source: doc/papers/concurrency/Paper.tex @ 80dbf6a

arm-ehjacob/cs343-translationjenkins-sandboxnew-astnew-ast-unique-expr
Last change on this file since 80dbf6a was 80dbf6a, checked in by Peter A. Buhr <pabuhr@…>, 22 months ago

updated concurrency paper

  • Property mode set to 100644
File size: 175.0 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{.}{\ETAL}%
135                {\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, resume, __restrict, __restrict__, __signed, __signed__, _Static_assert, suspend, 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{0.5ex}{\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
201% Go programming language: https://github.com/julienc91/listings-golang/blob/master/listings-golang.sty
202\lstdefinelanguage{Golang}{
203        morekeywords=[1]{package,import,func,type,struct,return,defer,panic,recover,select,var,const,iota,},
204        morekeywords=[2]{string,uint,uint8,uint16,uint32,uint64,int,int8,int16,int32,int64,
205                bool,float32,float64,complex64,complex128,byte,rune,uintptr, error,interface},
206        morekeywords=[3]{map,slice,make,new,nil,len,cap,copy,close,true,false,delete,append,real,imag,complex,chan,},
207        morekeywords=[4]{for,break,continue,range,goto,switch,case,fallthrough,if,else,default,},
208        morekeywords=[5]{Println,Printf,Error,},
209        sensitive=true,
210        morecomment=[l]{//},
211        morecomment=[s]{/*}{*/},
212        morestring=[b]',
213        morestring=[b]",
214        morestring=[s]{`}{`},
215        % replace/adjust listing characters that look bad in sanserif
216        literate={-}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.1ex}}}}1 {^}{\raisebox{0.6ex}{$\scriptstyle\land\,$}}1
217                {~}{\raisebox{0.3ex}{$\scriptstyle\sim\,$}}1 % {`}{\ttfamily\upshape\hspace*{-0.1ex}`}1
218                {<}{\textrm{\textless}}1 {>}{\textrm{\textgreater}}1
219                {<-}{\makebox[2ex][c]{\textrm{\textless}\raisebox{0.5ex}{\rule{0.8ex}{0.075ex}}}}2,
220}
221
222\lstnewenvironment{cfa}[1][]
223{\lstset{#1}}
224{}
225\lstnewenvironment{C++}[1][]                            % use C++ style
226{\lstset{language=C++,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
227{}
228\lstnewenvironment{uC++}[1][]
229{\lstset{language=uC++,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
230{}
231\lstnewenvironment{Go}[1][]
232{\lstset{language=Golang,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
233{}
234\lstnewenvironment{python}[1][]
235{\lstset{language=python,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
236{}
237
238% inline code @...@
239\lstMakeShortInline@%
240
241\let\OLDthebibliography\thebibliography
242\renewcommand\thebibliography[1]{
243  \OLDthebibliography{#1}
244  \setlength{\parskip}{0pt}
245  \setlength{\itemsep}{4pt plus 0.3ex}
246}
247
248\newbox\myboxA
249\newbox\myboxB
250\newbox\myboxC
251\newbox\myboxD
252
253\title{\texorpdfstring{Advanced Control-flow and Concurrency in \protect\CFA}{Advanced Control-flow in Cforall}}
254
255\author[1]{Thierry Delisle}
256\author[1]{Peter A. Buhr*}
257\authormark{DELISLE \textsc{et al.}}
258
259\address[1]{\orgdiv{Cheriton School of Computer Science}, \orgname{University of Waterloo}, \orgaddress{\state{Waterloo, ON}, \country{Canada}}}
260
261\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}}
262
263% \fundingInfo{Natural Sciences and Engineering Research Council of Canada}
264
265\abstract[Summary]{
266\CFA is a polymorphic, non-object-oriented, concurrent, backwards-compatible extension of the C programming language.
267This paper discusses the design philosophy and implementation of its advanced control-flow and concurrent/parallel features, along with the supporting runtime written in \CFA.
268These features are created from scratch as ISO C has only low-level and/or unimplemented concurrency, so C programmers continue to rely on library approaches like pthreads.
269\CFA introduces modern language-level control-flow mechanisms, like generators, coroutines, user-level threading, and monitors for mutual exclusion and synchronization.
270% Library extension for executors, futures, and actors are built on these basic mechanisms.
271The runtime provides significant programmer simplification and safety by eliminating spurious wakeup and monitor barging.
272The runtime also ensures multiple monitors can be safely acquired \emph{simultaneously} (deadlock free), and this feature is fully integrated with all monitor synchronization mechanisms.
273All control-flow features integrate with the \CFA polymorphic type-system and exception handling, while respecting the expectations and style of C programmers.
274Experimental results show comparable performance of the new features with similar mechanisms in other concurrent programming languages.
275}%
276
277\keywords{generator, coroutine, concurrency, parallelism, thread, monitor, runtime, C, \CFA (Cforall)}
278
279
280\begin{document}
281\linenumbers                            % comment out to turn off line numbering
282
283\maketitle
284
285
286\section{Introduction}
287
288\CFA~\cite{Moss18,Cforall} is a modern, polymorphic, non-object-oriented\footnote{
289\CFA has object-oriented features, such as constructors, destructors, virtuals and simple trait/interface inheritance.
290% Go interfaces, Rust traits, Swift Protocols, Haskell Type Classes and Java Interfaces.
291% "Trait inheritance" works for me. "Interface inheritance" might also be a good choice, and distinguish clearly from implementation inheritance.
292% You'll want to be a little bit careful with terms like "structural" and "nominal" inheritance as well. CFA has structural inheritance (I think Go as well) -- it's inferred based on the structure of the code. Java, Rust, and Haskell (not sure about Swift) have nominal inheritance, where there needs to be a specific statement that "this type inherits from this type".
293However, 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.},
294backwards-compatible extension of the C programming language.
295In 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\footnote{
296The TIOBE index~\cite{TIOBE} for December 2019 ranks the top five \emph{popular} programming languages as Java 17\%, C 16\%, Python 10\%, and \CC 6\%, \Csharp 5\% = 54\%, and over the past 30 years, C has always ranked either first or second in popularity.}
297allowing immediate dissemination.
298This paper discusses the design philosophy and implementation of advanced language-level control-flow and concurrent/parallel features in \CFA and its runtime, which is written entirely in \CFA.
299The \CFA control-flow framework extends ISO \Celeven~\cite{C11} with new call/return and concurrent/parallel control-flow.
300
301% The call/return extensions retain state between callee and caller versus losing the callee's state on return;
302% the concurrency extensions allow high-level management of threads.
303
304Call/return control-flow with argument/parameter passing appeared in the first programming languages.
305Over the past 50 years, call/return has been augmented with features like static/dynamic call, exceptions (multi-level return) and generators/coroutines (retain state between calls).
306While \CFA has mechanisms for dynamic call (algebraic effects) and exceptions\footnote{
307\CFA exception handling will be presented in a separate paper.
308The 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++}}, this work only discusses retaining state between calls via generators/coroutines.
309\newterm{Coroutining} was introduced by Conway~\cite{Conway63} (1963), discussed by Knuth~\cite[\S~1.4.2]{Knuth73V1}, implemented in Simula67~\cite{Simula67}, formalized by Marlin~\cite{Marlin80}, and is now popular and appears in old and new programming languages: CLU~\cite{CLU}, \Csharp~\cite{Csharp}, Ruby~\cite{Ruby}, Python~\cite{Python}, JavaScript~\cite{JavaScript}, Lua~\cite{Lua}, \CCtwenty~\cite{C++20Coroutine19}.
310Coroutining is sequential execution requiring direct handoff among coroutines, \ie only the programmer is controlling execution order.
311If coroutines transfer to an internal event-engine for scheduling the next coroutines, the program transitions into the realm of concurrency~\cite[\S~3]{Buhr05a}.
312Coroutines are only a stepping stone towards concurrency where the commonality is that coroutines and threads retain state between calls.
313
314\Celeven/\CCeleven define concurrency~\cite[\S~7.26]{C11}, but it is largely wrappers for a subset of the pthreads library~\cite{Pthreads}.\footnote{Pthreads concurrency is based on simple thread fork/join in a function and mutex/condition locks, which is low-level and error-prone}
315Interestingly, almost a decade after the \Celeven standard, neither gcc-9, clang-9 nor msvc-19 (most recent versions) support the \Celeven include @threads.h@, indicating no interest in the C11 concurrency approach (possibly because of the recent effort to add concurrency to \CC).
316While 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}, as for \CC.
317In contrast, there has been a renewed interest during the past decade in user-level (M:N, green) threading in old and new programming languages.
318As multi-core hardware became available in the 1980/90s, both user and kernel threading were examined.
319Kernel 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}.
320Libraries like pthreads were developed for C, and the Solaris operating-system switched from user (JDK 1.1~\cite{JDK1.1}) to kernel threads.
321As a result, many current languages implementations adopt the 1:1 kernel-threading model, like Java (Scala), Objective-C~\cite{obj-c-book}, \CCeleven~\cite{C11}, C\#~\cite{Csharp} and Rust~\cite{Rust}, with a variety of presentation mechanisms.
322From 2000 onwards, several language implementations have championed the M:N user-threading model, like Go~\cite{Go}, Erlang~\cite{Erlang}, Haskell~\cite{Haskell}, D~\cite{D}, and \uC~\cite{uC++,uC++book}, including putting green threads back into Java~\cite{Quasar}, and many user-threading libraries have appeared~\cite{Qthreads,MPC,Marcel}.
323The 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 encourages large numbers of threads performing medium-sized work to facilitate load balancing by the runtime~\cite{Verch12}.
324As well, user-threading facilitates a simpler concurrency approach using thread objects that leverage sequential patterns versus events with call-backs~\cite{Adya02,vonBehren03}.
325Finally, 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.
326
327A 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, \eg some language features are unsafe in the presence of aggressive sequential optimizations~\cite{Buhr95a,Boehm05}.
328The 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.
329One solution is low-level qualifiers and functions (\eg @volatile@ and atomics) allowing \emph{programmers} to explicitly write safe (race-free~\cite{Boehm12}) programs.
330A safer solution is high-level language constructs so the \emph{compiler} knows the concurrency boundaries (where mutual exclusion and synchronization are acquired/released) and provide implicit safety at and across these boundaries.
331While the optimization problem is best known with respect to concurrency, it applies to other complex control-flow, like exceptions and coroutines.
332As well, language solutions allow matching the language paradigm with the approach, \eg matching the functional paradigm with data-flow programming or the imperative paradigm with thread programming.
333
334Finally, it is important for a language to provide safety over performance \emph{as the default}, allowing careful reduction of safety (unsafe code) for performance when necessary.
335Two concurrency violations of this philosophy are \emph{spurious wakeup} (random wakeup~\cite[\S~9]{Buhr05a}) and \emph{barging}\footnote{
336Barging is competitive succession instead of direct handoff, \ie after a lock is released both arriving and preexisting waiter threads compete to acquire the lock.
337Hence, an arriving thread can temporally \emph{barge} ahead of threads already waiting for an event, which can repeat indefinitely leading to starvation of waiter threads.
338} (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.
339(Author experience teaching concurrency is that students are confused by these semantics.)
340However, spurious wakeup is \emph{not} a foundational concurrency property~\cite[\S~9]{Buhr05a};
341it is a performance design choice.
342We argue removing spurious wakeup and signals-as-hints make concurrent programming simpler and safer as there is less local non-determinism to manage.
343If barging acquisition is allowed, its specialized performance advantage should be available as an option not the default.
344
345\CFA embraces language extensions for advanced control-flow, user-level threading, and safety as the default.
346We present comparative examples to support our argument that the \CFA control-flow extensions are as expressive and safe as those in other concurrent imperative programming languages, and perform experiments to show the \CFA runtime is competitive with other similar mechanisms.
347The main contributions of this work are:
348\begin{itemize}[topsep=3pt,itemsep=0pt]
349\item
350a set of fundamental execution properties that dictate which language-level control-flow features need to be supported,
351
352\item
353integration of these language-level control-flow features, while respecting the style and expectations of C programmers,
354
355\item
356monitor 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,
357
358\item
359providing statically type-safe interfaces that integrate with the \CFA polymorphic type-system and other language features,
360
361% \item
362% library extensions for executors, futures, and actors built on the basic mechanisms.
363
364\item
365a runtime system without spurious wake-up and no performance loss,
366
367\item
368a dynamic partitioning mechanism to segregate groups of executing user and kernel threads performing specialized work (\eg web-server or compute engine) or requiring different scheduling (\eg NUMA or real-time).
369
370% \item
371% a non-blocking I/O library
372
373\item
374experimental results showing comparable performance of the \CFA features with similar mechanisms in other languages.
375\end{itemize}
376
377Section~\ref{s:FundamentalExecutionProperties} presents the compositional hierarchy of execution properties directing the design of control-flow features in \CFA.
378Section~\ref{s:StatefulFunction} begins advanced control by introducing sequential functions that retain data and execution state between calls producing constructs @generator@ and @coroutine@.
379Section~\ref{s:Concurrency} begins concurrency, or how to create (fork) and destroy (join) a thread producing the @thread@ construct.
380Section~\ref{s:MutualExclusionSynchronization} discusses the two mechanisms to restricted nondeterminism when controlling shared access to resources (mutual exclusion) and timing relationships among threads (synchronization).
381Section~\ref{s:Monitor} shows how both mutual exclusion and synchronization are safely embedded in the @monitor@ and @thread@ constructs.
382Section~\ref{s:CFARuntimeStructure} describes the large-scale mechanism to structure (cluster) threads and virtual processors (kernel threads).
383Section~\ref{s:Performance} uses a series of microbenchmarks to compare \CFA threading with pthreads, Java 11.0.6, Go 1.12.6, Rust 1.37.0, Python 3.7.6, Node.js 12.14.1, and \uC 7.0.0.
384
385
386\section{Fundamental Execution Properties}
387\label{s:FundamentalExecutionProperties}
388
389The features in a programming language should be composed from a set of fundamental properties rather than an ad hoc collection chosen by the designers.
390To this end, the control-flow features created for \CFA are based on the fundamental properties of any language with function-stack control-flow (see also \uC~\cite[pp.~140-142]{uC++}).
391The fundamental properties are execution state, thread, and mutual-exclusion/synchronization (MES).
392These independent properties can be used alone, in pairs, or in triplets to compose different language features, forming a compositional hierarchy where the most advanced feature has all the properties (state/thread/MES).
393While it is possible for a language to only support the most advanced feature~\cite{Hermes90}, this unnecessarily complicates and makes inefficient solutions to certain classes of problems.
394As is shown, each of the (non-rejected) composed features solves a particular set of problems, and hence, has a defensible position in a programming language.
395If a compositional feature is missing, a programmer has too few/many fundamental properties resulting in a complex and/or is inefficient solution.
396
397In detail, the fundamental properties are:
398\begin{description}[leftmargin=\parindent,topsep=3pt,parsep=0pt]
399\item[\newterm{execution state}:]
400is the state information needed by a control-flow feature to initialize, manage compute data and execution location(s), and de-initialize.
401State is retained in fixed-sized aggregate structures and dynamic-sized stack(s), often allocated in the heap(s) managed by the runtime system.
402The lifetime of the state varies with the control-flow feature, where longer life-time and dynamic size provide greater power but also increase usage complexity and cost.
403Control-flow transfers among execution states occurs in multiple ways, such as function call, context switch, asynchronous await, etc.
404Because the programming language determines what constitutes an execution state, implicitly manages this state, and defines movement mechanisms among states, execution state is an elementary property of the semantics of a programming language.
405% An execution-state is related to the notion of a process continuation \cite{Hieb90}.
406
407\item[\newterm{threading}:]
408is execution of code that occurs independently of other execution, \ie the execution resulting from a thread is sequential.
409Multiple threads provide \emph{concurrent execution};
410concurrent execution becomes parallel when run on multiple processing units (hyper-threading, cores, sockets).
411There must be language mechanisms to create, block/unblock, and join with a thread.
412
413\item[\newterm{MES}:]
414is the concurrency mechanisms to perform an action without interruption and establish timing relationships among multiple threads.
415These two properties are independent, \ie mutual exclusion cannot provide synchronization and vice versa without introducing additional threads~\cite[\S~4]{Buhr05a}.
416Limiting MES, \eg no access to shared data, results in contrived solutions and inefficiency on multi-core von Neumann computers where shared memory is a foundational aspect of its design.
417\end{description}
418These properties are fundamental because they cannot be built from existing language features, \eg a basic programming language like C99~\cite{C99} cannot create new control-flow features, concurrency, or provide MES using atomic hardware mechanisms.
419
420
421\subsection{Execution Properties}
422
423Table~\ref{t:ExecutionPropertyComposition} shows how the three fundamental execution properties: state, thread, and mutual exclusion compose a hierarchy of control-flow features needed in a programming language.
424(When doing case analysis, not all combinations are meaningful.)
425Note, basic von Neumann execution requires at least one thread and an execution state providing some form of call stack.
426For table entries missing these minimal components, the property is borrowed from the invoker (caller).
427
428Case 1 is a function that borrows storage for its state (stack frame/activation) and a thread from its invoker and retains this state across \emph{callees}, \ie function local-variables are retained on the stack across calls.
429Case 2 is case 1 with access to shared state so callers are restricted during update (mutual exclusion) and scheduling for other threads (synchronization).
430Case 3 is a stateful function supporting resume/suspend along with call/return to retain state across \emph{callers}, but has some restrictions because the function's state is stackless.
431Note, stackless functions still borrow the caller's stack and thread, where the stack is used to preserve state across its callees.
432Case 4 is cases 2 and 3 with protection to shared state for stackless functions.
433Cases 5 and 6 are the same as 3 and 4 but only the thread is borrowed as the function state is stackful, so resume/suspend is a context switch from the caller's to the function's stack.
434Cases 7 and 8 are rejected because a function that is given a new thread must have its own stack where the thread begins and stack frames are stored for calls, \ie there is no stack to borrow.
435Cases 9 and 10 are rejected because a thread with a fixed state (no stack) cannot accept calls, make calls, block, or be preempted, all of which require an unknown amount of additional dynamic state.
436Hence, once started, this kind of thread must execute to completion, \ie computation only, which severely restricts runtime management.
437Cases 11 and 12 have a stackful thread with and without safe access to shared state.
438Execution properties increase the cost of creation and execution along with complexity of usage.
439
440\begin{table}
441\caption{Execution property composition}
442\centering
443\label{t:ExecutionPropertyComposition}
444\renewcommand{\arraystretch}{1.25}
445%\setlength{\tabcolsep}{5pt}
446\begin{tabular}{c|c||l|l}
447\multicolumn{2}{c||}{execution properties} & \multicolumn{2}{c}{mutual exclusion / synchronization} \\
448\hline
449stateful                        & thread        & \multicolumn{1}{c|}{No} & \multicolumn{1}{c}{Yes} \\
450\hline   
451\hline   
452No                                      & No            & \textbf{1}\ \ \ function                              & \textbf{2}\ \ \ @monitor@ function    \\
453\hline   
454Yes (stackless)         & No            & \textbf{3}\ \ \ @generator@                   & \textbf{4}\ \ \ @monitor@ @generator@ \\
455\hline   
456Yes (stackful)          & No            & \textbf{5}\ \ \ @coroutine@                   & \textbf{6}\ \ \ @monitor@ @coroutine@ \\
457\hline   
458No                                      & Yes           & \textbf{7}\ \ \ {\color{red}rejected} & \textbf{8}\ \ \ {\color{red}rejected} \\
459\hline   
460Yes (stackless)         & Yes           & \textbf{9}\ \ \ {\color{red}rejected} & \textbf{10}\ \ \ {\color{red}rejected} \\
461\hline   
462Yes (stackful)          & Yes           & \textbf{11}\ \ \ @thread@                             & \textbf{12}\ \ @monitor@ @thread@             \\
463\end{tabular}
464\end{table}
465
466Given the execution-properties taxonomy, programmers can now answer three basic questions: is state necessary across calls and how much, is a separate thread necessary, is access to shared state necessary.
467The answers define the optimal language feature need for implementing a programming problem.
468The next sections discusses how \CFA fills in the table with language features, while other programming languages may only provide a subset of the table.
469
470
471\subsection{Design Requirements}
472
473The following design requirements largely stem from building \CFA on top of C.
474\begin{itemize}[topsep=3pt,parsep=0pt]
475\item
476All communication must be statically type checkable for early detection of errors and efficient code generation.
477This requirement is consistent with the fact that C is a statically-typed programming-language.
478
479\item
480Direct interaction among language features must be possible allowing any feature to be selected without restricting comm\-unication.
481For example, many concurrent languages do not provide direct communication (calls) among threads, \ie threads only communicate indirectly through monitors, channels, messages, and/or futures.
482Indirect communication increases the number of objects, consuming more resources, and require additional synchronization and possibly data transfer.
483
484\item
485All communication is performed using function calls, \ie data is transmitted from argument to parameter and results are returned from function calls.
486Alternative forms of communication, such as call-backs, message passing, channels, or communication ports, step outside of C's normal form of communication.
487
488\item
489All stateful features must follow the same declaration scopes and lifetimes as other language data.
490For C that means at program startup, during block and function activation, and on demand using dynamic allocation.
491
492\item
493MES must be available implicitly in language constructs as well as explicitly for specialized requirements, because requiring programmers to build MES using low-level locks often leads to incorrect programs.
494Furthermore, reducing synchronization scope by encapsulating it within language constructs further reduces errors in concurrent programs.
495
496\item
497Both synchronous and asynchronous communication are needed.
498However, we believe the best way to provide asynchrony, such as call-buffering/chaining and/or returning futures~\cite{multilisp}, is building it from expressive synchronous features.
499
500\item
501Synchronization must be able to control the service order of requests including prioritizing selection from different kinds of outstanding requests, and postponing a request for an unspecified time while continuing to accept new requests.
502Otherwise, certain concurrency problems are difficult, e.g.\ web server, disk scheduling, and the amount of concurrency is inhibited~\cite{Gentleman81}.
503\end{itemize}
504We have satisfied these requirements in \CFA while maintaining backwards compatibility with the huge body of legacy C programs.
505% In contrast, other new programming languages must still access C programs (\eg operating-system service routines), but do so through fragile C interfaces.
506
507
508\subsection{Asynchronous Await / Call}
509
510Asynchronous await/call is a caller mechanism for structuring programs and/or increasing concurrency, where the caller (client) postpones an action into the future, which is subsequently executed by a callee (server).
511The caller detects the action's completion through a \newterm{future}/\newterm{promise}.
512The benefit is asynchronous caller execution with respect to the callee until future resolution.
513For single-threaded languages like JavaScript, an asynchronous call passes a callee action, which is queued in the event-engine, and continues execution with a promise.
514When the caller needs the promise to be fulfilled, it executes @await@.
515A promise-completion call-back can be part of the callee action or the caller is rescheduled;
516in either case, the call back is executed after the promise is fulfilled.
517While asynchronous calls generate new callee (server) events, we content this mechanism is insufficient for advanced control-flow mechanisms like generators or coroutines (which are discussed next).
518Specifically, control between caller and callee occurs indirectly through the event-engine precluding direct handoff and cycling among events, and requires complex resolution of a control promise and data.
519Note, @async-await@ is just syntactic-sugar over the event engine so it does not solve these deficiencies.
520For multi-threaded languages like Java, the asynchronous call queues a callee action with an executor (server), which subsequently executes the work by a thread in the executor thread-pool.
521The problem is when concurrent work-units need to interact and/or block as this effects the executor, \eg stops threads.
522While it is possible to extend this approach to support the necessary mechanisms, \eg message passing in Actors, we show monitors and threads provide an equally competitive approach that does not deviate from normal call communication and can be used to build asynchronous call, as is done in Java.
523
524
525\section{Stateful Function}
526\label{s:StatefulFunction}
527
528A \emph{stateful function} has the ability to remember state between calls, where state can be either data or execution, \eg plugin, device driver, finite-state machine (FSM).
529A simple technique to retain data state between calls is @static@ declarations within a function, which is often implemented by hoisting the declarations to the global scope but hiding the names within the function using name mangling.
530However, each call starts the function at the top making it difficult to determine the last point of execution in an algorithm, and requiring multiple flag variables and testing to reestablish the continuation point.
531Hence, the next step of generalizing function state is implicitly remembering the return point between calls and reentering the function at this point rather than the top, called \emph{generators}\,/\,\emph{iterators} or \emph{stackless coroutines}.
532For example, a Fibonacci generator retains data and execution state allowing it to remember prior values needed to generate the next value and the location in the algorithm to compute that value.
533The next step of generalization is instantiating the function to allow multiple named instances, \eg multiple Fibonacci generators, where each instance has its own state, and hence, can generate an independent sequence of values.
534Note, a subset of generator state is a function \emph{closure}, \ie the technique of capturing lexical references when returning a nested function.
535A further generalization is adding a stack to a generator's state, called a \emph{coroutine}, so it can suspend outside of itself, \eg call helper functions to arbitrary depth before suspending back to its resumer without unwinding these calls.
536For example, a coroutine iterator for a binary tree can stop the traversal at the visit point (pre, infix, post traversal), return the node value to the caller, and then continue the recursive traversal from the current node on the next call.
537
538There are two styles of activating a stateful function, \emph{asymmetric} or \emph{symmetric}, identified by resume/suspend (no cycles) and resume/resume (cycles).
539These styles \emph{do not} cause incremental stack growth, \eg a million resume/suspend or resume/resume cycles do not remember each cycle just the last resumer for each cycle.
540Selecting between stackless/stackful semantics and asymmetric/symmetric style is a tradeoff between programming requirements, performance, and design, where stackless is faster and smaller (modified call/return between closures), stackful is more general but slower and larger (context switching between distinct stacks), and asymmetric is simpler control-flow than symmetric.
541Additionally, storage management for the closure/stack (especially in unmanaged languages, \ie no garbage collection) must be factored into design and performance.
542Note, creation cost (closure/stack) is amortized across usage, so activation cost (resume/suspend) is usually the dominant factor.
543
544% The 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.
545% Hence, 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.
546% If 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.
547% If 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.
548% Hence, refactoring a stackless coroutine may require changing it to stackful.
549% A 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.
550% As well, activating a stateful function is \emph{asymmetric} or \emph{symmetric}, identified by resume/suspend (no cycles) and resume/resume (cycles).
551% A fixed closure activated by modified call/return is faster than a variable closure activated by context switching.
552% Additionally, any storage management for the closure (especially in unmanaged languages, \ie no garbage collection) must also be factored into design and performance.
553% Therefore, selecting between stackless and stackful semantics is a tradeoff between programming requirements and performance, where stackless is faster and stackful is more general.
554% nppNote, creation cost is amortized across usage, so activation cost is usually the dominant factor.
555
556For example, Python presents asymmetric generators as a function object, \uC presents symmetric coroutines as a \lstinline[language=C++]|class|-like object, and many languages present threading using function pointers, @pthreads@~\cite{Butenhof97}, \Csharp~\cite{Csharp}, Go~\cite{Go}, and Scala~\cite{Scala}.
557\begin{center}
558\begin{tabular}{@{}l|l|l@{}}
559\multicolumn{1}{@{}c|}{Python asymmetric generator} & \multicolumn{1}{c|}{\uC symmetric coroutine} & \multicolumn{1}{c@{}}{Pthreads thread} \\
560\hline
561\begin{python}
562`def Gen():` $\LstCommentStyle{\color{red}// function}$
563        ... yield val ...
564gen = Gen()
565for i in range( 10 ):
566        print( next( gen ) )
567\end{python}
568&
569\begin{uC++}
570`_Coroutine Cycle {` $\LstCommentStyle{\color{red}// class}$
571        Cycle * p;
572        void main() { p->cycle(); }
573        void cycle() { resume(); }  `};`
574Cycle c1, c2; c1.p=&c2; c2.p=&c1; c1.cycle();
575\end{uC++}
576&
577\begin{cfa}
578void * rtn( void * arg ) { ... }
579int i = 3, rc;
580pthread_t t; $\C{// thread id}$
581$\LstCommentStyle{\color{red}// function pointer}$
582rc=pthread_create(&t, `rtn`, (void *)i);
583\end{cfa}
584\end{tabular}
585\end{center}
586\CFA's preferred presentation model for generators/coroutines/threads is a hybrid of functions and classes, giving an object-oriented flavour.
587Essentially, the generator/coroutine/thread function is semantically coupled with a generator/coroutine/thread custom type via the type's name.
588The custom type solves several issues, while accessing the underlying mechanisms used by the custom types is still allowed for flexibility reasons.
589Each custom type is discussed in detail in the following sections.
590
591
592\subsection{Generator}
593
594Stackless generators (Table~\ref{t:ExecutionPropertyComposition} case 3) have the potential to be very small and fast, \ie as small and fast as function call/return for both creation and execution.
595The \CFA goal is to achieve this performance target, possibly at the cost of some semantic complexity.
596A series of different kinds of generators and their implementation demonstrate how this goal is accomplished.\footnote{
597The \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()|.
598Operator \lstinline+|+ is overloaded for printing, like bit-shift \lstinline|<<| in \CC.
599The \CFA \lstinline|with| clause opens an aggregate scope making its fields directly accessible, like Pascal \lstinline|with|, but using parallel semantics;
600multiple aggregates may be opened.
601\CFA has rebindable references \lstinline|int i, & ip = i, j; `&ip = &j;`| and non-rebindable references \lstinline|int i, & `const` ip = i, j; `&ip = &j;` // disallowed|.
602}%
603
604\begin{figure}
605\centering
606\begin{lrbox}{\myboxA}
607\begin{cfa}[aboveskip=0pt,belowskip=0pt]
608typedef struct {
609        int fn1, fn;
610} Fib;
611#define FibCtor { 1, 0 }
612int fib( Fib * f ) {
613
614
615
616
617
618        int fn = f->fn; f->fn = f->fn1;
619                f->fn1 = f->fn + fn;
620        return fn;
621}
622int main() {
623        Fib f1 = FibCtor, f2 = FibCtor;
624        for ( int i = 0; i < 10; i += 1 )
625                printf( "%d %d\n",
626                           fib( &f1 ), fib( &f2 ) );
627}
628\end{cfa}
629\end{lrbox}
630
631\begin{lrbox}{\myboxB}
632\begin{cfa}[aboveskip=0pt,belowskip=0pt]
633`generator` Fib {
634        int fn1, fn;
635};
636
637void `main(Fib & fib)` with(fib) {
638
639
640        [fn1, fn] = [1, 0];
641        for () {
642                `suspend;`
643                [fn1, fn] = [fn, fn + fn1];
644
645        }
646}
647int main() {
648        Fib f1, f2;
649        for ( 10 )
650                sout | `resume( f1 )`.fn
651                         | `resume( f2 )`.fn;
652}
653\end{cfa}
654\end{lrbox}
655
656\begin{lrbox}{\myboxC}
657\begin{cfa}[aboveskip=0pt,belowskip=0pt]
658typedef struct {
659        int `restart`, fn1, fn;
660} Fib;
661#define FibCtor { `0`, 1, 0 }
662Fib * comain( Fib * f ) {
663        `static void * states[] = {&&s0, &&s1};`
664        `goto *states[f->restart];`
665  s0: f->`restart` = 1;
666        for ( ;; ) {
667                return f;
668          s1:; int fn = f->fn + f->fn1;
669                f->fn1 = f->fn; f->fn = fn;
670        }
671}
672int main() {
673        Fib f1 = FibCtor, f2 = FibCtor;
674        for ( int i = 0; i < 10; i += 1 )
675                printf("%d %d\n",comain(&f1)->fn,
676                                 comain(&f2)->fn);
677}
678\end{cfa}
679\end{lrbox}
680
681\subfloat[C]{\label{f:CFibonacci}\usebox\myboxA}
682\hspace{3pt}
683\vrule
684\hspace{3pt}
685\subfloat[\CFA]{\label{f:CFAFibonacciGen}\usebox\myboxB}
686\hspace{3pt}
687\vrule
688\hspace{3pt}
689\subfloat[C generated code for \CFA version]{\label{f:CFibonacciSim}\usebox\myboxC}
690\caption{Fibonacci (output) asymmetric generator}
691\label{f:FibonacciAsymmetricGenerator}
692
693\bigskip
694
695\begin{lrbox}{\myboxA}
696\begin{cfa}[aboveskip=0pt,belowskip=0pt]
697`generator Fmt` {
698        char ch;
699        int g, b;
700};
701void ?{}( Fmt & fmt ) { `resume(fmt);` } // constructor
702void ^?{}( Fmt & f ) with(f) { $\C[2.25in]{// destructor}$
703        if ( g != 0 || b != 0 ) sout | nl; }
704void `main( Fmt & f )` with(f) {
705        for () { $\C{// until destructor call}$
706                for ( ; g < 5; g += 1 ) { $\C{// groups}$
707                        for ( ; b < 4; b += 1 ) { $\C{// blocks}$
708                                do { `suspend;` $\C{// wait for character}$
709                                while ( ch == '\n' ); // ignore newline
710                                sout | ch;                      $\C{// print character}$
711                        } sout | " ";  $\C{// block separator}$
712                } sout | nl; $\C{// group separator}$
713        }
714}
715int main() {
716        Fmt fmt; $\C{// fmt constructor called}$
717        for () {
718                sin | fmt.ch; $\C{// read into generator}$
719          if ( eof( sin ) ) break;
720                `resume( fmt );`
721        }
722
723} $\C{// fmt destructor called}\CRT$
724\end{cfa}
725\end{lrbox}
726
727\begin{lrbox}{\myboxB}
728\begin{cfa}[aboveskip=0pt,belowskip=0pt]
729typedef struct {
730        int `restart`, g, b;
731        char ch;
732} Fmt;
733void comain( Fmt * f ) {
734        `static void * states[] = {&&s0, &&s1};`
735        `goto *states[f->restart];`
736  s0: f->`restart` = 1;
737        for ( ;; ) {
738                for ( f->g = 0; f->g < 5; f->g += 1 ) {
739                        for ( f->b = 0; f->b < 4; f->b += 1 ) {
740                                do { return;  s1: ;
741                                } while ( f->ch == '\n' );
742                                printf( "%c", f->ch );
743                        } printf( " " );
744                } printf( "\n" );
745        }
746}
747int main() {
748        Fmt fmt = { `0` };  comain( &fmt ); // prime
749        for ( ;; ) {
750                scanf( "%c", &fmt.ch );
751          if ( feof( stdin ) ) break;
752                comain( &fmt );
753        }
754        if ( fmt.g != 0 || fmt.b != 0 ) printf( "\n" );
755}
756\end{cfa}
757\end{lrbox}
758
759\subfloat[\CFA]{\label{f:CFAFormatGen}\usebox\myboxA}
760\hspace{35pt}
761\vrule
762\hspace{3pt}
763\subfloat[C generated code for \CFA version]{\label{f:CFormatGenImpl}\usebox\myboxB}
764\hspace{3pt}
765\caption{Formatter (input) asymmetric generator}
766\label{f:FormatterAsymmetricGenerator}
767\end{figure}
768
769Figure~\ref{f:FibonacciAsymmetricGenerator} shows an unbounded asymmetric generator for an infinite sequence of Fibonacci numbers written (left to right) in C, \CFA, and showing the underlying C implementation for the \CFA version.
770This generator is an \emph{output generator}, producing a new result on each resumption.
771To 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.
772An additional requirement is the ability to create an arbitrary number of generators (of any kind), \ie retaining one state in global variables is insufficient;
773hence, state is retained in a closure between calls.
774Figure~\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.
775The C version only has the middle execution state because the top execution state is declaration initialization.
776Figure~\ref{f:CFAFibonacciGen} shows the \CFA approach, which also has a manual closure, but replaces the structure with a custom \CFA @generator@ type.
777Each generator type must have a function named \lstinline|main|,
778% \footnote{
779% The name \lstinline|main| has special meaning in C, specifically the function where a program starts execution.
780% Leveraging starting semantics to this name for generator/coroutine/thread is a logical extension.}
781called a \emph{generator main} (leveraging the starting semantics for program @main@ in C), which is connected to the generator type via its single reference parameter.
782The generator main contains @suspend@ statements that suspend execution without ending the generator versus @return@.
783For the Fibonacci generator-main,
784the top initialization state appears at the start and the middle execution state is denoted by statement @suspend@.
785Any local variables in @main@ \emph{are not retained} between calls;
786hence local variables are only for temporary computations \emph{between} suspends.
787All retained state \emph{must} appear in the generator's type.
788As 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.
789The 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.
790Resuming an ended (returned) generator is undefined.
791Function @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.
792Figure~\ref{f:CFibonacciSim} shows the C implementation of the \CFA asymmetric generator.
793Only one execution-state field, @restart@, is needed to subscript the suspension points in the generator.
794At the start of the generator main, the @static@ declaration, @states@, is initialized to the N suspend points in the generator (where operator @&&@ dereferences/references a label~\cite{gccValueLabels}).
795Next, the computed @goto@ selects the last suspend point and branches to it.
796The  cost of setting @restart@ and branching via the computed @goto@ adds very little cost to the suspend/resume calls.
797
798An advantage of the \CFA explicit generator type is the ability to allow multiple type-safe interface functions taking and returning arbitrary types.
799\begin{cfa}
800int ?()( Fib & fib ) { return `resume( fib )`.fn; } $\C[3.9in]{// function-call interface}$
801int ?()( Fib & fib, int N ) { for ( N - 1 ) `fib()`; return `fib()`; } $\C{// add parameter to skip N values}$
802double ?()( Fib & fib ) { return (int)`fib()` / 3.14159; } $\C{// different return type, cast prevents recursive call}$
803Fib f;  int i;  double d;
804i = f();  i = f( 2 );  d = f();                                         $\C{// alternative interfaces}\CRT$
805\end{cfa}
806Now, the generator can be a separately compiled opaque-type only accessed through its interface functions.
807For 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.
808
809\begin{figure}
810%\centering
811\newbox\myboxA
812\begin{lrbox}{\myboxA}
813\begin{python}[aboveskip=0pt,belowskip=0pt]
814def Fib():
815        fn1, fn = 0, 1
816        while True:
817                `yield fn1`
818                fn1, fn = fn, fn1 + fn
819f1 = Fib()
820f2 = Fib()
821for i in range( 10 ):
822        print( next( f1 ), next( f2 ) )
823
824
825
826
827
828
829
830
831
832
833\end{python}
834\end{lrbox}
835
836\newbox\myboxB
837\begin{lrbox}{\myboxB}
838\begin{python}[aboveskip=0pt,belowskip=0pt]
839def Fmt():
840        try:
841                while True:                                             $\C[2.5in]{\# until destructor call}$
842                        for g in range( 5 ):            $\C{\# groups}$
843                                for b in range( 4 ):    $\C{\# blocks}$
844                                        while True:
845                                                ch = (yield)    $\C{\# receive from send}$
846                                                if '\n' not in ch: $\C{\# ignore newline}$
847                                                        break
848                                        print( ch, end='' )     $\C{\# print character}$
849                                print( '  ', end='' )   $\C{\# block separator}$
850                        print()                                         $\C{\# group separator}$
851        except GeneratorExit:                           $\C{\# destructor}$
852                if g != 0 | b != 0:                             $\C{\# special case}$
853                        print()
854fmt = Fmt()
855`next( fmt )`                                                   $\C{\# prime, next prewritten}$
856for i in range( 41 ):
857        `fmt.send( 'a' );`                                      $\C{\# send to yield}$
858\end{python}
859\end{lrbox}
860
861\hspace{30pt}
862\subfloat[Fibonacci]{\label{f:PythonFibonacci}\usebox\myboxA}
863\hspace{3pt}
864\vrule
865\hspace{3pt}
866\subfloat[Formatter]{\label{f:PythonFormatter}\usebox\myboxB}
867\caption{Python generator}
868\label{f:PythonGenerator}
869\end{figure}
870
871Having to manually create the generator closure by moving local-state variables into the generator type is an additional programmer burden (removed by the coroutine in Section~\ref{s:Coroutine}).
872This manual requirement follows from the generality of allowing variable-size local-state, \eg local state with a variable-length array requires dynamic allocation as the array size is unknown at compile time.
873However, dynamic allocation significantly increases the cost of generator creation/destruction and is a showstopper for embedded real-time programming.
874But 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.
875With respect to safety, we believe static analysis can discriminate persistent generator state from temporary generator-main state and raise a compile-time error for temporary usage spanning suspend points.
876Our experience using generators is that the problems have simple data state, including local state, but complex execution state, so the burden of creating the generator type is small.
877As well, C programmers are not afraid of this kind of semantic programming requirement, if it results in very small, fast generators.
878
879Figure~\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.
880\begin{center}
881\tt
882\begin{tabular}{@{}l|l@{}}
883\multicolumn{1}{c|}{\textbf{\textrm{input}}} & \multicolumn{1}{c}{\textbf{\textrm{output}}} \\
884\begin{tabular}[t]{@{}ll@{}}
885abcdefghijklmnopqrstuvwxyz \\
886abcdefghijklmnopqrstuvwxyz
887\end{tabular}
888&
889\begin{tabular}[t]{@{}lllll@{}}
890abcd    & efgh  & ijkl  & mnop  & qrst  \\
891uvwx    & yzab  & cdef  & ghij  & klmn  \\
892opqr    & stuv  & wxyz  &               &
893\end{tabular}
894\end{tabular}
895\end{center}
896The 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.
897The destructor provides a newline, if formatted text ends with a full line.
898Figure~\ref{f:CFormatGenImpl} shows the C implementation of the \CFA input generator with one additional field and the computed @goto@.
899For contrast, Figure~\ref{f:PythonFormatter} shows the equivalent Python format generator with the same properties as the format generator.
900
901% https://dl-acm-org.proxy.lib.uwaterloo.ca/
902
903Figure~\ref{f:DeviceDriverGen} shows an important application for an asymmetric generator, a device-driver, because device drivers are a significant source of operating-system errors: 85\% in Windows XP~\cite[p.~78]{Swift05} and 51.6\% in Linux~\cite[p.~1358,]{Xiao19}. %\cite{Palix11}
904Swift \etal~\cite[p.~86]{Swift05} restructure device drivers using the Extension Procedure Call (XPC) within the kernel via functions @nooks_driver_call@ and @nooks_kernel_call@, which have coroutine properties context switching to separate stacks with explicit hand-off calls;
905however, the calls do not retain execution state, and hence always start from the top.
906The alternative approach for implementing device drivers is using stack-ripping.
907However, Adya \etal~\cite{Adya02} argue against stack ripping in Section 3.2 and suggest a hybrid approach in Section 4 using cooperatively scheduled \emph{fibers}, which is coroutining.
908
909As an example, the following protocol:
910\begin{center}
911\ldots\, STX \ldots\, message \ldots\, ESC ETX \ldots\, message \ldots\, ETX 2-byte crc \ldots
912\end{center}
913is for a simple network message beginning with the control character STX, ending with an ETX, and followed by a 2-byte cyclic-redundancy check.
914Control characters may appear in a message if preceded by an ESC.
915When 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.
916The device driver returns a status code of its current state, and when a complete message is obtained, the operating system read the message accumulated in the supplied buffer.
917Hence, the device driver is an input/output generator, where the cost of resuming the device-driver generator is the same as call/return, so performance in an operating-system kernel is excellent.
918The key benefits of using a generator are correctness, safety, and maintenance because the execution states are transcribed directly into the programming language rather than table lookup or stack ripping.
919The conclusion is that FSMs are complex and occur in important domains, so direct generator support is important in a system programming language.
920
921\begin{figure}
922\centering
923\begin{tabular}{@{}l|l@{}}
924\begin{cfa}[aboveskip=0pt,belowskip=0pt]
925enum Status { CONT, MSG, ESTX,
926                                ELNTH, ECRC };
927`generator` Driver {
928        Status status;
929        char byte, * msg; // communication
930        int lnth, sum;      // local state
931        short int crc;
932};
933void ?{}( Driver & d, char * m ) { d.msg = m; }
934Status next( Driver & d, char b ) with( d ) {
935        byte = b; `resume( d );` return status;
936}
937void main( Driver & d ) with( d ) {
938        enum { STX = '\002', ESC = '\033',
939                        ETX = '\003', MaxMsg = 64 };
940  msg: for () { // parse message
941                status = CONT;
942                lnth = 0; sum = 0;
943                while ( byte != STX ) `suspend;`
944          emsg: for () {
945                        `suspend;` // process byte
946\end{cfa}
947&
948\begin{cfa}[aboveskip=0pt,belowskip=0pt]
949                        choose ( byte ) { // switch with implicit break
950                          case STX:
951                                status = ESTX; `suspend;` continue msg;
952                          case ETX:
953                                break emsg;
954                          case ESC:
955                                `suspend;`
956                        }
957                        if ( lnth >= MaxMsg ) { // buffer full ?
958                                status = ELNTH; `suspend;` continue msg; }
959                        msg[lnth++] = byte;
960                        sum += byte;
961                }
962                msg[lnth] = '\0'; // terminate string
963                `suspend;`
964                crc = byte << 8;
965                `suspend;`
966                status = (crc | byte) == sum ? MSG : ECRC;
967                `suspend;`
968        }
969}
970\end{cfa}
971\end{tabular}
972\caption{Device-driver generator for communication protocol}
973\label{f:DeviceDriverGen}
974\end{figure}
975
976Figure~\ref{f:CFAPingPongGen} shows a symmetric generator, where the generator resumes another generator, forming a resume/resume cycle.
977(The trivial cycle is a generator resuming itself.)
978This control flow is similar to recursion for functions but without stack growth.
979Figure~\ref{f:PingPongFullCoroutineSteps} shows the steps for symmetric control-flow are creating, executing, and terminating the cycle.
980Constructing 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.
981(This issue occurs for any cyclic data structure.)
982The example creates the generators, @ping@/@pong@, and then assigns the partners that form the cycle.
983% (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.)
984Once the cycle is formed, the program main resumes one of the generators, @ping@, and the generators can then traverse an arbitrary cycle using @resume@ to activate partner generator(s).
985Terminating 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).
986Note, the creator and starter may be different, \eg if the creator calls another function that starts the cycle.
987The starting stack-frame is below the last active generator because the resume/resume cycle does not grow the stack.
988Also, since local variables are not retained in the generator function, there are no objects with destructors to be called, so the cost is the same as a function return.
989Destructor cost occurs when the generator instance is deallocated by the creator.
990
991\begin{figure}
992\centering
993\begin{lrbox}{\myboxA}
994\begin{cfa}[aboveskip=0pt,belowskip=0pt]
995`generator PingPong` {
996        int N, i;                               // local state
997        const char * name;
998        PingPong & partner; // rebindable reference
999};
1000
1001void `main( PingPong & pp )` with(pp) {
1002
1003
1004        for ( ; i < N; i += 1 ) {
1005                sout | name | i;
1006                `resume( partner );`
1007        }
1008}
1009int main() {
1010        enum { N = 5 };
1011        PingPong ping = {"ping",N,0}, pong = {"pong",N,0};
1012        &ping.partner = &pong;  &pong.partner = &ping;
1013        `resume( ping );`
1014}
1015\end{cfa}
1016\end{lrbox}
1017
1018\begin{lrbox}{\myboxB}
1019\begin{cfa}[escapechar={},aboveskip=0pt,belowskip=0pt]
1020typedef struct PingPong {
1021        int restart, N, i;
1022        const char * name;
1023        struct PingPong * partner;
1024} PingPong;
1025#define PPCtor(name, N) {0, N, 0, name, NULL}
1026void comain( PingPong * pp ) {
1027        static void * states[] = {&&s0, &&s1};
1028        goto *states[pp->restart];
1029  s0: pp->restart = 1;
1030        for ( ; pp->i < pp->N; pp->i += 1 ) {
1031                printf( "%s %d\n", pp->name, pp->i );
1032                asm( "mov  %0,%%rdi" : "=m" (pp->partner) );
1033                asm( "mov  %rdi,%rax" );
1034                asm( "add  $16, %rsp" );
1035                asm( "popq %rbp" );
1036                asm( "jmp  comain" );
1037          s1: ;
1038        }
1039}
1040\end{cfa}
1041\end{lrbox}
1042
1043\subfloat[\CFA symmetric generator]{\label{f:CFAPingPongGen}\usebox\myboxA}
1044\hspace{3pt}
1045\vrule
1046\hspace{3pt}
1047\subfloat[C generator simulation]{\label{f:CPingPongSim}\usebox\myboxB}
1048\hspace{3pt}
1049\caption{Ping-Pong symmetric generator}
1050\label{f:PingPongSymmetricGenerator}
1051\end{figure}
1052
1053\begin{figure}
1054\centering
1055\input{FullCoroutinePhases.pstex_t}
1056\vspace*{-10pt}
1057\caption{Symmetric coroutine steps: Ping / Pong}
1058\label{f:PingPongFullCoroutineSteps}
1059\end{figure}
1060
1061Figure~\ref{f:CPingPongSim} shows the C implementation of the \CFA symmetric generator, where there is still only one additional field, @restart@, but @resume@ is more complex because it does a forward rather than backward jump.
1062Before the jump, the parameter for the next call @partner@ is placed into the register used for the first parameter, @rdi@, and the remaining registers are reset for a return.
1063The @jmp comain@ restarts the function but with a different parameter, so the new call's behaviour depends on the state of the coroutine type, i.e., branch to restart location with different data state.
1064While the semantics of call forward is a tail-call optimization, which compilers perform, the generator state is different on each call rather a common state for a tail-recursive function (i.e., the parameter to the function never changes during the forward calls.
1065However, this assembler code depends on what entry code is generated, specifically if there are local variables and the level of optimization.
1066Hence, internal compiler support is necessary for any forward call (or backwards return), \eg LLVM has various coroutine support~\cite{CoroutineTS}, and \CFA can leverage this support should it eventually fork @clang@.
1067For this reason, \CFA does not support general symmetric generators at this time, but, it is possible to hand generate any symmetric generators (as in Figure~\ref{f:CPingPongSim}) for proof of concept and performance testing.
1068
1069Finally, part of this generator work was inspired by the recent \CCtwenty coroutine proposal~\cite{C++20Coroutine19}, which uses the general term coroutine to mean generator.
1070Our work provides the same high-performance asymmetric generators as \CCtwenty, and extends their work with symmetric generators.
1071An 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:
1072\begin{cfa}
1073... suspend`{ ... }`;
1074... resume( C )`{ ... }` ...
1075\end{cfa}
1076Since the current generator's stack is released before calling the compound statement, the compound statement can only reference variables in the generator's type.
1077This 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.
1078Hence, this mechanism provides a general and safe handoff of the generator among competing threads.
1079
1080
1081\subsection{Coroutine}
1082\label{s:Coroutine}
1083
1084Stackful coroutines (Table~\ref{t:ExecutionPropertyComposition} case 5) extend generator semantics, \ie there is an implicit closure and @suspend@ may appear in a helper function called from the coroutine main.
1085A coroutine is specified by replacing @generator@ with @coroutine@ for the type.
1086Coroutine generality results in higher cost for creation, due to dynamic stack allocation, for execution, due to context switching among stacks, and for terminating, due to possible stack unwinding and dynamic stack deallocation.
1087A series of different kinds of coroutines and their implementations demonstrate how coroutines extend generators.
1088
1089First, 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.
1090\begin{center}
1091\begin{tabular}{@{}l|l|l|l@{}}
1092\multicolumn{1}{c|}{Fibonacci} & \multicolumn{1}{c|}{Formatter} & \multicolumn{1}{c|}{Device Driver} & \multicolumn{1}{c}{PingPong} \\
1093\hline
1094\begin{cfa}[xleftmargin=0pt]
1095void main( Fib & fib ) ...
1096        `int fn1;`
1097
1098
1099\end{cfa}
1100&
1101\begin{cfa}[xleftmargin=0pt]
1102for ( `g`; 5 ) {
1103        for ( `b`; 4 ) {
1104
1105
1106\end{cfa}
1107&
1108\begin{cfa}[xleftmargin=0pt]
1109status = CONT;
1110`int lnth = 0, sum = 0;`
1111...
1112`short int crc = byte << 8;`
1113\end{cfa}
1114&
1115\begin{cfa}[xleftmargin=0pt]
1116void main( PingPong & pp ) ...
1117        for ( `i`; N ) {
1118
1119
1120\end{cfa}
1121\end{tabular}
1122\end{center}
1123It 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.
1124\begin{cfa}
1125int Crc() {
1126        `suspend;`
1127        short int crc = byte << 8;
1128        `suspend;`
1129        status = (crc | byte) == sum ? MSG : ECRC;
1130        return crc;
1131}
1132\end{cfa}
1133A call to this function is placed at the end of the driver's coroutine-main.
1134For complex finite-state machines, refactoring is part of normal program abstraction, especially when code is used in multiple places.
1135Again, this complexity is usually associated with execution state rather than data state.
1136
1137\begin{comment}
1138Figure~\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 @restart@.
1139Like 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.
1140The 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@.
1141The interface function @restart@, takes a Fibonacci instance and context switches to it using @resume@;
1142on restart, the Fibonacci field, @fn@, contains the next value in the sequence, which is returned.
1143The first @resume@ is special because it allocates the coroutine stack and cocalls its coroutine main on that stack;
1144when the coroutine main returns, its stack is deallocated.
1145Hence, @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.
1146Figure~\ref{f:Coroutine1State} shows the coroutine version of the C version in Figure~\ref{f:ExternalState}.
1147Coroutine generators are called \newterm{output coroutines} because values are only returned.
1148
1149\begin{figure}
1150\centering
1151\newbox\myboxA
1152% \begin{lrbox}{\myboxA}
1153% \begin{cfa}[aboveskip=0pt,belowskip=0pt]
1154% `int fn1, fn2, state = 1;`   // single global variables
1155% int fib() {
1156%       int fn;
1157%       `switch ( state )` {  // explicit execution state
1158%         case 1: fn = 0;  fn1 = fn;  state = 2;  break;
1159%         case 2: fn = 1;  fn2 = fn1;  fn1 = fn;  state = 3;  break;
1160%         case 3: fn = fn1 + fn2;  fn2 = fn1;  fn1 = fn;  break;
1161%       }
1162%       return fn;
1163% }
1164% int main() {
1165%
1166%       for ( int i = 0; i < 10; i += 1 ) {
1167%               printf( "%d\n", fib() );
1168%       }
1169% }
1170% \end{cfa}
1171% \end{lrbox}
1172\begin{lrbox}{\myboxA}
1173\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1174#define FibCtor { 0, 1 }
1175typedef struct { int fn1, fn; } Fib;
1176int fib( Fib * f ) {
1177
1178        int ret = f->fn1;
1179        f->fn1 = f->fn;
1180        f->fn = ret + f->fn;
1181        return ret;
1182}
1183
1184
1185
1186int main() {
1187        Fib f1 = FibCtor, f2 = FibCtor;
1188        for ( int i = 0; i < 10; i += 1 ) {
1189                printf( "%d %d\n",
1190                                fib( &f1 ), fib( &f2 ) );
1191        }
1192}
1193\end{cfa}
1194\end{lrbox}
1195
1196\newbox\myboxB
1197\begin{lrbox}{\myboxB}
1198\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1199`coroutine` Fib { int fn1; };
1200void main( Fib & fib ) with( fib ) {
1201        int fn;
1202        [fn1, fn] = [0, 1];
1203        for () {
1204                `suspend;`
1205                [fn1, fn] = [fn, fn1 + fn];
1206        }
1207}
1208int ?()( Fib & fib ) with( fib ) {
1209        return `resume( fib )`.fn1;
1210}
1211int main() {
1212        Fib f1, f2;
1213        for ( 10 ) {
1214                sout | f1() | f2();
1215}
1216
1217
1218\end{cfa}
1219\end{lrbox}
1220
1221\newbox\myboxC
1222\begin{lrbox}{\myboxC}
1223\begin{python}[aboveskip=0pt,belowskip=0pt]
1224
1225def Fib():
1226
1227        fn1, fn = 0, 1
1228        while True:
1229                `yield fn1`
1230                fn1, fn = fn, fn1 + fn
1231
1232
1233// next prewritten
1234
1235
1236f1 = Fib()
1237f2 = Fib()
1238for i in range( 10 ):
1239        print( next( f1 ), next( f2 ) )
1240
1241
1242
1243\end{python}
1244\end{lrbox}
1245
1246\subfloat[C]{\label{f:GlobalVariables}\usebox\myboxA}
1247\hspace{3pt}
1248\vrule
1249\hspace{3pt}
1250\subfloat[\CFA]{\label{f:ExternalState}\usebox\myboxB}
1251\hspace{3pt}
1252\vrule
1253\hspace{3pt}
1254\subfloat[Python]{\label{f:ExternalState}\usebox\myboxC}
1255\caption{Fibonacci generator}
1256\label{f:C-fibonacci}
1257\end{figure}
1258
1259\bigskip
1260
1261\newbox\myboxA
1262\begin{lrbox}{\myboxA}
1263\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1264`coroutine` Fib { int fn; };
1265void main( Fib & fib ) with( fib ) {
1266        fn = 0;  int fn1 = fn; `suspend`;
1267        fn = 1;  int fn2 = fn1;  fn1 = fn; `suspend`;
1268        for () {
1269                fn = fn1 + fn2; fn2 = fn1; fn1 = fn; `suspend`; }
1270}
1271int next( Fib & fib ) with( fib ) { `resume( fib );` return fn; }
1272int main() {
1273        Fib f1, f2;
1274        for ( 10 )
1275                sout | next( f1 ) | next( f2 );
1276}
1277\end{cfa}
1278\end{lrbox}
1279\newbox\myboxB
1280\begin{lrbox}{\myboxB}
1281\begin{python}[aboveskip=0pt,belowskip=0pt]
1282
1283def Fibonacci():
1284        fn = 0; fn1 = fn; `yield fn`  # suspend
1285        fn = 1; fn2 = fn1; fn1 = fn; `yield fn`
1286        while True:
1287                fn = fn1 + fn2; fn2 = fn1; fn1 = fn; `yield fn`
1288
1289
1290f1 = Fibonacci()
1291f2 = Fibonacci()
1292for i in range( 10 ):
1293        print( `next( f1 )`, `next( f2 )` ) # resume
1294
1295\end{python}
1296\end{lrbox}
1297\subfloat[\CFA]{\label{f:Coroutine3States}\usebox\myboxA}
1298\qquad
1299\subfloat[Python]{\label{f:Coroutine1State}\usebox\myboxB}
1300\caption{Fibonacci input coroutine, 3 states, internal variables}
1301\label{f:cfa-fibonacci}
1302\end{figure}
1303\end{comment}
1304
1305\begin{figure}
1306\centering
1307\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
1308\begin{cfa}
1309`coroutine` Prod {
1310        Cons & c;                       $\C[1.5in]{// communication}$
1311        int N, money, receipt;
1312};
1313void main( Prod & prod ) with( prod ) {
1314        for ( i; N ) {          $\C{// 1st resume}\CRT$
1315                int p1 = random( 100 ), p2 = random( 100 );
1316                int status = delivery( c, p1, p2 );
1317                receipt += 1;
1318        }
1319        stop( c );
1320}
1321int payment( Prod & prod, int money ) {
1322        prod.money = money;
1323        `resume( prod );`
1324        return prod.receipt;
1325}
1326void start( Prod & prod, int N, Cons &c ) {
1327        &prod.c = &c;
1328        prod.[N, receipt] = [N, 0];
1329        `resume( prod );`
1330}
1331int main() {
1332        Prod prod;
1333        Cons cons = { prod };
1334        start( prod, 5, cons );
1335}
1336\end{cfa}
1337&
1338\begin{cfa}
1339`coroutine` Cons {
1340        Prod & p;                       $\C[1.5in]{// communication}$
1341        int p1, p2, status;
1342        bool done;
1343};
1344void ?{}( Cons & cons, Prod & p ) {
1345        &cons.p = &p;           $\C{// reassignable reference}$
1346        cons.[status, done ] = [0, false];
1347}
1348void main( Cons & cons ) with( cons ) {
1349        int money = 1, receipt; $\C{// 1st resume}\CRT$
1350        for ( ; ! done; ) {
1351                status += 1;
1352                receipt = payment( p, money );
1353                money += 1;
1354        }
1355}
1356int delivery( Cons & cons, int p1, int p2 ) {
1357        cons.[p1, p2] = [p1, p2];
1358        `resume( cons );`
1359        return cons.status;
1360}
1361void stop( Cons & cons ) {
1362        cons.done = true;
1363        `resume( cons );`
1364}
1365
1366\end{cfa}
1367\end{tabular}
1368\caption{Producer / consumer: resume-resume cycle, bidirectional communication}
1369\label{f:ProdCons}
1370\end{figure}
1371
1372Figure~\ref{f:ProdCons} shows the ping-pong example in Figure~\ref{f:CFAPingPongGen} extended into a producer/consumer symmetric-coroutine performing bidirectional communication.
1373This example is illustrative because both producer/consumer have two interface functions with @resume@s that suspend execution in these interface (helper) functions.
1374The 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.
1375The call to @start@ is the first @resume@ of @prod@, which remembers the program main as the starter and creates @prod@'s stack with a frame for @prod@'s coroutine main at the top, and context switches to it.
1376@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's @deliver@ function to transfer the values, and printing the status returned from the consumer.
1377The producer call to @delivery@ transfers values into the consumer's communication variables, resumes the consumer, and returns the consumer status.
1378Similarly on the first resume, @cons@'s stack is created and initialized, holding local-state variables retained between subsequent activations of the coroutine.
1379The symmetric coroutine cycle forms when the consumer calls the producer's @payment@ function, which resumes the producer in the consumer's delivery function.
1380When the producer calls @delivery@ again, it resumes the consumer in the @payment@ function.
1381Both interface function than return to the their corresponding coroutine-main functions for the next cycle.
1382Figure~\ref{f:ProdConsRuntimeStacks} shows the runtime stacks of the program main, and the coroutine mains for @prod@ and @cons@ during the cycling.
1383As a consequence of a coroutine retaining its last resumer for suspending back, these reverse pointers allow @suspend@ to cycle \emph{backwards} around a symmetric coroutine cycle.
1384
1385\begin{figure}
1386\begin{center}
1387\input{FullProdConsStack.pstex_t}
1388\end{center}
1389\vspace*{-10pt}
1390\caption{Producer / consumer runtime stacks}
1391\label{f:ProdConsRuntimeStacks}
1392\end{figure}
1393
1394Terminating 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.
1395Furthermore, each deallocated coroutine must execute all destructors 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.
1396In the example, termination begins with the producer's loop stopping after N iterations and calling the consumer's @stop@ function, which sets the @done@ flag, resumes the consumer in function @payment@, terminating the call, and the consumer's loop in its coroutine main.
1397% (Not shown is having @prod@ raise a nonlocal @stop@ exception at @cons@ after it finishes generating values and suspend back to @cons@, which catches the @stop@ exception to terminate its loop.)
1398When the consumer's main ends, its stack is already unwound so any stack allocated objects with destructors are finalized.
1399The question now is where does control continue?
1400
1401The na\"{i}ve semantics for coroutine-cycle termination is to context switch to the last resumer, like executing a @suspend@/@return@ in a generator.
1402However, for coroutines, the last resumer is \emph{not} implicitly below the current stack frame, as for generators, because each coroutine's stack is independent.
1403Unfortunately, it is impossible to determine statically if a coroutine is in a cycle and unrealistic to check dynamically (graph-cycle problem).
1404Hence, a compromise solution is necessary that works for asymmetric (acyclic) and symmetric (cyclic) coroutines.
1405Our solution is to retain a coroutine's starter (first resumer), and context switch back to the starter when the coroutine ends.
1406Hence, the consumer restarts its first resumer, @prod@, in @stop@, and when the producer ends, it restarts its first resumer, program main, in @start@ (see dashed lines from the end of the coroutine mains in Figure~\ref{f:ProdConsRuntimeStacks}).
1407This semantics works well for the most common asymmetric and symmetric coroutine usage patterns.
1408For asymmetric coroutines, it is common for the first resumer (starter) coroutine to be the only resumer;
1409for symmetric coroutines, it is common for the cycle creator to persist for the lifetime of the cycle.
1410For 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.
1411
1412Note, the producer/consumer example does not illustrate the full power of the starter semantics because @cons@ always ends first.
1413Assume generator @PingPong@ in Figure~\ref{f:PingPongSymmetricGenerator} is converted to a coroutine.
1414Unlike generators, coroutines have a starter structure with multiple levels, where the program main starts @ping@ and @ping@ starts @pong@.
1415By adjusting $N$ for either @ping@/@pong@, it is possible to have either finish first.
1416If @pong@ ends first, it resumes its starter @ping@ in its coroutine main, then @ping@ ends and resumes its starter the program main on return;
1417if @ping@ ends first, it resumes its starter the program main on return.
1418Regardless of the cycle complexity, the starter structure always leads back to the program main, but the path can be entered at an arbitrary point.
1419Once back at the program main (creator), coroutines @ping@ and @pong@ are deallocated, runnning any destructors for objects within the coroutine and possibly deallocating any coroutine stacks for non-terminated coroutines, where stack deallocation implies stack unwinding to find destructors for allocated objects on the stack.
1420Hence, the \CFA termination semantics for the generator and coroutine ensure correct deallocation semnatics, regardless of the coroutine's state (terminated or active), like any other aggregate object.
1421
1422
1423\subsection{Generator / Coroutine Implementation}
1424
1425A 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.
1426There are several solutions to these problem, which follow from the object-oriented flavour of adopting custom types.
1427
1428For object-oriented languages, inheritance is used to provide extra fields and code via explicit inheritance:
1429\begin{cfa}[morekeywords={class,inherits}]
1430class myCoroutine inherits baseCoroutine { ... }
1431\end{cfa}
1432% 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.
1433The 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.
1434Alternatives, such as explicitly starting threads as in Java, are repetitive and forgetting to call start is a common source of errors.
1435An alternative is composition:
1436\begin{cfa}
1437struct myCoroutine {
1438        ... // declaration/communication variables
1439        baseCoroutine dummy; // composition, last declaration
1440}
1441\end{cfa}
1442which also requires an explicit declaration that must be last to ensure correct initialization order.
1443However, there is nothing preventing wrong placement or multiple declarations.
1444
1445\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
1446% and when it is unsafe to perform certain optimizations,
1447and IDEs using simple parsing can find and manipulate types with special properties.
1448The downside of this approach is that it makes custom types a special case in the language.
1449Users wanting to extend custom types or build their own can only do so in ways offered by the language.
1450Furthermore, implementing custom types without language support may display the power of a programming language.
1451\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 (@java.util.concurrent@) monitors.
1452
1453Part 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.
1454\begin{cfa}
1455trait is_coroutine( `dtype` T ) {
1456        void main( T & );
1457        coroutine_desc * get_coroutine( T & );
1458};
1459forall( `dtype` T | is_coroutine(T) ) void $suspend$( T & ), resume( T & );
1460\end{cfa}
1461Note, copying generators/coroutines/threads is undefined because muliple objects cannot execute on a shared stack and stack copying does not work in unmanaged languages (no garbage collection), like C, because the stack may contain pointers to objects within it that require updating for the copy.
1462The \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).
1463The 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.
1464The @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.
1465The 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@.
1466
1467The \CFA custom-type @coroutine@ implicitly implements the getter and forward declarations for the coroutine main.
1468\begin{cquote}
1469\begin{tabular}{@{}ccc@{}}
1470\begin{cfa}
1471coroutine MyCor {
1472        int value;
1473
1474};
1475\end{cfa}
1476&
1477{\Large $\Rightarrow$}
1478&
1479\begin{tabular}{@{}ccc@{}}
1480\begin{cfa}
1481struct MyCor {
1482        int value;
1483        coroutine_desc cor;
1484};
1485\end{cfa}
1486&
1487\begin{cfa}
1488static inline coroutine_desc *
1489get_coroutine( MyCor & this ) {
1490        return &this.cor;
1491}
1492\end{cfa}
1493&
1494\begin{cfa}
1495void main( MyCor * this );
1496
1497
1498
1499\end{cfa}
1500\end{tabular}
1501\end{tabular}
1502\end{cquote}
1503The 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.
1504
1505Figure~\ref{f:CoroutineMemoryLayout} shows different memory-layout options for a coroutine (where a thread is similar).
1506The coroutine handle is the @coroutine@ instance containing programmer specified type global/communication variables across interface functions.
1507The coroutine descriptor contains all implicit declarations needed by the runtime, \eg @suspend@/@resume@, and can be part of the coroutine handle or separate.
1508The coroutine stack can appear in a number of locations and be fixed or variable sized.
1509Hence, the coroutine's stack could be a variable-length structure (VLS)\footnote{
1510We are examining VLSs, where fields can be variable-sized structures or arrays.
1511Once allocated, a VLS is fixed sized.}
1512on the allocating stack, provided the allocating stack is large enough.
1513For a VLS stack allocation/deallocation is an inexpensive adjustment of the stack pointer, modulo any stack constructor costs (\eg initial frame setup).
1514For stack allocation in the heap, allocation/deallocation is an expensive allocation, where the heap can be a shared resource, modulo any stack constructor costs.
1515It is also possible to use a split (segmented) stack calling convention, available with gcc and clang, allowing a variable-sized stack via a set of connected blocks in the heap.
1516Currently, \CFA supports stack/heap allocated descriptors but only fixed-sized heap allocated stacks.
1517In \CFA debug-mode, the fixed-sized stack is terminated with a write-only page, which catches most stack overflows.
1518Experience teaching concurrency with \uC~\cite{CS343} shows fixed-sized stacks are rarely an issue for students.
1519Split-stack allocation is under development but requires recompilation of legacy code, which is not always possible.
1520
1521\begin{figure}
1522\centering
1523\input{corlayout.pstex_t}
1524\caption{Coroutine memory layout}
1525\label{f:CoroutineMemoryLayout}
1526\end{figure}
1527
1528
1529\section{Concurrency}
1530\label{s:Concurrency}
1531
1532Concurrency is nondeterministic scheduling of independent sequential execution paths (threads), where each thread has its own stack.
1533A single thread with multiple stacks, \ie coroutining, does \emph{not} imply concurrency~\cite[\S~3]{Buhr05a}.
1534Coroutining self-schedule the thread across stacks so execution is deterministic.
1535(It is \emph{impossible} to generate a concurrency error when coroutining.)
1536
1537The 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}.
1538Therefore, a minimal concurrency system requires coroutines \emph{in conjunction with a nondeterministic scheduler}.
1539The resulting execution system now follows a cooperative threading-model~\cite{Adya02,libdill} because context-switching points to the scheduler (blocking) are known, but the next unblocking point is unknown due to the scheduler.
1540Adding \newterm{preemption} introduces \newterm{non-cooperative} or \newterm{preemptive} scheduling, where context switching points to the scheduler are unknown as they can occur randomly between any two instructions often based on a timer interrupt.
1541Uncertainty gives the illusion of parallelism on a single processor and provides a mechanism to access and increase performance on multiple processors.
1542The reason is that the scheduler/runtime have complete knowledge about resources and how to best utilized them.
1543However, the introduction of unrestricted nondeterminism results in the need for \newterm{mutual exclusion} and \newterm{synchronization}~\cite[\S~4]{Buhr05a}, which restrict nondeterminism for correctness;
1544otherwise, it is impossible to write meaningful concurrent programs.
1545Optimal concurrent performance is often obtained by having as much nondeterminism as mutual exclusion and synchronization correctness allow.
1546
1547A scheduler can also be stackless or stackful.
1548For 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.
1549For 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.
1550The \CFA runtime uses a stackful scheduler for uniformity and security.
1551
1552
1553\subsection{Thread}
1554\label{s:threads}
1555
1556Threading (Table~\ref{t:ExecutionPropertyComposition} case 11) needs the ability to start a thread and wait for its completion.
1557A common API for this ability is @fork@ and @join@.
1558\vspace{4pt}
1559\par\noindent
1560\begin{tabular}{@{}l|l|l@{}}
1561\multicolumn{1}{c|}{\textbf{Java}} & \multicolumn{1}{c|}{\textbf{\Celeven}} & \multicolumn{1}{c}{\textbf{pthreads}} \\
1562\hline
1563\begin{cfa}
1564class MyThread extends Thread {...}
1565mythread t = new MyThread(...);
1566`t.start();` // start
1567// concurrency
1568`t.join();` // wait
1569\end{cfa}
1570&
1571\begin{cfa}
1572class MyThread { ... } // functor
1573MyThread mythread;
1574`thread t( mythread, ... );` // start
1575// concurrency
1576`t.join();` // wait
1577\end{cfa}
1578&
1579\begin{cfa}
1580void * rtn( void * arg ) {...}
1581pthread_t t;  int i = 3;
1582`pthread_create( &t, rtn, (void *)i );` // start
1583// concurrency
1584`pthread_join( t, NULL );` // wait
1585\end{cfa}
1586\end{tabular}
1587\vspace{1pt}
1588\par\noindent
1589\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.
1590\begin{cfa}
1591thread MyThread {};
1592void main( MyThread & this ) { ... }
1593int main() {
1594        MyThread team`[10]`; $\C[2.5in]{// allocate stack-based threads, implicit start after construction}$
1595        // concurrency
1596} $\C{// deallocate stack-based threads, implicit joins before destruction}$
1597\end{cfa}
1598This semantic ensures a thread is started and stopped exactly once, eliminating some programming error, and scales to multiple threads for basic (termination) synchronization.
1599For block allocation to arbitrary depth, including recursion, threads are created/destroyed in a lattice structure (tree with top and bottom).
1600Arbitrary topologies are possible using dynamic allocation, allowing threads to outlive their declaration scope, identical to normal dynamic allocation.
1601\begin{cfa}
1602MyThread * factory( int N ) { ... return `anew( N )`; } $\C{// allocate heap-based threads, implicit start after construction}$
1603int main() {
1604        MyThread * team = factory( 10 );
1605        // concurrency
1606        `delete( team );` $\C{// deallocate heap-based threads, implicit joins before destruction}\CRT$
1607}
1608\end{cfa}
1609
1610Figure~\ref{s:ConcurrentMatrixSummation} shows concurrently adding the rows of a matrix and then totalling the subtotals sequentially, after all the row threads have terminated.
1611The program uses heap-based threads because each thread needs different constructor values.
1612(Python provides a simple iteration mechanism to initialize array elements to different values allowing stack allocation.)
1613The allocation/deallocation pattern appears unusual because allocated objects are immediately deallocated without any intervening code.
1614However, for threads, the deletion provides implicit synchronization, which is the intervening code.
1615% 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.
1616
1617\begin{figure}
1618\begin{cfa}
1619`thread` Adder { int * row, cols, & subtotal; } $\C{// communication variables}$
1620void ?{}( Adder & adder, int row[], int cols, int & subtotal ) {
1621        adder.[ row, cols, &subtotal ] = [ row, cols, &subtotal ];
1622}
1623void main( Adder & adder ) with( adder ) {
1624        subtotal = 0;
1625        for ( c; cols ) { subtotal += row[c]; }
1626}
1627int main() {
1628        const int rows = 10, cols = 1000;
1629        int matrix[rows][cols], subtotals[rows], total = 0;
1630        // read matrix
1631        Adder * adders[rows];
1632        for ( r; rows; ) { $\C{// start threads to sum rows}$
1633                adders[r] = `new( matrix[r], cols, &subtotals[r] );`
1634        }
1635        for ( r; rows ) { $\C{// wait for threads to finish}$
1636                `delete( adders[r] );` $\C{// termination join}$
1637                total += subtotals[r]; $\C{// total subtotal}$
1638        }
1639        sout | total;
1640}
1641\end{cfa}
1642\caption{Concurrent matrix summation}
1643\label{s:ConcurrentMatrixSummation}
1644\end{figure}
1645
1646
1647\subsection{Thread Implementation}
1648
1649Threads 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.
1650Like coroutines, and for the same design reasons, \CFA provides a custom @thread@ type and a @trait@ to enforce and restrict the thread-interface functions.
1651\begin{cquote}
1652\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1653\begin{cfa}
1654thread myThread {
1655        ... // declaration/communication variables
1656};
1657
1658
1659\end{cfa}
1660&
1661\begin{cfa}
1662trait is_thread( `dtype` T ) {
1663        void main( T & );
1664        thread_desc * get_thread( T & );
1665        void ^?{}( T & `mutex` );
1666};
1667\end{cfa}
1668\end{tabular}
1669\end{cquote}
1670Like 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).
1671Similarly, 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.
1672(The qualifier @mutex@ for the destructor parameter is discussed in Section~\ref{s:Monitor}.)
1673The 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;
1674whereas, a thread is scheduling for execution in @main@ immediately after its constructor is run.
1675No 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.
1676
1677
1678\section{Mutual Exclusion / Synchronization}
1679\label{s:MutualExclusionSynchronization}
1680
1681Unrestricted nondeterminism is meaningless as there is no way to know when a result is completed and safe to access.
1682To 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}.
1683The shared data protected by mutual exlusion is called a \newterm{critical section}~\cite{Dijkstra65}, and the protection can be simple (only 1 thread) or complex (only N kinds of threads, \eg group~\cite{Joung00} or readers/writer~\cite{Courtois71}).
1684Without synchronization control in a critical section, an arriving thread can barge ahead of preexisting waiter threads resulting in short/long-term starvation, staleness/freshness problems, and/or incorrect transfer of data.
1685Preventing or detecting barging is a challenge with low-level locks, but made easier through higher-level constructs.
1686This challenge is often split into two different approaches: barging \emph{avoidance} and \emph{prevention}.
1687Approaches that unconditionally releasing a lock for competing threads to acquire must use barging avoidance with flag/counter variable(s) to force barging threads to wait;
1688approaches that conditionally hold locks during synchronization, \eg baton-passing~\cite{Andrews89}, prevent barging completely.
1689
1690At the lowest level, concurrent control is provided by atomic operations, upon which different kinds of locking mechanisms are constructed, \eg spin locks, semaphores~\cite{Dijkstra68b}, barriers, and path expressions~\cite{Campbell74}.
1691However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
1692A significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
1693Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
1694Some concurrent systems eliminate mutable shared-state by switching to non-shared 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).
1695However, these approaches introduce a new communication mechanism for concurrency different from the standard communication using function call/return.
1696Hence, a programmer must learn and manipulate two sets of design/programming patterns.
1697While this distinction can be hidden away in library code, effective use of the library still has to take both paradigms into account.
1698In contrast, approaches based on shared-state models more closely resemble the standard call/return programming model, resulting in a single programming paradigm.
1699Finally, a newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
1700While 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~\cite{Cascaval08,Boehm09} to be the main concurrency paradigm for system languages.
1701
1702
1703\section{Monitor}
1704\label{s:Monitor}
1705
1706One of the most natural, elegant, efficient, high-level mechanisms for mutual exclusion and synchronization for shared-memory systems is the \emph{monitor} (Table~\ref{t:ExecutionPropertyComposition} case 2).
1707First 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}.
1708In 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 manually implement a monitor.
1709For these reasons, \CFA selected monitors as the core high-level concurrency construct, upon which higher-level approaches can be easily constructed.
1710
1711Specifically, a \textbf{monitor} is a set of functions that ensure mutual exclusion when accessing shared state.
1712More precisely, a monitor is a programming technique that implicitly binds mutual exclusion to static function scope by call/return, as opposed to locks, where mutual-exclusion is defined by acquire/release calls, independent of lexical context (analogous to block and heap storage allocation).
1713Restricting acquire/release points eases programming, comprehension, and maintenance, at a slight cost in flexibility and efficiency.
1714\CFA uses a custom @monitor@ type and leverages declaration semantics (deallocation) to protect active or waiting threads in a monitor.
1715
1716The following is a \CFA monitor implementation of an atomic counter.
1717\begin{cfa}
1718`monitor` Aint { int cnt; }; $\C[4.25in]{// atomic integer counter}$
1719int ++?( Aint & `mutex` this ) with( this ) { return ++cnt; } $\C{// increment}$
1720int ?=?( Aint & `mutex` lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions with int, mutex optional}\CRT$
1721int ?=?( int & lhs, Aint & `mutex` rhs ) with( rhs ) { lhs = cnt; }
1722\end{cfa}
1723The operators use the parameter-only declaration type-qualifier @mutex@ to mark which parameters require locking during function execution to protect from race conditions.
1724The assignment operators provide bidirectional conversion between an atomic and normal integer without accessing field @cnt@.
1725(These operations only need @mutex@, if reading/writing the implementation type is not atomic.)
1726The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics.
1727\begin{cfa}
1728int i = 0, j = 0, k = 5;
1729Aint x = { 0 }, y = { 0 }, z = { 5 }; $\C{// no mutex required}$
1730++x; ++y; ++z; $\C{// safe increment by multiple threads}$
1731x = 2; y = i; z = k; $\C{// conversions}$
1732i = x; j = y; k = z;
1733\end{cfa}
1734Note, like other concurrent programming languages, \CFA has specializations for the basic types using atomic instructions for performance and a general trait similar to the \CC template @std::atomic@.
1735
1736\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.
1737\newpage
1738\begin{cfa}
1739monitor M { ... } m;
1740void foo( M & mutex m ) { ... } $\C{// acquire mutual exclusion}$
1741void bar( M & mutex m ) { $\C{// acquire mutual exclusion}$
1742        ... `bar( m );` ... `foo( m );` ... $\C{// reacquire mutual exclusion}$
1743}
1744\end{cfa}
1745\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.
1746Similar safety is offered by \emph{explicit} opt-in disciplines like \CC RAII versus the monitor \emph{implicit} language-enforced safety guarantee ensuring no programmer usage errors.
1747Furthermore, 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;
1748RAII is purely a mutual-exclusion mechanism (see Section~\ref{s:Scheduling}).
1749
1750
1751\subsection{Monitor Implementation}
1752
1753For the same design reasons, \CFA provides a custom @monitor@ type and a @trait@ to enforce and restrict the monitor-interface functions.
1754\begin{cquote}
1755\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
1756\begin{cfa}
1757monitor M {
1758        ... // shared data
1759};
1760
1761\end{cfa}
1762&
1763\begin{cfa}
1764trait is_monitor( `dtype` T ) {
1765        monitor_desc * get_monitor( T & );
1766        void ^?{}( T & mutex );
1767};
1768\end{cfa}
1769\end{tabular}
1770\end{cquote}
1771The @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).
1772Similarly, 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.
1773The custom monitor type also inserts any locks needed to implement the mutual exclusion semantics.
1774
1775
1776\subsection{Mutex Acquisition}
1777\label{s:MutexAcquisition}
1778
1779While the monitor lock provides mutual exclusion for shared data, there are implementation options for when and where the locking/unlocking occurs.
1780(Much of this discussion also applies to basic locks.)
1781For example, a monitor may be passed through multiple helper functions before it is necessary to acquire the monitor's mutual exclusion.
1782
1783\CFA requires programmers to identify the kind of parameter with the @mutex@ keyword and uses no keyword to mean \lstinline[morekeywords=nomutex]@nomutex@, because @mutex@ parameters are rare and no keyword is the \emph{normal} parameter semantics.
1784Hence, @mutex@ parameters are documentation, at the function and its prototype, to both programmer and compiler, without other redundant keywords.
1785Furthermore, \CFA relies heavily on traits as an abstraction mechanism, so the @mutex@ qualifier prevents coincidentally matching of a monitor trait with a type that is not a monitor, similar to coincidental inheritance where a shape and playing card can both be drawable.
1786
1787The next semantic decision is establishing which parameter \emph{types} may be qualified with @mutex@.
1788The following has monitor parameter types that are composed of multiple objects.
1789\begin{cfa}
1790monitor M { ... }
1791int f1( M & mutex m ); $\C{// single parameter object}$
1792int f2( M * mutex m ); $\C{// single or multiple parameter object}$
1793int f3( M * mutex m[$\,$] ); $\C{// multiple parameter object}$
1794int f4( stack( M * ) & mutex m ); $\C{// multiple parameters object}$
1795\end{cfa}
1796Function @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.
1797Function @f3@ has a multiple object matrix, and @f4@ a multiple object data structure.
1798While shown shortly, multiple object acquisition is possible, but the number of objects must be statically known.
1799Therefore, \CFA only acquires one monitor per parameter with exactly one level of indirection, and exclude pointer types to unknown sized arrays.
1800
1801For object-oriented monitors, \eg Java, calling a mutex member \emph{implicitly} acquires mutual exclusion of the receiver object, @`rec`.foo(...)@.
1802\CFA has no receiver, and hence, the explicit @mutex@ qualifier is used to specify which objects acquire mutual exclusion.
1803A positive consequence of this design decision is the ability to support multi-monitor functions,\footnote{
1804While object-oriented monitors can be extended with a mutex qualifier for multiple-monitor members, no prior example of this feature could be found.}
1805called \newterm{bulk acquire}.
1806\CFA guarantees bulk acquisition order is consistent across calls to @mutex@ functions using the same monitors as arguments, so acquiring multiple monitors in a bulk acquire is safe from deadlock.
1807Figure~\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.
1808A \CFA programmer only has to manage when to acquire mutual exclusion;
1809a \CC programmer must select the correct lock and acquisition mechanism from a panoply of locking options.
1810Making good choices for common cases in \CFA simplifies the programming experience and enhances safety.
1811
1812\begin{figure}
1813\centering
1814\begin{lrbox}{\myboxA}
1815\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1816monitor BankAccount {
1817
1818        int balance;
1819} b1 = { 0 }, b2 = { 0 };
1820void deposit( BankAccount & `mutex` b,
1821                        int deposit ) with(b) {
1822        balance += deposit;
1823}
1824void transfer( BankAccount & `mutex` my,
1825        BankAccount & `mutex` your, int me2you ) {
1826        // bulk acquire
1827        deposit( my, -me2you ); // debit
1828        deposit( your, me2you ); // credit
1829}
1830`thread` Person { BankAccount & b1, & b2; };
1831void main( Person & person ) with(person) {
1832        for ( 10_000_000 ) {
1833                if ( random() % 3 ) deposit( b1, 3 );
1834                if ( random() % 3 ) transfer( b1, b2, 7 );
1835        }
1836}   
1837int main() {
1838        `Person p1 = { b1, b2 }, p2 = { b2, b1 };`
1839
1840} // wait for threads to complete
1841\end{cfa}
1842\end{lrbox}
1843
1844\begin{lrbox}{\myboxB}
1845\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1846struct BankAccount {
1847        `recursive_mutex m;`
1848        int balance = 0;
1849} b1, b2;
1850void deposit( BankAccount & b, int deposit ) {
1851        `scoped_lock lock( b.m );`
1852        b.balance += deposit;
1853}
1854void transfer( BankAccount & my,
1855                        BankAccount & your, int me2you ) {
1856        `scoped_lock lock( my.m, your.m );` // bulk acquire
1857        deposit( my, -me2you ); // debit
1858        deposit( your, me2you ); // credit
1859}
1860
1861void person( BankAccount & b1, BankAccount & b2 ) {
1862        for ( int i = 0; i < 10$'$000$'$000; i += 1 ) {
1863                if ( random() % 3 ) deposit( b1, 3 );
1864                if ( random() % 3 ) transfer( b1, b2, 7 );
1865        }
1866}   
1867int main() {
1868        `thread p1(person, ref(b1), ref(b2)), p2(person, ref(b2), ref(b1));`
1869        `p1.join(); p2.join();`
1870}
1871\end{cfa}
1872\end{lrbox}
1873
1874\subfloat[\CFA]{\label{f:CFABank}\usebox\myboxA}
1875\hspace{3pt}
1876\vrule
1877\hspace{3pt}
1878\subfloat[\CC]{\label{f:C++Bank}\usebox\myboxB}
1879\hspace{3pt}
1880\caption{Bank transfer problem}
1881\label{f:BankTransfer}
1882\end{figure}
1883
1884Users can still force the acquiring order by using or not using @mutex@.
1885\begin{cfa}
1886void foo( M & mutex m1, M & mutex m2 ); $\C{// acquire m1 and m2}$
1887void bar( M & mutex m1, M & m2 ) { $\C{// only acquire m1}$
1888        ... foo( m1, m2 ); ... $\C{// acquire m2}$
1889}
1890void baz( M & m1, M & mutex m2 ) { $\C{// only acquire m2}$
1891        ... foo( m1, m2 ); ... $\C{// acquire m1}$
1892}
1893\end{cfa}
1894The bulk-acquire semantics allow @bar@ or @baz@ to acquire a monitor lock and reacquire it in @foo@.
1895The calls to @bar@ and @baz@ acquired the monitors in opposite order, possibly resulting in deadlock.
1896However, this case is the simplest instance of the \emph{nested-monitor problem}~\cite{Lister77}, where monitors are acquired in sequence versus bulk.
1897Detecting the nested-monitor problem requires dynamic tracking of monitor calls, and dealing with it requires rollback semantics~\cite{Dice10}.
1898\CFA does not deal with this fundamental problem.
1899
1900Finally, like Java, \CFA offers an alternative @mutex@ statement to reduce refactoring and naming.
1901\begin{cquote}
1902\renewcommand{\arraystretch}{0.0}
1903\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
1904\multicolumn{1}{c}{\textbf{\lstinline@mutex@ call}} & \multicolumn{1}{c}{\lstinline@mutex@ \textbf{statement}} \\
1905\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1906monitor M { ... };
1907void foo( M & mutex m1, M & mutex m2 ) {
1908        // critical section
1909}
1910void bar( M & m1, M & m2 ) {
1911        foo( m1, m2 );
1912}
1913\end{cfa}
1914&
1915\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1916
1917void bar( M & m1, M & m2 ) {
1918        mutex( m1, m2 ) {       // remove refactoring and naming
1919                // critical section
1920        }
1921}
1922
1923\end{cfa}
1924\end{tabular}
1925\end{cquote}
1926
1927
1928\subsection{Scheduling}
1929\label{s:Scheduling}
1930
1931% 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.
1932% 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.
1933This section discusses scheduling for waiting threads eligible for monitor entry, \ie which user thread gets the shared resource next. (See Section~\ref{s:RuntimeStructureCluster} for scheduling kernel threads on virtual processors.)
1934While monitor mutual-exclusion provides safe access to its shared data, the data may indicate a thread cannot proceed, \eg a bounded buffer may be full/\-empty so produce/consumer threads must block.
1935Leaving the monitor and retrying (busy waiting) is impractical for high-level programming.
1936
1937Monitors eliminate busy waiting by providing synchronization within the monitor critical-section to schedule threads needing access to the shared data, where threads block versus spin.
1938Synchronization is generally achieved with internal~\cite{Hoare74} or external~\cite[\S~2.9.2]{uC++} scheduling.
1939\newterm{Internal} (largely) schedules threads located \emph{inside} the monitor and is accomplished using condition variables with signal and wait.
1940\newterm{External} (largely) schedules threads located \emph{outside} the monitor and is accomplished with the @waitfor@ statement.
1941Note, internal scheduling has a small amount of external scheduling and vice versus, so the naming denotes where the majority of the block threads reside (inside or outside) for scheduling.
1942For complex scheduling, the approaches can be combined, so there can be an equal number of threads waiting inside and outside.
1943
1944\CFA monitors do not allow calling threads to barge ahead of signalled threads (via barging prevention), which simplifies synchronization among threads in the monitor and increases correctness.
1945A direct consequence of this semantics is that unblocked waiting threads are not required to recheck the waiting condition, \ie waits are not in a starvation-prone busy-loop as required by the signals-as-hints style with barging.
1946Preventing barging comes directly from Hoare's semantics in the seminal paper on monitors~\cite[p.~550]{Hoare74}.
1947% \begin{cquote}
1948% 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.
1949% 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}
1950% \end{cquote}
1951Furthermore, \CFA concurrency has no spurious wakeup~\cite[\S~9]{Buhr05a}, which eliminates an implicit self barging.
1952
1953Monitor mutual-exclusion means signalling cannot have the signaller and signalled thread in the monitor simultaneously, so only the signaller or signallee can proceed.
1954Figure~\ref{f:MonitorScheduling} shows internal/external scheduling for the bounded-buffer examples in Figure~\ref{f:GenericBoundedBuffer}.
1955For internal scheduling in Figure~\ref{f:BBInt}, the @signal@ moves the signallee (front thread of the specified condition queue) to urgent and the signaller continues (solid line).
1956Multiple signals move multiple signallees to urgent until the condition queue is empty.
1957When the signaller exits or waits, a thread is implicitly unblocked from urgent (if available) before unblocking a calling thread to prevent barging.
1958(Java conceptually moves the signalled thread to the calling queue, and hence, allows barging.)
1959Signal is used when the signaller is providing the cooperation needed by the signallee (\eg creating an empty slot in a buffer for a producer) and the signaller immediately exits the monitor to run concurrently (consume the buffer element) and passes control of the monitor to the signalled thread, which can immediately take advantage of the state change.
1960Specifically, the @wait@ function atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the function's parameter list.
1961Signalling is unconditional because signalling an empty condition queue does nothing.
1962It is common to declare condition queues as monitor fields to prevent shared access, hence no locking is required for access as the queues are protected by the monitor lock.
1963In \CFA, a condition queue can be created/stored independently.
1964
1965\begin{figure}
1966\centering
1967% \subfloat[Scheduling Statements] {
1968% \label{fig:SchedulingStatements}
1969% {\resizebox{0.45\textwidth}{!}{\input{CondSigWait.pstex_t}}}
1970\input{CondSigWait.pstex_t}
1971% }% subfloat
1972% \quad
1973% \subfloat[Bulk acquire monitor] {
1974% \label{fig:BulkMonitor}
1975% {\resizebox{0.45\textwidth}{!}{\input{ext_monitor.pstex_t}}}
1976% }% subfloat
1977\caption{Monitor Scheduling}
1978\label{f:MonitorScheduling}
1979\end{figure}
1980
1981\begin{figure}
1982\centering
1983\newbox\myboxA
1984\begin{lrbox}{\myboxA}
1985\begin{cfa}[aboveskip=0pt,belowskip=0pt]
1986forall( otype T ) { // distribute forall
1987        monitor Buffer {
1988                `condition` full, empty;
1989                int front, back, count;
1990                T elements[10];
1991        };
1992        void ?{}( Buffer(T) & buf ) with(buf) {
1993                front = back = count = 0;
1994        }
1995
1996        void insert(Buffer(T) & mutex buf, T elm) with(buf){
1997                if ( count == 10 ) `wait( empty )`; // full ?
1998                // insert elm into buf
1999                `signal( full )`;
2000        }
2001        T remove( Buffer(T) & mutex buf ) with(buf) {
2002                if ( count == 0 ) `wait( full )`; // empty ?
2003                // remove elm from buf
2004                `signal( empty )`;
2005                return elm;
2006        }
2007}
2008\end{cfa}
2009\end{lrbox}
2010
2011\newbox\myboxB
2012\begin{lrbox}{\myboxB}
2013\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2014forall( otype T ) { // distribute forall
2015        monitor Buffer {
2016
2017                int front, back, count;
2018                T elements[10];
2019        };
2020        void ?{}( Buffer(T) & buf ) with(buf) {
2021                front = back = count = 0;
2022        }
2023        T remove( Buffer(T) & mutex buf ); // forward
2024        void insert(Buffer(T) & mutex buf, T elm) with(buf){
2025                if ( count == 10 ) `waitfor( remove : buf )`;
2026                // insert elm into buf
2027
2028        }
2029        T remove( Buffer(T) & mutex buf ) with(buf) {
2030                if ( count == 0 ) `waitfor( insert : buf )`;
2031                // remove elm from buf
2032
2033                return elm;
2034        }
2035}
2036\end{cfa}
2037\end{lrbox}
2038
2039\subfloat[Internal scheduling]{\label{f:BBInt}\usebox\myboxA}
2040\hspace{1pt}
2041\vrule
2042\hspace{3pt}
2043\subfloat[External scheduling]{\label{f:BBExt}\usebox\myboxB}
2044
2045\caption{Generic bounded buffer}
2046\label{f:GenericBoundedBuffer}
2047\end{figure}
2048
2049The @signal_block@ provides the opposite unblocking order, where the signaller is moved to urgent and the signallee continues and a thread is implicitly unblocked from urgent when the signallee exits or waits (dashed line).
2050Signal block is used when the signallee is providing the cooperation needed by the signaller (\eg if the buffer is removed and a producer hands off an item to a consumer, as in Figure~\ref{f:DatingSignalBlock}) so the signaller must wait until the signallee unblocks, provides the cooperation, exits the monitor to run concurrently, and passes control of the monitor to the signaller, which can immediately take advantage of the state change.
2051Using @signal@ or @signal_block@ can be a dynamic decision based on whether the thread providing the cooperation arrives before or after the thread needing the cooperation.
2052
2053External scheduling in Figure~\ref{f:BBExt} simplifies internal scheduling by eliminating condition queues and @signal@/@wait@ (cases where it cannot are discussed shortly), and has existed in the programming language Ada for almost 40 years with variants in other languages~\cite{SR,ConcurrentC++,uC++}.
2054While prior languages use external scheduling solely for thread interaction, \CFA generalizes it to both monitors and threads.
2055External scheduling allows waiting for events from other threads while restricting unrelated events, that would otherwise have to wait on condition queues in the monitor.
2056Scheduling 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.
2057Specifically, a thread calling the monitor is unblocked directly from the calling queue based on function names that can fulfill the cooperation required by the signaller.
2058(The linear search through the calling queue to locate a particular call can be reduced to $O(1)$.)
2059Hence, the @waitfor@ has the same semantics as @signal_block@, where the signallee thread from the calling queue executes before the signaller, which waits on urgent.
2060Now when a producer/consumer detects a full/empty buffer, the necessary cooperation for continuation is specified by indicating the next function call that can occur.
2061For example, a producer detecting a full buffer must have cooperation from a consumer to remove an item so function @remove@ is accepted, which prevents producers from entering the monitor, and after a consumer calls @remove@, the producer waiting on urgent is \emph{implicitly} unblocked because it can now continue its insert operation.
2062Hence, this mechanism is done in terms of control flow, next call, versus in terms of data, channels, as in Go/Rust @select@.
2063While both mechanisms have strengths and weaknesses, \CFA uses the control-flow mechanism to be consistent with other language features.
2064
2065Figure~\ref{f:ReadersWriterLock} shows internal/external scheduling for a readers/writer lock with no barging and threads are serviced in FIFO order to eliminate staleness/freshness among the reader/writer threads.
2066For internal scheduling in Figure~\ref{f:RWInt}, the readers and writers wait on the same condition queue in FIFO order, making it impossible to tell if a waiting thread is a reader or writer.
2067To clawback the kind of thread, a \CFA condition can store user data in the node for a blocking thread at the @wait@, \ie whether the thread is a @READER@ or @WRITER@.
2068An unblocked reader thread checks if the thread at the front of the queue is a reader and unblock it, \ie the readers daisy-chain signal the next group of readers demarcated by the next writer or end of the queue.
2069For external scheduling in Figure~\ref{f:RWExt}, a waiting reader checks if a writer is using the resource, and if so, restricts further calls until the writer exits by calling @EndWrite@.
2070The writer does a similar action for each reader or writer using the resource.
2071Note, no new calls to @StartRead@/@StartWrite@ may occur when waiting for the call to @EndRead@/@EndWrite@.
2072
2073\begin{figure}
2074\centering
2075\newbox\myboxA
2076\begin{lrbox}{\myboxA}
2077\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2078enum RW { READER, WRITER };
2079monitor ReadersWriter {
2080        int rcnt, wcnt; // readers/writer using resource
2081        `condition RWers;`
2082};
2083void ?{}( ReadersWriter & rw ) with(rw) {
2084        rcnt = wcnt = 0;
2085}
2086void EndRead( ReadersWriter & mutex rw ) with(rw) {
2087        rcnt -= 1;
2088        if ( rcnt == 0 ) `signal( RWers )`;
2089}
2090void EndWrite( ReadersWriter & mutex rw ) with(rw) {
2091        wcnt = 0;
2092        `signal( RWers );`
2093}
2094void StartRead( ReadersWriter & mutex rw ) with(rw) {
2095        if ( wcnt !=0 || ! empty( RWers ) )
2096                `wait( RWers, READER )`;
2097        rcnt += 1;
2098        if ( ! empty(RWers) && `front(RWers) == READER` )
2099                `signal( RWers )`;  // daisy-chain signalling
2100}
2101void StartWrite( ReadersWriter & mutex rw ) with(rw) {
2102        if ( wcnt != 0 || rcnt != 0 ) `wait( RWers, WRITER )`;
2103
2104        wcnt = 1;
2105}
2106\end{cfa}
2107\end{lrbox}
2108
2109\newbox\myboxB
2110\begin{lrbox}{\myboxB}
2111\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2112
2113monitor ReadersWriter {
2114        int rcnt, wcnt; // readers/writer using resource
2115
2116};
2117void ?{}( ReadersWriter & rw ) with(rw) {
2118        rcnt = wcnt = 0;
2119}
2120void EndRead( ReadersWriter & mutex rw ) with(rw) {
2121        rcnt -= 1;
2122
2123}
2124void EndWrite( ReadersWriter & mutex rw ) with(rw) {
2125        wcnt = 0;
2126
2127}
2128void StartRead( ReadersWriter & mutex rw ) with(rw) {
2129        if ( wcnt > 0 ) `waitfor( EndWrite : rw );`
2130
2131        rcnt += 1;
2132
2133
2134}
2135void StartWrite( ReadersWriter & mutex rw ) with(rw) {
2136        if ( wcnt > 0 ) `waitfor( EndWrite : rw );`
2137        else while ( rcnt > 0 ) `waitfor( EndRead : rw );`
2138        wcnt = 1;
2139}
2140\end{cfa}
2141\end{lrbox}
2142
2143\subfloat[Internal scheduling]{\label{f:RWInt}\usebox\myboxA}
2144\hspace{1pt}
2145\vrule
2146\hspace{3pt}
2147\subfloat[External scheduling]{\label{f:RWExt}\usebox\myboxB}
2148
2149\caption{Readers / writer lock}
2150\label{f:ReadersWriterLock}
2151\end{figure}
2152
2153Finally, external scheduling requires urgent to be a stack, because the signaller expects to execute immediately after the specified monitor call has exited or waited.
2154Internal schedulling performing multiple signalling results in unblocking from urgent in the reverse order from signalling.
2155It is rare for the unblocking order to be important as an unblocked thread can be time-sliced immediately after leaving the monitor.
2156If the unblocking order is important, multiple signalling can be restructured into daisy-chain signalling, where each thread signals the next thread.
2157Hence, \CFA uses a single urgent stack to correctly handle @waitfor@ and adequately support both forms of signalling.
2158(Advanced @waitfor@ features are discussed in Section~\ref{s:ExtendedWaitfor}.)
2159
2160\begin{figure}
2161\centering
2162\newbox\myboxA
2163\begin{lrbox}{\myboxA}
2164\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2165enum { CCodes = 20 };
2166monitor DS {
2167        int GirlPhNo, BoyPhNo;
2168        condition Girls[CCodes], Boys[CCodes];
2169        `condition exchange;`
2170};
2171int girl( DS & mutex ds, int phNo, int ccode ) {
2172        if ( empty( Boys[ccode] ) ) {
2173                wait( Girls[ccode] );
2174                GirlPhNo = phNo;
2175                `signal( exchange );`
2176        } else {
2177                GirlPhNo = phNo;
2178                `signal( Boys[ccode] );`
2179                `wait( exchange );`
2180        }
2181        return BoyPhNo;
2182}
2183int boy( DS & mutex ds, int phNo, int ccode ) {
2184        // as above with boy/girl interchanged
2185}
2186\end{cfa}
2187\end{lrbox}
2188
2189\newbox\myboxB
2190\begin{lrbox}{\myboxB}
2191\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2192
2193monitor DS {
2194        int GirlPhNo, BoyPhNo;
2195        condition Girls[CCodes], Boys[CCodes];
2196
2197};
2198int girl( DS & mutex ds, int phNo, int ccode ) {
2199        if ( empty( Boys[ccode] ) ) { // no compatible
2200                wait( Girls[ccode] ); // wait for boy
2201                GirlPhNo = phNo; // make phone number available
2202
2203        } else {
2204                GirlPhNo = phNo; // make phone number available
2205                `signal_block( Boys[ccode] );` // restart boy
2206
2207        } // if
2208        return BoyPhNo;
2209}
2210int boy( DS & mutex ds, int phNo, int ccode ) {
2211        // as above with boy/girl interchanged
2212}
2213\end{cfa}
2214\end{lrbox}
2215
2216\subfloat[\lstinline@signal@]{\label{f:DatingSignal}\usebox\myboxA}
2217\qquad
2218\subfloat[\lstinline@signal_block@]{\label{f:DatingSignalBlock}\usebox\myboxB}
2219\caption{Dating service Monitor}
2220\label{f:DatingServiceMonitor}
2221\end{figure}
2222
2223Figure~\ref{f:DatingServiceMonitor} shows a dating service demonstrating non-blocking and blocking signalling.
2224The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
2225A thread blocks until an appropriate partner arrives.
2226The complexity is exchanging phone numbers in the monitor because of the mutual-exclusion property.
2227For 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.
2228For signal-block scheduling, the implicit urgent-queue replaces the explicit @exchange@-condition and @signal_block@ puts the finding thread on the urgent stack and unblocks the matcher.
2229
2230The dating service is an important example of a monitor that cannot be written using external scheduling.
2231First, because scheduling requires knowledge of calling parameters to make matching decisions, and parameters of calling threads are unavailable within the monitor.
2232For example, a girl thread within the monitor cannot examine the @ccode@ of boy threads waiting on the calling queue to determine if there is a matching partner.
2233Second, because a scheduling decision may be delayed when there is no immediate match, which requires a condition queue for waiting, and condition queues imply internal scheduling.
2234For example, if a girl thread could determine there is no calling boy with the same @ccode@, it must wait until a matching boy arrives.
2235Finally, 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.
2236This situation shows rechecking the waiting condition and waiting again (signals-as-hints) fails, requiring significant restructured to account for barging.
2237
2238Both internal and external scheduling extend to multiple monitors in a natural way.
2239\begin{cquote}
2240\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
2241\begin{cfa}
2242monitor M { `condition e`; ... };
2243void foo( M & mutex m1, M & mutex m2 ) {
2244        ... wait( `e` ); ...   // wait( e, m1, m2 )
2245        ... wait( `e, m1` ); ...
2246        ... wait( `e, m2` ); ...
2247}
2248\end{cfa}
2249&
2250\begin{cfa}
2251void rtn$\(_1\)$( M & mutex m1, M & mutex m2 ); // overload rtn
2252void rtn$\(_2\)$( M & mutex m1 );
2253void bar( M & mutex m1, M & mutex m2 ) {
2254        ... waitfor( `rtn`${\color{red}\(_1\)}$ ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\) : m1, m2 )}$
2255        ... waitfor( `rtn${\color{red}\(_2\)}$ : m1` ); ...
2256}
2257\end{cfa}
2258\end{tabular}
2259\end{cquote}
2260For @wait( e )@, the default semantics is to atomically block the signaller and release all acquired mutex parameters, \ie @wait( e, m1, m2 )@.
2261To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @wait( e, m1 )@.
2262Wait cannot statically verify the released monitors are the acquired mutex-parameters without disallowing separately compiled helper functions calling @wait@.
2263While \CC supports bulk locking, @wait@ only accepts a single lock for a condition queue, so bulk locking with condition queues is asymmetric.
2264Finally, a signaller,
2265\begin{cfa}
2266void baz( M & mutex m1, M & mutex m2 ) {
2267        ... signal( e ); ...
2268}
2269\end{cfa}
2270must 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.
2271
2272Similarly, for @waitfor( rtn )@, the default semantics is to atomically block the acceptor and release all acquired mutex parameters, \ie @waitfor( rtn : m1, m2 )@.
2273To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn : m1 )@.
2274@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.
2275% When an overloaded function appears in an @waitfor@ statement, calls to any function with that name are accepted.
2276% The rationale is that members with the same name should perform a similar function, and therefore, all should be eligible to accept a call.
2277Overloaded functions can be disambiguated using a cast
2278\begin{cfa}
2279void rtn( M & mutex m );
2280`int` rtn( M & mutex m );
2281waitfor( (`int` (*)( M & mutex ))rtn : m );
2282\end{cfa}
2283
2284The ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock (see Section~\ref{s:MutexAcquisition}).
2285\newpage
2286\begin{cfa}
2287void foo( M & mutex m1, M & mutex m2 ) {
2288        ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired}$
2289void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2}$
2290        ... signal( `e` ); ...
2291\end{cfa}
2292The @wait@ only releases @m1@ so the signalling thread cannot acquire @m1@ and @m2@ to enter @bar@ and @signal@ the condition.
2293While deadlock can occur with multiple/nesting acquisition, this is a consequence of locks, and by extension monitor locking is not perfectly composable.
2294
2295
2296\subsection{\texorpdfstring{Extended \protect\lstinline@waitfor@}{Extended waitfor}}
2297\label{s:ExtendedWaitfor}
2298
2299Figure~\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.
2300For a @waitfor@ clause to be executed, its @when@ must be true and an outstanding call to its corresponding member(s) must exist.
2301The \emph{conditional-expression} of a @when@ may call a function, but the function must not block or context switch.
2302If 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@.
2303If 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.
2304If there is a @timeout@ clause, it provides an upper bound on waiting.
2305If 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.
2306Hence, the terminating @else@ clause allows a conditional attempt to accept a call without blocking.
2307If both @timeout@ and @else@ clause are present, the @else@ must be conditional, or the @timeout@ is never triggered.
2308There is also a traditional future wait queue (not shown) (\eg Microsoft @WaitForMultipleObjects@), to wait for a specified number of future elements in the queue.
2309Finally, there is a shorthand for specifying multiple functions using the same set of monitors: @waitfor( f, g, h : m1, m2, m3 )@.
2310
2311\begin{figure}
2312\centering
2313\begin{cfa}
2314`when` ( $\emph{conditional-expression}$ )      $\C{// optional guard}$
2315        waitfor( $\emph{mutex-member-name}$ ) $\emph{statement}$ $\C{// action after call}$
2316`or` `when` ( $\emph{conditional-expression}$ ) $\C{// any number of functions}$
2317        waitfor( $\emph{mutex-member-name}$ ) $\emph{statement}$
2318`or`    ...
2319`when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
2320        `timeout` $\emph{statement}$ $\C{// optional terminating timeout clause}$
2321`when` ( $\emph{conditional-expression}$ ) $\C{// optional guard}$
2322        `else`  $\emph{statement}$ $\C{// optional terminating clause}$
2323\end{cfa}
2324\caption{Extended \protect\lstinline@waitfor@}
2325\label{f:ExtendedWaitfor}
2326\end{figure}
2327
2328Note, a group of conditional @waitfor@ clauses is \emph{not} the same as a group of @if@ statements, \eg:
2329\begin{cfa}
2330if ( C1 ) waitfor( mem1 );                       when ( C1 ) waitfor( mem1 );
2331else if ( C2 ) waitfor( mem2 );         or when ( C2 ) waitfor( mem2 );
2332\end{cfa}
2333The left example only accepts @mem1@ if @C1@ is true or only @mem2@ if @C2@ is true.
2334The right example accepts either @mem1@ or @mem2@ if @C1@ and @C2@ are true.
2335
2336An interesting use of @waitfor@ is accepting the @mutex@ destructor to know when an object is deallocated, \eg assume the bounded buffer is restructured from a monitor to a thread with the following @main@.
2337\begin{cfa}
2338void main( Buffer(T) & buffer ) with(buffer) {
2339        for () {
2340                `waitfor( ^?{} : buffer )` break;
2341                or when ( count != 20 ) waitfor( insert : buffer ) { ... }
2342                or when ( count != 0 ) waitfor( remove : buffer ) { ... }
2343        }
2344        // clean up
2345}
2346\end{cfa}
2347When the program main deallocates the buffer, it first calls the buffer's destructor, which is accepted, the destructor runs, and the buffer is deallocated.
2348However, the buffer thread cannot continue after the destructor call because the object is gone;
2349hence, clean up in @main@ cannot occur, which means destructors for local objects are not run.
2350To 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.
2351Then, the destructor caller unblocks from urgent to deallocate the object.
2352Accepting the destructor is the idiomatic way in \CFA to terminate a thread performing direct communication.
2353
2354
2355\subsection{Bulk Barging Prevention}
2356
2357Figure~\ref{f:BulkBargingPrevention} shows \CFA code where bulk acquire adds complexity to the internal-signalling semantics.
2358The complexity begins at the end of the inner @mutex@ statement, where the semantics of internal scheduling need to be extended for multiple monitors.
2359The problem is that bulk acquire is used in the inner @mutex@ statement where one of the monitors is already acquired.
2360When 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.
2361However, both the signalling and waiting threads W1 and W2 need some subset of monitors @m1@ and @m2@.
2362\begin{cquote}
2363condition c: (order 1) W2(@m2@), W1(@m1@,@m2@)\ \ \ or\ \ \ (order 2) W1(@m1@,@m2@), W2(@m2@) \\
2364S: 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 \\
2365\hspace*{2.75in}$\rightarrow$ rel. @m1@ $\rightarrow$ pass @m1,m2@ unblock W1 (order 1)
2366\end{cquote}
2367
2368\begin{figure}
2369\newbox\myboxA
2370\begin{lrbox}{\myboxA}
2371\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2372monitor M m1, m2;
2373condition c;
2374mutex( m1 ) { // $\LstCommentStyle{\color{red}outer}$
2375        ...
2376        mutex( m1, m2 ) { // $\LstCommentStyle{\color{red}inner}$
2377                ... `signal( c )`; ...
2378                // m1, m2 still acquired
2379        } // $\LstCommentStyle{\color{red}release m2}$
2380        // m1 acquired
2381} // release m1
2382\end{cfa}
2383\end{lrbox}
2384
2385\newbox\myboxB
2386\begin{lrbox}{\myboxB}
2387\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2388
2389
2390mutex( m1 ) {
2391        ...
2392        mutex( m1, m2 ) {
2393                ... `wait( c )`; // release m1, m2
2394                // m1, m2 reacquired
2395        } // $\LstCommentStyle{\color{red}release m2}$
2396        // m1 acquired
2397} // release m1
2398\end{cfa}
2399\end{lrbox}
2400
2401\newbox\myboxC
2402\begin{lrbox}{\myboxC}
2403\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2404
2405
2406mutex( m2 ) {
2407        ... `wait( c )`; // release m2
2408        // m2 reacquired
2409} // $\LstCommentStyle{\color{red}release m2}$
2410
2411
2412
2413
2414\end{cfa}
2415\end{lrbox}
2416
2417\begin{cquote}
2418\subfloat[Signalling Thread (S)]{\label{f:SignallingThread}\usebox\myboxA}
2419\hspace{3\parindentlnth}
2420\subfloat[Waiting Thread (W1)]{\label{f:WaitingThread}\usebox\myboxB}
2421\hspace{2\parindentlnth}
2422\subfloat[Waiting Thread (W2)]{\label{f:OtherWaitingThread}\usebox\myboxC}
2423\end{cquote}
2424\caption{Bulk Barging Prevention}
2425\label{f:BulkBargingPrevention}
2426\end{figure}
2427
2428One 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.
2429However, 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.
2430If W1 waited first, the signaller must retain @m1@ amd @m2@ until completion of the outer mutex statement and then pass both to W1.
2431% Furthermore, there is an execution sequence where the signaller always finds waiter W2, and hence, waiter W1 starves.
2432To support this efficient semantics (and prevent barging), the implementation maintains a list of monitors acquired for each blocked thread.
2433When a signaller exits or waits in a monitor function/statement, the front waiter on urgent is unblocked if all its monitors are released.
2434Implementing a fast subset check for the necessary released monitors is important and discussed in the following sections.
2435% 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.
2436
2437
2438\subsection{\texorpdfstring{\protect\lstinline@waitfor@ Implementation}{waitfor Implementation}}
2439\label{s:waitforImplementation}
2440
2441In a statically-typed object-oriented programming language, a class has an exhaustive list of members, even when members are added via static inheritance (see Figure~\ref{f:uCinheritance}).
2442Knowing all members at compilation (even separate compilation) allows uniquely numbered them so the accept-statement implementation can use a fast/compact bit mask with $O(1)$ compare.
2443
2444\begin{figure}
2445\centering
2446\begin{lrbox}{\myboxA}
2447\begin{uC++}[aboveskip=0pt,belowskip=0pt]
2448$\emph{translation unit 1}$
2449_Monitor B { // common type in .h file
2450        _Mutex virtual void `f`( ... );
2451        _Mutex virtual void `g`( ... );
2452        _Mutex virtual void w1( ... ) { ... _Accept(`f`, `g`); ... }
2453};
2454$\emph{translation unit 2}$
2455// include B
2456_Monitor D : public B { // inherit
2457        _Mutex void `h`( ... ); // add
2458        _Mutex void w2( ... ) { ... _Accept(`f`, `h`); ... }
2459};
2460\end{uC++}
2461\end{lrbox}
2462
2463\begin{lrbox}{\myboxB}
2464\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2465$\emph{translation unit 1}$
2466monitor M { ... }; // common type in .h file
2467void `f`( M & mutex m, ... );
2468void `g`( M & mutex m, ... );
2469void w1( M & mutex m, ... ) { ... waitfor(`f`, `g` : m); ... }
2470
2471$\emph{translation unit 2}$
2472// include M
2473extern void `f`( M & mutex m, ... ); // import f but not g
2474void `h`( M & mutex m ); // add
2475void w2( M & mutex m, ... ) { ... waitfor(`f`, `h` : m); ... }
2476
2477\end{cfa}
2478\end{lrbox}
2479
2480\subfloat[\uC]{\label{f:uCinheritance}\usebox\myboxA}
2481\hspace{3pt}
2482\vrule
2483\hspace{3pt}
2484\subfloat[\CFA]{\label{f:CFinheritance}\usebox\myboxB}
2485\caption{Member / Function visibility}
2486\label{f:MemberFunctionVisibility}
2487\end{figure}
2488
2489However, the @waitfor@ statement in translation unit 2 (see Figure~\ref{f:CFinheritance}) cannot see function @g@ in translation unit 1 precluding a unique numbering for a bit-mask because the monitor type only carries the protected shared-data.
2490(A possible way to construct a dense mapping is at link or load-time.)
2491Hence, function pointers are used to identify the functions listed in the @waitfor@ statement, stored in a variable-sized array.
2492Then, the same implementation approach used for the urgent stack (see Section~\ref{s:Scheduling}) is used for the calling queue.
2493Each caller has a list of monitors acquired, and the @waitfor@ statement performs a (short) linear search matching functions in the @waitfor@ list with called functions, and then verifying the associated mutex locks can be transfers.
2494
2495
2496\subsection{Multi-Monitor Scheduling}
2497\label{s:Multi-MonitorScheduling}
2498
2499External scheduling, like internal scheduling, becomes significantly more complex for multi-monitor semantics.
2500Even in the simplest case, new semantics need to be established.
2501\begin{cfa}
2502monitor M { ... };
2503void f( M & mutex m1 );
2504void g( M & mutex m1, M & mutex m2 ) { `waitfor( f );` } $\C{// pass m1 or m2 to f?}$
2505\end{cfa}
2506The solution is for the programmer to disambiguate:
2507\begin{cfa}
2508waitfor( f : `m2` ); $\C{// wait for call to f with argument m2}$
2509\end{cfa}
2510Both 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@.
2511This behaviour can be extended to the multi-monitor @waitfor@ statement.
2512\begin{cfa}
2513monitor M { ... };
2514void f( M & mutex m1, M & mutex m2 );
2515void g( M & mutex m1, M & mutex m2 ) { waitfor( f : `m1, m2` ); $\C{// wait for call to f with arguments m1 and m2}$
2516\end{cfa}
2517Again, the set of monitors passed to the @waitfor@ statement must be entirely contained in the set of monitors already acquired by the accepting function.
2518% Also, the order of the monitors in a @waitfor@ statement must match the order of the mutex parameters.
2519
2520Figure~\ref{f:UnmatchedMutexSets} shows internal and external scheduling with multiple monitors that must match exactly with a signalling or accepting thread, \ie partial matching results in waiting.
2521In both cases, the set of monitors is disjoint so unblocking is impossible.
2522
2523\begin{figure}
2524\centering
2525\begin{lrbox}{\myboxA}
2526\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2527monitor M1 {} m11, m12;
2528monitor M2 {} m2;
2529condition c;
2530void f( M1 & mutex m1, M2 & mutex m2 ) {
2531        signal( c );
2532}
2533void g( M1 & mutex m1, M2 & mutex m2 ) {
2534        wait( c );
2535}
2536g( `m11`, m2 ); // block on wait
2537f( `m12`, m2 ); // cannot fulfil
2538\end{cfa}
2539\end{lrbox}
2540
2541\begin{lrbox}{\myboxB}
2542\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2543monitor M1 {} m11, m12;
2544monitor M2 {} m2;
2545
2546void f( M1 & mutex m1, M2 & mutex m2 ) {
2547
2548}
2549void g( M1 & mutex m1, M2 & mutex m2 ) {
2550        waitfor( f : m1, m2 );
2551}
2552g( `m11`, m2 ); // block on accept
2553f( `m12`, m2 ); // cannot fulfil
2554\end{cfa}
2555\end{lrbox}
2556\subfloat[Internal scheduling]{\label{f:InternalScheduling}\usebox\myboxA}
2557\hspace{3pt}
2558\vrule
2559\hspace{3pt}
2560\subfloat[External scheduling]{\label{f:ExternalScheduling}\usebox\myboxB}
2561\caption{Unmatched \protect\lstinline@mutex@ sets}
2562\label{f:UnmatchedMutexSets}
2563\end{figure}
2564
2565\begin{figure}
2566\centering
2567\begin{lrbox}{\myboxA}
2568\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2569
2570struct Msg { int i, j; };
2571monitor thread GoRtn { int i;  float f;  Msg m; };
2572void mem1( GoRtn & mutex gortn, int i ) { gortn.i = i; }
2573void mem2( GoRtn & mutex gortn, float f ) { gortn.f = f; }
2574void mem3( GoRtn & mutex gortn, Msg m ) { gortn.m = m; }
2575void ^?{}( GoRtn & mutex ) {}
2576
2577void main( GoRtn & gortn ) with( gortn ) {  // thread starts
2578
2579        for () {
2580
2581                `waitfor( mem1 : gortn )` sout | i;  // wait for calls
2582                or `waitfor( mem2 : gortn )` sout | f;
2583                or `waitfor( mem3 : gortn )` sout | m.i | m.j;
2584                or `waitfor( ^?{} : gortn )` break; // low priority
2585
2586        }
2587
2588}
2589int main() {
2590        GoRtn gortn; $\C[2.0in]{// start thread}$
2591        `mem1( gortn, 0 );` $\C{// different calls}\CRT$
2592        `mem2( gortn, 2.5 );`
2593        `mem3( gortn, (Msg){1, 2} );`
2594
2595
2596} // wait for completion
2597\end{cfa}
2598\end{lrbox}
2599
2600\begin{lrbox}{\myboxB}
2601\begin{Go}[aboveskip=0pt,belowskip=0pt]
2602func main() {
2603        type Msg struct{ i, j int }
2604
2605        ch1 := make( chan int )
2606        ch2 := make( chan float32 )
2607        ch3 := make( chan Msg )
2608        hand := make( chan string )
2609        shake := make( chan string )
2610        gortn := func() { $\C[1.5in]{// thread starts}$
2611                var i int;  var f float32;  var m Msg
2612                L: for {
2613                        select { $\C{// wait for messages}$
2614                          case `i = <- ch1`: fmt.Println( i )
2615                          case `f = <- ch2`: fmt.Println( f )
2616                          case `m = <- ch3`: fmt.Println( m )
2617                          case `<- hand`: break L $\C{// sentinel}$
2618                        }
2619                }
2620                `shake <- "SHAKE"` $\C{// completion}$
2621        }
2622
2623        go gortn() $\C{// start thread}$
2624        `ch1 <- 0` $\C{// different messages}$
2625        `ch2 <- 2.5`
2626        `ch3 <- Msg{1, 2}`
2627        `hand <- "HAND"` $\C{// sentinel value}$
2628        `<- shake` $\C{// wait for completion}\CRT$
2629}
2630\end{Go}
2631\end{lrbox}
2632
2633\subfloat[\CFA]{\label{f:CFAwaitfor}\usebox\myboxA}
2634\hspace{3pt}
2635\vrule
2636\hspace{3pt}
2637\subfloat[Go]{\label{f:Gochannel}\usebox\myboxB}
2638\caption{Direct versus indirect communication}
2639\label{f:DirectCommunicationComparison}
2640
2641\medskip
2642
2643\begin{cfa}
2644monitor thread DatingService {
2645        condition Girls[CompCodes], Boys[CompCodes];
2646        int girlPhoneNo, boyPhoneNo, ccode;
2647};
2648int girl( DatingService & mutex ds, int phoneno, int code ) with( ds ) {
2649        girlPhoneNo = phoneno;  ccode = code;
2650        `wait( Girls[ccode] );`                                                         $\C{// wait for boy}$
2651        girlPhoneNo = phoneno;  return boyPhoneNo;
2652}
2653int boy( DatingService & mutex ds, int phoneno, int code ) with( ds ) {
2654        boyPhoneNo = phoneno;  ccode = code;
2655        `wait( Boys[ccode] );`                                                          $\C{// wait for girl}$
2656        boyPhoneNo = phoneno;  return girlPhoneNo;
2657}
2658void main( DatingService & ds ) with( ds ) {                    $\C{// thread starts, ds defaults to mutex}$
2659        for () {
2660                waitfor( ^?{} ) break;                                                  $\C{// high priority}$
2661                or waitfor( girl )                                                              $\C{// girl called, compatible boy ? restart boy then girl}$
2662                        if ( ! is_empty( Boys[ccode] ) ) { `signal_block( Boys[ccode] );  signal_block( Girls[ccode] );` }
2663                or waitfor( boy ) {                                                             $\C{// boy called, compatible girl ? restart girl then boy}$
2664                        if ( ! is_empty( Girls[ccode] ) ) { `signal_block( Girls[ccode] );  signal_block( Boys[ccode] );` }
2665        }
2666}
2667\end{cfa}
2668\caption{Direct communication dating service}
2669\label{f:DirectCommunicationDatingService}
2670\end{figure}
2671
2672\begin{comment}
2673The following shows an example of two threads directly calling each other and accepting calls from each other in a cycle.
2674\begin{cfa}
2675\end{cfa}
2676\vspace{-0.8\baselineskip}
2677\begin{cquote}
2678\begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
2679\begin{cfa}
2680thread Ping {} pi;
2681void ping( Ping & mutex ) {}
2682void main( Ping & pi ) {
2683        for ( 10 ) {
2684                `waitfor( ping : pi );`
2685                `pong( po );`
2686        }
2687}
2688int main() {}
2689\end{cfa}
2690&
2691\begin{cfa}
2692thread Pong {} po;
2693void pong( Pong & mutex ) {}
2694void main( Pong & po ) {
2695        for ( 10 ) {
2696                `ping( pi );`
2697                `waitfor( pong : po );`
2698        }
2699}
2700
2701\end{cfa}
2702\end{tabular}
2703\end{cquote}
2704% \lstMakeShortInline@%
2705% \caption{Threads ping/pong using external scheduling}
2706% \label{f:pingpong}
2707% \end{figure}
2708Note, the ping/pong threads are globally declared, @pi@/@po@, and hence, start (and possibly complete) before the program main starts.
2709\end{comment}
2710
2711
2712\subsection{\texorpdfstring{\protect\lstinline@monitor@ Generators / Coroutines / Threads}{monitor Generators / Coroutines / Threads}}
2713
2714\CFA generators, coroutines, and threads can also be monitors (Table~\ref{t:ExecutionPropertyComposition} cases 4, 6, 12) allowing safe \emph{direct communication} with threads, \ie the custom types can have mutex functions that are called by other threads.
2715All monitor features are available within these mutex functions.
2716For example, if the formatter generator (or coroutine equivalent) in Figure~\ref{f:CFAFormatGen} is extended with the monitor property and this interface function is used to communicate with the formatter:
2717\begin{cfa}
2718void fmt( Fmt & mutex fmt, char ch ) { fmt.ch = ch; resume( fmt ) }
2719\end{cfa}
2720multiple threads can safely pass characters for formatting.
2721
2722Figure~\ref{f:DirectCommunicationComparison} shows a comparison of direct call-communication in \CFA versus indirect channel-communication in Go.
2723(Ada has a similar mechanism to \CFA direct communication.)
2724The program thread in \CFA @main@ uses the call/return paradigm to directly communicate with the @GoRtn main@, whereas Go switches to the channel paradigm to indirectly communicate with the goroutine.
2725Communication by multiple threads is safe for the @gortn@ thread via mutex calls in \CFA or channel assignment in Go.
2726
2727Figure~\ref{f:DirectCommunicationDatingService} shows the dating-service problem in Figure~\ref{f:DatingServiceMonitor} extended from indirect monitor communication to direct thread communication.
2728When converting a monitor to a thread (server), the coding pattern is to move as much code as possible from the accepted members into the thread main so it does an much work as possible.
2729Notice, the dating server is postponing requests for an unspecified time while continuing to accept new requests.
2730For complex servers (web-servers), there can be hundreds of lines of code in the thread main and safe interaction with clients can be complex.
2731
2732
2733\subsection{Low-level Locks}
2734
2735For 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.
2736Some of these low-level mechanism are used to build the \CFA runtime, but we always advocate using high-level mechanisms whenever possible.
2737
2738
2739% \section{Parallelism}
2740% \label{s:Parallelism}
2741%
2742% Historically, computer performance was about processor speeds.
2743% However, with heat dissipation being a direct consequence of speed increase, parallelism is the new source for increased performance~\cite{Sutter05, Sutter05b}.
2744% Therefore, high-performance applications must care about parallelism, which requires concurrency.
2745% The lowest-level approach of parallelism is to use \newterm{kernel threads} in combination with semantics like @fork@, @join@, \etc.
2746% However, kernel threads are better as an implementation tool because of complexity and higher cost.
2747% Therefore, different abstractions are often layered onto kernel threads to simplify them, \eg pthreads.
2748%
2749%
2750% \subsection{User Threads}
2751%
2752% A direct improvement on kernel threads is user threads, \eg Erlang~\cite{Erlang} and \uC~\cite{uC++book}.
2753% 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.
2754% In many cases, user threads can be used on a much larger scale (100,000 threads).
2755% Like kernel threads, user threads support preemption, which maximizes nondeterminism, but increases the potential for concurrency errors: race, livelock, starvation, and deadlock.
2756% \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}.
2757%
2758% A variant of user thread is \newterm{fibres}, which removes preemption, \eg Go~\cite{Go} @goroutine@s.
2759% 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.
2760% However, preemption is necessary for fairness and to reduce tail-latency.
2761% For concurrency that relies on spinning, if all cores spin the system is livelocked, whereas preemption breaks the livelock.
2762
2763
2764\begin{comment}
2765\subsection{Thread Pools}
2766
2767In 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.
2768If the jobs are dependent, \ie interact, there is an implicit/explicit dependency graph that ties them together.
2769While removing direct concurrency, and hence the amount of context switching, thread pools significantly limit the interaction that can occur among jobs.
2770Indeed, jobs should not block because that also blocks the underlying thread, which effectively means the CPU utilization, and therefore throughput, suffers.
2771While 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.
2772As well, concurrency errors return, which threads pools are suppose to mitigate.
2773
2774\begin{figure}
2775\centering
2776\begin{tabular}{@{}l|l@{}}
2777\begin{cfa}
2778struct Adder {
2779        int * row, cols;
2780};
2781int operator()() {
2782        subtotal = 0;
2783        for ( int c = 0; c < cols; c += 1 )
2784                subtotal += row[c];
2785        return subtotal;
2786}
2787void ?{}( Adder * adder, int row[$\,$], int cols, int & subtotal ) {
2788        adder.[rows, cols, subtotal] = [rows, cols, subtotal];
2789}
2790
2791
2792
2793
2794\end{cfa}
2795&
2796\begin{cfa}
2797int main() {
2798        const int rows = 10, cols = 10;
2799        int matrix[rows][cols], subtotals[rows], total = 0;
2800        // read matrix
2801        Executor executor( 4 ); // kernel threads
2802        Adder * adders[rows];
2803        for ( r; rows ) { // send off work for executor
2804                adders[r] = new( matrix[r], cols, &subtotal[r] );
2805                executor.send( *adders[r] );
2806        }
2807        for ( r; rows ) {       // wait for results
2808                delete( adders[r] );
2809                total += subtotals[r];
2810        }
2811        sout | total;
2812}
2813\end{cfa}
2814\end{tabular}
2815\caption{Executor}
2816\end{figure}
2817\end{comment}
2818
2819
2820\section{Runtime Structure}
2821\label{s:CFARuntimeStructure}
2822
2823Figure~\ref{f:RunTimeStructure} illustrates the runtime structure of a \CFA program.
2824In 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.
2825An executing thread is illustrated by its containment in a processor.
2826
2827\begin{figure}
2828\centering
2829\input{RunTimeStructure}
2830\caption{\CFA Runtime structure}
2831\label{f:RunTimeStructure}
2832\end{figure}
2833
2834
2835\subsection{Cluster}
2836\label{s:RuntimeStructureCluster}
2837
2838A \newterm{cluster} is a collection of user and kernel threads, where the kernel threads run the user threads from the cluster's ready queue, and the operating system runs the kernel threads on the processors from its ready queue.
2839The term \newterm{virtual processor} is introduced as a synonym for kernel thread to disambiguate between user and kernel thread.
2840From the language perspective, a virtual processor is an actual processor (core).
2841
2842The purpose of a cluster is to control the amount of parallelism that is possible among threads, plus scheduling and other execution defaults.
2843The default cluster-scheduler is single-queue multi-server, which provides automatic load-balancing of threads on processors.
2844However, the design allows changing the scheduler, \eg multi-queue multi-server with work-stealing/sharing across the virtual processors.
2845If several clusters exist, both threads and virtual processors, can be explicitly migrated from one cluster to another.
2846No automatic load balancing among clusters is performed by \CFA.
2847
2848When 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.
2849The user cluster is created to contain the application user-threads.
2850Having all threads execute on the one cluster often maximizes utilization of processors, which minimizes runtime.
2851However, because of limitations of scheduling requirements (real-time), NUMA architecture, heterogeneous hardware, or issues with the underlying operating system, multiple clusters are sometimes necessary.
2852
2853
2854\subsection{Virtual Processor}
2855\label{s:RuntimeStructureProcessor}
2856
2857A 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.
2858Programs may use more virtual processors than hardware processors.
2859On a multiprocessor, kernel threads are distributed across the hardware processors resulting in virtual processors executing in parallel.
2860(It is possible to use affinity to lock a virtual processor onto a particular hardware processor~\cite{affinityLinux,affinityWindows}, which is used when caching issues occur or for heterogeneous hardware processors.) %, affinityFreebsd, affinityNetbsd, affinityMacosx
2861The \CFA runtime attempts to block unused processors and unblock processors as the system load increases;
2862balancing the workload with processors is difficult because it requires future knowledge, \ie what will the application workload do next.
2863Preemption occurs on virtual processors rather than user threads, via operating-system interrupts.
2864Thus virtual processors execute user threads, where preemption frequency applies to a virtual processor, so preemption occurs randomly across the executed user threads.
2865Turning off preemption transforms user threads into fibres.
2866
2867
2868\begin{comment}
2869\section{Implementation}
2870\label{s:Implementation}
2871
2872A primary implementation challenge is avoiding contention from dynamically allocating memory because of bulk acquire, \eg the internal-scheduling design is (almost) free of allocations.
2873All 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.
2874Furthermore, several bulk-acquire operations need a variable amount of memory.
2875This 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.
2876
2877In \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.
2878When a mutex call is made, pointers to the concerned monitors are aggregated into a variable-length array and sorted.
2879This array persists for the entire duration of the mutual exclusion and is used extensively for synchronization operations.
2880
2881To 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;
2882the corresponding registers are then restored for the other context.
2883Note, the instruction pointer is untouched since the context switch is always inside the same function.
2884Experimental 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.
2885
2886All kernel threads (@pthreads@) created a stack.
2887Each \CFA virtual processor is implemented as a coroutine and these coroutines run directly on the kernel-thread stack, effectively stealing this stack.
2888The exception to this rule is the program main, \ie the initial kernel thread that is given to any program.
2889In 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.
2890\end{comment}
2891
2892
2893\subsection{Preemption}
2894
2895Nondeterministic 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.
2896This atomic reliance can fail on multi-core machines, because execution across cores is nondeterministic.
2897A different reason for not supporting preemption is that it significantly complicates the runtime system, \eg Windows runtime does not support interrupts and on Linux systems, interrupts are complex (see below).
2898Preemption is normally handled by setting a countdown timer on each virtual processor.
2899When the timer expires, an interrupt is delivered, and its signal 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.
2900Multiple signal handlers may be pending.
2901When 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.
2902The 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;
2903therefore, the same signal mask is required for all virtual processors in a cluster.
2904Because preemption interval is usually long (1 millisecond) performance cost is negligible.
2905
2906Linux switched a decade ago from specific to arbitrary virtual-processor signal-delivery for applications with multiple kernel threads.
2907In the new semantics, a virtual-processor directed signal may be delivered to any virtual processor created by the application that does not have the signal blocked.
2908Hence, 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).
2909To 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.
2910Virtual processors register an expiration time with the discrete-event simulator, which is inserted in sorted order.
2911The 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.
2912Processing a preemption event sends an \emph{internal} @SIGUSR1@ signal to the registered virtual processor, which is always delivered to that processor.
2913
2914
2915\subsection{Debug Kernel}
2916
2917There are two versions of the \CFA runtime kernel: debug and non-debug.
2918The 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.
2919After a program is debugged, the non-debugging version can be used to significantly decrease space and increase performance.
2920
2921
2922\section{Performance}
2923\label{s:Performance}
2924
2925To test the performance of the \CFA runtime, a series of microbenchmarks are used to compare \CFA with pthreads, Java 11.0.6, Go 1.12.6, Rust 1.37.0, Python 3.7.6, Node.js 12.14.1, and \uC 7.0.0.
2926For 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.
2927The benchmark computer is an AMD Opteron\texttrademark\ 6380 NUMA 64-core, 8 socket, 2.5 GHz processor, running Ubuntu 16.04.6 LTS, and pthreads/\CFA/\uC are compiled with gcc 9.2.1.
2928
2929All benchmarks are run using the following harness. (The Java harness is augmented to circumvent JIT issues.)
2930\begin{cfa}
2931#define BENCH( `run` ) uint64_t start = cputime_ns();  `run;`  double result = (double)(cputime_ns() - start) / N;
2932\end{cfa}
2933where CPU time in nanoseconds is from the appropriate language clock.
2934Each benchmark is performed @N@ times, where @N@ is selected so the benchmark runs in the range of 2--20 seconds for the specific programming language.
2935The total time is divided by @N@ to obtain the average time for a benchmark.
2936Each benchmark experiment is run 13 times and the average appears in the table.
2937All omitted tests for other languages are functionally identical to the \CFA tests and available online~\cite{CforallBenchMarks}.
2938% tar --exclude-ignore=exclude -cvhf benchmark.tar benchmark
2939
2940\paragraph{Context Switching}
2941
2942In procedural programming, the cost of a function call is important as modularization (refactoring) increases.
2943(In many cases, a compiler inlines function calls to increase the size and number of basic blocks for optimizing.)
2944Similarly, when modularization extends to coroutines/threads, the time for a context switch becomes a relevant factor.
2945The coroutine test is from resumer to suspender and from suspender to resumer, which is two context switches.
2946%For async-await systems, the test is scheduling and fulfilling @N@ empty promises, where all promises are allocated before versus interleaved with fulfillment to avoid garbage collection.
2947For async-await systems, the test measures the cost of the @await@ expression entering the event engine by awaiting @N@ promises, where each created promise is resolved by an immediate event in the engine (using Node.js @setImmediate@).
2948The thread test is using yield to enter and return from the runtime kernel, which is two context switches.
2949The difference in performance between coroutine and thread context-switch is the cost of scheduling for threads, whereas coroutines are self-scheduling.
2950Figure~\ref{f:ctx-switch} shows the \CFA code for a coroutine/thread with results in Table~\ref{t:ctx-switch}.
2951
2952% From: Gregor Richards <gregor.richards@uwaterloo.ca>
2953% To: "Peter A. Buhr" <pabuhr@plg2.cs.uwaterloo.ca>
2954% Date: Fri, 24 Jan 2020 13:49:18 -0500
2955%
2956% I can also verify that the previous version, which just tied a bunch of promises together, *does not* go back to the
2957% event loop at all in the current version of Node. Presumably they're taking advantage of the fact that the ordering of
2958% events is intentionally undefined to just jump right to the next 'then' in the chain, bypassing event queueing
2959% entirely. That's perfectly correct behavior insofar as its difference from the specified behavior isn't observable, but
2960% it isn't typical or representative of much anything useful, because most programs wouldn't have whole chains of eager
2961% promises. Also, it's not representative of *anything* you can do with async/await, as there's no way to encode such an
2962% eager chain that way.
2963
2964\begin{multicols}{2}
2965\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
2966\begin{cfa}[aboveskip=0pt,belowskip=0pt]
2967@coroutine@ C {} c;
2968void main( C & ) { while () { @suspend;@ } }
2969int main() { // coroutine test
2970        BENCH( for ( N ) { @resume( c );@ } )
2971        sout | result;
2972}
2973int main() { // thread test
2974        BENCH( for ( N ) { @yield();@ } )
2975        sout | result;
2976}
2977\end{cfa}
2978\captionof{figure}{\CFA context-switch benchmark}
2979\label{f:ctx-switch}
2980
2981\columnbreak
2982
2983\vspace*{-16pt}
2984\captionof{table}{Context switch comparison (nanoseconds)}
2985\label{t:ctx-switch}
2986\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
2987\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
2988C function                      & 1.8           & 1.8           & 0.0   \\
2989\CFA generator          & 1.8           & 1.8           & 0.1   \\
2990\CFA coroutine          & 32.5          & 32.9          & 0.8   \\
2991\CFA thread                     & 93.8          & 93.6          & 2.2   \\
2992\uC coroutine           & 50.3          & 50.3          & 0.2   \\
2993\uC thread                      & 97.3          & 97.4          & 1.0   \\
2994Python generator        & 40.9          & 41.3          & 1.5   \\
2995Node.js generator       & 32.6          & 32.2          & 1.0   \\
2996Node.js await           & 1852.2        & 1854.7        & 16.4  \\
2997Goroutine thread        & 143.0         & 143.3         & 1.1   \\
2998Rust thread                     & 332.0         & 331.4         & 2.4   \\
2999Java thread                     & 405.0         & 415.0         & 17.6  \\
3000Pthreads thread         & 334.3         & 335.2         & 3.9
3001\end{tabular}
3002\end{multicols}
3003
3004\paragraph{Internal Scheduling}
3005
3006Internal scheduling is measured using a cycle of two threads signalling and waiting.
3007Figure~\ref{f:schedint} shows the code for \CFA, with results in Table~\ref{t:schedint}.
3008Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
3009Java 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.
3010
3011\begin{multicols}{2}
3012\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
3013\begin{cfa}
3014volatile int go = 0;
3015@condition c;@
3016@monitor@ M {} m1/*, m2, m3, m4*/;
3017void call( M & @mutex p1/*, p2, p3, p4*/@ ) {
3018        @signal( c );@
3019}
3020void wait( M & @mutex p1/*, p2, p3, p4*/@ ) {
3021        go = 1; // continue other thread
3022        for ( N ) { @wait( c );@ } );
3023}
3024thread T {};
3025void main( T & ) {
3026        while ( go == 0 ) { yield(); } // waiter must start first
3027        BENCH( for ( N ) { call( m1/*, m2, m3, m4*/ ); } )
3028        sout | result;
3029}
3030int main() {
3031        T t;
3032        wait( m1/*, m2, m3, m4*/ );
3033}
3034\end{cfa}
3035\captionof{figure}{\CFA Internal-scheduling benchmark}
3036\label{f:schedint}
3037
3038\columnbreak
3039
3040\vspace*{-16pt}
3041\captionof{table}{Internal-scheduling comparison (nanoseconds)}
3042\label{t:schedint}
3043\bigskip
3044
3045\begin{tabular}{@{}r*{3}{D{.}{.}{5.2}}@{}}
3046\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} & \multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
3047\CFA @signal@, 1 monitor        & 364.4         & 364.2         & 4.4           \\
3048\CFA @signal@, 2 monitor        & 484.4         & 483.9         & 8.8           \\
3049\CFA @signal@, 4 monitor        & 709.1         & 707.7         & 15.0          \\
3050\uC @signal@ monitor            & 328.3         & 327.4         & 2.4           \\
3051Rust cond. variable                     & 7514.0        & 7437.4        & 397.2         \\
3052Java @notify@ monitor           & 9623.0        & 9654.6        & 236.2         \\
3053Pthreads cond. variable         & 5553.7        & 5576.1        & 345.6
3054\end{tabular}
3055\end{multicols}
3056
3057
3058\paragraph{External Scheduling}
3059
3060External scheduling is measured using a cycle of two threads calling and accepting the call using the @waitfor@ statement.
3061Figure~\ref{f:schedext} shows the code for \CFA with results in Table~\ref{t:schedext}.
3062Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
3063
3064\begin{multicols}{2}
3065\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
3066\vspace*{-16pt}
3067\begin{cfa}
3068@monitor@ M {} m1/*, m2, m3, m4*/;
3069void call( M & @mutex p1/*, p2, p3, p4*/@ ) {}
3070void wait( M & @mutex p1/*, p2, p3, p4*/@ ) {
3071        for ( N ) { @waitfor( call : p1/*, p2, p3, p4*/ );@ }
3072}
3073thread T {};
3074void main( T & ) {
3075        BENCH( for ( N ) { call( m1/*, m2, m3, m4*/ ); } )
3076        sout | result;
3077}
3078int main() {
3079        T t;
3080        wait( m1/*, m2, m3, m4*/ );
3081}
3082\end{cfa}
3083\captionof{figure}{\CFA external-scheduling benchmark}
3084\label{f:schedext}
3085
3086\columnbreak
3087
3088\vspace*{-16pt}
3089\captionof{table}{External-scheduling comparison (nanoseconds)}
3090\label{t:schedext}
3091\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
3092\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
3093\CFA @waitfor@, 1 monitor       & 367.1 & 365.3 & 5.0   \\
3094\CFA @waitfor@, 2 monitor       & 463.0 & 464.6 & 7.1   \\
3095\CFA @waitfor@, 4 monitor       & 689.6 & 696.2 & 21.5  \\
3096\uC \lstinline[language=uC++]|_Accept| monitor  & 328.2 & 329.1 & 3.4   \\
3097Go \lstinline[language=Golang]|select| channel  & 365.0 & 365.5 & 1.2
3098\end{tabular}
3099\end{multicols}
3100
3101\paragraph{Mutual-Exclusion}
3102
3103Uncontented mutual exclusion, which frequently occurs, is measured by entering/leaving a critical section.
3104For monitors, entering and leaving a monitor function is measured, otherwise the language-appropriate mutex-lock is measured.
3105For comparison, a spinning (versus blocking) test-and-test-set lock is presented.
3106Figure~\ref{f:mutex} shows the code for \CFA with results in Table~\ref{t:mutex}.
3107Note the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
3108
3109\begin{multicols}{2}
3110\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
3111\begin{cfa}
3112@monitor@ M {} m1/*, m2, m3, m4*/;
3113call( M & @mutex p1/*, p2, p3, p4*/@ ) {}
3114int main() {
3115        BENCH( for( N ) call( m1/*, m2, m3, m4*/ ); )
3116        sout | result;
3117}
3118\end{cfa}
3119\captionof{figure}{\CFA acquire/release mutex benchmark}
3120\label{f:mutex}
3121
3122\columnbreak
3123
3124\vspace*{-16pt}
3125\captionof{table}{Mutex comparison (nanoseconds)}
3126\label{t:mutex}
3127\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
3128\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
3129test-and-test-set lock                  & 19.1  & 18.9  & 0.4   \\
3130\CFA @mutex@ function, 1 arg.   & 48.3  & 47.8  & 0.9   \\
3131\CFA @mutex@ function, 2 arg.   & 86.7  & 87.6  & 1.9   \\
3132\CFA @mutex@ function, 4 arg.   & 173.4 & 169.4 & 5.9   \\
3133\uC @monitor@ member rtn.               & 54.8  & 54.8  & 0.1   \\
3134Goroutine mutex lock                    & 34.0  & 34.0  & 0.0   \\
3135Rust mutex lock                                 & 33.0  & 33.2  & 0.8   \\
3136Java synchronized method                & 31.0  & 31.0  & 0.0   \\
3137Pthreads mutex Lock                             & 31.0  & 31.1  & 0.4
3138\end{tabular}
3139\end{multicols}
3140
3141\paragraph{Creation}
3142
3143Creation is measured by creating/deleting a specific kind of control-flow object.
3144Figure~\ref{f:creation} shows the code for \CFA with results in Table~\ref{t:creation}.
3145Note, the call stacks of \CFA coroutines are lazily created on the first resume, therefore the cost of creation with and without a stack are presented.
3146
3147\begin{multicols}{2}
3148\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
3149\begin{cfa}
3150@coroutine@ MyCoroutine {};
3151void ?{}( MyCoroutine & this ) {
3152#ifdef EAGER
3153        resume( this );
3154#endif
3155}
3156void main( MyCoroutine & ) {}
3157int main() {
3158        BENCH( for ( N ) { @MyCoroutine c;@ } )
3159        sout | result;
3160}
3161\end{cfa}
3162\captionof{figure}{\CFA creation benchmark}
3163\label{f:creation}
3164
3165\columnbreak
3166
3167\vspace*{-16pt}
3168\captionof{table}{Creation comparison (nanoseconds)}
3169\label{t:creation}
3170
3171\begin{tabular}[t]{@{}r*{3}{D{.}{.}{5.2}}@{}}
3172\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} & \multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
3173\CFA generator                  & 0.6           & 0.6           & 0.0           \\
3174\CFA coroutine lazy             & 13.4          & 13.1          & 0.5           \\
3175\CFA coroutine eager    & 144.7         & 143.9         & 1.5           \\
3176\CFA thread                             & 466.4         & 468.0         & 11.3          \\
3177\uC coroutine                   & 155.6         & 155.7         & 1.7           \\
3178\uC thread                              & 523.4         & 523.9         & 7.7           \\
3179Python generator                & 123.2         & 124.3         & 4.1           \\
3180Node.js generator               & 32.3          & 32.2          & 0.3           \\
3181Goroutine thread                & 751.0         & 750.5         & 3.1           \\
3182Rust thread                             & 53801.0       & 53896.8       & 274.9         \\
3183Java thread                             & 120274.0      & 120722.9      & 2356.7        \\
3184Pthreads thread                 & 31465.5       & 31419.5       & 140.4
3185\end{tabular}
3186\end{multicols}
3187
3188
3189\subsection{Discussion}
3190
3191Languages using 1:1 threading based on pthreads can at best meet or exceed (due to language overhead) the pthread results.
3192Note, pthreads has a fast zero-contention mutex lock checked in user space.
3193Languages with M:N threading have better performance than 1:1 because there is no operating-system interactions.
3194Languages with stackful coroutines have higher cost than stackless coroutines because of stack allocation and context switching;
3195however, stackful \uC and \CFA coroutines have approximately the same performance as stackless Python and Node.js generators.
3196The \CFA stackless generator is approximately 25 times faster for suspend/resume and 200 times faster for creation than stackless Python and Node.js generators.
3197
3198
3199\section{Conclusion}
3200
3201Advanced control-flow will always be difficult, especially when there is temporal ordering and nondeterminism.
3202However, many systems exacerbate the difficulty through their presentation mechanisms.
3203This paper shows it is possible to understand high-level control-flow using three properties: statefulness, thread, mutual-exclusion/synchronization.
3204Combining these properties creates a number of high-level, efficient, and maintainable control-flow types: generator, coroutine, thread, each of which can be a monitor.
3205Eliminated from \CFA are barging and spurious wakeup, which are nonintuitive and lead to errors, and having to work with a bewildering set of low-level locks and acquisition techniques.
3206\CFA high-level race-free monitors and threads provide the core mechanisms for mutual exclusion and synchronization, without having to resort to magic qualifiers like @volatile@/@atomic@.
3207Extending these mechanisms to handle high-level deadlock-free bulk acquire across both mutual exclusion and synchronization is a unique contribution.
3208The \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.
3209The M:N model is judged to be efficient and provide greater flexibility than a 1:1 threading model.
3210These 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.
3211Performance comparisons with other concurrent systems/languages show the \CFA approach is competitive across all basic operations, which translates directly into good performance in well-written applications with advanced control-flow.
3212C 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 using only calling communication.
3213
3214
3215\section{Future Work}
3216
3217While control flow in \CFA has a strong start, development is still underway to complete a number of missing features.
3218
3219\paragraph{Flexible Scheduling}
3220\label{futur:sched}
3221
3222An important part of concurrency is scheduling.
3223Different scheduling algorithms can affect performance (both in terms of average and variation).
3224However, no single scheduler is optimal for all workloads and therefore there is value in being able to change the scheduler for given programs.
3225One solution is to offer various tuning options, allowing the scheduler to be adjusted to the requirements of the workload.
3226However, to be truly flexible, a pluggable scheduler is necessary.
3227Currently, 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}.
3228
3229\paragraph{Non-Blocking I/O}
3230\label{futur:nbio}
3231
3232Many modern workloads are not bound by computation but IO operations, common cases being web servers and XaaS~\cite{XaaS} (anything as a service).
3233These types of workloads require significant engineering to amortizing costs of blocking IO-operations.
3234At 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.
3235Current 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.
3236However, these solutions lead to code that is hard to create, read and maintain.
3237A 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.
3238A non-blocking I/O library is currently under development for \CFA.
3239
3240\paragraph{Other Concurrency Tools}
3241\label{futur:tools}
3242
3243While monitors offer flexible and powerful concurrency for \CFA, other concurrency tools are also necessary for a complete multi-paradigm concurrency package.
3244Examples of such tools can include futures and promises~\cite{promises}, executors and actors.
3245These additional features are useful for applications that can be constructed without shared data and direct blocking.
3246As well, new \CFA extensions should make it possible to create a uniform interface for virtually all mutual exclusion, including monitors and low-level locks.
3247
3248\paragraph{Implicit Threading}
3249\label{futur:implcit}
3250
3251Basic 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.
3252This type of concurrency can be achieved both at the language level and at the library level.
3253The canonical example of implicit concurrency is concurrent nested @for@ loops, which are amenable to divide and conquer algorithms~\cite{uC++book}.
3254The \CFA language features should make it possible to develop a reasonable number of implicit concurrency mechanism to solve basic HPC data-concurrency problems.
3255However, implicit concurrency is a restrictive solution with significant limitations, so it can never replace explicit concurrent programming.
3256
3257
3258\section{Acknowledgements}
3259
3260The authors recognize the design assistance of Aaron Moss, Rob Schluntz, Andrew Beach, and Michael Brooks; David Dice for commenting and helping with the Java benchmarks; and Gregor Richards for helping with the Node.js benchmarks.
3261This research is funded by a grant from Waterloo-Huawei (\url{http://www.huawei.com}) Joint Innovation Lab. %, and Peter Buhr is partially funded by the Natural Sciences and Engineering Research Council of Canada.
3262
3263{%
3264\fontsize{9bp}{11.5bp}\selectfont%
3265\bibliography{pl,local}
3266}%
3267
3268\end{document}
3269
3270% Local Variables: %
3271% tab-width: 4 %
3272% fill-column: 120 %
3273% compile-command: "make" %
3274% End: %
Note: See TracBrowser for help on using the repository browser.