Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!unix.cis.pitt.edu!dsinc!bagate!cbmvax!cbmehq!cbmdeo!jsmami!jsmoller
From: jsmoller@jsmami.UUCP (Jesper Steen Moller)
Newsgroups: comp.sys.amiga.programmer
Subject: Re: > SAS gripes (was DICE vs GCC)
Message-ID: <18f1e3ae.ARN128f@jsmami.UUCP>
Date: 6 Apr 91 19:33:06 GMT
References: <9104021420.AA10848@thunder.LakeheadU.Ca> <1991Apr4.034920.16298@marlin.jcu.edu.au>
 <1991Apr4.180217.19773@nntp-server.caltech.edu> <1991Apr5.030228.28756@marlin.jcu.edu.au>
 <1991Apr5.173845.4404@nntp-server.caltech.edu> <20421@cbmvax.commodore.com>
Reply-To: cbmehq!cbmdeo!jsmami!jsmoller (Jesper S. Moller)
Followup-To: comp.sys.amiga.programmer
Organization: Danish SofTech
Lines: 51

In article <20421@cbmvax.commodore.com>, Randell Jesup writes:

> In article <1991Apr5.173845.4404@nntp-server.caltech.edu> tll@nntp-server.caltech.edu (Tal Lewis Lancaster) writes:
> >cpca@marlin.jcu.edu.au (Colin Adams) writes:
> >>  Because I am doing things that just can't be done with SAS or Aztec.
> >
> >>Why not?
> >
> >Well the main reason is I am creating object files greater than 32K (actually
> >some are around 80K).  SAS and Aztec can not handle object files > 32K.  Or to
> >be more precise a function call to another function in the same object file must
> >be < 32K apart.
> 
> 	Easy, just use -r0 to turn off pc-relative bsr's (-b0 to use long
> data addressing).

Not easy, unfortunately. The branches aren't correct (in fact, the compiler
will produce some utter crap if the branches exceed 32k).
-b0 will turn JSR xx(pc) into JSR xx, but only for outside functions...
To optimize, the compiler changes all jsr's to BRA.(B|W), somthing
it's done ever since version 3.03. These BSRs are not affected by -r0
or -r1... A bug to me...

To prove that, I made a file of:

void f1() {f1();}
void f2() {f1();}
void f3() {f1();}
void f4() {f1();}
void f5() {f1();}
/* etc... 8000 functions all in all */

And compiled with -r0 (with lc1b, lc1 simply can't do it...)
And -v for lc2 of course...

Try disassembling this (no go with OMD, but Matt's dobj does this
just fine :) ). It's ok for the first 32k, but from there on,
there's garbage instead (as a "text-section exceeds 32k,
branches might be corrupt (or whatever) warns the user).
The brances are wrong and there's garbage after them...

Hope you don't rely on SAS/C too much...

> Randell Jesup, Keeper of AmigaDos, Commodore Engineering.

Greets, Jesper
--                     __
Jesper Steen Moller   ///  VOICE: +45 31 62 46 45
Maglemosevej 52  __  ///  USENET: cbmehq!cbmdeo!jsmoller
DK-2920 Charl    \\\///  FIDONET: 2:231/84.45
Denmark           \XX/