFA,escapechar={},moredelim=**[is][\protect\color{red}]{`}{`}}% allow $
1115\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
1116\begin{cfa}
1117`coroutine` Prod {
1118        Cons & c;                       // communication
1119        int N, money, receipt;
1120};
1121void main( Prod & prod ) with( prod ) {
1122        // 1st resume starts here
1123        for ( i; N ) {
1124                int p1 = random( 100 ), p2 = random( 100 );
1125                sout | p1 | " " | p2;
1126                int status = delivery( c, p1, p2 );
1127                sout | " $" | money | nl | status;
1128                receipt += 1;
1129        }
1130        stop( c );
1131        sout | "prod stops";
1132}
1133int payment( Prod & prod, int money ) {
1134        prod.money = money;
1135        `resume( prod );`
1136        return prod.receipt;
1137}
1138void start( Prod & prod, int N, Cons &c ) {
1139        &prod.c = &c;
1140        prod.[N, receipt] = [N, 0];
1141        `resume( prod );`
1142}
1143int main() {
1144        Prod prod;
1145        Cons cons = { prod };
1146        start( prod, 5, cons );
1147}
1148\end{cfa}
1149&
1150\begin{cfa}
1151`coroutine` Cons {
1152        Prod & p;                       // communication
1153        int p1, p2, status;
1154        bool done;
1155};
1156void ?{}( Cons & cons, Prod & p ) {
1157        &cons.p = &p; // reassignable reference
1158        cons.[status, done ] = [0, false];
1159}
1160void main( Cons & cons ) with( cons ) {
1161        // 1st resume starts here
1162        int money = 1, receipt;
1163        for ( ; ! done; ) {
1164                sout | p1 | " " | p2 | nl | " $" | money;
1165                status += 1;
1166                receipt = payment( p, money );
1167                sout | " #" | receipt;
1168                money += 1;
1169        }
1170        sout | "cons stops";
1171}
1172int delivery( Cons & cons, int p1, int p2 ) {
1173        cons.[p1, p2] = [p1, p2];
1174        `resume( cons );`
1175        return cons.status;
1176}
1177void stop( Cons & cons ) {
1178        cons.done = true;
1179        `resume( cons );`
1180}
1181
1182\end{cfa}
1183\end{tabular}
1184\caption{Producer / consumer: resume-resume cycle, bidirectional communication}
1185\label{f:ProdCons}
1186\end{figure}
1187
1188Figure~\ref{f:ProdCons} shows the ping-pong example in Figure~\ref{f:CFAPingPongGen} extended into a producer/consumer symmetric-coroutine performing bidirectional communication.
1189This example is illustrative because both producer/consumer have two interface functions with @resume@s that suspend execution in these interface (helper) functions.
1190The program main creates the producer coroutine, passes it to the consumer coroutine in its initialization, and closes the cycle at the call to @start@ along with the number of items to be produced.
1191The first @resume@ of @prod@ creates @prod@'s stack with a frame for @prod@'s coroutine main at the top, and context switches to it.
1192@prod@'s coroutine main starts, creates local-state 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.
1193
1194The producer call to @delivery@ transfers values into the consumer's communication variables, resumes the consumer, and returns the consumer status.
1195On the first resume, @cons@'s stack is created and initialized, holding local-state variables retained between subsequent activations of the coroutine.
1196The 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).
1197The call from the consumer to @payment@ introduces the cycle between producer and consumer.
1198When @payment@ is called, the consumer copies values into the producer's communication variable and a resume is executed.
1199The context switch restarts the producer at the point where it last context switched, so it continues in @delivery@ after the resume.
1200@delivery@ returns the status value in @prod@'s coroutine main, where the status is printed.
1201The loop then repeats calling @delivery@, where each call resumes the consumer coroutine.
1202The context switch to the consumer continues in @payment@.
1203The consumer increments and returns the receipt to the call in @cons@'s coroutine main.
1204The loop then repeats calling @payment@, where each call resumes the producer coroutine.
1205Figure~\ref{f:ProdConsRuntimeStacks} shows the runtime stacks of the program main, and the coroutine mains for @prod@ and @cons@ during the cycling.
1206
1207\begin{figure}
1208\begin{center}
1209\input{FullProdConsStack.pstex_t}
1210\end{center}
1211\vspace*{-10pt}
1212\caption{Producer / consumer runtime stacks}
1213\label{f:ProdConsRuntimeStacks}
1214
1215\medskip
1216
1217\begin{center}
1218\input{FullCoroutinePhases.pstex_t}
1219\end{center}
1220\vspace*{-10pt}
1221\caption{Ping / Pong coroutine steps}
1222\label{f:PingPongFullCoroutineSteps}
1223\end{figure}
1224
1225Terminating a coroutine cycle is more complex than a generator cycle, because it requires context switching to the program main's \emph{stack} to shutdown the program, whereas generators started by the program main run on its stack.
1226Furthermore, each deallocated coroutine must guarantee all destructors are run for object allocated in the coroutine type \emph{and} allocated on the coroutine's stack at the point of suspension, which can be arbitrarily deep.
1227When a coroutine's main ends, its stack is already unwound so any stack allocated objects with destructors have been finalized.
1228The na\"{i}ve semantics for coroutine-cycle termination is to context switch to the last resumer, like executing a @suspend@/@return@ in a generator.
1229However, for coroutines, the last resumer is \emph{not} implicitly below the current stack frame, as for generators, because each coroutine's stack is independent.
1230Unfortunately, it is impossible to determine statically if a coroutine is in a cycle and unrealistic to check dynamically (graph-cycle problem).
1231Hence, a compromise solution is necessary that works for asymmetric (acyclic) and symmetric (cyclic) coroutines.
1232
1233Our solution is to context switch back to the first resumer (starter) once the coroutine ends.
1234This semantics works well for the most common asymmetric and symmetric coroutine usage patterns.
1235For asymmetric coroutines, it is common for the first resumer (starter) coroutine to be the only resumer.
1236All previous generators converted to coroutines have this property.
1237For symmetric coroutines, it is common for the cycle creator to persist for the lifetime of the cycle.
1238Hence, the starter coroutine is remembered on the first resume and ending the coroutine resumes the starter.
1239Figure~\ref{f:ProdConsRuntimeStacks} shows this semantic by the dashed lines from the end of the coroutine mains: @prod@ starts @cons@ so @cons@ resumes @prod@ at the end, and the program main starts @prod@ so @prod@ resumes the program main at the end.
1240For other scenarios, it is always possible to devise a solution with additional programming effort, such as forcing the cycle forward (backward) to a safe point before starting termination.
1241
1242The producer/consumer example does not illustrate the full power of the starter semantics because @cons@ always ends first.
1243Assume generator @PingPong@ is converted to a coroutine.
1244Figure~\ref{f:PingPongFullCoroutineSteps} shows the creation, starter, and cyclic execution steps of the coroutine version.
1245The program main creates (declares) coroutine instances @ping@ and @pong@.
1246Next, program main resumes @ping@, making it @ping@'s starter, and @ping@'s main resumes @pong@'s main, making it @pong@'s starter.
1247Execution forms a cycle when @pong@ resumes @ping@, and cycles $N$ times.
1248By adjusting $N$ for either @ping@/@pong@, it is possible to have either one finish first, instead of @pong@ always ending first.
1249If @pong@ ends first, it resumes its starter @ping@ in its coroutine main, then @ping@ ends and resumes its starter the program main in function @start@.
1250If @ping@ ends first, it resumes its starter the program main in function @start@.
1251Regardless of the cycle complexity, the starter stack always leads back to the program main, but the stack can be entered at an arbitrary point.
1252Once back at the program main, coroutines @ping@ and @pong@ are deallocated.
1253For generators, deallocation runs the destructors for all objects in the generator type.
1254For coroutines, deallocation deals with objects in the coroutine type and must also run the destructors for any objects pending on the coroutine's stack for any unterminated coroutine.
1255Hence, if a coroutine's destructor detects the coroutine is not ended, it implicitly raises a cancellation exception (uncatchable exception) at the coroutine and resumes it so the cancellation exception can propagate to the root of the coroutine's stack destroying all local variable on the stack.
1256So the \CFA semantics for the generator and coroutine, ensure both can be safely deallocated at any time, regardless of their current state, like any other aggregate object.
1257Explicitly raising normal exceptions at another coroutine can replace flag variables, like @stop@, \eg @prod@ raises a @stop@ exception at @cons@ after it finishes generating values and resumes @cons@, which catches the @stop@ exception to terminate its loop.
1258
1259Finally, there is an interesting effect for @suspend@ with symmetric coroutines.
1260A coroutine must retain its last resumer to suspend back because the resumer is on a different stack.
1261These reverse pointers allow @suspend@ to cycle \emph{backwards}, which may be useful in certain cases.
1262However, there is an anomaly if a coroutine resumes itself, because it overwrites its last resumer with itself, losing the ability to resume the last external resumer.
1263To prevent losing this information, a self-resume does not overwrite the last resumer.
1264
1265
1266\subsection{Generator / Coroutine Implementation}
1267
1268A significant implementation challenge for generators/coroutines (and threads in Section~\ref{s:threads}) is adding extra fields to the custom types and related functions, \eg inserting code after/before the coroutine constructor/destructor and @main@ to create/initialize/de-initialize/destroy any extra fields, \eg stack.
1269There are several solutions to these problem, which follow from the object-oriented flavour of adopting custom types.
1270
1271For object-oriented languages, inheritance is used to provide extra fields and code via explicit inheritance:
1272\begin{cfa}[morekeywords={class,inherits}]
1273class myCoroutine inherits baseCoroutine { ... }
1274\end{cfa}
1275% The problem is that the programming language and its tool chain, \eg debugger, @valgrind@, need to understand @baseCoroutine@ because it infers special property, so type @baseCoroutine@ becomes a de facto keyword and all types inheriting from it are implicitly custom types.
1276The problem is that some special properties are not handled by existing language semantics, \eg the execution of constructors/destructors is in the wrong order to implicitly start threads because the thread must start \emph{after} all constructors as it relies on a completely initialized object, but the inherited constructor runs \emph{before} the derived.
1277Alternatives, such as explicitly starting threads as in Java, are repetitive and forgetting to call start is a common source of errors.
1278An alternative is composition:
1279\begin{cfa}
1280struct myCoroutine {
1281        ... // declaration/communication variables
1282        baseCoroutine dummy; // composition, last declaration
1283}
1284\end{cfa}
1285which also requires an explicit declaration that must be last to ensure correct initialization order.
1286However, there is nothing preventing wrong placement or multiple declarations.
1287
1288\CFA custom types make any special properties explicit to the language and its tool chain, \eg the language code-generator knows where to inject code
1289% and when it is unsafe to perform certain optimizations,
1290and IDEs using simple parsing can find and manipulate types with special properties.
1291The downside of this approach is that it makes custom types a special case in the language.
1292Users wanting to extend custom types or build their own can only do so in ways offered by the language.
1293Furthermore, implementing custom types without language support may display the power of a programming language.
1294\CFA blends the two approaches, providing custom type for idiomatic \CFA code, while extending and building new custom types is still possible, similar to Java concurrency with builtin and library.
1295
1296Part of the mechanism to generalize custom types is the \CFA trait~\cite[\S~2.3]{Moss18}, \eg the definition for custom-type @coroutine@ is anything satisfying the trait @is_coroutine@, and this trait both enforces and restricts the coroutine-interface functions.
1297\begin{cfa}
1298trait is_coroutine( `dtype` T ) {
1299        void main( T & );
1300        coroutine_desc * get_coroutine( T & );
1301};
1302forall( `dtype` T | is_coroutine(T) ) void $suspend$( T & ), resume( T & );
1303\end{cfa}
1304Note, copying generators/coroutines/threads is not meaningful.
1305For example, both the resumer and suspender descriptors can have bidirectional pointers;
1306copying these coroutines does not update the internal pointers so behaviour of both copies would be difficult to understand.
1307Furthermore, two coroutines cannot logically execute on the same stack.
1308A deep coroutine copy, which copies the stack, is also meaningless in an unmanaged language (no garbage collection), like C, because the stack may contain pointers to object within it that require updating for the copy.
1309The \CFA @dtype@ property provides no \emph{implicit} copying operations and the @is_coroutine@ trait provides no \emph{explicit} copying operations, so all coroutines must be passed by reference (pointer).
1310The function definitions ensure there is a statically typed @main@ function that is the starting point (first stack frame) of a coroutine, and a mechanism to get (read) the coroutine descriptor from its handle.
1311The @main@ function has no return value or additional parameters because the coroutine type allows an arbitrary number of interface functions with corresponding arbitrary typed input/output values versus fixed ones.
1312The 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@ function, and possibly redefining \textsf{suspend} and @resume@.
1313
1314The \CFA custom-type @coroutine@ implicitly implements the getter and forward declarations for the coroutine main.
1315\begin{cquote}
1316\begin{tabular}{@{}ccc@{}}
1317\begin{cfa}
1318coroutine MyCor {
1319        int value;
1320
1321};
1322\end{cfa}
1323&
1324{\Large $\Rightarrow$}
1325&
1326\begin{tabular}{@{}ccc@{}}
1327\begin{cfa}
1328struct MyCor {
1329        int value;
1330        coroutine_desc cor;
1331};
1332\end{cfa}
1333&
1334\begin{cfa}
1335static inline coroutine_desc *
1336get_coroutine( MyCor & this ) {
1337        return &this.cor;
1338}
1339\end{cfa}
1340&
1341\begin{cfa}
1342void main( MyCor * this );
1343
1344
1345
1346\end{cfa}
1347\end{tabular}
1348\end{tabular}
1349\end{cquote}
1350The combination of custom types and fundamental @trait@ description of these types allows a concise specification for programmers and tools, while more advanced programmers can have tighter control over memory layout and initialization.
1351
1352Figure~\ref{f:CoroutineMemoryLayout} shows different memory-layout options for a coroutine (where a task is similar).
1353The coroutine handle is the @coroutine@ instance containing programmer specified type global/communication variables across interface functions.
1354The coroutine descriptor contains all implicit declarations needed by the runtime, \eg @suspend@/@resume@, and can be part of the coroutine handle or separate.
1355The coroutine stack can appear in a number of locations and be fixed or variable sized.
1356Hence, the coroutine's stack could be a VLS\footnote{
1357We are examining variable-sized structures (VLS), where fields can be variable-sized structures or arrays.
1358Once allocated, a VLS is fixed sized.}
1359on the allocating stack, provided the allocating stack is large enough.
1360For a VLS stack allocation/deallocation is an inexpensive adjustment of the stack pointer, modulo any stack constructor costs (\eg initial frame setup).
1361For heap stack allocation, allocation/deallocation is an expensive heap allocation (where the heap can be a shared resource), modulo any stack constructor costs.
1362With heap stack allocation, it is also possible to use a split (segmented) stack calling convention, available with gcc and clang, so the stack is variable sized.
1363Currently, \CFA supports stack/heap allocated descriptors but only fixed-sized heap allocated stacks.
1364In \CFA debug-mode, the fixed-sized stack is terminated with a write-only page, which catches most stack overflows.
1365Experience teaching concurrency with \uC~\cite{CS343} shows fixed-sized stacks are rarely an issue for students.
1366Split-stack allocation is under development but requires recompilation of legacy code, which may be impossible.
1367
1368\begin{figure}
1369\centering
1370\input{corlayout.pstex_t}
1371\caption{Coroutine memory layout}
1372\label{f:CoroutineMemoryLayout}
1373\end{figure}
1374
1375
1376\section{Concurrency}
1377\label{s:Concurrency}
1378
1379Concurrency is nondeterministic scheduling of independent sequential execution paths (threads), where each thread has its own stack.
1380A single thread with multiple call stacks, \newterm{coroutining}~\cite{Conway63,Marlin80}, does \emph{not} imply concurrency~\cite[\S~2]{Buhr05a}.
1381In coroutining, coroutines self-schedule the thread across stacks so execution is deterministic.
1382(It is \emph{impossible} to generate a concurrency error when coroutining.)
1383However, coroutines are a stepping stone towards concurrency.
1384
1385The transition to concurrency, even for a single thread with multiple stacks, occurs when coroutines context switch to a \newterm{scheduling coroutine}, introducing non-determinism from the coroutine perspective~\cite[\S~3,]{Buhr05a}.
1386Therefore, a minimal concurrency system requires coroutines \emph{in conjunction with a nondeterministic scheduler}.
1387The resulting execution system now follows a cooperative threading model~\cite{Adya02,libdill}, called \newterm{non-preemptive scheduling}.
1388Adding \newterm{preemption} introduces non-cooperative scheduling, where context switching occurs randomly between any two instructions often based on a timer interrupt, called \newterm{preemptive scheduling}.
1389While a scheduler introduces uncertain execution among explicit context switches, preemption introduces uncertainty by introducing implicit context switches.
1390Uncertainty gives the illusion of parallelism on a single processor and provides a mechanism to access and increase performance on multiple processors.
1391The reason is that the scheduler/runtime have complete knowledge about resources and how to best utilized them.
1392However, the introduction of unrestricted nondeterminism results in the need for \newterm{mutual exclusion} and \newterm{synchronization}, which restrict nondeterminism for correctness;
1393otherwise, it is impossible to write meaningful concurrent programs.
1394Optimal concurrent performance is often obtained by having as much nondeterminism as mutual exclusion and synchronization correctness allow.
1395
1396A scheduler can either be a stackless or stackful.
1397For 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.
1398For 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.
1399The \CFA runtime uses a stackful scheduler for uniformity and security.
1400
1401
1402\subsection{Thread}
1403\label{s:threads}
1404
1405Threading needs the ability to start a thread and wait for its completion.
1406A common API for this ability is @fork@ and @join@.
1407\begin{cquote}
1408\begin{tabular}{@{}lll@{}}
1409\multicolumn{1}{c}{\textbf{Java}} & \multicolumn{1}{c}{\textbf{\Celeven}} & \multicolumn{1}{c}{\textbf{pthreads}} \\
1410\begin{cfa}
1411class MyTask extends Thread {...}
1412mytask t = new MyTask(...);
1413`t.start();` // start
1414// concurrency
1415`t.join();` // wait
1416\end{cfa}
1417&
1418\begin{cfa}
1419class MyTask { ... } // functor
1420MyTask mytask;
1421`thread t( mytask, ... );` // start
1422// concurrency
1423`t.join();` // wait
1424\end{cfa}
1425&
1426\begin{cfa}
1427void * rtn( void * arg ) {...}
1428pthread_t t;  int i = 3;
1429`pthread_create( &t, rtn, (void *)i );` // start
1430// concurrency
1431`pthread_join( t, NULL );` // wait
1432\end{cfa}
1433\end{tabular}
1434\end{cquote}
1435\CFA has a simpler approach using a custom @thread@ type and leveraging declaration semantics (allocation/deallocation), where threads implicitly @fork@ after construction and @join@ before destruction.
1436\begin{cfa}
1437thread MyTask {};
1438void main( MyTask & this ) { ... }
1439int main() {
1440        MyTask team`[10]`; $\C[2.5in]{// allocate stack-based threads, implicit start after construction}$
1441        // concurrency
1442} $\C{// deallocate stack-based threads, implicit joins before destruction}$
1443\end{cfa}
1444This semantic ensures a thread is started and stopped exactly once, eliminating some programming error, and scales to multiple threads for basic (termination) synchronization.
1445For block allocation to arbitrary depth, including recursion, threads are created/destroyed in a lattice structure (tree with top and bottom).
1446Arbitrary topologies are possible using dynamic allocation, allowing threads to outlive their declaration scope, identical to normal dynamic allocation.
1447\begin{cfa}
1448MyTask * factory( int N ) { ... return `anew( N )`; } $\C{// allocate heap-based threads, implicit start after construction}$
1449int main() {
1450        MyTask * team = factory( 10 );
1451        // concurrency
1452        `delete( team );` $\C{// deallocate heap-based threads, implicit joins before destruction}\CRT$
1453}
1454\end{cfa}
1455
1456Figure~\ref{s:ConcurrentMatrixSummation} shows concurrently adding the rows of a matrix and then totalling the subtotals sequentially, after all the row threads have terminated.
1457The program uses heap-based threads because each thread needs different constructor values.
1458(Python provides a simple iteration mechanism to initialize array elements to different values allowing stack allocation.)
1459The allocation/deallocation pattern appears unusual because allocated objects are immediately deallocated without any intervening code.
1460However, for threads, the deletion provides implicit synchronization, which is the intervening code.
1461% While 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.
1462
1463\begin{figure}
1464\begin{cfa}
1465`thread` Adder { int * row, cols, & subtotal; } $\C{// communication variables}$
1466void ?{}( Adder & adder, int row[], int cols, int & subtotal ) {
1467        adder.[ row, cols, &subtotal ] = [ row, cols, &subtotal ];
1468}
1469void main( Adder & adder ) with( adder ) {
1470        subtotal = 0;
1471        for ( c; cols ) { subtotal += row[c]; }
1472}
1473int main() {
1474        const int rows = 10, cols = 1000;
1475        int matrix[rows][cols], subtotals[rows], total = 0;
1476        // read matrix
1477        Adder * adders[rows];
1478        for ( r; rows; ) { $\C{// start threads to sum rows}$
1479                adders[r] = `new( matrix[r], cols, &subtotals[r] );`
1480        }
1481        for ( r; rows ) { $\C{// wait for threads to finish}$
1482                `delete( adders[r] );` $\C{// termination join}$
1483                total += subtotals[r]; $\C{// total subtotal}$
1484        }
1485        sout | total;
1486}
1487\end{cfa}
1488\caption{Concurrent matrix summation}
1489\label{s:ConcurrentMatrixSummation}
1490\end{figure}
1491
1492
1493\subsection{Thread Implementation}
1494
1495Threads in \CFA are user level run by runtime kernel threads (see Section~\ref{s:CFARuntimeStructure}), where user threads provide concurrency and kernel threads provide parallelism.
1496Like coroutines, and for the same design reasons, \CFA provides a custom @thread@ type and a @trait@ to enforce and restrict the task-interface functions.
1497\begin{cquote}
1498\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1499\begin{cfa}
1500thread myThread {
1501        ... // declaration/communication variables
1502};
1503
1504
1505\end{cfa}
1506&
1507\begin{cfa}
1508trait is_thread( `dtype` T ) {
1509        void main( T & );
1510        thread_desc * get_thread( T & );
1511        void ^?{}( T & `mutex` );
1512};
1513\end{cfa}
1514\end{tabular}
1515\end{cquote}
1516Like coroutines, the @dtype@ property prevents \emph{implicit} copy operations and the @is_thread@ trait provides no \emph{explicit} copy operations, so threads must be passed by reference (pointer).
1517Similarly, the function definitions ensure there is a statically typed @main@ function that is the thread starting point (first stack frame), a mechanism to get (read) the thread descriptor from its handle, and a special destructor to prevent deallocation while the thread is executing.
1518(The qualifier @mutex@ for the destructor parameter is discussed in Section~\ref{s:Monitor}.)
1519The difference between the coroutine and thread is that a coroutine borrows a thread from its caller, so the first thread resuming a coroutine creates the coroutine's stack and starts running the coroutine main on the stack;
1520whereas, a thread is scheduling for execution in @main@ immediately after its constructor is run.
1521No return value or additional parameters are necessary for this function because the @thread@ type allows an arbitrary number of interface functions with corresponding arbitrary typed input/output values.
1522
1523
1524\section{Mutual Exclusion / Synchronization}
1525\label{s:MutualExclusionSynchronization}
1526
1527Unrestricted nondeterminism is meaningless as there is no way to know when the result is completed without synchronization.
1528To produce meaningful execution requires clawing back some determinism using mutual exclusion and synchronization, where mutual exclusion provides access control for threads using shared data, and synchronization is a timing relationship among threads~\cite[\S~4]{Buhr05a}.
1529Some concurrent systems eliminate mutable shared-state by switching to stateless communication like message passing~\cite{Thoth,Harmony,V-Kernel,MPI} (Erlang, MPI), channels~\cite{CSP} (CSP,Go), actors~\cite{Akka} (Akka, Scala), or functional techniques (Haskell).
1530However, these approaches introduce a new communication mechanism for concurrency different from the standard communication using function call/return.
1531Hence, a programmer must learn and manipulate two sets of design/programming patterns.
1532While this distinction can be hidden away in library code, effective use of the library still has to take both paradigms into account.
1533In contrast, approaches based on stateful models more closely resemble the standard call/return programming model, resulting in a single programming paradigm.
1534
1535At 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}.
1536However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
1537A newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
1538While 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.
1539
1540One of the most natural, elegant, and efficient mechanisms for mutual exclusion and synchronization for shared-memory systems is the \emph{monitor}.
1541First 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}.
1542In 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.
1543For these reasons, \CFA selected monitors as the core high-level concurrency construct, upon which higher-level approaches can be easily constructed.
1544
1545
1546\subsection{Mutual Exclusion}
1547
1548A group of instructions manipulating a specific instance of shared data that must be performed atomically is called a \newterm{critical section}~\cite{Dijkstra65}, which is enforced by \newterm{simple mutual-exclusion}.
1549The generalization is called a \newterm{group critical-section}~\cite{Joung00}, where multiple tasks with the same session use the resource simultaneously and different sessions are segregated, which is enforced by \newterm{complex mutual-exclusion} providing the correct kind and number of threads using a group critical-section.
1550The readers/writer problem~\cite{Courtois71} is an instance of a group critical-section, where readers share a session but writers have a unique session.
1551
1552However, many solutions exist for mutual exclusion, which vary in terms of performance, flexibility and ease of use.
1553Methods 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.
1554Ease 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.
1555For 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.
1556However, a significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
1557Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
1558
1559
1560\subsection{Synchronization}
1561
1562Synchronization enforces relative ordering of execution, and synchronization tools provide numerous mechanisms to establish these timing relationships.
1563Low-level synchronization primitives offer good performance and flexibility at the cost of ease of use;
1564higher-level mechanisms often simplify usage by adding better coupling between synchronization and data, \eg receive-specific versus receive-any thread in message passing or offering specialized solutions, \eg barrier lock.
1565Often synchronization is used to order access to a critical section, \eg ensuring a waiting writer thread enters the critical section before a calling reader thread.
1566If the calling reader is scheduled before the waiting writer, the reader has barged.
1567Barging 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).
1568Preventing or detecting barging is an involved challenge with low-level locks, which is made easier through higher-level constructs.
1569This challenge is often split into two different approaches: barging avoidance and prevention.
1570Algorithms that unconditionally releasing a lock for competing threads to acquire use barging avoidance during synchronization to force a barging thread to wait;
1571algorithms that conditionally hold locks during synchronization, \eg baton-passing~\cite{Andrews89}, prevent barging completely.
1572
1573
1574\section{Monitor}
1575\label{s:Monitor}
1576
1577A \textbf{monitor} is a set of functions that ensure mutual exclusion when accessing shared state.
1578More precisely, a monitor is a programming technique that implicitly binds mutual exclusion to static function 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).
1579Restricting acquire/release points eases programming, comprehension, and maintenance, at a slight cost in flexibility and efficiency.
1580\CFA uses a custom @monitor@ type and leverages declaration semantics (deallocation) to protect active or waiting threads in a monitor.
1581
1582The following is a \CFA monitor implementation of an atomic counter.
1583\begin{cfa}[morekeywords=nomutex]
1584`monitor` Aint { int cnt; }; $\C[4.25in]{// atomic integer counter}$
1585int ++?( Aint & `mutex`$\(_{opt}\)$ this ) with( this ) { return ++cnt; } $\C{// increment}$
1586int ?=?( Aint & `mutex`$\(_{opt}\)$ lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions with int}\CRT$
1587int ?=?( int & lhs, Aint & `mutex`$\(_{opt}\)$ rhs ) with( rhs ) { lhs = cnt; }
1588\end{cfa}
1589% The @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.
1590% (While a constructor may publish its address into a global variable, doing so generates a race-condition.)
1591The prefix increment operation, @++?@, is normally @mutex@, indicating mutual exclusion is necessary during function execution, to protect the incrementing from race conditions, unless there is an atomic increment instruction for the implementation type.
1592The assignment operators provide bidirectional conversion between an atomic and normal integer without accessing field @cnt@;
1593these operations only need @mutex@, if reading/writing the implementation type is not atomic.
1594The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics, which is similar to the \CC template @std::atomic@.
1595\begin{cfa}
1596int i = 0, j = 0, k = 5;
1597Aint x = { 0 }, y = { 0 }, z = { 5 }; $\C{// no mutex required}$
1598++x; ++y; ++z; $\C{// safe increment by multiple threads}$
1599x = 2; y = i; z = k; $\C{// conversions}$
1600i = x; j = y; k = z;
1601\end{cfa}
1602
1603\CFA monitors have \newterm{multi-acquire} semantics so the thread in the monitor may acquire it multiple times without deadlock, allowing recursion and calling other interface functions.
1604\begin{cfa}
1605monitor M { ... } m;
1606void foo( M & mutex m ) { ... } $\C{// acquire mutual exclusion}$
1607void bar( M & mutex m ) { $\C{// acquire mutual exclusion}$
1608        ... `bar( m );` ... `foo( m );` ... $\C{// reacquire mutual exclusion}$
1609}
1610\end{cfa}
1611\CFA monitors also ensure the monitor lock is released regardless of how an acquiring function ends (normal or exceptional), and returning a shared variable is safe via copying before the lock is released.
1612Similar safety is offered by \emph{explicit} mechanisms like \CC RAII;
1613monitor \emph{implicit} safety ensures no programmer usage errors.
1614Furthermore, RAII mechanisms cannot handle complex synchronization within a monitor, where the monitor lock may not be released on function exit because it is passed to an unblocking thread;
1615RAII is purely a mutual-exclusion mechanism (see Section~\ref{s:Scheduling}).
1616
1617
1618\subsection{Monitor Implementation}
1619
1620For the same design reasons, \CFA provides a custom @monitor@ type and a @trait@ to enforce and restrict the monitor-interface functions.
1621\begin{cquote}
1622\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1623\begin{cfa}
1624monitor M {
1625        ... // shared data
1626};
1627
1628\end{cfa}
1629&
1630\begin{cfa}
1631trait is_monitor( `dtype` T ) {
1632        monitor_desc * get_monitor( T & );
1633        void ^?{}( T & mutex );
1634};
1635\end{cfa}
1636\end{tabular}
1637\end{cquote}
1638The @dtype@ property prevents \emph{implicit} copy operations and the @is_monitor@ trait provides no \emph{explicit} copy operations, so monitors must be passed by reference (pointer).
1639% Copying 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.
1640% Copying 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.
1641Similarly, the function definitions ensures there is a mechanism to get (read) the monitor descriptor from its handle, and a special destructor to prevent deallocation if a thread using the shared data.
1642The custom monitor type also inserts any locks needed to implement the mutual exclusion semantics.
1643
1644
1645\subsection{Mutex Acquisition}
1646\label{s:MutexAcquisition}
1647
1648While the monitor lock provides mutual exclusion for shared data, there are implementation options for when and where the locking/unlocking occurs.
1649(Much of this discussion also applies to basic locks.)
1650For example, a monitor may be passed through multiple helper functions before it is necessary to acquire the monitor's mutual exclusion.
1651
1652The benefit of mandatory monitor qualifiers is self-documentation, but requiring both @mutex@ and \lstinline[morekeywords=nomutex]@nomutex@ for all monitor parameters is redundant.
1653Instead, the semantics has one qualifier as the default and the other required.
1654For example, make the safe @mutex@ qualifier the default because assuming \lstinline[morekeywords=nomutex]@nomutex@ may cause subtle errors.
1655Alternatively, make the unsafe \lstinline[morekeywords=nomutex]@nomutex@ qualifier the default because it is the \emph{normal} parameter semantics while @mutex@ parameters are rare.
1656Providing a default qualifier implies knowing whether a parameter is a monitor.
1657Since \CFA relies heavily on traits as an abstraction mechanism, types can coincidentally match the monitor trait but not be a monitor, similar to inheritance where a shape and playing card can both be drawable.
1658For 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@.
1659
1660The next semantic decision is establishing which parameter \emph{types} may be qualified with @mutex@.
1661The following has monitor parameter types that are composed of multiple objects.
1662\begin{cfa}
1663monitor M { ... }
1664int f1( M & mutex m ); $\C{// single parameter object}$
1665int f2( M * mutex m ); $\C{// single or multiple parameter object}$
1666int f3( M * mutex m[$\,$] ); $\C{// multiple parameter object}$
1667int f4( stack( M * ) & mutex m ); $\C{// multiple parameters object}$
1668\end{cfa}
1669Function @f1@ has a single parameter object, while @f2@'s indirection could be a single or multi-element array, where static array size is often unknown in C.
1670Function @f3@ has a multiple object matrix, and @f4@ a multiple object data structure.
1671While shown shortly, multiple object acquisition is possible, but the number of objects must be statically known.
1672Therefore, \CFA only acquires one monitor per parameter with at most one level of indirection, excluding pointers as it is impossible to statically determine the size.
1673
1674For object-oriented monitors, \eg Java, calling a mutex member \emph{implicitly} acquires mutual exclusion of the receiver object, @`rec`.foo(...)@.
1675\CFA has no receiver, and hence, the explicit @mutex@ qualifier is used to specify which objects acquire mutual exclusion.
1676A positive consequence of this design decision is the ability to support multi-monitor functions,\footnote{
1677While object-oriented monitors can be extended with a mutex qualifier for multiple-monitor members, no prior example of this feature could be found.}
1678called \newterm{bulk acquire}.
1679\CFA guarantees acquisition order is consistent across calls to @mutex@ functions using the same monitors as arguments, so acquiring multiple monitors is safe from deadlock.
1680Figure~\ref{f:BankTransfer} shows a trivial solution to the bank transfer problem~\cite{BankTransfer}, where two resources must be locked simultaneously, using \CFA monitors with implicit locking and \CC with explicit locking.
1681A \CFA programmer only has to manage when to acquire mutual exclusion;
1682a \CC programmer must select the correct lock and acquisition mechanism from a panoply of locking options.
1683Making good choices for common cases in \CFA simplifies the programming experience and enhances safety.
1684
1685\begin{figure}
1686\centering
1687\begin{lrbox}{\myboxA}
1688\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1689monitor BankAccount {
1690
1691        int balance;
1692} b1 = { 0 }, b2 = { 0 };
1693void deposit( BankAccount & `mutex` b,
1694                        int deposit ) with(b) {
1695        balance += deposit;
1696}
1697void transfer( BankAccount & `mutex` my,
1698        BankAccount & `mutex` your, int me2you ) {
1699
1700        deposit( my, -me2you ); // debit
1701        deposit( your, me2you ); // credit
1702}
1703`thread` Person { BankAccount & b1, & b2; };
1704void main( Person & person ) with(person) {
1705        for ( 10_000_000 ) {
1706                if ( random() % 3 ) deposit( b1, 3 );
1707                if ( random() % 3 ) transfer( b1, b2, 7 );
1708        }
1709}   
1710int main() {
1711        `Person p1 = { b1, b2 }, p2 = { b2, b1 };`
1712
1713} // wait for threads to complete
1714\end{cfa}
1715\end{lrbox}
1716
1717\begin{lrbox}{\myboxB}
1718\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1719struct BankAccount {
1720        `recursive_mutex m;`
1721        int balance = 0;
1722} b1, b2;
1723void deposit( BankAccount & b, int deposit ) {
1724        `scoped_lock lock( b.m );`
1725        b.balance += deposit;
1726}
1727void transfer( BankAccount & my,
1728                        BankAccount & your, int me2you ) {
1729        `scoped_lock lock( my.m, your.m );`
1730        deposit( my, -me2you ); // debit
1731        deposit( your, me2you ); // credit
1732}
1733
1734void person( BankAccount & b1, BankAccount & b2 ) {
1735        for ( int i = 0; i < 10$'$000$'$000; i += 1 ) {
1736                if ( random() % 3 ) deposit( b1, 3 );
1737                if ( random() % 3 ) transfer( b1, b2, 7 );
1738        }
1739}   
1740int main() {
1741        `thread p1(person, ref(b1), ref(b2)), p2(person, ref(b2), ref(b1));`
1742        `p1.join(); p2.join();`
1743}
1744\end{cfa}
1745\end{lrbox}
1746
1747\subfloat[\CFA]{\label{f:CFABank}\usebox\myboxA}
1748\hspace{3pt}
1749\vrule
1750\hspace{3pt}
1751\subfloat[\CC]{\label{f:C++Bank}\usebox\myboxB}
1752\hspace{3pt}
1753\caption{Bank transfer problem}
1754\label{f:BankTransfer}
1755\end{figure}
1756
1757Users can still force the acquiring order by using @mutex@/\lstinline[morekeywords=nomutex]@nomutex@.
1758\begin{cfa}
1759void foo( M & mutex m1, M & mutex m2 ); $\C{// acquire m1 and m2}$
1760void bar( M & mutex m1, M & /* nomutex */ m2 ) { $\C{// acquire m1}$
1761        ... foo( m1, m2 ); ... $\C{// acquire m2}$
1762}
1763void baz( M & /* nomutex */ m1, M & mutex m2 ) { $\C{// acquire m2}$
1764        ... foo( m1, m2 ); ... $\C{// acquire m1}$
1765}
1766\end{cfa}
1767The bulk-acquire semantics allow @bar@ or @baz@ to acquire a monitor lock and reacquire it in @foo@.
1768The calls to @bar@ and @baz@ acquired the monitors in opposite order, possibly resulting in deadlock.
1769However, this case is the simplest instance of the \emph{nested-monitor problem}~\cite{Lister77}, where monitors are acquired in sequence versus bulk.
1770Detecting the nested-monitor problem requires dynamic tracking of monitor calls, and dealing with it requires rollback semantics~\cite{Dice10}.
1771\CFA does not deal with this fundamental problem.
1772
1773Finally, like Java, \CFA offers an alternative @mutex@ statement to reduce refactoring and naming.
1774\begin{cquote}
1775\renewcommand{\arraystretch}{0.0}
1776\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1777\multicolumn{1}{c}{\textbf{\lstinline@mutex@ call}} & \multicolumn{1}{c}{\lstinline@mutex@ \textbf{statement}} \\
1778\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1779monitor M { ... };
1780void foo( M & mutex m1, M & mutex m2 ) {
1781        // critical section
1782}
1783void bar( M & m1, M & m2 ) {
1784        foo( m1, m2 );
1785}
1786\end{cfa}
1787&
1788\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1789
1790void bar( M & m1, M & m2 ) {
1791        mutex( m1, m2 ) {       // remove refactoring and naming
1792                // critical section
1793        }
1794}
1795
1796\end{cfa}
1797\end{tabular}
1798\end{cquote}
1799
1800
1801\subsection{Scheduling}
1802\label{s:Scheduling}
1803
1804% There are many aspects of scheduling in a concurrency system, all related to resource utilization by waiting threads, \ie which thread gets the resource next.
1805% Different forms of scheduling include access to processors by threads (see Section~\ref{s:RuntimeStructureCluster}), another is access to a shared resource by a lock or monitor.
1806This section discusses monitor scheduling for waiting threads eligible for entry, \ie which thread gets the shared resource next. (See Section~\ref{s:RuntimeStructureCluster} for scheduling threads on virtual processors.)
1807While monitor mutual-exclusion provides safe access to shared data, the monitor data may indicate that a thread accessing it cannot proceed, \eg a bounded buffer may be full/empty so produce/consumer threads must block.
1808Leaving the monitor and trying again (busy waiting) is impractical for high-level programming.
1809Monitors eliminate busy waiting by providing synchronization to schedule threads needing access to the shared data, where threads block versus spinning.
1810Synchronization is generally achieved with internal~\cite{Hoare74} or external~\cite[\S~2.9.2]{uC++} scheduling.
1811\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.
1812Finally, \CFA monitors do not allow calling threads to barge ahead of signalled threads, which simplifies synchronization among threads in the monitor and increases correctness.
1813If barging is allowed, synchronization between a signaller and signallee is difficult, often requiring additional flags and multiple unblock/block cycles.
1814In fact, signals-as-hints is completely opposite from that proposed by Hoare in the seminal paper on monitors~\cite[p.~550]{Hoare74}.
1815% \begin{cquote}
1816% However, 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.
1817% It 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}
1818% \end{cquote}
1819Furthermore, \CFA concurrency has no spurious wakeup~\cite[\S~9]{Buhr05a}, which eliminates an implicit form of self barging.
1820Hence, a \CFA @wait@ statement is not enclosed in a @while@ loop retesting a blocking predicate, which can cause thread starvation due to barging.
1821
1822Figure~\ref{f:MonitorScheduling} shows general internal/external scheduling (for the bounded-buffer example in Figure~\ref{f:InternalExternalScheduling}).
1823External calling threads block on the calling queue, if the monitor is occupied, otherwise they enter in FIFO order.
1824Internal threads block on condition queues via @wait@ and reenter from the condition in FIFO order.
1825Alternatively, internal threads block on urgent from the @signal_block@ or @waitfor@, and reenter implicitly when the monitor becomes empty, \ie, the thread in the monitor exits or waits.
1826
1827There are three signalling mechanisms to unblock waiting threads to enter the monitor.
1828Note, signalling cannot have the signaller and signalled thread in the monitor simultaneously because of the mutual exclusion, so either the signaller or signallee can proceed.
1829For internal scheduling, threads are unblocked from condition queues using @signal@, where the signallee is moved to urgent and the signaller continues (solid line).
1830Multiple signals move multiple signallees to urgent until the condition is empty.
1831When the signaller exits or waits, a thread blocked on urgent is processed before calling threads to prevent barging.
1832(Java conceptually moves the signalled thread to the calling queue, and hence, allows barging.)
1833The alternative unblock is in the opposite order using @signal_block@, where the signaller is moved to urgent and the signallee continues (dashed line), and is implicitly unblocked from urgent when the signallee exits or waits.
1834
1835For external scheduling, the condition queues are not used;
1836instead threads are unblocked directly from the calling queue using @waitfor@ based on function names requesting mutual exclusion.
1837(The linear search through the calling queue to locate a particular call can be reduced to $O(1)$.)
1838The @waitfor@ has the same semantics as @signal_block@, where the signalled thread executes before the signallee, which waits on urgent.
1839Executing multiple @waitfor@s from different signalled functions causes the calling threads to move to urgent.
1840External scheduling requires urgent to be a stack, because the signaller expects to execute immediately after the specified monitor call has exited or waited.
1841Internal scheduling behaves the same for an urgent stack or queue, except for multiple signalling, where the threads unblock from urgent in reverse order from signalling.
1842If the restart order is important, multiple signalling by a signal thread can be transformed into daisy-chain signalling among threads, where each thread signals the next thread.
1843We tried both a stack for @waitfor@ and queue for signalling, but that resulted in complex semantics about which thread enters next.
1844Hence, \CFA uses a single urgent stack to correctly handle @waitfor@ and adequately support both forms of signalling.
1845
1846\begin{figure}
1847\centering
1848% \subfloat[Scheduling Statements] {
1849% \label{fig:SchedulingStatements}
1850% {\resizebox{0.45\textwidth}{!}{\input{CondSigWait.pstex_t}}}
1851\input{CondSigWait.pstex_t}
1852% }% subfloat
1853% \quad
1854% \subfloat[Bulk acquire monitor] {
1855% \label{fig:BulkMonitor}
1856% {\resizebox{0.45\textwidth}{!}{\input{ext_monitor.pstex_t}}}
1857% }% subfloat
1858\caption{Monitor Scheduling}
1859\label{f:MonitorScheduling}
1860\end{figure}
1861
1862Figure~\ref{f:BBInt} shows a \CFA generic bounded-buffer with internal scheduling, where producers/consumers enter the monitor, detect the buffer is full/empty, and block on an appropriate condition variable, @full@/@empty@.
1863The @wait@ function atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the function's parameter list.
1864The appropriate condition variable is signalled to unblock an opposite kind of thread after an element is inserted/removed from the buffer.
1865Signalling is unconditional, because signalling an empty condition variable does nothing.
1866It is common to declare condition variables as monitor fields to prevent shared access, hence no locking is required for access as the conditions are protected by the monitor lock.
1867In \CFA, a condition variable can be created/stored independently.
1868% To still prevent expensive locking on access, a condition variable is tied to a \emph{group} of monitors on first use, called \newterm{branding}, resulting in a low-cost boolean test to detect sharing from other monitors.
1869
1870% Signalling semantics cannot have the signaller and signalled thread in the monitor simultaneously, which means:
1871% \begin{enumerate}
1872% \item
1873% The signalling thread returns immediately and the signalled thread continues.
1874% \item
1875% The signalling thread continues and the signalled thread is marked for urgent unblocking at the next scheduling point (exit/wait).
1876% \item
1877% The signalling thread blocks but is marked for urgent unblocking at the next scheduling point and the signalled thread continues.
1878% \end{enumerate}
1879% The first approach is too restrictive, as it precludes solving a reasonable class of problems, \eg dating service (see Figure~\ref{f:DatingService}).
1880% \CFA supports the next two semantics as both are useful.
1881
1882\begin{figure}
1883\centering
1884\newbox\myboxA
1885\begin{lrbox}{\myboxA}
1886\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1887forall( otype T ) { // distribute forall
1888        monitor Buffer {
1889                `condition` full, empty;
1890                int front, back, count;
1891                T elements[10];
1892        };
1893        void ?{}( Buffer(T) & buffer ) with(buffer) {
1894                front = back = count = 0;
1895        }
1896        void insert( Buffer(T) & mutex buffer, T elem )
1897                                with(buffer) {
1898                if ( count == 10 ) `wait( empty )`;
1899                // insert elem into buffer
1900                `signal( full )`;
1901        }
1902        T remove( Buffer(T) & mutex buffer ) with(buffer) {
1903                if ( count == 0 ) `wait( full )`;
1904                // remove elem from buffer
1905                `signal( empty )`;
1906                return elem;
1907        }
1908}
1909\end{cfa}
1910\end{lrbox}
1911
1912% \newbox\myboxB
1913% \begin{lrbox}{\myboxB}
1914% \begin{cfa}[aboveskip=0pt,belowskip=0pt]
1915% forall( otype T ) { // distribute forall
1916%       monitor Buffer {
1917%
1918%               int front, back, count;
1919%               T elements[10];
1920%       };
1921%       void ?{}( Buffer(T) & buffer ) with(buffer) {
1922%               [front, back, count] = 0;
1923%       }
1924%       T remove( Buffer(T) & mutex buffer ); // forward
1925%       void insert( Buffer(T) & mutex buffer, T elem )
1926%                               with(buffer) {
1927%               if ( count == 10 ) `waitfor( remove, buffer )`;
1928%               // insert elem into buffer
1929%
1930%       }
1931%       T remove( Buffer(T) & mutex buffer ) with(buffer) {
1932%               if ( count == 0 ) `waitfor( insert, buffer )`;
1933%               // remove elem from buffer
1934%
1935%               return elem;
1936%       }
1937% }
1938% \end{cfa}
1939% \end{lrbox}
1940
1941\newbox\myboxB
1942\begin{lrbox}{\myboxB}
1943\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1944monitor ReadersWriter {
1945        int rcnt, wcnt; // readers/writer using resource
1946};
1947void ?{}( ReadersWriter & rw ) with(rw) {
1948        rcnt = wcnt = 0;
1949}
1950void EndRead( ReadersWriter & mutex rw ) with(rw) {
1951        rcnt -= 1;
1952}
1953void EndWrite( ReadersWriter & mutex rw ) with(rw) {
1954        wcnt = 0;
1955}
1956void StartRead( ReadersWriter & mutex rw ) with(rw) {
1957        if ( wcnt > 0 ) `waitfor( EndWrite, rw );`
1958        rcnt += 1;
1959}
1960void StartWrite( ReadersWriter & mutex rw ) with(rw) {
1961        if ( wcnt > 0 ) `waitfor( EndWrite, rw );`
1962        else while ( rcnt > 0 ) `waitfor( EndRead, rw );`
1963        wcnt = 1;
1964}
1965
1966\end{cfa}
1967\end{lrbox}
1968
1969\subfloat[Generic bounded buffer, internal scheduling]{\label{f:BBInt}\usebox\myboxA}
1970\hspace{3pt}
1971\vrule
1972\hspace{3pt}
1973\subfloat[Readers / writer lock, external scheduling]{\label{f:RWExt}\usebox\myboxB}
1974
1975\caption{Internal / external scheduling}
1976\label{f:InternalExternalScheduling}
1977\end{figure}
1978
1979Figure~\ref{f:BBInt} can be transformed into external scheduling by removing the condition variables and signals/waits, and adding the following lines at the locations of the current @wait@s in @insert@/@remove@, respectively.
1980\begin{cfa}[aboveskip=2pt,belowskip=1pt]
1981if ( count == 10 ) `waitfor( remove, buffer )`;       |      if ( count == 0 ) `waitfor( insert, buffer )`;
1982\end{cfa}
1983Here, the producers/consumers detects a full/\-empty buffer and prevents more producers/consumers from entering the monitor until there is a free/empty slot in the buffer.
1984External scheduling is controlled by the @waitfor@ statement, which atomically blocks the calling thread, releases the monitor lock, and restricts the function calls that can next acquire mutual exclusion.
1985If 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.
1986Threads calling excluded functions block outside of (external to) the monitor on the calling queue, versus blocking on condition queues inside of (internal to) the monitor.
1987Figure~\ref{f:RWExt} shows a readers/writer lock written using external scheduling, where a waiting reader detects a writer using the resource and restricts further calls until the writer exits by calling @EndWrite@.
1988The writer does a similar action for each reader or writer using the resource.
1989Note, no new calls to @StarRead@/@StartWrite@ may occur when waiting for the call to @EndRead@/@EndWrite@.
1990External scheduling allows waiting for events from other threads while restricting unrelated events, that would otherwise have to wait on conditions in the monitor.
1991The mechnaism can be done in terms of control flow, \eg Ada @accept@ or \uC @_Accept@, or in terms of data, \eg Go @select@ on channels.
1992While both mechanisms have strengths and weaknesses, this project uses the control-flow mechanism to be consistent with other language features.
1993% Two challenges specific to \CFA for external scheduling are loose object-definitions (see Section~\ref{s:LooseObjectDefinitions}) and multiple-monitor functions (see Section~\ref{s:Multi-MonitorScheduling}).
1994
1995Figure~\ref{f:DatingService} shows a dating service demonstrating non-blocking and blocking signalling.
1996The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
1997A thread blocks until an appropriate partner arrives.
1998The complexity is exchanging phone numbers in the monitor because of the mutual-exclusion property.
1999For 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.
2000For 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.
2001The 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;
2002as well, an arriving thread may not find a partner and must wait, which requires a condition variable, and condition variables imply internal scheduling.
2003Furthermore, barging corrupts the dating service during an exchange because a barger may also match and change the phone numbers, invalidating the previous exchange phone number.
2004Putting loops around the @wait@s does not correct the problem;
2005the simple solution must be restructured to account for barging.
2006
2007\begin{figure}
2008\centering
2009\newbox\myboxA
2010\begin{lrbox}{\myboxA}
2011\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2012enum { CCodes = 20 };
2013monitor DS {
2014        int GirlPhNo, BoyPhNo;
2015        condition Girls[CCodes], Boys[CCodes];
2016        `condition exchange;`
2017};
2018int girl( DS & mutex ds, int phNo, int ccode ) {
2019        if ( is_empty( Boys[ccode] ) ) {
2020                wait( Girls[ccode] );
2021                GirlPhNo = phNo;
2022                `signal( exchange );`
2023        } else {
2024                GirlPhNo = phNo;
2025                `signal( Boys[ccode] );`
2026                `wait( exchange );`
2027        }
2028        return BoyPhNo;
2029}
2030int boy( DS & mutex ds, int phNo, int ccode ) {
2031        // as above with boy/girl interchanged
2032}
2033\end{cfa}
2034\end{lrbox}
2035
2036\newbox\myboxB
2037\begin{lrbox}{\myboxB}
2038\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2039
2040monitor DS {
2041        int GirlPhNo, BoyPhNo;
2042        condition Girls[CCodes], Boys[CCodes];
2043
2044};
2045int girl( DS & mutex ds, int phNo, int ccode ) {
2046        if ( is_empty( Boys[ccode] ) ) { // no compatible
2047                wait( Girls[ccode] ); // wait for boy
2048                GirlPhNo = phNo; // make phone number available
2049
2050        } else {
2051                GirlPhNo = phNo; // make phone number available
2052                `signal_block( Boys[ccode] );` // restart boy
2053
2054        } // if
2055        return BoyPhNo;
2056}
2057int boy( DS & mutex ds, int phNo, int ccode ) {
2058        // as above with boy/girl interchanged
2059}
2060\end{cfa}
2061\end{lrbox}
2062
2063\subfloat[\lstinline@signal@]{\label{f:DatingSignal}\usebox\myboxA}
2064\qquad
2065\subfloat[\lstinline@signal_block@]{\label{f:DatingSignalBlock}\usebox\myboxB}
2066\caption{Dating service}
2067\label{f:DatingService}
2068\end{figure}
2069
2070In summation, for internal scheduling, non-blocking signalling (as in the producer/consumer example) is used when the signaller is providing the cooperation for a waiting thread;
2071the 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.
2072The waiter unblocks next from the urgent queue, uses/takes the state, and exits the monitor.
2073Blocking signal is the reverse, where the waiter is providing the cooperation for the signalling thread;
2074the 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.
2075The waiter changes state and exits the monitor, and the signaller unblocks next from the urgent queue to use/take the state.
2076
2077Both internal and external scheduling extend to multiple monitors in a natural way.
2078\begin{cquote}
2079\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
2080\begin{cfa}
2081monitor M { `condition e`; ... };
2082void foo( M & mutex m1, M & mutex m2 ) {
2083        ... wait( `e` ); ...   // wait( e, m1, m2 )
2084        ... wait( `e, m1` ); ...
2085        ... wait( `e, m2` ); ...
2086}
2087\end{cfa}
2088&
2089\begin{cfa}
2090void rtn$\(_1\)$( M & mutex m1, M & mutex m2 );
2091void rtn$\(_2\)$( M & mutex m1 );
2092void bar( M & mutex m1, M & mutex m2 ) {
2093        ... waitfor( `rtn` ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\), m1, m2 )}$
2094        ... waitfor( `rtn, m1` ); ... // $\LstCommentStyle{waitfor( rtn\(_2\), m1 )}$
2095}
2096\end{cfa}
2097\end{tabular}
2098\end{cquote}
2099For @wait( e )@, the default semantics is to atomically block the signaller and release all acquired mutex parameters, \ie @wait( e, m1, m2 )@.
2100To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @wait( e, m1 )@.
2101Wait cannot statically verifies the released monitors are the acquired mutex-parameters without disallowing separately compiled helper functions calling @wait@.
2102While \CC supports bulk locking, @wait@ only accepts a single lock for a condition variable, so bulk locking with condition variables is asymmetric.
2103Finally, a signaller,
2104\begin{cfa}
2105void baz( M & mutex m1, M & mutex m2 ) {
2106        ... signal( e ); ...
2107}
2108\end{cfa}
2109must have acquired at least the same locks as the waiting thread signalled from a condition queue to allow the locks to be passed, and hence, prevent barging.
2110
2111Similarly, for @waitfor( rtn )@, the default semantics is to atomically block the acceptor and release all acquired mutex parameters, \ie @waitfor( rtn, m1, m2 )@.
2112To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn, m1 )@.
2113@waitfor@ does statically verify the monitor types passed are the same as the acquired mutex-parameters of the given function or function pointer, hence the function (pointer) prototype must be accessible.
2114% When an overloaded function appears in an @waitfor@ statement, calls to any function with that name are accepted.
2115% The rationale is that members with the same name should perform a similar function, and therefore, all should be eligible to accept a call.
2116Overloaded functions can be disambiguated using a cast
2117\begin{cfa}
2118void rtn( M & mutex m );
2119`int` rtn( M & mutex m );
2120waitfor( (`int` (*)( M & mutex ))rtn, m );
2121\end{cfa}
2122
2123The ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock.
2124\begin{cfa}
2125void foo( M & mutex m1, M & mutex m2 ) {
2126        ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired )}$
2127void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2 )}$
2128        ... signal( `e` ); ...
2129\end{cfa}
2130The @wait@ only releases @m1@ so the signalling thread cannot acquire @m1@ and @m2@ to enter @bar@ and @signal@ the condition.
2131While deadlock can occur with multiple/nesting acquisition, this is a consequence of locks, and by extension monitors, not being perfectly composable.
2132
2133
2134
2135\subsection{\texorpdfstring{Extended \protect\lstinline@waitfor@}{Extended waitfor}}
2136
2137Figure~\ref{f:ExtendedWaitfor} shows the extended form of the @waitfor@ statement to conditionally accept one of a group of mutex functions, with an optional statement to be performed \emph{after} the mutex function finishes.
2138For a @waitfor@ clause to be executed, its @when@ must be true and an outstanding call to its corresponding member(s) must exist.
2139The \emph{conditional-expression} of a @when@ may call a function, but the function must not block or context switch.
2140If there are multiple acceptable mutex calls, selection occurs top-to-bottom (prioritized) among the @waitfor@ clauses, whereas some programming languages with similar mechanisms accept nondeterministically for this case, \eg Go \lstinline[morekeywords=select]@select@.
2141If some accept guards are true and there are no outstanding calls to these members, the acceptor is blocked until a call to one of these members is made.
2142If there is a @timeout@ clause, it provides an upper bound on waiting.
2143If 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.
2144Hence, the terminating @else@ clause allows a conditional attempt to accept a call without blocking.
2145If both @timeout@ and @else@ clause are present, the @else@ must be conditional, or the @timeout@ is never triggered.
2146There is also a traditional future wait queue (not shown) (\eg Microsoft (@WaitForMultipleObjects@)), to wait for a specified number of future elements in the queue.
2147
2148\begin{figure}
2149\centering
2150\begin{cfa}
2151`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
2152        waitfor( $\emph{mutex-member-name}$ ) $\emph{statement}$ $\C{// action after call}$
2153`or` `when` ( $\emph{conditional-expression}$ ) $\C{// any number of functions}$
2154        waitfor( $\emph{mutex-member-name}$ ) $\emph{statement}$
2155`or`    ...
2156`when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
2157        `timeout` $\emph{statement}$ $\C{// optional terminating timeout clause}$
2158`when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
2159        `else`  $\emph{statement}$ $\C{// optional terminating clause}$
2160\end{cfa}
2161\caption{Extended \protect\lstinline@waitfor@}
2162\label{f:ExtendedWaitfor}
2163\end{figure}
2164
2165Note, a group of conditional @waitfor@ clauses is \emph{not} the same as a group of @if@ statements, \eg:
2166\begin{cfa}
2167if ( C1 ) waitfor( mem1 );                       when ( C1 ) waitfor( mem1 );
2168else if ( C2 ) waitfor( mem2 );         or when ( C2 ) waitfor( mem2 );
2169\end{cfa}
2170The left example only accepts @mem1@ if @C1@ is true or only @mem2@ if @C2@ is true.
2171The right example accepts either @mem1@ or @mem2@ if @C1@ and @C2@ are true.
2172
2173An interesting use of @waitfor@ is accepting the @mutex@ destructor to know when an object is deallocated, \eg assume the bounded buffer is restructred from a monitor to a thread with the following @main@.
2174\begin{cfa}
2175void main( Buffer(T) & buffer ) with(buffer) {
2176        for () {
2177                `waitfor( ^?{}, buffer )` break;
2178                or when ( count != 20 ) waitfor( insert, buffer ) { ... }
2179                or when ( count != 0 ) waitfor( remove, buffer ) { ... }
2180        }
2181        // clean up
2182}
2183\end{cfa}
2184When the program main deallocates the buffer, it first calls the buffer's destructor, which is accepted, the destructor runs, and the buffer is deallocated.
2185However, the buffer thread cannot continue after the destructor call because the object is gone;
2186hence, clean up in @main@ cannot occur, which means destructors for local objects are not run.
2187To make this useful capability work, the semantics for accepting the destructor is the same as @signal@, \ie the destructor call is placed on urgent and the acceptor continues execution, which ends the loop, cleans up, and the thread terminates.
2188Then, the destructor caller unblocks from urgent to deallocate the object.
2189Accepting the destructor is the idiomatic way in \CFA to terminate a thread performing direct communication.
2190
2191
2192\subsection{Bulk Barging Prevention}
2193
2194Figure~\ref{f:BulkBargingPrevention} shows \CFA code where bulk acquire adds complexity to the internal-signalling semantics.
2195The complexity begins at the end of the inner @mutex@ statement, where the semantics of internal scheduling need to be extended for multiple monitors.
2196The problem is that bulk acquire is used in the inner @mutex@ statement where one of the monitors is already acquired.
2197When 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.
2198However, both the signalling and waiting threads W1 and W2 need some subset of monitors @m1@ and @m2@.
2199\begin{cquote}
2200condition c: (order 1) W2(@m2@), W1(@m1@,@m2@)\ \ \ or\ \ \ (order 2) W1(@m1@,@m2@), W2(@m2@) \\
2201S: acq. @m1@ $\rightarrow$ acq. @m1,m2@ $\rightarrow$ @signal(c)@ $\rightarrow$ rel. @m2@ $\rightarrow$ pass @m2@ unblock W2 (order 2) $\rightarrow$ rel. @m1@ $\rightarrow$ pass @m1,m2@ unblock W1 \\
2202\hspace*{2.75in}$\rightarrow$ rel. @m1@ $\rightarrow$ pass @m1,m2@ unblock W1 (order 1)
2203\end{cquote}
2204
2205\begin{figure}
2206\newbox\myboxA
2207\begin{lrbox}{\myboxA}
2208\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2209monitor M m1, m2;
2210condition c;
2211mutex( m1 ) { // $\LstCommentStyle{\color{red}outer}$
2212        ...
2213        mutex( m1, m2 ) { // $\LstCommentStyle{\color{red}inner}$
2214                ... `signal( c )`; ...
2215                // m1, m2 still acquired
2216        } // $\LstCommentStyle{\color{red}release m2}$
2217        // m1 acquired
2218} // release m1
2219\end{cfa}
2220\end{lrbox}
2221
2222\newbox\myboxB
2223\begin{lrbox}{\myboxB}
2224\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2225
2226
2227mutex( m1 ) {
2228        ...
2229        mutex( m1, m2 ) {
2230                ... `wait( c )`; // release m1, m2
2231                // m1, m2 reacquired
2232        } // $\LstCommentStyle{\color{red}release m2}$
2233        // m1 acquired
2234} // release m1
2235\end{cfa}
2236\end{lrbox}
2237
2238\newbox\myboxC
2239\begin{lrbox}{\myboxC}
2240\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2241
2242
2243mutex( m2 ) {
2244        ... `wait( c )`; // release m2
2245        // m2 reacquired
2246} // $\LstCommentStyle{\color{red}release m2}$
2247
2248
2249
2250
2251\end{cfa}
2252\end{lrbox}
2253
2254\begin{cquote}
2255\subfloat[Signalling Thread (S)]{\label{f:SignallingThread}\usebox\myboxA}
2256\hspace{3\parindentlnth}
2257\subfloat[Waiting Thread (W1)]{\label{f:WaitingThread}\usebox\myboxB}
2258\hspace{2\parindentlnth}
2259\subfloat[Waiting Thread (W2)]{\label{f:OtherWaitingThread}\usebox\myboxC}
2260\end{cquote}
2261\caption{Bulk Barging Prevention}
2262\label{f:BulkBargingPrevention}
2263\end{figure}
2264
2265One scheduling solution is for the signaller S 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.
2266However, this solution is inefficient if W2 waited first and can be immediate passed @m2@ when released, while S retains @m1@ until completion of the outer mutex statement.
2267If W1 waited first, the signaller must retain @m1@ amd @m2@ until completion of the outer mutex statement and then pass both to W1.
2268% Furthermore, there is an execution sequence where the signaller always finds waiter W2, and hence, waiter W1 starves.
2269To support this efficient semantics (and prevent barging), the implementation maintains a list of monitors acquired for each blocked thread.
2270When a signaller exits or waits in a monitor function/statement, the front waiter on urgent is unblocked if all its monitors are released.
2271Implementing a fast subset check for the necessary released monitors is important.
2272% The 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.
2273
2274
2275\subsection{Loose Object Definitions}
2276\label{s:LooseObjectDefinitions}
2277
2278In an object-oriented programming language, a class includes an exhaustive list of operations.
2279A new class can add members via static inheritance but the subclass still has an exhaustive list of operations.
2280(Dynamic member adding, \eg JavaScript~\cite{JavaScript}, is not considered.)
2281In the object-oriented scenario, the type and all its operators are always present at compilation (even separate compilation), so it is possible to number the operations in a bit mask and use an $O(1)$ compare with a similar bit mask created for the operations specified in a @waitfor@.
2282
2283However, in \CFA, monitor functions can be statically added/removed in translation units, making a fast subset check difficult.
2284\begin{cfa}
2285        monitor M { ... }; // common type, included in .h file
2286translation unit 1
2287        void `f`( M & mutex m );
2288        void g( M & mutex m ) { waitfor( `f`, m ); }
2289translation unit 2
2290        void `f`( M & mutex m ); $\C{// replacing f and g for type M in this translation unit}$
2291        void `g`( M & mutex m );
2292        void h( M & mutex m ) { waitfor( `f`, m ) or waitfor( `g`, m ); } $\C{// extending type M in this translation unit}$
2293\end{cfa}
2294The @waitfor@ statements in each translation unit cannot form a unique bit-mask because the monitor type does not carry that information.
2295Hence, function pointers are used to identify the functions listed in the @waitfor@ statement, stored in a variable-sized array.
2296Then, the same implementation approach used for the urgent stack is used for the calling queue.
2297Each caller has a list of monitors acquired, and the @waitfor@ statement performs a (usually short) linear search matching functions in the @waitfor@ list with called functions, and then verifying the associated mutex locks can be transfers.
2298(A possible way to construct a dense mapping is at link or load-time.)
2299
2300
2301\subsection{Multi-Monitor Scheduling}
2302\label{s:Multi-MonitorScheduling}
2303
2304External scheduling, like internal scheduling, becomes significantly more complex for multi-monitor semantics.
2305Even in the simplest case, new semantics need to be established.
2306\begin{cfa}
2307monitor M { ... };
2308void f( M & mutex m1 );
2309void g( M & mutex m1, M & mutex m2 ) { `waitfor( f );` } $\C{// pass m1 or m2 to f?}$
2310\end{cfa}
2311The solution is for the programmer to disambiguate:
2312\begin{cfa}
2313waitfor( f, `m2` ); $\C{// wait for call to f with argument m2}$
2314\end{cfa}
2315Both locks are acquired by function @g@, so when function @f@ is called, the lock for monitor @m2@ is passed from @g@ to @f@, while @g@ still holds lock @m1@.
2316This behaviour can be extended to the multi-monitor @waitfor@ statement.
2317\begin{cfa}
2318monitor M { ... };
2319void f( M & mutex m1, M & mutex m2 );
2320void g( M & mutex m1, M & mutex m2 ) { waitfor( f, `m1, m2` ); $\C{// wait for call to f with arguments m1 and m2}$
2321\end{cfa}
2322Again, the set of monitors passed to the @waitfor@ statement must be entirely contained in the set of monitors already acquired by the accepting function.
2323Also, the order of the monitors in a @waitfor@ statement is unimportant.
2324
2325Figure~\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.
2326For both examples, the set of monitors is disjoint so unblocking is impossible.
2327
2328\begin{figure}
2329\centering
2330\begin{lrbox}{\myboxA}
2331\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2332monitor M1 {} m11, m12;
2333monitor M2 {} m2;
2334condition c;
2335void f( M1 & mutex m1, M2 & mutex m2 ) {
2336        signal( c );
2337}
2338void g( M1 & mutex m1, M2 & mutex m2 ) {
2339        wait( c );
2340}
2341g( `m11`, m2 ); // block on wait
2342f( `m12`, m2 ); // cannot fulfil
2343\end{cfa}
2344\end{lrbox}
2345
2346\begin{lrbox}{\myboxB}
2347\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2348monitor M1 {} m11, m12;
2349monitor M2 {} m2;
2350
2351void f( M1 & mutex m1, M2 & mutex m2 ) {
2352
2353}
2354void g( M1 & mutex m1, M2 & mutex m2 ) {
2355        waitfor( f, m1, m2 );
2356}
2357g( `m11`, m2 ); // block on accept
2358f( `m12`, m2 ); // cannot fulfil
2359\end{cfa}
2360\end{lrbox}
2361\subfloat[Internal scheduling]{\label{f:InternalScheduling}\usebox\myboxA}
2362\hspace{3pt}
2363\vrule
2364\hspace{3pt}
2365\subfloat[External scheduling]{\label{f:ExternalScheduling}\usebox\myboxB}
2366\caption{Unmatched \protect\lstinline@mutex@ sets}
2367\label{f:UnmatchedMutexSets}
2368\end{figure}
2369
2370
2371\subsection{\texorpdfstring{\protect\lstinline@mutex@ Threads}{mutex Threads}}
2372
2373Threads in \CFA can also be monitors to allow \emph{direct communication} among threads, \ie threads can have mutex functions that are called by other threads.
2374Hence, all monitor features are available when using threads.
2375Figure~\ref{f:DirectCommunication} shows a comparison of direct call communication in \CFA with direct channel communication in Go.
2376(Ada provides a similar mechanism to the \CFA direct communication.)
2377The program main in both programs communicates directly with the other thread versus indirect communication where two threads interact through a passive monitor.
2378Both direct and indirection thread communication are valuable tools in structuring concurrent programs.
2379
2380\begin{figure}
2381\centering
2382\begin{lrbox}{\myboxA}
2383\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2384
2385struct Msg { int i, j; };
2386thread GoRtn { int i;  float f;  Msg m; };
2387void mem1( GoRtn & mutex gortn, int i ) { gortn.i = i; }
2388void mem2( GoRtn & mutex gortn, float f ) { gortn.f = f; }
2389void mem3( GoRtn & mutex gortn, Msg m ) { gortn.m = m; }
2390void ^?{}( GoRtn & mutex ) {}
2391
2392void main( GoRtn & gortn ) with( gortn ) {  // thread starts
2393
2394        for () {
2395
2396                `waitfor( mem1, gortn )` sout | i;  // wait for calls
2397                or `waitfor( mem2, gortn )` sout | f;
2398                or `waitfor( mem3, gortn )` sout | m.i | m.j;
2399                or `waitfor( ^?{}, gortn )` break;
2400
2401        }
2402
2403}
2404int main() {
2405        GoRtn gortn; $\C[2.0in]{// start thread}$
2406        `mem1( gortn, 0 );` $\C{// different calls}\CRT$
2407        `mem2( gortn, 2.5 );`
2408        `mem3( gortn, (Msg){1, 2} );`
2409
2410
2411} // wait for completion
2412\end{cfa}
2413\end{lrbox}
2414
2415\begin{lrbox}{\myboxB}
2416\begin{Go}[aboveskip=0pt,belowskip=0pt]
2417func main() {
2418        type Msg struct{ i, j int }
2419
2420        ch1 := make( chan int )
2421        ch2 := make( chan float32 )
2422        ch3 := make( chan Msg )
2423        hand := make( chan string )
2424        shake := make( chan string )
2425        gortn := func() { $\C[1.5in]{// thread starts}$
2426                var i int;  var f float32;  var m Msg
2427                L: for {
2428                        select { $\C{// wait for messages}$
2429                          case `i = <- ch1`: fmt.Println( i )
2430                          case `f = <- ch2`: fmt.Println( f )
2431                          case `m = <- ch3`: fmt.Println( m )
2432                          case `<- hand`: break L $\C{// sentinel}$
2433                        }
2434                }
2435                `shake <- "SHAKE"` $\C{// completion}$
2436        }
2437
2438        go gortn() $\C{// start thread}$
2439        `ch1 <- 0` $\C{// different messages}$
2440        `ch2 <- 2.5`
2441        `ch3 <- Msg{1, 2}`
2442        `hand <- "HAND"` $\C{// sentinel value}$
2443        `<- shake` $\C{// wait for completion}\CRT$
2444}
2445\end{Go}
2446\end{lrbox}
2447
2448\subfloat[\CFA]{\label{f:CFAwaitfor}\usebox\myboxA}
2449\hspace{3pt}
2450\vrule
2451\hspace{3pt}
2452\subfloat[Go]{\label{f:Gochannel}\usebox\myboxB}
2453\caption{Direct communication}
2454\label{f:DirectCommunication}
2455\end{figure}
2456
2457\begin{comment}
2458The following shows an example of two threads directly calling each other and accepting calls from each other in a cycle.
2459\begin{cfa}
2460\end{cfa}
2461\vspace{-0.8\baselineskip}
2462\begin{cquote}
2463\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
2464\begin{cfa}
2465thread Ping {} pi;
2466void ping( Ping & mutex ) {}
2467void main( Ping & pi ) {
2468        for ( 10 ) {
2469                `waitfor( ping, pi );`
2470                `pong( po );`
2471        }
2472}
2473int main() {}
2474\end{cfa}
2475&
2476\begin{cfa}
2477thread Pong {} po;
2478void pong( Pong & mutex ) {}
2479void main( Pong & po ) {
2480        for ( 10 ) {
2481                `ping( pi );`
2482                `waitfor( pong, po );`
2483        }
2484}
2485
2486\end{cfa}
2487\end{tabular}
2488\end{cquote}
2489% \lstMakeShortInline@%
2490% \caption{Threads ping/pong using external scheduling}
2491% \label{f:pingpong}
2492% \end{figure}
2493Note, the ping/pong threads are globally declared, @pi@/@po@, and hence, start (and possibly complete) before the program main starts.
2494\end{comment}
2495
2496
2497\subsection{Execution Properties}
2498
2499Table~\ref{t:ObjectPropertyComposition} shows how the \CFA high-level constructs cover 3 fundamental execution properties: thread, stateful function, and mutual exclusion.
2500Case 1 is a basic object, with none of the new execution properties.
2501Case 2 allows @mutex@ calls to Case 1 to protect shared data.
2502Case 3 allows stateful functions to suspend/resume but restricts operations because the state is stackless.
2503Case 4 allows @mutex@ calls to Case 3 to protect shared data.
2504Cases 5 and 6 are the same as 3 and 4 without restriction because the state is stackful.
2505Cases 7 and 8 are rejected because a thread cannot execute without a stackful state in a preemptive environment when context switching from the signal handler.
2506Cases 9 and 10 have a stackful thread without and with @mutex@ calls.
2507For situations where threads do not require direct communication, case 9 provides faster creation/destruction by eliminating @mutex@ setup.
2508
2509\begin{table}
2510\caption{Object property composition}
2511\centering
2512\label{t:ObjectPropertyComposition}
2513\renewcommand{\arraystretch}{1.25}
2514%\setlength{\tabcolsep}{5pt}
2515\begin{tabular}{c|c||l|l}
2516\multicolumn{2}{c||}{object properties} & \multicolumn{2}{c}{mutual exclusion} \\
2517\hline
2518thread  & stateful                              & \multicolumn{1}{c|}{No} & \multicolumn{1}{c}{Yes} \\
2519\hline
2520\hline
2521No              & No                                    & \textbf{1}\ \ \ aggregate type                & \textbf{2}\ \ \ @monitor@ aggregate type \\
2522\hline
2523No              & Yes (stackless)               & \textbf{3}\ \ \ @generator@                   & \textbf{4}\ \ \ @monitor@ @generator@ \\
2524\hline
2525No              & Yes (stackful)                & \textbf{5}\ \ \ @coroutine@                   & \textbf{6}\ \ \ @monitor@ @coroutine@ \\
2526\hline
2527Yes             & No / Yes (stackless)  & \textbf{7}\ \ \ {\color{red}rejected} & \textbf{8}\ \ \ {\color{red}rejected} \\
2528\hline
2529Yes             & Yes (stackful)                & \textbf{9}\ \ \ @thread@                              & \textbf{10}\ \ @monitor@ @thread@ \\
2530\end{tabular}
2531\end{table}
2532
2533
2534\subsection{Low-level Locks}
2535
2536For 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.
2537Some of these low-level mechanism are used in the \CFA runtime, but we strongly advocate using high-level mechanisms whenever possible.
2538
2539
2540% \section{Parallelism}
2541% \label{s:Parallelism}
2542%
2543% Historically, computer performance was about processor speeds.
2544% However, with heat dissipation being a direct consequence of speed increase, parallelism is the new source for increased performance~\cite{Sutter05, Sutter05b}.
2545% Therefore, high-performance applications must care about parallelism, which requires concurrency.
2546% The lowest-level approach of parallelism is to use \newterm{kernel threads} in combination with semantics like @fork@, @join@, \etc.
2547% However, kernel threads are better as an implementation tool because of complexity and higher cost.
2548% Therefore, different abstractions are often layered onto kernel threads to simplify them, \eg pthreads.
2549%
2550%
2551% \subsection{User Threads}
2552%
2553% A direct improvement on kernel threads is user threads, \eg Erlang~\cite{Erlang} and \uC~\cite{uC++book}.
2554% This approach provides an interface that matches the language paradigms, gives more control over concurrency by the language runtime, and an abstract (and portable) interface to the underlying kernel threads across operating systems.
2555% In many cases, user threads can be used on a much larger scale (100,000 threads).
2556% Like kernel threads, user threads support preemption, which maximizes nondeterminism, but increases the potential for concurrency errors: race, livelock, starvation, and deadlock.
2557% \CFA adopts user-threads to provide more flexibility and a low-cost mechanism to build any other concurrency approach, \eg thread pools and actors~\cite{Actors}.
2558%
2559% A variant of user thread is \newterm{fibres}, which removes preemption, \eg Go~\cite{Go} @goroutine@s.
2560% Like functional programming, which removes mutation and its associated problems, removing preemption from concurrency reduces nondeterminism, making race and deadlock errors more difficult to generate.
2561% However, preemption is necessary for fairness and to reduce tail-latency.
2562% For concurrency that relies on spinning, if all cores spin the system is livelocked, whereas preemption breaks the livelock.
2563
2564
2565\begin{comment}
2566\subsection{Thread Pools}
2567
2568In contrast to direct threading is indirect \newterm{thread pools}, \eg Java @executor@, where small jobs (work units) are inserted into a work pool for execution.
2569If the jobs are dependent, \ie interact, there is an implicit/explicit dependency graph that ties them together.
2570While removing direct concurrency, and hence the amount of context switching, thread pools significantly limit the interaction that can occur among jobs.
2571Indeed, jobs should not block because that also blocks the underlying thread, which effectively means the CPU utilization, and therefore throughput, suffers.
2572While 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.
2573As well, concurrency errors return, which threads pools are suppose to mitigate.
2574
2575\begin{figure}
2576\centering
2577\begin{tabular}{@{}l|l@{}}
2578\begin{cfa}
2579struct Adder {
2580    int * row, cols;
2581};
2582int operator()() {
2583        subtotal = 0;
2584        for ( int c = 0; c < cols; c += 1 )
2585                subtotal += row[c];
2586        return subtotal;
2587}
2588void ?{}( Adder * adder, int row[$\,$], int cols, int & subtotal ) {
2589        adder.[rows, cols, subtotal] = [rows, cols, subtotal];
2590}
2591
2592
2593
2594
2595\end{cfa}
2596&
2597\begin{cfa}
2598int main() {
2599        const int rows = 10, cols = 10;
2600        int matrix[rows][cols], subtotals[rows], total = 0;
2601        // read matrix
2602        Executor executor( 4 ); // kernel threads
2603        Adder * adders[rows];
2604        for ( r; rows ) { // send off work for executor
2605                adders[r] = new( matrix[r], cols, &subtotal[r] );
2606                executor.send( *adders[r] );
2607        }
2608        for ( r; rows ) {       // wait for results
2609                delete( adders[r] );
2610                total += subtotals[r];
2611        }
2612        sout | total;
2613}
2614\end{cfa}
2615\end{tabular}
2616\caption{Executor}
2617\end{figure}
2618\end{comment}
2619
2620
2621\section{Runtime Structure}
2622\label{s:CFARuntimeStructure}
2623
2624Figure~\ref{f:RunTimeStructure} illustrates the runtime structure of a \CFA program.
2625In 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.
2626An executing thread is illustrated by its containment in a processor.
2627
2628\begin{figure}
2629\centering
2630\input{RunTimeStructure}
2631\caption{\CFA Runtime structure}
2632\label{f:RunTimeStructure}
2633\end{figure}
2634
2635
2636\subsection{Cluster}
2637\label{s:RuntimeStructureCluster}
2638
2639A \newterm{cluster} is a collection of threads and virtual processors (abstract kernel-thread) that execute the (user) threads from its own ready queue (like an OS executing kernel threads).
2640The purpose of a cluster is to control the amount of parallelism that is possible among threads, plus scheduling and other execution defaults.
2641The default cluster-scheduler is single-queue multi-server, which provides automatic load-balancing of threads on processors.
2642However, the design allows changing the scheduler, \eg multi-queue multi-server with work-stealing/sharing across the virtual processors.
2643If several clusters exist, both threads and virtual processors, can be explicitly migrated from one cluster to another.
2644No automatic load balancing among clusters is performed by \CFA.
2645
2646When 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.
2647The user cluster is created to contain the application user-threads.
2648Having all threads execute on the one cluster often maximizes utilization of processors, which minimizes runtime.
2649However, because of limitations of scheduling requirements (real-time), NUMA architecture, heterogeneous hardware, or issues with the underlying operating system, multiple clusters are sometimes necessary.
2650
2651
2652\subsection{Virtual Processor}
2653\label{s:RuntimeStructureProcessor}
2654
2655A virtual processor is implemented by a kernel thread (\eg UNIX process), which are scheduled for execution on a hardware processor by the underlying operating system.
2656Programs may use more virtual processors than hardware processors.
2657On a multiprocessor, kernel threads are distributed across the hardware processors resulting in virtual processors executing in parallel.
2658(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.)
2659The \CFA runtime attempts to block unused processors and unblock processors as the system load increases;
2660balancing the workload with processors is difficult because it requires future knowledge, \ie what will the applicaton workload do next.
2661Preemption occurs on virtual processors rather than user threads, via operating-system interrupts.
2662Thus virtual processors execute user threads, where preemption frequency applies to a virtual processor, so preemption occurs randomly across the executed user threads.
2663Turning off preemption transforms user threads into fibres.
2664
2665
2666\begin{comment}
2667\section{Implementation}
2668\label{s:Implementation}
2669
2670A primary implementation challenge is avoiding contention from dynamically allocating memory because of bulk acquire, \eg the internal-scheduling design is (almost) free of allocations.
2671All 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.
2672Furthermore, several bulk-acquire operations need a variable amount of memory.
2673This 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.
2674
2675In \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.
2676When a mutex call is made, pointers to the concerned monitors are aggregated into a variable-length array and sorted.
2677This array persists for the entire duration of the mutual exclusion and is used extensively for synchronization operations.
2678
2679To improve performance and simplicity, context switching occurs inside a function call, so only callee-saved registers are copied onto the stack and then the stack register is switched;
2680the corresponding registers are then restored for the other context.
2681Note, the instruction pointer is untouched since the context switch is always inside the same function.
2682Experimental results (not presented) for a stackless or stackful scheduler (1 versus 2 context switches) (see Section~\ref{s:Concurrency}) show the performance is virtually equivalent, because both approaches are dominated by locking to prevent a race condition.
2683
2684All kernel threads (@pthreads@) created a stack.
2685Each \CFA virtual processor is implemented as a coroutine and these coroutines run directly on the kernel-thread stack, effectively stealing this stack.
2686The exception to this rule is the program main, \ie the initial kernel thread that is given to any program.
2687In 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.
2688\end{comment}
2689
2690
2691\subsection{Preemption}
2692
2693Nondeterministic preemption provides fairness from long-running threads, and forces concurrent programmers to write more robust programs, rather than relying on code between cooperative scheduling to be atomic.
2694This atomic reliance can fail on multi-core machines, because execution across cores is nondeterministic.
2695A different reason for not supporting preemption is that it significantly complicates the runtime system, \eg Microsoft runtime does not support interrupts and on Linux systems, interrupts are complex (see below).
2696Preemption is normally handled by setting a countdown timer on each virtual processor.
2697When the timer expires, an interrupt is delivered, and the interrupt handler resets the countdown 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.
2698Multiple signal handlers may be pending.
2699When 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.
2700The 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;
2701therefore, the same signal mask is required for all virtual processors in a cluster.
2702Because preemption frequency is usually long (1 millisecond) performance cost is negligible.
2703
2704Linux switched a decade ago from specific to arbitrary process signal-delivery for applications with multiple kernel threads.
2705\begin{cquote}
2706A process-directed signal may be delivered to any one of the threads that does not currently have the signal blocked.
2707If more than one of the threads has the signal unblocked, then the kernel chooses an arbitrary thread to which it will deliver the signal.
2708SIGNAL(7) - Linux Programmer's Manual
2709\end{cquote}
2710Hence, the timer-expiry signal, which is generated \emph{externally} by the Linux kernel to an application, is delivered to any of its Linux subprocesses (kernel threads).
2711To ensure each virtual processor receives a preemption signal, a discrete-event simulation is run on a special virtual processor, and only it sets and receives timer events.
2712Virtual processors register an expiration time with the discrete-event simulator, which is inserted in sorted order.
2713The simulation sets the countdown 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.
2714Processing a preemption event sends an \emph{internal} @SIGUSR1@ signal to the registered virtual processor, which is always delivered to that processor.
2715
2716
2717\subsection{Debug Kernel}
2718
2719There are two versions of the \CFA runtime kernel: debug and non-debug.
2720The 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.
2721After a program is debugged, the non-debugging version can be used to significantly decrease space and increase performance.
2722
2723
2724\section{Performance}
2725\label{s:Performance}
2726
2727To verify the implementation of the \CFA runtime, a series of microbenchmarks are performed comparing \CFA with pthreads, Java OpenJDK-9, Go 1.12.6 and \uC 7.0.0.
2728For comparison, the package must be multi-processor (M:N), which excludes libdill/libmil~\cite{libdill} (M:1)), and use a shared-memory programming model, \eg not message passing.
2729The benchmark computer is an AMD Opteron\texttrademark\ 6380 NUMA 64-core, 8 socket, 2.5 GHz processor, running Ubuntu 16.04.6 LTS, and \CFA/\uC are compiled with gcc 6.5.
2730
2731All benchmarks are run using the following harness. (The Java harness is augmented to circumvent JIT issues.)
2732\begin{cfa}
2733unsigned int N = 10_000_000;
2734#define BENCH( `run` ) Time before = getTimeNsec();  `run;`  Duration result = (getTimeNsec() - before) / N;
2735\end{cfa}
2736The method used to get time is @clock_gettime( CLOCK_REALTIME )@.
2737Each benchmark is performed @N@ times, where @N@ varies depending on the benchmark;
2738the total time is divided by @N@ to obtain the average time for a benchmark.
2739Each benchmark experiment is run 31 times.
2740All omitted tests for other languages are functionally identical to the \CFA tests and available online~\cite{CforallBenchMarks}.
2741% tar --exclude=.deps --exclude=Makefile --exclude=Makefile.in --exclude=c.c --exclude=cxx.cpp --exclude=fetch_add.c -cvhf benchmark.tar benchmark
2742
2743\paragraph{Object Creation}
2744
2745Object creation is measured by creating/deleting the specific kind of concurrent object.
2746Figure~\ref{f:creation} shows the code for \CFA, with results in Table~\ref{tab:creation}.
2747The 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.
2748
2749\begin{multicols}{2}
2750\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2751\begin{cfa}
2752@thread@ MyThread {};
2753void @main@( MyThread & ) {}
2754int main() {
2755        BENCH( for ( N ) { @MyThread m;@ } )
2756        sout | result`ns;
2757}
2758\end{cfa}
2759\captionof{figure}{\CFA object-creation benchmark}
2760\label{f:creation}
2761
2762\columnbreak
2763
2764\vspace*{-16pt}
2765\captionof{table}{Object creation comparison (nanoseconds)}
2766\label{tab:creation}
2767
2768\begin{tabular}[t]{@{}r*{3}{D{.}{.}{5.2}}@{}}
2769\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} & \multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
2770\CFA Coroutine Lazy             & 13.2          & 13.1          & 0.44          \\
2771\CFA Coroutine Eager    & 531.3         & 536.0         & 26.54         \\
2772\CFA Thread                             & 2074.9        & 2066.5        & 170.76        \\
2773\uC Coroutine                   & 89.6          & 90.5          & 1.83          \\
2774\uC Thread                              & 528.2         & 528.5         & 4.94          \\
2775Goroutine                               & 4068.0        & 4113.1        & 414.55        \\
2776Java Thread                             & 103848.5      & 104295.4      & 2637.57       \\
2777Pthreads                                & 33112.6       & 33127.1       & 165.90
2778\end{tabular}
2779\end{multicols}
2780
2781
2782\paragraph{Context-Switching}
2783
2784In procedural programming, the cost of a function call is important as modularization (refactoring) increases.
2785(In many cases, a compiler inlines function calls to eliminate this cost.)
2786Similarly, when modularization extends to coroutines/tasks, the time for a context switch becomes a relevant factor.
2787The coroutine test is from resumer to suspender and from suspender to resumer, which is two context switches.
2788The thread test is using yield to enter and return from the runtime kernel, which is two context switches.
2789The difference in performance between coroutine and thread context-switch is the cost of scheduling for threads, whereas coroutines are self-scheduling.
2790Figure~\ref{f:ctx-switch} only shows the \CFA code for coroutines/threads (other systems are similar) with all results in Table~\ref{tab:ctx-switch}.
2791
2792\begin{multicols}{2}
2793\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2794\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2795@coroutine@ C {} c;
2796void main( C & ) { for ( ;; ) { @suspend;@ } }
2797int main() { // coroutine test
2798        BENCH( for ( N ) { @resume( c );@ } )
2799        sout | result`ns;
2800}
2801int main() { // task test
2802        BENCH( for ( N ) { @yield();@ } )
2803        sout | result`ns;
2804}
2805\end{cfa}
2806\captionof{figure}{\CFA context-switch benchmark}
2807\label{f:ctx-switch}
2808
2809\columnbreak
2810
2811\vspace*{-16pt}
2812\captionof{table}{Context switch comparison (nanoseconds)}
2813\label{tab:ctx-switch}
2814\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
2815\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
2816C function              & 1.8   & 1.8   & 0.01  \\
2817\CFA generator  & 2.4   & 2.2   & 0.25  \\
2818\CFA Coroutine  & 36.2  & 36.2  & 0.25  \\
2819\CFA Thread             & 93.2  & 93.5  & 2.09  \\
2820\uC Coroutine   & 52.0  & 52.1  & 0.51  \\
2821\uC Thread              & 96.2  & 96.3  & 0.58  \\
2822Goroutine               & 141.0 & 141.3 & 3.39  \\
2823Java Thread             & 374.0 & 375.8 & 10.38 \\
2824Pthreads Thread & 361.0 & 365.3 & 13.19
2825\end{tabular}
2826\end{multicols}
2827
2828
2829\paragraph{Mutual-Exclusion}
2830
2831Uncontented mutual exclusion, which frequently occurs, is measured by entering/leaving a critical section.
2832For monitors, entering and leaving a monitor function is measured.
2833To put the results in context, the cost of entering a non-inline function and the cost of acquiring and releasing a @pthread_mutex@ lock is also measured.
2834Figure~\ref{f:mutex} shows the code for \CFA with all results in Table~\ref{tab:mutex}.
2835Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2836
2837\begin{multicols}{2}
2838\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2839\begin{cfa}
2840@monitor@ M {} m1/*, m2, m3, m4*/;
2841void __attribute__((noinline))
2842do_call( M & @mutex m/*, m2, m3, m4*/@ ) {}
2843int main() {
2844        BENCH(
2845                for( N ) do_call( m1/*, m2, m3, m4*/ );
2846        )
2847        sout | result`ns;
2848}
2849\end{cfa}
2850\captionof{figure}{\CFA acquire/release mutex benchmark}
2851\label{f:mutex}
2852
2853\columnbreak
2854
2855\vspace*{-16pt}
2856\captionof{table}{Mutex comparison (nanoseconds)}
2857\label{tab:mutex}
2858\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
2859\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
2860test and test-and-test lock             & 19.1  & 18.9  & 0.40  \\
2861\CFA @mutex@ function, 1 arg.   & 45.9  & 46.6  & 1.45  \\
2862\CFA @mutex@ function, 2 arg.   & 105.0 & 104.7 & 3.08  \\
2863\CFA @mutex@ function, 4 arg.   & 165.0 & 167.6 & 5.65  \\
2864\uC @monitor@ member rtn.               & 54.0  & 53.7  & 0.82  \\
2865Java synchronized method                & 31.0  & 31.1  & 0.50  \\
2866Pthreads Mutex Lock                             & 33.6  & 32.6  & 1.14
2867\end{tabular}
2868\end{multicols}
2869
2870
2871\paragraph{External Scheduling}
2872
2873External scheduling is measured using a cycle of two threads calling and accepting the call using the @waitfor@ statement.
2874Figure~\ref{f:ext-sched} shows the code for \CFA, with results in Table~\ref{tab:ext-sched}.
2875Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2876
2877\begin{multicols}{2}
2878\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2879\vspace*{-16pt}
2880\begin{cfa}
2881volatile int go = 0;
2882@monitor@ M {} m;
2883thread T {};
2884void __attribute__((noinline))
2885do_call( M & @mutex@ ) {}
2886void main( T & ) {
2887        while ( go == 0 ) { yield(); }
2888        while ( go == 1 ) { do_call( m ); }
2889}
2890int __attribute__((noinline))
2891do_wait( M & @mutex@ m ) {
2892        go = 1; // continue other thread
2893        BENCH( for ( N ) { @waitfor( do_call, m );@ } )
2894        go = 0; // stop other thread
2895        sout | result`ns;
2896}
2897int main() {
2898        T t;
2899        do_wait( m );
2900}
2901\end{cfa}
2902\captionof{figure}{\CFA external-scheduling benchmark}
2903\label{f:ext-sched}
2904
2905\columnbreak
2906
2907\vspace*{-16pt}
2908\captionof{table}{External-scheduling comparison (nanoseconds)}
2909\label{tab:ext-sched}
2910\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
2911\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
2912\CFA @waitfor@, 1 @monitor@     & 376.4 & 376.8 & 7.63  \\
2913\CFA @waitfor@, 2 @monitor@     & 491.4 & 492.0 & 13.31 \\
2914\CFA @waitfor@, 4 @monitor@     & 681.0 & 681.7 & 19.10 \\
2915\uC @_Accept@                           & 331.1 & 331.4 & 2.66
2916\end{tabular}
2917\end{multicols}
2918
2919
2920\paragraph{Internal Scheduling}
2921
2922Internal scheduling is measured using a cycle of two threads signalling and waiting.
2923Figure~\ref{f:int-sched} shows the code for \CFA, with results in Table~\ref{tab:int-sched}.
2924Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
2925Java 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.
2926
2927\begin{multicols}{2}
2928\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2929\begin{cfa}
2930volatile int go = 0;
2931@monitor@ M { @condition c;@ } m;
2932void __attribute__((noinline))
2933do_call( M & @mutex@ a1 ) { @signal( c );@ }
2934thread T {};
2935void main( T & this ) {
2936        while ( go == 0 ) { yield(); }
2937        while ( go == 1 ) { do_call( m ); }
2938}
2939int  __attribute__((noinline))
2940do_wait( M & mutex m ) with(m) {
2941        go = 1; // continue other thread
2942        BENCH( for ( N ) { @wait( c );@ } );
2943        go = 0; // stop other thread
2944        sout | result`ns;
2945}
2946int main() {
2947        T t;
2948        do_wait( m );
2949}
2950\end{cfa}
2951\captionof{figure}{\CFA Internal-scheduling benchmark}
2952\label{f:int-sched}
2953
2954\columnbreak
2955
2956\vspace*{-16pt}
2957\captionof{table}{Internal-scheduling comparison (nanoseconds)}
2958\label{tab:int-sched}
2959\bigskip
2960
2961\begin{tabular}{@{}r*{3}{D{.}{.}{5.2}}@{}}
2962\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} & \multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
2963\CFA @signal@, 1 @monitor@      & 372.6         & 374.3         & 14.17         \\
2964\CFA @signal@, 2 @monitor@      & 492.7         & 494.1         & 12.99         \\
2965\CFA @signal@, 4 @monitor@      & 749.4         & 750.4         & 24.74         \\
2966\uC @signal@                            & 320.5         & 321.0         & 3.36          \\
2967Java @notify@                           & 10160.5       & 10169.4       & 267.71        \\
2968Pthreads Cond. Variable         & 4949.6        & 5065.2        & 363
2969\end{tabular}
2970\end{multicols}
2971
2972
2973\section{Conclusion}
2974
2975Advanced control-flow will always be difficult, especially when there is temporal ordering and nondeterminism.
2976However, many systems exacerbate the difficulty through their presentation mechanisms.
2977This paper shows it is possible to present a hierarchy of control-flow features, generator, coroutine, thread, and monitor, providing an integrated set of high-level, efficient, and maintainable control-flow features.
2978Eliminated from \CFA are spurious wakeup and barging, which are nonintuitive and lead to errors, and having to work with a bewildering set of low-level locks and acquisition techniques.
2979\CFA high-level race-free monitors and tasks provide the core mechanisms for mutual exclusion and synchronization, without having to resort to magic qualifiers like @volatile@/@atomic@.
2980Extending these mechanisms to handle high-level deadlock-free bulk acquire across both mutual exclusion and synchronization is a unique contribution.
2981The \CFA runtime 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.
2982The M:N model is judged to be efficient and provide greater flexibility than a 1:1 threading model.
2983These concepts and the \CFA runtime-system are written in the \CFA language, extensively leveraging the \CFA type-system, which demonstrates the expressiveness of the \CFA language.
2984Performance 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.
2985C programmers should feel comfortable using these mechanisms for developing complex control-flow in applications, with the ability to obtain maximum available performance by selecting mechanisms at the appropriate level of need.
2986
2987
2988\section{Future Work}
2989
2990While control flow in \CFA has a strong start, development is still underway to complete a number of missing features.
2991
2992\paragraph{Flexible Scheduling}
2993\label{futur:sched}
2994
2995An important part of concurrency is scheduling.
2996Different scheduling algorithms can affect performance (both in terms of average and variation).
2997However, no single scheduler is optimal for all workloads and therefore there is value in being able to change the scheduler for given programs.
2998One solution is to offer various tuning options, allowing the scheduler to be adjusted to the requirements of the workload.
2999However, to be truly flexible, a pluggable scheduler is necessary.
3000Currently, 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~\cite{Buhr00b}.
3001
3002\paragraph{Non-Blocking I/O}
3003\label{futur:nbio}
3004
3005Many modern workloads are not bound by computation but IO operations, a common case being web servers and XaaS~\cite{XaaS} (anything as a service).
3006These types of workloads require significant engineering to amortizing costs of blocking IO-operations.
3007At 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.
3008Current 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.
3009However, these solutions lead to code that is hard to create, read and maintain.
3010A 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.
3011A non-blocking I/O library is currently under development for \CFA.
3012
3013\paragraph{Other Concurrency Tools}
3014\label{futur:tools}
3015
3016While monitors offer flexible and powerful concurrency for \CFA, other concurrency tools are also necessary for a complete multi-paradigm concurrency package.
3017Examples of such tools can include futures and promises~\cite{promises}, executors and actors.
3018These additional features are useful for applications that can be constructed without shared data and direct blocking.
3019As well, new \CFA extensions should make it possible to create a uniform interface for virtually all mutual exclusion, including monitors and low-level locks.
3020
3021\paragraph{Implicit Threading}
3022\label{futur:implcit}
3023
3024Basic 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.
3025This type of concurrency can be achieved both at the language level and at the library level.
3026The canonical example of implicit concurrency is concurrent nested @for@ loops, which are amenable to divide and conquer algorithms~\cite{uC++book}.
3027The \CFA language features should make it possible to develop a reasonable number of implicit concurrency mechanism to solve basic HPC data-concurrency problems.
3028However, implicit concurrency is a restrictive solution with significant limitations, so it can never replace explicit concurrent programming.
3029
3030
3031\section{Acknowledgements}
3032
3033The authors would like to recognize the design assistance of Aaron Moss, Rob Schluntz, Andrew Beach and Michael Brooks on the features described in this paper.
3034Funding 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.
3035
3036{%
3037\fontsize{9bp}{12bp}\selectfont%
3038\bibliography{pl,local}
3039}%
3040
3041\end{document}
3042
3043% Local Variables: %
3044% tab-width: 4 %
3045% fill-column: 120 %
3046% compile-command: "make" %
3047% End: %
Note: See TracBrowser for help on using the repository browser.