mul6 in assembler?

[Heaven/TQA]

how would you do a*6 in 6502?

 

i need that to calc an offset for sprite background buffer... i have to expand from *4 to *6 but do not want to go for *8 instead...

 

any quick code to do that?

[Bruce Tomlin]

Multiplying by a constant is just a series of shifts and adds.

 

To multiply by 6, you add *2 and *4. So it's shift, save, shift, add.

 

ASLA (A=A*2)

STA temp (save A*2)

ASLA (A=A*4)

ADC temp (A*4+A*2=A*6, and assuming that the carry was clear from the previous ASLA)

 

Multiply by ten is *2 + *8.

 

ASLA

STA temp

ASLA

ASLA

ADC temp

 

Of course it gets a bit trickier when your result is greater than 255. On the Z80, 16-bit constant multiplies are easy when you use a bunch of ADD HL,HL and ADD HL,DE instructions, but on the 6502 it's a lot of extra work to do 16-bit shifts and adds, and you will probably need 4 bytes of temp.

Edited July 5, 2007 by Bruce Tomlin

[Rybags]

Shift left to get 2*, keep result

Shift left again to get 4*, add to result

 

How big will the initial number be? If it's a single byte, then maybe table lookup would be economical.

 

Or maybe, another way of looking at it:

TAY ; save original number

AND #$3F

TAX

LDA MUL6TABL,X

STA RESULT

LDA #0

CPX #43 ; 6*43 is an overflow, so we set the high byte to 1

BCC NOTBIG

LDA #1

NOTBIG STA RESULT+1

TYA ; get number back

AND #$C0

ROL A

ROL A

ROL A

TAX

LDA MUL6TABH,X

ADC RESULT+1

STA RESULT+1

RTS

MUL6TABL .byte 0,6,12,18 ... ; etc... table has 64 entries, each one = (N*6) AND 255

MUL6TABH .byte 0,1,3,3 ; 4 entries

 

I think that should work. Could be optimised a bit.

 

Cost - about 100 or so bytes.

Time - might just be quicker to do the shift method

Edited July 5, 2007 by Rybags

[Rybags]

TAY ; save original number

AND #$3F

TAX

LDA MUL6TABL,X

STA RESULT

TYA ; get number back

AND #$C0

ROL A

ROL A

ROL A

TAY

LDA MUL6TABH,Y

CPX #43

ADC #0

STA RESULT+1

RTS

MUL6TABL .byte 0,6,12,18 ... ; etc... table has 64 entries, each one = (N*6) AND 255

MUL6TABH .byte 0,1,3,3 ; 4 entries

 

optimized a bit - CPX sets the carry to what we want for the add anyway, so saved some time.

[Heaven/TQA]

thanks guys... the result will be not greater than 255 as my sprite buffer is one page... so i can handle 42 sprites with that...which i never will need...