Re: Verified compilers?

["Randy Brukardt" <randy@rrsoftware.com>]

"Shark8" <onewingedshark@gmail.com> wrote in message 
news:9207716.776.1331054644462.JavaMail.geo-discussion-forums@ynaz38...
> On Monday, March 5, 2012 7:27:25 PM UTC-6, Randy Brukardt wrote:
>>
>> In any case, no one is going to run out and totally rebuild their 
>> compiler
>> because someone has found a supposedly "better" architecture for it. 
>> There
>> were many such better architectures when I started out (supposely 
>> everything
>> was going to be table-driven and automatically generated), and they're 
>> all
>> long gone. No reason to assume that anything is different here.
>
> I seem to recall someone on this forum doing a table-driven approach, 
> though it might have
> only been WRT tokenization...

The table-driven approaches I was thinking of were involved with semantics 
and code generation. Table-driver parsers and tokenizers are common 
(Janus/Ada uses a table-driven parser, for instance.)

And to be fair, we did use a tiny corner of a table-driver approach in our 
code generator (there is a pattern-matching table for generating improved 
code for some sequences). We've move away from that in recent updates (we 
used to handle all numeric literals using the table, nowdays they are 
handled as special stack contents, a more general approach allowing more use 
of indexed addressing modes and instructions with embedded constants).

                        Randy.