Passing constants and addresses to assembler

scotta

I was wondering if anyone has figured out a way to pass constants
and references (addresses) from Spin to assembler without the
use of the PAR variable ?

Scott

deSilva

There is nothing to figure out
- You can use constants "as is". They are fully integrated into the Propeller Tool concept.
- You can set or preset any address you want in a DAT cell loaded up with the COG code. See the examples in my Tutorial.

scotta

The following code shows what I have been doing.

I think its easier than using PAR, and uses the same code
space and execution time (as PAR).

Note how the LCDx_BAUD variable is being changed
from spin (in Main method), on a separate cog. The
assembler LCD object dereferences this on the fly.

I have used this technique for simple assembler interfaces
with 1 variable to 40 incoming references and
(about) 10 pre-set constants.

<
*************************
****    MAIN.spin FILE   
*************************
OBJ
    LCD1     : "LCD"
    LCD2     : "LCD"

VAR
    long LCD1_BAUD
    long LCD2_BAUD

PUB Main
    LCD1_BAUD:=9600
    LCD2_BAUD:=19200

    LCD1.Start(1,@LCD1_BAUD)
    LCD2.Start(2,@LCD2_BAUD)

    repeat
         LCD1_BAUD:=LCD1_BAUD+10 
         LCD2_BAUD:=LCD2_BAUD+10
    
         DoMoreStuff(....)



*************************
****    LCD.spin FILE   
*************************
PUB Setup(lcd_pin_, lcd_baud_)
  lcd_pin               := lcd_pin_             'copy constant to assembler space
  lcd_baud_ref      := lcd_baud_ref_     'copy reference to assembler space
  cognew(@Run, 0 )
 

DAT ORG 0
RUN
'setup lcd_pin as a mask (note: lcd_pin is a constant and will never change)
     mov t1,#1
     shl   t1,lcd_pin
     mov lcd_pin,t1

'read the baud rate, compute bit-ticks on the fly
     rdlong t1,lcd_baud_ref
     call #do_bit_ticks

'more program
....


t1 long 0
lcd_pin long 0
lcd_baud_ref long 0


      


>

deSilva

Many people do it that way. Fine you give this additional example.

I think "Start" = "SetUp" ?

Edit: An additional advantage is that you can sometimes trade expensive COG space against cheap HUB space:

  lcd_pin               := |< lcd_pin_

will save you 3 COG instructions..

Post Edited (deSilva) : 9/26/2007 10:49:05 PM GMT

scotta

Parallax should do an example on this method, its
much easier to read than:

mov t1,par
rdlong a,t1
add t1,#4
rdlong b,t1
...

deSilva

DeSilva did On page 11.

But the PAR method is much more systematic and can have advantages in case of MANY parameters.
Also, when you want to load many COGs with the same code a PAR parametrisation avoids the waiting for the end of loading of each of those COGs....

The main pitfall is, that when using VAR variables as target, the mixing of LONGs, WORDs, and BYTEs can lead to much confusion

scotta

[noparse]:)[/noparse]

I should Read the Funny Manual

Thanks !