Changes in / [ef42e764:e109716]


Ignore:
Files:
8 edited

Legend:

Unmodified
Added
Removed

  • doc/aaron_comp_II/comp_II.tex

    ref42e764 re109716  
    6262%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
    6363
    64 \newcommand{\bigO}[1]{O\left( #1 \right)}
    65 
    6664\begin{document}
    6765\pagestyle{headings}
    … …  
    118116The ©identity© function above can be applied to any complete object type (or ``©otype©'').
    119117The type variable ©T© is transformed into a set of additional implicit parameters to ©identity© which encode sufficient information about ©T© to create and return a variable of that type.
    120 The current \CFA implementation passes the size and alignment of the type represented by an ©otype© parameter, as well as an assignment operator, constructor, copy constructor and destructor.
    121 Here, the runtime cost of polymorphism is spread over each polymorphic call, due to passing more arguments to polymorphic functions; preliminary experiments have shown this overhead to be similar to \CC virtual function calls.
    122 Determining if packaging all polymorphic arguments to a function into a virtual function table would reduce the runtime overhead of polymorphic calls is an open research question.
     118The current \CFA implementation passes the size and alignment of the type represented by an ©otype© parameter, as well as an assignment operator, constructor, copy constructor and destructor.
    123119
    124120Since bare polymorphic types do not provide a great range of available operations, \CFA also provides a \emph{type assertion} mechanism to provide further information about a type:
    … …  
    133129double magic = four_times(10.5); // T is bound to double, uses (1) to satisfy type assertion
    134130\end{lstlisting}
    135 These type assertions may be either variable or function declarations that depend on a polymorphic type variable.
    136 ©four_times© can only be called with an argument for which there exists a function named ©twice© that can take that argument and return another value of the same type; a pointer to the appropriate ©twice© function is passed as an additional implicit parameter to the call to ©four_times©.
     131These type assertions may be either variable or function declarations which depend on a polymorphic type variable.
     132©four_times© can only be called with an argument for which there exists a function named ©twice© that can take that argument and return another value of the same type; a pointer to the appropriate ©twice© function will be passed as an additional implicit parameter to the call to ©four_times©.
    137133
    138134Monomorphic specializations of polymorphic functions can themselves be used to satisfy type assertions.
    139 For instance, ©twice© could have been defined using the \CFA syntax for operator overloading as:
    140 \begin{lstlisting}
    141 forall(otype S | { ®S ?+?(S, S);® })
     135For instance, ©twice© could have been defined as below, using the \CFA syntax for operator overloading:
     136\begin{lstlisting}
     137forall(otype S | { S ?+?(S, S); })
    142138S twice(S x) { return x + x; }  // (2)
    143139\end{lstlisting}
    144 This version of ©twice© works for any type ©S© that has an addition operator defined for it, and it could have been used to satisfy the type assertion on ©four_times©.
     140This version of ©twice© will work for any type ©S© that has an addition operator defined for it, and it could have been used to satisfy the type assertion on ©four_times©.
    145141The compiler accomplishes this by creating a wrapper function calling ©twice // (2)© with ©S© bound to ©double©, then providing this wrapper function to ©four_times©\footnote{©twice // (2)© could also have had a type parameter named ©T©; \CFA specifies renaming of the type parameters, which would avoid the name conflict with the type variable ©T© of ©four_times©.}.
    146142
    147143Finding appropriate functions to satisfy type assertions is essentially a recursive case of expression resolution, as it takes a name (that of the type assertion) and attempts to match it to a suitable declaration in the current scope.
    148 If a polymorphic function can be used to satisfy one of its own type assertions, this recursion may not terminate, as it is possible that function is examined as a candidate for its own type assertion unboundedly repeatedly.
    149 To avoid infinite loops, the current CFA compiler imposes a fixed limit on the possible depth of recursion, similar to that employed by most \CC compilers for template expansion; this restriction means that there are some semantically well-typed expressions that cannot be resolved by CFA.
    150 One area of potential improvement this project proposes to investigate is the possibility of using the compiler's knowledge of the current set of declarations to more precicely determine when further type assertion satisfaction recursion does not produce a well-typed expression.
     144If a polymorphic function can be used to satisfy one of its own type assertions, this recursion may not terminate, as it is possible that function will be examined as a candidate for its own type assertion unboundedly repeatedly.
     145To avoid infinite loops, the current CFA compiler imposes a fixed limit on the possible depth of recursion, similar to that employed by most \CC compilers for template expansion; this restriction means that there are some semantically well-typed expressions which cannot be resolved by CFA.
     146One area of potential improvement this project proposes to investigate is the possibility of using the compiler's knowledge of the current set of declarations to more precicely determine when further type assertion satisfaction recursion will not produce a well-typed expression.
    151147
    152148\subsubsection{Traits}
    153149\CFA provides \emph{traits} as a means to name a group of type assertions, as in the example below:
    154150\begin{lstlisting}
    155 ®trait has_magnitude(otype T)® {
     151trait has_magnitude(otype T) {
    156152    bool ?<?(T, T);        // comparison operator for T
    157153    T -?(T);               // negation operator for T
    … …  
    172168\end{lstlisting}
    173169
    174 Semantically, traits are simply a named lists of type assertions, but they may be used for many of the same purposes that interfaces in Java or abstract base classes in \CC are used for.
     170Semantically, a trait is merely a named list of type assertions, but they can be used in many of the same situations where an interface in Java or an abstract base class in \CC would be used.
    175171Unlike Java interfaces or \CC base classes, \CFA types do not explicitly state any inheritance relationship to traits they satisfy; this can be considered a form of structural inheritance, similar to interface implementation in Go, as opposed to the nominal inheritance model of Java and \CC.
    176 Nominal inheritance can be simulated with traits using marker variables or functions:
    177 \begin{lstlisting}
    178 trait nominal(otype T) {
    179     ®T is_nominal;®
    180 };
    181 
    182 int is_nominal;  // int now satisfies the nominal trait
    183 {
    184     char is_nominal; // char satisfies the nominal trait
    185 }
    186 // char no longer satisfies the nominal trait here 
    187 \end{lstlisting}
    188 
    189 Traits, however, are significantly more powerful than nominal-inheritance interfaces; firstly, due to the scoping rules of the declarations which satisfy a trait's type assertions, a type may not satisfy a trait everywhere that the type is declared, as with ©char© and the ©nominal© trait above.
    190 Secondly, traits may be used to declare a relationship between multiple types, a property which may be difficult or impossible to represent in nominal-inheritance type systems:
    191 \begin{lstlisting}
    192 trait pointer_like(®otype Ptr, otype El®) {
    193     lvalue El *?(Ptr); // Ptr can be dereferenced into a modifiable value of type El
    194 }
    195 
    196 struct list {
    197     int value;
    198     list *next;  // may omit "struct" on type names
    199 };
    200 
    201 typedef list* list_iterator;
    202 
    203 lvalue int *?( list_iterator it ) {
    204     return it->value;
    205 }
    206 \end{lstlisting}
    207 
    208 In the example above, ©(list_iterator, int)© satisfies ©pointer_like© by the given function, and ©(list_iterator, list)© also satisfies ©pointer_like© by the built-in pointer dereference operator.
    209 While a nominal-inheritance system with associated types could model one of those two relationships by making ©El© an associated type of ©Ptr© in the ©pointer_like© implementation, few such systems could model both relationships simultaneously.
    210 
    211 The flexibility of \CFA's implicit trait satisfaction mechanism provides user programmers with a great deal of power, but also blocks some optimization approaches for expression resolution.
    212 The ability of types to begin to or cease to satisfy traits when declarations go into or out of scope makes caching of trait satisfaction judgements difficult, and the ability of traits to take multiple type parameters could lead to a combinatorial explosion of work in any attempt to pre-compute trait satisfaction relationships.
    213 On the other hand, the addition of a nominal inheritance mechanism to \CFA's type system or replacement of \CFA's trait satisfaction system with a more object-oriented inheritance model and investigation of possible expression resolution optimizations for such a system may be an interesting avenue of further research.
     172% TODO talk about modelling of nominal inheritance with structural inheritance, possibility of investigating some resolver algorithms that require nominal
    214173
    215174\subsection{Name Overloading}
    216 In C, no more than one variable or function in the same scope may share the same name\footnote{Technically, C has multiple separated namespaces, one holding ©struct©, ©union©, and ©enum© tags, one holding labels, one holding typedef names, variable, function, and enumerator identifiers, and one for each ©struct© or ©union© type holding the field names.}, and variable or function declarations in inner scopes with the same name as a declaration in an outer scope hide the outer declaration.
    217 This makes finding the proper declaration to match to a variable expression or function application a simple matter of symbol table lookup, which can be easily and efficiently implemented.
    218 \CFA, on the other hand, allows overloading of variable and function names, so long as the overloaded declarations do not have the same type, avoiding the multiplication of variable and function names for different types common in the C standard library, as in the following example:
    219 \begin{lstlisting}
    220 #include <limits.h>
    221 
    222 int max(int a, int b) { return a < b ? b : a; }  // (1)
    223 double max(double a, double b) { return a < b ? b : a; }  // (2)
    224 
    225 int max = INT_MAX;     // (3)
    226 double max = DBL_MAX;  // (4)
    227 
    228 max(7, -max);   // uses (1) and (3), by matching int type of 7
    229 max(max, 3.14); // uses (2) and (4), by matching double type of 3.14
    230 
    231 max(max, -max);  // ERROR: ambiguous
    232 int m = max(max, -max); // uses (1) and (3) twice, by return type
    233 \end{lstlisting}
    234 
    235 The presence of name overloading in \CFA means that simple table lookup is insufficient to match identifiers to declarations, and a type matching algorithm must be part of expression resolution.
     175In C, no more than one function or variable in the same scope may share the same name, and function or variable declarations in inner scopes with the same name as a declaration in an outer scope hide the outer declaration. 
     176This makes finding the proper declaration to match to a function application or variable expression a simple matter of symbol table lookup, which can be easily and efficiently implemented.
     177\CFA, on the other hand, allows overloading of variable and function names, so long as the overloaded declarations do not have the same type, avoiding the multiplication of function names for different types common in the C standard library, as in the following example:
     178\begin{lstlisting}
     179int three = 3;
     180double three = 3.0;
     181
     182int thrice(int i) { return i * three; } // uses int three
     183double thrice(double d) { return d * three; } // uses double three
     184
     185// thrice(three); // ERROR: ambiguous
     186int nine = thrice(three);    // uses int thrice and three, based on return type
     187double nine = thrice(three); // uses double thrice and three, based on return type
     188\end{lstlisting}
     189
     190The presence of name overloading in \CFA means that simple table lookup is not sufficient to match identifiers to declarations, and a type matching algorithm must be part of expression resolution.
    236191
    237192\subsection{Implicit Conversions}
    … …  
    239194C does not have a traditionally-defined inheritance hierarchy of types, but the C standard's rules for the ``usual arithmetic conversions'' define which of the built-in types are implicitly convertable to which other types, and the relative cost of any pair of such conversions from a single source type.
    240195\CFA adds to the usual arithmetic conversions rules for determining the cost of binding a polymorphic type variable in a function call; such bindings are cheaper than any \emph{unsafe} (narrowing) conversion, \eg ©int© to ©char©, but more expensive than any \emph{safe} (widening) conversion, \eg ©int© to ©double©.
    241 
    242196The expression resolution problem, then, is to find the unique minimal-cost interpretation of each expression in the program, where all identifiers must be matched to a declaration and implicit conversions or polymorphic bindings of the result of an expression may increase the cost of the expression.
    243 Note that which subexpression interpretation is minimal-cost may require contextual information to disambiguate.
    244 For instance, in the example in the previous subsection, ©max(max, -max)© cannot be unambiguously resolved, but ©int m = max(max, -max);© has a single minimal-cost resolution.
    245 ©int m = (int)max((double)max, -(double)max)© is also be a valid interpretation, but is not minimal-cost due to the unsafe cast from the ©double© result of ©max© to ©int© (the two ©double© casts function as type ascriptions selecting ©double max© rather than casts from ©int max© to ©double©, and as such are zero-cost).
     197Note that which subexpression interpretation is minimal-cost may require contextual information to disambiguate.
    246198
    247199\subsubsection{User-generated Implicit Conversions}
    … …  
    249201Such a conversion system should be simple for user programmers to utilize, and fit naturally with the existing design of implicit conversions in C; ideally it would also be sufficiently powerful to encode C's usual arithmetic conversions itself, so that \CFA only has one set of rules for conversions.
    250202
    251 Ditchfield~\cite{Ditchfield:conversions} has laid out a framework for using polymorphic-conversion-constructor functions to create a directed acyclic graph (DAG) of conversions.
     203Ditchfield\cite{Ditchfield:conversions} has laid out a framework for using polymorphic conversion constructor functions to create a directed acyclic graph (DAG) of conversions.
    252204A monomorphic variant of these functions can be used to mark a conversion arc in the DAG as only usable as the final step in a conversion.
    253205With these two types of conversion arcs, separate DAGs can be created for the safe and the unsafe conversions, and conversion cost can be represented as path length through the DAG.
    254 \begin{figure}[h]
    255 \centering
    256 \includegraphics{conversion_dag}
    257 \caption{A portion of the implicit conversion DAG for built-in types.}
    258 \end{figure}
    259 As can be seen in the example DAG above, there are either safe or unsafe paths between each of the arithmetic types listed; the ``final'' arcs are important both to avoid creating cycles in the signed-unsigned conversions, and to disambiguate potential diamond conversions (\eg, if the ©int© to ©unsigned int© conversion was not marked final there would be two length-two paths from ©int© to ©unsigned long©, and it would be impossible to choose which one; however, since the ©unsigned int© to ©unsigned long© arc can not be traversed after the final ©int© to ©unsigned int© arc, there is a single unambiguous conversion path from ©int© to ©unsigned long©).
    260 
    261 Open research questions on this topic include:
    262 \begin{itemize}
    263 \item Can a conversion graph be generated that represents each allowable conversion in C with a unique minimal-length path such that the path lengths accurately represent the relative costs of the conversions?
    264 \item Can such a graph representation can be usefully augmented to include user-defined types as well as built-in types?
    265 \item Can the graph can be efficiently represented and used in the expression resolver?
    266 \end{itemize}
     206Open research questions on this topic include whether a conversion graph can be generated that represents each allowable conversion in C with a unique minimal-length path, such that the path lengths accurately represent the relative costs of the conversions, whether such a graph representation can be usefully augmented to include user-defined types as well as built-in types, and whether the graph can be efficiently represented and included in the expression resolver.
    267207
    268208\subsection{Constructors and Destructors}
    269209Rob Shluntz, a current member of the \CFA research team, has added constructors and destructors to \CFA.
    270 Each type has an overridable default-generated zero-argument constructor, copy constructor, assignment operator, and destructor; for ©struct© types these functions each call their equivalents on each field of the ©struct©.
     210Each type has an overridable default-generated zero-argument constructor, copy constructor, assignment operator, and destructor; for struct types these functions each call their equivalents on each field of the struct.
    271211This affects expression resolution because an ©otype© type variable ©T© implicitly adds four type assertions, one for each of these four functions, so assertion resolution is pervasive in \CFA polymorphic functions, even those without any explicit type assertions.
    272 The following example shows the implicitly-generated code in green:
    273 \begin{lstlisting}
    274 struct kv {
    275     int key;
    276     char *value;
    277 };
    278 
    279 ¢void ?{}(kv *this) {
    280     ?{}(&this->key);
    281     ?{}(&this->value);
    282 }
    283 void ?{}(kv *this, kv that) {
    284     ?{}(&this->key, that.key);
    285     ?{}(&this->value, that.value);
    286 }
    287 kv ?=?(kv *this, kv that) {
    288     ?=?(&this->key, that.key);
    289     ?=?(&this->value, that.value);
    290     return *this;
    291 }
    292 void ^?{}(kv *this) {
    293     ^?{}(&this->key);
    294     ^?{}(&this->value);
    295 }¢
    296 
    297 forall(otype T ¢| { void ?{}(T*); void ?{}(T*, T); T ?=?(T*, T); void ^?{}(T*); }¢)
    298 void foo(T);
    299 \end{lstlisting}
    300212
    301213\subsection{Generic Types}
    302214I have already added a generic type capability to \CFA, designed to efficiently and naturally integrate with \CFA's existing polymorphic functions.
    303 A generic type can be declared by placing a ©forall© specifier on a ©struct© or ©union© declaration, and instantiated using a parenthesized list of types after the type name:
     215A generic type can be declared by placing a ©forall© specifier on a struct or union declaration, and instantiated using a parenthesized list of types after the type name:
    304216\begin{lstlisting}
    305217forall(otype R, otype S) struct pair {
    … …  
    317229The default-generated constructors, destructor and assignment operator for a generic type are polymorphic functions with the same list of type parameters as the generic type definition.
    318230
    319 Aside from giving users the ability to create more parameterized types than just the built-in pointer, array and function types, the combination of generic types with polymorphic functions and implicit conversions makes the edge case where the resolver may enter an infinite loop much more common, as in the following code example:
    320 \begin{lstlisting}
    321 forall(otype T) struct box { T x; };
    322 
    323 void f(void*); // (1)
    324 
    325 forall(otype S)
    326 void f(box(S)* b) { // (2)
    327         f(®(void*)0®);
    328 }
    329 \end{lstlisting}
    330 
    331 The loop in the resolver happens as follows:
     231Aside from giving users the ability to create more parameterized types than just the built-in pointer, array and function types, the combination of generic types with polymorphic functions and implicit conversions makes the edge case where a polymorphic function can match its own assertions much more common, as follows:
    332232\begin{itemize}
    333 \item Since there is an implicit conversion from ©void*© to any pointer type, the highlighted expression can be interpreted as either a ©void*©, matching ©f // (1)©, or a ©box(S)*© for some type ©S©, matching ©f // (2)©.
    334 \item To determine the cost of the ©box(S)© interpretation, a type must be found for ©S© which satisfies the ©otype© implicit type assertions (assignment operator, default and copy constructors, and destructor); one option is ©box(S2)© for some type ©S2©.
    335 \item The assignment operator, default and copy constructors, and destructor of ©box(T)© are also polymorphic functions, each of which require the type parameter ©T© to have an assignment operator, default and copy constructors, and destructor. When choosing an interpretation for ©S2©, one option is ©box(S3)©, for some type ©S3©.
    336 \item The previous step repeats until stopped, with four times as much work performed at each step.
     233\item Given an expression in an untyped context, such as a top-level function call with no assignment of return values, apply a polymorphic implicit conversion to the expression that can produce multiple types (the built-in conversion from ©void*© to any other pointer type is one, but not the only).
     234\item When attempting to use a generic type with ©otype© parameters (such as ©box© above) for the result type of the expression, the resolver will also need to decide what type to use for the ©otype© parameters on the constructors and related functions, and will have no constraints on what they may be.
     235\item Attempting to match some yet-to-be-determined specialization of the generic type to this ©otype© parameter will create a recursive case of the default constructor, \etc matching their own type assertions, creating an unboundedly deep nesting of the generic type inside itself.
    337236\end{itemize}
    338 This problem can occur in any resolution context where a polymorphic function that can satisfy its own type assertions is required for a possible interpretation of an expression with no constraints on its type, and is thus not limited to combinations of generic types with ©void*© conversions, though constructors for generic types often satisfy their own assertions and a polymorphic conversion such as the ©void*© conversion to a polymorphic variable can create an expression with no constraints on its type.
    339 As discussed above, the \CFA expression resolver must handle this possible infinite recursion somehow, and it occurs fairly naturally in code like the above that uses generic types.
     237As discussed above, any \CFA expression resolver must handle this possible infinite recursion somehow, but the combination of generic types with other language features makes this particular edge case occur somewhat frequently in user code.
    340238
    341239\subsection{Tuple Types}
    342 \CFA adds \emph{tuple types} to C, a syntactic facility for referring to lists of values anonymously or with a single identifier.
    343 An identifier may name a tuple, and a function may return one.
     240\CFA adds \emph{tuple types} to C, a facility for referring to multiple values with a single identifier.
     241A variable may name a tuple, and a function may return one.
    344242Particularly relevantly for resolution, a tuple may be implicitly \emph{destructured} into a list of values, as in the call to ©swap© below:
    345243\begin{lstlisting}
    … …  
    350248
    351249x = swap( x ); // destructure [char, char] x into two elements of parameter list
    352 // cannot use int x for parameter, not enough arguments to swap
     250// can't use int x for parameter, not enough arguments to swap
    353251\end{lstlisting}
    354252Tuple destructuring means that the mapping from the position of a subexpression in the argument list to the position of a paramter in the function declaration is not straightforward, as some arguments may be expandable to different numbers of parameters, like ©x© above.
    … …  
    358256Given some type ©T©, a ©T&© (``reference to ©T©'') is essentially an automatically dereferenced pointer; with these semantics most of the C standard's discussions of lvalues can be expressed in terms of references instead, with the benefit of being able to express the difference between the reference and non-reference version of a type in user code.
    359257References preserve C's existing qualifier-dropping lvalue-to-rvalue conversion (\eg a ©const volatile int&© can be implicitly converted to a bare ©int©); the reference proposal also adds a rvalue-to-lvalue conversion to \CFA, implemented by storing the value in a new compiler-generated temporary and passing a reference to the temporary.
    360 These two conversions can chain, producing a qualifier-dropping conversion for references, for instance converting a reference to a ©const int© into a reference to a non-©const int© by copying the originally refered to value into a fresh temporary and taking a reference to this temporary, as below:
    361 \begin{lstlisting}
    362 const int magic = 42;
    363 
    364 void inc_print( int& x ) { printf("%d\n", ++x); }
    365 
    366 print_inc( magic ); // legal; implicitly generated code in green below:
    367 
    368 ¢int tmp = magic;¢ // copies to safely strip const-qualifier
    369 ¢print_inc( tmp );¢ // tmp is incremented, magic is unchanged
    370 \end{lstlisting}
     258These two conversions can chain, producing a qualifier-dropping conversion for references, for instance converting a reference to a ©const int© into a reference to a non-©const int© by copying the originally refered to value into a fresh temporary and taking a reference to this temporary.
    371259These reference conversions may also chain with the other implicit type conversions.
    372260The main implication of this for expression resolution is the multiplication of available implicit conversions, though in a restricted context that may be able to be treated efficiently as a special case.
    373261
    374 \subsection{Special Literal Types}
    375 Another proposal currently under consideration for the \CFA type-system is assigning special types to the literal values ©0© and ©1©.
    376 Implicit conversions from these types allow ©0© and ©1© to be considered as values of many different types, depending on context, allowing expression desugarings like ©if ( x ) {}© $\Rightarrow$ ©if ( x != 0 ) {}© to be implemented efficiently and precicely.
    377 This approach is a generalization of C's existing behaviour of treating ©0© as either an integer zero or a null pointer constant, and treating either of those values as boolean false.
    378 The main implication for expression resolution is that the frequently encountered expressions ©0© and ©1© may have a large number of valid interpretations.
     262\subsection{Literal Types}
     263Another proposal currently under consideration for the \CFA type-system is assigning special types to the literal values ©0© and ©1©.%, say ©zero_t© and ©one_t©.
     264Implicit conversions from these types would allow ©0© and ©1© to be considered as values of many different types, depending on context, allowing expression desugarings like ©if ( x ) {}© $\Rightarrow$ ©if ( x != 0 ) {}© to be implemented efficiently and precicely.
     265This is a generalization of C's existing behaviour of treating ©0© as either an integer zero or a null pointer constant, and treating either of those values as boolean false.
     266The main implication for expression resolution is that the frequently encountered expressions ©0© and ©1© may have a significant number of valid interpretations.
    379267
    380268\subsection{Deleted Function Declarations}
    … …  
    383271int somefn(char) = delete;
    384272\end{lstlisting}
    385 This feature is typically used in \CCeleven to make a type non-copyable by deleting its copy constructor and assignment operator, or forbidding some interpretations of a polymorphic function by specifically deleting the forbidden overloads.
    386273To add a similar feature to \CFA would involve including the deleted function declarations in expression resolution along with the normal declarations, but producing a compiler error if the deleted function was the best resolution.
    387274How conflicts should be handled between resolution of an expression to both a deleted and a non-deleted function is a small but open research question.
    388275
    389276\section{Expression Resolution}
    390 The expression resolution problem is determining an optimal match between some combination of argument interpretations and the parameter list of some overloaded instance of a function; the argument interpretations are produced by recursive invocations of expression resolution, where the base case is zero-argument functions (which are, for purposes of this discussion, semantically equivalent to named variables or constant literal expressions).
    391 Assuming that the matching between a function's parameter list and a combination of argument interpretations can be done in $\bigO{p^k}$ time, where $p$ is the number of parameters and $k$ is some positive number, if there are $\bigO{i}$ valid interpretations for each subexpression, there will be $\bigO{i}$ candidate functions and $\bigO{i^p}$ possible argument combinations for each expression, so for a single recursive call expression resolution takes $\bigO{i^{p+1} \cdot p^k}$ time if it must compare all combinations, or $\bigO{i(p+1) \cdot p^k}$ time if argument-parameter matches can be chosen independently of each other.
    392 Given these bounds, resolution of a single top-level expression tree of depth $d$ takes $\bigO{i^{p+1} \cdot p^{k \cdot d}}$ time under full-combination matching, or $\bigO{i(p+1) \cdot p^{k \cdot d}}$ time for independent-parameter matching\footnote{A call tree has leaves at depth $\bigO{d}$, and each internal node has $\bigO{p}$ fan-out, producing $\bigO{p^d}$ total recursive calls.}.
    393 
    394 Expression resolution is somewhat unavoidably exponential in $d$, the depth of the expression tree, and if arguments cannot be matched to parameters independently of each other, expression resolution is also exponential in $p$.
    395 However, both $d$ and $p$ are fixed by the user programmer, and generally bounded by reasonably small constants.
    396 $k$, on the other hand, is mostly dependent on the representation of types in the system and the efficiency of type assertion checking; if a candidate argument combination can be compared to a function parameter list in linear time in the length of the list (\ie $k = 1$), then the $p^{k \cdot d}$ factor is linear in the input size of the source code for the expression, otherwise the resolution algorithm exibits sub-linear performance scaling on code containing more-deeply nested expressions.
     277The expression resolution problem is essentially to determine an optimal matching between some combination of argument interpretations and the parameter list of some overloaded instance of a function; the argument interpretations are produced by recursive invocations of expression resolution, where the base case is zero-argument functions (which are, for purposes of this discussion, semantically equivalent to named variables or constant literal expressions).
     278Assuming that the matching between a function's parameter list and a combination of argument interpretations can be done in $O(p^k)$ time, where $p$ is the number of parameters and $k$ is some positive number, if there are $O(i)$ valid interpretations for each subexpression, there will be $O(i)$ candidate functions and $O(i^p)$ possible argument combinations for each expression, so a single recursive call to expression resolution will take $O(i^{p+1} \cdot p^k)$ time if it compares all combinations.
     279Given this bound, resolution of a single top-level expression tree of depth $d$ takes $O(i^{p+1} \cdot p^{k \cdot d})$ time\footnote{The call tree will have leaves at depth $O(d)$, and each internal node will have $O(p)$ fan-out, producing $O(p^d)$ total recursive calls.}.
     280Expression resolution is somewhat unavoidably exponential in $p$, the number of function parameters, and $d$, the depth of the expression tree, but these values are fixed by the user programmer, and generally bounded by reasonably small constants.
     281$k$, on the other hand, is mostly dependent on the representation of types in the system and the efficiency of type assertion checking; if a candidate argument combination can be compared to a function parameter list in linear time in the length of the list (\ie $k = 1$), then the $p^{k \cdot d}$ term is linear in the input size of the source code for the expression, otherwise the resolution algorithm will exibit sub-linear performance scaling on code containing more-deeply nested expressions.
    397282The number of valid interpretations of any subexpression, $i$, is bounded by the number of types in the system, which is possibly infinite, though practical resolution algorithms for \CFA must be able to place some finite bound on $i$, possibly at the expense of type-system completeness.
    398283
    399 The research goal of this project is to develop a performant expression resolver for \CFA; this analysis suggests three primary areas of investigation to accomplish that end.
    400 The first area of investigation is efficient argument-parameter matching; Bilson~\cite{Bilson03} mentions significant optimization opportunities available in the current literature to improve on the existing CFA compiler.
     284The research goal of this project is to develop a performant expression resolver for \CFA; this analysis suggests two primary areas of investigation to accomplish that end.
     285The first is efficient argument-parameter matching; Bilson\cite{Bilson03} mentions significant optimization opportunities available in the current literature to improve on the existing CFA compiler.
    401286%TODO: look up and lit review
    402 The second area of investigation is minimizing dependencies between argument-parameter matches; the current CFA compiler attempts to match entire argument combinations against functions at once, potentially attempting to match the same argument against the same parameter multiple times.
    403 Whether the feature set of \CFA admits an expression resolution algorithm where arguments can be matched to parameters independently of other arguments in the same function application is an area of open research; polymorphic type paramters produce enough of a cross-argument dependency that the problem is not trivial.
    404 If cross-argument resolution dependencies cannot be completely eliminated, effective caching strategies to reduce duplicated work between equivalent argument-parameter matches in different combinations may mitigate the asymptotic defecits of the whole-combination matching approach.
    405 The final area of investigation is heuristics and algorithmic approaches to reduce the number of argument interpretations considered in the common case; if argument-parameter matches cannot be made independent, even small reductions in $i$ should yield significant reductions in the $i^{p+1}$ resolver runtime factor.
     287The second, and likely more fruitful, area of investigation is heuristics and algorithmic approaches to reduce the number of argument interpretations considered in the common case; given the large ($p+1$) exponent on number of interpretations considered in the runtime analysis, even small reductions here could have a significant effect on overall resolver runtime.
    406288The discussion below presents a number of largely orthagonal axes for expression resolution algorithm design to be investigated, noting prior work where applicable.
    407289

  • src/GenPoly/Box.cc

    ref42e764 re109716  
    104104                        Type *replaceWithConcrete( ApplicationExpr *appExpr, Type *type, bool doClone = true );
    105105                        /// wraps a function application returning a polymorphic type with a new temporary for the out-parameter return value
    106                         Expression *addDynRetParam( ApplicationExpr *appExpr, FunctionType *function, ReferenceToType *polyType, std::list< Expression *>::iterator &arg );
     106                        Expression *addPolyRetParam( ApplicationExpr *appExpr, FunctionType *function, ReferenceToType *polyType, std::list< Expression *>::iterator &arg );
    107107                        Expression *applyAdapter( ApplicationExpr *appExpr, FunctionType *function, std::list< Expression *>::iterator &arg, const TyVarMap &exprTyVars );
    108108                        void boxParam( Type *formal, Expression *&arg, const TyVarMap &exprTyVars );
    … …  
    661661                                // process polymorphic return value
    662662                                retval = 0;
    663                                 if ( isDynRet( functionDecl->get_functionType() ) && functionDecl->get_linkage() == LinkageSpec::Cforall ) {
     663                                if ( isPolyRet( functionDecl->get_functionType() ) && functionDecl->get_linkage() == LinkageSpec::Cforall ) {
    664664                                        retval = functionDecl->get_functionType()->get_returnVals().front();
    665665
    … …  
    868868                }
    869869
    870                 Expression *Pass1::addDynRetParam( ApplicationExpr *appExpr, FunctionType *function, ReferenceToType *dynType, std::list< Expression *>::iterator &arg ) {
     870                Expression *Pass1::addPolyRetParam( ApplicationExpr *appExpr, FunctionType *function, ReferenceToType *polyType, std::list< Expression *>::iterator &arg ) {
    871871                        assert( env );
    872                         Type *concrete = replaceWithConcrete( appExpr, dynType );
     872                        Type *concrete = replaceWithConcrete( appExpr, polyType );
    873873                        // add out-parameter for return value
    874874                        return addRetParam( appExpr, function, concrete, arg );
    … …  
    877877                Expression *Pass1::applyAdapter( ApplicationExpr *appExpr, FunctionType *function, std::list< Expression *>::iterator &arg, const TyVarMap &tyVars ) {
    878878                        Expression *ret = appExpr;
    879 //                      if ( ! function->get_returnVals().empty() && isPolyType( function->get_returnVals().front()->get_type(), tyVars ) ) {
    880                         if ( isDynRet( function, tyVars ) ) {
     879                        if ( ! function->get_returnVals().empty() && isPolyType( function->get_returnVals().front()->get_type(), tyVars ) ) {
    881880                                ret = addRetParam( appExpr, function, function->get_returnVals().front()->get_type(), arg );
    882881                        } // if
    … …  
    969968                        // actually make the adapter type
    970969                        FunctionType *adapter = adaptee->clone();
    971 //                      if ( ! adapter->get_returnVals().empty() && isPolyType( adapter->get_returnVals().front()->get_type(), tyVars ) ) {
    972                         if ( isDynRet( adapter, tyVars ) ) {
     970                        if ( ! adapter->get_returnVals().empty() && isPolyType( adapter->get_returnVals().front()->get_type(), tyVars ) ) {
    973971                                makeRetParm( adapter );
    974972                        } // if
    … …  
    10321030                                addAdapterParams( adapteeApp, arg, param, adapterType->get_parameters().end(), realParam, tyVars );
    10331031                                bodyStmt = new ExprStmt( noLabels, adapteeApp );
    1034 //                      } else if ( isPolyType( adaptee->get_returnVals().front()->get_type(), tyVars ) ) {
    1035                         } else if ( isDynType( adaptee->get_returnVals().front()->get_type(), tyVars ) ) {
     1032                        } else if ( isPolyType( adaptee->get_returnVals().front()->get_type(), tyVars ) ) {
    10361033                                // return type T
    10371034                                if ( (*param)->get_name() == "" ) {
    … …  
    12801277                        TyVarMap exprTyVars( (TypeDecl::Kind)-1 );
    12811278                        makeTyVarMap( function, exprTyVars );
    1282                         ReferenceToType *dynRetType = isDynRet( function, exprTyVars );
    1283 
    1284                         if ( dynRetType ) {
    1285                                 ret = addDynRetParam( appExpr, function, dynRetType, arg );
     1279                        ReferenceToType *polyRetType = isPolyRet( function );
     1280
     1281                        if ( polyRetType ) {
     1282                                ret = addPolyRetParam( appExpr, function, polyRetType, arg );
    12861283                        } else if ( needsAdapter( function, scopeTyVars ) ) {
    12871284                                // std::cerr << "needs adapter: ";
    … …  
    12931290                        arg = appExpr->get_args().begin();
    12941291
    1295                         passTypeVars( appExpr, dynRetType, arg, exprTyVars );
     1292                        passTypeVars( appExpr, polyRetType, arg, exprTyVars );
    12961293                        addInferredParams( appExpr, function, arg, exprTyVars );
    12971294
    … …  
    15801577
    15811578                        // move polymorphic return type to parameter list
    1582                         if ( isDynRet( funcType ) ) {
     1579                        if ( isPolyRet( funcType ) ) {
    15831580                                DeclarationWithType *ret = funcType->get_returnVals().front();
    15841581                                ret->set_type( new PointerType( Type::Qualifiers(), ret->get_type() ) );

  • src/GenPoly/GenPoly.cc

    ref42e764 re109716  
    2323
    2424namespace GenPoly {
     25        bool needsAdapter( FunctionType *adaptee, const TyVarMap &tyVars ) {
     26                if ( ! adaptee->get_returnVals().empty() && isPolyType( adaptee->get_returnVals().front()->get_type(), tyVars ) ) {
     27                        return true;
     28                } // if
     29                for ( std::list< DeclarationWithType* >::const_iterator innerArg = adaptee->get_parameters().begin(); innerArg != adaptee->get_parameters().end(); ++innerArg ) {
     30                        if ( isPolyType( (*innerArg)->get_type(), tyVars ) ) {
     31                                return true;
     32                        } // if
     33                } // for
     34                return false;
     35        }
     36
     37        ReferenceToType *isPolyRet( FunctionType *function ) {
     38                if ( ! function->get_returnVals().empty() ) {
     39                        TyVarMap forallTypes( (TypeDecl::Kind)-1 );
     40                        makeTyVarMap( function, forallTypes );
     41                        return (ReferenceToType*)isPolyType( function->get_returnVals().front()->get_type(), forallTypes );
     42                } // if
     43                return 0;
     44        }
     45
    2546        namespace {
    2647                /// Checks a parameter list for polymorphic parameters; will substitute according to env if present
    … …  
    4364                        return false;
    4465                }
    45 
    46                 /// Checks a parameter list for dynamic-layout parameters from tyVars; will substitute according to env if present
    47                 bool hasDynParams( std::list< Expression* >& params, const TyVarMap &tyVars, const TypeSubstitution *env ) {
    48                         for ( std::list< Expression* >::iterator param = params.begin(); param != params.end(); ++param ) {
    49                                 TypeExpr *paramType = dynamic_cast< TypeExpr* >( *param );
    50                                 assert(paramType && "Aggregate parameters should be type expressions");
    51                                 if ( isDynType( paramType->get_type(), tyVars, env ) ) return true;
    52                         }
    53                         return false;
    54                 }
    5566        }
    5667
    … …  
    90101                }
    91102                return 0;
    92         }
    93 
    94         Type *isDynType( Type *type, const TyVarMap &tyVars, const TypeSubstitution *env ) {
    95                 type = replaceTypeInst( type, env );
    96 
    97                 if ( TypeInstType *typeInst = dynamic_cast< TypeInstType * >( type ) ) {
    98                         auto var = tyVars.find( typeInst->get_name() );
    99                         if ( var != tyVars.end() && var->second == TypeDecl::Any ) {
    100                                 return type;
    101                         }
    102                 } else if ( StructInstType *structType = dynamic_cast< StructInstType* >( type ) ) {
    103                         if ( hasDynParams( structType->get_parameters(), tyVars, env ) ) return type;
    104                 } else if ( UnionInstType *unionType = dynamic_cast< UnionInstType* >( type ) ) {
    105                         if ( hasDynParams( unionType->get_parameters(), tyVars, env ) ) return type;
    106                 }
    107                 return 0;
    108         }
    109 
    110         ReferenceToType *isDynRet( FunctionType *function, const TyVarMap &forallTypes ) {
    111                 if ( function->get_returnVals().empty() ) return 0;
    112                
    113                 return (ReferenceToType*)isDynType( function->get_returnVals().front()->get_type(), forallTypes );
    114         }
    115 
    116         ReferenceToType *isDynRet( FunctionType *function ) {
    117                 if ( function->get_returnVals().empty() ) return 0;
    118 
    119                 TyVarMap forallTypes( (TypeDecl::Kind)-1 );
    120                 makeTyVarMap( function, forallTypes );
    121                 return (ReferenceToType*)isDynType( function->get_returnVals().front()->get_type(), forallTypes );
    122         }
    123 
    124         bool needsAdapter( FunctionType *adaptee, const TyVarMap &tyVars ) {
    125 //              if ( ! adaptee->get_returnVals().empty() && isPolyType( adaptee->get_returnVals().front()->get_type(), tyVars ) ) {
    126 //                      return true;
    127 //              } // if
    128                 if ( isDynRet( adaptee, tyVars ) ) return true;
    129                
    130                 for ( std::list< DeclarationWithType* >::const_iterator innerArg = adaptee->get_parameters().begin(); innerArg != adaptee->get_parameters().end(); ++innerArg ) {
    131 //                      if ( isPolyType( (*innerArg)->get_type(), tyVars ) ) {
    132                         if ( isDynType( (*innerArg)->get_type(), tyVars ) ) {
    133                                 return true;
    134                         } // if
    135                 } // for
    136                 return false;
    137103        }
    138104

  • src/GenPoly/GenPoly.h

    ref42e764 re109716  
    3131namespace GenPoly {
    3232        typedef ErasableScopedMap< std::string, TypeDecl::Kind > TyVarMap;
     33       
     34        /// A function needs an adapter if it returns a polymorphic value or if any of its
     35        /// parameters have polymorphic type
     36        bool needsAdapter( FunctionType *adaptee, const TyVarMap &tyVarr );
     37
     38        /// true iff function has polymorphic return type
     39        ReferenceToType *isPolyRet( FunctionType *function );
    3340
    3441        /// Replaces a TypeInstType by its referrent in the environment, if applicable
    … …  
    4047        /// returns polymorphic type if is polymorphic type in tyVars, NULL otherwise; will look up substitution in env if provided
    4148        Type *isPolyType( Type *type, const TyVarMap &tyVars, const TypeSubstitution *env = 0 );
    42 
    43         /// returns dynamic-layout type if is dynamic-layout type in tyVars, NULL otherwise; will look up substitution in env if provided
    44         Type *isDynType( Type *type, const TyVarMap &tyVars, const TypeSubstitution *env = 0 );
    45 
    46         /// true iff function has dynamic-layout return type under the given type variable map
    47         ReferenceToType *isDynRet( FunctionType *function, const TyVarMap &tyVars );
    48 
    49         /// true iff function has dynamic-layout return type under the type variable map generated from its forall-parameters
    50         ReferenceToType *isDynRet( FunctionType *function );
    51 
    52         /// A function needs an adapter if it returns a dynamic-layout value or if any of its parameters have dynamic-layout type
    53         bool needsAdapter( FunctionType *adaptee, const TyVarMap &tyVarr );
    5449
    5550        /// returns polymorphic type if is pointer to polymorphic type, NULL otherwise; will look up substitution in env if provided

  • src/GenPoly/InstantiateGeneric.cc

    ref42e764 re109716  
    2424#include "GenPoly.h"
    2525#include "ScopedMap.h"
    26 #include "ScopedSet.h"
    2726
    2827#include "ResolvExpr/typeops.h"
    … …  
    123122                }
    124123        };
    125 
    126         /// Possible options for a given specialization of a generic type
    127         enum class genericType {
    128                 dtypeStatic,  ///< Concrete instantiation based solely on {d,f}type-to-void conversions
    129                 concrete,     ///< Concrete instantiation requiring at least one parameter type
    130                 dynamic       ///< No concrete instantiation
    131         };
    132 
    133         genericType& operator |= ( genericType& gt, const genericType& ht ) {
    134                 switch ( gt ) {
    135                 case genericType::dtypeStatic:
    136                         gt = ht;
    137                         break;
    138                 case genericType::concrete:
    139                         if ( ht == genericType::dynamic ) { gt = genericType::dynamic; }
    140                         break;
    141                 case genericType::dynamic:
    142                         // nothing possible
    143                         break;
    144                 }
    145                 return gt;
    146         }
    147124       
    148125        /// Mutator pass that replaces concrete instantiations of generic types with actual struct declarations, scoped appropriately
    … …  
    150127                /// Map of (generic type, parameter list) pairs to concrete type instantiations
    151128                InstantiationMap< AggregateDecl, AggregateDecl > instantiations;
    152                 /// Set of types which are dtype-only generic (and therefore have static layout)
    153                 ScopedSet< AggregateDecl* > dtypeStatics;
    154129                /// Namer for concrete types
    155130                UniqueName typeNamer;
    156131
    157132        public:
    158                 GenericInstantiator() : DeclMutator(), instantiations(), dtypeStatics(), typeNamer("_conc_") {}
     133                GenericInstantiator() : DeclMutator(), instantiations(), typeNamer("_conc_") {}
    159134
    160135                virtual Type* mutate( StructInstType *inst );
    … …  
    172147                /// Wrap instantiation insertion for unions
    173148                void insert( UnionInstType *inst, const std::list< TypeExpr* > &typeSubs, UnionDecl *decl ) { instantiations.insert( inst->get_baseUnion(), typeSubs, decl ); }
    174 
    175                 /// Strips a dtype-static aggregate decl of its type parameters, marks it as stripped
    176                 void stripDtypeParams( AggregateDecl *base, std::list< TypeDecl* >& baseParams, const std::list< TypeExpr* >& typeSubs );
    177149        };
    178150
    … …  
    182154        }
    183155
    184         /// Makes substitutions of params into baseParams; returns dtypeStatic if there is a concrete instantiation based only on {d,f}type-to-void conversions,
    185         /// concrete if there is a concrete instantiation requiring at least one parameter type, and dynamic if there is no concrete instantiation
     156        //////////////////////////////////////// GenericInstantiator //////////////////////////////////////////////////
     157
     158        /// Possible options for a given specialization of a generic type
     159        enum class genericType {
     160                dtypeStatic,  ///< Concrete instantiation based solely on {d,f}type-to-void conversions
     161                concrete,     ///< Concrete instantiation requiring at least one parameter type
     162                dynamic       ///< No concrete instantiation
     163        };
     164
     165        genericType& operator |= ( genericType& gt, const genericType& ht ) {
     166                switch ( gt ) {
     167                case genericType::dtypeStatic:
     168                        gt = ht;
     169                        break;
     170                case genericType::concrete:
     171                        if ( ht == genericType::dynamic ) { gt = genericType::dynamic; }
     172                        break;
     173                case genericType::dynamic:
     174                        // nothing possible
     175                        break;
     176                }
     177                return gt;
     178        }
     179
     180        /// Makes substitutions of params into baseParams; returns true if all parameters substituted for a concrete type
    186181        genericType makeSubstitutions( const std::list< TypeDecl* >& baseParams, const std::list< Expression* >& params, std::list< TypeExpr* >& out ) {
    187182                genericType gt = genericType::dtypeStatic;
    … …  
    228223        }
    229224
    230         /// Substitutes types of members according to baseParams => typeSubs, working in-place
    231         void substituteMembers( std::list< Declaration* >& members, const std::list< TypeDecl* >& baseParams, const std::list< TypeExpr* >& typeSubs ) {
    232                 // substitute types into new members
    233                 TypeSubstitution subs( baseParams.begin(), baseParams.end(), typeSubs.begin() );
    234                 for ( std::list< Declaration* >::iterator member = members.begin(); member != members.end(); ++member ) {
    235                         subs.apply(*member);
    236                 }
    237         }
    238 
    239         /// Strips the instances's type parameters
    240         void stripInstParams( ReferenceToType *inst ) {
    241                 deleteAll( inst->get_parameters() );
    242                 inst->get_parameters().clear();
    243         }
    244        
    245         void GenericInstantiator::stripDtypeParams( AggregateDecl *base, std::list< TypeDecl* >& baseParams, const std::list< TypeExpr* >& typeSubs ) {
    246                 substituteMembers( base->get_members(), baseParams, typeSubs );
    247 
    248                 deleteAll( baseParams );
    249                 baseParams.clear();
    250                
    251                 dtypeStatics.insert( base );
    252         }
    253 
    254225        Type* GenericInstantiator::mutate( StructInstType *inst ) {
    255226                // mutate subtypes
    … …  
    260231                // exit early if no need for further mutation
    261232                if ( inst->get_parameters().empty() ) return inst;
    262 
    263                 // check for an already-instantiatiated dtype-static type
    264                 if ( dtypeStatics.find( inst->get_baseStruct() ) != dtypeStatics.end() ) {
    265                         stripInstParams( inst );
    266                         return inst;
    267                 }
    268                
     233                assert( inst->get_baseParameters() && "Base struct has parameters" );
     234
    269235                // check if type can be concretely instantiated; put substitutions into typeSubs
    270                 assert( inst->get_baseParameters() && "Base struct has parameters" );
    271236                std::list< TypeExpr* > typeSubs;
    272237                genericType gt = makeSubstitutions( *inst->get_baseParameters(), inst->get_parameters(), typeSubs );
    273238                switch ( gt ) {
    274                 case genericType::dtypeStatic:
    275                         stripDtypeParams( inst->get_baseStruct(), *inst->get_baseParameters(), typeSubs );
    276                         stripInstParams( inst );
    277                         break;
    278                
    279                 case genericType::concrete: {
     239                case genericType::dtypeStatic: // TODO strip params off original decl and reuse here
     240                case genericType::concrete:
     241                {
    280242                        // make concrete instantiation of generic type
    281243                        StructDecl *concDecl = lookup( inst, typeSubs );
    … …  
    312274                // exit early if no need for further mutation
    313275                if ( inst->get_parameters().empty() ) return inst;
    314 
    315                 // check for an already-instantiatiated dtype-static type
    316                 if ( dtypeStatics.find( inst->get_baseUnion() ) != dtypeStatics.end() ) {
    317                         stripInstParams( inst );
    318                         return inst;
    319                 }
     276                assert( inst->get_baseParameters() && "Base union has parameters" );
    320277
    321278                // check if type can be concretely instantiated; put substitutions into typeSubs
    322                 assert( inst->get_baseParameters() && "Base union has parameters" );
    323279                std::list< TypeExpr* > typeSubs;
    324280                genericType gt = makeSubstitutions( *inst->get_baseParameters(), inst->get_parameters(), typeSubs );
    325281                switch ( gt ) {
    326                 case genericType::dtypeStatic:
    327                         stripDtypeParams( inst->get_baseUnion(), *inst->get_baseParameters(), typeSubs );
    328                         stripInstParams( inst );
    329                         break;
    330                        
     282                case genericType::dtypeStatic:  // TODO strip params off original decls and reuse here
    331283                case genericType::concrete:
    332284                {
    … …  
    359311                DeclMutator::doBeginScope();
    360312                instantiations.beginScope();
    361                 dtypeStatics.beginScope();
    362313        }
    363314
    … …  
    365316                DeclMutator::doEndScope();
    366317                instantiations.endScope();
    367                 dtypeStatics.endScope();
    368318        }
    369319

  • src/GenPoly/ScrubTyVars.cc

    ref42e764 re109716  
    4545
    4646        Type * ScrubTyVars::mutateAggregateType( Type *ty ) {
    47                 if ( shouldScrub( ty ) ) {
     47                if ( isPolyType( ty, tyVars ) ) {
    4848                        PointerType *ret = new PointerType( Type::Qualifiers(), new VoidType( ty->get_qualifiers() ) );
    4949                        delete ty;
    … …  
    6363        Expression * ScrubTyVars::mutate( SizeofExpr *szeof ) {
    6464                // sizeof( T ) => _sizeof_T parameter, which is the size of T
    65                 if ( Type *dynType = shouldScrub( szeof->get_type() ) ) {
    66                         Expression *expr = new NameExpr( sizeofName( mangleType( dynType ) ) );
     65                if ( Type *polyType = isPolyType( szeof->get_type() ) ) {
     66                        Expression *expr = new NameExpr( sizeofName( mangleType( polyType ) ) );
    6767                        return expr;
    6868                } else {
    … …  
    7373        Expression * ScrubTyVars::mutate( AlignofExpr *algnof ) {
    7474                // alignof( T ) => _alignof_T parameter, which is the alignment of T
    75                 if ( Type *dynType = shouldScrub( algnof->get_type() ) ) {
    76                         Expression *expr = new NameExpr( alignofName( mangleType( dynType ) ) );
     75                if ( Type *polyType = isPolyType( algnof->get_type() ) ) {
     76                        Expression *expr = new NameExpr( alignofName( mangleType( polyType ) ) );
    7777                        return expr;
    7878                } else {
    … …  
    8282
    8383        Type * ScrubTyVars::mutate( PointerType *pointer ) {
    84 //              // special case of shouldScrub that takes all TypeInstType pointer bases, even if they're not dynamic
    85 //              Type *base = pointer->get_base();
    86 //              Type *dynType = 0;
    87 //              if ( dynamicOnly ) {
    88 //                      if ( TypeInstType *typeInst = dynamic_cast< TypeInstType* >( base ) ) {
    89 //                              if ( tyVars.find( typeInst->get_name() ) != tyVars.end() ) { dynType = typeInst; }
    90 //                      } else {
    91 //                              dynType = isDynType( base, tyVars );
    92 //                      }
    93 //              } else {
    94 //                      dynType = isPolyType( base, tyVars );
    95 //              }
    96 //              if ( dynType ) {
    97                 if ( Type *dynType = shouldScrub( pointer->get_base() ) ) {
    98                         Type *ret = dynType->acceptMutator( *this );
     84                if ( Type *polyType = isPolyType( pointer->get_base(), tyVars ) ) {
     85                        Type *ret = polyType->acceptMutator( *this );
    9986                        ret->get_qualifiers() += pointer->get_qualifiers();
    10087                        pointer->set_base( 0 );

  • src/GenPoly/ScrubTyVars.h

    ref42e764 re109716  
    2727        class ScrubTyVars : public Mutator {
    2828          public:
    29                 ScrubTyVars( const TyVarMap &tyVars, bool dynamicOnly = false ): tyVars( tyVars ), dynamicOnly( dynamicOnly ) {}
     29                ScrubTyVars( const TyVarMap &tyVars ): tyVars( tyVars ) {}
    3030
    3131                /// For all polymorphic types with type variables in `tyVars`, replaces generic types, dtypes, and ftypes with the appropriate void type,
    … …  
    3333                template< typename SynTreeClass >
    3434                static SynTreeClass *scrub( SynTreeClass *target, const TyVarMap &tyVars );
    35 
    36                 /// For all dynamic-layout types with type variables in `tyVars`, replaces generic types, dtypes, and ftypes with the appropriate void type,
    37                 /// and sizeof/alignof expressions with the proper variable
    38                 template< typename SynTreeClass >
    39                 static SynTreeClass *scrubDynamic( SynTreeClass *target, const TyVarMap &tyVars );
    4035
    4136                virtual Type* mutate( TypeInstType *typeInst );
    … …  
    4742
    4843          private:
    49                 /// Returns the type if it should be scrubbed, NULL otherwise.
    50                 Type* shouldScrub( Type *ty ) {
    51                         return dynamicOnly ? isDynType( ty, tyVars ) : isPolyType( ty, tyVars );
    52 //                      if ( ! dynamicOnly ) return isPolyType( ty, tyVars );
    53 //
    54 //                      if ( TypeInstType *typeInst = dynamic_cast< TypeInstType* >( ty ) ) {
    55 //                              return tyVars.find( typeInst->get_name() ) != tyVars.end() ? ty : 0;
    56 //                      }
    57 //
    58 //                      return isDynType( ty, tyVars );
    59                 }
    60                
    6144                /// Mutates (possibly generic) aggregate types appropriately
    6245                Type* mutateAggregateType( Type *ty );
    6346               
    64                 const TyVarMap &tyVars;  ///< Type variables to scrub
    65                 bool dynamicOnly;        ///< only scrub the types with dynamic layout? [false]
     47                const TyVarMap &tyVars;
    6648        };
    6749
     50        /* static class method */
    6851        template< typename SynTreeClass >
    6952        SynTreeClass * ScrubTyVars::scrub( SynTreeClass *target, const TyVarMap &tyVars ) {
    7053                ScrubTyVars scrubber( tyVars );
    71                 return static_cast< SynTreeClass * >( target->acceptMutator( scrubber ) );
    72         }
    73 
    74         template< typename SynTreeClass >
    75         SynTreeClass * ScrubTyVars::scrubDynamic( SynTreeClass *target, const TyVarMap &tyVars ) {
    76                 ScrubTyVars scrubber( tyVars, true );
    7754                return static_cast< SynTreeClass * >( target->acceptMutator( scrubber ) );
    7855        }

  • src/SymTab/Autogen.cc

    ref42e764 re109716  
    174174
    175175        void makeStructMemberOp( ObjectDecl * dstParam, Expression * src, DeclarationWithType * field, FunctionDecl * func, TypeSubstitution & genericSubs, bool isDynamicLayout, bool forward = true ) {
    176 //              if ( isDynamicLayout && src ) {
    177 //                      genericSubs.apply( src );
    178 //              }
     176                if ( isDynamicLayout && src ) {
     177                        genericSubs.apply( src );
     178                }
    179179
    180180                ObjectDecl * returnVal = NULL;
Note: See TracChangeset for help on using the changeset viewer.