Jump to content
Sign in to follow this  
Lillapojkenpåön

temp7's purpose?

Recommended Posts

Posted (edited)

What is temp7's purpose when bankswitching?

 

.L08 ;  goto _bank2 bank2

 

 sta temp7
 lda #>(._bank2-1)
 pha
 lda #<(._bank2-1)
 pha
 lda temp7
 pha
 txa
 pha
 ldx #2
 jmp BS_jsr

 

 

BS_jsr
 lda bankswitch_hotspot-1,x
 pla
 tax
 pla
 rts

 

It looks to me like it's just used to push what's currently in the accumulator to the stack, and that you also push x to the stack, is that just so you can continue your code in another bank? 

There's no way to ldx or lda (without also assigning it to a variable) in bB, and even if there where, wouldn't it be faster to just

sta temp1

stx temp2

and load them after the bankswitching

lda temp1

ldx temp2

 

If I don't need to keep what's in a and x would this work

 lda #>(._bank2-1)
 pha
 lda #<(._bank2-1)
 pha

 ldx #2
 lda bankswitch_hotspot-1,x
 rts

 

Just tried it and I got error: Unknown Mnemonic 'bankswitch_hotspot-1,x'.

Is bankswitch_hotspot a table? I can't find it anywhere.

Edited by Lillapojkenpåön

Share this post


Link to post
Share on other sites
Posted (edited)

Woops answered here 

But I still need to know how to put the hotspots in a table or something?

 

hotspots

  $FFF6

  $FFF7

  $FFF8

  $FFF9

  $FFFA

  $FFFB

 

I tried this  

lda $fff6,x

and -1 to but didn't work

Edited by Lillapojkenpåön

Share this post


Link to post
Share on other sites

The specific hotspot for bankswitching depends on the size of your project, so using the bankswitch_hotspot constant instead of a hardcoded value is a good idea. I don't know why the stack is used to preserve register values instead of temp variables; maybe @RevEng can answer that part.

 

Hotspots are not in a table. They are addresses that trigger a bankswitch in the hardware if they are accessed (read from or written to). If you just do a "lda bankswitch_hotspot" the code would switch to the first bank. Each address following bankswitch_hotspot would switch to the following bank.  The X register contains the bank you are switching to.

 

bB starts counting banks at 1 instead of 0, so that's why it uses "lda bankswitch_hotspot-1,x".

  • Thanks 1

Share this post


Link to post
Share on other sites

it appears to work with just a little trickery

   set romsize 16k
   set optimization inlinerand 

   const targetlo = <(.dloop-1)
   const targethi = >(.dloop-1)

   def gt_bank = callmacro bsm


  macro bsm
  asm
  lda #{2}
  pha
  lda #{1}
  pha
  ldx {3}
  lda bankswitch_hotspot-1,x
  jmp BS_jsr+6
end
end


  q = $55

  gt_bank targetlo targethi 2


 bank 2

 a = a

dloop
  c = c + 1
  if !c & $7F then q = q ^ $FF : COLUBK = q

  if e && joy0fire then q = q + 4 : COLUBK = q
  if joy0fire then e = 0 else e = 1

  drawscreen
  goto dloop


 bank 3

 bank 4


 

Share this post


Link to post
Share on other sites

The intent of temp7 is indeed to save save A. As to the question of why other temp variables weren't used, I believe the intent was to eventually bankswitch-enable certain bB functions that are already using the temps. e.g. the div_mul libraries.

  • Thanks 1

Share this post


Link to post
Share on other sites


if I do this

  def gt_bank = callmacro bsm


  macro bsm
  asm
  lda #>(.{1}-1)
  pha
  lda #<(.{1}-1)
  pha
  ldx {2}
  lda bankswitch_hotspot-1,x
  jmp BS_jsr+6
end
end


  gt_bank dloop 2

it complains about uresolved symbol

1.dloop

I don't know where that comes from

it seems to produce correct code but won't assemble

     90  1004                   .L08         ;  gt_bank dloop 2

     91  1004
      0  1004                          bsm    dloop, #2,
      1  1004
      2  1004                   .L06
      3  1004
      4  1004               a9 ff              lda    #>(.dloop-1)
      5  1006
      6  1006               48              pha
      7  1007
      8  1007               a9 ff              lda    #<(.dloop-1)
      9  1009
     10  1009               48              pha
     11  100a
     12  100a               a2 02              ldx    #2
     13  100c
     14  100c               bd f5 1f           lda    bankswitch_hotspot-1,x
     15  100f
     16  100f               4c f1 ff           jmp    BS_jsr+6


copy and paste errror in the previous code (although it seems to work)

should have been

  macro bsm
  asm
  lda {2}
  pha
  lda {1}
  pha
  ldx {3}
  lda bankswitch_hotspot-1,x
  jmp BS_jsr+6
end
end

Share this post


Link to post
Share on other sites
17 minutes ago, bogax said:

It complains about uresolved symbol

1.dloop

I don't know where that comes from

it seems to produce correct code but won't assemble

Dasm considers dot symbols to be limited in scope. I'm pretty sure that limits the dasm in-macro access to bB labels.

  • Like 1

Share this post


Link to post
Share on other sites
18 minutes ago, RevEng said:

Dasm considers dot symbols to be limited in scope. I'm pretty sure that limits the dasm in-macro access to bB labels.


well in that case

   set romsize 16k
   set optimization inlinerand
   set optimization noinlinedata

   def gt_bank = callmacro bsm


  macro bsm
  asm
  lda #>({1}-1)
  pha
  lda #<({1}-1)
  pha
  ldx {2}
  lda bankswitch_hotspot-1,x
  jmp BS_jsr+6
end
end


  q = $55

  gt_bank dloop 2


 bank 2

 a = a

  data dloop
end

dloop
  c = c + 1
  if !c & $7F then q = q ^ $FF : COLUBK = q

  if e && joy0fire then q = q + 4 : COLUBK = q
  if joy0fire then e = 0 else e = 1


  drawscreen
  goto dloop


 bank 3

 bank 4

(yes , it works)

  • Haha 1

Share this post


Link to post
Share on other sites

IObBCC..

I've read several times that macros are slow, so I guess when using it you don't input a copy of that code with the parameters you give?

Or if you do does that happen when it turns into machine language?

 

18 hours ago, Karl G said:

The specific hotspot for bankswitching depends on the size of your project, so using the bankswitch_hotspot constant instead of a hardcoded value is a good idea.

 

Hotspots are not in a table. They are addresses that trigger a bankswitch in the hardware if they are accessed (read from or written to). 

 

bB starts counting banks at 1 instead of 0, so that's why it uses "lda bankswitch_hotspot-1,x".

I couldn't use the bankswitch_hotspot constant before without getting an error, now I can for some reason.

maybe the problem is that I'm trying to jump to another bank and return to the old bank again, on gosub with bankswitch, which bB doesn't support, this is the first part

 

 asm

   lda #>(._ongosub_return_-1)

   PHA

   lda #<(._ongosub_return_-1)

   PHA

 

   LDX bankNumber   ; 4 in this case

   LDA _jumptablehi_,x

   PHA

   LDA _jumptablelo_,x

   PHA

 

   lda bankswitch_hotspot-1,x

   rts

 

_jumptablehi_

   .byte #0

   .byte #0

   .byte #0

   .byte >(._bank3data-1)

   .byte >(._bank4data-1)

   .byte >(._bank5data-1)

   .byte >(._bank6data-1)

_jumptablelo_

   .byte #0

   .byte #0

   .byte #0

   .byte <(._bank3data-1)

   .byte <(._bank4data-1)

   .byte <(._bank5data-1)

   .byte <(._bank6data-1)

end

_ongosub_return_

 

 

I probably need to make a custom return otherbank code also, 

what do I change? getting rid of the a and x stuff didn't work

 

 asm

 pha ;push a to stack

 txa

 pha ; push x to stack

 tsx ; Transfer Stack ptr to X

 

 if bankswitch != 64

   lda 4,x ; get high byte of return address

 

   rol

   rol

   rol

   rol

   and #bs_mask ;1 3 or 7 for F8/F6/F4

   tax

   inx

 else

   lda 4,x ; get high byte of return address

   tay

   ora #$10 ; change our bank nibble into a valid rom mirror

   sta 4,x

   tya

   lsr 

   lsr 

   lsr 

   lsr 

   tax

   inx

 endif


 

 lda bankswitch_hotspot-1,x

 pla

 tax ; *** restore X

 pla ; *** restore A

 rts ; *** RTS will "return" to the address we pushed onto the stack earlier

end

Share this post


Link to post
Share on other sites
Posted (edited)

If someone can get this to work, I would be happy

you gosub to bank 4 by pressing joy0fire,

it should change bgcolor to blue

bankswitch2.bas

 

and is the  jmp BS_jsr+6 in bogax code because you have to do the rts at that address?

Edited by Lillapojkenpåön

Share this post


Link to post
Share on other sites

it doesn't work as advertised

the BS_jsr+6 comes after the bank switch and so never executes

it just so happens that the bank switch ends up in the dloop

I haven't figured out how to fix it

 

Share this post


Link to post
Share on other sites

if you're just trying to compute your target you can set something up with on gosub and on goto

put gotos for each bank (or each possible target bank) in each bank tha you might be coming from

and a dispatch on goto for each possible target in the bank

in other words on gosub to a goto bank (in the current bank) that goes to an on goto in the target bank that selects the routine you want in that bank

Share this post


Link to post
Share on other sites

I did that, but think I did on gosub to gosub to another bank, so I could return, then I just merged the codes in assembly

 

 sta temp7

   lda #>(._ongosub_return_-1)

   PHA

   lda #<(._ongosub_return_-1)

   PHA

   LDX bankNumber

   LDA _jumptablehi_,x

   PHA

   LDA _jumptablelo_,x

   PHA

 lda temp7

 pha

 txa

 pha

 ldx bankNumber

 jmp BS_jsr

 

this works perfectly with a regular return otherbank

So I'm not trying to get it to work, I'm trying to optimize it by removing as much excess as possible and hopefully get a little bit closer to understanding bankswitching in the process.

 

I don't need to save a and x, but BS_jsr and BS_return (that falls into BS_JSR) tries to pull them from the stack, I think that's what the problem is.

Share this post


Link to post
Share on other sites

this works, but it's getting messy

the to_bank macro just loads the target address to the stack

you then have to goto the bank switch code at the beginning of the bank
the location of the bank switch code has to be the same in all banks
and the beginning of the bank seemed simplest

since you have to jump around it in the first bank there's padding in the rest
(in place of the jump/goto)

   set romsize 16k
   set optimization inlinerand
   set optimization noinlinedata

   def to_bank = callmacro bsm
   def go_to = callmacro gtom


  macro bsm
  asm
  lda #>({2}-1)
  pha
  lda #<({2}-1)
  pha
  ldx {1}
end
end

  macro gtom
  asm
  jmp {1}
end
end

  def bswitch = callmacro bswm
  macro bswm
  asm
  nop
  nop
  nop
  lda bankswitch_hotspot-1,x
  rts
end
end

  goto skip_BS_code

bs1
  asm
  lda bankswitch_hotspot-1,x
  rts
end

skip_BS_code

  gosub init

  to_bank 2 dloop
  goto bs1

init
  q = $55 
  COLUBK = q
  COLUPF = $33
  return thisbank

 bank 2

bs2
  bswitch

  data dloop
end

  c = c + 1
  if !c & $7F then q = q ^ $FF : COLUBK = q

  if e && joy0fire then q = q + 4 : COLUBK = q
  if joy0fire then e = 0 else e = 1


  drawscreen
  go_to dloop


 bank 3

 bank 4

  • Thanks 1

Share this post


Link to post
Share on other sites

Aah, now I get it, ofcourse that code has to be in the same place in all banks 💡

Thanks, it doesn't support returning but I can add that.

 

I'm also trying to save space and I got my previous code down to this

   lda #>(._ongosub_return2_-1)

   PHA

   lda #<(._ongosub_return2_-1)

   PHA

   LDX bankNumber

   LDA _jumptablehi_,x

   PHA

   LDA _jumptablelo_,x

   PHA

 

 pha

 pha

 

 jmp BS_jsr

 

just two extra pushes to get the existing bs code to work and return, not shure it would be worth adding another slightly modified version to every bank just to get rid of the extra pushing and pulling, but I sure will try it.

What the hell is a macro anyway?

Share this post


Link to post
Share on other sites

you're indexing your table with the bank number(?)

 

in this case a macro is sort of like an inlined function

 

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...