[virtmach] Virtual Machines and Code Representation

Peter_Murray@allegiance.com.au Peter_Murray@allegiance.com.au
Mon, 24 Jul 2000 10:51:31 +1000

Great message!  The rationale for using a parse tree rather than byte-codes
is that more information is available to a specific compiler to generate
"better" machine code, in whatever way "better" is defined
(faster/smaller/safer/more parrallel/etc...).

But can this be taken even further?  The more information that is available
to a sufficiently smart compiler, the better, and even a parse-tree throws
away information.  Would a theoretically smart compiler be able to
determine the intent of the programmer from such things as code comments,
design documentation, and any other information that may have been captured
in the development process, and from this information produce machine-level
code way beyond what could be produced from a simple parse-tree?  This
probably sounds ridiculous now, but in another decade or two the available
computing resources will be so much greater that this sort of thing may
become a possibility.

- Peter M.