Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!think.com!mintaka!pogo.ai.mit.edu!rjc
From: rjc@pogo.ai.mit.edu (Ray Cromwell)
Newsgroups: comp.sys.amiga.emulations
Subject: Re: Emulator Mechanics (sorry long post)
Message-ID: <1991Mar9.033450.9362@mintaka.lcs.mit.edu>
Date: 9 Mar 91 03:34:50 GMT
References: <1303@macuni.mqcc.mq.oz> <39935@cup.portal.com> <10775@dog.ee.lbl.gov>
Sender: daemon@mintaka.lcs.mit.edu (Lucifer Maleficius)
Distribution: na
Organization: None
Lines: 39


  Besides inline data, don't forget this old trick

  lda choice
  asl
  tax
  lda jumptable+1,x
  pha
  lda jumptable,x
  pha
  php
  rti

Or what about this

entrypoint1 lda #data1
            .byte 2c
entrypoint2 lda #data2
   	    .byte 2c
entrypoint3 lda #data3
            .byte 2c
etc,etc

2c is the 6502 instruction 'bit'. The operand to bit is a SHORT 
pointing to an address. LDA # is 2 bytes long, so entering at
any point causes the following instructions to do nothing.


 How are compilers supposed to be smart enough to recognize such
clever optimizations and translate them to their 68000 equivelent?
The 6502 could only address 64k, space savers like this were used all
the time. Then of course, there weren't any rules of how to call
the OS, so most 6502 programs would JUMP directly into rom at different
addresses depending on how much of the function they wanted to use.

I think compilers would do a much better job on 16bit micro's where
C is used. I don't know a single C compiler that generates self modifying
code, or 'do-nothing' instructions between opcodes to save 
bytes whenentering functions.