|
|
Post by VDC 8x2 on Jul 19, 2014 18:48:51 GMT
Thought of a way to pass augments to a subroutine via the stack.
start lda #$03
pha ;put 3 on stack
jsr subroutine
pla ;clean up stack
rts
subroutine tsx
lda $103,x ;we got the value
sta $20 ;do something with a
rts
Any other stack tricks to try?
|
|
|
|
Post by hydrophilic on Jul 20, 2014 1:20:39 GMT
That's pretty nice and easy. One of the most practical ways to use the stack with limitations of a 6502.
Another slightly more complicated way allows you to indirectly pass a large a number of arguments without push and pop. It is the method used by KERNAL PrImm. Just JSR to a subroutine and place all your data after the JSR.
jsr MySub
.byte x1
.byte y1
.byte x2
.byte y2
lda #blue ;continue here... what ever
jsr color
...
MySub:
tsx
lda $101,x
sta $20
lda $102,x
sta $21
ldy #1,2,3,4 ;access any of the 4 bytes
lda ($20),y
nop ;do something with it
lda #4 ;#bytes on stack
tsx ;index bad 'return address'
clc
adc $101,x ;fix return address
sta $101,x
bcc +
inc $102,x
+ rts
Compared to your first method, it takes about 16 cycles setup + 17 cycles of clean-up (33 overhead), while your first method takes 9 cycles setup + 4 cleanup (13 overhead). Of course this method has constant overhead, while your method would need more cycles with more arguments... I'm guessing four or more bytes and this method would be better... Well lots of other trade-offs to consider too... but something to think about.
Only other stack thing I can think of now is thread-safe exchange registers...
;swap X and Y
pha
tya
pha
txa
tay
pla
tax
pla
;swap A and X
pha
pha
txa
tsx
sta $102,x
pla
tax
pla
;swap A and Y
pha
pha
txa
pha
tya
tsx
sta $103,x
pla
tax
pla
tay
pla
|
|
|
|
Post by VDC 8x2 on Jul 20, 2014 12:18:09 GMT
A way to do it without clean up. I think...
start tsx
dex
dex
dex
lda #01
sta $100,x
dex
lda #$02
sta $100,x
jsr subroutine
do stuff with having to clean up
subroutine tsx
lda $100,x
sta somewhere
dex
lda $100,x
sta someplace
do stuff
rts
|
|
|
|
Post by hydrophilic on Jul 20, 2014 13:03:34 GMT
I think you have too many DEX after start (3... should be 2?)
More importantly, the stuff you store at $100,x would get clobbered by an IRQ. If interrupts are impossible, then it may be useful...
Edit
Also, if you're subroutine pushed any data on the stack or called a subroutine, the data stored at $100,x by the original caller would get clobbered. Of course you could save the data at $100,x before doing any push or JSR.
|
|
|
|
Post by VDC 8x2 on Jul 20, 2014 21:35:12 GMT
what about using the mmu to make a separate stack for parameters.
would have the sei while working with separate stack.
|
|
|
|
Post by hydrophilic on Jul 21, 2014 0:26:23 GMT
Sure, that is what the MMU is good for  Just need to maintain a variable storing the stack pointer for your data stack. You shouldn't need to disable interrupts if properly written. But then, it is about as much work as a software stack... tsx
stx cpu_stack ;save 'real' / cpu stack pointer
lda #dummy ;for example $0b.... any unused memory page
sta $d509
ldx data_stk ;get 'data' stack pointer
txs
lda #data ;for example $0c, stack for data in $c00~cff
sta $d509
;repeat as much as you want
lda some_data
pha
;
lda #dummy
sta $d509
ldx cpu_stack
txs
lda #1 ;the CPU stack page
sta $d509
jsr subroutine
Well that is a bit verbose because it does not disable IRQs. Here is alternate version that does: tsx
stx cpu_stack ;save 'real' / cpu stack pointer
ldx data_stk ;get 'data' stack pointer
lda #data ;for example $0c, stack for data in $c00~cff
sei
txs
sta $d509
cli
;repeat as much as you want
lda some_data
pha
;
ldx cpu_stack
lda #1 ;the CPU stack page
sei
txs
sta $d509
cli
jsr subroutineNot shown is similar code the 'subroutine' would need to access the data stack. In summary, I don't this would be any better (probably slower) than using a software stack. I think the MMU method is better for multiple threads, because the MMU only gets updated on a task swap which happens far less frequently than calling subroutines. But maybe I am missing something? |
|
