PASM to spin

UltraLazer · 2009-12-05 00:02

Is there a good way to send a long from a PASM cog back to a spin cog.

I am using a

COGINIT (SEQ_cog,@Step000,@SEQ)

to lunch my PASM cog so par is pointing at the beginning of a

SEQ long 0,0,0,0,0,0....

in DAT...

I have a changing variable (long) in my PASM that I want to

wrlong

to update the value in hub ram 1 x per loop of the pasm.

should I reserve space in DAT for the value to be shared? or declare it in spin?

can I "reuse" the par register later in the pasm, now that the cog has been initialized to the sequence?

Graham Stabler · 2009-12-05 00:12

You can declare seq in the VAR section then access it using par as you say:

mov   p,par      'make a copy of par
wrlong   data,p  'write to hub
add   p,#4     'address of next long
wrlong   data,p  'write to next position

par will just stay as par I am not sure what you mean by resuse

JonnyMac · 2009-12-05 00:18

When you start a PASM cog you can only pass one parameter (usually an address, as you're doing with Step000) -- this goes across in par. If you're dealing with more than one address in the hub then your PASM code needs to understand the relationship to the address passed in par.

Let's say you have three longs; you really should put them into contiguous locations and then pass the address of the first with coginit. In your PASM code you can start off like this:

dat

                        org     0

asm                     mov     val1pntr, par
                        mov     val2pntr, val1pntr
                        add     val2pntr, #4
                        mov     val3pntr, val2pntr
                        add     val3pntr, #4

This a simple way to handle it; there are some crafty programmer here that have come up with other solutions that are very clever -- I tend to stick with simple (keeps my brain from melting).

UltraLazer · 2009-12-05 21:26

Thanks for the tips guys, although I think my question was unclear. Sorry Im a PASM newb.

I am actually experimenting with Ray Tracy's, Mother of all LED sequencers (M.O.A.L.S.?)(http://obex.parallax.com/objects/240/)

I have a spin cog that reads a sequence from an eeprom, writes it to some predefined long array space called "SEQ" in DAT and launches the M.O.A.L.S. SEQ_Cog.

This works well, and looks Something like this:

Pub start
    
    i2c.ToRam(@SEQ,@SEQ[noparse][[/noparse]SEQ_num]+3,tmp_MV_Address) 
    COGINIT (SEQ_cog,@Step000,@SEQ) 



DAT
        org     0
{   Step000 initialize   }
Step000 or      DIRA, ledmask           ' Set output lines in the direction register
            mov     Time, delay             ' Setup arbitrary initial $01312D00
            add     Time, cnt               

{ Step010 setup the first sequence }
:Step010
        mov     P0,par                  ' Get data patterns starting address 
        mov     P1,P0                   ' Setup working pointer
        
        rdlong Reps,P1                  ' Get Repetition count
        add     P1,#4                   ' Advance pointer to sequence timing
        rdlong  Dly,P1                  ' Get sequence timing

{ Step020 set working pointer to start of sequence }
:Step020
        mov     P2,P1                    'P1 points to sequence timing
        add     P2,#4                    'P2 now points to element count

{ Step030 get element count }
:Step030
        rdlong  Ectr,P2               ' Get element count
        
{ Step040 get next pattern element }
:Step040
        add     P2,#4                 ' Advance pointer to next element
        
        rdlong T1,P2          ' Get the pattern
        'wrlong T1,SEQ <--- this does not work; AREA IN QUESTION - As far as I can tell I need write value of T1 to the hub about here

{ Step050 turn on leds }
:Step050
        xor     T1,#0                  ' For the Demo Board 03ff0000 (ledmask) for the PropRPM
     
        mov     outa,T1

      ...  jmp     #:Step020 .........and so on.....

        cogid ID
        cogstop ID         
                
        FIT $1F0

SEQ long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
       long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
       long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
       long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
       long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
       long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
       long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0

I am hoping to get the M.O.A.L.S. to update a long in hub ram with its current element as it chunks through the SEQ data.
the par I was referring to is the one already used in the sequencer to set the data patterns adress. I was wondering if I have to reference the address of the shared SEQ data in order to write a long to the hub? my head hurts.

JonnyMac · 2009-12-05 21:39

In your top level program you could declare two longs: one holds the address of the DAT table with the sequence data (this will be passed in par, the next is the variable that holds the current step your sequencer is running. Like this:

var

  long  seqstart
  long  seqstep

Your top-level code would write the address (@) of the data table to seqstart before calling coginit:

pub main

  seqstart := @seq
  coginit(@moas, @seqstart)

The MOAS code needs to capture the value par and add four to it to know the address of seqstep -- this is the address you will write back to that tells the top level code what step MOAS is presently running. Make sense? I would do something like this at the top of MOAS:

dat

                        org     0

moas                    mov     startpntr, par
                        mov     steppntr, startpntr
                        add     steppntr, #4

Remember... you can only pass one parameter when using coginit with an assembly cog. This is why the order of your variables is important; the cog needs to match the spin code that launches it. Now that know where the step address is held in the hub you can write the step back with wrlong (to steppntr).

Post Edited (JonnyMac) : 12/5/2009 9:52:52 PM GMT

UltraLazer · 2009-12-05 22:12

seqstart := @seq etc work well thanks!, but the PASM bit seems to crash the SEQuencer...


Step000 or      DIRA, ledmask           ' Set output lines in the direction register
        mov     Time, delay             ' Setup arbitrary initial $01312D00
        add     Time, cnt               

        mov    startpntr, par                 <--- 
        mov    steppntr, startpntr   <--- this seems to crash the sequncer
        add    steppntr, #4                   <--- 
        
{ Step010 setup the first sequence }
:Step010
        mov     P0,par                  ' Get data patterns starting address
        mov     P1,P0                   ' Setup working pointer
        
        rdlong Reps,P1                  ' Get Repetition count
        add     P1,#4                   ' Advance pointer to sequence timing
        rdlong  Dly,P1 

.....
wrlong T1,steppntr   'T1 holds current SEQ value
...
steppntr res  1
startpntr res 1

The first 3 values (0,1,2) in the SEQ data are required for reps, timing and # of elements. Is this an issue with the suggested fix?

MagIO2 · 2009-12-05 23:40

What Jonny did not tell you is that you have to read the SEQ address from HUB RAM then.

{ Step010 setup the first sequence }
:Step010
mov P0,par ' Get data patterns starting address

needs to be changed to
{ Step010 setup the first sequence }
:Step010
rdlong P0,startpntr ' Get data patterns starting address

Ehmmm ... and currently the sequencer does not wait. My expectation is that this will be changed, as you want to see the sequence?! But·at the moment·your SPIN code will miss a lot of data.

Post Edited (MagIO2) : 12/5/2009 11:48:20 PM GMT

PASM to spin

Comments