Assembly programming problem with CNT

HGPL

Hi,
I worry about the code below:

                        mov     wcol,#1
                        shl     wcol,#28
...
                        call    #waitcol                   or       call #waitcol2
...
waitcol2
                        mov     tnextcol,cnt
                        add     tnextcol,wcol                        
                        cmp     cnt,tnextcol wc   'c=1; cnt < tnextcol
                 if_c   jmp     waitcol2
waitcol2_RET            ret


waitcol                 
                        mov     clk,cnt
                        add     clk,wcol
                        waitcnt clk,wcol
waitcol_ret             ret

The waitcol is working. The waitcol2 not working...
I think make the fixrate wait with a pin verifying capability, and the waitcnt not good here...

Anybody know help for me...?

thanks, Gabor

Post Edited (HGPL) : 11/6/2007 4:18:41 PM GMT

deSilva

waitcol2
       mov   tnextcol, cnt
       add   tnextcol, wcol
       cmp   cnt,      tnextcol wc    'c=1; cnt < tnextcol
  if_c jmp   waitcol2
waitcol2_RET
       ret

(1) JMP #waitcol2B
(2) Set the label waitcol2B before the CMP; you have I fine moving target with your original code

HGPL

Very thanks... [noparse]:)[/noparse] last hour was very poor for me...
HVALA
The # is a very high blame... [noparse]:)[/noparse]

but... [noparse]:)[/noparse]

'first column delay init 
                        mov     wcol,#1
                        shl     wcol,#19
                        mov     tnextcol,cnt
                        add     tnextcol,wcol
...
                        call    #waitcol2
...
waitcol2
                        cmp     cnt,tnextcol wc   'c=1; cnt < nextcnt
                 if_c   jmp     #waitcol2                 
                        mov     tnextcol,cnt
                        add     tnextcol,wcol                                         
waitcol2_RET            ret

... and not working yet...

Post Edited (HGPL) : 11/6/2007 4:37:53 PM GMT

HGPL

huh... The newest solution is below...
It is working, but I don't know why need the T1 temporarily variable...
Is CNT not working with CMP command?

waitcol2
                        mov     t1,cnt
                        cmp     t1,tnextcol wc   'c=1; cnt < nextcnt
                 if_c   jmp     #waitcol2
                 
                        mov     tnextcol,t1
                        add     tnextcol,wcol                        
                 
waitcol2_RET            ret

deSilva

Sorry, I overlooked it!

The rule is: Never use a "special register" at the DEST operand unless you exactly know what you are doing..

The reason: You will not read:
- CNT
- INA
- PAR
- PHSA
- PHSB

as expected!
Rather you will read what you last have written to it. This is well know as "The Behaviour Of The Shadow Registers"

You can fix it in your second example by just reversing CNT and TNEXTCOL and jumping IF_NC

Another note: Once every 40 to 50 seconds your program will fail, when the CNT overflows from $FFFF_FFFF to $0

HGPL

Very Thanks!
I will read deep the documentation [noparse]:)[/noparse] thanks.

My setup is _CLKMODE = RCFAST
The fail wait about 6 minutes and i will correct the code but this mistake not matter for my app.

thanks

Beau Schwabe

HGPL,
You can avoid using the temp variable 't1' if you use cmpsub instead of cmp

waitcol2
                        cmpsub  tnextcol, cnt    wc, [color=#0]nr[/color]   'c=1 if tnextcnt is greater than or equal to cnt
                 if_c   jmp     #waitcol2
                 
                        mov     tnextcol, cnt
                        add     tnextcol, wcol                        
                 
waitcol2_RET            ret

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe

IC Layout Engineer
Parallax, Inc.

Post Edited (Beau Schwabe (Parallax)) : 11/7/2007 7:55:08 PM GMT

HGPL

thanks ,

but this last code is not working...
I read the CMPSUB doc, and I think the code is ok, but the CNT usage is not working with this command...

Post Edited (HGPL) : 11/6/2007 6:06:27 PM GMT

deSilva

Don't get confused!

(1) You have to read the "Erratum" www.parallax.com/dl/docs/prod/prop/PMv1.0Supplement-v1.3.pdf

The description of the CMPSUB instruction is not only absolutely confusing in the Manual, it is WRONG!

(2) You don't need CMPSUB! Your program will work fine, when you just reverse the operands of CMP and JUMP if NC, as I described above...

BTW: The Eratum 1.3 now contains detailed hints wrt the "Shadow Registers"

Edited...

Post Edited (deSilva) : 11/6/2007 11:09:41 PM GMT

HGPL

thanks ok