1

I am writing a compiler from Java 1.0 to x86. I have built the Abstract Syntax Tree (using lex + yacc) and the symbol table. Now, in order to do semantic analysis I plan on traversing the AST in a Depth-first order , type-checking each syntactic construct as I reach its node while also performing necessary type conversions. However, before proceeding to intermediate code generation, what more should be done in the semantic analysis phase.

Shashank Kumar
  • 65
  • 8
  • 1
    It's up to you, really. Typical checks would include checks for missing or duplicate identifiers, range checks that can be performed at compile-time (arrays, constant assignment, and so on). This can also include live variable analysis (i.e. is a variable referenced before it's initialized, and so on), but this is not strictly required, just a nice-to-have. – 500 - Internal Server Error Apr 17 '20 at 08:47
  • @500-InternalServerError People have said that scope analysis is a part of semantic analysis. But since scope analysis I believe involves creation of global and local scopes and linking these scopes together to form a tree structure, I think it can be done while generating the syntax tree itself and it is not a part of semantic analysis since we are not doing anything meaning related here. – Shashank Kumar Apr 17 '20 at 08:58
  • @500-InternalServerError Also, what do you mean by missing or duplicate identifiers? – Shashank Kumar Apr 17 '20 at 08:59
  • Identifiers that are either not defined at all or defined twice in the same scope. – 500 - Internal Server Error Apr 17 '20 at 11:45
  • I don't know that there is a limit to what constitutes semantic analysis, but constant folding, elimination of redundant evaluation of common subexpressions and dead code elimination are pretty common. There are also a lot of analyses based on control flow analysis and symbolic execution. – rici Apr 17 '20 at 15:52

0 Answers0