mirror of
https://github.com/deater/dos33fsprogs.git
synced 2024-12-26 11:30:12 +00:00
60 lines
1.7 KiB
ArmAsm
60 lines
1.7 KiB
ArmAsm
; Fast mutiply
|
|
|
|
; Note for our purposes we only care about 8.8 x 8.8 fixed point
|
|
; with 8.8 result, which means we only care about the middle two bytes
|
|
; of the 32 bit result. So we disable generation of the high and low byte
|
|
; to save some cycles.
|
|
|
|
;
|
|
; The old routine took around 700 cycles for a 16bitx16bit=32bit mutiply
|
|
; This routine, at an expense of 2kB of looku tables, takes around 250
|
|
; If you reuse a term the next time this drops closer to 200
|
|
|
|
; This routine was described by Stephen Judd and found
|
|
; in The Fridge and in the C=Hacking magazine
|
|
; http://codebase64.org/doku.php?id=base:seriously_fast_multiplication
|
|
|
|
; The key thing to note is that
|
|
; (a+b)^2 (a-b)^2
|
|
; a*b = ------- - --------
|
|
; 4 4
|
|
; So if you have tables of the squares of 0..511 you can lookup and subtract
|
|
; instead of multiplying.
|
|
|
|
; Table generation: I:0..511
|
|
; square1_lo = <((I*I)/4)
|
|
; square1_hi = >((I*I)/4)
|
|
; square2_lo = <(((I-255)*(I-255))/4)
|
|
; square2_hi = >(((I-255)*(I-255))/4)
|
|
|
|
|
|
; Fast 8x8 bit unsigned multiplication, 16-bit result
|
|
; Input: M1xM2
|
|
; Result: M2:M1
|
|
;
|
|
multiply_u8x8:
|
|
stx TEMP
|
|
|
|
lda M1
|
|
sta sm1a+1 ; 3
|
|
sta sm3a+1 ; 3
|
|
eor #$ff ; invert the bits for subtracting ; 2
|
|
sta sm2a+1 ; 3
|
|
sta sm4a+1 ; 3
|
|
|
|
ldx M2
|
|
sec
|
|
sm1a:
|
|
lda square1_lo,X
|
|
sm2a:
|
|
sbc square2_lo,X
|
|
sta M1
|
|
sm3a:
|
|
lda square1_hi,X
|
|
sm4a:
|
|
sbc square2_hi,X
|
|
sta M2
|
|
|
|
ldx TEMP
|
|
rts
|