Handout 7
CMSC 22610, W2011

Abstract Machine Design

What are the requirements?

* data representation
  - primitive data
    - integers (big-endian vs little-endian)
    - strings (pointers to heap-allocated char vectors)
  - structured data
    - tuples (pointers to heap-allocated word vectors)
    - datatype values (constructions)
      - tags and headers
  - boxed and unboxed values
    tagged values
  - heap allocation of structured data

* environments
  - function closure environments
  - local environments (let decls)
  - global environments
  - structure and access
    - vector/array accessed by indexing
    - name mapping accessed by name lookup

* variables
  - representation  (indexes or offsets)
  - access

* temporary storage of intermediate results
  (i.e. values that are never bound to a variable)
  - value stack
  
* function representation (closures)
  - code
  - environment

* function call and return
  - argument passing
  - returning results
  - "stack frames" or "activation records"
    - "return address"
    - local variable storage
  - special functions provided by the a.m. (C calls)

* a.m. code
  - instructions
    - representation ("symbolic" or bytes)
    - primitive ops (e.g. integer addition)
    - control flow (e.g. if, case, (jump, conditional jump))
    - book-keeping (e.g. stack manipulation)
  - linear vs tree structured code
  - labels vs code string pointers
  - instruction size: uniform or variable

* a.m. structures
  - value stack, used for intermediate results
  - stack pointer register (SP -- top of stack)
  - current environment register (EV)
  - current "call frame" pointer (FP)
  - multiple stacks?
  - the heap (where compound data structures are
    allocated)

* garbage collection of the heap (?)