[Sdcc-user] page swapping

[Joris v. d. S. <vd...@pu...>]

(My ISP bounced back my previous attempt to send this 
 mail. My apologies if this posting should end up twice 
 on the list. -- Joris)

At 18:15 10-11-00 -0800, you wrote:

> I have already added a __page_no__ variable
> using the scheme you described will be executed at the entry
> of a function with uses xstack.

Woeps. I have just studied the data-sheets for the AT89C52,
and found out I was wrong. The example I gave earlier won't
work:

    mov   dph,_stack_page
    movx  a,@dptr   ; WRONG! Won't permanently latch P2!
    mov   r0,_spx
    mov   a,r2
    movx  @r0,a     ; push r2 on external stack

When executing a 'movx @r0,a' or 'movx a,@r0' instruction, the
microcontroller will *not* multiplex port P2 with the MSB of
the address generator! Therefore port P2 will have the last
value latched into its internal buffer. However, 
'movx a,@dptr' will not latch the buffer, it will only
multiplex temporarily with the address generator. Therefore,
we should latch the page *explicitly* into P2:

    mov   P2,_stack_page  ; this will *really* latch the page
    mov   r0,_spx
    mov   a,r2
    movx  @r0,a           ; push r2 on external stack

Also, I said '_stack_page' was stored in 'idata', but that had
to be 'data' of course. Sorry, it had been over 36 hours since
I had some sleep. Doing much better today ;-)

> using the scheme you described will be executed at the entry
> of a function with uses xstack.

I don't think that is necessary. You only need to do this once
in the startup-code. I am not aware of any boards using a
multiplexed P2 / external ram. Therefore, I think it is safe
to assume that programs using external ram almost certainly
won't modify P2 themselves.


Kind regards,

Joris van der Sande
vd...@pu...