SCOOP-A3: Compiler adaptation

Revision as of 02:21, 25 August 2010 by Alexander Kogtenkov (Talk | contribs) (Introduced several new sections with more details)

Construction.png Not Ready for Review: This Page is Under Development!

Team: Alexander Kogtenkov, Ted

Summary

  • Parser
    • Support only separate type declarations (exclude separate class declarations).
    • Provide syntax to refer to the processors (not supported in 6.7).
    • Take into account the possibility to use anchors.
  • Type checking introduces new
    • Conformance rules that affect type system
    • Validity rules that require good description of errors
  • Code generation is closely connected with work on the scheduler (A1, A2) and should target
    • bytecode
    • C code
    • CIL code

Decisions

  1. Syntax
    • Drop support of separate classes, only separate types are supported.
    • Separate mark is allowed not only for class types, but also for other types.
  2. Semantics
    • Default formal generic constraint is detachable separate ANY.
    • Formal generic is considered non-separate if at least one constraint is not separate, otherwise it is considered separate.
  3. Code generation
    • Separate and non-separate class types are generated separately, i.e. ANY and separate ANY get different code.
    • Generic derivations with separate and non-separate generics are generated separately, i.e. ARRAY [G] and ARRAY [separate G] get different code.

Open questions

  • Shall separate types be supported in CECIL? If yes, how this is done?

Progress

SCOOP support in compiler
Functionality Status
Syntax for separate types rev#84252
Separate status in conformance tests
Validity rules specific to separate types
Bytecode generation
C code generation
CIL code generation