ASM wont work...Help Please !!

BTX · 2006-12-26 16:03

Hi all !!
I've just beggining trying to understand assembly code, and I try with this.
Talking some PWM examples from Counters doccument, I want to·dimmer two differents LED at· different pins, I just write this simple spin code trying to use two cogs one for each LED, like still I can't see..how to pass different pins from spin to asm code,·I duplicate twice the PWM code with differents pin assigment, but this wont work !!.
First question is : ...what is the concept ? that I'm not seeing ? If I'm starting only two cogs with different pins ...why it seems to be that the first call to PWM1.go never returns..·and let start the PWM2 object ??
Second question is :· The parameter address is the only way to pass data from spin to asm started cog ?, I'd try to understand FullDuplex Serial object...where you must pass some·variables (like rxpin, txpin, mode, etc), and I can't, these variables appears "for me" like from the "sky" in the cog. (making· "understood" some calculations with PAR register...)
In this PWM example @parameter address, is passed to the cog, and then taken by vallue variable...but·it is only one...how to pass two variables ?

This is the code that I'm talking.
Thanks in advance for your time !!

CON _clkmode = xtal1 + pll16x
    _xinfreq = 5_000_000
OBJ
   R : "PWM1"
   B : "PWM2"
  
PUB Vamos
   R.go
   B.go
      
repeat

VAR long parameter
PUB go | x
  cognew(@entry, @parameter)
  repeat
    repeat x from 0 to period          'linearly advance parameter from 0 to 100
      parameter := x                   'a constant here locks to value x percent high
      waitcnt(1_000_000 + cnt)         'wait a little while before next update
DAT
        org
entry   mov dira, diraval              'set APIN to output
        mov ctra, ctraval              'establish counter A mode and APIN
        mov frqa, #1                   'set counter to increment 1 each cycle
        mov time, cnt                  'record current time
        add time, period               'establish next period
:loop   rdlong value, par              'get an up to date pulse width
        waitcnt time, period           'wait until next period
        neg phsa, value                'back up phsa so that it
        jmp #:loop                     'loop for next cycle
diraval long |< 6                      'APIN=6
ctraval long %00100 << 26 + 6          'NCO/PWM APIN=6
period  long 100                       '800kHz (1.25µS period) (_clkfreq / period)                      
time    res 1
value   res 1

VAR long parameter
PUB go | x
  cognew(@entry, @parameter)
  repeat
    repeat x from 0 to period          'linearly advance parameter from 0 to 100
      parameter := x                   'a constant here locks to value x percent high
      waitcnt(1_000_000 + cnt)         'wait a little while before next update
DAT
        org
entry   mov dira, diraval              'set APIN to output
        mov ctra, ctraval              'establish counter A mode and APIN
        mov frqa, #1                   'set counter to increment 1 each cycle
        mov time, cnt                  'record current time
        add time, period               'establish next period
:loop   rdlong value, par              'get an up to date pulse width
        waitcnt time, period           'wait until next period
        neg phsa, value                'back up phsa so that it
        jmp #:loop                     'loop for next cycle
diraval long |< 5                      'APIN=5
ctraval long %00100 << 26 + 5          'NCO/PWM APIN=5
period  long 100                       '800kHz (1.25µS period) (_clkfreq / period)                      
time    res 1
value   res 1

Regards
Alberto.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

CJ · 2006-12-26 16:17

try using cognew on the first go call, your code as it is now, the main cog gets into the infinite loop in the go method and never gets to the second go call

you will also need some stack space set aside for it

cognew(R.go, @stack_array)

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Who says you have to have knowledge to use it?

I've killed a fly with my bare mind.

BTX · 2006-12-26 16:32

Hi Cj.
Thanks for your help.
I'd try with this...but doesn't work

Is this correct·? .. only one LED continue fadding ..like in my original code....

CON _clkmode = xtal1 + pll16x
    _xinfreq = 5_000_000
VAR
   long Stack[noparse][[/noparse]20]
   long Dstack[noparse][[/noparse]20]
   
OBJ
   R : "PWM1"
   B : "PWM2"
  
PUB Vamos
   cognew(R.go, @Stack)
   cognew(B.go, @Dstack)

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

CJ · 2006-12-26 16:56

I don't see anything wrong with the code, you may want to check your wiring, possibly switch the LEDs to make sure they both work

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Who says you have to have knowledge to use it?

I've killed a fly with my bare mind.

Mike Green · 2006-12-26 17:49

I don't think you can use COGNEW with methods in other objects (R.go/B.go). In your example, you don't need the COGNEW anyway since it is already provided in PWM1 and PWM2. Just use

PUB Vamos
  R.go
  B.go

BTX · 2006-12-26 18:10

Right Mike.
Just see my original code, it's like you said but doesn't work.

With :

PUB vamos
 
    R.go
    B.go

· Please !! read my first question to see all original code posted. Only R.go works and never return.
Regards
Alberto.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

Charlie Johnson · 2006-12-26 18:22

Could it be the repeat?

PUB Vamos
R.go
B.go

repeat <

Maybe try removing this.

Just a SWAG

Charlie

nutson · 2006-12-26 18:54

repeat
    repeat x from 0 to period          'linearly advance parameter from 0 to 100
      parameter := x                   'a constant here locks to value x percent high
      waitcnt(1_000_000 + cnt)         'wait a little while before next update

This is an endless loop,·you never return to the main program to execute the "B.GO" statement.

Nico

BTX · 2006-12-27 00:57

Charlie
I'd try that ...but stil wont work...

Nico.
You're right with the effect, but I'm starting a new cog !! in B.go...first "go" is not the same as 2nd.....

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

BTX · 2006-12-27 04:14

Original code wont work.......(at first post)

And this simple code works.

CON
  _CLKMODE = XTAL1 + PLL16X     'Set to ext low-speed crystal, 4x PLL
  _XINFREQ = 5_000_000         'Frequency on XIN pin is 5 MHz
OBJ
   R : "Pru1"
   B : "Pru2"
PUB Main
   R.go                       
   B.go

{Pru1.spin}
VAR
  long  Stack[noparse][[/noparse]9]                       'Stack space for new cog
PUB go
  cognew(Toggle, @Stack) 

PRI Toggle                       ' BLINK LED
  dira[noparse][[/noparse]16]~~                     'Set I/O pin to output direction 
  repeat                         'Repeat the following
    !outa[noparse][[/noparse]16]                    'Toggle I/O Pin 
    waitcnt(clkfreq / 10 + cnt)  'Wait

{Pru2.spin}
VAR
  long  Stack[noparse][[/noparse]9]                       'Stack space for new cog
PUB go
  cognew(Toggle, @Stack) 

PRI Toggle                       ' BLINK LED
  dira[noparse][[/noparse]23]~~                     'Set I/O pin to output direction 
  repeat                         'Repeat the following
    !outa[noparse][[/noparse]23]                    'Toggle I/O Pin
    waitcnt(clkfreq / 20 + cnt)  'Wait

In fact, the problem seems to be, the repeat loop into the original code, but it hasn't the same effect, into this last code.

When I call "R.go" in the spin version it start a new cog, and continue running from it in spin, then return to the Main method, but when I do the same with the assembler code, it wont return !!

WHY PLEASE !! How to solve this ???

My target is to fade two differents LED,·at different pins, and with different speed, just only to learn, for future objects.

Regards.

Alberto.

EDIT: All wired is fine, the last code works with the demo board LED, Pins 16 and 23, when the first post code, is executed, only one LED works fading.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

Envio editado por (BTX) : 12/27/2006 4:20:40 AM GMT

bambino · 2006-12-28 15:41

Mike,BTX,
I think what we are not seeing here is the fact that each oject is running a cognew(assembly with Stack space) AND some spin(without Stack space from the TopObject).
Also, each par return is going to write to one global variable"Paremater"

BTX,
Try moving:

· repeat
··· repeat x from 0 to period········· 'linearly advance parameter from 0 to 100
····· parameter := x·················· 'a constant here locks to value x percent high
····· waitcnt(1_000_000 + cnt)········ 'wait a little while before next update

: up to the top object and monitor your variables from there.

Post Edited (bambino) : 12/28/2006 4:35:07 PM GMT

BTX · 2006-12-28 17:14

Hi Bambino.

Thanks for the suggest, I did all test that I can imagine, and never got a correct work for this.
After 3 or 4 days trying, I was getting some frustrating, just do it, only for learn to deal with cogs, and some of asm code.
I thought that it could be a simple task...but nobody has solution for that problem...It seems that Parallax people is at "vacations" or similar and can't help...and, at this morning (here), I can solve it.
Just about a few minutes I get the correct work...I've now the solution....but I'm waiting to finish a complete task to post it code.
1- The problem solved code.
2- A demo code, of how to fade 3 LEDS (RGB) at differents rates, at different times, and at different pins.

Thanks again.
Regards.
Alberto.

PS: Chi vediamo dopo...

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

bambino · 2006-12-28 17:24

Glad to hear it,
And yes I am glad to hear from anyone actually, I thought I might have become a ghost writter.
Tell me, was it this

  repeat
    repeat x from 0 to period          'linearly advance parameter from 0 to 100
      parameter := x                   'a constant here locks to value x percent high
      waitcnt(1_000_000 + cnt)         'wait a little while before next update

causing your problem?

BTX · 2006-12-28 18:17

Hi Bambino.

Bambino said...
And yes I am glad to hear from anyone actually, I thought I might have become a ghost writter.

· hahaha...me too

·The answer is:

PUB start(Level,Pin) : Pass 
 
    Stop
    Pass := (Cog := cognew(Auto(Level,Pin), @Stack) + 1)
 
 
PRI Auto(Level,Pin)   | x
 
  cognew(@entry, @Level_Cog)
    Level_Cog := Level
    pin_cog   := Pin
 
  repeat
    repeat x from 0 to period                 'linearly advance level_cog from 0 to 100
      Level_Cog := x                          'a constant here locks to value x percent high
      waitcnt(1_000_000 + cnt)                'wait before next update
DAT

entry "asm code here"

So when you call "start", it returns, an let original spin code to continue running.

If you run Auto method like me "before", it never ends....simple but my confusion was that I have one cog for asm code and now I have one more to lunch it. In a minutes I'll be uploading final RGB LED fade code.

I could say that is a "mix" of many answers...that I have.

Be patient....here's is 15:15 and I've not lunch...still....

Regards.

Alberto.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

BTX · 2006-12-28 18:26

Hi all.

Here's the final working code !!!!

Regards.
ALberto.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

BTX · 2006-12-28 21:01

Hi.
Now it has a bit more to play, Using the unused variable, you can fix the bright at each LED.
Take a look at the code now and try it.

A good question for experts could be .....is it·possible to do the same ..but with no so many cog's ?? easily ........
It is too much, to fade 3 led, using 6 cog's...

I think to include the "repeat loops" into the assembler code.....but it sounds me, hard to do, while maintain the counters working.
And it seems a commonly serial programming....just not for propeller.

Regards.
Alberto.

Here's the code:

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

Mike Green · 2006-12-28 22:22

Alberto,
You are actually doing 6 independent functions, but one of them is being done by a CTRx register with your assembly routine doing the reloading of the PHSx register. It would be possible to combine the PWM and fading functions into one cog. You have:

:loop   rdlong value, par              'get an up to date pulse width
        waitcnt time, period           'wait until next period
        neg phsa, value                'back up phsa so that it
        jmp #:loop                     'loop for next cycle

You could count periods and decrease the count with each period (or something similar):

:get  rdlong value,par            ' Upper word has pulse width
        mov    delta,value         ' Lower word has period decrease
        shr      value,#16
        and     value,mask
        and     delta,mask
:loop waitcnt time,period       ' Wait for period
        neg     phsa,value         ' Restart pulse width
        sub     value,delta         ' Decrease pulse width
        tjnz     value,#:loop      ' Stop when width is zero
        jmp    #:get                 ' Start over again
mask long   $FFFF
period long  100
value res    1
delta  res    1
time  res    1

This only goes from on to off. You can add another loop that goes from off to on and stops when the width is the same value as supplied in the upper word of the long passed to the routine.

Mike Green · 2006-12-28 22:26

If you have only one value to pass to an assembly routine, PAR is convenient. If you want to pass several values or a "control block", you pass the address of the first value to the assembly program in PAR, then do arithmetic on that value to get to other parameters like:

begin     mov   temp,par         ' address of first long
             rdlong value1,temp
             add    temp,#4         ' address of 2nd long
             rdlong value2,temp
             add    temp,#4         ' address of 3rd long
'           ......                          ' more if you need it

BTX · 2006-12-29 00:58

Oh Mike...thanks so much.

I understand perfect your example.... I wrote that could be possible to do some similar, but at this point of my learn.. I wont, cause I was 3 or 4 days to get it work, "with 6 cogs".
(Hope you understand me).
In fact, I was discover your example method, to pass many variables to the asm code, but you write it generically very well !!
This is my first try in asm...I took the CTRs example and modify it a bit.. I'll be continue trying every time I'd have some time...

Thanks so much again
Alberto.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

ASM wont work...Help Please !!

Comments