- File:
-
- 1 edited
-
doc/proposals/concurrency/text/cforall.tex (modified) (3 diffs)
Legend:
- Unmodified
- Added
- Removed
-
doc/proposals/concurrency/text/cforall.tex
r3628765 r21a1efb 5 5 % ====================================================================== 6 6 7 This thesis presents the design for a set of concurrency features in \CFA. Since it is a new dialect of C, the following is a quick introduction to the language,specifically tailored to the features needed to support concurrency.7 As mentionned in the introduction, the document presents the design for the concurrency features in \CFA. Since it is a new language here is a quick review of the language specifically tailored to the features needed to support concurrency. 8 8 9 \CFA is a extension of ISO -C and therefore supports all of the same paradigms as C. It is a non-object oriented system language, meaning most of the major abstractions have either no runtime overhead or can be opt-out easily. Like C, the basics of \CFA revolve around structures and routines, which are thin abstractions over machine code. The vast majority of the code produced by the \CFA translator respects memory-layouts and calling-conventions laid out by C. Interestingly, while \CFA is not an object-oriented language, lacking the concept of a received (e.g.: this), it does have some notion of objects\footnote{C defines the term objects as : [Where to I get the C11 reference manual?]}, most importantly construction and destruction of objects. Most of the following pieces of code can be found on the \CFA website\cite{www-cfa}9 \CFA is a extension of ISO C and therefore supports much of the same paradigms as C. It is a non-object oriented system level language, meaning it has very most of the major abstractions have either no runtime cost or can be opt-out easily. Like C, the basics of \CFA revolve around structures and routines, which are thin abstractions over assembly. The vast majority of the code produced by a \CFA compiler respects memory-layouts and calling-conventions laid out by C. However, while \CFA is not an object-oriented language according to a strict definition. It does have some notion of objects, most importantly construction and destruction of objects. Most of the following pieces of code can be found as is on the \CFA website : \cite{www-cfa} 10 10 11 11 \section{References} 12 12 13 Like \CC, \CFA introduces references as an alternative to pointers. In regards to concurrency, the semantics difference between pointers and references are not particularly relevant but since this document uses mostly references here is a quick overview of the semantics :13 Like \CC, \CFA introduces references as an alternative to pointers. In regards to concurrency, the semantics difference between pointers and references aren't particularly relevant but since this document uses mostly references here is a quick overview of the semantics : 14 14 \begin{cfacode} 15 15 int x, *p1 = &x, **p2 = &p1, ***p3 = &p2, 16 16 &r1 = x, &&r2 = r1, &&&r3 = r2; 17 ***p3 = 3; //change x 18 r3 = 3; //change x, ***r3 19 **p3 = ...; //change p1 20 *p3 = ...; //change p2 21 int y, z, & ar[3] = {x, y, z}; //initialize array of references 22 typeof( ar[1]) p; //is int, i.e., the type of referenced object 23 typeof(&ar[1]) q; //is int &, i.e., the type of reference 24 sizeof( ar[1]) == sizeof(int); //is true, i.e., the size of referenced object 25 sizeof(&ar[1]) == sizeof(int *); //is true, i.e., the size of a reference 17 ***p3 = 3; // change x 18 r3 = 3; // change x, ***r3 19 **p3 = ...; // change p1 20 &r3 = ...; // change r1, (&*)**r3 21 *p3 = ...; // change p2 22 &&r3 = ...; // change r2, (&(&*)*)*r3 23 &&&r3 = p3; // change r3 to p3, (&(&(&*)*)*)r3 24 int y, z, & ar[3] = { x, y, z }; // initialize array of references 25 &ar[1] = &z; // change reference array element 26 typeof( ar[1] ) p; // is int, i.e., the type of referenced object 27 typeof( &ar[1] ) q; // is int &, i.e., the type of reference 28 sizeof( ar[1] ) == sizeof( int ); // is true, i.e., the size of referenced object 29 sizeof( &ar[1] ) == sizeof( int *); // is true, i.e., the size of a reference 26 30 \end{cfacode} 27 31 The important thing to take away from this code snippet is that references offer a handle to an object much like pointers but which is automatically derefferenced when convinient. … … 29 33 \section{Overloading} 30 34 31 Another important feature of \CFA is function overloading as in Java and \CC, where routine with the same name are selected based on the numbers and type of the arguments. As well, \CFA uses the return type as part of the selection criteria, as in Ada\cite{Ada}. For routines with multiple parameters and returns, the selection is complex.35 Another important feature \CFA has in common with \CC is function overloading : 32 36 \begin{cfacode} 33 // selection based on type and number of parameters34 void f( void); //(1)35 void f( char); //(2)36 void f( int, double); //(3)37 f(); //select (1)38 f( 'a'); //select (2)39 f( 3, 5.2); //select (3)37 // selection based on type and number of parameters 38 void f( void ); // (1) 39 void f( char ); // (2) 40 void f( int, double ); // (3) 41 f(); // select (1) 42 f( 'a' ); // select (2) 43 f( 3, 5.2 ); // select (3) 40 44 41 //selection based on type and number of returns 42 char f(int); //(1) 43 double f(int); //(2) 44 char c = f(3); //select (1) 45 double d = f(4); //select (2) 45 // selection based on type and number of returns 46 char f( int ); // (1) 47 double f( int ); // (2) 48 [ int, double ] f( int ); // (3) 49 char c = f( 3 ); // select (1) 50 double d = f( 4 ); // select (2) 51 [ int, double ] t = f( 5 ); // select (3) 46 52 \end{cfacode} 47 This feature is particularly important for concurrency since the runtime system relies on creating different types to represent concurrency objects. Therefore, overloading is necessary to prevent the need for long prefixes and other naming conventions that prevent name clashes. As seen in chapter \ref{basics}, routines main is an example that benefits from overloading.53 This feature is particularly important for concurrency since the runtime system relies on creating different types do represent concurrency objects. Therefore, overloading is necessary to prevent the need for long prefixes and other naming conventions that prevent clashes. As seen in chapter \ref{basics}, the main is an example of routine that benefits from overloading when concurrency in introduced. 48 54 49 55 \section{Operators} 50 56 Overloading also extends to operators. The syntax for denoting operator-overloading is to name a routine with the symbol of the operator and question marks where the arguments of the operation would be, like so : 51 57 \begin{cfacode} 52 int ++? (int op); //unary prefix increment53 int ?++ (int op); //unary postfix increment54 int ?+? (int op1, int op2); //binary plus55 int ?<=?( int op1, int op2); //binary less than56 int ?=? (int & op1, int op2); //binary assignment57 int ?+=?( int & op1, int op2); //binary plus-assignment58 int ++?( int op ); // unary prefix increment 59 int ?++( int op ); // unary postfix increment 60 int ?+?( int op1, int op2 ); // binary plus 61 int ?<=?( int op1, int op2 ); // binary less than 62 int ?=?( int & op1, int op2 ); // binary assignment 63 int ?+=?( int & op1, int op2 ); // binary plus-assignment 58 64 59 struct S { int i, j;};60 S ?+?( S op1, S op2) { //add two structures61 return (S){ op1.i + op2.i, op1.j + op2.j};65 struct S { int i, j; }; 66 S ?+?( S op1, S op2 ) { // add two structures 67 return (S){ op1.i + op2.i, op1.j + op2.j }; 62 68 } 63 S s1 = { 1, 2}, s2 = {2, 3}, s3;64 s3 = s1 + s2; //compute sum: s3 == {2, 5}69 S s1 = { 1, 2 }, s2 = { 2, 3 }, s3; 70 s3 = s1 + s2; // compute sum: s3 == { 2, 5 } 65 71 \end{cfacode} 66 While concurrency does not use operator overloading directly, this feature is more important as an introduction for the syntax of constructors. 72 73 Since concurrency does not use operator overloading, this feature is more important as an introduction for the syntax of constructors. 67 74 68 75 \section{Constructors/Destructors} 69 Object life-time is often a challenge in concurrency. \CFA uses the approach of giving concurrent meaning to object life-time as a mean of synchronization and/or mutual exclusion. Since \CFA relies heavily on the life time of objects, constructors and destructors are a core feature required for concurrency and parallelism.\CFA uses the following syntax for constructors and destructors :76 \CFA uses the following syntax for constructors and destructors : 70 77 \begin{cfacode} 71 78 struct S { … … 73 80 int * ia; 74 81 }; 75 void ?{}( S & s, int asize) { //constructor operator76 s .size = asize; //initialize fields77 s.ia = calloc(size, sizeof(S));82 void ?{}( S & s, int asize ) with s { // constructor operator 83 size = asize; // initialize fields 84 ia = calloc( size, sizeof( S ) ); 78 85 } 79 void ^?{}( S & s) { //destructor operator80 free( ia); //de-initialization fields86 void ^?{}( S & s ) with s { // destructor operator 87 free( ia ); // de-initialization fields 81 88 } 82 89 int main() { 83 S x = { 10}, y = {100}; //implict calls: ?{}(x, 10), ?{}(y, 100)84 ... //use x and y85 ^x{}; ^y{}; // explicit calls to de-initialize86 x{ 20}; y{200}; //explicit calls to reinitialize87 ... //reuse x and y88 } //implict calls: ^?{}(y), ^?{}(x)90 S x = { 10 }, y = { 100 }; // implict calls: ?{}( x, 10 ), ?{}( y, 100 ) 91 ... // use x and y 92 ^x{}; ^y{}; // explicit calls to de-initialize 93 x{ 20 }; y{ 200 }; // explicit calls to reinitialize 94 ... // reuse x and y 95 } // implict calls: ^?{}( y ), ^?{}( x ) 89 96 \end{cfacode} 90 The language guarantees that every object and all their fields are constructed. Like \CC, construction of an object is automatically done on allocation and destruction of the object is done on deallocation. Allocation and deallocation can occur on the stack or on the heap. 91 \begin{cfacode} 92 { 93 struct S s = {10}; //allocation, call constructor 94 ... 95 } //deallocation, call destructor 96 struct S * s = new(); //allocation, call constructor 97 ... 98 delete(s); //deallocation, call destructor 99 \end{cfacode} 100 Note that like \CC, \CFA introduces \code{new} and \code{delete}, which behave like \code{malloc} and \code{free} in addition to constructing and destructing objects, after calling \code{malloc} and before calling \code{free} respectively. 97 The language guarantees that every object and all their fields are constructed. Like \CC construction is automatically done on declaration and destruction done when the declared variables reach the end of its scope. 101 98 102 \section{Parametric Polymorphism} 103 Routines in \CFA can also be reused for multiple types. This is done using the \code{forall} clause which gives \CFA it's name. \code{forall} clauses allow seperatly compiled routines to support generic usage over multiple types. For example, the following sum function will work for any type which support construction from 0 and addition : 104 \begin{cfacode} 105 //constraint type, 0 and + 106 forall(otype T | { void ?{}(T *, zero_t); T ?+?(T, T); }) 107 T sum(T a[ ], size_t size) { 108 T total = 0; //construct T from 0 109 for(size_t i = 0; i < size; i++) 110 total = total + a[i]; //select appropriate + 111 return total; 112 } 113 114 S sa[5]; 115 int i = sum(sa, 5); //use S's 0 construction and + 116 \end{cfacode} 117 118 Since writing constraints on types can become cumbersome for more constrained functions, \CFA also has the concept of traits. Traits are named collection of constraints which can be used both instead and in addition to regular constraints: 119 \begin{cfacode} 120 trait sumable( otype T ) { 121 void ?{}(T *, zero_t); //constructor from 0 literal 122 T ?+?(T, T); //assortment of additions 123 T ?+=?(T *, T); 124 T ++?(T *); 125 T ?++(T *); 126 }; 127 forall( otype T | sumable(T) ) //use trait 128 T sum(T a[], size_t size); 129 \end{cfacode} 130 131 \section{with Clause/Statement} 132 Since \CFA lacks the concept of a receiver, certain functions end-up needing to repeat variable names often, to solve this \CFA offers the \code{with} statement which opens an aggregate scope making its fields directly accessible (like Pascal). 133 \begin{cfacode} 134 struct S { int i, j; }; 135 int mem(S & this) with this //with clause 136 i = 1; //this->i 137 j = 2; //this->j 138 } 139 int foo() { 140 struct S1 { ... } s1; 141 struct S2 { ... } s2; 142 with s1 //with statement 143 { 144 //access fields of s1 145 //without qualification 146 with s2 //nesting 147 { 148 //access fields of s1 and s2 149 //without qualification 150 } 151 } 152 with s1, s2 //scopes open in parallel 153 { 154 //access fields of s1 and s2 155 //without qualification 156 } 157 } 158 \end{cfacode} 159 160 For more information on \CFA see \cite{cforall-ug,rob-thesis,www-cfa}. 99 For more information see \cite{cforall-ug,rob-thesis,www-cfa}.
Note:
See TracChangeset
for help on using the changeset viewer.
