Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: Notesfiles $Revision: 1.7.0.10 $; site ccvaxa
Path: utzoo!watmath!clyde!cbosgd!ihnp4!inuxc!pur-ee!uiucdcs!ccvaxa!aglew
From: aglew@ccvaxa.UUCP
Newsgroups: net.arch
Subject: Re: Addressing modes (conditional branc
Message-ID: <5100019@ccvaxa>
Date: Mon, 3-Mar-86 01:24:00 EST
Article-I.D.: ccvaxa.5100019
Posted: Mon Mar  3 01:24:00 1986
Date-Received: Tue, 4-Mar-86 05:07:23 EST
References: <546@tekcrl.UUCP>
Lines: 18
Nf-ID: #R:tekcrl.UUCP:546:ccvaxa:5100019:000:852
Nf-From: ccvaxa.UUCP!aglew    Mar  3 00:24:00 1986


We DO have the formal discipline to use self-modifying code:
it's called incremental compilation.

For some specific instance of a general routine with all sorts of
ifs depending on parameters values, take a general prototype with all
the ifs in it, compile only the parts you need into a buffer, and then
execute that buffer.

Note: this is at the module level, not the machine instruction level.
Self-modifying code at the instruction level will just always be too
machine dependent to become formalized as standard practice. Doing things
at the module level doesn't mean that we're losing too much efficiency,
on the new machines that support function calls well.

By the way, such compilations into buffers DO NOT make programs non-reentrant
- you just have to have a concept of more than one code (text) segment,
some pure and shared, others not.