How do you find the hub address of initialized data?

pedward · 2013-04-14 13:57

If I include initialized data in a SPIN program (PASM at this point), how do I find the hub address of that initialized data? I looked at Chips ROM monitor code and he takes some very elaborate steps to statically calculate the offsets and store them as immediates, but I want to know the right way of finding the hub address of a variable.

potatohead · 2013-04-14 14:26

Do you mean the @operator?

What I do is look at the object listing in Pnut, then add the load address to it, which is usually $e80, but can be $1000, or something else

TYPE: 4B   VALUE: 013402B8   NAME: XMIN
TYPE: 4B   VALUE: 013802BC   NAME: YMIN
TYPE: 4B   VALUE: 013C02C0   NAME: XMAX
TYPE: 4B   VALUE: 014002C4   NAME: YMAX
TYPE: 4B   VALUE: 014402C8   NAME: MAX_COMPUTE
TYPE: 4B   VALUE: 014802CC   NAME: COLOR_MUL

That is from the object listing for the fractal thing I just did. The VALUE: field is two words. For YMAX, these are 0140 02C4. The lower word is the HUB address, starting from 0. The upper word is the address relative to the ORG last issued.

So then, YMAX lives at $0e80+$02c4 absolute HUB address.

The "@" operator generally just needs the load address added to it for things to work nicely, if you've not padded the code with an array.

IMHO, the monitor being in low RAM warrants and ORG and OBJ operator... or some other thing that specifies an address offset to work with.

There is a more detailed explanation aimed at using the monitor to modify values, start COGS, etc... You can find that document here: http://forums.parallax.com/attachment.php?attachmentid=98587&d=1357887830

Cluso99 · 2013-04-14 19:27

pedward: If you fill the $e80 are with 0's and start your cog code at $e80, or fill to $1000 and start your cog code there, and use p2load to load it with the -s and optonally -h modifiers, you can do a straight @label. I do this throughout my P2 debugger so take a look there.

pedward · 2013-04-14 20:23

Well, I did get it to work this way:

		setptrb	buffer
		mov	buf, #fox
		shl	buf, #2
		add	buf, cogbase
		call	#print_string

print_string

		setptra	buf
:loop		rdbytec	c, PTRA++	wz
if_nz		or	c, color
if_nz		wrword	c, PTRB++
if_nz		jmp	#:loop		
		
print_string_ret	ret


cogbase		long	$E80
fox		byte	"The quick brown fox jumped over the lazy dog.",0

Cluso99 · 2013-04-14 22:11

If you compile like this...

CON
  _ROM_SIZE     = $E80                          ' length of ROM (i.e. the start of Hub Ram)
  _RESERVED     = $180                          ' reserved for drivers $00E80-$00FFF
  _COG_BASE     = _ROM_SIZE+_RESERVED           ' cog code entry

  _clkmode = xinput
  _xinfreq = 60_000_000
''============================[ DAT ]============================================================
DAT
                byte    0[_ROM_SIZE]            ' filler for ROM space $e80
                byte    0[_RESERVED]            ' filler for reserved HUB space $180
'$01000
                
                org     0
entry
                setptrb buffer
                mov     buf, ptrfox
                call    #print_string
print_string
                setptra buf
:loop           rdbytec c, PTRA++       wz
        if_nz   or      c, color
        if_nz   wrword  c, PTRB++
        if_nz   jmp     #:loop          
                
print_string_ret
                ret

fox             byte    "The quick brown fox jumped over the lazy dog.",0 .
ptrfox          long     @fox                   ' the hub address of string1

and use p2load like this...

p2load -v -s xxxx.obj -h -T

p2load (-s) skips the first $e80 longs, and starts cog execution at (-h) $1000 hub.

This method saves worrying about the $e80 offset.

If you remove the -h, then execution will start at $e80, so delete the line allocating the
byte 0[_RESERVED]

How do you find the hub address of initialized data?

Comments