Here is draft of a possible code. It compiles but it's not tested.
As you can see the Munching squares calculation is easy, just an XOR and an ADD.
The plot subroutine is a bit more challenging... and you have to pass the screen address to the PASM cog, fortunatly there is getter methode for that in the VGA code.
{{ Munching Squares - VGA }}
CON
_clkmode = xtal1 + pll16x
_xinfreq = 5_000_000
OBJ
pix: "VGA64.spin"
PUB demo
pix.PIXEngineStart(2) 'start VGA driver
waitcnt(clkfreq/100 + cnt) 'wait 10ms
cognew(@munch_asm, pix.displayPointer) 'start PASM cog
DAT
munch_asm mov screen,par 'pointer to sreenbuffer
mov mt,#0 'init t,x,y
loopt mov my,#0
loopy mov mx,#0 'loop over x,y
loopx mov mval,mx 'calc: val := (x ^ y) + t
xor mval,my
add mval,mt
call #plot
add mx,#1
cmp mx,mwidth wc
if_b jmp #loopx
add my,#1
cmp my,mheight wc
if_b jmp #loopy
add mt,#1
jmp #loopt
plot mov tmp,my 'calc screen address
shl my,#5
shl tmp,#7
add tmp,my 'my*32 + my*128 =^= mx*160
shr my,#5 'restore my
add tmp,mx '+mx
add tmp,screen '+screen base addr
shl mval,#2
or mval,#3 'color
wrbyte mval,tmp 'write into screen
plot_ret ret
tmp long 0
mt long 0
mx long 0
my long 0
mval long 0
mheight long 120
mwidth long 160
screen long 0
The propeller does not have multiply, divide, or modulo PASM instructions. You have to use subroutines for them.
What base modulo are you trying to do? If the base is constant, you can make a custom routine that will be a lot faster. If the base is a power of 2, it's as simple as ANDing your value with (base - 1).
You need to use a modulo subroutine. Propeller assembly instructions do not include more complex ones like multiply/divide/modulo. Those functions can be done with subroutines using adds/shifts/etc.
This is a simple divide subroutine from Chip. The remainder is the Modulo result, so you need to shift the result right by 16 (shr x,#16).
It works only with positive numbers and max. a 16bit divisor, but if you need it for graphics this should not be a problem.
' Divide x[31..0] by y[15..0] (y[16] must be 0)
' on exit, quotient is in x[15..0] and remainder is in x[31..16]
'
divide shl y,#15 'get divisor into y[30..15]
mov t,#16 'ready for 16 quotient bits
:loop cmpsub x,y wc 'if y =< x then subtract it, quotient bit into c
rcl x,#1 'rotate c into quotient, shift dividend
djnz t,#:loop 'loop until done
divide_ret ret 'quotient in x[15..0], remainder in x[31..16]
example: x=30 ,y=7
30-7=23, 23-7=16, 16-7=9, 9-7=2
Next cmpsub will not be approved for subtraction and c-flag will not be set, result modulo=2
modulo 'x original value, y the divider
:loop cmpsub x,y wc 'if y =< x then subtract it,
if_c jmp #:loop
modulo_ret ret 'x=modulo
As you don't need to know how many times, un-roll the loop a little will speed it up most of the time
modulo 'x original value, y the divider
:loop cmpsub x,y 'if y =< x then subtract it,
cmpsub x,y 'if y =< x then subtract it,
cmpsub x,y 'if y =< x then subtract it,
cmpsub x,y wc 'if y =< x then subtract it,
if_c jmp #:loop
modulo_ret ret 'x=modulo
Comments
As you can see the Munching squares calculation is easy, just an XOR and an ADD.
The plot subroutine is a bit more challenging... and you have to pass the screen address to the PASM cog, fortunatly there is getter methode for that in the VGA code.
{{ Munching Squares - VGA }} CON _clkmode = xtal1 + pll16x _xinfreq = 5_000_000 OBJ pix: "VGA64.spin" PUB demo pix.PIXEngineStart(2) 'start VGA driver waitcnt(clkfreq/100 + cnt) 'wait 10ms cognew(@munch_asm, pix.displayPointer) 'start PASM cog DAT munch_asm mov screen,par 'pointer to sreenbuffer mov mt,#0 'init t,x,y loopt mov my,#0 loopy mov mx,#0 'loop over x,y loopx mov mval,mx 'calc: val := (x ^ y) + t xor mval,my add mval,mt call #plot add mx,#1 cmp mx,mwidth wc if_b jmp #loopx add my,#1 cmp my,mheight wc if_b jmp #loopy add mt,#1 jmp #loopt plot mov tmp,my 'calc screen address shl my,#5 shl tmp,#7 add tmp,my 'my*32 + my*128 =^= mx*160 shr my,#5 'restore my add tmp,mx '+mx add tmp,screen '+screen base addr shl mval,#2 or mval,#3 'color wrbyte mval,tmp 'write into screen plot_ret ret tmp long 0 mt long 0 mx long 0 my long 0 mval long 0 mheight long 120 mwidth long 160 screen long 0you can copy and paste this into the Prop-Tool.Andy
Many many thanks this is working great :-)
And this is my start point to learn with pasm !
Simple question - is possible use modulo function in the pasm or must use modulo sub routine ?
Many thanks !
Kamil
What base modulo are you trying to do? If the base is constant, you can make a custom routine that will be a lot faster. If the base is a power of 2, it's as simple as ANDing your value with (base - 1).
Do exist simple modulo subroutine ?
It works only with positive numbers and max. a 16bit divisor, but if you need it for graphics this should not be a problem.
' Divide x[31..0] by y[15..0] (y[16] must be 0) ' on exit, quotient is in x[15..0] and remainder is in x[31..16] ' divide shl y,#15 'get divisor into y[30..15] mov t,#16 'ready for 16 quotient bits :loop cmpsub x,y wc 'if y =< x then subtract it, quotient bit into c rcl x,#1 'rotate c into quotient, shift dividend djnz t,#:loop 'loop until done divide_ret ret 'quotient in x[15..0], remainder in x[31..16]Andy
See my example XOR Fractal
30-7=23, 23-7=16, 16-7=9, 9-7=2
Next cmpsub will not be approved for subtraction and c-flag will not be set, result modulo=2
As you don't need to know how many times, un-roll the loop a little will speed it up most of the time
modulo 'x original value, y the divider :loop cmpsub x,y 'if y =< x then subtract it, cmpsub x,y 'if y =< x then subtract it, cmpsub x,y 'if y =< x then subtract it, cmpsub x,y wc 'if y =< x then subtract it, if_c jmp #:loop modulo_ret ret 'x=modulonice code. the second one is almost mean. love it!
Thanks!
Mike