[virtmach] Virtual Machines and Code Representation

[Peter William Lount]

Hi,

I'll let the paper and other info from the Oberon folks on "slim binaries"
do the talking for me. ;--)

http://www.ics.uci.edu/~franz/SlimBinaries.html

http://www.ussc.alltheweb.com/cgi-bin/search?query=%22Slim+Binaries%22

All the best,

Peter William Lount
peter@smalltalk.org

----------
From: Luk Stoops <lstoops@vub.ac.be>
To: virtmach@iecc.com
Subject: Re: [virtmach] Virtual Machines and Code Representation
Date: July 26, 2000 7:56 AM



Peter William Lount wrote:

> 
> The advantage of the "parse-tree" representation of programs and program
> fragements is that there is more information avaliable about the context
> that the code is in that gets lost with byte codes. Simply put,
byte-codes
> lose information. They have enough information to compile to native code,
> but if that extra information was still avalible then it's possible that
> additional optimizations could be applied to the code.
> 
> 
> The pase tree representation is also elegant in that it simplifies the
> number of representations from four (source text, parse tree, byte-codes,
> native code) to three (source text, parse tree and native code). Simple
but
> not simplistic is better. Compressed parse-trees offer better space
> efficiency than byte-codes. Smaller sizes means faster loading times.
> 

Do you compare compressed parse-trees against uncompressed byte-codes ?

Otherwise it seems difficult to compress more information into less
space?



greetings

Luk

----------------------------------------------------------
[ Luk Stoops                           [lstoops@vub.ac.be]
----------------------------------------------------------