RE: Pcode definition wanted (Peter Thiemann)
Wed, 16 Oct 91 09:41:52 MEZ

          From comp.compilers

Related articles
Pcode definition wanted (1991-10-15)
RE: Pcode definition wanted (1991-10-16)
| List of all articles for this month |

Newsgroups: comp.compilers
From: (Peter Thiemann)
Keywords: pascal, optimize
Organization: Compilers Central
References: 91-10-053
Date: Wed, 16 Oct 91 09:41:52 MEZ

When reading about pcode, it always occurs to me that a single pcode
instruction is too powerful to be used as an intermediate representation
for code generation. For example, the load and store instructions take the
difference of the declaration level and the current nesting level and an
offset. Such an instruction must be translated into a _sequence_ of
indirect memory accesses and should thus be considered as a MACRO for such
a sequence. (in my opinion, static link chains should really be replaced
by use of a display, anyway...)

>From this point of view, all the load & store instructions you need are
stack frame relative. I feel that this kind of representation has a lot
more potential for optimization than the pcode representation. This kind
of intermediate code looks like a flat representation of an intermediate
tree notation?!


Peter Thiemann, Wilh.-Schickard-Institut, | Phone: x49 7071 29 5467
Sand 13, D-7400 Tuebingen, Germany | Fax: x49 7071 29 5958
Email: |

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.