Changes in / [26a09f92:6565321]


Ignore:
Files:
30 added
2 deleted
72 edited

Legend:

Unmodified
Added
Removed
  • benchmark/ctxswitch/cfa_cor.cfa

    r26a09f92 r6565321  
    77void main( __attribute__((unused)) C & ) {
    88        while () {
    9                 suspend();
     9                suspend;
    1010        }
    1111}
  • doc/bibliography/pl.bib

    r26a09f92 r6565321  
    99%    Predefined journal names:
    1010%  acmcs: Computing Surveys             acta: Acta Infomatica
    11 @string{acta="Acta Infomatica"}
    1211%  cacm: Communications of the ACM
    1312%  ibmjrd: IBM J. Research & Development ibmsj: IBM Systems Journal
     
    2221%  tcs: Theoretical Computer Science
    2322
     23@string{acta="Acta Infomatica"}
    2424string{ieeepds="IEEE Transactions on Parallel and Distributed Systems"}
    2525@string{ieeepds="IEEE Trans. Parallel Distrib. Syst."}
     
    124124    series      = {ACM Distinguished Dissertations},
    125125    year        = 1983,
     126}
     127
     128@article{Zhang19,
     129    keywords    = {Algebraic effects, dynamic scoping, exceptions, parametricity, type systems},
     130    author      = {Zhang, Yizhou and Myers, Andrew C.},
     131    title       = {Abstraction-safe Effect Handlers via Tunneling},
     132    journal     = {Proc. ACM Program. Lang.},
     133    issue_date  = {January 2019},
     134    volume      = {3},
     135    number      = {POPL},
     136    month       = jan,
     137    year        = {2019},
     138    issn        = {2475-1421},
     139    pages       = {5:1--5:29},
     140    articleno   = {5},
     141    publisher   = {ACM},
     142    address     = {New York, NY, USA},
     143}
     144
     145@inproceedings{Zhang16,
     146    keywords    = {Exception tunneling, Genus, exception handling},
     147    author      = {Zhang, Yizhou and Salvaneschi, Guido and Beightol, Quinn and Liskov, Barbara and Myers, Andrew C.},
     148    title       = {Accepting Blame for Safe Tunneled Exceptions},
     149    booktitle   = {Proceedings of the 37th ACM SIGPLAN Conference on Programming Language Design and Implementation},
     150    series      = {PLDI'16},
     151    year        = {2016},
     152    location    = {Santa Barbara, CA, USA},
     153    pages       = {281--295},
     154    publisher   = {ACM},
     155    address     = {New York, NY, USA},
    126156}
    127157
     
    398428    journal     = sigplan,
    399429    year        = 1981,
    400     month       = feb, volume = 16, number = 2, pages = {48-52},
     430    month       = feb,
     431    volume      = 16,
     432    number      = 2,
     433    pages       = {48-52},
    401434    comment     = {
    402435        A one-pass, top-down algorithm for overload resolution.  Input is a
     
    477510    title       = {An Alternative to Subclassing},
    478511    journal     = sigplan,
    479     volume      = {21},    number = {11},
     512    volume      = {21},
     513    number      = {11},
    480514    pages       = {424-428},
    481     month       = nov, year = 1986,
     515    month       = nov,
     516    year        = 1986,
    482517    comment     = {
    483518        The Smalltalk class hierarchy has three uses: factoring out code;
     
    533568    isbn        = {3-540-66538-2},
    534569    location    = {Toulouse, France},
    535     doi         = {http://doi.acm.org/10.1145/318773.319251},
    536570    publisher   = {Springer},
    537571    address     = {London, UK},
     
    631665    year        = 2010,
    632666    pages       = {39--50},
    633     numpages    = {12},
    634667    publisher   = {IEEE Computer Society},
    635668    address     = {Washington, DC, USA},
     
    922955}
    923956
     957@manual{C99,
     958    keywords    = {ISO/IEC C 9899},
     959    contributer = {pabuhr@plg},
     960    key         = {C99},
     961    title       = {C Programming Language {ISO/IEC} 9899:1999(E)},
     962    edition     = {2nd},
     963    publisher   = {International Standard Organization},
     964    address     = {\href{https://webstore.ansi.org/Standards/INCITS/INCITSISOIEC98991999R2005}{https://webstore.ansi.org/\-Standards/\-INCITS/\-INCITSISOIEC98991999R2005}},
     965    year        = 1999,
     966}
     967
    924968@manual{C11,
    925969    keywords    = {ISO/IEC C 11},
     
    13051349    location    = {London, United Kingdom},
    13061350    pages       = {41--53},
    1307     numpages    = {13},
    1308     url         = {http://doi.acm.org/10.1145/360204.360207},
    1309     doi         = {10.1145/360204.360207},
    1310     acmid       = {360207},
    13111351    publisher   = {ACM},
    13121352    address     = {New York, NY, USA},
     
    24082448    year        = 1993,
    24092449    pages       = {201--208},
    2410     url         = {http://doi.acm.org/10.1145/155360.155580},
    24112450    publisher   = {ACM},
    24122451    address     = {New York, NY, USA},
     
    26062645    location    = {Boulder, Colorado, USA},
    26072646    pages       = {91--97},
    2608     numpages    = {7},
    26092647    publisher   = {ACM},
    26102648    address     = {New York, NY, USA},
     
    26372675    issn        = {0004-5411},
    26382676    pages       = {215--225},
    2639     numpages    = {11},
    2640     url         = {http://doi.acm.org/10.1145/321879.321884},
    2641     doi         = {10.1145/321879.321884},
    2642     acmid       = {321884},
    26432677    publisher   = {ACM},
    26442678    address     = {New York, NY, USA},
     
    27082742}
    27092743
     2744@misc{Drepper13,
     2745    keywords    = {thread-local storage},
     2746    contributer = {pabuhr@plg},
     2747    author      = {Ulrich Drepper},
     2748    title       = {{ELF} Handling For Thread-Local Storage},
     2749    year        = 2013,
     2750    month       = aug,
     2751    note        = {WikipediA},
     2752    howpublished= {\href{http://www.akkadia.org/drepper/tls.pdf}
     2753                  {http://\-www.akkadia.org/\-drepper/\-tls.pdf}},
     2754}
     2755
    27102756@misc{Turley99,
    27112757    keywords    = {embedded system, micrprocessor},
     
    27182764    howpublished= {\href{https://www.eetimes.com/author.asp?sectionid=36&doc_id=1287712}
    27192765                  {https://\-www.eetimes.com/\-author.asp?sectionid=\-36&doc_id=1287712}},
     2766}
     2767
     2768@article{Xiao19,
     2769    keywords    = {bug classification, fault trigger, Linux operating system, regression bug},
     2770    contributer = {pabuhr@plg},
     2771    author      = {Guanping Xiao and Zheng Zheng and Beibei Yin and Kishor S. Trivedi and Xiaoting Du and Kai-Yuan Cai},
     2772    title       = {An Empirical Study of Fault Triggers in the Linux Operating System: An Evolutionary Perspective},
     2773    journal     = {IEEE Transactions on Reliability},
     2774    month       = dec,
     2775    year        = 2019,
     2776    volume      = 68,
     2777    number      = 4,
     2778    pages       = {1356-1383},
    27202779}
    27212780
     
    31373196}
    31383197
     3198@inproceedings{Palix11,
     3199    keywords    = {Linux, fault-finding tools},
     3200    contributer = {pabuhr@plg},
     3201    author      = {Nicolas Palix and Ga\"el Thomas and Suman Saha and Christophe Calv\`es and Julia Lawall and Gilles Muller},
     3202    title       = {Faults in Linux: Ten Years Later},
     3203    booktitle   = {Proc. of the 16 International Conf. on Arch. Support for Prog. Lang. and Oper. Sys.},
     3204    series      = {ASPLOS'11},
     3205    month       = mar,
     3206    year        = 2011,
     3207    location    = {Newport Beach, California, USA},
     3208    pages       = {305-318},
     3209    publisher   = {ACM},
     3210    address     = {New York, NY, USA},
     3211}
     3212
    31393213@article{Lamport87,
    31403214    keywords    = {software solutions, mutual exclusion, fast},
     
    32583332    issn        = {0001-0782},
    32593333    pages       = {107--115},
    3260     numpages    = {9},
    3261     url         = {http://doi.acm.org/10.1145/1538788.1538814},
    3262     doi         = {10.1145/1538788.1538814},
    3263     acmid       = {1538814},
    32643334    publisher   = {ACM},
    32653335    address     = {New York, NY, USA},
     
    36643734}
    36653735
     3736@mastersthesis{Radhakrishnan19,
     3737    author      = {Srihari Radhakrishnan},
     3738    title       = {High Performance Web Servers: A Study In Concurrent Programming Models},
     3739    school      = {School of Computer Sc., University of Waterloo},
     3740    year        = 2019,
     3741    optaddress  = {Waterloo, Ontario, Canada, N2L 3G1},
     3742    note        = {\href{https://uwspace.uwaterloo.ca/handle/10012/14706}{https://\-uwspace.uwaterloo.ca/\-handle/\-10012/\-14706}},
     3743}
     3744
    36663745@article{katzenelson83b,
    36673746    contributer = {gjditchfield@plg},
     
    36973776    pages       = {115-138},
    36983777    year        = 1971,
     3778}
     3779
     3780@inproceedings{Hagersten03,
     3781    keywords    = {cache storage, parallel architectures, performance evaluation, shared memory systems},
     3782    author      = {Zoran Radovi\'{c} and Erik Hagersten},
     3783    title       = {Hierarchical backoff locks for nonuniform communication architectures},
     3784    booktitle   = {Proceedings of the Ninth International Symposium on High-Performance Computer Architecture},
     3785    year        = {2003},
     3786    location    = {Anaheim, CA, USA},
     3787    pages       = {241-252},
     3788    publisher   = {IEEE},
    36993789}
    37003790
     
    43654455}
    43664456
     4457@misc{gccValueLabels,
     4458    keywords    = {gcc extension, value labels},
     4459    contributer = {pabuhr@plg},
     4460    key         = {Labels as Values},
     4461    author      = {{gcc Extension}},
     4462    title       = {Labels as Values},
     4463    year        = {since gcc-3},
     4464    howpublished= {\href{https://gcc.gnu.org/onlinedocs/gcc/Labels-as-Values.html}
     4465                  {https:\-//gcc.gnu.org/\-onlinedocs/\-gcc/\-Labels-as-Values.html}},
     4466}
     4467
    43674468@mastersthesis{Clarke90,
    43684469    keywords    = {concurrency, postponing requests},
     
    44574558
    44584559@article{Pierce00,
    4459     keywords    = {Scala},
     4560    keywords    = {Scala, polymorphism, subtyping, type inference},
    44604561    contributer = {a3moss@uwaterloo.ca},
    44614562    author      = {Pierce, Benjamin C. and Turner, David N.},
     
    44694570    issn        = {0164-0925},
    44704571    pages       = {1--44},
    4471     numpages    = {44},
    4472     url         = {http://doi.acm.org/10.1145/345099.345100},
    4473     doi         = {10.1145/345099.345100},
    4474     acmid       = {345100},
    44754572    publisher   = {ACM},
    44764573    address     = {New York, NY, USA},
    4477     keywords    = {polymorphism, subtyping, type inference},
    44784574}
     4575
     4576@article{Dice15,
     4577    keywords    = {Concurrency, NUMA, hierarchical locks, locks, multicore, mutex, mutual exclusion, spin locks},
     4578    author      = {Dice, David and Marathe, Virendra J. and Shavit, Nir},
     4579    title       = {Lock Cohorting: A General Technique for Designing NUMA Locks},
     4580    journal     = {ACM Trans. Parallel Comput.},
     4581    issue_date  = {January 2015},
     4582    volume      = 1,
     4583    number      = 2,
     4584    month       = feb,
     4585    year        = 2015,
     4586    pages       = {13:1--13:42},
     4587    publisher   = {ACM},
     4588    address     = {New York, NY, USA},
     4589}
    44794590
    44804591@article{Sundell08,
     
    45544665    journal     = sigplan,
    45554666    year        = 1989,
    4556     month       = jun, volume = 24, number = 6, pages = {37-48},
     4667    month       = jun,
     4668    volume      = 24,
     4669    number      = 6,
     4670    pages       = {37-48},
    45574671    abstract    = {
    45584672        This paper describes a scheme we have used to manage a large
     
    49955109    year        = 1986,
    49965110    pages       = {313--326},
    4997     numpages    = {14},
    49985111    publisher   = {ACM},
    49995112    address     = {New York, NY, USA},
     
    50115124    year        = 1986,
    50125125    pages       = {327--348},
    5013     numpages    = {22},
    50145126    publisher   = {ACM},
    50155127    address     = {New York, NY, USA},
     
    52085320    year        = 2005,
    52095321    pages       = {146-196},
    5210     numpages    = {51},
    52115322    publisher   = {ACM},
    52125323    address     = {New York, NY, USA},
     
    53545465    year        = 2000,
    53555466    pages       = {29-46},
    5356     note        = {OOPSLA'00, Oct. 15--19, 2000, Minneapolis, Minnesota, U.S.A.},
     5467    note        = {OOPSLA'00, Oct. 15--19, 2000, Minneapolis, Minn., U.S.A.},
    53575468}
    53585469
     
    54685579    location    = {San Diego, California, USA},
    54695580    pages       = {101--112},
    5470     numpages    = {12},
    5471     url         = {http://doi.acm.org/10.1145/2535838.2535878},
    5472     doi         = {10.1145/2535838.2535878},
    5473     acmid       = {2535878},
    54745581    publisher   = {ACM},
    54755582    address     = {New York, NY, USA},
     
    55755682    issn        = {0362-1340},
    55765683    pages       = {30--42},
    5577     numpages    = {13},
    5578     url         = {http://doi.acm.org/10.1145/947586.947589},
    5579     doi         = {10.1145/947586.947589},
    55805684    publisher   = {ACM},
    55815685    address     = {New York, NY, USA}
     
    61126216    month       = 9,
    61136217    year        = 2005,
     6218}
     6219
     6220@article{Bauer15,
     6221    keywords    = {resumption exceptions, theory},
     6222    contributer = {pabuhr@plg},
     6223    author      = {Andrej Bauer and Matija Pretnar},
     6224    title       = {Programming with Algebraic Effects and Handlers},
     6225    journal     = {Journal of Logical and Algebraic Methods in Programming},
     6226    publisher   = {Elsevier BV},
     6227    volume      = 84,
     6228    number      = 1,
     6229    month       = jan,
     6230    year        = 2015,
     6231    pages       = {108-123},
    61146232}
    61156233
     
    64996617    issn        = {0164-0925},
    65006618    pages       = {429-475},
    6501     url         = {http://doi.acm.org/10.1145/1133651.1133653},
    6502     doi         = {10.1145/1133651.1133653},
    6503     acmid       = {1133653},
    65046619    publisher   = {ACM},
    65056620    address     = {New York, NY, USA},
     
    68796994    issn        = {0001-0782},
    68806995    pages       = {565--569},
    6881     numpages    = {5},
    6882     url         = {http://doi.acm.org/10.1145/359545.359566},
    6883     doi         = {10.1145/359545.359566},
    6884     acmid       = {359566},
    68856996    publisher   = {ACM},
    68866997    address     = {New York, NY, USA}
     
    69007011    issn        = {0362-1340},
    69017012    pages       = {145--147},
    6902     numpages    = {3},
    6903     url         = {http://doi.acm.org/10.1145/122598.122614},
    6904     doi         = {10.1145/122598.122614},
    6905     acmid       = {122614},
    69067013    publisher   = {ACM},
    69077014    address     = {New York, NY, USA},
     
    70067113    issn        = {0362-1340},
    70077114    pages       = {82--87},
    7008     numpages    = {6},
    7009     url         = {http://doi.acm.org/10.1145/947680.947688},
    7010     doi         = {10.1145/947680.947688},
    70117115    publisher   = {ACM},
    70127116    address     = {New York, NY, USA},
     
    71537257}
    71547258
     7259@article{Cascaval08,
     7260    author      = {Cascaval, Calin and Blundell, Colin and Michael, Maged and Cain, Harold W. and Wu, Peng and Chiras, Stefanie and Chatterjee, Siddhartha},
     7261    title       = {Software Transactional Memory: Why Is It Only a Research Toy?},
     7262    journal     = {Queue},
     7263    volume      = {6},
     7264    number      = {5},
     7265    month       = sep,
     7266    year        = {2008},
     7267    pages       = {40:46--40:58},
     7268    publisher   = {ACM},
     7269    address     = {New York, NY, USA},
     7270}
     7271
    71557272@article{Dijkstra65a,
    71567273    keywords    = {N-thread software-solution mutual exclusion},
     
    73637480    year        = 1974,
    73647481    pages       = {261-301},
    7365     issn        = {0360-0300},
    7366     doi         = {http://doi.acm.org/10.1145/356635.356640},
    73677482    publisher   = {ACM},
    73687483    address     = {New York, NY, USA},
     
    74547569    publisher   = {ACM Press},
    74557570    address     = {New York, NY, USA},
    7456     doi         = {http://doi.acm.org/10.1145/356586.356588},
    74577571}
    74587572
     
    77557869    howpublished= {\href{https://projects.eclipse.org/proposals/trace-compass}{https://\-projects.eclipse.org/\-proposals/\-trace-compass}},
    77567870}
    7757  
     7871
     7872@inproceedings{Boehm09,
     7873    author      = {Boehm, Hans-J.},
     7874    title       = {Transactional Memory Should Be an Implementation Technique, Not a Programming Interface},
     7875    booktitle   = {Proceedings of the First USENIX Conference on Hot Topics in Parallelism},
     7876    series      = {HotPar'09},
     7877    year        = {2009},
     7878    location    = {Berkeley, California},
     7879    publisher   = {USENIX Association},
     7880    address     = {Berkeley, CA, USA},
     7881}
     7882
    77587883@article{Leroy00,
    77597884    keywords    = {type-systems, exceptions},
     
    78057930    number      = {2},
    78067931    pages       = {204-214},
    7807     month       = apr, year = 1988,
     7932    month       = apr,
     7933    year        = 1988,
    78087934    comment     = {
    78097935        Extended record types add fields to their base record.  Assignment
     
    81108236    issn        = {0004-5411},
    81118237    pages       = {245--281},
    8112     numpages    = {37},
    8113     url         = {http://doi.acm.org/10.1145/62.2160},
    8114     doi         = {10.1145/62.2160},
    8115     acmid       = {2160},
    81168238    publisher   = {ACM},
    81178239    address     = {New York, NY, USA},
     
    81268248    contributer = {pabuhr@plg},
    81278249    author      = {Boehm, Hans-J. and Adve, Sarita V.},
    8128     title       = {You Don'T Know Jack About Shared Variables or Memory Models},
     8250    title       = {You Don't Know Jack About Shared Variables or Memory Models},
    81298251    journal     = cacm,
    81308252    volume      = 55,
  • doc/papers/concurrency/Paper.tex

    r26a09f92 r6565321  
    6161\newcommand{\CCseventeen}{\textrm{C}\kern-.1em\hbox{+\kern-.25em+}17\xspace} % C++17 symbolic name
    6262\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
     63\newcommand{\Csharp}{C\raisebox{-0.7ex}{\large$^\sharp$}\xspace} % C# symbolic name
    6464
    6565%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     
    127127\newcommand*{\etc}{%
    128128        \@ifnextchar{.}{\ETC}%
    129         {\ETC.\xspace}%
     129                {\ETC.\xspace}%
    130130}}{}%
    131131\@ifundefined{etal}{
    132132\newcommand{\ETAL}{\abbrevFont{et}~\abbrevFont{al}}
    133133\newcommand*{\etal}{%
    134         \@ifnextchar{.}{\protect\ETAL}%
    135                 {\protect\ETAL.\xspace}%
     134        \@ifnextchar{.}{\ETAL}%
     135                {\ETAL.\xspace}%
    136136}}{}%
    137137\@ifundefined{viz}{
     
    163163                __float80, float80, __float128, float128, forall, ftype, generator, _Generic, _Imaginary, __imag, __imag__,
    164164                inline, __inline, __inline__, __int128, int128, __label__, monitor, mutex, _Noreturn, one_t, or,
    165                 otype, restrict, __restrict, __restrict__, __signed, __signed__, _Static_assert, thread,
     165                otype, restrict, resume, __restrict, __restrict__, __signed, __signed__, _Static_assert, suspend, thread,
    166166                _Thread_local, throw, throwResume, timeout, trait, try, ttype, typeof, __typeof, __typeof__,
    167167                virtual, __volatile, __volatile__, waitfor, when, with, zero_t},
    168168        moredirectives={defined,include_next},
    169169        % replace/adjust listing characters that look bad in sanserif
    170         literate={-}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.1ex}}}}1 {^}{\raisebox{0.6ex}{$\scriptstyle\land\,$}}1
     170        literate={-}{\makebox[1ex][c]{\raisebox{0.5ex}{\rule{0.8ex}{0.1ex}}}}1 {^}{\raisebox{0.6ex}{$\scriptstyle\land\,$}}1
    171171                {~}{\raisebox{0.3ex}{$\scriptstyle\sim\,$}}1 % {`}{\ttfamily\upshape\hspace*{-0.1ex}`}1
    172172                {<}{\textrm{\textless}}1 {>}{\textrm{\textgreater}}1
     
    197197                _Else, _Enable, _Event, _Finally, _Monitor, _Mutex, _Nomutex, _PeriodicTask, _RealTimeTask,
    198198                _Resume, _Select, _SporadicTask, _Task, _Timeout, _When, _With, _Throw},
    199 }
    200 \lstdefinelanguage{Golang}{
    201         morekeywords=[1]{package,import,func,type,struct,return,defer,panic,recover,select,var,const,iota,},
    202         morekeywords=[2]{string,uint,uint8,uint16,uint32,uint64,int,int8,int16,int32,int64,
    203                 bool,float32,float64,complex64,complex128,byte,rune,uintptr, error,interface},
    204         morekeywords=[3]{map,slice,make,new,nil,len,cap,copy,close,true,false,delete,append,real,imag,complex,chan,},
    205         morekeywords=[4]{for,break,continue,range,goto,switch,case,fallthrough,if,else,default,},
    206         morekeywords=[5]{Println,Printf,Error,},
    207         sensitive=true,
    208         morecomment=[l]{//},
    209         morecomment=[s]{/*}{*/},
    210         morestring=[b]',
    211         morestring=[b]",
    212         morestring=[s]{`}{`},
    213199}
    214200
     
    241227{}
    242228\lstnewenvironment{uC++}[1][]
    243 {\lstset{#1}}
     229{\lstset{language=uC++,moredelim=**[is][\protect\color{red}]{`}{`},#1}\lstset{#1}}
    244230{}
    245231\lstnewenvironment{Go}[1][]
     
    282268\CFA is a polymorphic, non-object-oriented, concurrent, backwards-compatible extension of the C programming language.
    283269This paper discusses the design philosophy and implementation of its advanced control-flow and concurrent/parallel features, along with the supporting runtime written in \CFA.
    284 These features are created from scratch as ISO C has only low-level and/or unimplemented concurrency, so C programmers continue to rely on library features like pthreads.
     270These 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.
    285271\CFA introduces modern language-level control-flow mechanisms, like generators, coroutines, user-level threading, and monitors for mutual exclusion and synchronization.
    286272% Library extension for executors, futures, and actors are built on these basic mechanisms.
     
    295281
    296282\begin{document}
    297 \linenumbers                                            % comment out to turn off line numbering
     283\linenumbers                            % comment out to turn off line numbering
    298284
    299285\maketitle
     
    302288\section{Introduction}
    303289
    304 This paper discusses the design philosophy and implementation of advanced language-level control-flow and concurrent/parallel features in \CFA~\cite{Moss18,Cforall} and its runtime, which is written entirely in \CFA.
    305 \CFA is a modern, polymorphic, non-object-oriented\footnote{
    306 \CFA has features often associated with object-oriented programming languages, such as constructors, destructors, virtuals and simple inheritance.
     290\CFA~\cite{Moss18,Cforall} is a modern, polymorphic, non-object-oriented\footnote{
     291\CFA has object-oriented features, such as constructors, destructors, virtuals and simple trait/interface inheritance.
     292% Go interfaces, Rust traits, Swift Protocols, Haskell Type Classes and Java Interfaces.
     293% "Trait inheritance" works for me. "Interface inheritance" might also be a good choice, and distinguish clearly from implementation inheritance.
     294% 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".
    307295However, 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.},
    308296backwards-compatible extension of the C programming language.
    309 In many ways, \CFA is to C as Scala~\cite{Scala} is to Java, providing a \emph{research vehicle} for new typing and control-flow capabilities on top of a highly popular programming language allowing immediate dissemination.
    310 Within the \CFA framework, new control-flow features are created from scratch because ISO \Celeven defines only a subset of the \CFA extensions, where the overlapping features are concurrency~\cite[\S~7.26]{C11}.
    311 However, \Celeven concurrency is largely wrappers for a subset of the pthreads library~\cite{Butenhof97,Pthreads}, and \Celeven and pthreads concurrency is simple, based on thread fork/join in a function and mutex/condition locks, which is low-level and error-prone;
    312 no high-level language concurrency features are defined.
    313 Interestingly, almost a decade after publication of the \Celeven standard, neither gcc-8, clang-9 nor msvc-19 (most recent versions) support the \Celeven include @threads.h@, indicating little interest in the C11 concurrency approach (possibly because the effort to add concurrency to \CC).
    314 Finally, while the \Celeven standard does not state a threading model, the historical association with pthreads suggests implementations would adopt kernel-level threading (1:1)~\cite{ThreadModel}.
    315 
     297In 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{
     298The 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.}
     299allowing immediate dissemination.
     300This 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.
     301The \CFA control-flow framework extends ISO \Celeven~\cite{C11} with new call/return and concurrent/parallel control-flow.
     302
     303% The call/return extensions retain state between callee and caller versus losing the callee's state on return;
     304% the concurrency extensions allow high-level management of threads.
     305
     306Call/return control-flow with argument/parameter passing appeared in the first programming languages.
     307Over 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).
     308While \CFA has mechanisms for dynamic call (algebraic effects) and exceptions\footnote{
     309\CFA exception handling will be presented in a separate paper.
     310The 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.
     311\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}.
     312Coroutining is sequential execution requiring direct handoff among coroutines, \ie only the programmer is controlling execution order.
     313If coroutines transfer to an internal event-engine for scheduling the next coroutines, the program transitions into the realm of concurrency~\cite[\S~3]{Buhr05a}.
     314Coroutines are only a stepping stone towards concurrency where the commonality is that coroutines and threads retain state between calls.
     315
     316\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}
     317Interestingly, 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).
     318While 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.
    316319In contrast, there has been a renewed interest during the past decade in user-level (M:N, green) threading in old and new programming languages.
    317320As multi-core hardware became available in the 1980/90s, both user and kernel threading were examined.
    318321Kernel 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}.
    319322Libraries like pthreads were developed for C, and the Solaris operating-system switched from user (JDK 1.1~\cite{JDK1.1}) to kernel threads.
    320 As a result, languages like Java, Scala, Objective-C~\cite{obj-c-book}, \CCeleven~\cite{C11}, and C\#~\cite{Csharp} adopt the 1:1 kernel-threading model, with a variety of presentation mechanisms.
    321 From 2000 onwards, languages like Go~\cite{Go}, Erlang~\cite{Erlang}, Haskell~\cite{Haskell}, D~\cite{D}, and \uC~\cite{uC++,uC++book} have championed the M:N user-threading model, and many user-threading libraries have appeared~\cite{Qthreads,MPC,Marcel}, including putting green threads back into Java~\cite{Quasar}.
    322 The main argument for user-level threading is that it is lighter weight than kernel threading (locking and context switching do not cross the kernel boundary), so there is less restriction on programming styles that encourage large numbers of threads performing medium work units to facilitate load balancing by the runtime~\cite{Verch12}.
     323As 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.
     324From 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}.
     325The 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}.
    323326As well, user-threading facilitates a simpler concurrency approach using thread objects that leverage sequential patterns versus events with call-backs~\cite{Adya02,vonBehren03}.
    324327Finally, 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.
    325328
    326 A further effort over the past two decades is the development of language memory models to deal with the conflict between language features and compiler/hardware optimizations, \ie some language features are unsafe in the presence of aggressive sequential optimizations~\cite{Buhr95a,Boehm05}.
     329A 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}.
    327330The 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.
    328331One solution is low-level qualifiers and functions (\eg @volatile@ and atomics) allowing \emph{programmers} to explicitly write safe (race-free~\cite{Boehm12}) programs.
    329 A safer solution is high-level language constructs so the \emph{compiler} knows the optimization boundaries, and hence, provides implicit safety.
    330 This problem is best known with respect to concurrency, but applies to other complex control-flow, like exceptions\footnote{
    331 \CFA exception handling will be presented in a separate paper.
    332 The 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++}
    333 } and coroutines.
    334 Finally, language solutions allow matching constructs with language paradigm, \ie imperative and functional languages often have different presentations of the same concept to fit their programming model.
    335 
    336 Finally, it is important for a language to provide safety over performance \emph{as the default}, allowing careful reduction of safety for performance when necessary.
    337 Two concurrency violations of this philosophy are \emph{spurious wakeup} (random wakeup~\cite[\S~8]{Buhr05a}) and \emph{barging}\footnote{
    338 The notion of competitive succession instead of direct handoff, \ie a lock owner releases the lock and an arriving thread acquires it ahead of preexisting waiter threads.
     332A 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.
     333While the optimization problem is best known with respect to concurrency, it applies to other complex control-flow, like exceptions and coroutines.
     334As 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.
     335
     336Finally, 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.
     337Two concurrency violations of this philosophy are \emph{spurious wakeup} (random wakeup~\cite[\S~9]{Buhr05a}) and \emph{barging}\footnote{
     338Barging is competitive succession instead of direct handoff, \ie after a lock is released both arriving and preexisting waiter threads compete to acquire the lock.
     339Hence, 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.
    339340} (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.
    340 However, spurious wakeup is \emph{not} a foundational concurrency property~\cite[\S~8]{Buhr05a}, it is a performance design choice.
    341 Similarly, signals-as-hints are often a performance decision.
    342 We argue removing spurious wakeup and signals-as-hints make concurrent programming significantly safer because it removes local non-determinism and matches with programmer expectation.
    343 (Author experience teaching concurrency is that students are highly confused by these semantics.)
    344 Clawing back performance, when local non-determinism is unimportant, should be an option not the default.
    345 
    346 \begin{comment}
    347 Most augmented traditional (Fortran 18~\cite{Fortran18}, Cobol 14~\cite{Cobol14}, Ada 12~\cite{Ada12}, Java 11~\cite{Java11}) and new languages (Go~\cite{Go}, Rust~\cite{Rust}, and D~\cite{D}), except \CC, diverge from C with different syntax and semantics, only interoperate indirectly with C, and are not systems languages, for those with managed memory.
    348 As a result, there is a significant learning curve to move to these languages, and C legacy-code must be rewritten.
    349 While \CC, like \CFA, takes an evolutionary approach to extend C, \CC's constantly growing complex and interdependent features-set (\eg objects, inheritance, templates, etc.) mean idiomatic \CC code is difficult to use from C, and C programmers must expend significant effort learning \CC.
    350 Hence, rewriting and retraining costs for these languages, even \CC, are prohibitive for companies with a large C software-base.
    351 \CFA with its orthogonal feature-set, its high-performance runtime, and direct access to all existing C libraries circumvents these problems.
    352 \end{comment}
    353 
    354 \CFA embraces user-level threading, language extensions for advanced control-flow, and safety as the default.
    355 We present comparative examples so the reader can judge if the \CFA control-flow extensions are better and safer than those in other concurrent, imperative programming languages, and perform experiments to show the \CFA runtime is competitive with other similar mechanisms.
     341(Author experience teaching concurrency is that students are confused by these semantics.)
     342However, spurious wakeup is \emph{not} a foundational concurrency property~\cite[\S~9]{Buhr05a};
     343it is a performance design choice.
     344We argue removing spurious wakeup and signals-as-hints make concurrent programming simpler and safer as there is less local non-determinism to manage.
     345If barging acquisition is allowed, its specialized performance advantage should be available as an option not the default.
     346
     347\CFA embraces language extensions for advanced control-flow, user-level threading, and safety as the default.
     348We 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.
    356349The main contributions of this work are:
    357 \begin{itemize}[topsep=3pt,itemsep=1pt]
     350\begin{itemize}[topsep=3pt,itemsep=0pt]
    358351\item
    359 language-level generators, coroutines and user-level threading, which respect the expectations of C programmers.
     352a set of fundamental execution properties that dictate which language-level control-flow features need to be supported,
     353
    360354\item
    361 monitor 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.
     355integration of these language-level control-flow features, while respecting the style and expectations of C programmers,
     356
    362357\item
    363 providing statically type-safe interfaces that integrate with the \CFA polymorphic type-system and other language features.
     358monitor 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,
     359
     360\item
     361providing statically type-safe interfaces that integrate with the \CFA polymorphic type-system and other language features,
     362
    364363% \item
    365364% library extensions for executors, futures, and actors built on the basic mechanisms.
     365
    366366\item
    367 a runtime system with no spurious wakeup.
     367a runtime system without spurious wake-up and no performance loss,
     368
    368369\item
    369 a dynamic partitioning mechanism to segregate the execution environment for specialized requirements.
     370a 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).
     371
    370372% \item
    371373% a non-blocking I/O library
     374
    372375\item
    373 experimental results showing comparable performance of the new features with similar mechanisms in other programming languages.
     376experimental results showing comparable performance of the \CFA features with similar mechanisms in other languages.
    374377\end{itemize}
    375378
    376 Section~\ref{s:StatefulFunction} begins advanced control by introducing sequential functions that retain data and execution state between calls, which produces constructs @generator@ and @coroutine@.
    377 Section~\ref{s:Concurrency} begins concurrency, or how to create (fork) and destroy (join) a thread, which produces the @thread@ construct.
     379Section~\ref{s:FundamentalExecutionProperties} presents the compositional hierarchy of execution properties directing the design of control-flow features in \CFA.
     380Section~\ref{s:StatefulFunction} begins advanced control by introducing sequential functions that retain data and execution state between calls producing constructs @generator@ and @coroutine@.
     381Section~\ref{s:Concurrency} begins concurrency, or how to create (fork) and destroy (join) a thread producing the @thread@ construct.
    378382Section~\ref{s:MutualExclusionSynchronization} discusses the two mechanisms to restricted nondeterminism when controlling shared access to resources (mutual exclusion) and timing relationships among threads (synchronization).
    379383Section~\ref{s:Monitor} shows how both mutual exclusion and synchronization are safely embedded in the @monitor@ and @thread@ constructs.
    380384Section~\ref{s:CFARuntimeStructure} describes the large-scale mechanism to structure (cluster) threads and virtual processors (kernel threads).
    381 Section~\ref{s:Performance} uses a series of microbenchmarks to compare \CFA threading with pthreads, Java OpenJDK-9, Go 1.12.6 and \uC 7.0.0.
     385Section~\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.
     386
     387
     388\section{Fundamental Execution Properties}
     389\label{s:FundamentalExecutionProperties}
     390
     391The features in a programming language should be composed from a set of fundamental properties rather than an ad hoc collection chosen by the designers.
     392To 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++}).
     393The fundamental properties are execution state, thread, and mutual-exclusion/synchronization (MES).
     394These 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).
     395While 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.
     396As 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.
     397If a compositional feature is missing, a programmer has too few/many fundamental properties resulting in a complex and/or is inefficient solution.
     398
     399In detail, the fundamental properties are:
     400\begin{description}[leftmargin=\parindent,topsep=3pt,parsep=0pt]
     401\item[\newterm{execution state}:]
     402is the state information needed by a control-flow feature to initialize, manage compute data and execution location(s), and de-initialize.
     403State is retained in fixed-sized aggregate structures and dynamic-sized stack(s), often allocated in the heap(s) managed by the runtime system.
     404The 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.
     405Control-flow transfers among execution states occurs in multiple ways, such as function call, context switch, asynchronous await, etc.
     406Because 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.
     407% An execution-state is related to the notion of a process continuation \cite{Hieb90}.
     408
     409\item[\newterm{threading}:]
     410is execution of code that occurs independently of other execution, \ie the execution resulting from a thread is sequential.
     411Multiple threads provide \emph{concurrent execution};
     412concurrent execution becomes parallel when run on multiple processing units (hyper-threading, cores, sockets).
     413There must be language mechanisms to create, block/unblock, and join with a thread.
     414
     415\item[\newterm{MES}:]
     416is the concurrency mechanisms to perform an action without interruption and establish timing relationships among multiple threads.
     417These two properties are independent, \ie mutual exclusion cannot provide synchronization and vice versa without introducing additional threads~\cite[\S~4]{Buhr05a}.
     418Limiting 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.
     419\end{description}
     420These 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.
     421
     422
     423\subsection{Execution Properties}
     424
     425Table~\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.
     426(When doing case analysis, not all combinations are meaningful.)
     427Note, basic von Neumann execution requires at least one thread and an execution state providing some form of call stack.
     428For table entries missing these minimal components, the property is borrowed from the invoker (caller).
     429
     430Case 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.
     431Case 2 is case 1 with access to shared state so callers are restricted during update (mutual exclusion) and scheduling for other threads (synchronization).
     432Case 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.
     433Note, stackless functions still borrow the caller's stack and thread, where the stack is used to preserve state across its callees.
     434Case 4 is cases 2 and 3 with protection to shared state for stackless functions.
     435Cases 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.
     436Cases 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.
     437Cases 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.
     438Hence, once started, this kind of thread must execute to completion, \ie computation only, which severely restricts runtime management.
     439Cases 11 and 12 have a stackful thread with and without safe access to shared state.
     440Execution properties increase the cost of creation and execution along with complexity of usage.
     441
     442\begin{table}
     443\caption{Execution property composition}
     444\centering
     445\label{t:ExecutionPropertyComposition}
     446\renewcommand{\arraystretch}{1.25}
     447%\setlength{\tabcolsep}{5pt}
     448\begin{tabular}{c|c||l|l}
     449\multicolumn{2}{c||}{execution properties} & \multicolumn{2}{c}{mutual exclusion / synchronization} \\
     450\hline
     451stateful                        & thread        & \multicolumn{1}{c|}{No} & \multicolumn{1}{c}{Yes} \\
     452\hline   
     453\hline   
     454No                                      & No            & \textbf{1}\ \ \ function                              & \textbf{2}\ \ \ @monitor@ function    \\
     455\hline   
     456Yes (stackless)         & No            & \textbf{3}\ \ \ @generator@                   & \textbf{4}\ \ \ @monitor@ @generator@ \\
     457\hline   
     458Yes (stackful)          & No            & \textbf{5}\ \ \ @coroutine@                   & \textbf{6}\ \ \ @monitor@ @coroutine@ \\
     459\hline   
     460No                                      & Yes           & \textbf{7}\ \ \ {\color{red}rejected} & \textbf{8}\ \ \ {\color{red}rejected} \\
     461\hline   
     462Yes (stackless)         & Yes           & \textbf{9}\ \ \ {\color{red}rejected} & \textbf{10}\ \ \ {\color{red}rejected} \\
     463\hline   
     464Yes (stackful)          & Yes           & \textbf{11}\ \ \ @thread@                             & \textbf{12}\ \ @monitor@ @thread@             \\
     465\end{tabular}
     466\end{table}
     467
     468Given 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.
     469The answers define the optimal language feature need for implementing a programming problem.
     470The next sections discusses how \CFA fills in the table with language features, while other programming languages may only provide a subset of the table.
     471
     472
     473\subsection{Design Requirements}
     474
     475The following design requirements largely stem from building \CFA on top of C.
     476\begin{itemize}[topsep=3pt,parsep=0pt]
     477\item
     478All communication must be statically type checkable for early detection of errors and efficient code generation.
     479This requirement is consistent with the fact that C is a statically-typed programming-language.
     480
     481\item
     482Direct interaction among language features must be possible allowing any feature to be selected without restricting comm\-unication.
     483For example, many concurrent languages do not provide direct communication (calls) among threads, \ie threads only communicate indirectly through monitors, channels, messages, and/or futures.
     484Indirect communication increases the number of objects, consuming more resources, and require additional synchronization and possibly data transfer.
     485
     486\item
     487All communication is performed using function calls, \ie data is transmitted from argument to parameter and results are returned from function calls.
     488Alternative forms of communication, such as call-backs, message passing, channels, or communication ports, step outside of C's normal form of communication.
     489
     490\item
     491All stateful features must follow the same declaration scopes and lifetimes as other language data.
     492For C that means at program startup, during block and function activation, and on demand using dynamic allocation.
     493
     494\item
     495MES 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.
     496Furthermore, reducing synchronization scope by encapsulating it within language constructs further reduces errors in concurrent programs.
     497
     498\item
     499Both synchronous and asynchronous communication are needed.
     500However, 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.
     501
     502\item
     503Synchronization 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.
     504Otherwise, certain concurrency problems are difficult, e.g.\ web server, disk scheduling, and the amount of concurrency is inhibited~\cite{Gentleman81}.
     505\end{itemize}
     506We have satisfied these requirements in \CFA while maintaining backwards compatibility with the huge body of legacy C programs.
     507% In contrast, other new programming languages must still access C programs (\eg operating-system service routines), but do so through fragile C interfaces.
     508
     509
     510\subsection{Asynchronous Await / Call}
     511
     512Asynchronous 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).
     513The caller detects the action's completion through a \newterm{future}/\newterm{promise}.
     514The benefit is asynchronous caller execution with respect to the callee until future resolution.
     515For 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.
     516When the caller needs the promise to be fulfilled, it executes @await@.
     517A promise-completion call-back can be part of the callee action or the caller is rescheduled;
     518in either case, the call back is executed after the promise is fulfilled.
     519While 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).
     520Specifically, 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.
     521Note, @async-await@ is just syntactic-sugar over the event engine so it does not solve these deficiencies.
     522For 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.
     523The problem is when concurrent work-units need to interact and/or block as this effects the executor, \eg stops threads.
     524While 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.
    382525
    383526
     
    385528\label{s:StatefulFunction}
    386529
    387 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.
    388 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.
    389 This capability is accomplished by retaining a data/execution \emph{closure} between invocations.
    390 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.
    391 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.
    392 Hence, refactoring a stackless coroutine may require changing it to stackful.
    393 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.
    394 As well, activating a stateful function is \emph{asymmetric} or \emph{symmetric}, identified by resume/suspend (no cycles) and resume/resume (cycles).
    395 A fixed closure activated by modified call/return is faster than a variable closure activated by context switching.
    396 Additionally, any storage management for the closure (especially in unmanaged languages, \ie no garbage collection) must also be factored into design and performance.
    397 Therefore, selecting between stackless and stackful semantics is a tradeoff between programming requirements and performance, where stackless is faster and stackful is more general.
    398 Note, creation cost is amortized across usage, so activation cost is usually the dominant factor.
     530A \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).
     531A 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.
     532However, 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.
     533Hence, 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}.
     534For 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.
     535The 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.
     536Note, a subset of generator state is a function \emph{closure}, \ie the technique of capturing lexical references when returning a nested function.
     537A 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.
     538For 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.
     539
     540There are two styles of activating a stateful function, \emph{asymmetric} or \emph{symmetric}, identified by resume/suspend (no cycles) and resume/resume (cycles).
     541These 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.
     542Selecting 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.
     543Additionally, storage management for the closure/stack (especially in unmanaged languages, \ie no garbage collection) must be factored into design and performance.
     544Note, creation cost (closure/stack) is amortized across usage, so activation cost (resume/suspend) is usually the dominant factor.
     545
     546% 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.
     547% 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.
     548% 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.
     549% 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.
     550% Hence, refactoring a stackless coroutine may require changing it to stackful.
     551% 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.
     552% As well, activating a stateful function is \emph{asymmetric} or \emph{symmetric}, identified by resume/suspend (no cycles) and resume/resume (cycles).
     553% A fixed closure activated by modified call/return is faster than a variable closure activated by context switching.
     554% Additionally, any storage management for the closure (especially in unmanaged languages, \ie no garbage collection) must also be factored into design and performance.
     555% Therefore, selecting between stackless and stackful semantics is a tradeoff between programming requirements and performance, where stackless is faster and stackful is more general.
     556% nppNote, creation cost is amortized across usage, so activation cost is usually the dominant factor.
     557
     558For 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}.
     559\begin{center}
     560\begin{tabular}{@{}l|l|l@{}}
     561\multicolumn{1}{@{}c|}{Python asymmetric generator} & \multicolumn{1}{c|}{\uC symmetric coroutine} & \multicolumn{1}{c@{}}{Pthreads thread} \\
     562\hline
     563\begin{python}
     564`def Gen():` $\LstCommentStyle{\color{red}// function}$
     565        ... yield val ...
     566gen = Gen()
     567for i in range( 10 ):
     568        print( next( gen ) )
     569\end{python}
     570&
     571\begin{uC++}
     572`_Coroutine Cycle {` $\LstCommentStyle{\color{red}// class}$
     573        Cycle * p;
     574        void main() { p->cycle(); }
     575        void cycle() { resume(); }  `};`
     576Cycle c1, c2; c1.p=&c2; c2.p=&c1; c1.cycle();
     577\end{uC++}
     578&
     579\begin{cfa}
     580void * rtn( void * arg ) { ... }
     581int i = 3, rc;
     582pthread_t t; $\C{// thread id}$
     583$\LstCommentStyle{\color{red}// function pointer}$
     584rc=pthread_create(&t, `rtn`, (void *)i);
     585\end{cfa}
     586\end{tabular}
     587\end{center}
     588\CFA's preferred presentation model for generators/coroutines/threads is a hybrid of functions and classes, giving an object-oriented flavour.
     589Essentially, the generator/coroutine/thread function is semantically coupled with a generator/coroutine/thread custom type via the type's name.
     590The custom type solves several issues, while accessing the underlying mechanisms used by the custom types is still allowed for flexibility reasons.
     591Each custom type is discussed in detail in the following sections.
     592
     593
     594\subsection{Generator}
     595
     596Stackless 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.
     597The \CFA goal is to achieve this performance target, possibly at the cost of some semantic complexity.
     598A series of different kinds of generators and their implementation demonstrate how this goal is accomplished.\footnote{
     599The \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()|.
     600Operator \lstinline+|+ is overloaded for printing, like bit-shift \lstinline|<<| in \CC.
     601The \CFA \lstinline|with| clause opens an aggregate scope making its fields directly accessible, like Pascal \lstinline|with|, but using parallel semantics;
     602multiple aggregates may be opened.
     603\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|.
     604}%
    399605
    400606\begin{figure}
     
    410616
    411617
     618
     619
    412620        int fn = f->fn; f->fn = f->fn1;
    413621                f->fn1 = f->fn + fn;
    414622        return fn;
    415 
    416623}
    417624int main() {
     
    432639void `main(Fib & fib)` with(fib) {
    433640
     641
    434642        [fn1, fn] = [1, 0];
    435643        for () {
     
    451659\begin{cfa}[aboveskip=0pt,belowskip=0pt]
    452660typedef struct {
    453         int fn1, fn;  void * `next`;
     661        int `restart`, fn1, fn;
    454662} Fib;
    455 #define FibCtor { 1, 0, NULL }
     663#define FibCtor { `0`, 1, 0 }
    456664Fib * comain( Fib * f ) {
    457         if ( f->next ) goto *f->next;
    458         f->next = &&s1;
     665        `static void * states[] = {&&s0, &&s1};`
     666        `goto *states[f->restart];`
     667  s0: f->`restart` = 1;
    459668        for ( ;; ) {
    460669                return f;
    461670          s1:; int fn = f->fn + f->fn1;
    462                         f->fn1 = f->fn; f->fn = fn;
     671                f->fn1 = f->fn; f->fn = fn;
    463672        }
    464673}
     
    472681\end{lrbox}
    473682
    474 \subfloat[C asymmetric generator]{\label{f:CFibonacci}\usebox\myboxA}
     683\subfloat[C]{\label{f:CFibonacci}\usebox\myboxA}
    475684\hspace{3pt}
    476685\vrule
    477686\hspace{3pt}
    478 \subfloat[\CFA asymmetric generator]{\label{f:CFAFibonacciGen}\usebox\myboxB}
     687\subfloat[\CFA]{\label{f:CFAFibonacciGen}\usebox\myboxB}
    479688\hspace{3pt}
    480689\vrule
    481690\hspace{3pt}
    482 \subfloat[C generator implementation]{\label{f:CFibonacciSim}\usebox\myboxC}
     691\subfloat[C generated code for \CFA version]{\label{f:CFibonacciSim}\usebox\myboxC}
    483692\caption{Fibonacci (output) asymmetric generator}
    484693\label{f:FibonacciAsymmetricGenerator}
     
    493702};
    494703void ?{}( Fmt & fmt ) { `resume(fmt);` } // constructor
    495 void ^?{}( Fmt & f ) with(f) { $\C[1.75in]{// destructor}$
     704void ^?{}( Fmt & f ) with(f) { $\C[2.25in]{// destructor}$
    496705        if ( g != 0 || b != 0 ) sout | nl; }
    497706void `main( Fmt & f )` with(f) {
     
    499708                for ( ; g < 5; g += 1 ) { $\C{// groups}$
    500709                        for ( ; b < 4; b += 1 ) { $\C{// blocks}$
    501                                 `suspend;` $\C{// wait for character}$
    502                                 while ( ch == '\n' ) `suspend;` // ignore
    503                                 sout | ch;                                              // newline
    504                         } sout | " ";  // block spacer
    505                 } sout | nl; // group newline
     710                                do { `suspend;` $\C{// wait for character}$
     711                                while ( ch == '\n' ); // ignore newline
     712                                sout | ch;                      $\C{// print character}$
     713                        } sout | " ";  $\C{// block separator}$
     714                } sout | nl; $\C{// group separator}$
    506715        }
    507716}
     
    521730\begin{cfa}[aboveskip=0pt,belowskip=0pt]
    522731typedef struct {
    523         void * next;
     732        int `restart`, g, b;
    524733        char ch;
    525         int g, b;
    526734} Fmt;
    527735void comain( Fmt * f ) {
    528         if ( f->next ) goto *f->next;
    529         f->next = &&s1;
     736        `static void * states[] = {&&s0, &&s1};`
     737        `goto *states[f->restart];`
     738  s0: f->`restart` = 1;
    530739        for ( ;; ) {
    531740                for ( f->g = 0; f->g < 5; f->g += 1 ) {
    532741                        for ( f->b = 0; f->b < 4; f->b += 1 ) {
    533                                 return;
    534                           s1:;  while ( f->ch == '\n' ) return;
     742                                do { return;  s1: ;
     743                                } while ( f->ch == '\n' );
    535744                                printf( "%c", f->ch );
    536745                        } printf( " " );
     
    539748}
    540749int main() {
    541         Fmt fmt = { NULL };  comain( &fmt ); // prime
     750        Fmt fmt = { `0` };  comain( &fmt ); // prime
    542751        for ( ;; ) {
    543752                scanf( "%c", &fmt.ch );
     
    550759\end{lrbox}
    551760
    552 \subfloat[\CFA asymmetric generator]{\label{f:CFAFormatGen}\usebox\myboxA}
    553 \hspace{3pt}
     761\subfloat[\CFA]{\label{f:CFAFormatGen}\usebox\myboxA}
     762\hspace{35pt}
    554763\vrule
    555764\hspace{3pt}
    556 \subfloat[C generator simulation]{\label{f:CFormatSim}\usebox\myboxB}
     765\subfloat[C generated code for \CFA version]{\label{f:CFormatGenImpl}\usebox\myboxB}
    557766\hspace{3pt}
    558767\caption{Formatter (input) asymmetric generator}
     
    560769\end{figure}
    561770
    562 Stateful functions appear as generators, coroutines, and threads, where presentations are based on function objects or pointers~\cite{Butenhof97, C++14, MS:VisualC++, BoostCoroutines15}.
    563 For example, Python presents generators as a function object:
    564 \begin{python}
    565 def Gen():
    566         ... `yield val` ...
    567 gen = Gen()
    568 for i in range( 10 ):
    569         print( next( gen ) )
    570 \end{python}
    571 Boost presents coroutines in terms of four functor object-types:
    572 \begin{cfa}
    573 asymmetric_coroutine<>::pull_type
    574 asymmetric_coroutine<>::push_type
    575 symmetric_coroutine<>::call_type
    576 symmetric_coroutine<>::yield_type
    577 \end{cfa}
    578 and many languages present threading using function pointers, @pthreads@~\cite{Butenhof97}, \Csharp~\cite{Csharp}, Go~\cite{Go}, and Scala~\cite{Scala}, \eg pthreads:
    579 \begin{cfa}
    580 void * rtn( void * arg ) { ... }
    581 int i = 3, rc;
    582 pthread_t t; $\C{// thread id}$
    583 `rc = pthread_create( &t, rtn, (void *)i );` $\C{// create and initialized task, type-unsafe input parameter}$
    584 \end{cfa}
    585 % void mycor( pthread_t cid, void * arg ) {
    586 %       int * value = (int *)arg;                               $\C{// type unsafe, pointer-size only}$
    587 %       // thread body
    588 % }
    589 % int main() {
    590 %       int input = 0, output;
    591 %       coroutine_t cid = coroutine_create( &mycor, (void *)&input ); $\C{// type unsafe, pointer-size only}$
    592 %       coroutine_resume( cid, (void *)input, (void **)&output ); $\C{// type unsafe, pointer-size only}$
    593 % }
    594 \CFA's preferred presentation model for generators/coroutines/threads is a hybrid of objects and functions, with an object-oriented flavour.
    595 Essentially, the generator/coroutine/thread function is semantically coupled with a generator/coroutine/thread custom type.
    596 The custom type solves several issues, while accessing the underlying mechanisms used by the custom types is still allowed.
    597 
    598 
    599 \subsection{Generator}
    600 
    601 Stackless generators have the potential to be very small and fast, \ie as small and fast as function call/return for both creation and execution.
    602 The \CFA goal is to achieve this performance target, possibly at the cost of some semantic complexity.
    603 A series of different kinds of generators and their implementation demonstrate how this goal is accomplished.
    604 
    605 Figure~\ref{f:FibonacciAsymmetricGenerator} shows an unbounded asymmetric generator for an infinite sequence of Fibonacci numbers written in C and \CFA, with a simple C implementation for the \CFA version.
     771Figure~\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.
    606772This generator is an \emph{output generator}, producing a new result on each resumption.
    607773To 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.
     
    611777The C version only has the middle execution state because the top execution state is declaration initialization.
    612778Figure~\ref{f:CFAFibonacciGen} shows the \CFA approach, which also has a manual closure, but replaces the structure with a custom \CFA @generator@ type.
    613 This generator type is then connected to a function that \emph{must be named \lstinline|main|},\footnote{
    614 The name \lstinline|main| has special meaning in C, specifically the function where a program starts execution.
    615 Hence, overloading this name for other starting points (generator/coroutine/thread) is a logical extension.}
    616 called a \emph{generator main},which takes as its only parameter a reference to the generator type.
     779Each generator type must have a function named \lstinline|main|,
     780% \footnote{
     781% The name \lstinline|main| has special meaning in C, specifically the function where a program starts execution.
     782% Leveraging starting semantics to this name for generator/coroutine/thread is a logical extension.}
     783called 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.
    617784The generator main contains @suspend@ statements that suspend execution without ending the generator versus @return@.
    618 For the Fibonacci generator-main,\footnote{
    619 The \CFA \lstinline|with| opens an aggregate scope making its fields directly accessible, like Pascal \lstinline|with|, but using parallel semantics.
    620 Multiple aggregates may be opened.}
     785For the Fibonacci generator-main,
    621786the top initialization state appears at the start and the middle execution state is denoted by statement @suspend@.
    622787Any local variables in @main@ \emph{are not retained} between calls;
     
    627792Resuming an ended (returned) generator is undefined.
    628793Function @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.
    629 Figure~\ref{f:CFibonacciSim} shows the C implementation of the \CFA generator only needs one additional field, @next@, to handle retention of execution state.
    630 The computed @goto@ at the start of the generator main, which branches after the previous suspend, adds very little cost to the resume call.
    631 Finally, an explicit generator type provides both design and performance benefits, such as multiple type-safe interface functions taking and returning arbitrary types.\footnote{
    632 The \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()|.
    633 }%
     794Figure~\ref{f:CFibonacciSim} shows the C implementation of the \CFA asymmetric generator.
     795Only one execution-state field, @restart@, is needed to subscript the suspension points in the generator.
     796At 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}).
     797Next, the computed @goto@ selects the last suspend point and branches to it.
     798The  cost of setting @restart@ and branching via the computed @goto@ adds very little cost to the suspend/resume calls.
     799
     800An advantage of the \CFA explicit generator type is the ability to allow multiple type-safe interface functions taking and returning arbitrary types.
    634801\begin{cfa}
    635802int ?()( Fib & fib ) { return `resume( fib )`.fn; } $\C[3.9in]{// function-call interface}$
    636 int ?()( Fib & fib, int N ) { for ( N - 1 ) `fib()`; return `fib()`; } $\C{// use function-call interface to skip N values}$
    637 double ?()( Fib & fib ) { return (int)`fib()` / 3.14159; } $\C{// different return type, cast prevents recursive call}\CRT$
    638 sout | (int)f1() | (double)f1() | f2( 2 ); // alternative interface, cast selects call based on return type, step 2 values
     803int ?()( Fib & fib, int N ) { for ( N - 1 ) `fib()`; return `fib()`; } $\C{// add parameter to skip N values}$
     804double ?()( Fib & fib ) { return (int)`fib()` / 3.14159; } $\C{// different return type, cast prevents recursive call}$
     805Fib f;  int i;  double d;
     806i = f();  i = f( 2 );  d = f();                                         $\C{// alternative interfaces}\CRT$
    639807\end{cfa}
    640808Now, the generator can be a separately compiled opaque-type only accessed through its interface functions.
    641809For 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.
    642810
    643 Having to manually create the generator closure by moving local-state variables into the generator type is an additional programmer burden.
    644 (This restriction is removed by the coroutine in Section~\ref{s:Coroutine}.)
    645 This requirement follows from the generality of variable-size local-state, \eg local state with a variable-length array requires dynamic allocation because the array size is unknown at compile time.
     811\begin{figure}
     812%\centering
     813\newbox\myboxA
     814\begin{lrbox}{\myboxA}
     815\begin{python}[aboveskip=0pt,belowskip=0pt]
     816def Fib():
     817        fn1, fn = 0, 1
     818        while True:
     819                `yield fn1`
     820                fn1, fn = fn, fn1 + fn
     821f1 = Fib()
     822f2 = Fib()
     823for i in range( 10 ):
     824        print( next( f1 ), next( f2 ) )
     825
     826
     827
     828
     829
     830
     831
     832
     833
     834
     835\end{python}
     836\end{lrbox}
     837
     838\newbox\myboxB
     839\begin{lrbox}{\myboxB}
     840\begin{python}[aboveskip=0pt,belowskip=0pt]
     841def Fmt():
     842        try:
     843                while True:                                             $\C[2.5in]{\# until destructor call}$
     844                        for g in range( 5 ):            $\C{\# groups}$
     845                                for b in range( 4 ):    $\C{\# blocks}$
     846                                        while True:
     847                                                ch = (yield)    $\C{\# receive from send}$
     848                                                if '\n' not in ch: $\C{\# ignore newline}$
     849                                                        break
     850                                        print( ch, end='' )     $\C{\# print character}$
     851                                print( '  ', end='' )   $\C{\# block separator}$
     852                        print()                                         $\C{\# group separator}$
     853        except GeneratorExit:                           $\C{\# destructor}$
     854                if g != 0 | b != 0:                             $\C{\# special case}$
     855                        print()
     856fmt = Fmt()
     857`next( fmt )`                                                   $\C{\# prime, next prewritten}$
     858for i in range( 41 ):
     859        `fmt.send( 'a' );`                                      $\C{\# send to yield}$
     860\end{python}
     861\end{lrbox}
     862
     863\hspace{30pt}
     864\subfloat[Fibonacci]{\label{f:PythonFibonacci}\usebox\myboxA}
     865\hspace{3pt}
     866\vrule
     867\hspace{3pt}
     868\subfloat[Formatter]{\label{f:PythonFormatter}\usebox\myboxB}
     869\caption{Python generator}
     870\label{f:PythonGenerator}
     871\end{figure}
     872
     873Having 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}).
     874This 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.
    646875However, dynamic allocation significantly increases the cost of generator creation/destruction and is a showstopper for embedded real-time programming.
    647876But 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.
    648 With respect to safety, we believe static analysis can discriminate local state from temporary variables in a generator, \ie variable usage spanning @suspend@, and generate a compile-time error.
    649 Finally, our current experience is that most generator problems have simple data state, including local state, but complex execution state, so the burden of creating the generator type is small.
     877With 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.
     878Our 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.
    650879As well, C programmers are not afraid of this kind of semantic programming requirement, if it results in very small, fast generators.
    651880
     
    669898The 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.
    670899The destructor provides a newline, if formatted text ends with a full line.
    671 Figure~\ref{f:CFormatSim} shows the C implementation of the \CFA input generator with one additional field and the computed @goto@.
    672 For contrast, Figure~\ref{f:PythonFormatter} shows the equivalent Python format generator with the same properties as the Fibonacci generator.
    673 
    674 Figure~\ref{f:DeviceDriverGen} shows a \emph{killer} asymmetric generator, a device-driver, because device drivers caused 70\%-85\% of failures in Windows/Linux~\cite{Swift05}.
    675 Device drives follow the pattern of simple data state but complex execution state, \ie finite state-machine (FSM) parsing a protocol.
    676 For example, the following protocol:
     900Figure~\ref{f:CFormatGenImpl} shows the C implementation of the \CFA input generator with one additional field and the computed @goto@.
     901For contrast, Figure~\ref{f:PythonFormatter} shows the equivalent Python format generator with the same properties as the format generator.
     902
     903% https://dl-acm-org.proxy.lib.uwaterloo.ca/
     904
     905Figure~\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}
     906Swift \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;
     907however, the calls do not retain execution state, and hence always start from the top.
     908The alternative approach for implementing device drivers is using stack-ripping.
     909However, 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.
     910
     911As an example, the following protocol:
    677912\begin{center}
    678913\ldots\, STX \ldots\, message \ldots\, ESC ETX \ldots\, message \ldots\, ETX 2-byte crc \ldots
    679914\end{center}
    680 is a network message beginning with the control character STX, ending with an ETX, and followed by a 2-byte cyclic-redundancy check.
     915is for a simple network message beginning with the control character STX, ending with an ETX, and followed by a 2-byte cyclic-redundancy check.
    681916Control characters may appear in a message if preceded by an ESC.
    682917When 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.
    683 The device driver returns a status code of its current state, and when a complete message is obtained, the operating system knows the message is in the message buffer.
    684 Hence, the device driver is an input/output generator.
    685 
    686 Note, the cost of creating and resuming the device-driver generator, @Driver@, is virtually identical to call/return, so performance in an operating-system kernel is excellent.
    687 As well, the data state is small, where variables @byte@ and @msg@ are communication variables for passing in message bytes and returning the message, and variables @lnth@, @crc@, and @sum@ are local variable that must be retained between calls and are manually hoisted into the generator type.
    688 % Manually, detecting and hoisting local-state variables is easy when the number is small.
    689 In contrast, the execution state is large, with one @resume@ and seven @suspend@s.
    690 Hence, the key benefits of the generator are correctness, safety, and maintenance because the execution states are transcribed directly into the programming language rather than using a table-driven approach.
    691 Because FSMs can be complex and frequently occur in important domains, direct generator support is important in a system programming language.
     918The 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.
     919Hence, 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.
     920The 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.
     921The conclusion is that FSMs are complex and occur in important domains, so direct generator support is important in a system programming language.
    692922
    693923\begin{figure}
    694924\centering
    695 \newbox\myboxA
    696 \begin{lrbox}{\myboxA}
    697 \begin{python}[aboveskip=0pt,belowskip=0pt]
    698 def Fib():
    699         fn1, fn = 0, 1
    700         while True:
    701                 `yield fn1`
    702                 fn1, fn = fn, fn1 + fn
    703 f1 = Fib()
    704 f2 = Fib()
    705 for i in range( 10 ):
    706         print( next( f1 ), next( f2 ) )
    707 
    708 
    709 
    710 
    711 
    712 
    713 \end{python}
    714 \end{lrbox}
    715 
    716 \newbox\myboxB
    717 \begin{lrbox}{\myboxB}
    718 \begin{python}[aboveskip=0pt,belowskip=0pt]
    719 def Fmt():
    720         try:
    721                 while True:
    722                         for g in range( 5 ):
    723                                 for b in range( 4 ):
    724                                         print( `(yield)`, end='' )
    725                                 print( '  ', end='' )
    726                         print()
    727         except GeneratorExit:
    728                 if g != 0 | b != 0:
    729                         print()
    730 fmt = Fmt()
    731 `next( fmt )`                    # prime, next prewritten
    732 for i in range( 41 ):
    733         `fmt.send( 'a' );`      # send to yield
    734 \end{python}
    735 \end{lrbox}
    736 \subfloat[Fibonacci]{\label{f:PythonFibonacci}\usebox\myboxA}
    737 \hspace{3pt}
    738 \vrule
    739 \hspace{3pt}
    740 \subfloat[Formatter]{\label{f:PythonFormatter}\usebox\myboxB}
    741 \caption{Python generator}
    742 \label{f:PythonGenerator}
    743 
    744 \bigskip
    745 
    746925\begin{tabular}{@{}l|l@{}}
    747926\begin{cfa}[aboveskip=0pt,belowskip=0pt]
     
    750929`generator` Driver {
    751930        Status status;
    752         unsigned char byte, * msg; // communication
    753         unsigned int lnth, sum;      // local state
    754         unsigned short int crc;
     931        char byte, * msg; // communication
     932        int lnth, sum;      // local state
     933        short int crc;
    755934};
    756935void ?{}( Driver & d, char * m ) { d.msg = m; }
     
    800979(The trivial cycle is a generator resuming itself.)
    801980This control flow is similar to recursion for functions but without stack growth.
    802 The steps for symmetric control-flow are creating, executing, and terminating the cycle.
     981Figure~\ref{f:PingPongFullCoroutineSteps} shows the steps for symmetric control-flow are creating, executing, and terminating the cycle.
    803982Constructing 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.
    804983(This issue occurs for any cyclic data structure.)
    805 % The example creates all the generators and then assigns the partners that form the cycle.
    806 % 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.
    807 Once the cycle is formed, the program main resumes one of the generators, and the generators can then traverse an arbitrary cycle using @resume@ to activate partner generator(s).
     984The example creates the generators, @ping@/@pong@, and then assigns the partners that form the cycle.
     985% (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.)
     986Once 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).
    808987Terminating 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).
     988Note, the creator and starter may be different, \eg if the creator calls another function that starts the cycle.
    809989The starting stack-frame is below the last active generator because the resume/resume cycle does not grow the stack.
    810 Also, since local variables are not retained in the generator function, it does not contain any objects with destructors that must be called, so the  cost is the same as a function return.
    811 Destructor cost occurs when the generator instance is deallocated, which is easily controlled by the programmer.
    812 
    813 Figure~\ref{f:CPingPongSim} shows the implementation of the symmetric generator, where the complexity is the @resume@, which needs an extension to the calling convention to perform a forward rather than backward jump.
    814 This jump-starts at the top of the next generator main to re-execute the normal calling convention to make space on the stack for its local variables.
    815 However, before the jump, the caller must reset its stack (and any registers) equivalent to a @return@, but subsequently jump forward.
    816 This semantics is basically a tail-call optimization, which compilers already perform.
    817 The example shows the assembly code to undo the generator's entry code before the direct jump.
    818 This assembly code depends on what entry code is generated, specifically if there are local variables and the level of optimization.
    819 To provide this new calling convention requires a mechanism built into the compiler, which is beyond the scope of \CFA at this time.
    820 Nevertheless, it is possible to hand generate any symmetric generators for proof of concept and performance testing.
    821 A compiler could also eliminate other artifacts in the generator simulation to further increase performance, \eg LLVM has various coroutine support~\cite{CoroutineTS}, and \CFA can leverage this support should it fork @clang@.
     990Also, 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.
     991Destructor cost occurs when the generator instance is deallocated by the creator.
    822992
    823993\begin{figure}
     
    826996\begin{cfa}[aboveskip=0pt,belowskip=0pt]
    827997`generator PingPong` {
     998        int N, i;                               // local state
    828999        const char * name;
    829         int N;
    830         int i;                          // local state
    8311000        PingPong & partner; // rebindable reference
    8321001};
    8331002
    8341003void `main( PingPong & pp )` with(pp) {
     1004
     1005
    8351006        for ( ; i < N; i += 1 ) {
    8361007                sout | name | i;
     
    8501021\begin{cfa}[escapechar={},aboveskip=0pt,belowskip=0pt]
    8511022typedef struct PingPong {
     1023        int restart, N, i;
    8521024        const char * name;
    853         int N, i;
    8541025        struct PingPong * partner;
    855         void * next;
    8561026} PingPong;
    857 #define PPCtor(name, N) {name,N,0,NULL,NULL}
     1027#define PPCtor(name, N) {0, N, 0, name, NULL}
    8581028void comain( PingPong * pp ) {
    859         if ( pp->next ) goto *pp->next;
    860         pp->next = &&cycle;
     1029        static void * states[] = {&&s0, &&s1};
     1030        goto *states[pp->restart];
     1031  s0: pp->restart = 1;
    8611032        for ( ; pp->i < pp->N; pp->i += 1 ) {
    8621033                printf( "%s %d\n", pp->name, pp->i );
    8631034                asm( "mov  %0,%%rdi" : "=m" (pp->partner) );
    8641035                asm( "mov  %rdi,%rax" );
    865                 asm( "popq %rbx" );
     1036                asm( "add  $16, %rsp" );
     1037                asm( "popq %rbp" );
    8661038                asm( "jmp  comain" );
    867           cycle: ;
     1039          s1: ;
    8681040        }
    8691041}
     
    8811053\end{figure}
    8821054
    883 Finally, part of this generator work was inspired by the recent \CCtwenty generator proposal~\cite{C++20Coroutine19} (which they call coroutines).
     1055\begin{figure}
     1056\centering
     1057\input{FullCoroutinePhases.pstex_t}
     1058\vspace*{-10pt}
     1059\caption{Symmetric coroutine steps: Ping / Pong}
     1060\label{f:PingPongFullCoroutineSteps}
     1061\end{figure}
     1062
     1063Figure~\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.
     1064Before 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.
     1065The @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.
     1066While 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.
     1067However, this assembler code depends on what entry code is generated, specifically if there are local variables and the level of optimization.
     1068Hence, 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@.
     1069For 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.
     1070
     1071Finally, 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.
    8841072Our work provides the same high-performance asymmetric generators as \CCtwenty, and extends their work with symmetric generators.
    8851073An 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:
     
    8961084\label{s:Coroutine}
    8971085
    898 Stackful coroutines extend generator semantics, \ie there is an implicit closure and @suspend@ may appear in a helper function called from the coroutine main.
     1086Stackful 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.
    8991087A coroutine is specified by replacing @generator@ with @coroutine@ for the type.
    900 Coroutine generality results in higher cost for creation, due to dynamic stack allocation, execution, due to context switching among stacks, and terminating, due to possible stack unwinding and dynamic stack deallocation.
     1088Coroutine 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.
    9011089A series of different kinds of coroutines and their implementations demonstrate how coroutines extend generators.
    9021090
    9031091First, 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.
    904 \begin{description}
    905 \item[Fibonacci]
    906 Move the declaration of @fn1@ to the start of coroutine main.
     1092\begin{center}
     1093\begin{tabular}{@{}l|l|l|l@{}}
     1094\multicolumn{1}{c|}{Fibonacci} & \multicolumn{1}{c|}{Formatter} & \multicolumn{1}{c|}{Device Driver} & \multicolumn{1}{c}{PingPong} \\
     1095\hline
    9071096\begin{cfa}[xleftmargin=0pt]
    908 void main( Fib & fib ) with(fib) {
     1097void main( Fib & fib ) ...
    9091098        `int fn1;`
    910 \end{cfa}
    911 \item[Formatter]
    912 Move the declaration of @g@ and @b@ to the for loops in the coroutine main.
     1099
     1100
     1101\end{cfa}
     1102&
    9131103\begin{cfa}[xleftmargin=0pt]
    9141104for ( `g`; 5 ) {
    9151105        for ( `b`; 4 ) {
    916 \end{cfa}
    917 \item[Device Driver]
    918 Move the declaration of @lnth@ and @sum@ to their points of initialization.
     1106
     1107
     1108\end{cfa}
     1109&
    9191110\begin{cfa}[xleftmargin=0pt]
    920         status = CONT;
    921         `unsigned int lnth = 0, sum = 0;`
    922         ...
    923         `unsigned short int crc = byte << 8;`
    924 \end{cfa}
    925 \item[PingPong]
    926 Move the declaration of @i@ to the for loop in the coroutine main.
     1111status = CONT;
     1112`int lnth = 0, sum = 0;`
     1113...
     1114`short int crc = byte << 8;`
     1115\end{cfa}
     1116&
    9271117\begin{cfa}[xleftmargin=0pt]
    928 void main( PingPong & pp ) with(pp) {
     1118void main( PingPong & pp ) ...
    9291119        for ( `i`; N ) {
    930 \end{cfa}
    931 \end{description}
     1120
     1121
     1122\end{cfa}
     1123\end{tabular}
     1124\end{center}
    9321125It 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.
    9331126\begin{cfa}
    934 unsigned int Crc() {
     1127int Crc() {
    9351128        `suspend;`
    936         unsigned short int crc = byte << 8;
     1129        short int crc = byte << 8;
    9371130        `suspend;`
    9381131        status = (crc | byte) == sum ? MSG : ECRC;
     
    9451138
    9461139\begin{comment}
    947 Figure~\ref{f:Coroutine3States} creates a @coroutine@ type, @`coroutine` Fib { int fn; }@, which provides communication, @fn@, for the \newterm{coroutine main}, @main@, which runs on the coroutine stack, and possibly multiple interface functions, \eg @next@.
     1140Figure~\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@.
    9481141Like 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.
    9491142The 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@.
    950 The interface function @next@, takes a Fibonacci instance and context switches to it using @resume@;
     1143The interface function @restart@, takes a Fibonacci instance and context switches to it using @resume@;
    9511144on restart, the Fibonacci field, @fn@, contains the next value in the sequence, which is returned.
    9521145The first @resume@ is special because it allocates the coroutine stack and cocalls its coroutine main on that stack;
     
    11141307\begin{figure}
    11151308\centering
    1116 \lstset{language=CFA,escapechar={},moredelim=**[is][\protect\color{red}]{`}{`}}% allow $
    11171309\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
    11181310\begin{cfa}
    11191311`coroutine` Prod {
    1120         Cons & c;                       // communication
     1312        Cons & c;                       $\C[1.5in]{// communication}$
    11211313        int N, money, receipt;
    11221314};
    11231315void main( Prod & prod ) with( prod ) {
    1124         // 1st resume starts here
    1125         for ( i; N ) {
     1316        for ( i; N ) {          $\C{// 1st resume}\CRT$
    11261317                int p1 = random( 100 ), p2 = random( 100 );
    1127                 sout | p1 | " " | p2;
    11281318                int status = delivery( c, p1, p2 );
    1129                 sout | " $" | money | nl | status;
    11301319                receipt += 1;
    11311320        }
    11321321        stop( c );
    1133         sout | "prod stops";
    11341322}
    11351323int payment( Prod & prod, int money ) {
     
    11521340\begin{cfa}
    11531341`coroutine` Cons {
    1154         Prod & p;                       // communication
     1342        Prod & p;                       $\C[1.5in]{// communication}$
    11551343        int p1, p2, status;
    11561344        bool done;
    11571345};
    11581346void ?{}( Cons & cons, Prod & p ) {
    1159         &cons.p = &p; // reassignable reference
     1347        &cons.p = &p;           $\C{// reassignable reference}$
    11601348        cons.[status, done ] = [0, false];
    11611349}
    11621350void main( Cons & cons ) with( cons ) {
    1163         // 1st resume starts here
    1164         int money = 1, receipt;
     1351        int money = 1, receipt; $\C{// 1st resume}\CRT$
    11651352        for ( ; ! done; ) {
    1166                 sout | p1 | " " | p2 | nl | " $" | money;
    11671353                status += 1;
    11681354                receipt = payment( p, money );
    1169                 sout | " #" | receipt;
    11701355                money += 1;
    11711356        }
    1172         sout | "cons stops";
    11731357}
    11741358int delivery( Cons & cons, int p1, int p2 ) {
     
    11911375This example is illustrative because both producer/consumer have two interface functions with @resume@s that suspend execution in these interface (helper) functions.
    11921376The 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.
    1193 The first @resume@ of @prod@ creates @prod@'s stack with a frame for @prod@'s coroutine main at the top, and context switches to it.
    1194 @prod@'s coroutine main starts, creates local-state variables that are retained between coroutine activations, and executes $N$ iterations, each generating two random values, calling the consumer to deliver the values, and printing the status returned from the consumer.
    1195 
     1377The 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.
     1378@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.
    11961379The producer call to @delivery@ transfers values into the consumer's communication variables, resumes the consumer, and returns the consumer status.
    1197 On the first resume, @cons@'s stack is created and initialized, holding local-state variables retained between subsequent activations of the coroutine.
    1198 The consumer iterates until the @done@ flag is set, prints the values delivered by the producer, increments status, and calls back to the producer via @payment@, and on return from @payment@, prints the receipt from the producer and increments @money@ (inflation).
    1199 The call from the consumer to @payment@ introduces the cycle between producer and consumer.
    1200 When @payment@ is called, the consumer copies values into the producer's communication variable and a resume is executed.
    1201 The context switch restarts the producer at the point where it last context switched, so it continues in @delivery@ after the resume.
    1202 @delivery@ returns the status value in @prod@'s coroutine main, where the status is printed.
    1203 The loop then repeats calling @delivery@, where each call resumes the consumer coroutine.
    1204 The context switch to the consumer continues in @payment@.
    1205 The consumer increments and returns the receipt to the call in @cons@'s coroutine main.
    1206 The loop then repeats calling @payment@, where each call resumes the producer coroutine.
     1380Similarly on the first resume, @cons@'s stack is created and initialized, holding local-state variables retained between subsequent activations of the coroutine.
     1381The symmetric coroutine cycle forms when the consumer calls the producer's @payment@ function, which resumes the producer in the consumer's delivery function.
     1382When the producer calls @delivery@ again, it resumes the consumer in the @payment@ function.
     1383Both interface function than return to the their corresponding coroutine-main functions for the next cycle.
    12071384Figure~\ref{f:ProdConsRuntimeStacks} shows the runtime stacks of the program main, and the coroutine mains for @prod@ and @cons@ during the cycling.
     1385As 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.
    12081386
    12091387\begin{figure}
     
    12141392\caption{Producer / consumer runtime stacks}
    12151393\label{f:ProdConsRuntimeStacks}
    1216 
    1217 \medskip
    1218 
    1219 \begin{center}
    1220 \input{FullCoroutinePhases.pstex_t}
    1221 \end{center}
    1222 \vspace*{-10pt}
    1223 \caption{Ping / Pong coroutine steps}
    1224 \label{f:PingPongFullCoroutineSteps}
    12251394\end{figure}
    12261395
    12271396Terminating 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.
    1228 Furthermore, each deallocated coroutine must guarantee all destructors are run for object allocated in the coroutine type \emph{and} allocated on the coroutine's stack at the point of suspension, which can be arbitrarily deep.
    1229 When a coroutine's main ends, its stack is already unwound so any stack allocated objects with destructors have been finalized.
     1397Furthermore, 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.
     1398In 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.
     1399% (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.)
     1400When the consumer's main ends, its stack is already unwound so any stack allocated objects with destructors are finalized.
     1401The question now is where does control continue?
     1402
    12301403The na\"{i}ve semantics for coroutine-cycle termination is to context switch to the last resumer, like executing a @suspend@/@return@ in a generator.
    12311404However, for coroutines, the last resumer is \emph{not} implicitly below the current stack frame, as for generators, because each coroutine's stack is independent.
    12321405Unfortunately, it is impossible to determine statically if a coroutine is in a cycle and unrealistic to check dynamically (graph-cycle problem).
    12331406Hence, a compromise solution is necessary that works for asymmetric (acyclic) and symmetric (cyclic) coroutines.
    1234 
    1235 Our solution is to context switch back to the first resumer (starter) once the coroutine ends.
     1407Our solution is to retain a coroutine's starter (first resumer), and context switch back to the starter when the coroutine ends.
     1408Hence, 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}).
    12361409This semantics works well for the most common asymmetric and symmetric coroutine usage patterns.
    1237 For asymmetric coroutines, it is common for the first resumer (starter) coroutine to be the only resumer.
    1238 All previous generators converted to coroutines have this property.
    1239 For symmetric coroutines, it is common for the cycle creator to persist for the lifetime of the cycle.
    1240 Hence, the starter coroutine is remembered on the first resume and ending the coroutine resumes the starter.
    1241 Figure~\ref{f:ProdConsRuntimeStacks} shows this semantic by the dashed lines from the end of the coroutine mains: @prod@ starts @cons@ so @cons@ resumes @prod@ at the end, and the program main starts @prod@ so @prod@ resumes the program main at the end.
     1410For asymmetric coroutines, it is common for the first resumer (starter) coroutine to be the only resumer;
     1411for symmetric coroutines, it is common for the cycle creator to persist for the lifetime of the cycle.
    12421412For 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.
    12431413
    1244 The producer/consumer example does not illustrate the full power of the starter semantics because @cons@ always ends first.
    1245 Assume generator @PingPong@ is converted to a coroutine.
    1246 Figure~\ref{f:PingPongFullCoroutineSteps} shows the creation, starter, and cyclic execution steps of the coroutine version.
    1247 The program main creates (declares) coroutine instances @ping@ and @pong@.
    1248 Next, program main resumes @ping@, making it @ping@'s starter, and @ping@'s main resumes @pong@'s main, making it @pong@'s starter.
    1249 Execution forms a cycle when @pong@ resumes @ping@, and cycles $N$ times.
    1250 By adjusting $N$ for either @ping@/@pong@, it is possible to have either one finish first, instead of @pong@ always ending first.
    1251 If @pong@ ends first, it resumes its starter @ping@ in its coroutine main, then @ping@ ends and resumes its starter the program main in function @start@.
    1252 If @ping@ ends first, it resumes its starter the program main in function @start@.
    1253 Regardless of the cycle complexity, the starter stack always leads back to the program main, but the stack can be entered at an arbitrary point.
    1254 Once back at the program main, coroutines @ping@ and @pong@ are deallocated.
    1255 For generators, deallocation runs the destructors for all objects in the generator type.
    1256 For coroutines, deallocation deals with objects in the coroutine type and must also run the destructors for any objects pending on the coroutine's stack for any unterminated coroutine.
    1257 Hence, if a coroutine's destructor detects the coroutine is not ended, it implicitly raises a cancellation exception (uncatchable exception) at the coroutine and resumes it so the cancellation exception can propagate to the root of the coroutine's stack destroying all local variable on the stack.
    1258 So the \CFA semantics for the generator and coroutine, ensure both can be safely deallocated at any time, regardless of their current state, like any other aggregate object.
    1259 Explicitly raising normal exceptions at another coroutine can replace flag variables, like @stop@, \eg @prod@ raises a @stop@ exception at @cons@ after it finishes generating values and resumes @cons@, which catches the @stop@ exception to terminate its loop.
    1260 
    1261 Finally, there is an interesting effect for @suspend@ with symmetric coroutines.
    1262 A coroutine must retain its last resumer to suspend back because the resumer is on a different stack.
    1263 These reverse pointers allow @suspend@ to cycle \emph{backwards}, which may be useful in certain cases.
    1264 However, there is an anomaly if a coroutine resumes itself, because it overwrites its last resumer with itself, losing the ability to resume the last external resumer.
    1265 To prevent losing this information, a self-resume does not overwrite the last resumer.
     1414Note, the producer/consumer example does not illustrate the full power of the starter semantics because @cons@ always ends first.
     1415Assume generator @PingPong@ in Figure~\ref{f:PingPongSymmetricGenerator} is converted to a coroutine.
     1416Unlike generators, coroutines have a starter structure with multiple levels, where the program main starts @ping@ and @ping@ starts @pong@.
     1417By adjusting $N$ for either @ping@/@pong@, it is possible to have either finish first.
     1418If @pong@ ends first, it resumes its starter @ping@ in its coroutine main, then @ping@ ends and resumes its starter the program main on return;
     1419if @ping@ ends first, it resumes its starter the program main on return.
     1420Regardless of the cycle complexity, the starter structure always leads back to the program main, but the path can be entered at an arbitrary point.
     1421Once 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.
     1422Hence, 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.
    12661423
    12671424
     
    12941451Users wanting to extend custom types or build their own can only do so in ways offered by the language.
    12951452Furthermore, implementing custom types without language support may display the power of a programming language.
    1296 \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.
     1453\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.
    12971454
    12981455Part 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.
     
    13041461forall( `dtype` T | is_coroutine(T) ) void $suspend$( T & ), resume( T & );
    13051462\end{cfa}
    1306 Note, copying generators/coroutines/threads is not meaningful.
    1307 For example, both the resumer and suspender descriptors can have bidirectional pointers;
    1308 copying these coroutines does not update the internal pointers so behaviour of both copies would be difficult to understand.
    1309 Furthermore, two coroutines cannot logically execute on the same stack.
    1310 A deep coroutine copy, which copies the stack, is also meaningless in an unmanaged language (no garbage collection), like C, because the stack may contain pointers to object within it that require updating for the copy.
     1463Note, 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.
    13111464The \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).
    13121465The 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.
     
    13521505The 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.
    13531506
    1354 Figure~\ref{f:CoroutineMemoryLayout} shows different memory-layout options for a coroutine (where a task is similar).
     1507Figure~\ref{f:CoroutineMemoryLayout} shows different memory-layout options for a coroutine (where a thread is similar).
    13551508The coroutine handle is the @coroutine@ instance containing programmer specified type global/communication variables across interface functions.
    13561509The coroutine descriptor contains all implicit declarations needed by the runtime, \eg @suspend@/@resume@, and can be part of the coroutine handle or separate.
    13571510The coroutine stack can appear in a number of locations and be fixed or variable sized.
    1358 Hence, the coroutine's stack could be a VLS\footnote{
    1359 We are examining variable-sized structures (VLS), where fields can be variable-sized structures or arrays.
     1511Hence, the coroutine's stack could be a variable-length structure (VLS)\footnote{
     1512We are examining VLSs, where fields can be variable-sized structures or arrays.
    13601513Once allocated, a VLS is fixed sized.}
    13611514on the allocating stack, provided the allocating stack is large enough.
    13621515For a VLS stack allocation/deallocation is an inexpensive adjustment of the stack pointer, modulo any stack constructor costs (\eg initial frame setup).
    1363 For heap stack allocation, allocation/deallocation is an expensive heap allocation (where the heap can be a shared resource), modulo any stack constructor costs.
    1364 With heap stack allocation, it is also possible to use a split (segmented) stack calling convention, available with gcc and clang, so the stack is variable sized.
     1516For stack allocation in the heap, allocation/deallocation is an expensive allocation, where the heap can be a shared resource, modulo any stack constructor costs.
     1517It 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.
    13651518Currently, \CFA supports stack/heap allocated descriptors but only fixed-sized heap allocated stacks.
    13661519In \CFA debug-mode, the fixed-sized stack is terminated with a write-only page, which catches most stack overflows.
    13671520Experience teaching concurrency with \uC~\cite{CS343} shows fixed-sized stacks are rarely an issue for students.
    1368 Split-stack allocation is under development but requires recompilation of legacy code, which may be impossible.
     1521Split-stack allocation is under development but requires recompilation of legacy code, which is not always possible.
    13691522
    13701523\begin{figure}
     
    13801533
    13811534Concurrency is nondeterministic scheduling of independent sequential execution paths (threads), where each thread has its own stack.
    1382 A single thread with multiple call stacks, \newterm{coroutining}~\cite{Conway63,Marlin80}, does \emph{not} imply concurrency~\cite[\S~2]{Buhr05a}.
    1383 In coroutining, coroutines self-schedule the thread across stacks so execution is deterministic.
     1535A single thread with multiple stacks, \ie coroutining, does \emph{not} imply concurrency~\cite[\S~3]{Buhr05a}.
     1536Coroutining self-schedule the thread across stacks so execution is deterministic.
    13841537(It is \emph{impossible} to generate a concurrency error when coroutining.)
    1385 However, coroutines are a stepping stone towards concurrency.
    1386 
    1387 The 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}.
     1538
     1539The 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}.
    13881540Therefore, a minimal concurrency system requires coroutines \emph{in conjunction with a nondeterministic scheduler}.
    1389 The resulting execution system now follows a cooperative threading model~\cite{Adya02,libdill}, called \newterm{non-preemptive scheduling}.
    1390 Adding \newterm{preemption} introduces non-cooperative scheduling, where context switching occurs randomly between any two instructions often based on a timer interrupt, called \newterm{preemptive scheduling}.
    1391 While a scheduler introduces uncertain execution among explicit context switches, preemption introduces uncertainty by introducing implicit context switches.
     1541The 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.
     1542Adding \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.
    13921543Uncertainty gives the illusion of parallelism on a single processor and provides a mechanism to access and increase performance on multiple processors.
    13931544The reason is that the scheduler/runtime have complete knowledge about resources and how to best utilized them.
    1394 However, the introduction of unrestricted nondeterminism results in the need for \newterm{mutual exclusion} and \newterm{synchronization}, which restrict nondeterminism for correctness;
     1545However, 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;
    13951546otherwise, it is impossible to write meaningful concurrent programs.
    13961547Optimal concurrent performance is often obtained by having as much nondeterminism as mutual exclusion and synchronization correctness allow.
    13971548
    1398 A scheduler can either be a stackless or stackful.
     1549A scheduler can also be stackless or stackful.
    13991550For 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.
    14001551For 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.
     
    14051556\label{s:threads}
    14061557
    1407 Threading needs the ability to start a thread and wait for its completion.
     1558Threading (Table~\ref{t:ExecutionPropertyComposition} case 11) needs the ability to start a thread and wait for its completion.
    14081559A common API for this ability is @fork@ and @join@.
    1409 \begin{cquote}
    1410 \begin{tabular}{@{}lll@{}}
    1411 \multicolumn{1}{c}{\textbf{Java}} & \multicolumn{1}{c}{\textbf{\Celeven}} & \multicolumn{1}{c}{\textbf{pthreads}} \\
    1412 \begin{cfa}
    1413 class MyTask extends Thread {...}
    1414 mytask t = new MyTask(...);
     1560\vspace{4pt}
     1561\par\noindent
     1562\begin{tabular}{@{}l|l|l@{}}
     1563\multicolumn{1}{c|}{\textbf{Java}} & \multicolumn{1}{c|}{\textbf{\Celeven}} & \multicolumn{1}{c}{\textbf{pthreads}} \\
     1564\hline
     1565\begin{cfa}
     1566class MyThread extends Thread {...}
     1567mythread t = new MyThread(...);
    14151568`t.start();` // start
    14161569// concurrency
     
    14191572&
    14201573\begin{cfa}
    1421 class MyTask { ... } // functor
    1422 MyTask mytask;
    1423 `thread t( mytask, ... );` // start
     1574class MyThread { ... } // functor
     1575MyThread mythread;
     1576`thread t( mythread, ... );` // start
    14241577// concurrency
    14251578`t.join();` // wait
     
    14341587\end{cfa}
    14351588\end{tabular}
    1436 \end{cquote}
     1589\vspace{1pt}
     1590\par\noindent
    14371591\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.
    14381592\begin{cfa}
    1439 thread MyTask {};
    1440 void main( MyTask & this ) { ... }
     1593thread MyThread {};
     1594void main( MyThread & this ) { ... }
    14411595int main() {
    1442         MyTask team`[10]`; $\C[2.5in]{// allocate stack-based threads, implicit start after construction}$
     1596        MyThread team`[10]`; $\C[2.5in]{// allocate stack-based threads, implicit start after construction}$
    14431597        // concurrency
    14441598} $\C{// deallocate stack-based threads, implicit joins before destruction}$
     
    14481602Arbitrary topologies are possible using dynamic allocation, allowing threads to outlive their declaration scope, identical to normal dynamic allocation.
    14491603\begin{cfa}
    1450 MyTask * factory( int N ) { ... return `anew( N )`; } $\C{// allocate heap-based threads, implicit start after construction}$
     1604MyThread * factory( int N ) { ... return `anew( N )`; } $\C{// allocate heap-based threads, implicit start after construction}$
    14511605int main() {
    1452         MyTask * team = factory( 10 );
     1606        MyThread * team = factory( 10 );
    14531607        // concurrency
    14541608        `delete( team );` $\C{// deallocate heap-based threads, implicit joins before destruction}\CRT$
     
    14961650
    14971651Threads 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.
    1498 Like coroutines, and for the same design reasons, \CFA provides a custom @thread@ type and a @trait@ to enforce and restrict the task-interface functions.
     1652Like coroutines, and for the same design reasons, \CFA provides a custom @thread@ type and a @trait@ to enforce and restrict the thread-interface functions.
    14991653\begin{cquote}
    15001654\begin{tabular}{@{}c@{\hspace{3\parindentlnth}}c@{}}
     
    15271681\label{s:MutualExclusionSynchronization}
    15281682
    1529 Unrestricted nondeterminism is meaningless as there is no way to know when the result is completed without synchronization.
     1683Unrestricted nondeterminism is meaningless as there is no way to know when a result is completed and safe to access.
    15301684To 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}.
    1531 Some concurrent systems eliminate mutable shared-state by switching to stateless communication like message passing~\cite{Thoth,Harmony,V-Kernel,MPI} (Erlang, MPI), channels~\cite{CSP} (CSP,Go), actors~\cite{Akka} (Akka, Scala), or functional techniques (Haskell).
     1685The 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}).
     1686Without 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.
     1687Preventing or detecting barging is a challenge with low-level locks, but made easier through higher-level constructs.
     1688This challenge is often split into two different approaches: barging \emph{avoidance} and \emph{prevention}.
     1689Approaches 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;
     1690approaches that conditionally hold locks during synchronization, \eg baton-passing~\cite{Andrews89}, prevent barging completely.
     1691
     1692At 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}.
     1693However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
     1694A significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
     1695Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
     1696Some 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).
    15321697However, these approaches introduce a new communication mechanism for concurrency different from the standard communication using function call/return.
    15331698Hence, a programmer must learn and manipulate two sets of design/programming patterns.
    15341699While this distinction can be hidden away in library code, effective use of the library still has to take both paradigms into account.
    1535 In contrast, approaches based on stateful models more closely resemble the standard call/return programming model, resulting in a single programming paradigm.
    1536 
    1537 At the lowest level, concurrent control is implemented by atomic operations, upon which different kinds of locking mechanisms are constructed, \eg semaphores~\cite{Dijkstra68b}, barriers, and path expressions~\cite{Campbell74}.
    1538 However, for productivity it is always desirable to use the highest-level construct that provides the necessary efficiency~\cite{Hochstein05}.
    1539 A newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
    1540 While this approach is pursued in hardware~\cite{Nakaike15} and system languages, like \CC~\cite{Cpp-Transactions}, the performance and feature set is still too restrictive to be the main concurrency paradigm for system languages, which is why it is rejected as the core paradigm for concurrency in \CFA.
    1541 
    1542 One of the most natural, elegant, and efficient mechanisms for mutual exclusion and synchronization for shared-memory systems is the \emph{monitor}.
    1543 First 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}.
    1544 In addition, operating-system kernels and device drivers have a monitor-like structure, although they often use lower-level primitives such as mutex locks or semaphores to simulate monitors.
    1545 For these reasons, \CFA selected monitors as the core high-level concurrency construct, upon which higher-level approaches can be easily constructed.
    1546 
    1547 
    1548 \subsection{Mutual Exclusion}
    1549 
    1550 A group of instructions manipulating a specific instance of shared data that must be performed atomically is called a \newterm{critical section}~\cite{Dijkstra65}, which is enforced by \newterm{simple mutual-exclusion}.
    1551 The generalization is called a \newterm{group critical-section}~\cite{Joung00}, where multiple tasks with the same session use the resource simultaneously and different sessions are segregated, which is enforced by \newterm{complex mutual-exclusion} providing the correct kind and number of threads using a group critical-section.
    1552 The readers/writer problem~\cite{Courtois71} is an instance of a group critical-section, where readers share a session but writers have a unique session.
    1553 
    1554 However, many solutions exist for mutual exclusion, which vary in terms of performance, flexibility and ease of use.
    1555 Methods range from low-level locks, which are fast and flexible but require significant attention for correctness, to higher-level concurrency techniques, which sacrifice some performance to improve ease of use.
    1556 Ease of use comes by either guaranteeing some problems cannot occur, \eg deadlock free, or by offering a more explicit coupling between shared data and critical section.
    1557 For example, the \CC @std::atomic<T>@ offers an easy way to express mutual-exclusion on a restricted set of operations, \eg reading/writing, for numerical types.
    1558 However, a significant challenge with locks is composability because it takes careful organization for multiple locks to be used while preventing deadlock.
    1559 Easing composability is another feature higher-level mutual-exclusion mechanisms can offer.
    1560 
    1561 
    1562 \subsection{Synchronization}
    1563 
    1564 Synchronization enforces relative ordering of execution, and synchronization tools provide numerous mechanisms to establish these timing relationships.
    1565 Low-level synchronization primitives offer good performance and flexibility at the cost of ease of use;
    1566 higher-level mechanisms often simplify usage by adding better coupling between synchronization and data, \eg receive-specific versus receive-any thread in message passing or offering specialized solutions, \eg barrier lock.
    1567 Often synchronization is used to order access to a critical section, \eg ensuring a waiting writer thread enters the critical section before a calling reader thread.
    1568 If the calling reader is scheduled before the waiting writer, the reader has barged.
    1569 Barging can result in staleness/freshness problems, where a reader barges ahead of a writer and reads temporally stale data, or a writer barges ahead of another writer overwriting data with a fresh value preventing the previous value from ever being read (lost computation).
    1570 Preventing or detecting barging is an involved challenge with low-level locks, which is made easier through higher-level constructs.
    1571 This challenge is often split into two different approaches: barging avoidance and prevention.
    1572 Algorithms that unconditionally releasing a lock for competing threads to acquire use barging avoidance during synchronization to force a barging thread to wait;
    1573 algorithms that conditionally hold locks during synchronization, \eg baton-passing~\cite{Andrews89}, prevent barging completely.
     1700In contrast, approaches based on shared-state models more closely resemble the standard call/return programming model, resulting in a single programming paradigm.
     1701Finally, a newer approach for restricting non-determinism is transactional memory~\cite{Herlihy93}.
     1702While 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.
    15741703
    15751704
     
    15771706\label{s:Monitor}
    15781707
    1579 A \textbf{monitor} is a set of functions that ensure mutual exclusion when accessing shared state.
    1580 More precisely, a monitor is a programming technique that implicitly binds mutual exclusion to static function scope, as opposed to locks, where mutual-exclusion is defined by acquire/release calls, independent of lexical context (analogous to block and heap storage allocation).
     1708One 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).
     1709First 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}.
     1710In 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.
     1711For these reasons, \CFA selected monitors as the core high-level concurrency construct, upon which higher-level approaches can be easily constructed.
     1712
     1713Specifically, a \textbf{monitor} is a set of functions that ensure mutual exclusion when accessing shared state.
     1714More 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).
    15811715Restricting acquire/release points eases programming, comprehension, and maintenance, at a slight cost in flexibility and efficiency.
    15821716\CFA uses a custom @monitor@ type and leverages declaration semantics (deallocation) to protect active or waiting threads in a monitor.
    15831717
    15841718The following is a \CFA monitor implementation of an atomic counter.
    1585 \begin{cfa}[morekeywords=nomutex]
     1719\begin{cfa}
    15861720`monitor` Aint { int cnt; }; $\C[4.25in]{// atomic integer counter}$
    1587 int ++?( Aint & `mutex`$\(_{opt}\)$ this ) with( this ) { return ++cnt; } $\C{// increment}$
    1588 int ?=?( Aint & `mutex`$\(_{opt}\)$ lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions with int}\CRT$
    1589 int ?=?( int & lhs, Aint & `mutex`$\(_{opt}\)$ rhs ) with( rhs ) { lhs = cnt; }
    1590 \end{cfa}
    1591 % The @Aint@ constructor, @?{}@, uses the \lstinline[morekeywords=nomutex]@nomutex@ qualifier indicating mutual exclusion is unnecessary during construction because an object is inaccessible (private) until after it is initialized.
    1592 % (While a constructor may publish its address into a global variable, doing so generates a race-condition.)
    1593 The prefix increment operation, @++?@, is normally @mutex@, indicating mutual exclusion is necessary during function execution, to protect the incrementing from race conditions, unless there is an atomic increment instruction for the implementation type.
    1594 The assignment operators provide bidirectional conversion between an atomic and normal integer without accessing field @cnt@;
    1595 these operations only need @mutex@, if reading/writing the implementation type is not atomic.
    1596 The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics, which is similar to the \CC template @std::atomic@.
     1721int ++?( Aint & `mutex` this ) with( this ) { return ++cnt; } $\C{// increment}$
     1722int ?=?( Aint & `mutex` lhs, int rhs ) with( lhs ) { cnt = rhs; } $\C{// conversions with int, mutex optional}\CRT$
     1723int ?=?( int & lhs, Aint & `mutex` rhs ) with( rhs ) { lhs = cnt; }
     1724\end{cfa}
     1725The operators use the parameter-only declaration type-qualifier @mutex@ to mark which parameters require locking during function execution to protect from race conditions.
     1726The assignment operators provide bidirectional conversion between an atomic and normal integer without accessing field @cnt@.
     1727(These operations only need @mutex@, if reading/writing the implementation type is not atomic.)
     1728The atomic counter is used without any explicit mutual-exclusion and provides thread-safe semantics.
    15971729\begin{cfa}
    15981730int i = 0, j = 0, k = 5;
     
    16021734i = x; j = y; k = z;
    16031735\end{cfa}
     1736Note, 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@.
    16041737
    16051738\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.
     1739\newpage
    16061740\begin{cfa}
    16071741monitor M { ... } m;
     
    16121746\end{cfa}
    16131747\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.
    1614 Similar safety is offered by \emph{explicit} mechanisms like \CC RAII;
    1615 monitor \emph{implicit} safety ensures no programmer usage errors.
     1748Similar 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.
    16161749Furthermore, 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;
    16171750RAII is purely a mutual-exclusion mechanism (see Section~\ref{s:Scheduling}).
     
    16391772\end{cquote}
    16401773The @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).
    1641 % Copying a lock is insecure because it is possible to copy an open lock and then use the open copy when the original lock is closed to simultaneously access the shared data.
    1642 % Copying a monitor is secure because both the lock and shared data are copies, but copying the shared data is meaningless because it no longer represents a unique entity.
    16431774Similarly, 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.
    16441775The custom monitor type also inserts any locks needed to implement the mutual exclusion semantics.
     
    16521783For example, a monitor may be passed through multiple helper functions before it is necessary to acquire the monitor's mutual exclusion.
    16531784
    1654 The benefit of mandatory monitor qualifiers is self-documentation, but requiring both @mutex@ and \lstinline[morekeywords=nomutex]@nomutex@ for all monitor parameters is redundant.
    1655 Instead, the semantics has one qualifier as the default and the other required.
    1656 For example, make the safe @mutex@ qualifier the default because assuming \lstinline[morekeywords=nomutex]@nomutex@ may cause subtle errors.
    1657 Alternatively, make the unsafe \lstinline[morekeywords=nomutex]@nomutex@ qualifier the default because it is the \emph{normal} parameter semantics while @mutex@ parameters are rare.
    1658 Providing a default qualifier implies knowing whether a parameter is a monitor.
    1659 Since \CFA relies heavily on traits as an abstraction mechanism, types can coincidentally match the monitor trait but not be a monitor, similar to inheritance where a shape and playing card can both be drawable.
    1660 For this reason, \CFA requires programmers to identify the kind of parameter with the @mutex@ keyword and uses no keyword to mean \lstinline[morekeywords=nomutex]@nomutex@.
     1785\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.
     1786Hence, @mutex@ parameters are documentation, at the function and its prototype, to both programmer and compiler, without other redundant keywords.
     1787Furthermore, \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.
    16611788
    16621789The next semantic decision is establishing which parameter \emph{types} may be qualified with @mutex@.
     
    16721799Function @f3@ has a multiple object matrix, and @f4@ a multiple object data structure.
    16731800While shown shortly, multiple object acquisition is possible, but the number of objects must be statically known.
    1674 Therefore, \CFA only acquires one monitor per parameter with at most one level of indirection, excluding pointers as it is impossible to statically determine the size.
     1801Therefore, \CFA only acquires one monitor per parameter with exactly one level of indirection, and exclude pointer types to unknown sized arrays.
    16751802
    16761803For object-oriented monitors, \eg Java, calling a mutex member \emph{implicitly} acquires mutual exclusion of the receiver object, @`rec`.foo(...)@.
     
    16791806While object-oriented monitors can be extended with a mutex qualifier for multiple-monitor members, no prior example of this feature could be found.}
    16801807called \newterm{bulk acquire}.
    1681 \CFA guarantees acquisition order is consistent across calls to @mutex@ functions using the same monitors as arguments, so acquiring multiple monitors is safe from deadlock.
     1808\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.
    16821809Figure~\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.
    16831810A \CFA programmer only has to manage when to acquire mutual exclusion;
     
    16991826void transfer( BankAccount & `mutex` my,
    17001827        BankAccount & `mutex` your, int me2you ) {
    1701 
     1828        // bulk acquire
    17021829        deposit( my, -me2you ); // debit
    17031830        deposit( your, me2you ); // credit
     
    17291856void transfer( BankAccount & my,
    17301857                        BankAccount & your, int me2you ) {
    1731         `scoped_lock lock( my.m, your.m );`
     1858        `scoped_lock lock( my.m, your.m );` // bulk acquire
    17321859        deposit( my, -me2you ); // debit
    17331860        deposit( your, me2you ); // credit
     
    17571884\end{figure}
    17581885
    1759 Users can still force the acquiring order by using @mutex@/\lstinline[morekeywords=nomutex]@nomutex@.
     1886Users can still force the acquiring order by using or not using @mutex@.
    17601887\begin{cfa}
    17611888void foo( M & mutex m1, M & mutex m2 ); $\C{// acquire m1 and m2}$
    1762 void bar( M & mutex m1, M & /* nomutex */ m2 ) { $\C{// acquire m1}$
     1889void bar( M & mutex m1, M & m2 ) { $\C{// only acquire m1}$
    17631890        ... foo( m1, m2 ); ... $\C{// acquire m2}$
    17641891}
    1765 void baz( M & /* nomutex */ m1, M & mutex m2 ) { $\C{// acquire m2}$
     1892void baz( M & m1, M & mutex m2 ) { $\C{// only acquire m2}$
    17661893        ... foo( m1, m2 ); ... $\C{// acquire m1}$
    17671894}
     
    18061933% 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.
    18071934% 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.
    1808 This section discusses monitor scheduling for waiting threads eligible for entry, \ie which thread gets the shared resource next. (See Section~\ref{s:RuntimeStructureCluster} for scheduling threads on virtual processors.)
    1809 While monitor mutual-exclusion provides safe access to shared data, the monitor data may indicate that a thread accessing it cannot proceed, \eg a bounded buffer may be full/empty so produce/consumer threads must block.
    1810 Leaving the monitor and trying again (busy waiting) is impractical for high-level programming.
    1811 Monitors eliminate busy waiting by providing synchronization to schedule threads needing access to the shared data, where threads block versus spinning.
     1935This 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.)
     1936While 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.
     1937Leaving the monitor and retrying (busy waiting) is impractical for high-level programming.
     1938
     1939Monitors 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.
    18121940Synchronization is generally achieved with internal~\cite{Hoare74} or external~\cite[\S~2.9.2]{uC++} scheduling.
    1813 \newterm{Internal scheduling} is characterized by each thread entering the monitor and making an individual decision about proceeding or blocking, while \newterm{external scheduling} is characterized by an entering thread making a decision about proceeding for itself and on behalf of other threads attempting entry.
    1814 Finally, \CFA monitors do not allow calling threads to barge ahead of signalled threads, which simplifies synchronization among threads in the monitor and increases correctness.
    1815 If barging is allowed, synchronization between a signaller and signallee is difficult, often requiring additional flags and multiple unblock/block cycles.
    1816 In fact, signals-as-hints is completely opposite from that proposed by Hoare in the seminal paper on monitors~\cite[p.~550]{Hoare74}.
     1941\newterm{Internal} (largely) schedules threads located \emph{inside} the monitor and is accomplished using condition variables with signal and wait.
     1942\newterm{External} (largely) schedules threads located \emph{outside} the monitor and is accomplished with the @waitfor@ statement.
     1943Note, 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.
     1944For complex scheduling, the approaches can be combined, so there can be an equal number of threads waiting inside and outside.
     1945
     1946\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.
     1947A 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.
     1948Preventing barging comes directly from Hoare's semantics in the seminal paper on monitors~\cite[p.~550]{Hoare74}.
    18171949% \begin{cquote}
    18181950% 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.
    18191951% 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}
    18201952% \end{cquote}
    1821 Furthermore, \CFA concurrency has no spurious wakeup~\cite[\S~9]{Buhr05a}, which eliminates an implicit form of self barging.
    1822 Hence, a \CFA @wait@ statement is not enclosed in a @while@ loop retesting a blocking predicate, which can cause thread starvation due to barging.
    1823 
    1824 Figure~\ref{f:MonitorScheduling} shows general internal/external scheduling (for the bounded-buffer example in Figure~\ref{f:InternalExternalScheduling}).
    1825 External calling threads block on the calling queue, if the monitor is occupied, otherwise they enter in FIFO order.
    1826 Internal threads block on condition queues via @wait@ and reenter from the condition in FIFO order.
    1827 Alternatively, internal threads block on urgent from the @signal_block@ or @waitfor@, and reenter implicitly when the monitor becomes empty, \ie, the thread in the monitor exits or waits.
    1828 
    1829 There are three signalling mechanisms to unblock waiting threads to enter the monitor.
    1830 Note, signalling cannot have the signaller and signalled thread in the monitor simultaneously because of the mutual exclusion, so either the signaller or signallee can proceed.
    1831 For internal scheduling, threads are unblocked from condition queues using @signal@, where the signallee is moved to urgent and the signaller continues (solid line).
    1832 Multiple signals move multiple signallees to urgent until the condition is empty.
    1833 When the signaller exits or waits, a thread blocked on urgent is processed before calling threads to prevent barging.
     1953Furthermore, \CFA concurrency has no spurious wakeup~\cite[\S~9]{Buhr05a}, which eliminates an implicit self barging.
     1954
     1955Monitor mutual-exclusion means signalling cannot have the signaller and signalled thread in the monitor simultaneously, so only the signaller or signallee can proceed.
     1956Figure~\ref{f:MonitorScheduling} shows internal/external scheduling for the bounded-buffer examples in Figure~\ref{f:GenericBoundedBuffer}.
     1957For 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).
     1958Multiple signals move multiple signallees to urgent until the condition queue is empty.
     1959When the signaller exits or waits, a thread is implicitly unblocked from urgent (if available) before unblocking a calling thread to prevent barging.
    18341960(Java conceptually moves the signalled thread to the calling queue, and hence, allows barging.)
    1835 The alternative unblock is in the opposite order using @signal_block@, where the signaller is moved to urgent and the signallee continues (dashed line), and is implicitly unblocked from urgent when the signallee exits or waits.
    1836 
    1837 For external scheduling, the condition queues are not used;
    1838 instead threads are unblocked directly from the calling queue using @waitfor@ based on function names requesting mutual exclusion.
    1839 (The linear search through the calling queue to locate a particular call can be reduced to $O(1)$.)
    1840 The @waitfor@ has the same semantics as @signal_block@, where the signalled thread executes before the signallee, which waits on urgent.
    1841 Executing multiple @waitfor@s from different signalled functions causes the calling threads to move to urgent.
    1842 External scheduling requires urgent to be a stack, because the signaller expects to execute immediately after the specified monitor call has exited or waited.
    1843 Internal scheduling behaves the same for an urgent stack or queue, except for multiple signalling, where the threads unblock from urgent in reverse order from signalling.
    1844 If the restart order is important, multiple signalling by a signal thread can be transformed into daisy-chain signalling among threads, where each thread signals the next thread.
    1845 We tried both a stack for @waitfor@ and queue for signalling, but that resulted in complex semantics about which thread enters next.
    1846 Hence, \CFA uses a single urgent stack to correctly handle @waitfor@ and adequately support both forms of signalling.
     1961Signal 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.
     1962Specifically, the @wait@ function atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the function's parameter list.
     1963Signalling is unconditional because signalling an empty condition queue does nothing.
     1964It 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.
     1965In \CFA, a condition queue can be created/stored independently.
    18471966
    18481967\begin{figure}
     
    18621981\end{figure}
    18631982
    1864 Figure~\ref{f:BBInt} shows a \CFA generic bounded-buffer with internal scheduling, where producers/consumers enter the monitor, detect the buffer is full/empty, and block on an appropriate condition variable, @full@/@empty@.
    1865 The @wait@ function atomically blocks the calling thread and implicitly releases the monitor lock(s) for all monitors in the function's parameter list.
    1866 The appropriate condition variable is signalled to unblock an opposite kind of thread after an element is inserted/removed from the buffer.
    1867 Signalling is unconditional, because signalling an empty condition variable does nothing.
    1868 It is common to declare condition variables as monitor fields to prevent shared access, hence no locking is required for access as the conditions are protected by the monitor lock.
    1869 In \CFA, a condition variable can be created/stored independently.
    1870 % To still prevent expensive locking on access, a condition variable is tied to a \emph{group} of monitors on first use, called \newterm{branding}, resulting in a low-cost boolean test to detect sharing from other monitors.
    1871 
    1872 % Signalling semantics cannot have the signaller and signalled thread in the monitor simultaneously, which means:
    1873 % \begin{enumerate}
    1874 % \item
    1875 % The signalling thread returns immediately and the signalled thread continues.
    1876 % \item
    1877 % The signalling thread continues and the signalled thread is marked for urgent unblocking at the next scheduling point (exit/wait).
    1878 % \item
    1879 % The signalling thread blocks but is marked for urgent unblocking at the next scheduling point and the signalled thread continues.
    1880 % \end{enumerate}
    1881 % The first approach is too restrictive, as it precludes solving a reasonable class of problems, \eg dating service (see Figure~\ref{f:DatingService}).
    1882 % \CFA supports the next two semantics as both are useful.
    1883 
    18841983\begin{figure}
    18851984\centering
     
    18931992                T elements[10];
    18941993        };
    1895         void ?{}( Buffer(T) & buffer ) with(buffer) {
     1994        void ?{}( Buffer(T) & buf ) with(buf) {
    18961995                front = back = count = 0;
    18971996        }
    1898         void insert( Buffer(T) & mutex buffer, T elem )
    1899                                 with(buffer) {
    1900                 if ( count == 10 ) `wait( empty )`;
    1901                 // insert elem into buffer
     1997
     1998        void insert(Buffer(T) & mutex buf, T elm) with(buf){
     1999                if ( count == 10 ) `wait( empty )`; // full ?
     2000                // insert elm into buf
    19022001                `signal( full )`;
    19032002        }
    1904         T remove( Buffer(T) & mutex buffer ) with(buffer) {
    1905                 if ( count == 0 ) `wait( full )`;
    1906                 // remove elem from buffer
     2003        T remove( Buffer(T) & mutex buf ) with(buf) {
     2004                if ( count == 0 ) `wait( full )`; // empty ?
     2005                // remove elm from buf
    19072006                `signal( empty )`;
    1908                 return elem;
     2007                return elm;
    19092008        }
    19102009}
    19112010\end{cfa}
    19122011\end{lrbox}
    1913 
    1914 % \newbox\myboxB
    1915 % \begin{lrbox}{\myboxB}
    1916 % \begin{cfa}[aboveskip=0pt,belowskip=0pt]
    1917 % forall( otype T ) { // distribute forall
    1918 %       monitor Buffer {
    1919 %
    1920 %               int front, back, count;
    1921 %               T elements[10];
    1922 %       };
    1923 %       void ?{}( Buffer(T) & buffer ) with(buffer) {
    1924 %               [front, back, count] = 0;
    1925 %       }
    1926 %       T remove( Buffer(T) & mutex buffer ); // forward
    1927 %       void insert( Buffer(T) & mutex buffer, T elem )
    1928 %                               with(buffer) {
    1929 %               if ( count == 10 ) `waitfor( remove, buffer )`;
    1930 %               // insert elem into buffer
    1931 %
    1932 %       }
    1933 %       T remove( Buffer(T) & mutex buffer ) with(buffer) {
    1934 %               if ( count == 0 ) `waitfor( insert, buffer )`;
    1935 %               // remove elem from buffer
    1936 %
    1937 %               return elem;
    1938 %       }
    1939 % }
    1940 % \end{cfa}
    1941 % \end{lrbox}
    19422012
    19432013\newbox\myboxB
    19442014\begin{lrbox}{\myboxB}
    19452015\begin{cfa}[aboveskip=0pt,belowskip=0pt]
     2016forall( otype T ) { // distribute forall
     2017        monitor Buffer {
     2018
     2019                int front, back, count;
     2020                T elements[10];
     2021        };
     2022        void ?{}( Buffer(T) & buf ) with(buf) {
     2023                front = back = count = 0;
     2024        }
     2025        T remove( Buffer(T) & mutex buf ); // forward
     2026        void insert(Buffer(T) & mutex buf, T elm) with(buf){
     2027                if ( count == 10 ) `waitfor( remove : buf )`;
     2028                // insert elm into buf
     2029
     2030        }
     2031        T remove( Buffer(T) & mutex buf ) with(buf) {
     2032                if ( count == 0 ) `waitfor( insert : buf )`;
     2033                // remove elm from buf
     2034
     2035                return elm;
     2036        }
     2037}
     2038\end{cfa}
     2039\end{lrbox}
     2040
     2041\subfloat[Internal scheduling]{\label{f:BBInt}\usebox\myboxA}
     2042\hspace{1pt}
     2043\vrule
     2044\hspace{3pt}
     2045\subfloat[External scheduling]{\label{f:BBExt}\usebox\myboxB}
     2046
     2047\caption{Generic bounded buffer}
     2048\label{f:GenericBoundedBuffer}
     2049\end{figure}
     2050
     2051The @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).
     2052Signal 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.
     2053Using @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.
     2054
     2055External 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++}.
     2056While prior languages use external scheduling solely for thread interaction, \CFA generalizes it to both monitors and threads.
     2057External scheduling allows waiting for events from other threads while restricting unrelated events, that would otherwise have to wait on condition queues in the monitor.
     2058Scheduling 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.
     2059Specifically, 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.
     2060(The linear search through the calling queue to locate a particular call can be reduced to $O(1)$.)
     2061Hence, 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.
     2062Now 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.
     2063For 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.
     2064Hence, this mechanism is done in terms of control flow, next call, versus in terms of data, channels, as in Go/Rust @select@.
     2065While both mechanisms have strengths and weaknesses, \CFA uses the control-flow mechanism to be consistent with other language features.
     2066
     2067Figure~\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.
     2068For 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.
     2069To 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@.
     2070An 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.
     2071For 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@.
     2072The writer does a similar action for each reader or writer using the resource.
     2073Note, no new calls to @StartRead@/@StartWrite@ may occur when waiting for the call to @EndRead@/@EndWrite@.
     2074
     2075\begin{figure}
     2076\centering
     2077\newbox\myboxA
     2078\begin{lrbox}{\myboxA}
     2079\begin{cfa}[aboveskip=0pt,belowskip=0pt]
     2080enum RW { READER, WRITER };
    19462081monitor ReadersWriter {
    1947         int rcnt, wcnt; // readers/writer using resource
     2082        int rcnt, wcnt; // readers/writer using resource
     2083        `condition RWers;`
    19482084};
    19492085void ?{}( ReadersWriter & rw ) with(rw) {
     
    19522088void EndRead( ReadersWriter & mutex rw ) with(rw) {
    19532089        rcnt -= 1;
     2090        if ( rcnt == 0 ) `signal( RWers )`;
    19542091}
    19552092void EndWrite( ReadersWriter & mutex rw ) with(rw) {
    19562093        wcnt = 0;
     2094        `signal( RWers );`
    19572095}
    19582096void StartRead( ReadersWriter & mutex rw ) with(rw) {
    1959         if ( wcnt > 0 ) `waitfor( EndWrite, rw );`
     2097        if ( wcnt !=0 || ! empty( RWers ) )
     2098                `wait( RWers, READER )`;
    19602099        rcnt += 1;
     2100        if ( ! empty(RWers) && `front(RWers) == READER` )
     2101                `signal( RWers )`;  // daisy-chain signalling
    19612102}
    19622103void StartWrite( ReadersWriter & mutex rw ) with(rw) {
    1963         if ( wcnt > 0 ) `waitfor( EndWrite, rw );`
    1964         else while ( rcnt > 0 ) `waitfor( EndRead, rw );`
     2104        if ( wcnt != 0 || rcnt != 0 ) `wait( RWers, WRITER )`;
     2105
    19652106        wcnt = 1;
    19662107}
    1967 
    19682108\end{cfa}
    19692109\end{lrbox}
    19702110
    1971 \subfloat[Generic bounded buffer, internal scheduling]{\label{f:BBInt}\usebox\myboxA}
    1972 \hspace{3pt}
     2111\newbox\myboxB
     2112\begin{lrbox}{\myboxB}
     2113\begin{cfa}[aboveskip=0pt,belowskip=0pt]
     2114
     2115monitor ReadersWriter {
     2116        int rcnt, wcnt; // readers/writer using resource
     2117
     2118};
     2119void ?{}( ReadersWriter & rw ) with(rw) {
     2120        rcnt = wcnt = 0;
     2121}
     2122void EndRead( ReadersWriter & mutex rw ) with(rw) {
     2123        rcnt -= 1;
     2124
     2125}
     2126void EndWrite( ReadersWriter & mutex rw ) with(rw) {
     2127        wcnt = 0;
     2128
     2129}
     2130void StartRead( ReadersWriter & mutex rw ) with(rw) {
     2131        if ( wcnt > 0 ) `waitfor( EndWrite : rw );`
     2132
     2133        rcnt += 1;
     2134
     2135
     2136}
     2137void StartWrite( ReadersWriter & mutex rw ) with(rw) {
     2138        if ( wcnt > 0 ) `waitfor( EndWrite : rw );`
     2139        else while ( rcnt > 0 ) `waitfor( EndRead : rw );`
     2140        wcnt = 1;
     2141}
     2142\end{cfa}
     2143\end{lrbox}
     2144
     2145\subfloat[Internal scheduling]{\label{f:RWInt}\usebox\myboxA}
     2146\hspace{1pt}
    19732147\vrule
    19742148\hspace{3pt}
    1975 \subfloat[Readers / writer lock, external scheduling]{\label{f:RWExt}\usebox\myboxB}
    1976 
    1977 \caption{Internal / external scheduling}
    1978 \label{f:InternalExternalScheduling}
     2149\subfloat[External scheduling]{\label{f:RWExt}\usebox\myboxB}
     2150
     2151\caption{Readers / writer lock}
     2152\label{f:ReadersWriterLock}
    19792153\end{figure}
    19802154
    1981 Figure~\ref{f:BBInt} can be transformed into external scheduling by removing the condition variables and signals/waits, and adding the following lines at the locations of the current @wait@s in @insert@/@remove@, respectively.
    1982 \begin{cfa}[aboveskip=2pt,belowskip=1pt]
    1983 if ( count == 10 ) `waitfor( remove, buffer )`;       |      if ( count == 0 ) `waitfor( insert, buffer )`;
    1984 \end{cfa}
    1985 Here, the producers/consumers detects a full/\-empty buffer and prevents more producers/consumers from entering the monitor until there is a free/empty slot in the buffer.
    1986 External scheduling is controlled by the @waitfor@ statement, which atomically blocks the calling thread, releases the monitor lock, and restricts the function calls that can next acquire mutual exclusion.
    1987 If the buffer is full, only calls to @remove@ can acquire the buffer, and if the buffer is empty, only calls to @insert@ can acquire the buffer.
    1988 Threads calling excluded functions block outside of (external to) the monitor on the calling queue, versus blocking on condition queues inside of (internal to) the monitor.
    1989 Figure~\ref{f:RWExt} shows a readers/writer lock written using external scheduling, where a waiting reader detects a writer using the resource and restricts further calls until the writer exits by calling @EndWrite@.
    1990 The writer does a similar action for each reader or writer using the resource.
    1991 Note, no new calls to @StarRead@/@StartWrite@ may occur when waiting for the call to @EndRead@/@EndWrite@.
    1992 External scheduling allows waiting for events from other threads while restricting unrelated events, that would otherwise have to wait on conditions in the monitor.
    1993 The mechnaism can be done in terms of control flow, \eg Ada @accept@ or \uC @_Accept@, or in terms of data, \eg Go @select@ on channels.
    1994 While both mechanisms have strengths and weaknesses, this project uses the control-flow mechanism to be consistent with other language features.
    1995 % Two challenges specific to \CFA for external scheduling are loose object-definitions (see Section~\ref{s:LooseObjectDefinitions}) and multiple-monitor functions (see Section~\ref{s:Multi-MonitorScheduling}).
    1996 
    1997 Figure~\ref{f:DatingService} shows a dating service demonstrating non-blocking and blocking signalling.
    1998 The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
    1999 A thread blocks until an appropriate partner arrives.
    2000 The complexity is exchanging phone numbers in the monitor because of the mutual-exclusion property.
    2001 For 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.
    2002 For signal-block scheduling, the implicit urgent-queue replaces the explict @exchange@-condition and @signal_block@ puts the finding thread on the urgent condition and unblocks the matcher.
    2003 The dating service is an example of a monitor that cannot be written using external scheduling because it requires knowledge of calling parameters to make scheduling decisions, and parameters of waiting threads are unavailable;
    2004 as well, an arriving thread may not find a partner and must wait, which requires a condition variable, and condition variables imply internal scheduling.
    2005 Furthermore, 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.
    2006 Putting loops around the @wait@s does not correct the problem;
    2007 the simple solution must be restructured to account for barging.
     2155Finally, external scheduling requires urgent to be a stack, because the signaller expects to execute immediately after the specified monitor call has exited or waited.
     2156Internal schedulling performing multiple signalling results in unblocking from urgent in the reverse order from signalling.
     2157It is rare for the unblocking order to be important as an unblocked thread can be time-sliced immediately after leaving the monitor.
     2158If the unblocking order is important, multiple signalling can be restructured into daisy-chain signalling, where each thread signals the next thread.
     2159Hence, \CFA uses a single urgent stack to correctly handle @waitfor@ and adequately support both forms of signalling.
     2160(Advanced @waitfor@ features are discussed in Section~\ref{s:ExtendedWaitfor}.)
    20082161
    20092162\begin{figure}
     
    20192172};
    20202173int girl( DS & mutex ds, int phNo, int ccode ) {
    2021         if ( is_empty( Boys[ccode] ) ) {
     2174        if ( empty( Boys[ccode] ) ) {
    20222175                wait( Girls[ccode] );
    20232176                GirlPhNo = phNo;
     
    20462199};
    20472200int girl( DS & mutex ds, int phNo, int ccode ) {
    2048         if ( is_empty( Boys[ccode] ) ) { // no compatible
     2201        if ( empty( Boys[ccode] ) ) { // no compatible
    20492202                wait( Girls[ccode] ); // wait for boy
    20502203                GirlPhNo = phNo; // make phone number available
     
    20662219\qquad
    20672220\subfloat[\lstinline@signal_block@]{\label{f:DatingSignalBlock}\usebox\myboxB}
    2068 \caption{Dating service}
    2069 \label{f:DatingService}
     2221\caption{Dating service Monitor}
     2222\label{f:DatingServiceMonitor}
    20702223\end{figure}
    20712224
    2072 In summation, for internal scheduling, non-blocking signalling (as in the producer/consumer example) is used when the signaller is providing the cooperation for a waiting thread;
    2073 the signaller enters the monitor and changes state, detects a waiting threads that can use the state, performs a non-blocking signal on the condition queue for the waiting thread, and exits the monitor to run concurrently.
    2074 The waiter unblocks next from the urgent queue, uses/takes the state, and exits the monitor.
    2075 Blocking signal is the reverse, where the waiter is providing the cooperation for the signalling thread;
    2076 the signaller enters the monitor, detects a waiting thread providing the necessary state, performs a blocking signal to place it on the urgent queue and unblock the waiter.
    2077 The waiter changes state and exits the monitor, and the signaller unblocks next from the urgent queue to use/take the state.
     2225Figure~\ref{f:DatingServiceMonitor} shows a dating service demonstrating non-blocking and blocking signalling.
     2226The dating service matches girl and boy threads with matching compatibility codes so they can exchange phone numbers.
     2227A thread blocks until an appropriate partner arrives.
     2228The complexity is exchanging phone numbers in the monitor because of the mutual-exclusion property.
     2229For 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.
     2230For 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.
     2231
     2232The dating service is an important example of a monitor that cannot be written using external scheduling.
     2233First, because scheduling requires knowledge of calling parameters to make matching decisions, and parameters of calling threads are unavailable within the monitor.
     2234For 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.
     2235Second, 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.
     2236For example, if a girl thread could determine there is no calling boy with the same @ccode@, it must wait until a matching boy arrives.
     2237Finally, 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.
     2238This situation shows rechecking the waiting condition and waiting again (signals-as-hints) fails, requiring significant restructured to account for barging.
    20782239
    20792240Both internal and external scheduling extend to multiple monitors in a natural way.
    20802241\begin{cquote}
    2081 \begin{tabular}{@{}l@{\hspace{3\parindentlnth}}l@{}}
     2242\begin{tabular}{@{}l@{\hspace{2\parindentlnth}}l@{}}
    20822243\begin{cfa}
    20832244monitor M { `condition e`; ... };
     
    20902251&
    20912252\begin{cfa}
    2092 void rtn$\(_1\)$( M & mutex m1, M & mutex m2 );
     2253void rtn$\(_1\)$( M & mutex m1, M & mutex m2 ); // overload rtn
    20932254void rtn$\(_2\)$( M & mutex m1 );
    20942255void bar( M & mutex m1, M & mutex m2 ) {
    2095         ... waitfor( `rtn` ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\), m1, m2 )}$
    2096         ... waitfor( `rtn, m1` ); ... // $\LstCommentStyle{waitfor( rtn\(_2\), m1 )}$
     2256        ... waitfor( `rtn`${\color{red}\(_1\)}$ ); ...       // $\LstCommentStyle{waitfor( rtn\(_1\) : m1, m2 )}$
     2257        ... waitfor( `rtn${\color{red}\(_2\)}$ : m1` ); ...
    20972258}
    20982259\end{cfa}
     
    21012262For @wait( e )@, the default semantics is to atomically block the signaller and release all acquired mutex parameters, \ie @wait( e, m1, m2 )@.
    21022263To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @wait( e, m1 )@.
    2103 Wait cannot statically verifies the released monitors are the acquired mutex-parameters without disallowing separately compiled helper functions calling @wait@.
    2104 While \CC supports bulk locking, @wait@ only accepts a single lock for a condition variable, so bulk locking with condition variables is asymmetric.
     2264Wait cannot statically verify the released monitors are the acquired mutex-parameters without disallowing separately compiled helper functions calling @wait@.
     2265While \CC supports bulk locking, @wait@ only accepts a single lock for a condition queue, so bulk locking with condition queues is asymmetric.
    21052266Finally, a signaller,
    21062267\begin{cfa}
     
    21112272must 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.
    21122273
    2113 Similarly, for @waitfor( rtn )@, the default semantics is to atomically block the acceptor and release all acquired mutex parameters, \ie @waitfor( rtn, m1, m2 )@.
    2114 To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn, m1 )@.
     2274Similarly, for @waitfor( rtn )@, the default semantics is to atomically block the acceptor and release all acquired mutex parameters, \ie @waitfor( rtn : m1, m2 )@.
     2275To override the implicit multi-monitor wait, specific mutex parameter(s) can be specified, \eg @waitfor( rtn : m1 )@.
    21152276@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.
    21162277% When an overloaded function appears in an @waitfor@ statement, calls to any function with that name are accepted.
     
    21202281void rtn( M & mutex m );
    21212282`int` rtn( M & mutex m );
    2122 waitfor( (`int` (*)( M & mutex ))rtn, m );
    2123 \end{cfa}
    2124 
    2125 The ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock.
     2283waitfor( (`int` (*)( M & mutex ))rtn : m );
     2284\end{cfa}
     2285
     2286The ability to release a subset of acquired monitors can result in a \newterm{nested monitor}~\cite{Lister77} deadlock (see Section~\ref{s:MutexAcquisition}).
     2287\newpage
    21262288\begin{cfa}
    21272289void foo( M & mutex m1, M & mutex m2 ) {
    2128         ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired )}$
    2129 void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2 )}$
     2290        ... wait( `e, m1` ); ...                                $\C{// release m1, keeping m2 acquired}$
     2291void bar( M & mutex m1, M & mutex m2 ) {        $\C{// must acquire m1 and m2}$
    21302292        ... signal( `e` ); ...
    21312293\end{cfa}
    21322294The @wait@ only releases @m1@ so the signalling thread cannot acquire @m1@ and @m2@ to enter @bar@ and @signal@ the condition.
    2133 While deadlock can occur with multiple/nesting acquisition, this is a consequence of locks, and by extension monitors, not being perfectly composable.
    2134 
     2295While deadlock can occur with multiple/nesting acquisition, this is a consequence of locks, and by extension monitor locking is not perfectly composable.
    21352296
    21362297
    21372298\subsection{\texorpdfstring{Extended \protect\lstinline@waitfor@}{Extended waitfor}}
     2299\label{s:ExtendedWaitfor}
    21382300
    21392301Figure~\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.
     
    21462308Hence, the terminating @else@ clause allows a conditional attempt to accept a call without blocking.
    21472309If both @timeout@ and @else@ clause are present, the @else@ must be conditional, or the @timeout@ is never triggered.
    2148 There is also a traditional future wait queue (not shown) (\eg Microsoft (@WaitForMultipleObjects@)), to wait for a specified number of future elements in the queue.
     2310There is also a traditional future wait queue (not shown) (\eg Microsoft @WaitForMultipleObjects@), to wait for a specified number of future elements in the queue.
     2311Finally, there is a shorthand for specifying multiple functions using the same set of monitors: @waitfor( f, g, h : m1, m2, m3 )@.
    21492312
    21502313\begin{figure}
     
    21732336The right example accepts either @mem1@ or @mem2@ if @C1@ and @C2@ are true.
    21742337
    2175 An interesting use of @waitfor@ is accepting the @mutex@ destructor to know when an object is deallocated, \eg assume the bounded buffer is restructred from a monitor to a thread with the following @main@.
     2338An 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@.
    21762339\begin{cfa}
    21772340void main( Buffer(T) & buffer ) with(buffer) {
    21782341        for () {
    2179                 `waitfor( ^?{}, buffer )` break;
    2180                 or when ( count != 20 ) waitfor( insert, buffer ) { ... }
    2181                 or when ( count != 0 ) waitfor( remove, buffer ) { ... }
     2342                `waitfor( ^?{} : buffer )` break;
     2343                or when ( count != 20 ) waitfor( insert : buffer ) { ... }
     2344                or when ( count != 0 ) waitfor( remove : buffer ) { ... }
    21822345        }
    21832346        // clean up
     
    22712434To support this efficient semantics (and prevent barging), the implementation maintains a list of monitors acquired for each blocked thread.
    22722435When a signaller exits or waits in a monitor function/statement, the front waiter on urgent is unblocked if all its monitors are released.
    2273 Implementing a fast subset check for the necessary released monitors is important.
     2436Implementing a fast subset check for the necessary released monitors is important and discussed in the following sections.
    22742437% 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.
    22752438
    22762439
    2277 \subsection{Loose Object Definitions}
    2278 \label{s:LooseObjectDefinitions}
    2279 
    2280 In an object-oriented programming language, a class includes an exhaustive list of operations.
    2281 A new class can add members via static inheritance but the subclass still has an exhaustive list of operations.
    2282 (Dynamic member adding, \eg JavaScript~\cite{JavaScript}, is not considered.)
    2283 In the object-oriented scenario, the type and all its operators are always present at compilation (even separate compilation), so it is possible to number the operations in a bit mask and use an $O(1)$ compare with a similar bit mask created for the operations specified in a @waitfor@.
    2284 
    2285 However, in \CFA, monitor functions can be statically added/removed in translation units, making a fast subset check difficult.
    2286 \begin{cfa}
    2287         monitor M { ... }; // common type, included in .h file
    2288 translation unit 1
    2289         void `f`( M & mutex m );
    2290         void g( M & mutex m ) { waitfor( `f`, m ); }
    2291 translation unit 2
    2292         void `f`( M & mutex m ); $\C{// replacing f and g for type M in this translation unit}$
    2293         void `g`( M & mutex m );
    2294         void h( M & mutex m ) { waitfor( `f`, m ) or waitfor( `g`, m ); } $\C{// extending type M in this translation unit}$
    2295 \end{cfa}
    2296 The @waitfor@ statements in each translation unit cannot form a unique bit-mask because the monitor type does not carry that information.
     2440\subsection{\texorpdfstring{\protect\lstinline@waitfor@ Implementation}{waitfor Implementation}}
     2441\label{s:waitforImplementation}
     2442
     2443In 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}).
     2444Knowing 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.
     2445
     2446\begin{figure}
     2447\centering
     2448\begin{lrbox}{\myboxA}
     2449\begin{uC++}[aboveskip=0pt,belowskip=0pt]
     2450$\emph{translation unit 1}$
     2451_Monitor B { // common type in .h file
     2452        _Mutex virtual void `f`( ... );
     2453        _Mutex virtual void `g`( ... );
     2454        _Mutex virtual void w1( ... ) { ... _Accept(`f`, `g`); ... }
     2455};
     2456$\emph{translation unit 2}$
     2457// include B
     2458_Monitor D : public B { // inherit
     2459        _Mutex void `h`( ... ); // add
     2460        _Mutex void w2( ... ) { ... _Accept(`f`, `h`); ... }
     2461};
     2462\end{uC++}
     2463\end{lrbox}
     2464
     2465\begin{lrbox}{\myboxB}
     2466\begin{cfa}[aboveskip=0pt,belowskip=0pt]
     2467$\emph{translation unit 1}$
     2468monitor M { ... }; // common type in .h file
     2469void `f`( M & mutex m, ... );
     2470void `g`( M & mutex m, ... );
     2471void w1( M & mutex m, ... ) { ... waitfor(`f`, `g` : m); ... }
     2472
     2473$\emph{translation unit 2}$
     2474// include M
     2475extern void `f`( M & mutex m, ... ); // import f but not g
     2476void `h`( M & mutex m ); // add
     2477void w2( M & mutex m, ... ) { ... waitfor(`f`, `h` : m); ... }
     2478
     2479\end{cfa}
     2480\end{lrbox}
     2481
     2482\subfloat[\uC]{\label{f:uCinheritance}\usebox\myboxA}
     2483\hspace{3pt}
     2484\vrule
     2485\hspace{3pt}
     2486\subfloat[\CFA]{\label{f:CFinheritance}\usebox\myboxB}
     2487\caption{Member / Function visibility}
     2488\label{f:MemberFunctionVisibility}
     2489\end{figure}
     2490
     2491However, 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.
     2492(A possible way to construct a dense mapping is at link or load-time.)
    22972493Hence, function pointers are used to identify the functions listed in the @waitfor@ statement, stored in a variable-sized array.
    2298 Then, the same implementation approach used for the urgent stack is used for the calling queue.
    2299 Each caller has a list of monitors acquired, and the @waitfor@ statement performs a (usually short) linear search matching functions in the @waitfor@ list with called functions, and then verifying the associated mutex locks can be transfers.
    2300 (A possible way to construct a dense mapping is at link or load-time.)
     2494Then, the same implementation approach used for the urgent stack (see Section~\ref{s:Scheduling}) is used for the calling queue.
     2495Each 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.
    23012496
    23022497
     
    23132508The solution is for the programmer to disambiguate:
    23142509\begin{cfa}
    2315 waitfor( f, `m2` ); $\C{// wait for call to f with argument m2}$
     2510waitfor( f : `m2` ); $\C{// wait for call to f with argument m2}$
    23162511\end{cfa}
    23172512Both 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@.
     
    23202515monitor M { ... };
    23212516void f( M & mutex m1, M & mutex m2 );
    2322 void g( M & mutex m1, M & mutex m2 ) { waitfor( f, `m1, m2` ); $\C{// wait for call to f with arguments m1 and m2}$
     2517void g( M & mutex m1, M & mutex m2 ) { waitfor( f : `m1, m2` ); $\C{// wait for call to f with arguments m1 and m2}$
    23232518\end{cfa}
    23242519Again, the set of monitors passed to the @waitfor@ statement must be entirely contained in the set of monitors already acquired by the accepting function.
    2325 Also, the order of the monitors in a @waitfor@ statement is unimportant.
    2326 
    2327 Figure~\ref{f:UnmatchedMutexSets} shows an example where, for internal and external scheduling with multiple monitors, a signalling or accepting thread must match exactly, \ie partial matching results in waiting.
    2328 For both examples, the set of monitors is disjoint so unblocking is impossible.
     2520% Also, the order of the monitors in a @waitfor@ statement must match the order of the mutex parameters.
     2521
     2522Figure~\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.
     2523In both cases, the set of monitors is disjoint so unblocking is impossible.
    23292524
    23302525\begin{figure}
     
    23552550}
    23562551void g( M1 & mutex m1, M2 & mutex m2 ) {
    2357         waitfor( f, m1, m2 );
     2552        waitfor( f : m1, m2 );
    23582553}
    23592554g( `m11`, m2 ); // block on accept
     
    23702565\end{figure}
    23712566
    2372 
    2373 \subsection{\texorpdfstring{\protect\lstinline@mutex@ Threads}{mutex Threads}}
    2374 
    2375 Threads in \CFA can also be monitors to allow \emph{direct communication} among threads, \ie threads can have mutex functions that are called by other threads.
    2376 Hence, all monitor features are available when using threads.
    2377 Figure~\ref{f:DirectCommunication} shows a comparison of direct call communication in \CFA with direct channel communication in Go.
    2378 (Ada provides a similar mechanism to the \CFA direct communication.)
    2379 The program main in both programs communicates directly with the other thread versus indirect communication where two threads interact through a passive monitor.
    2380 Both direct and indirection thread communication are valuable tools in structuring concurrent programs.
    2381 
    23822567\begin{figure}
    23832568\centering
     
    23862571
    23872572struct Msg { int i, j; };
    2388 thread GoRtn { int i;  float f;  Msg m; };
     2573monitor thread GoRtn { int i;  float f;  Msg m; };
    23892574void mem1( GoRtn & mutex gortn, int i ) { gortn.i = i; }
    23902575void mem2( GoRtn & mutex gortn, float f ) { gortn.f = f; }
     
    23962581        for () {
    23972582
    2398                 `waitfor( mem1, gortn )` sout | i;  // wait for calls
    2399                 or `waitfor( mem2, gortn )` sout | f;
    2400                 or `waitfor( mem3, gortn )` sout | m.i | m.j;
    2401                 or `waitfor( ^?{}, gortn )` break;
     2583                `waitfor( mem1 : gortn )` sout | i;  // wait for calls
     2584                or `waitfor( mem2 : gortn )` sout | f;
     2585                or `waitfor( mem3 : gortn )` sout | m.i | m.j;
     2586                or `waitfor( ^?{} : gortn )` break; // low priority
    24022587
    24032588        }
     
    24532638\hspace{3pt}
    24542639\subfloat[Go]{\label{f:Gochannel}\usebox\myboxB}
    2455 \caption{Direct communication}
    2456 \label{f:DirectCommunication}
     2640\caption{Direct versus indirect communication}
     2641\label{f:DirectCommunicationComparison}
     2642
     2643\medskip
     2644
     2645\begin{cfa}
     2646monitor thread DatingService {
     2647        condition Girls[CompCodes], Boys[CompCodes];
     2648        int girlPhoneNo, boyPhoneNo, ccode;
     2649};
     2650int girl( DatingService & mutex ds, int phoneno, int code ) with( ds ) {
     2651        girlPhoneNo = phoneno;  ccode = code;
     2652        `wait( Girls[ccode] );`                                                         $\C{// wait for boy}$
     2653        girlPhoneNo = phoneno;  return boyPhoneNo;
     2654}
     2655int boy( DatingService & mutex ds, int phoneno, int code ) with( ds ) {
     2656        boyPhoneNo = phoneno;  ccode = code;
     2657        `wait( Boys[ccode] );`                                                          $\C{// wait for girl}$
     2658        boyPhoneNo = phoneno;  return girlPhoneNo;
     2659}
     2660void main( DatingService & ds ) with( ds ) {                    $\C{// thread starts, ds defaults to mutex}$
     2661        for () {
     2662                waitfor( ^?{} ) break;                                                  $\C{// high priority}$
     2663                or waitfor( girl )                                                              $\C{// girl called, compatible boy ? restart boy then girl}$
     2664                        if ( ! is_empty( Boys[ccode] ) ) { `signal_block( Boys[ccode] );  signal_block( Girls[ccode] );` }
     2665                or waitfor( boy ) {                                                             $\C{// boy called, compatible girl ? restart girl then boy}$
     2666                        if ( ! is_empty( Girls[ccode] ) ) { `signal_block( Girls[ccode] );  signal_block( Boys[ccode] );` }
     2667        }
     2668}
     2669\end{cfa}
     2670\caption{Direct communication dating service}
     2671\label{f:DirectCommunicationDatingService}
    24572672\end{figure}
    24582673
     
    24692684void main( Ping & pi ) {
    24702685        for ( 10 ) {
    2471                 `waitfor( ping, pi );`
     2686                `waitfor( ping : pi );`
    24722687                `pong( po );`
    24732688        }
     
    24822697        for ( 10 ) {
    24832698                `ping( pi );`
    2484                 `waitfor( pong, po );`
     2699                `waitfor( pong : po );`
    24852700        }
    24862701}
     
    24972712
    24982713
    2499 \subsection{Execution Properties}
    2500 
    2501 Table~\ref{t:ObjectPropertyComposition} shows how the \CFA high-level constructs cover 3 fundamental execution properties: thread, stateful function, and mutual exclusion.
    2502 Case 1 is a basic object, with none of the new execution properties.
    2503 Case 2 allows @mutex@ calls to Case 1 to protect shared data.
    2504 Case 3 allows stateful functions to suspend/resume but restricts operations because the state is stackless.
    2505 Case 4 allows @mutex@ calls to Case 3 to protect shared data.
    2506 Cases 5 and 6 are the same as 3 and 4 without restriction because the state is stackful.
    2507 Cases 7 and 8 are rejected because a thread cannot execute without a stackful state in a preemptive environment when context switching from the signal handler.
    2508 Cases 9 and 10 have a stackful thread without and with @mutex@ calls.
    2509 For situations where threads do not require direct communication, case 9 provides faster creation/destruction by eliminating @mutex@ setup.
    2510 
    2511 \begin{table}
    2512 \caption{Object property composition}
    2513 \centering
    2514 \label{t:ObjectPropertyComposition}
    2515 \renewcommand{\arraystretch}{1.25}
    2516 %\setlength{\tabcolsep}{5pt}
    2517 \begin{tabular}{c|c||l|l}
    2518 \multicolumn{2}{c||}{object properties} & \multicolumn{2}{c}{mutual exclusion} \\
    2519 \hline
    2520 thread  & stateful                              & \multicolumn{1}{c|}{No} & \multicolumn{1}{c}{Yes} \\
    2521 \hline
    2522 \hline
    2523 No              & No                                    & \textbf{1}\ \ \ aggregate type                & \textbf{2}\ \ \ @monitor@ aggregate type \\
    2524 \hline
    2525 No              & Yes (stackless)               & \textbf{3}\ \ \ @generator@                   & \textbf{4}\ \ \ @monitor@ @generator@ \\
    2526 \hline
    2527 No              & Yes (stackful)                & \textbf{5}\ \ \ @coroutine@                   & \textbf{6}\ \ \ @monitor@ @coroutine@ \\
    2528 \hline
    2529 Yes             & No / Yes (stackless)  & \textbf{7}\ \ \ {\color{red}rejected} & \textbf{8}\ \ \ {\color{red}rejected} \\
    2530 \hline
    2531 Yes             & Yes (stackful)                & \textbf{9}\ \ \ @thread@                              & \textbf{10}\ \ @monitor@ @thread@ \\
    2532 \end{tabular}
    2533 \end{table}
     2714\subsection{\texorpdfstring{\protect\lstinline@monitor@ Generators / Coroutines / Threads}{monitor Generators / Coroutines / Threads}}
     2715
     2716\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.
     2717All monitor features are available within these mutex functions.
     2718For 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:
     2719\begin{cfa}
     2720void fmt( Fmt & mutex fmt, char ch ) { fmt.ch = ch; resume( fmt ) }
     2721\end{cfa}
     2722multiple threads can safely pass characters for formatting.
     2723
     2724Figure~\ref{f:DirectCommunicationComparison} shows a comparison of direct call-communication in \CFA versus indirect channel-communication in Go.
     2725(Ada has a similar mechanism to \CFA direct communication.)
     2726The 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.
     2727Communication by multiple threads is safe for the @gortn@ thread via mutex calls in \CFA or channel assignment in Go.
     2728
     2729Figure~\ref{f:DirectCommunicationDatingService} shows the dating-service problem in Figure~\ref{f:DatingServiceMonitor} extended from indirect monitor communication to direct thread communication.
     2730When 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.
     2731Notice, the dating server is postponing requests for an unspecified time while continuing to accept new requests.
     2732For complex servers (web-servers), there can be hundreds of lines of code in the thread main and safe interaction with clients can be complex.
    25342733
    25352734
     
    25372736
    25382737For 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.
    2539 Some of these low-level mechanism are used in the \CFA runtime, but we strongly advocate using high-level mechanisms whenever possible.
     2738Some of these low-level mechanism are used to build the \CFA runtime, but we always advocate using high-level mechanisms whenever possible.
    25402739
    25412740
     
    25802779\begin{cfa}
    25812780struct Adder {
    2582     int * row, cols;
     2781        int * row, cols;
    25832782};
    25842783int operator()() {
     
    26392838\label{s:RuntimeStructureCluster}
    26402839
    2641 A \newterm{cluster} is a collection of threads and virtual processors (abstract kernel-thread) that execute the (user) threads from its own ready queue (like an OS executing kernel threads).
     2840A \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.
     2841The term \newterm{virtual processor} is introduced as a synonym for kernel thread to disambiguate between user and kernel thread.
     2842From the language perspective, a virtual processor is an actual processor (core).
     2843
    26422844The purpose of a cluster is to control the amount of parallelism that is possible among threads, plus scheduling and other execution defaults.
    26432845The default cluster-scheduler is single-queue multi-server, which provides automatic load-balancing of threads on processors.
     
    26582860Programs may use more virtual processors than hardware processors.
    26592861On a multiprocessor, kernel threads are distributed across the hardware processors resulting in virtual processors executing in parallel.
    2660 (It is possible to use affinity to lock a virtual processor onto a particular hardware processor~\cite{affinityLinux, affinityWindows, affinityFreebsd, affinityNetbsd, affinityMacosx}, which is used when caching issues occur or for heterogeneous hardware processors.)
     2862(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
    26612863The \CFA runtime attempts to block unused processors and unblock processors as the system load increases;
    2662 balancing the workload with processors is difficult because it requires future knowledge, \ie what will the applicaton workload do next.
     2864balancing the workload with processors is difficult because it requires future knowledge, \ie what will the application workload do next.
    26632865Preemption occurs on virtual processors rather than user threads, via operating-system interrupts.
    26642866Thus virtual processors execute user threads, where preemption frequency applies to a virtual processor, so preemption occurs randomly across the executed user threads.
     
    26952897Nondeterministic 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.
    26962898This atomic reliance can fail on multi-core machines, because execution across cores is nondeterministic.
    2697 A different reason for not supporting preemption is that it significantly complicates the runtime system, \eg Microsoft runtime does not support interrupts and on Linux systems, interrupts are complex (see below).
     2899A 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).
    26982900Preemption is normally handled by setting a countdown timer on each virtual processor.
    2699 When the timer expires, an interrupt is delivered, and the interrupt handler resets the countdown timer, and if the virtual processor is executing in user code, the signal handler performs a user-level context-switch, or if executing in the language runtime kernel, the preemption is ignored or rolled forward to the point where the runtime kernel context switches back to user code.
     2901When 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.
    27002902Multiple signal handlers may be pending.
    27012903When 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.
    27022904The 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;
    27032905therefore, the same signal mask is required for all virtual processors in a cluster.
    2704 Because preemption frequency is usually long (1 millisecond) performance cost is negligible.
    2705 
    2706 Linux switched a decade ago from specific to arbitrary process signal-delivery for applications with multiple kernel threads.
    2707 \begin{cquote}
    2708 A process-directed signal may be delivered to any one of the threads that does not currently have the signal blocked.
    2709 If more than one of the threads has the signal unblocked, then the kernel chooses an arbitrary thread to which it will deliver the signal.
    2710 SIGNAL(7) - Linux Programmer's Manual
    2711 \end{cquote}
     2906Because preemption interval is usually long (1 millisecond) performance cost is negligible.
     2907
     2908Linux switched a decade ago from specific to arbitrary virtual-processor signal-delivery for applications with multiple kernel threads.
     2909In 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.
    27122910Hence, 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).
    27132911To 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.
     
    27272925\label{s:Performance}
    27282926
    2729 To verify the implementation of the \CFA runtime, a series of microbenchmarks are performed comparing \CFA with pthreads, Java OpenJDK-9, Go 1.12.6 and \uC 7.0.0.
     2927To 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.
    27302928For 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.
    2731 The benchmark computer is an AMD Opteron\texttrademark\ 6380 NUMA 64-core, 8 socket, 2.5 GHz processor, running Ubuntu 16.04.6 LTS, and \CFA/\uC are compiled with gcc 6.5.
     2929The 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.
    27322930
    27332931All benchmarks are run using the following harness. (The Java harness is augmented to circumvent JIT issues.)
    27342932\begin{cfa}
    2735 unsigned int N = 10_000_000;
    2736 #define BENCH( `run` ) Time before = getTimeNsec();  `run;`  Duration result = (getTimeNsec() - before) / N;
    2737 \end{cfa}
    2738 The method used to get time is @clock_gettime( CLOCK_REALTIME )@.
    2739 Each benchmark is performed @N@ times, where @N@ varies depending on the benchmark;
    2740 the total time is divided by @N@ to obtain the average time for a benchmark.
    2741 Each benchmark experiment is run 31 times.
     2933#define BENCH( `run` ) uint64_t start = cputime_ns();  `run;`  double result = (double)(cputime_ns() - start) / N;
     2934\end{cfa}
     2935where CPU time in nanoseconds is from the appropriate language clock.
     2936Each 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.
     2937The total time is divided by @N@ to obtain the average time for a benchmark.
     2938Each benchmark experiment is run 13 times and the average appears in the table.
    27422939All omitted tests for other languages are functionally identical to the \CFA tests and available online~\cite{CforallBenchMarks}.
    2743 % tar --exclude=.deps --exclude=Makefile --exclude=Makefile.in --exclude=c.c --exclude=cxx.cpp --exclude=fetch_add.c -cvhf benchmark.tar benchmark
    2744 
    2745 \paragraph{Object Creation}
    2746 
    2747 Object creation is measured by creating/deleting the specific kind of concurrent object.
    2748 Figure~\ref{f:creation} shows the code for \CFA, with results in Table~\ref{tab:creation}.
    2749 The only note here is that the call stacks of \CFA coroutines are lazily created, therefore without priming the coroutine to force stack creation, the creation cost is artificially low.
    2750 
    2751 \begin{multicols}{2}
    2752 \lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
    2753 \begin{cfa}
    2754 @thread@ MyThread {};
    2755 void @main@( MyThread & ) {}
    2756 int main() {
    2757         BENCH( for ( N ) { @MyThread m;@ } )
    2758         sout | result`ns;
    2759 }
    2760 \end{cfa}
    2761 \captionof{figure}{\CFA object-creation benchmark}
    2762 \label{f:creation}
    2763 
    2764 \columnbreak
    2765 
    2766 \vspace*{-16pt}
    2767 \captionof{table}{Object creation comparison (nanoseconds)}
    2768 \label{tab:creation}
    2769 
    2770 \begin{tabular}[t]{@{}r*{3}{D{.}{.}{5.2}}@{}}
    2771 \multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} & \multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
    2772 \CFA Coroutine Lazy             & 13.2          & 13.1          & 0.44          \\
    2773 \CFA Coroutine Eager    & 531.3         & 536.0         & 26.54         \\
    2774 \CFA Thread                             & 2074.9        & 2066.5        & 170.76        \\
    2775 \uC Coroutine                   & 89.6          & 90.5          & 1.83          \\
    2776 \uC Thread                              & 528.2         & 528.5         & 4.94          \\
    2777 Goroutine                               & 4068.0        & 4113.1        & 414.55        \\
    2778 Java Thread                             & 103848.5      & 104295.4      & 2637.57       \\
    2779 Pthreads                                & 33112.6       & 33127.1       & 165.90
    2780 \end{tabular}
    2781 \end{multicols}
    2782 
    2783 
    2784 \paragraph{Context-Switching}
     2940% tar --exclude-ignore=exclude -cvhf benchmark.tar benchmark
     2941
     2942\paragraph{Context Switching}
    27852943
    27862944In procedural programming, the cost of a function call is important as modularization (refactoring) increases.
    2787 (In many cases, a compiler inlines function calls to eliminate this cost.)
    2788 Similarly, when modularization extends to coroutines/tasks, the time for a context switch becomes a relevant factor.
     2945(In many cases, a compiler inlines function calls to increase the size and number of basic blocks for optimizing.)
     2946Similarly, when modularization extends to coroutines/threads, the time for a context switch becomes a relevant factor.
    27892947The coroutine test is from resumer to suspender and from suspender to resumer, which is two context switches.
     2948%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.
     2949For 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@).
    27902950The thread test is using yield to enter and return from the runtime kernel, which is two context switches.
    27912951The difference in performance between coroutine and thread context-switch is the cost of scheduling for threads, whereas coroutines are self-scheduling.
    2792 Figure~\ref{f:ctx-switch} only shows the \CFA code for coroutines/threads (other systems are similar) with all results in Table~\ref{tab:ctx-switch}.
     2952Figure~\ref{f:ctx-switch} shows the \CFA code for a coroutine/thread with results in Table~\ref{t:ctx-switch}.
     2953
     2954% From: Gregor Richards <gregor.richards@uwaterloo.ca>
     2955% To: "Peter A. Buhr" <pabuhr@plg2.cs.uwaterloo.ca>
     2956% Date: Fri, 24 Jan 2020 13:49:18 -0500
     2957%
     2958% I can also verify that the previous version, which just tied a bunch of promises together, *does not* go back to the
     2959% event loop at all in the current version of Node. Presumably they're taking advantage of the fact that the ordering of
     2960% events is intentionally undefined to just jump right to the next 'then' in the chain, bypassing event queueing
     2961% entirely. That's perfectly correct behavior insofar as its difference from the specified behavior isn't observable, but
     2962% it isn't typical or representative of much anything useful, because most programs wouldn't have whole chains of eager
     2963% promises. Also, it's not representative of *anything* you can do with async/await, as there's no way to encode such an
     2964% eager chain that way.
    27932965
    27942966\begin{multicols}{2}
     
    27962968\begin{cfa}[aboveskip=0pt,belowskip=0pt]
    27972969@coroutine@ C {} c;
    2798 void main( C & ) { for ( ;; ) { @suspend;@ } }
     2970void main( C & ) { while () { @suspend;@ } }
    27992971int main() { // coroutine test
    28002972        BENCH( for ( N ) { @resume( c );@ } )
    2801         sout | result`ns;
    2802 }
    2803 int main() { // task test
     2973        sout | result;
     2974}
     2975int main() { // thread test
    28042976        BENCH( for ( N ) { @yield();@ } )
    2805         sout | result`ns;
     2977        sout | result;
    28062978}
    28072979\end{cfa}
     
    28132985\vspace*{-16pt}
    28142986\captionof{table}{Context switch comparison (nanoseconds)}
    2815 \label{tab:ctx-switch}
     2987\label{t:ctx-switch}
    28162988\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
    28172989\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
    2818 C function              & 1.8   & 1.8   & 0.01  \\
    2819 \CFA generator  & 2.4   & 2.2   & 0.25  \\
    2820 \CFA Coroutine  & 36.2  & 36.2  & 0.25  \\
    2821 \CFA Thread             & 93.2  & 93.5  & 2.09  \\
    2822 \uC Coroutine   & 52.0  & 52.1  & 0.51  \\
    2823 \uC Thread              & 96.2  & 96.3  & 0.58  \\
    2824 Goroutine               & 141.0 & 141.3 & 3.39  \\
    2825 Java Thread             & 374.0 & 375.8 & 10.38 \\
    2826 Pthreads Thread & 361.0 & 365.3 & 13.19
     2990C function                      & 1.8           & 1.8           & 0.0   \\
     2991\CFA generator          & 1.8           & 1.8           & 0.1   \\
     2992\CFA coroutine          & 32.5          & 32.9          & 0.8   \\
     2993\CFA thread                     & 93.8          & 93.6          & 2.2   \\
     2994\uC coroutine           & 50.3          & 50.3          & 0.2   \\
     2995\uC thread                      & 97.3          & 97.4          & 1.0   \\
     2996Python generator        & 40.9          & 41.3          & 1.5   \\
     2997Node.js generator       & 32.6          & 32.2          & 1.0   \\
     2998Node.js await           & 1852.2        & 1854.7        & 16.4  \\
     2999Goroutine thread        & 143.0         & 143.3         & 1.1   \\
     3000Rust thread                     & 332.0         & 331.4         & 2.4   \\
     3001Java thread                     & 405.0         & 415.0         & 17.6  \\
     3002Pthreads thread         & 334.3         & 335.2         & 3.9
    28273003\end{tabular}
    28283004\end{multicols}
    28293005
    2830 
    2831 \paragraph{Mutual-Exclusion}
    2832 
    2833 Uncontented mutual exclusion, which frequently occurs, is measured by entering/leaving a critical section.
    2834 For monitors, entering and leaving a monitor function is measured.
    2835 To put the results in context, the cost of entering a non-inline function and the cost of acquiring and releasing a @pthread_mutex@ lock is also measured.
    2836 Figure~\ref{f:mutex} shows the code for \CFA with all results in Table~\ref{tab:mutex}.
     3006\paragraph{Internal Scheduling}
     3007
     3008Internal scheduling is measured using a cycle of two threads signalling and waiting.
     3009Figure~\ref{f:schedint} shows the code for \CFA, with results in Table~\ref{t:schedint}.
    28373010Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
     3011Java 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.
    28383012
    28393013\begin{multicols}{2}
    28403014\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
    28413015\begin{cfa}
     3016volatile int go = 0;
     3017@condition c;@
    28423018@monitor@ M {} m1/*, m2, m3, m4*/;
    2843 void __attribute__((noinline))
    2844 do_call( M & @mutex m/*, m2, m3, m4*/@ ) {}
     3019void call( M & @mutex p1/*, p2, p3, p4*/@ ) {
     3020        @signal( c );@
     3021}
     3022void wait( M & @mutex p1/*, p2, p3, p4*/@ ) {
     3023        go = 1; // continue other thread
     3024        for ( N ) { @wait( c );@ } );
     3025}
     3026thread T {};
     3027void main( T & ) {
     3028        while ( go == 0 ) { yield(); } // waiter must start first
     3029        BENCH( for ( N ) { call( m1/*, m2, m3, m4*/ ); } )
     3030        sout | result;
     3031}
    28453032int main() {
    2846         BENCH(
    2847                 for( N ) do_call( m1/*, m2, m3, m4*/ );
    2848         )
    2849         sout | result`ns;
    2850 }
    2851 \end{cfa}
    2852 \captionof{figure}{\CFA acquire/release mutex benchmark}
    2853 \label{f:mutex}
     3033        T t;
     3034        wait( m1/*, m2, m3, m4*/ );
     3035}
     3036\end{cfa}
     3037\captionof{figure}{\CFA Internal-scheduling benchmark}
     3038\label{f:schedint}
    28543039
    28553040\columnbreak
    28563041
    28573042\vspace*{-16pt}
    2858 \captionof{table}{Mutex comparison (nanoseconds)}
    2859 \label{tab:mutex}
    2860 \begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
    2861 \multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
    2862 test and test-and-test lock             & 19.1  & 18.9  & 0.40  \\
    2863 \CFA @mutex@ function, 1 arg.   & 45.9  & 46.6  & 1.45  \\
    2864 \CFA @mutex@ function, 2 arg.   & 105.0 & 104.7 & 3.08  \\
    2865 \CFA @mutex@ function, 4 arg.   & 165.0 & 167.6 & 5.65  \\
    2866 \uC @monitor@ member rtn.               & 54.0  & 53.7  & 0.82  \\
    2867 Java synchronized method                & 31.0  & 31.1  & 0.50  \\
    2868 Pthreads Mutex Lock                             & 33.6  & 32.6  & 1.14
     3043\captionof{table}{Internal-scheduling comparison (nanoseconds)}
     3044\label{t:schedint}
     3045\bigskip
     3046
     3047\begin{tabular}{@{}r*{3}{D{.}{.}{5.2}}@{}}
     3048\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} & \multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
     3049\CFA @signal@, 1 monitor        & 364.4         & 364.2         & 4.4           \\
     3050\CFA @signal@, 2 monitor        & 484.4         & 483.9         & 8.8           \\
     3051\CFA @signal@, 4 monitor        & 709.1         & 707.7         & 15.0          \\
     3052\uC @signal@ monitor            & 328.3         & 327.4         & 2.4           \\
     3053Rust cond. variable                     & 7514.0        & 7437.4        & 397.2         \\
     3054Java @notify@ monitor           & 9623.0        & 9654.6        & 236.2         \\
     3055Pthreads cond. variable         & 5553.7        & 5576.1        & 345.6
    28693056\end{tabular}
    28703057\end{multicols}
     
    28743061
    28753062External scheduling is measured using a cycle of two threads calling and accepting the call using the @waitfor@ statement.
    2876 Figure~\ref{f:ext-sched} shows the code for \CFA, with results in Table~\ref{tab:ext-sched}.
     3063Figure~\ref{f:schedext} shows the code for \CFA with results in Table~\ref{t:schedext}.
    28773064Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
    28783065
     
    28813068\vspace*{-16pt}
    28823069\begin{cfa}
    2883 volatile int go = 0;
    2884 @monitor@ M {} m;
     3070@monitor@ M {} m1/*, m2, m3, m4*/;
     3071void call( M & @mutex p1/*, p2, p3, p4*/@ ) {}
     3072void wait( M & @mutex p1/*, p2, p3, p4*/@ ) {
     3073        for ( N ) { @waitfor( call : p1/*, p2, p3, p4*/ );@ }
     3074}
    28853075thread T {};
    2886 void __attribute__((noinline))
    2887 do_call( M & @mutex@ ) {}
    28883076void main( T & ) {
    2889         while ( go == 0 ) { yield(); }
    2890         while ( go == 1 ) { do_call( m ); }
    2891 }
    2892 int __attribute__((noinline))
    2893 do_wait( M & @mutex@ m ) {
    2894         go = 1; // continue other thread
    2895         BENCH( for ( N ) { @waitfor( do_call, m );@ } )
    2896         go = 0; // stop other thread
    2897         sout | result`ns;
     3077        BENCH( for ( N ) { call( m1/*, m2, m3, m4*/ ); } )
     3078        sout | result;
    28983079}
    28993080int main() {
    29003081        T t;
    2901         do_wait( m );
     3082        wait( m1/*, m2, m3, m4*/ );
    29023083}
    29033084\end{cfa}
    29043085\captionof{figure}{\CFA external-scheduling benchmark}
    2905 \label{f:ext-sched}
     3086\label{f:schedext}
    29063087
    29073088\columnbreak
     
    29093090\vspace*{-16pt}
    29103091\captionof{table}{External-scheduling comparison (nanoseconds)}
    2911 \label{tab:ext-sched}
     3092\label{t:schedext}
    29123093\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
    29133094\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
    2914 \CFA @waitfor@, 1 @monitor@     & 376.4 & 376.8 & 7.63  \\
    2915 \CFA @waitfor@, 2 @monitor@     & 491.4 & 492.0 & 13.31 \\
    2916 \CFA @waitfor@, 4 @monitor@     & 681.0 & 681.7 & 19.10 \\
    2917 \uC @_Accept@                           & 331.1 & 331.4 & 2.66
     3095\CFA @waitfor@, 1 monitor       & 367.1 & 365.3 & 5.0   \\
     3096\CFA @waitfor@, 2 monitor       & 463.0 & 464.6 & 7.1   \\
     3097\CFA @waitfor@, 4 monitor       & 689.6 & 696.2 & 21.5  \\
     3098\uC \lstinline[language=uC++]|_Accept| monitor  & 328.2 & 329.1 & 3.4   \\
     3099Go \lstinline[language=Golang]|select| channel  & 365.0 & 365.5 & 1.2
    29183100\end{tabular}
    29193101\end{multicols}
    29203102
    2921 
    2922 \paragraph{Internal Scheduling}
    2923 
    2924 Internal scheduling is measured using a cycle of two threads signalling and waiting.
    2925 Figure~\ref{f:int-sched} shows the code for \CFA, with results in Table~\ref{tab:int-sched}.
    2926 Note, the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
    2927 Java 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.
     3103\paragraph{Mutual-Exclusion}
     3104
     3105Uncontented mutual exclusion, which frequently occurs, is measured by entering/leaving a critical section.
     3106For monitors, entering and leaving a monitor function is measured, otherwise the language-appropriate mutex-lock is measured.
     3107For comparison, a spinning (versus blocking) test-and-test-set lock is presented.
     3108Figure~\ref{f:mutex} shows the code for \CFA with results in Table~\ref{t:mutex}.
     3109Note the incremental cost of bulk acquire for \CFA, which is largely a fixed cost for small numbers of mutex objects.
    29283110
    29293111\begin{multicols}{2}
    29303112\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
    29313113\begin{cfa}
    2932 volatile int go = 0;
    2933 @monitor@ M { @condition c;@ } m;
    2934 void __attribute__((noinline))
    2935 do_call( M & @mutex@ a1 ) { @signal( c );@ }
    2936 thread T {};
    2937 void main( T & this ) {
    2938         while ( go == 0 ) { yield(); }
    2939         while ( go == 1 ) { do_call( m ); }
    2940 }
    2941 int  __attribute__((noinline))
    2942 do_wait( M & mutex m ) with(m) {
    2943         go = 1; // continue other thread
    2944         BENCH( for ( N ) { @wait( c );@ } );
    2945         go = 0; // stop other thread
    2946         sout | result`ns;
    2947 }
     3114@monitor@ M {} m1/*, m2, m3, m4*/;
     3115call( M & @mutex p1/*, p2, p3, p4*/@ ) {}
    29483116int main() {
    2949         T t;
    2950         do_wait( m );
    2951 }
    2952 \end{cfa}
    2953 \captionof{figure}{\CFA Internal-scheduling benchmark}
    2954 \label{f:int-sched}
     3117        BENCH( for( N ) call( m1/*, m2, m3, m4*/ ); )
     3118        sout | result;
     3119}
     3120\end{cfa}
     3121\captionof{figure}{\CFA acquire/release mutex benchmark}
     3122\label{f:mutex}
    29553123
    29563124\columnbreak
    29573125
    29583126\vspace*{-16pt}
    2959 \captionof{table}{Internal-scheduling comparison (nanoseconds)}
    2960 \label{tab:int-sched}
    2961 \bigskip
    2962 
    2963 \begin{tabular}{@{}r*{3}{D{.}{.}{5.2}}@{}}
    2964 \multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} & \multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
    2965 \CFA @signal@, 1 @monitor@      & 372.6         & 374.3         & 14.17         \\
    2966 \CFA @signal@, 2 @monitor@      & 492.7         & 494.1         & 12.99         \\
    2967 \CFA @signal@, 4 @monitor@      & 749.4         & 750.4         & 24.74         \\
    2968 \uC @signal@                            & 320.5         & 321.0         & 3.36          \\
    2969 Java @notify@                           & 10160.5       & 10169.4       & 267.71        \\
    2970 Pthreads Cond. Variable         & 4949.6        & 5065.2        & 363
     3127\captionof{table}{Mutex comparison (nanoseconds)}
     3128\label{t:mutex}
     3129\begin{tabular}{@{}r*{3}{D{.}{.}{3.2}}@{}}
     3130\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} &\multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
     3131test-and-test-set lock                  & 19.1  & 18.9  & 0.4   \\
     3132\CFA @mutex@ function, 1 arg.   & 48.3  & 47.8  & 0.9   \\
     3133\CFA @mutex@ function, 2 arg.   & 86.7  & 87.6  & 1.9   \\
     3134\CFA @mutex@ function, 4 arg.   & 173.4 & 169.4 & 5.9   \\
     3135\uC @monitor@ member rtn.               & 54.8  & 54.8  & 0.1   \\
     3136Goroutine mutex lock                    & 34.0  & 34.0  & 0.0   \\
     3137Rust mutex lock                                 & 33.0  & 33.2  & 0.8   \\
     3138Java synchronized method                & 31.0  & 31.0  & 0.0   \\
     3139Pthreads mutex Lock                             & 31.0  & 31.1  & 0.4
    29713140\end{tabular}
    29723141\end{multicols}
    29733142
     3143\paragraph{Creation}
     3144
     3145Creation is measured by creating/deleting a specific kind of control-flow object.
     3146Figure~\ref{f:creation} shows the code for \CFA with results in Table~\ref{t:creation}.
     3147Note, 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.
     3148
     3149\begin{multicols}{2}
     3150\lstset{language=CFA,moredelim=**[is][\color{red}]{@}{@},deletedelim=**[is][]{`}{`}}
     3151\begin{cfa}
     3152@coroutine@ MyCoroutine {};
     3153void ?{}( MyCoroutine & this ) {
     3154#ifdef EAGER
     3155        resume( this );
     3156#endif
     3157}
     3158void main( MyCoroutine & ) {}
     3159int main() {
     3160        BENCH( for ( N ) { @MyCoroutine c;@ } )
     3161        sout | result;
     3162}
     3163\end{cfa}
     3164\captionof{figure}{\CFA creation benchmark}
     3165\label{f:creation}
     3166
     3167\columnbreak
     3168
     3169\vspace*{-16pt}
     3170\captionof{table}{Creation comparison (nanoseconds)}
     3171\label{t:creation}
     3172
     3173\begin{tabular}[t]{@{}r*{3}{D{.}{.}{5.2}}@{}}
     3174\multicolumn{1}{@{}c}{} & \multicolumn{1}{c}{Median} & \multicolumn{1}{c}{Average} & \multicolumn{1}{c@{}}{Std Dev} \\
     3175\CFA generator                  & 0.6           & 0.6           & 0.0           \\
     3176\CFA coroutine lazy             & 13.4          & 13.1          & 0.5           \\
     3177\CFA coroutine eager    & 144.7         & 143.9         & 1.5           \\
     3178\CFA thread                             & 466.4         & 468.0         & 11.3          \\
     3179\uC coroutine                   & 155.6         & 155.7         & 1.7           \\
     3180\uC thread                              & 523.4         & 523.9         & 7.7           \\
     3181Python generator                & 123.2         & 124.3         & 4.1           \\
     3182Node.js generator               & 32.3          & 32.2          & 0.3           \\
     3183Goroutine thread                & 751.0         & 750.5         & 3.1           \\
     3184Rust thread                             & 53801.0       & 53896.8       & 274.9         \\
     3185Java thread                             & 120274.0      & 120722.9      & 2356.7        \\
     3186Pthreads thread                 & 31465.5       & 31419.5       & 140.4
     3187\end{tabular}
     3188\end{multicols}
     3189
     3190
     3191\subsection{Discussion}
     3192
     3193Languages using 1:1 threading based on pthreads can at best meet or exceed (due to language overhead) the pthread results.
     3194Note, pthreads has a fast zero-contention mutex lock checked in user space.
     3195Languages with M:N threading have better performance than 1:1 because there is no operating-system interactions.
     3196Languages with stackful coroutines have higher cost than stackless coroutines because of stack allocation and context switching;
     3197however, stackful \uC and \CFA coroutines have approximately the same performance as stackless Python and Node.js generators.
     3198The \CFA stackless generator is approximately 25 times faster for suspend/resume and 200 times faster for creation than stackless Python and Node.js generators.
     3199
    29743200
    29753201\section{Conclusion}
     
    29773203Advanced control-flow will always be difficult, especially when there is temporal ordering and nondeterminism.
    29783204However, many systems exacerbate the difficulty through their presentation mechanisms.
    2979 This paper shows it is possible to present a hierarchy of control-flow features, generator, coroutine, thread, and monitor, providing an integrated set of high-level, efficient, and maintainable control-flow features.
    2980 Eliminated from \CFA are spurious wakeup and barging, which are nonintuitive and lead to errors, and having to work with a bewildering set of low-level locks and acquisition techniques.
    2981 \CFA high-level race-free monitors and tasks provide the core mechanisms for mutual exclusion and synchronization, without having to resort to magic qualifiers like @volatile@/@atomic@.
     3205This paper shows it is possible to understand high-level control-flow using three properties: statefulness, thread, mutual-exclusion/synchronization.
     3206Combining these properties creates a number of high-level, efficient, and maintainable control-flow types: generator, coroutine, thread, each of which can be a monitor.
     3207Eliminated 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.
     3208\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@.
    29823209Extending these mechanisms to handle high-level deadlock-free bulk acquire across both mutual exclusion and synchronization is a unique contribution.
    29833210The \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.
    29843211The M:N model is judged to be efficient and provide greater flexibility than a 1:1 threading model.
    29853212These 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.
    2986 Performance comparisons with other concurrent systems/languages show the \CFA approach is competitive across all low-level operations, which translates directly into good performance in well-written concurrent applications.
    2987 C 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.
     3213Performance 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.
     3214C 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.
    29883215
    29893216
     
    30053232\label{futur:nbio}
    30063233
    3007 Many modern workloads are not bound by computation but IO operations, a common case being web servers and XaaS~\cite{XaaS} (anything as a service).
     3234Many modern workloads are not bound by computation but IO operations, common cases being web servers and XaaS~\cite{XaaS} (anything as a service).
    30083235These types of workloads require significant engineering to amortizing costs of blocking IO-operations.
    30093236At 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.
     
    30333260\section{Acknowledgements}
    30343261
    3035 The authors would like to recognize the design assistance of Aaron Moss, Rob Schluntz, Andrew Beach and Michael Brooks on the features described in this paper.
    3036 Funding for this project has been provided by Huawei Ltd.\ (\url{http://www.huawei.com}). %, and Peter Buhr is partially funded by the Natural Sciences and Engineering Research Council of Canada.
     3262The 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.
     3263This 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.
    30373264
    30383265{%
    3039 \fontsize{9bp}{12bp}\selectfont%
     3266\fontsize{9bp}{11.5bp}\selectfont%
    30403267\bibliography{pl,local}
    30413268}%
  • doc/papers/concurrency/examples/Fib.py

    r26a09f92 r6565321  
    44        while True:
    55                fn = fn1 + fn2; fn2 = fn1; fn1 = fn; yield fn
    6 
    7 
    86
    97f1 = Fib()
     
    1412# Local Variables: #
    1513# tab-width: 4 #
    16 # compile-command: "python3.5 Fib.py" #
     14# compile-command: "python3.7 Fib.py" #
    1715# End: #
  • doc/papers/concurrency/examples/Fib2.c

    r26a09f92 r6565321  
    11#include <stdio.h>
    22
    3 void mary() {
    4         printf( "MARY\n" );
    5 }
    6 
    73#define FIB_INIT { 0 }
    8 typedef struct { int next; int fn1, fn2; } Fib;
     4typedef struct { int restart; int fn1, fn2; } Fib;
    95int fib( Fib * f ) {
    10         static void * states[] = { &&s1, &&s2, &&s3 };
    11         goto *states[f->next];
     6        static void * states[] = { &&s0, &&s1, &&s2 };
     7        goto *states[f->restart];
     8  s0:
     9        f->fn1 = 0;
     10        f->restart = 1;
     11        return f->fn1;
    1212  s1:
    13         mary();
    14         f->fn1 = 0;
    15         f->next = 1;
    16         return f->fn1;
    17   s2:
    18         mary();
    1913        f->fn2 = f->fn1;
    2014        f->fn1 = 1;
    21         f->next = 2;
     15        f->restart = 2;
    2216        return f->fn1;
    23   s3:;
    24         mary();
     17  s2:;
    2518        int fn = f->fn1 + f->fn2;
    2619        f->fn2 = f->fn1;
  • doc/papers/concurrency/examples/Fib2.py

    r26a09f92 r6565321  
    11def Fib():
    2     fn1, fn = 0, 1
     2    fn1, fn = 1, 0
    33    while True:
    4         yield fn1
     4        yield fn
    55        fn1, fn = fn, fn1 + fn
    66
     
    1212# Local Variables: #
    1313# tab-width: 4 #
    14 # compile-command: "python3.5 Fib2.py" #
     14# compile-command: "python3.7 Fib2.py" #
    1515# End: #
  • doc/papers/concurrency/examples/Fib3.c

    r26a09f92 r6565321  
    22
    33typedef struct {
    4         int fn1, fn;
    5         void * next;
     4        int restart, fn1, fn;
    65} Fib;
    7 #define FibCtor { 1, 0, NULL }
     6#define FibCtor { 0, 1, 0 }
    87
    98Fib * comain( Fib * f ) {
    10         if ( __builtin_expect(f->next != 0, 1) ) goto *f->next;
    11         f->next = &&s1;
     9        static void * states[] = {&&s0, &&s1};
     10        goto *states[f->restart];
     11  s0: f->restart = 1;
    1212        for ( ;; ) {
    1313                return f;
  • doc/papers/concurrency/examples/FibRefactor.py

    r26a09f92 r6565321  
    2222# Local Variables: #
    2323# tab-width: 4 #
    24 # compile-command: "python3.5 FibRefactor.py" #
     24# compile-command: "python3.7 FibRefactor.py" #
    2525# End: #
  • doc/papers/concurrency/examples/Format.c

    r26a09f92 r6565321  
    22
    33typedef struct {
    4         void * next;
     4        int restart, g, b;
    55        char ch;
    6         int g, b;
    76} Fmt;
    87
    98void comain( Fmt * f ) {
    10         if ( __builtin_expect(f->next != 0, 1) ) goto *f->next;
    11         f->next = &&s1;
     9        static void * states[] = {&&s0, &&s1};
     10        goto *states[f->restart];
     11  s0: f->restart = 1;
    1212        for ( ;; ) {
    1313                for ( f->g = 0; f->g < 5; f->g += 1 ) {                 // groups
    1414                        for ( f->b = 0; f->b < 4; f->b += 1 ) {         // blocks
    15                                 return;
    16                           s1:;  while ( f->ch == '\n' ) return;         // ignore
     15                                do {
     16                                        return;  s1: ;
     17                                } while ( f->ch == '\n' );                              // ignore
    1718                                printf( "%c", f->ch );                                  // print character
    1819                        }
     
    2425
    2526int main() {
    26         Fmt fmt = { NULL };
     27        Fmt fmt = { 0 };
    2728        comain( &fmt );                                                                         // prime
    2829        for ( ;; ) {
  • doc/papers/concurrency/examples/Format.cc

    r26a09f92 r6565321  
    66                        for ( g = 0; g < 5; g += 1 ) { // groups of 5 blocks
    77                                for ( b = 0; b < 4; b += 1 ) { // blocks of 4 characters
    8 //                                      for ( ;; ) { // for newline characters
     8                                        for ( ;; ) { // for newline characters
    99                                                suspend();
    10 //                                              if ( ch != '\n' ) break; // ignore newline
    11 //                                      }
     10                                                if ( ch != '\n' ) break; // ignore newline
     11                                        }
    1212//                                      cout << ch; // print character
    1313                                }
     
    3131// Local Variables: //
    3232// tab-width: 4 //
    33 // compile-command: "u++-work -O2 -nodebubg Format.cc" //
     33// compile-command: "u++-work -O2 -nodebug Format.cc" //
    3434// End: //
  • doc/papers/concurrency/examples/Format.cfa

    r26a09f92 r6565321  
    1111                for ( g = 0; g < 5; g += 1 ) {          // groups of 5 blocks
    1212                        for ( b = 0; b < 4; b += 1 ) {  // blocks of 4 characters
    13 //                              do {
     13                                do {
    1414                                        suspend();
    15 //                              } while ( ch == '\n' || ch == '\t' );
     15                                } while ( ch == '\n' || ch == '\t' );
    1616                                sout | ch;                                      // print character
    1717                        }
  • doc/papers/concurrency/examples/Format.data

    r26a09f92 r6565321  
    1 abcdefghijklmnopqrstuvwxyzxxxxxxxxxxxxxx
     1abcdefghijklmnop
     2qrstuvwxyzx
     3xxxxxxxxxxxxx
  • doc/papers/concurrency/examples/Format.py

    r26a09f92 r6565321  
    44                        for g in range( 5 ):    # groups of 5 blocks
    55                                for b in range( 4 ): # blocks of 4 characters
    6                                         print( (yield), end='' ) # receive from send
     6                                        while True:
     7                                                ch = (yield) # receive from send
     8                                                if '\n' not in ch:
     9                                                        break
     10                                        print( ch, end='' ) # receive from send
    711                                print( '  ', end='' ) # block separator
    812                        print()                                 # group separator
     
    1115                        print()
    1216
     17input = "abcdefghijklmnop\nqrstuvwx\nyzxxxxxxxxxxxxxx\n"
     18
    1319fmt = Format()
    1420next( fmt )                                                     # prime generator
    15 for i in range( 41 ):
    16         fmt.send( 'a' );                                # send to yield
     21for i in input:
     22        fmt.send( i );                          # send to yield
    1723
    1824# Local Variables: #
    1925# tab-width: 4 #
    20 # compile-command: "python3.5 Format.py" #
     26# compile-command: "python3.7 Format.py" #
    2127# End: #
  • doc/papers/concurrency/examples/Format1.c

    r26a09f92 r6565321  
    22
    33typedef struct {
    4         void * next;
     4        int restart, g, b;
    55        char ch;
    6         int g, b;
    76} Fmt;
    87
    98void format( Fmt * f ) {
    10         if ( __builtin_expect(f->next != 0, 1) ) goto *f->next;
    11         f->next = &&s1;
     9        static void * states[] = {&&s0, &&s1};
     10        goto *states[f->restart];
     11  s0: f->restart = 1;
    1212        for ( ;; ) {
    1313                for ( f->g = 0; f->g < 5; f->g += 1 ) {                 // groups
    1414                        for ( f->b = 0; f->b < 4; f->b += 1 ) {         // blocks
    1515                                return;
    16                           s1: ;
    17                                 if ( f->ch == '\0' ) goto fini;                 // EOF ?
     16                          s1: if ( f->ch == '\0' ) goto fini;           // EOF ?
    1817                                while ( f->ch == '\n' ) return;                 // ignore
    19                                 printf( "%c", f->ch );                                  // print character
     18//                              printf( "%c", f->ch );                                  // print character
    2019                        }
    21                         printf( " " );                                                          // block separator
     20//                      printf( " " );                                                          // block separator
    2221                }
    23                 printf( "\n" );                                                                 // group separator
     22//              printf( "\n" );                                                                 // group separator
    2423        }
    25   fini:
    26         if ( f->g != 0 || f->b != 0 ) printf( "\n" );
     24  fini:;
     25//      if ( f->g != 0 || f->b != 0 ) printf( "\n" );
    2726}
    2827
    2928int main() {
    30         Fmt fmt = { NULL };
     29        Fmt fmt = { 0 };
    3130        format( &fmt );                                                                         // prime
    32         for ( ;; ) {
    33                 scanf( "%c", &fmt.ch );                                                 // direct read into communication variable
    34           if ( feof( stdin ) ) break;
     31        fmt.ch = 'a';
     32        for ( long int i = 0; i < 1000000000; i += 1 ) {
     33//              scanf( "%c", &fmt.ch );                                                 // direct read into communication variable
     34//        if ( feof( stdin ) ) break;
    3535                format( &fmt );
    3636        }
    37         fmt.ch = '\0';
     37        fmt.ch = '\0';                                                                          // sentential (EOF)
    3838        format( &fmt );
    3939}
  • doc/papers/concurrency/examples/PingPong.c

    r26a09f92 r6565321  
    22
    33typedef struct PingPong {
     4        int restart;                                                                            // style 1
     5        int N, i;
    46        const char * name;
    5         int N, i;
    67        struct PingPong * partner;
    7         void * next;
     8        void * next;                                                                            // style 2
    89} PingPong;
    9 #define PPCtor( name, N ) { name, N, 0, NULL, NULL }
     10#define PPCtor( name, N ) { 0, N, 0, name, NULL, NULL }
     11
    1012void comain( PingPong * pp ) __attribute__(( noinline ));
    1113void comain( PingPong * pp ) {
     14#if 0
    1215        if ( __builtin_expect(pp->next != 0, 1) ) goto *pp->next;
    13 #if 0
    14         pp->next = &&here;
    15                 asm( "mov  %0,%%rdi" : "=m" (pp) );
    16                 asm( "mov  %rdi,%rax" );
    17 #ifndef OPT
    18 #ifdef PRINT
    19                 asm( "add  $16, %rsp" );
    20 #endif // PRINT
    21                 asm( "popq %rbp" );
    22 #endif // ! OPT
    23 
    24 #ifdef OPT
    25 #ifdef PRINT
    26                 asm( "popq %rbx" );
    27 #endif // PRINT
    28 #endif // OPT
    29                 asm( "jmp  comain" );
    30   here: ;
    31 #endif // 0
    32 
    3316        pp->next = &&cycle;
    3417        for ( ; pp->i < pp->N; pp->i += 1 ) {
     
    5336          cycle: ;
    5437        } // for
     38#endif // 0
     39
     40#if 1
     41        static void * states[] = {&&s0, &&s1};
     42        goto *states[pp->restart];
     43  s0: pp->restart = 1;
     44        for ( ; pp->i < pp->N; pp->i += 1 ) {
     45#ifdef PRINT
     46                printf( "%s %d\n", pp->name, pp->i );
     47#endif // PRINT
     48                asm( "mov  %0,%%rdi" : "=m" (pp->partner) );
     49                asm( "mov  %rdi,%rax" );
     50#ifndef OPT
     51#ifdef PRINT
     52                asm( "add  $16, %rsp" );
     53#endif // PRINT
     54                asm( "popq %rbp" );
     55#endif // ! OPT
     56
     57#ifdef OPT
     58#ifdef PRINT
     59                asm( "popq %rbx" );
     60#endif // PRINT
     61#endif // OPT
     62                asm( "jmp  comain" );
     63          s1: ;
     64        } // for
     65#endif // 0
    5566}
    5667
     
    7081// Local Variables: //
    7182// tab-width: 4 //
    72 // compile-command: "gcc-8 -g -DPRINT PingPong.c" //
     83// compile-command: "gcc-9 -g -DPRINT PingPong.c" //
    7384// End: //
  • doc/papers/concurrency/examples/Pingpong.py

    r26a09f92 r6565321  
    11def PingPong( name, N ):
    2         partner = (yield)           # get partner
    3         yield                       # resume scheduler
     2        partner = yield                         # get partner
     3        yield                                           # resume scheduler
    44        for i in range( N ):
    55                print( name )
    6                 yield partner           # execute next
     6                yield partner                   # execute next
    77        print( "end", name )
    88
    99def Scheduler():
    10         n = (yield)                 # starting coroutine
    11         while True:
    12                 n = next( n )           # schedule coroutine
     10        n = yield                                       # starting coroutine
     11        try:
     12                while True:
     13                        n = next( n )           # schedule coroutine
     14        except StopIteration:
     15                pass
    1316
    1417pi = PingPong( "ping", 5 )
    1518po = PingPong( "pong", 5 )
    16 next( pi )                      # prime
    17 pi.send( po )                   # send partner
    18 next( po )                      # prime
    19 po.send( pi )                   # send partner
     19next( pi )                                              # prime
     20pi.send( po )                                   # send partner
     21next( po )                                              # prime
     22po.send( pi )                                   # send partner
    2023
    2124s = Scheduler();
    22 next( s )                       # prime
     25next( s )                                               # prime
    2326try:
    2427        s.send( pi )                            # start cycle
    25 except StopIteration:
    26         print( "scheduler stop" )
     28except StopIteration:                   # scheduler stopped
     29        pass
    2730print( "stop" )
    2831
    2932# Local Variables: #
    3033# tab-width: 4 #
    31 # compile-command: "python3.5 Pingpong.py" #
     34# compile-command: "python3.7 Pingpong.py" #
    3235# End: #
  • doc/papers/concurrency/examples/ProdCons.py

    r26a09f92 r6565321  
    11def Prod( N ):
    2         cons = (yield)              # get cons
    3         yield                       # resume scheduler
     2        cons = yield                            # get cons
     3        yield                                           # resume scheduler
    44        for i in range( N ):
    55                print( "prod" )
    6                 yield cons              # execute next
     6                yield cons                              # execute next
    77        print( "end", "prod" )
    88
    99def Cons( N ):
    10         prod = (yield)              # get prod
    11         yield                       # resume scheduler
     10        prod = yield                            # get prod
     11        yield                                           # resume scheduler
    1212        for i in range( N ):
    1313                print( "cons" )
    14                 yield prod              # execute next
     14                yield prod                              # execute next
    1515        print( "end", "cons" )
    1616
    1717def Scheduler():
    18         n = (yield)                 # starting coroutine
    19         while True:
    20                 n = next( n )           # schedule coroutine
     18        n = yield                                       # starting coroutine
     19        try:
     20                while True:
     21                        n = next( n )           # schedule coroutine
     22        except StopIteration:
     23                pass
    2124
    2225prod = Prod( 5 )
    2326cons = Cons( 5 )
    24 next( prod )                    # prime
    25 prod.send( cons )               # send cons
    26 next( cons )                    # prime
    27 cons.send( prod )               # send prod
     27next( prod )                                    # prime
     28prod.send( cons )                               # send cons
     29next( cons )                                    # prime
     30cons.send( prod )                               # send prod
    2831
    2932s = Scheduler();
    30 next( s )                       # prime
     33next( s )                                               # prime
    3134try:
    3235        s.send( prod )                          # start cycle
    33 except StopIteration:
    34         print( "scheduler stop" )
     36except StopIteration:                   # scheduler stopped
     37        pass
    3538print( "stop" )
    3639
    3740# Local Variables: #
    3841# tab-width: 4 #
    39 # compile-command: "python3.5 ProdCons.py" #
     42# compile-command: "python3.7 ProdCons.py" #
    4043# End: #
  • doc/papers/concurrency/examples/Refactor.py

    r26a09f92 r6565321  
    2626# Local Variables: #
    2727# tab-width: 4 #
    28 # compile-command: "python3.5 Refactor.py" #
     28# compile-command: "python3.7 Refactor.py" #
    2929# End: #
  • doc/papers/concurrency/figures/FullCoroutinePhases.fig

    r26a09f92 r6565321  
    88-2
    991200 2
    10 5 1 0 1 0 7 100 0 -1 0.000 0 0 1 0 4575.000 2437.500 4275 1875 4575 1800 4875 1875
     105 1 0 1 0 7 100 0 -1 0.000 0 0 1 0 5175.000 2437.500 4875 1875 5175 1800 5475 1875
    1111        1 1 1.00 45.00 90.00
    12 5 1 0 1 0 7 100 0 -1 0.000 0 0 1 0 4575.000 1537.500 4875 2100 4575 2175 4275 2100
     125 1 0 1 0 7 100 0 -1 0.000 0 0 1 0 5175.000 1537.500 5475 2100 5175 2175 4875 2100
    1313        1 1 1.00 45.00 90.00
    14 5 1 0 1 0 7 50 -1 -1 0.000 0 1 1 0 4207.500 1642.500 4125 1425 3975 1650 4200 1875
     145 1 0 1 0 7 50 -1 -1 0.000 0 1 1 0 4807.500 1642.500 4725 1425 4575 1650 4800 1875
    1515        1 1 1.00 45.00 90.00
     166 1575 1575 2700 2025
    16172 1 0 1 0 7 100 0 -1 0.000 0 0 -1 1 0 2
    1718        1 1 1.00 45.00 90.00
     
    2021        1 1 1.00 45.00 90.00
    2122         2175 1575 2400 1800
     234 1 0 100 0 4 10 0.0000 2 165 300 1725 1950 ping\001
     244 1 0 100 0 4 10 0.0000 2 135 360 2475 1950 pong\001
     25-6
     266 3075 1575 4200 2025
     276 3075 1575 4200 2025
    22282 1 0 1 0 7 100 0 -1 0.000 0 0 -1 1 0 2
    2329        1 1 1.00 45.00 90.00
    24          3300 1575 3300 1800
     30         3525 1575 3300 1800
    25312 1 0 1 0 7 100 0 -1 0.000 0 0 -1 1 0 2
    2632        1 1 1.00 45.00 90.00
    27          3300 2025 3300 2250
    28 4 1 0 100 0 0 10 0.0000 2 105 555 2100 1200 creation\001
    29 4 1 0 100 0 4 10 0.0000 2 165 300 1725 1950 ping\001
    30 4 1 0 100 0 4 10 0.0000 2 135 360 2475 1950 pong\001
    31 4 1 0 100 0 4 10 0.0000 2 165 300 3300 1950 ping\001
    32 4 1 0 100 0 4 10 0.0000 2 135 360 3300 2400 pong\001
    33 4 1 0 100 0 0 10 0.0000 2 105 675 4575 1200 execution\001
    34 4 1 0 100 0 4 10 0.0000 2 165 300 4275 2025 ping\001
    35 4 1 0 100 0 4 10 0.0000 2 135 360 4875 2025 pong\001
    36 4 1 0 100 0 0 10 0.0000 2 90 420 3300 1200 starter\001
     33         3675 1575 3900 1800
     344 1 0 100 0 4 10 0.0000 2 165 300 3225 1950 ping\001
     354 1 0 100 0 4 10 0.0000 2 135 360 3975 1950 pong\001
     36-6
     37-6
    37384 1 0 100 0 4 10 0.0000 2 165 705 2100 1500 pgm main\001
    38 4 1 0 100 0 4 10 0.0000 2 165 705 3300 1500 pgm main\001
    39 4 1 0 100 0 4 10 0.0000 2 165 705 4500 1500 pgm main\001
     394 1 0 100 0 4 10 0.0000 2 165 705 3600 1500 pgm main\001
     404 1 0 100 0 4 10 0.0000 2 165 300 4875 2025 ping\001
     414 1 0 100 0 4 10 0.0000 2 135 360 5475 2025 pong\001
     424 1 0 100 0 4 10 0.0000 2 165 705 5100 1500 pgm main\001
     434 1 0 100 0 2 10 0.0000 2 105 540 2100 1275 creator\001
     444 1 0 100 0 2 10 0.0000 2 105 495 3600 1275 starter\001
     454 1 0 100 0 2 10 0.0000 2 105 690 5175 1275 execution\001
  • doc/papers/concurrency/figures/RunTimeStructure.fig

    r26a09f92 r6565321  
    36361 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 4500 3600 15 15 4500 3600 4515 3615
    3737-6
    38 6 2175 4650 7050 4950
    39 1 3 0 1 0 0 0 0 0 0.000 1 0.0000 2250 4830 30 30 2250 4830 2280 4860
    40 1 1 0 1 -1 -1 0 0 -1 0.000 1 0.0000 4200 4800 150 75 4200 4800 4350 4875
    41 1 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 3275 4800 100 100 3275 4800 3375 4800
     386 3225 4125 4650 4425
     396 4350 4200 4650 4350
     401 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 4425 4275 15 15 4425 4275 4440 4290
     411 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 4500 4275 15 15 4500 4275 4515 4290
     421 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 4575 4275 15 15 4575 4275 4590 4290
     43-6
     441 1 0 1 -1 -1 0 0 -1 0.000 1 0.0000 3450 4275 225 150 3450 4275 3675 4425
     451 1 0 1 -1 -1 0 0 -1 0.000 1 0.0000 4050 4275 225 150 4050 4275 4275 4425
     46-6
     476 6675 4125 7500 4425
     486 7200 4200 7500 4350
     491 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 7275 4275 15 15 7275 4275 7290 4290
     501 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 7350 4275 15 15 7350 4275 7365 4290
     511 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 7425 4275 15 15 7425 4275 7440 4290
     52-6
     531 1 0 1 -1 -1 0 0 -1 0.000 1 0.0000 6900 4275 225 150 6900 4275 7125 4425
     54-6
     556 6675 3525 8025 3975
     562 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 1 0 2
     57        1 1 1.00 45.00 90.00
     58         6675 3750 6975 3750
     592 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 1 0 2
     60        1 1 1.00 45.00 90.00
     61         7125 3750 7350 3750
    42622 2 0 1 -1 -1 0 0 -1 0.000 0 0 0 0 0 5
    43          5400 4950 5400 4725 5175 4725 5175 4950 5400 4950
    44 2 2 1 1 -1 -1 0 0 -1 3.000 0 0 0 0 0 5
    45          6525 4950 6300 4950 6300 4725 6525 4725 6525 4950
    46 4 0 -1 0 0 0 10 0.0000 2 105 450 6600 4875 cluster\001
    47 4 0 -1 0 0 0 10 0.0000 2 105 660 5475 4875 processor\001
    48 4 0 -1 0 0 0 10 0.0000 2 105 555 4425 4875 monitor\001
    49 4 0 -1 0 0 0 10 0.0000 2 120 270 3450 4875 task\001
    50 4 0 -1 0 0 0 10 0.0000 2 105 660 2325 4875 coroutine\001
    51 -6
    52 6 3450 1275 3750 1425
    53 1 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 3525 1350 15 15 3525 1350 3540 1365
    54 1 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 3600 1350 15 15 3600 1350 3615 1365
    55 1 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 3675 1350 15 15 3675 1350 3690 1365
    56 -6
    57 6 5550 1275 5850 1425
    58 1 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 5625 1350 15 15 5625 1350 5640 1365
    59 1 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 5700 1350 15 15 5700 1350 5715 1365
    60 1 3 0 1 -1 -1 0 0 20 0.000 1 0.0000 5775 1350 15 15 5775 1350 5790 1365
     63         7800 3975 7800 3525 7350 3525 7350 3975 7800 3975
     642 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 1 0 2
     65        1 1 1.00 45.00 90.00
     66         7800 3750 8025 3750
    6167-6
    62681 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 5550 2625 150 150 5550 2625 5700 2625
     
    67731 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 4425 2850 150 150 4425 2850 4575 2850
    68741 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 4650 2475 150 150 4650 2475 4800 2475
    69 1 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 3525 3600 150 150 3525 3600 3675 3600
    70751 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 3975 3600 150 150 3975 3600 4125 3600
    71761 3 0 1 0 0 0 0 0 0.000 1 0.0000 3525 3600 30 30 3525 3600 3555 3630
     
    74791 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 3975 2850 150 150 3975 2850 4125 2850
    75801 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 7200 2775 150 150 7200 2775 7350 2775
    76 1 1 0 1 -1 -1 0 0 -1 0.000 1 0.0000 4650 1350 225 150 4650 1350 4875 1500
    77 1 1 0 1 -1 -1 0 0 -1 0.000 1 0.0000 5250 1350 225 150 5250 1350 5475 1500
    78 1 1 0 1 -1 -1 0 0 -1 0.000 1 0.0000 4050 1350 225 150 4050 1350 4275 1500
     811 3 0 1 0 0 0 0 0 0.000 1 0.0000 2250 4830 30 30 2250 4830 2280 4860
     821 3 0 1 0 0 0 0 0 0.000 1 0.0000 7200 2775 30 30 7200 2775 7230 2805
     831 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 3525 3600 150 150 3525 3600 3675 3600
     841 3 0 1 -1 -1 0 0 -1 0.000 1 0.0000 3875 4800 100 100 3875 4800 3975 4800
     851 1 0 1 -1 -1 0 0 -1 0.000 1 0.0000 4650 4800 150 75 4650 4800 4800 4875
    79862 2 0 1 -1 -1 0 0 -1 0.000 0 0 0 0 0 5
    8087         2400 4200 2400 3750 1950 3750 1950 4200 2400 4200
     
    1401472 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 1 0 2
    141148        1 1 1.00 45.00 90.00
    142          6675 3975 6975 3975
    143 2 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 1 0 2
    144         1 1 1.00 45.00 90.00
    145149         7050 2775 6825 2775
    1461502 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 0 0 2
    147          6825 2775 6825 3975
    148 2 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 1 0 2
    149         1 1 1.00 45.00 90.00
    150          7125 3975 7350 3975
    151 2 2 0 1 -1 -1 0 0 -1 0.000 0 0 0 0 0 5
    152          7800 4200 7800 3750 7350 3750 7350 4200 7800 4200
    153 2 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 1 0 2
    154         1 1 1.00 45.00 90.00
    155          7800 3975 8025 3975
     151         6825 2775 6825 3750
    1561522 1 0 1 -1 -1 0 0 -1 0.000 0 0 -1 1 0 4
    157153        1 1 1.00 45.00 90.00
    158          7875 3975 7875 2325 7200 2325 7200 2550
     154         7875 3750 7875 2325 7200 2325 7200 2550
     1552 2 0 1 -1 -1 0 0 -1 0.000 0 0 0 0 0 5
     156         5850 4950 5850 4725 5625 4725 5625 4950 5850 4950
     1572 2 1 1 -1 -1 0 0 -1 3.000 0 0 0 0 0 5
     158         6975 4950 6750 4950 6750 4725 6975 4725 6975 4950
    1591594 1 -1 0 0 0 10 0.0000 2 105 720 5550 4425 Processors\001
    1601604 1 -1 0 0 0 10 0.0000 2 120 1005 4200 3225 Blocked Tasks\001
     
    1651654 1 -1 0 0 0 10 0.0000 2 105 990 2175 3525 Discrete-event\001
    1661664 1 -1 0 0 0 10 0.0000 2 135 795 2175 4350 preemption\001
     1674 0 -1 0 0 0 10 0.0000 2 150 1290 2325 4875 genrator/coroutine\001
     1684 0 -1 0 0 0 10 0.0000 2 120 270 4050 4875 task\001
     1694 0 -1 0 0 0 10 0.0000 2 105 450 7050 4875 cluster\001
     1704 0 -1 0 0 0 10 0.0000 2 105 660 5925 4875 processor\001
     1714 0 -1 0 0 0 10 0.0000 2 105 555 4875 4875 monitor\001
  • doc/papers/concurrency/mail2

    r26a09f92 r6565321  
    2222Software: Practice and Experience Editorial Office
    2323
     24
     25
     26Date: Tue, 12 Nov 2019 22:25:17 +0000
     27From: Richard Jones <onbehalfof@manuscriptcentral.com>
     28Reply-To: R.E.Jones@kent.ac.uk
     29To: tdelisle@uwaterloo.ca, pabuhr@uwaterloo.ca
     30Subject: Software: Practice and Experience - Decision on Manuscript ID
     31 SPE-19-0219
     32
     3312-Nov-2019
     34
     35Dear Dr Buhr,
     36
     37Many thanks for submitting SPE-19-0219 entitled "Advanced Control-flow and Concurrency in Cforall" to Software: Practice and Experience. The paper has now been reviewed and the comments of the referees are included at the bottom of this letter.
     38
     39The decision on this paper is that it requires substantial further work is required. The referees have a number of substantial concerns. All the reviewers found the submission very hard to read; two of the reviewers state that it needs very substantial restructuring. These concerns must be addressed before your submission can be considered further.
     40
     41A revised version of your manuscript that takes into account the comments of the referees will be reconsidered for publication.
     42
     43Please note that submitting a revision of your manuscript does not guarantee eventual acceptance, and that your revision will be subject to re-review by the referees before a decision is rendered.
     44
     45You have 90 days from the date of this email to submit your revision. If you are unable to complete the revision within this time, please contact me to request an extension.
     46
     47You can upload your revised manuscript and submit it through your Author Center. Log into https://mc.manuscriptcentral.com/spe  and enter your Author Center, where you will find your manuscript title listed under "Manuscripts with Decisions".
     48
     49When submitting your revised manuscript, you will be able to respond to the comments made by the referee(s) in the space provided.  You can use this space to document any changes you make to the original manuscript.
     50
     51If you feel that your paper could benefit from English language polishing, you may wish to consider having your paper professionally edited for English language by a service such as Wiley's at http://wileyeditingservices.com. Please note that while this service will greatly improve the readability of your paper, it does not guarantee acceptance of your paper by the journal.
     52 
     53Once again, thank you for submitting your manuscript to Software: Practice and Experience and I look forward to receiving your revision.
     54
     55
     56Sincerely,
     57
     58Prof. Richard Jones
     59Software: Practice and Experience
     60R.E.Jones@kent.ac.uk
     61
     62
     63Referee(s)' Comments to Author:
     64
     65Reviewing: 1
     66
     67Comments to the Author
     68This article presents the design and rationale behind the various
     69threading and synchronization mechanisms of C-forall, a new low-level
     70programming language.  This paper is very similar to a companion paper
     71which I have also received: as the papers are similar, so will these
     72reviews be --- in particular any general comments from the other
     73review apply to this paper also.
     74
     75As far as I can tell, the article contains three main ideas: an
     76asynchronous execution / threading model; a model for monitors to
     77provide mutual exclusion; and an implementation.  The first two ideas
     78are drawn together in Table 1: unfortunately this is on page 25 of 30
     79pages of text. Implementation choices and descriptions are scattered
     80throughout the paper - and the sectioning of the paper seems almost
     81arbitrary.
     82
     83The article is about its contributions.  Simply adding feature X to
     84language Y isn't by itself a contribution, (when feature X isn't
     85already a contribution).  The contribution can be in the design: the
     86motivation, the space of potential design options, the particular
     87design chosen and the rationale for that choice, or the resulting
     88performance.  For example: why support two kinds of generators as well
     89as user-level threads?  Why support both low and high level
     90synchronization constructs?  Similarly I would have found the article
     91easier to follow if it was written top down, presenting the design
     92principles, present the space of language features, justify chosen
     93language features (and rationale) and those excluded, and then present
     94implementation, and performance.
     95
     96Then the writing of the article is often hard to follow, to say the
     97least. Two examples: section 3 "stateful functions" - I've some idea
     98what that is (a function with Algol's "own" or C's "static" variables?
     99but in fact the paper has a rather more specific idea than that. The
     100top of page 3 throws a whole lot of defintions at the reader
     101"generator" "coroutine" "stackful" "stackless" "symmetric"
     102"asymmetric" without every stopping to define each one --- but then in
     103footnote "C" takes the time to explain what C's "main" function is?  I
     104cannot imagine a reader of this paper who doesn't know what "main" is
     105in C; especially if they understand the other concepts already
     106presented in the paper.  The start of section 3 then does the same
     107thing: putting up a whole lot of definitions, making distinctions and
     108comparisons, even talking about some runtime details, but the critical
     109definition of a monitor doesn't appear until three pages later, at the
     110start of section 5 on p15, lines 29-34 are a good, clear, description
     111of what a monitor actually is.  That needs to come first, rather than
     112being buried again after two sections of comparisons, discussions,
     113implementations, and options that are ungrounded because they haven't
     114told the reader what they are actually talking about.  First tell the
     115reader what something is, then how they might use it (as programmers:
     116what are the rules and restrictions) and only then start comparison
     117with other things, other approaches, other languages, or
     118implementations.
     119
     120The description of the implementation is similarly lost in the trees
     121without ever really seeing the wood. Figure 19 is crucial here, but
     122it's pretty much at the end of the paper, and comments about
     123implementations are threaded throughout the paper without the context
     124(fig 19) to understand what's going on.   The protocol for performance
     125testing may just about suffice for C (although is N constantly ten
     126million, or does it vary for each benchmark) but such evaluation isn't
     127appropriate for garbage-collected or JITTed languages like Java or Go.
     128
     129other comments working through the paper - these are mostly low level
     130and are certainly not comprehensive.
     131
     132p1 only a subset of C-forall extensions?
     133
     134p1 "has features often associated with object-oriented programming
     135languages, such as constructors, destructors, virtuals and simple
     136inheritance."   There's no need to quibble about this. Once a language
     137has inheritance, it's hard to claim it's not object-oriented.
     138
     139
     140p2 barging? signals-as-hints?
     141
     142p3 start your discussion of generations with a simple example of a
     143C-forall generator.  Fig 1(b) might do: but put it inline instead of
     144the python example - and explain the key rules and restrictions on the
     145construct.  Then don't even start to compare with coroutines until
     146you've presented, described and explained your coroutines...
     147p3 I'd probably leave out the various "C" versions unless there are
     148key points to make you can't make in C-forall. All the alternatives
     149are just confusing.
     150
     151
     152p4 but what's that "with" in Fig 1(B)
     153
     154p5 start with the high level features of C-forall generators...
     155
     156p5 why is the paper explaining networking protocols?
     157
     158p7 lines 1-9 (transforming generator to coroutine - why would I do any
     159of this? Why would I want one instead of the other (do not use "stack"
     160in your answer!)
     161
     162p10 last para "A coroutine must retain its last resumer to suspend
     163back because the resumer is on a different stack. These reverse
     164pointers allow suspend to cycle backwards, "  I've no idea what is
     165going on here?  why should I care?  Shouldn't I just be using threads
     166instead?  why not?
     167
     168p16 for the same reasons - what reasons?
     169
     170p17 if the multiple-monitor entry procedure really is novel, write a
     171paper about that, and only about that.
     172
     173p23 "Loose Object Definitions" - no idea what that means.  in that
     174section: you can't leave out JS-style dynamic properties.  Even in
     175OOLs that (one way or another) allow separate definitions of methods
     176(like Objective-C, Swift, Ruby, C#) at any time a runtime class has a
     177fixed definition.  Quite why the detail about bit mask implementation
     178is here anyway, I've no idea.
     179
     180p25 this cluster isn't a CLU cluster then?
     181
     182* conclusion should conclude the paper, not the related.
     183
     184
     185Reviewing: 2
     186
     187Comments to the Author
     188This paper describes the concurrency features of an extension of C (whose name I will write as "C\/" here, for convenience), including much design-level discussion of the coroutine- and monitor-based features and some microbenchmarks exploring the current implementation's performance. The key message of the latter is that the system's concurrency abstractions are much lighter-weight than the threading found in mainstream C or Java implementations.
     189
     190There is much description of the system and its details, but nothing about (non-artificial) uses of it. Although the microbenchmark data is encouraging, arguably not enough practical experience with the system has been reported here to say much about either its usability advantages or its performance.
     191
     192As such, the main contribution of the paper seem to be to document the existence of the described system and to provide a detailed design rationale and (partial) tutorial. I believe that could be of interest to some readers, so an acceptable manuscript is lurking in here somewhere.
     193
     194Unfortunately, at present the writing style is somewhere between unclear and infuriating. It omits to define terms; it uses needlessly many terms for what are apparently (but not clearly) the same things; it interrupts itself rather than deliver the natural consequent of whatever it has just said; and so on. Section 5 is particularly bad in these regards -- see my detailed comments below. Fairly major additional efforts will be needed to turn the present text into a digestible design-and-tutorial document. I suspect that a shorter paper could do this job better than the present manuscript, which is overwrought in parts.
     195
     196p2: lines 4--9 are a little sloppy. It is not the languages but their popular implementations which "adopt" the 1:1 kernel threading model.
     197
     198line 10: "medium work" -- "medium-sized work"?
     199
     200line 18: "is all sequential to the compiler" -- not true in modern compilers, and in 2004 H-J Boehm wrote a tech report describing exactly why ("Threads cannot be implemented as a library", HP Labs).
     201
     202line 20: "knows the optimization boundaries" -- I found this vague. What's an example?
     203
     204line 31: this paragraph has made a lot of claims. Perhaps forward-reference to the parts of the paper that discuss each one.
     205
     206line 33: "so the reader can judge if" -- this reads rather passive-aggressively. Perhaps better: "... to support our argument that..."
     207
     208line 41: "a dynamic partitioning mechanism" -- I couldn't tell what this meant
     209
     210p3. Presenting concept of a "stateful function" as a new language feature seems odd. In C, functions often have local state thanks to static local variables (or globals, indeed). Of course, that has several limitations. Can you perhaps present your contributions by enumerating these limitations? See also my suggestion below about a possible framing centred on a strawman.
     211
     212line 2: "an old idea that is new again" -- this is too oblique
     213
     214lines 2--15: I found this to be a word/concept soup. Stacks, closures, generators, stackless stackful, coroutine, symmetric, asymmetric, resume/suspend versus resume/resume... there needs to be a more gradual and structured way to introduce all this, and ideally one that minimises redundancy. Maybe present it as a series of "definitions" each with its own heading, e.g. "A closure is stackless if its local state has statically known fixed size"; "A generator simply means a stackless closure." And so on. Perhaps also strongly introduce the word "activate" as a direct contrast with resume and suspend. These are just a flavour of the sort of changes that might make this paragraph into something readable.
     215
     216Continuing the thought: I found it confusing that by these definitinos, a stackful closure is not a stack, even though logically the stack *is* a kind of closure (it is a representation of the current thread's continuation).
     217
     218lines 24--27: without explaining what the boost functor types mean, I don't think the point here comes across.
     219
     220line 34: "semantically coupled" -- I wasn't surew hat this meant
     221
     222p4: the point of Figure 1 (C) was not immediately clear. It seem to be showing how one might "compile down" Figure 1 (B). Or is that Figure 1 (A)?
     223
     224It's right that the incidental language features of the system are not front-and-centre, but I'd appreciate some brief glossing of non-C languages features as they appear. Examples are the square bracket notation, the pipe notation and the constructor syntax. These explanations could go in the caption of the figure which first uses them, perhaps. Overall I found the figure captions to be terse, and a missed opportunity to explain clearly what was going on.
     225
     226p5 line 23: "This restriction is removed..." -- give us some up-front summary of your contributions and the elements of the language design that will be talked about, so that this isn't an aside. This will reduce the "twisty passages" feeling that characterises much of the paper.
     227
     228line 40: "a killer asymmetric generator" -- this is stylistically odd, and the sentence about failures doesn't convincigly argue that C\/ will help with them. Have you any experience writing device drivers using C\/? Or any argument that the kinds of failures can be traced to the "stack-ripping" style that one is forced to use without coroutines? Also, a typo on line 41: "device drives". And saying "Windows/Linux" is sloppy... what does the cited paper actually say?
     229
     230p6 lines 13--23: this paragraph is difficult to understand. It seems to be talking about a control-flow pattern roughly equivalent to tail recursion. What is the high-level point, other than that this is possible?
     231
     232line 34: "which they call coroutines" -- a better way to make this point is presumably that the C++20 proposal only provides a specialised kind of coroutine, namely generators, despite its use of the more general word.
     233
     234line 47: "... due to dynamic stack allocation, execution..." -- this sentence doesn't scan. I suggest adding "and for" in the relevant places where currently there are only commas.
     235
     236p8 / Figure 5 (B) -- the GNU C extension of unary "&&" needs to be explained. The whole figure needs a better explanation, in fact.
     237
     238p9, lines 1--10: I wasn't sure this stepping-through really added much value. What are the truly important points to note about this code?
     239
     240p10: similarly, lines 3--27 again are somewhere between tedious and confusing. I'm sure the motivation and details of "starter semantics" can both be stated much more pithily.
     241
     242line 32: "a self-resume does not overwrite the last resumer" -- is this a hack or a defensible principled decision?
     243
     244p11: "a common source of errors" -- among beginners or among production code? Presumably the former.
     245
     246line 23: "with builtin and library" -- not sure what this means
     247
     248lines 31--36: these can be much briefer. The only important point here seems to be that coroutines cannot be copied.
     249
     250p12: line 1: what is a "task"? Does it matter?
     251
     252line 7: calling it "heap stack" seems to be a recipe for confusion. "Stack-and-heap" might be better, and contrast with "stack-and-VLS" perhaps. When "VLS" is glossed, suggest actually expanding its initials: say "length" not "size".
     253
     254line 21: are you saying "cooperative threading" is the same as "non-preemptive scheduling", or that one is a special case (kind) of the other? Both are defensible, but be clear.
     255
     256line 27: "mutual exclusion and synchronization" -- the former is a kind of the latter, so I suggest "and other forms of synchronization".
     257
     258line 30: "can either be a stackless or stackful" -- stray "a", but also, this seems to be switching from generic/background terminology to C\/-specific terminology.
     259
     260An expositional idea occurs: start the paper with a strawman naive/limited realisation of coroutines -- say, Simon Tatham's popular "Coroutines in C" web page -- and identify point by point what the limitations are and how C\/ overcomes them. Currently the presentation is often flat (lacking motivating contrasts) and backwards (stating solutions before problems). The foregoing approach might fix both of these.
     261
     262page 13: line 23: it seems a distraction to mention the Python feature here.
     263
     264p14 line 5: it seems odd to describe these as "stateless" just because they lack shared mutable state. It means the code itself is even more stateful. Maybe the "stack ripping" argument could usefully be given here.
     265
     266line 16: "too restrictive" -- would be good to have a reference to justify this, or at least give a sense of what the state-of-the-art performance in transactional memory systems is (both software and hardware)
     267
     268line 22: "simulate monitors" -- what about just *implementing* monitors? isn't that what these systems do? or is the point more about refining them somehow into something more specialised?
     269
     270p15: sections 4.1 and 4.2 seem adrift and misplaced. Split them into basic parts (which go earlier) and more advanced parts (e.g. barging, which can be explained later).
     271
     272line 31: "acquire/release" -- misses an opportunity to contrast the monitor's "enter/exit" abstraction with the less structured acquire/release of locks.
     273
     274p16 line 12: the "implicit" versus "explicit" point is unclear. Is it perhaps about the contract between an opt-in *discipline* and a language-enforced *guarantee*?
     275
     276line 28: no need to spend ages dithering about which one is default and which one is the explicit qualifier. Tell us what you decided, briefly justify it, and move on.
     277
     278p17: Figure 11: since the main point seems to be to highlight bulk acquire, include a comment which identifies the line where this is happening.
     279
     280line 2: "impossible to statically..." -- or dynamically. Doing it dynamically would be perfectly acceptable (locking is a dynamic operation after all)
     281
     282"guarantees acquisition order is consistent" -- assuming it's done in a single bulk acquire.
     283
     284p18: section 5.3: the text here is a mess. The explanations of "internal" versus "external" scheduling are unclear, and "signals as hints" is not explained. "... can cause thread starvation" -- means including a while loop, or not doing so? "There are three signalling mechanisms.." but the text does not follow that by telling us what they are. My own scribbled attempt at unpicking the internal/external thing: "threads already in the monitor, albeit waiting, have priority over those trying to enter".
     285
     286p19: line 3: "empty condition" -- explain that condition variables don't store anything. So being "empty" means that the queue of waiting threads (threads waiting to be signalled that the condition has become true) is empty.
     287
     288line 6: "... can be transformed into external scheduling..." -- OK, but give some motivation.
     289
     290p20: line 6: "mechnaism"
     291
     292lines 16--20: this is dense and can probably only be made clear with an example
     293
     294p21 line 21: clarify that nested monitor deadlock was describe earlier (in 5.2). (Is the repetition necessary?)
     295
     296line 27: "locks, and by extension monitors" -- this is true but the "by extension" argument is faulty. It is perfectly possible to use locks as a primitive and build a compositional mechanism out of them, e.g. transactions.
     297
     298p22 line 2: should say "restructured"
     299
     300line 33: "Implementing a fast subset check..." -- make clear that the following section explains how to do this. Restructuring the sections themselves could do this, or noting in the text.
     301
     302p23: line 3: "dynamic member adding, eg, JavaScript" -- needs to say "as permitted in JavaScript", and "dynamically adding members" is stylistically better
     303
     304p23: line 18: "urgent stack" -- back-reference to where this was explained before
     305
     306p24 line 7: I did not understand what was more "direct" about "direct communication". Also, what is a "passive monitor" -- just a monitor, given that monitors are passive by design?
     307
     308line 14 / section 5.9: this table was useful and it (or something like it) could be used much earlier on to set the structure of the rest of the paper. The explanation at present is too brief, e.g. I did not really understand the point about cases 7 and 8.
     309
     310p25 line 2: instead of casually dropping in a terse explanation for the newly intrdouced term "virtual processor", introduce it properly. Presumably the point is to give a less ambiguous meaning to "thread" by reserving it only for C\/'s green threads.
     311
     312Table 1: what does "No / Yes" mean?
     313
     314p26 line 15: "transforms user threads into fibres" -- a reference is needed to explain what "fibres" means... guessing it's in the sense of Adya et al.
     315
     316line 20: "Microsoft runtime" -- means Windows?
     317
     318lines 21--26: don't say "interrupt" to mean "signal", especially not without clear introduction. You can use "POSIX signal" to disambiguate from condition variables' "signal".
     319
     320p27 line 3: "frequency is usually long" -- that's a "time period" or "interval", not a frequency
     321
     322line 5: the lengthy quotation is not really necessary; just paraphrase the first sentence and move on.
     323
     324line 20: "to verify the implementation" -- I don't think that means what is intended
     325
     326Tables in section 7 -- too many significant figures. How many overall runs are described? What is N in each case?
     327
     328p29 line 2: "to eliminate this cost" -- arguably confusing since nowadays on commodity CPUs most of the benefits of inlining are not to do with call overheads, but from later optimizations enabled as a consequence of the inlining
     329
     330line 41: "a hierarchy" -- are they a hierarchy? If so, this could be explained earlier. Also, to say these make up "an integrated set... of control-flow features" verges on the tautologous.
     331
     332p30 line 15: "a common case being web servers and XaaS" -- that's two cases
     333
     334
     335Reviewing: 3
     336
     337Comments to the Author
     338# Cforall review
     339
     340Overall, I quite enjoyed reading the paper. Cforall has some very interesting ideas. I did have some suggestions that I think would be helpful before final publication. I also left notes on various parts of the paper that I find confusing when reading, in hopes that it may be useful to you.
     341
     342## Summary
     343
     344* Expand on the motivations for including both generator and coroutines, vs trying to build one atop the other
     345* Expand on the motivations for having Why both symmetric and asymettric coroutines?
     346* Comparison to async-await model adopted by other languages
     347    * C#, JS
     348    * Rust and its async/await model
     349* Consider performance comparisons against node.js and Rust frameworks
     350* Discuss performance of monitors vs finer-grained memory models and atomic operations found in other languages
     351* Why both internal/external scheduling for synchronization?
     352
     353## Generator/coroutines
     354
     355In general, this section was clear, but I thought it would be useful to provide a somewhat deeper look into why Cforall opted for the particular combination of features that it offers. I see three main differences from other languages:
     356
     357* Generators are not exposed as a "function" that returns a generator object, but rather as a kind of struct, with communication happening via mutable state instead of "return values". That is, the generator must be manually resumed and (if I understood) it is expected to store values that can then later be read (perhaps via methods), instead of having a `yield <Expr>` statement that yields up a value explicitly.
     358* Both "symmetric" and "asymmetric" generators are supported, instead of only asymmetric.
     359* Coroutines (multi-frame generators) are an explicit mechanism.
     360
     361In most other languages, coroutines are rather built by layering single-frame generators atop one another (e.g., using a mechanism like async-await), and symmetric coroutines are basically not supported. I'd like to see a bit more justification for Cforall including all the above mechanisms -- it seemed like symmetric coroutines were a useful building block for some of the user-space threading and custom scheduler mechanisms that were briefly mentioned later in the paper.
     362
     363In the discussion of coroutines, I would have expected a bit more of a comparison to the async-await mechanism offered in other languages. Certainly the semantics of async-await in JavaScript implies significantly more overhead (because each async fn is a distinct heap object). [Rust's approach avoids this overhead][zc], however, and might be worthy of a comparison (see the Performance section).
     364
     365## Locks and threading
     366
     367### Comparison to atomics overlooks performance
     368
     369There are several sections in the paper that compare against atomics -- for example, on page 15, the paper shows a simple monitor that encapsulates an integer and compares that to C++ atomics. Later, the paper compares the simplicity of monitors against the `volatile` quantifier from Java. The conclusion in section 8 also revisits this point.
     370
     371While I agree that monitors are simpler, they are obviously also significantly different from a performance perspective -- the paper doesn't seem to address this at all. It's plausible that (e.g.) the `Aint` monitor type described in the paper can be compiled and mapped to the specialized instructions offered by hardware, but I didn't see any mention of how this would be done. There is also no mention of the more nuanced memory ordering relations offered by C++11 and how one might achieve similar performance characteristics in Cforall (perhaps the answer is that one simply doesn't need to; I think that's defensible, but worth stating explicitly).
     372
     373### Justification for external scheduling feels lacking
     374
     375Cforall includes both internal and external scheduling; I found the explanation for the external scheduling mechanism to be lacking in justification. Why include both mechanisms when most languages seem to make do with only internal scheduling? It would be useful to show some scenarios where external scheduling is truly more powerful.
     376
     377I would have liked to see some more discussion of external scheduling and how it  interacts with software engineering best practices. It seems somewhat similar to AOP in certain regards. It seems to add a bit of "extra semantics" to monitor methods, in that any method may now also become a kind of synchronization point. The "open-ended" nature of this feels like it could easily lead to subtle bugs, particularly when code refactoring occurs (which may e.g. split an existing method into two). This seems particularly true if external scheduling can occur across compilation units -- the paper suggested that this is true, but I wasn't entirely clear.
     378
     379I would have also appreciated a few more details on how external scheduling is implemented. It seems to me that there must be some sort of "hooks" on mutex methods so that they can detect whether some other function is waiting on them and awaken those blocked threads. I'm not sure how such hooks are inserted, particularly across compilation units. The material in Section 5.6 didn't quite clarify the matter for me. For example, it left me somewhat confused about whether the `f` and `g` functions declared were meant to be local to a translation unit, or shared with other unit.
     380
     381### Presentation of monitors is somewhat confusing
     382
     383I found myself confused fairly often in the section on monitors. I'm just going to leave some notes here on places that I got confused in how that it could be useful to you as feedback on writing that might want to be clarified.
     384
     385To start, I did not realize that the `mutex_opt` notation was a keyword, I thought it was a type annotation. I think this could be called out more explicitly.
     386
     387Later, in section 5.2, the paper discusses `nomutex` annotations, which initially threw me, as they had not been introduced (now I realize that this paragraph is there to justify why there is no such keyword). The paragraph might be rearranged to make that clearer, perhaps by leading with the choice that Cforall made.
     388
     389On page 17, the paper states that "acquiring multiple monitors is safe from deadlock", but this could be stated a bit more precisely: acquiring multiple monitors in a bulk-acquire is safe from deadlock (deadlock can still result from nested acquires).
     390
     391On page 18, the paper states that wait states do not have to be enclosed in loops, as there is no concern of barging. This seems true but there are also other reasons to use loops (e.g., if there are multiple reasons to notify on the same condition). Thus the statement initially surprised me, as barging is only one of many reasons that I typically employ loops around waits.
     392
     393I did not understand the diagram in Figure 12 for some time. Initially, I thought that it was generic to all monitors, and I could not understand the state space. It was only later that I realized it was specific to your example. Updating the caption from "Monitor scheduling to "Monitor scheduling in the example from Fig 13" might have helped me quite a bit.
     394
     395I spent quite some time reading the boy/girl dating example (\*) and I admit I found it somewhat confusing. For example, I couldn't tell whether there were supposed to be many "girl" threads executing at once, or if there was only supposed to be one girl and one boy thread executing in a loop. Are the girl/boy threads supposed to invoke the girl/boy methods or vice versa? Surely there is some easier way to set this up? I believe that when reading the paper I convinced myself of how it was supposed to be working, but I'm writing this review some days later, and I find myself confused all over again and not able to easily figure it out.
     396
     397(\*) as an aside, I would consider modifying the example to some other form of matching, like customers and support personnel.
     398
     399## Related work
     400
     401The paper offered a number of comparisons to Go, C#, Scala, and so forth, but seems to have overlooked another recent language, Rust. In many ways, Rust seems to be closest in philosophy to Cforall, so it seems like an odd omission. I already mentioned above that Rust is in the process of shipping [async-await syntax][aa], which is definitely an alternative to the generator/coroutine approach in Cforall (though one with clear pros/cons).
     402
     403## Performance
     404
     405In the performance section in particular, you might consider comparing against some of the Rust web servers and threading systems. For example, actix is top of the [single query TechEmpower Framework benchmarks], and tokio is near the top of the [plainthreading benchmarks][pt] (hyper, the top, is more of an HTTP framework, though it is also written in Rust). It would seem worth trying to compare their "context switching" costs as well -- I believe both actix and tokio have a notion of threads that could be readily compared.
     406
     407Another addition that might be worth considering is to compare against node.js promises, although I think the comparison to process creation is not as clean.
     408
     409That said, I think that the performance comparison is not a big focus of the paper, so it may not be necessary to add anything to it.
     410
     411## Authorship of this review
     412
     413I'm going to sign this review. This review was authored by Nicholas D. Matsakis. In the intrerest of full disclosure, I'm heavily involved in the Rust project, although I dont' think that influenced this review in particular. Feel free to reach out to me for clarifying questions.
     414
     415## Links
     416
     417[aa]: https://blog.rust-lang.org/2019/09/30/Async-await-hits-beta.html
     418[zc]: https://aturon.github.io/blog/2016/08/11/futures/
     419[sq]: https://www.techempower.com/benchmarks/#section=data-r18&hw=ph&test=db
     420[pt]: https://www.techempower.com/benchmarks/#section=data-r18&hw=ph&test=plaintext
     421
     422
     423
     424Subject: Re: manuscript SPE-19-0219
     425To: "Peter A. Buhr" <pabuhr@uwaterloo.ca>
     426From: Richard Jones <R.E.Jones@kent.ac.uk>
     427Date: Tue, 12 Nov 2019 22:43:55 +0000
     428
     429Dear Dr Buhr
     430
     431Your should have received a decision letter on this today. I am sorry that this
     432has taken so long. Unfortunately SP&E receives a lot of submissions and getting
     433reviewers is a perennial problem.
     434
     435Regards
     436Richard
     437
     438Peter A. Buhr wrote on 11/11/2019 13:10:
     439>     26-Jun-2019
     440>     Your manuscript entitled "Advanced Control-flow and Concurrency in Cforall"
     441>     has been received by Software: Practice and Experience. It will be given
     442>     full consideration for publication in the journal.
     443>
     444> Hi, it has been over 4 months since submission of our manuscript SPE-19-0219
     445> with no response.
     446>
     447> Currently, I am refereeing a paper for IEEE that already cites our prior SP&E
     448> paper and the Master's thesis forming the bases of the SP&E paper under
     449> review. Hence our work is apropos and we want to get it disseminates as soon as
     450> possible.
     451>
     452> [3] A. Moss, R. Schluntz, and P. A. Buhr, "Cforall: Adding modern programming
     453>      language features to C," Software - Practice and Experience, vol. 48,
     454>      no. 12, pp. 2111-2146, 2018.
     455>
     456> [4] T. Delisle, "Concurrency in C for all," Master's thesis, University of
     457>      Waterloo, 2018.  [Online].  Available:
     458>      https://uwspace.uwaterloo.ca/bitstream/handle/10012/12888
     459
     460
     461
     462Date: Mon, 13 Jan 2020 05:33:15 +0000
     463From: Richard Jones <onbehalfof@manuscriptcentral.com>
     464Reply-To: R.E.Jones@kent.ac.uk
     465To: pabuhr@uwaterloo.ca
     466Subject: Revision reminder - SPE-19-0219
     467
     46813-Jan-2020
     469Dear Dr Buhr
     470SPE-19-0219
     471
     472This is a reminder that your opportunity to revise and re-submit your
     473manuscript will expire 28 days from now. If you require more time please
     474contact me directly and I may grant an extension to this deadline, otherwise
     475the option to submit a revision online, will not be available.
     476
     477I look forward to receiving your revision.
     478
     479Sincerely,
     480
     481Prof. Richard Jones
     482Editor, Software: Practice and Experience
     483https://mc.manuscriptcentral.com/spe
     484
     485
     486
     487Date: Wed, 5 Feb 2020 04:22:18 +0000
     488From: Aaron Thomas <onbehalfof@manuscriptcentral.com>
     489Reply-To: speoffice@wiley.com
     490To: tdelisle@uwaterloo.ca, pabuhr@uwaterloo.ca
     491Subject: SPE-19-0219.R1 successfully submitted
     492
     49304-Feb-2020
     494
     495Dear Dr Buhr,
     496
     497Your manuscript entitled "Advanced Control-flow and Concurrency in Cforall" has
     498been successfully submitted online and is presently being given full
     499consideration for publication in Software: Practice and Experience.
     500
     501Your manuscript number is SPE-19-0219.R1.  Please mention this number in all
     502future correspondence regarding this submission.
     503
     504You can view the status of your manuscript at any time by checking your Author
     505Center after logging into https://mc.manuscriptcentral.com/spe.  If you have
     506difficulty using this site, please click the 'Get Help Now' link at the top
     507right corner of the site.
     508
     509Thank you for submitting your manuscript to Software: Practice and Experience.
     510
     511Sincerely,
     512Software: Practice and Experience Editorial Office
     513
  • doc/theses/thierry_delisle_PhD/.gitignore

    r26a09f92 r6565321  
    88
    99comp_II/build/
     10comp_II/img/*.fig.bak
    1011comp_II/comp_II.pdf
    1112comp_II/comp_II.ps
  • doc/theses/thierry_delisle_PhD/comp_II/Makefile

    r26a09f92 r6565321  
    1818
    1919FIGURES = ${addsuffix .tex, \
     20        base \
     21        empty \
     22        emptybit \
     23        emptytree \
     24        resize \
    2025}
    2126
     
    7075        mkdir -p ${Build}
    7176
    72 %.tex : %.fig ${Build}
     77%.tex : img/%.fig ${Build}
    7378        fig2dev -L eepic $< > ${Build}/$@
    7479
    75 %.ps : %.fig | ${Build}
     80%.ps : img/%.fig | ${Build}
    7681        fig2dev -L ps $< > ${Build}/$@
    7782
    78 %.pstex : %.fig | ${Build}
     83%.pstex : img/%.fig | ${Build}
    7984        fig2dev -L pstex $< > ${Build}/$@
    8085        fig2dev -L pstex_t -p ${Build}/$@ $< > ${Build}/$@_t
  • doc/theses/thierry_delisle_PhD/comp_II/comp_II.tex

    r26a09f92 r6565321  
    1 \documentclass[11pt,fullpage]{article}
     1\documentclass[11pt]{article}
     2\usepackage{fullpage}
    23\usepackage[T1]{fontenc}
    34\usepackage[utf8]{inputenc}
     
    67\usepackage{xcolor}
    78\usepackage{graphicx}
    8 \usepackage[hidelinks]{hyperref}
     9\usepackage{epic,eepic}
    910\usepackage{glossaries}
    1011\usepackage{textcomp}
    11 \usepackage{geometry}
     12\usepackage[hidelinks]{hyperref}
     13%\usepackage[margin=1in]{geometry}
     14%\usepackage{float}
    1215
    1316% cfa macros used in the document
     
    5154\section{Introduction}
    5255\subsection{\CFA and the \CFA concurrency package}
    53 \CFA\cit is a modern, polymorphic, non-object-oriented, backwards-compatible extension of the C programming language. It aims to add high productivity features while maintaning the predictible performance of C. As such concurrency in \CFA\cit aims to offer simple and safe high-level tools while still allowing performant code. Concurrent code is written in the syncrhonous programming paradigm but uses \glspl{uthrd} in order to achieve the simplicity and maintainability of synchronous programming without sacrificing the efficiency of asynchronous programing. As such the \CFA scheduler is a user-level scheduler that maps \glspl{uthrd} onto \glspl{kthrd}.
    54 
    55 The goal of this research is to produce a scheduler that is simple to use and offers acceptable performance in all cases. Here simplicity does not refer to the API but to how much scheduling concerns programmers need to take into account when using the \CFA concurrency package. Therefore, the main goal of this proposal is as follows :
     56\CFA\cit is a modern, polymorphic, non-object-oriented, backwards-compatible extension of the C programming language. It aims to add high-productivity features while maintaning the predictible performance of C. As such, concurrency in \CFA\cit aims to offer simple and safe high-level tools while still allowing performant code. \CFA concurrrent code is written in the synchronous programming paradigm but uses \glspl{uthrd} in order to achieve the simplicity and maintainability of synchronous programming without sacrificing the efficiency of asynchronous programing. As such, the \CFA \emph{scheduler} is a preemptive user-level scheduler that maps \glspl{uthrd} onto \glspl{kthrd}.
     57
     58Scheduling occurs when execution switches from one thread to another, where the second thread is implicitly chosen by the scheduler. This scheduling is an indirect handoff, as opposed to generators and coroutines which explicitly switch to the next generator and coroutine respectively. The cost of switching between two threads for an indirect handoff has two components : the cost of actually context-switching, i.e., changing the relevant registers to move execution from one thread to the other, and the cost of scheduling, i.e., deciding which thread to run next among all the threads ready to run. The first cost is generally constant and fixed, while the scheduling cost can vary based on the system state\footnote{Affecting the context-switch cost is whether it is done in one step, after the scheduling, or in two steps, context-switching to a fixed third-thread before scheduling.}. Adding multiple \glspl{kthrd} does not fundamentally change the scheduler semantics or requirements, it simply adds new correctness requirements, i.e. \textit{linearizability}, and a new dimension to performance: scalability, where scheduling cost now also depends on contention.
     59
     60The more threads switch, the more the administrating cost of scheduling becomes noticeable. It is therefore important to build a scheduler with the lowest possible cost and latency. Another important consideration is \emph{fairness}. In principle, scheduling should give the illusion of perfect fairness, where all threads ready to run are running \emph{simultaneously}. While the illusion of simultaneity is easier to reason about, it can break down if the scheduler allows to much unfairness. Therefore, the scheduler should offer as much fairness as needed to guarantee eventual progress, but use unfairness to help performance. In practice, threads must wait in turn but there can be advantages to unfair scheduling, e.g., the express cash register at a grocery store.
     61
     62The goal of this research is to produce a scheduler that is simple for programmers to understand and offers good performance. Here understandability does not refer to the API but to how much scheduling concerns programmers need to take into account when writing a \CFA concurrent package. Therefore, the main goal of this proposal is :
    5663\begin{quote}
    57 The \CFA scheduler should be \emph{viable} for any workload.
     64The \CFA scheduler should be \emph{viable} for \emph{any} workload.
    5865\end{quote}
    5966
    60 This objective includes producing a scheduling strategy with minimal fairness guarantees, creating an abstraction layer over the operating system to handle kernel-threads spinning unnecessarily and hide blocking I/O operations and, writing sufficient library tools to allow developpers to properly use the scheduler.
     67For a general purpose scheduler, it is impossible to produce an optimal algorithm as it would require knowledge of the future behaviour of threads. As such, scheduling performance is generally either defined by the best case scenario, a workload to which the scheduler is tailored, or the worst case scenario, i.e., the scheduler behaves no worst than \emph{X}. For this proposal, the performance is evaluated using the second approach to allow \CFA programmers to rely on scheduling performance. A solution to this impossibility is to allow programmers to write their own scheduler, that is not the subject of this proposal, which considers only the default scheduler. As such, it is important that only programmers with exceptionally high performance requirements should need to write their own scheduler and replace the scheduler in this proposal.
     68
     69This objective includes producing a scheduling strategy with sufficient fairness guarantees, creating an abstraction layer over the operating system to handle kernel-threads spinning unnecessarily and hide blocking I/O operations, and writing sufficient library tools to allow developers to indirectly use the scheduler.
    6170
    6271% ===============================================================================
     
    6473
    6574\section{Scheduling for \CFA}
    66 While the \CFA concurrency package doesn't have any particular scheduling needs beyond those of any concurrency package which uses \glspl{uthrd}, it is important that the default \CFA Scheduler be viable in general. Indeed, since the \CFA Scheduler does not target any specific workloads, it is unrealistic to demand that it use the best scheduling strategy in all cases. However, it should offer a viable ``out of the box'' solution for most scheduling problems so that programmers can quickly write performant concurrent without needed to think about which scheduling strategy is more appropriate for their workload. Indeed, only programmers with exceptionnaly high performance requirements should need to write their own scheduler. More specifically, two broad types of schedulering strategies should be avoided in order to avoid penalizing certain types of workloads : feedback-based and priority schedulers.
     75While the \CFA concurrency package does not have any particular scheduling requirements beyond supporting \glspl{uthrd}. Therefore, the detailed requirements of the \CFA scheduler are :
     76
     77\paragraph{Correctness} As with any other concurrent data structure or algorithm, the correctness requirement is paramount. The scheduler cannot allow threads to be dropped from the ready-queue, i.e., scheduled but never run, or be executed multiple times when only being scheduled once. Since \CFA concurrency has no spurious wakeup, this definition of correctness also means the scheduler should have no spurious wakeup. The \CFA scheduler must be correct.
     78
     79\paragraph{Performance} The performance of a scheduler can generally be mesured in terms of scheduling cost, scalability and latency. Scheduling cost is the cost to switch from one thread to another, as mentioned above. For simple applications where a single kernel thread does most of the scheduling, it is generally the dominating cost. When adding many kernel threads, scalability can become an issue, effectively increasing the cost of context-switching when contention is high. Finally, a third axis of performance is tail latency. This measurement is related to fairness and mesures how long is needed for a thread to be run once scheduled but evaluated in the worst cases. The \CFA scheduler should offer good performance in all three metrics.
     80
     81\paragraph{Fairness} Like performance, this requirements has several aspect : eventual progress, predictability and performance reliablility. As a hard requirement, the \CFA scheduler must guarantee eventual progress, i.e., prevent starvation, otherwise the above mentioned illusion of simultaneous execution is broken and the scheduler becomes much more complext to reason about. Beyond this requirement, performance should be predictible and reliable, which means similar workloads achieve similar performance and programmer intuition is respected. An example of this is : a thread that yields agressively should not run more often then other tasks. While this is intuitive, it does not hold true for many work-stealing or feedback based schedulers. The \CFA scheduler must guarantee eventual progress and should be predictible and offer reliable performance.
     82
     83\paragraph{Efficiency} Finally, efficient usage of CPU resources is also an important requirement. This issue is discussed more in depth towards the end of this proposal. It effectively refers to avoiding using CPU power when there are no threads to run, and conversely, use all CPUs available when the workload can benefit from it. Balancing these two states is where the complexity lies. The \CFA scheduler should be efficient with respect to the underlying (shared) computer.
     84
     85To achieve these requirements, I can reject two broad types of scheduling strategies : feedback-based and priority schedulers.
    6786
    6887\subsection{Feedback-Based Schedulers}
    69 Many operating systems use schedulers based on feadback loops in some form, they measure how much CPU a particular thread has used\footnote{Different metrics can be used to here but it is not relevant to the discussion.} and schedule threads based on this metric. These strategies are sensible for operating systems but rely on two assumptions on the workload :
     88Many operating systems use schedulers based on feedback in some form, e.g., measuring how much CPU a particular thread has used\footnote{Different metrics can measured here but it is not relevant to the discussion.} and schedule threads based on this metric. These strategies are sensible for operating systems but rely on two assumptions on the workload :
    7089
    7190\begin{enumerate}
    72         \item Threads live long enough to be scheduled many times.
    73         \item Cooperation among all threads is not simply infeasible, it is a security risk.
     91        \item Threads live long enough for useful feedback information to be to gathered.
     92        \item Threads belong to multiple users so fairness across threads is insufficient.
    7493\end{enumerate}
    7594
    76 While these two assumptions generally hold for operating systems, they may not for \CFA programs. In fact, \CFA uses \glspl{uthrd} which have the explicit goal of reducing the cost of threading primitives to allow many smaller threads. This can naturally lead to have threads with much shorter lifetime and only being scheduled a few times. Scheduling strategies based on feadback loops cannot be effective in these cases because they will not have the opportunity to measure the metrics that underlay the algorithm. Note that the problem of feadback loop convergence (reacting too slowly to scheduling events) is not specific to short lived threads but can also occur with threads that show drastic changes in scheduling event, e.g., threads running for long periods of time and then suddenly blocking and unblocking quickly and repeatedly.
    77 
    78 In the context of operating systems, these concerns can be overshadowed by a more pressing concern : security. When multiple users are involved, it is possible that some users are malevolent and try to exploit the scheduling strategy in order to achieve some nefarious objective. Security concerns mean that more precise and robust fairness metrics must be used. In the case of the \CFA scheduler, every thread runs in the same user-space and are controlled from the same user. It is then possible to safely ignore the possibility that threads are malevolent and assume that all threads will ignore or cooperate with each other. This allows for a much simpler fairness metric and in this proposal ``fairness'' will be considered as equal opportunities to run once scheduled.
    79 
    80 Since feadback is not necessarily feasible within the lifetime of all threads and a simple fairness metric can be used, the scheduling strategy proposed for the \CFA runtime does not user per-threads feedback. Feedback loops in general are not rejected for secondary concerns like idle sleep, but no feedback loop is used to decide which thread to run next.
     95While these two assumptions generally hold for operating systems, they may not for user-level threading. Since \CFA has the explicit goal of allowing many smaller threads, this can naturally lead to threads with much shorter lifetime, only being scheduled a few times. Scheduling strategies based on feedback cannot be effective in these cases because they do not have the opportunity to measure the metrics that underlie the algorithm. Note that the problem of feedback convergence (reacting too slowly to scheduling events) is not specific to short lived threads but can also occur with threads that show drastic changes in scheduling, e.g., threads running for long periods of time and then suddenly blocking and unblocking quickly and repeatedly.
     96
     97In the context of operating systems, these concerns can be overshadowed by a more pressing concern : security. When multiple users are involved, it is possible that some users are malevolent and try to exploit the scheduling strategy in order to achieve some nefarious objective. Security concerns mean that more precise and robust fairness metrics must be used to guarantee fairness across processes created by users as well as threads created within a process. In the case of the \CFA scheduler, every thread runs in the same user-space and are controlled by the same user. Fairness across users is therefore a given and it is then possible to safely ignore the possibility that threads are malevolent. This approach allows for a much simpler fairness metric and in this proposal ``fairness'' is considered as follows : when multiple threads are cycling through the system, the total ordering of threads being scheduled, i.e., pushed onto the ready-queue, should not differ much from the total ordering of threads being executed, i.e., popped from the ready-queue.
     98
     99Since feedback is not necessarily feasible within the lifetime of all threads and a simple fairness metric can be used, the scheduling strategy proposed for the \CFA runtime does not use per-threads feedback. Feedback in general is not rejected for secondary concerns like idle sleep, but no feedback is used to decide which thread to run next.
    81100
    82101\subsection{Priority Schedulers}
    83 Another broad category of schedulers are priority schedulers. In these scheduling strategies threads have priorities and the runtime schedules the threads with the highest priority before scheduling other threads. Threads with equal priority are scheduled using a secondary strategy, often something simple like round-robin or FIFO. These priority mean that, as long as there is a thread with a higher priority that desires to run, a thread with a lower priority will not run. This possible starving of threads can dramatically increase programming complexity since starving threads and priority inversion (prioritising a lower priority thread) can both lead to serious problems, leaving programmers between a rock and a hard place.
    84 
    85 An important observation to make is that threads do not need to have explicit priorities for problems to be possible. Indeed, any system with multiple ready-queues and attempts to exhaust one queue before accessing the other queues, could encounter starvation problems. A popular scheduling strategy that suffers from implicit priorities is work-stealing. Work-stealing is generally presented as follows :
    86 
    87 \begin{itemize}
    88         \item Each processor has a list of threads.
    89 \end{itemize}
     102Another broad category of schedulers are priority schedulers. In these scheduling strategies, threads have priorities and the runtime schedules the threads with the highest priority before scheduling other threads. Threads with equal priority are scheduled using a secondary strategy, often something simple like round-robin or FIFO. These priority mean that, as long as there is a thread with a higher priority that desires to run, a thread with a lower priority does not run. This possible starving of threads can dramatically increase programming complexity since starving threads and priority inversion (prioritizing a lower priority thread) can both lead to serious problems.
     103
     104An important observation to make is that threads do not need to have explicit priorities for problems to occur. Indeed, any system with multiple ready-queues and attempts to exhaust one queue before accessing the other queues, can encounter starvation problems. A popular scheduling strategy that suffers from implicit priorities is work-stealing. Work-stealing is generally presented as follows, each processor has a list of ready threads.
    90105\begin{enumerate}
    91106        \item Run threads from ``this'' processor's list.
     
    93108\end{enumerate}
    94109
    95 In a loaded system\footnote{A loaded system is a system where threads are being run at the same rate they are scheduled}, if a thread does not yield or block for an extended period of time, threads on the same processor list will starve if no other processors can exhaust their list.
     110In a loaded system\footnote{A loaded system is a system where threads are being run at the same rate they are scheduled.}, if a thread does not yield, block or preempt for an extended period of time, threads on the same processor list starve if no other processors exhaust their list.
    96111
    97112Since priorities can be complex to handle for programmers, the scheduling strategy proposed for the \CFA runtime does not use a strategy with either implicit or explicit thread priorities.
    98113
    99 \subsection{Schedulers without feadback or priorities}
    100 I claim that the ideal default scheduler for the \CFA runtime is a scheduler that offers good scalability and a simple fairness guarantee that is easy for programmers to reason about. The simplest fairness guarantee is to guarantee FIFO ordering, i.e., threads scheduled first will run first. However, enforcing FIFO ordering generally conflicts with scalability across multiple processors because of the additionnal synchronization. Thankfully, strict FIFO is not needed for scheduling. Since concurrency is inherently non-deterministic, fairness concerns in scheduling are only a problem if a thread repeatedly runs before another thread can run\footnote{This is because the non-determinism means that programmers must already handle ordering problems in order to produce correct code and already must rely on weak guarantees, for example that a specific thread will \emph{eventually} run.}. This need for unfairness to persist before problems occur means that the FIFO fairness guarantee can be significantly relaxed without causing problems. For this proposal, the target guarantee is that the \CFA scheduler guarantees \emph{probable} FIFO ordering, which is defined as follows :
     114\subsection{Schedulers without feedback or priorities}
     115This proposal conjectures that is is possible to construct a default scheduler for the \CFA runtime that offers good scalability and a simple fairness guarantee that is easy for programmers to reason about. The simplest fairness guarantee is FIFO ordering, i.e., threads scheduled first run first. However, enforcing FIFO ordering generally conflicts with scalability across multiple processors because of the additionnal synchronization. Thankfully, strict FIFO is not needed for sufficient fairness. Since concurrency is inherently non-deterministic, fairness concerns in scheduling are only a problem if a thread repeatedly runs before another thread can run. This is because the non-determinism means that programmers must already handle ordering problems in order to produce correct code and already must rely on weak guarantees, for example that a specific thread will \emph{eventually} run. Since some reordering does not break correctness, the FIFO fairness guarantee can be significantly relaxed without causing problems. For this proposal, the target guarantee is that the \CFA scheduler provides \emph{probable} FIFO ordering, which allows reordering but makes it improbable that threads are reordered far from their position in total ordering.
     116
     117Scheduling is defined as follows :
    101118\begin{itemize}
    102         \item Given two threads $X$ and $Y$, the odds that thread $X$ runs $N$ times \emph{after} thread $Y$ is scheduled but \emph{before} it is run, decreases exponentially with regards to $N$.
     119        \item Given two threads $X$ and $Y$, the odds that thread $X$ runs $N$ times \emph{after} thread $Y$ is scheduled but \emph{before} it is run, decreases exponentially with regard to $N$.
    103120\end{itemize}
    104121
    105 While this is not a strong guarantee, the probability that problems persist for long period of times decreases exponentially, making persisting problems virtually impossible.
    106 
    107 \subsection{Real-Time}
    108 While the objective of this proposed scheduler is similar to the objective of real-time scheduling, this proposal is not a proposal for real-time scheduler and as such makes no attempt to offer either soft or hard guarantees on scheduling delays.
     122While this is not a bounded guarantee, the probability that unfairness persist for long periods of times decreases exponentially, making persisting unfairness virtually impossible.
    109123
    110124% ===============================================================================