[Coco] 6309 Weirdness ??

[Robert Gault]

Darren A. wrote:
>> From: Robert Gault
>> Subject: Re: [Coco] 6309 Weirdness ??
>> Date: Tue, 15 May 2007 22:04:53 -0400
>>
>> <snip>
>>
>> My tests indicate that you are misinterpreting your results.
>>
>> When I run the program
>>     org $7000
>> start    ldx #$1234
>> $7003    addr pc,x
>> $7006    swi
>>     end
>> the result is that regX contains $1234+$7003+3=$823A.
>> <snip>
>> Your example LDX #-4  ADDR X,PC  will lead to disaster. Using -3, the 
>> program should loop indefinitely at the addr x,pc instruction. But 
>> with -4, the program will jump into the middle of the ldx #-4 
>> instruction. The result in most cases will be an illegal instruction 
>> resulting in the 6309 jumping to a special trap which in most cases 
>> will not have a vector installed.
>>
>> -- 
> 
> -
> 
> Robert,
> 
> Did you run your test on real 6309 hardware or with an emulator?  When I 
> run the example you provided:
> 
>  ORG   $7000
> 7000  LDX   #$1234
> 7003  ADDR  PC,X
> 
> ..I get the same results you did if I run it on MESS (X = $823A), but I 
> get a different result (X = $823B) when I run it on real hardware. This 
> indicates to me that the CPU is calculating it as ($1234 + $7003 + 4).
> 
> Thus the example I gave:
> 
>      LDX   #-4
>      ADDR  X,PC
> 
> ...will, as you say, lead to disaster if run on MESS, but works on a 
> real 6309 (as long as interrupts are masked).  That was the point of my 
> post; that the real CPU isn't behaving the way I was led to believe. I 
> even tried it with two different processors (a 63B09E and a 63C09E) just 
> to see if I had a defective one.
> 
> Darren
> 

I had tried first on MESS. I just tried the code on a real Coco3 but 
still don't get your results although there is a problem.

  ORG $7000
7000 LDX #$1234
7003 ADDR X,PC
7006 SWI
results in regPC=$823A which is what it should be but still unexpected. 
I ran the test in the Debug module of EDTASM and would have expected the 
code to jump to $823A which is RORB and then attempt $823B which is an 
illegal instruction. That was what happened with MESS. However, on a 
real Coco EDTASM issued a BRK @ 823A with regPC=$823A even though there 
was no breakpoint set there.
  ORG $7000
7000 LDX #$1234
7003 ADDR PC,X
7006 SWI
results in regX=$823B and regPC=$7006. Here the addition is wrong even 
though regPC stepped through the code correctly.

The code
  LDX #-3
  ADDR X,PC
when run from EDTASM with the G (go command) works correctly but stepped 
through with "," results in CAN'T CONTINUE regardless of the value in regX.

To remove any effects from EDTASM the following code was assembled
  ORG $7000
START LDX #-3
  ORCC #$50
  ADDR X,PC
  RTS START
This when executed, immediately came back to the Basic prompt OK. Same 
thing happened with LDX #-4. This opcode ADDR X,PC seems to have been 
ignored. Change the RTS to SWI with the code run from Basic caused a 
system crash. This seems to comfirm that ADDR X,PC was skipped as Basic 
does not like SWI in this context.