// // Cforall Version 1.0.0 Copyright (C) 2015 University of Waterloo // // The contents of this file are covered under the licence agreement in the // file "LICENCE" distributed with Cforall. // // CodeGenerator.cc -- // // Author : Richard C. Bilson // Created On : Mon May 18 07:44:20 2015 // Last Modified By : Peter A. Buhr // Last Modified On : Thu Mar 30 16:38:01 2017 // Update Count : 482 // #include #include #include #include #include "Parser/ParseNode.h" #include "SynTree/Declaration.h" #include "SynTree/Expression.h" #include "SynTree/Initializer.h" #include "SynTree/Statement.h" #include "SynTree/Type.h" #include "SynTree/Attribute.h" #include "Common/utility.h" #include "Common/UnimplementedError.h" #include "CodeGenerator.h" #include "OperatorTable.h" #include "GenType.h" #include "InitTweak/InitTweak.h" using namespace std; namespace CodeGen { int CodeGenerator::tabsize = 4; // the kinds of statements that would ideally be followed by whitespace bool wantSpacing( Statement * stmt) { return dynamic_cast< IfStmt * >( stmt ) || dynamic_cast< CompoundStmt * >( stmt ) || dynamic_cast< WhileStmt * >( stmt ) || dynamic_cast< ForStmt * >( stmt ) || dynamic_cast< SwitchStmt *>( stmt ); } void CodeGenerator::extension( Expression * expr ) { if ( expr->get_extension() ) { output << "__extension__ "; } // if } // extension void CodeGenerator::extension( Declaration * decl ) { if ( decl->get_extension() ) { output << "__extension__ "; } // if } // extension void CodeGenerator::asmName( DeclarationWithType * decl ) { if ( ConstantExpr * asmName = decl->get_asmName() ) { output << " asm ( " << asmName->get_constant()->get_value() << " )"; } // if } // extension ostream & CodeGenerator::Indenter::operator()( ostream & output ) const { return output << string( cg.cur_indent, ' ' ); } ostream & operator<<( ostream & output, const CodeGenerator::Indenter &indent ) { return indent( output ); } CodeGenerator::LabelPrinter & CodeGenerator::LabelPrinter::operator()( std::list< Label > & l ) { labels = &l; return *this; } ostream & operator<<( ostream & output, CodeGenerator::LabelPrinter & printLabels ) { std::list< Label > & labs = *printLabels.labels; // l.unique(); // assumes a sorted list. Why not use set? Does order matter? for ( Label & l : labs ) { output << l.get_name() + ": "; printLabels.cg.genAttributes( l.get_attributes() ); } // for return output; } CodeGenerator::CodeGenerator( std::ostream & os, bool pretty, bool genC ) : indent( *this), cur_indent( 0 ), insideFunction( false ), output( os ), printLabels( *this ), pretty( pretty ), genC( genC ) {} CodeGenerator::CodeGenerator( std::ostream & os, std::string init, int indentation, bool infunp ) : indent( *this), cur_indent( indentation ), insideFunction( infunp ), output( os ), printLabels( *this ) { //output << std::string( init ); } CodeGenerator::CodeGenerator( std::ostream & os, char * init, int indentation, bool infunp ) : indent( *this ), cur_indent( indentation ), insideFunction( infunp ), output( os ), printLabels( *this ) { //output << std::string( init ); } string CodeGenerator::mangleName( DeclarationWithType * decl ) { if ( pretty ) return decl->get_name(); if ( decl->get_mangleName() != "" ) { // need to incorporate scope level in order to differentiate names for destructors return decl->get_scopedMangleName(); } else { return decl->get_name(); } // if } void CodeGenerator::genAttributes( list< Attribute * > & attributes ) { if ( attributes.empty() ) return; output << "__attribute__ (("; for ( list< Attribute * >::iterator attr( attributes.begin() );; ) { output << (*attr)->get_name(); if ( ! (*attr)->get_parameters().empty() ) { output << "("; genCommaList( (*attr)->get_parameters().begin(), (*attr)->get_parameters().end() ); output << ")"; } // if if ( ++attr == attributes.end() ) break; output << ","; // separator } // for output << ")) "; } // CodeGenerator::genAttributes //*** Declarations void CodeGenerator::visit( FunctionDecl * functionDecl ) { extension( functionDecl ); genAttributes( functionDecl->get_attributes() ); handleStorageClass( functionDecl ); functionDecl->get_funcSpec().print( output ); output << genType( functionDecl->get_functionType(), mangleName( functionDecl ), pretty, genC ); asmName( functionDecl ); // acceptAll( functionDecl->get_oldDecls(), *this ); if ( functionDecl->get_statements() ) { functionDecl->get_statements()->accept( *this ); } // if } void CodeGenerator::visit( ObjectDecl * objectDecl ) { if (objectDecl->get_name().empty() && genC ) { // only generate an anonymous name when generating C code, otherwise it clutters the output too much static UniqueName name = { "__anonymous_object" }; objectDecl->set_name( name.newName() ); } extension( objectDecl ); genAttributes( objectDecl->get_attributes() ); handleStorageClass( objectDecl ); output << genType( objectDecl->get_type(), mangleName( objectDecl ), pretty, genC ); asmName( objectDecl ); if ( objectDecl->get_init() ) { output << " = "; objectDecl->get_init()->accept( *this ); } // if if ( objectDecl->get_bitfieldWidth() ) { output << ":"; objectDecl->get_bitfieldWidth()->accept( *this ); } // if } void CodeGenerator::handleAggregate( AggregateDecl * aggDecl ) { genAttributes( aggDecl->get_attributes() ); if( ! aggDecl->get_parameters().empty() && ! genC ) { // assertf( ! genC, "Aggregate type parameters should not reach code generation." ); output << "forall("; genCommaList( aggDecl->get_parameters().begin(), aggDecl->get_parameters().end() ); output << ")" << endl; } if ( aggDecl->get_name() != "" ) output << aggDecl->get_name(); if ( aggDecl->has_body() ) { std::list< Declaration * > & memb = aggDecl->get_members(); output << " {" << endl; cur_indent += CodeGenerator::tabsize; for ( std::list< Declaration* >::iterator i = memb.begin(); i != memb.end(); i++ ) { output << indent; (*i)->accept( *this ); output << ";" << endl; } // for cur_indent -= CodeGenerator::tabsize; output << indent << "}"; } // if } void CodeGenerator::visit( StructDecl * structDecl ) { extension( structDecl ); output << "struct "; handleAggregate( structDecl ); } void CodeGenerator::visit( UnionDecl * unionDecl ) { extension( unionDecl ); output << "union "; handleAggregate( unionDecl ); } void CodeGenerator::visit( EnumDecl * enumDecl ) { extension( enumDecl ); output << "enum "; genAttributes( enumDecl->get_attributes() ); if ( enumDecl->get_name() != "" ) output << enumDecl->get_name(); std::list< Declaration* > &memb = enumDecl->get_members(); if ( ! memb.empty() ) { output << " {" << endl; cur_indent += CodeGenerator::tabsize; for ( std::list< Declaration* >::iterator i = memb.begin(); i != memb.end(); i++) { ObjectDecl * obj = dynamic_cast< ObjectDecl* >( *i ); assert( obj ); output << indent << mangleName( obj ); if ( obj->get_init() ) { output << " = "; obj->get_init()->accept( *this ); } // if output << "," << endl; } // for cur_indent -= CodeGenerator::tabsize; output << indent << "}"; } // if } void CodeGenerator::visit( TraitDecl * traitDecl ) {} void CodeGenerator::visit( TypedefDecl * typeDecl ) { assertf( ! genC, "Typedefs are removed and substituted in earlier passes." ); output << "typedef "; output << genType( typeDecl->get_base(), typeDecl->get_name(), pretty, genC ) << endl; } void CodeGenerator::visit( TypeDecl * typeDecl ) { if ( genC ) { // really, we should mutate this into something that isn't a TypeDecl but that requires large-scale changes, // still to be done extension( typeDecl ); output << "extern unsigned long " << typeDecl->get_name(); if ( typeDecl->get_base() ) { output << " = sizeof( " << genType( typeDecl->get_base(), "", pretty, genC ) << " )"; } // if } else { output << typeDecl->typeString() << " " << typeDecl->get_name(); if ( ! typeDecl->get_assertions().empty() ) { output << " | { "; genCommaList( typeDecl->get_assertions().begin(), typeDecl->get_assertions().end() ); output << " }"; } } } void CodeGenerator::printDesignators( std::list< Expression * > & designators ) { typedef std::list< Expression * > DesignatorList; if ( designators.size() == 0 ) return; for ( DesignatorList::iterator iter = designators.begin(); iter != designators.end(); ++iter ) { if ( dynamic_cast< NameExpr * >( *iter ) ) { // if expression is a name, then initializing aggregate member output << "."; (*iter)->accept( *this ); } else { // if not a simple name, it has to be a constant expression, i.e. an array designator output << "["; (*iter)->accept( *this ); output << "]"; } // if } // for output << " = "; } void CodeGenerator::visit( SingleInit * init ) { printDesignators( init->get_designators() ); init->get_value()->accept( *this ); } void CodeGenerator::visit( ListInit * init ) { printDesignators( init->get_designators() ); output << "{ "; if ( init->begin() == init->end() ) { // illegal to leave initializer list empty for scalar initializers, but always legal to have 0 output << "0"; } else { genCommaList( init->begin(), init->end() ); } // if output << " }"; } void CodeGenerator::visit( ConstructorInit * init ){ assertf( ! genC, "ConstructorInit nodes should not reach code generation." ); // xxx - generate something reasonable for constructor/destructor pairs output << ""; } void CodeGenerator::visit( Constant * constant ) { output << constant->get_value() ; } //*** Expressions void CodeGenerator::visit( ApplicationExpr * applicationExpr ) { extension( applicationExpr ); if ( VariableExpr * varExpr = dynamic_cast< VariableExpr* >( applicationExpr->get_function() ) ) { OperatorInfo opInfo; if ( varExpr->get_var()->get_linkage() == LinkageSpec::Intrinsic && operatorLookup( varExpr->get_var()->get_name(), opInfo ) ) { std::list< Expression* >::iterator arg = applicationExpr->get_args().begin(); switch ( opInfo.type ) { case OT_PREFIXASSIGN: case OT_POSTFIXASSIGN: case OT_INFIXASSIGN: case OT_CTOR: case OT_DTOR: { assert( arg != applicationExpr->get_args().end() ); if ( AddressExpr * addrExpr = dynamic_cast< AddressExpr * >( *arg ) ) { // remove & from first assignment/ctor argument *arg = addrExpr->get_arg(); } else { // no address-of operator, so must be a pointer - add dereference // NOTE: if the assertion starts to trigger, check that the application expr isn't being shared. // Since its arguments are modified here, this assertion most commonly triggers when the application // is visited multiple times. UntypedExpr * newExpr = new UntypedExpr( new NameExpr( "*?" ) ); newExpr->get_args().push_back( *arg ); Type * type = InitTweak::getPointerBase( (*arg)->get_result() ); assertf( type, "First argument to a derefence must be a pointer. Ensure that expressions are not being shared." ); newExpr->set_result( type->clone() ); *arg = newExpr; } // if break; } default: // do nothing ; } // switch switch ( opInfo.type ) { case OT_INDEX: assert( applicationExpr->get_args().size() == 2 ); (*arg++)->accept( *this ); output << "["; (*arg)->accept( *this ); output << "]"; break; case OT_CALL: // there are no intrinsic definitions of the function call operator assert( false ); break; case OT_CTOR: case OT_DTOR: if ( applicationExpr->get_args().size() == 1 ) { // the expression fed into a single parameter constructor or destructor may contain side // effects, so must still output this expression output << "("; (*arg++)->accept( *this ); output << ") /* " << opInfo.inputName << " */"; } else if ( applicationExpr->get_args().size() == 2 ) { // intrinsic two parameter constructors are essentially bitwise assignment output << "("; (*arg++)->accept( *this ); output << opInfo.symbol; (*arg)->accept( *this ); output << ") /* " << opInfo.inputName << " */"; } else { // no constructors with 0 or more than 2 parameters assert( false ); } // if break; case OT_PREFIX: case OT_PREFIXASSIGN: assert( applicationExpr->get_args().size() == 1 ); output << "("; output << opInfo.symbol; (*arg)->accept( *this ); output << ")"; break; case OT_POSTFIX: case OT_POSTFIXASSIGN: assert( applicationExpr->get_args().size() == 1 ); (*arg)->accept( *this ); output << opInfo.symbol; break; case OT_INFIX: case OT_INFIXASSIGN: assert( applicationExpr->get_args().size() == 2 ); output << "("; (*arg++)->accept( *this ); output << opInfo.symbol; (*arg)->accept( *this ); output << ")"; break; case OT_CONSTANT: case OT_LABELADDRESS: // there are no intrinsic definitions of 0/1 or label addresses as functions assert( false ); } // switch } else { varExpr->accept( *this ); output << "("; genCommaList( applicationExpr->get_args().begin(), applicationExpr->get_args().end() ); output << ")"; } // if } else { applicationExpr->get_function()->accept( *this ); output << "("; genCommaList( applicationExpr->get_args().begin(), applicationExpr->get_args().end() ); output << ")"; } // if } void CodeGenerator::visit( UntypedExpr * untypedExpr ) { extension( untypedExpr ); if ( NameExpr * nameExpr = dynamic_cast< NameExpr* >( untypedExpr->get_function() ) ) { OperatorInfo opInfo; if ( operatorLookup( nameExpr->get_name(), opInfo ) ) { std::list< Expression* >::iterator arg = untypedExpr->get_args().begin(); switch ( opInfo.type ) { case OT_INDEX: assert( untypedExpr->get_args().size() == 2 ); (*arg++)->accept( *this ); output << "["; (*arg)->accept( *this ); output << "]"; break; case OT_CALL: assert( false ); case OT_CTOR: case OT_DTOR: if ( untypedExpr->get_args().size() == 1 ) { // the expression fed into a single parameter constructor or destructor may contain side // effects, so must still output this expression output << "("; (*arg++)->accept( *this ); output << ") /* " << opInfo.inputName << " */"; } else if ( untypedExpr->get_args().size() == 2 ) { // intrinsic two parameter constructors are essentially bitwise assignment output << "("; (*arg++)->accept( *this ); output << opInfo.symbol; (*arg)->accept( *this ); output << ") /* " << opInfo.inputName << " */"; } else { // no constructors with 0 or more than 2 parameters assert( false ); } // if break; case OT_PREFIX: case OT_PREFIXASSIGN: case OT_LABELADDRESS: assert( untypedExpr->get_args().size() == 1 ); output << "("; output << opInfo.symbol; (*arg)->accept( *this ); output << ")"; break; case OT_POSTFIX: case OT_POSTFIXASSIGN: assert( untypedExpr->get_args().size() == 1 ); (*arg)->accept( *this ); output << opInfo.symbol; break; case OT_INFIX: case OT_INFIXASSIGN: assert( untypedExpr->get_args().size() == 2 ); output << "("; (*arg++)->accept( *this ); output << opInfo.symbol; (*arg)->accept( *this ); output << ")"; break; case OT_CONSTANT: // there are no intrinsic definitions of 0 or 1 as functions assert( false ); } // switch } else { if ( nameExpr->get_name() == "..." ) { // case V1 ... V2 or case V1~V2 assert( untypedExpr->get_args().size() == 2 ); (*untypedExpr->get_args().begin())->accept( *this ); output << " ... "; (*--untypedExpr->get_args().end())->accept( *this ); } else { // builtin routines nameExpr->accept( *this ); output << "("; genCommaList( untypedExpr->get_args().begin(), untypedExpr->get_args().end() ); output << ")"; } // if } // if } else { untypedExpr->get_function()->accept( *this ); output << "("; genCommaList( untypedExpr->get_args().begin(), untypedExpr->get_args().end() ); output << ")"; } // if } void CodeGenerator::visit( RangeExpr * rangeExpr ) { rangeExpr->get_low()->accept( *this ); output << " ... "; rangeExpr->get_high()->accept( *this ); } void CodeGenerator::visit( NameExpr * nameExpr ) { extension( nameExpr ); OperatorInfo opInfo; if ( operatorLookup( nameExpr->get_name(), opInfo ) ) { assert( opInfo.type == OT_CONSTANT ); output << opInfo.symbol; } else { output << nameExpr->get_name(); } // if } void CodeGenerator::visit( AddressExpr * addressExpr ) { extension( addressExpr ); output << "(&"; // this hack makes sure that we don't convert "constant_zero" to "0" if we're taking its address if ( VariableExpr * variableExpr = dynamic_cast< VariableExpr* >( addressExpr->get_arg() ) ) { output << mangleName( variableExpr->get_var() ); } else { addressExpr->get_arg()->accept( *this ); } // if output << ")"; } void CodeGenerator::visit( CastExpr * castExpr ) { extension( castExpr ); output << "("; if ( castExpr->get_result()->isVoid() ) { output << "(void)" ; } else if ( ! castExpr->get_result()->get_lvalue() ) { // at least one result type of cast, but not an lvalue output << "("; output << genType( castExpr->get_result(), "", pretty, genC ); output << ")"; } else { // otherwise, the cast is to an lvalue type, so the cast should be dropped, since the result of a cast is // never an lvalue in C } // if castExpr->get_arg()->accept( *this ); output << ")"; } void CodeGenerator::visit( UntypedMemberExpr * memberExpr ) { assertf( ! genC, "UntypedMemberExpr should not reach code generation." ); extension( memberExpr ); memberExpr->get_aggregate()->accept( *this ); output << "." << memberExpr->get_member(); } void CodeGenerator::visit( MemberExpr * memberExpr ) { extension( memberExpr ); memberExpr->get_aggregate()->accept( *this ); output << "." << mangleName( memberExpr->get_member() ); } void CodeGenerator::visit( VariableExpr * variableExpr ) { extension( variableExpr ); OperatorInfo opInfo; if ( variableExpr->get_var()->get_linkage() == LinkageSpec::Intrinsic && operatorLookup( variableExpr->get_var()->get_name(), opInfo ) && opInfo.type == OT_CONSTANT ) { output << opInfo.symbol; } else { output << mangleName( variableExpr->get_var() ); } // if } void CodeGenerator::visit( ConstantExpr * constantExpr ) { assert( constantExpr->get_constant() ); extension( constantExpr ); constantExpr->get_constant()->accept( *this ); } void CodeGenerator::visit( SizeofExpr * sizeofExpr ) { extension( sizeofExpr ); output << "sizeof("; if ( sizeofExpr->get_isType() ) { output << genType( sizeofExpr->get_type(), "", pretty, genC ); } else { sizeofExpr->get_expr()->accept( *this ); } // if output << ")"; } void CodeGenerator::visit( AlignofExpr * alignofExpr ) { // use GCC extension to avoid bumping std to C11 extension( alignofExpr ); output << "__alignof__("; if ( alignofExpr->get_isType() ) { output << genType( alignofExpr->get_type(), "", pretty, genC ); } else { alignofExpr->get_expr()->accept( *this ); } // if output << ")"; } void CodeGenerator::visit( UntypedOffsetofExpr * offsetofExpr ) { assertf( ! genC, "UntypedOffsetofExpr should not reach code generation." ); output << "offsetof("; output << genType( offsetofExpr->get_type(), "", pretty, genC ); output << ", " << offsetofExpr->get_member(); output << ")"; } void CodeGenerator::visit( OffsetofExpr * offsetofExpr ) { // use GCC builtin output << "__builtin_offsetof("; output << genType( offsetofExpr->get_type(), "", pretty, genC ); output << ", " << mangleName( offsetofExpr->get_member() ); output << ")"; } void CodeGenerator::visit( OffsetPackExpr * offsetPackExpr ) { assertf( ! genC, "OffsetPackExpr should not reach code generation." ); output << "__CFA_offsetpack(" << genType( offsetPackExpr->get_type(), "", pretty, genC ) << ")"; } void CodeGenerator::visit( LogicalExpr * logicalExpr ) { extension( logicalExpr ); output << "("; logicalExpr->get_arg1()->accept( *this ); if ( logicalExpr->get_isAnd() ) { output << " && "; } else { output << " || "; } // if logicalExpr->get_arg2()->accept( *this ); output << ")"; } void CodeGenerator::visit( ConditionalExpr * conditionalExpr ) { extension( conditionalExpr ); output << "("; conditionalExpr->get_arg1()->accept( *this ); output << " ? "; conditionalExpr->get_arg2()->accept( *this ); output << " : "; conditionalExpr->get_arg3()->accept( *this ); output << ")"; } void CodeGenerator::visit( CommaExpr * commaExpr ) { extension( commaExpr ); output << "("; commaExpr->get_arg1()->accept( *this ); output << " , "; commaExpr->get_arg2()->accept( *this ); output << ")"; } void CodeGenerator::visit( UntypedTupleExpr * tupleExpr ) { assertf( ! genC, "UntypedTupleExpr should not reach code generation." ); output << "["; genCommaList( tupleExpr->get_exprs().begin(), tupleExpr->get_exprs().end() ); output << "]"; } void CodeGenerator::visit( TupleExpr * tupleExpr ) { assertf( ! genC, "TupleExpr should not reach code generation." ); output << "["; genCommaList( tupleExpr->get_exprs().begin(), tupleExpr->get_exprs().end() ); output << "]"; } void CodeGenerator::visit( TypeExpr * typeExpr ) { assertf( ! genC, "TypeExpr should not reach code generation." ); output<< genType( typeExpr->get_type(), "", pretty, genC ); } void CodeGenerator::visit( AsmExpr * asmExpr ) { if ( asmExpr->get_inout() ) { output << "[ "; asmExpr->get_inout()->accept( *this ); output << " ] "; } // if asmExpr->get_constraint()->accept( *this ); output << " ( "; asmExpr->get_operand()->accept( *this ); output << " )"; } void CodeGenerator::visit( CompoundLiteralExpr *compLitExpr ) { assert( compLitExpr->get_result() && dynamic_cast< ListInit * > ( compLitExpr->get_initializer() ) ); output << "(" << genType( compLitExpr->get_result(), "", pretty, genC ) << ")"; compLitExpr->get_initializer()->accept( *this ); } void CodeGenerator::visit( StmtExpr * stmtExpr ) { std::list< Statement * > & stmts = stmtExpr->get_statements()->get_kids(); output << "({" << std::endl; cur_indent += CodeGenerator::tabsize; unsigned int numStmts = stmts.size(); unsigned int i = 0; for ( Statement * stmt : stmts ) { output << indent << printLabels( stmt->get_labels() ); if ( i+1 == numStmts ) { // last statement in a statement expression needs to be handled specially - // cannot cast to void, otherwise the expression statement has no value if ( ExprStmt * exprStmt = dynamic_cast< ExprStmt * >( stmt ) ) { exprStmt->get_expr()->accept( *this ); output << ";" << endl; ++i; break; } } stmt->accept( *this ); output << endl; if ( wantSpacing( stmt ) ) { output << endl; } // if ++i; } cur_indent -= CodeGenerator::tabsize; output << indent << "})"; } //*** Statements void CodeGenerator::visit( CompoundStmt * compoundStmt ) { std::list ks = compoundStmt->get_kids(); output << "{" << endl; cur_indent += CodeGenerator::tabsize; for ( std::list::iterator i = ks.begin(); i != ks.end(); i++ ) { output << indent << printLabels( (*i)->get_labels() ); (*i)->accept( *this ); output << endl; if ( wantSpacing( *i ) ) { output << endl; } // if } // for cur_indent -= CodeGenerator::tabsize; output << indent << "}"; } void CodeGenerator::visit( ExprStmt * exprStmt ) { assert( exprStmt ); // cast the top-level expression to void to reduce gcc warnings. Expression * expr = new CastExpr( exprStmt->get_expr() ); expr->accept( *this ); output << ";"; } void CodeGenerator::visit( AsmStmt * asmStmt ) { output << "asm "; if ( asmStmt->get_voltile() ) output << "volatile "; if ( ! asmStmt->get_gotolabels().empty() ) output << "goto "; output << "( "; if ( asmStmt->get_instruction() ) asmStmt->get_instruction()->accept( *this ); output << " : "; genCommaList( asmStmt->get_output().begin(), asmStmt->get_output().end() ); output << " : "; genCommaList( asmStmt->get_input().begin(), asmStmt->get_input().end() ); output << " : "; genCommaList( asmStmt->get_clobber().begin(), asmStmt->get_clobber().end() ); if ( ! asmStmt->get_gotolabels().empty() ) { output << " : "; for ( std::list