编译原理 中间代码生成.pptVIP

中间代码生成;编译器的前端与后端 ;3;语法树, DAG 后缀表达式可以看做语法树的一种线性表示 a b c uminus * b c uminus * + assign ;语法树能通过语法制导定义（SDD）生成. E.g: ;6;三地址代码;翻译时生成临时变量: x + y * z t1 := y * z t2 := x + t1 DAG;表达式的三地址代码;newtemp() 生成临时变量: t1, t2, … gen() 生成中间代码 ;生成四元式;E.g. 关于数组 x[i] := y 或者 x := y[i]三元式;间接三元式 ;过程声明;Integers have width=4 and reals have the width=8;跟踪记录过程中变量的作用域;嵌套过程的符号表; 过程 : mktable(previous) - create a new symbol table and return a pointer to the new table (placed in the header) enter(table, name, type, offset) – create a new entry name addwidth(table, width) enterproc(table, name, newtable) – create a new entry for the procedure name in the symbol table pointed by table ;For each procedure generated by this grammar, the translation scheme sets up a separate symbol table ;20;记录类型;赋值语句;lookup() procedure is modified: ;By modifying the lookup procedure, we can apply the translation scheme if the most closely nested scope rule applies to nonlocal names (like Pascal) – Fig. 8.15 ;e.g. pp 479, last paragraph;数组的地址计算;For row-major form, A[i1, i2] is stored at: Some languages give the possibility to specify the size of the array dynamically, when a procedure is called run-time (see sect. 7.3, pp. 408). The formulas above remain valid, but the upper and lower limits are not known at compile time. Generalizations for row-major form for A[i1, i2, ..,ik] ((…((i1n2 + i2)n3+i3)…)nk+ik) ? w + base –((…((low1n2+low2)n3+low3)…)nk+lowk) ? w In order to generate code for array reference we must do a connection of the computation above with the grammar for array reference ;The key problem in generating code for array references: relate the computation below to a grammar for array references;类型转换;30;布尔表达式 ;32;布尔表达式的SDT:;34;控制流语句;36;37;Control-Flow Translation of Boolean Expressions (E.code) ;(syntax directed definition to produce TAS code for Booleans) ;例子: ;Revisited: E ? E1 + E2;Case 语句翻译;43;回填技术;Three functions (for code generation in one pas): makelist