🚀 go-pugleaf

In article <1989Nov6.172043.10373@world.std.com> bzs@world.std.com (Barry
Shein) writes:
%A common, general usage of self-modifying code is seen in interpreter
%(not necessarily, but often, programming languages) loops. Suppose you
%have the ability to trace or otherwise flag certain events. This is
%typically done by a check at the top of the loop:

%	IF request_is_flagged THEN do_flag_action

%The check for "request_is_flagged" can be quite expensive (say,
%hashing into a symbol table to see if a bit is set) and interpreters
%might have to go through this loop for every minor operation (eg. add
%two numbers.) If there are no items flagged or only certain subsets
%need to be checked one way to speed up this loop is to simply modify
%the top-level code so the check isn't done or loops back to a
%different check. ...

Some time ago I made a posting about the connection between partial
evaluation and self-modifying code. (Partial evaluation is simplifying
functions, or code, when the input is partially known. Self-modifying code
can be seen as partial evaluation at runtime.) The above was an instance I
though of when writing that posting. The rule

IF true THEN S => S

is used to simplify conditionals where the condition is known. This can be
done either at compile-time, if the condition is known then, or at run-time,
when the condition becomes known (and we know it will not change value).

Actually, I think partial evaluation would provide a quite clean theory for
an interesting class of self-modifying programs.

Bjorn Lisper