Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!jarthur!nntp-server.caltech.edu!toddpw
From: toddpw@nntp-server.caltech.edu (Todd P. Whitesel)
Newsgroups: comp.sys.apple2
Subject: Re: Orca/c bug followup
Message-ID: <1991Feb5.020407.24118@nntp-server.caltech.edu>
Date: 5 Feb 91 02:04:07 GMT
References: <9102032044.AA05781@relay.CDNnet.CA> <11676@ucrmath.ucr.edu>
Organization: California Institute of Technology, Pasadena
Lines: 34

bazyar@chip (Jawaid Bazyar) writes:

>  If anyone would be interested in helping! just drop me a note. I'm 
>looking for someone who knows C in and out, who knows '816 assembly
>in and out, and who has experience with Orca's code generation models
>(I don't want to redo that stuff from scratch, and it should also be
>compatible with Orca code).

Gee, sounds like me... although the only Orca code generation model I
would stick with is the subroutine calling protocol. I've looked at some
of Orca's output code and I'm convinced that Mike is having problems because
his expression evaluation and l-value computation models are constantly
at odds with each other. I've got an idea for code generation and I think it's
a major winner, but I don't have the time right now to write all the front end
code required before I could actually try it out -- it's a variation on P-code
that is geared for peephole optimization DURING the translation to native code.
I believe I can even get away with some simple register scheduling and still
keep the model reasonable -- and I know of quite a few simple cases in which
this would yield rather nice improvements over Orca's code.

I suggest you DO NOT use the Orca library AT ALL. I believe we should construct
a new library. I've had some fun playing with the .ROOT code (Twilight blankers
in C, finder message to argc/argv converters, and so on) and I'm convinced that
an awesome library could be written in 'a GS/OS state of mind' so to speak.

>  We'll let you know how it goes- I ran out of memory before I could
>see if cpp would compile, but there was only one error I saw, so that
>bodes well.

I've got 2 megs -- will that help? Or does the source just need to be
split up into manageable pieces?

Todd Whitesel
toddpw @ tybalt.caltech.edu