prog8/compiler/res/prog8lib/math.p8

488 lines
12 KiB
Plaintext
Raw Normal View History

2021-01-07 00:56:31 +00:00
; Internal Math library routines - always included by the compiler
2023-10-27 20:36:43 +00:00
; note: some functions you might expect here are builtin functions,
; such as abs, sqrt, clamp, min, max for example.
2019-07-29 21:11:13 +00:00
math {
%option no_symbol_prefixing, ignore_unused
2023-06-29 22:29:50 +00:00
%asminclude "library:math.asm"
asmsub sin8u(ubyte angle @A) clobbers(Y) -> ubyte @A {
%asm {{
tay
lda _sinecos8u,y
rts
_sinecos8u .byte trunc(128.0 + 127.5 * sin(range(256+64) * rad(360.0/256.0)))
}}
}
asmsub cos8u(ubyte angle @A) clobbers(Y) -> ubyte @A {
%asm {{
tay
lda sin8u._sinecos8u+64,y
rts
}}
}
asmsub sin8(ubyte angle @A) clobbers(Y) -> byte @A {
%asm {{
tay
lda _sinecos8,y
rts
_sinecos8 .char trunc(127.0 * sin(range(256+64) * rad(360.0/256.0)))
}}
}
asmsub cos8(ubyte angle @A) clobbers(Y) -> byte @A {
%asm {{
tay
lda sin8._sinecos8+64,y
rts
}}
}
asmsub sinr8u(ubyte radians @A) clobbers(Y) -> ubyte @A {
%asm {{
tay
lda _sinecosR8u,y
rts
_sinecosR8u .byte trunc(128.0 + 127.5 * sin(range(180+45) * rad(360.0/180.0)))
}}
}
asmsub cosr8u(ubyte radians @A) clobbers(Y) -> ubyte @A {
%asm {{
tay
lda sinr8u._sinecosR8u+45,y
rts
}}
}
asmsub sinr8(ubyte radians @A) clobbers(Y) -> byte @A {
%asm {{
tay
lda _sinecosR8,y
rts
_sinecosR8 .char trunc(127.0 * sin(range(180+45) * rad(360.0/180.0)))
}}
}
asmsub cosr8(ubyte radians @A) clobbers(Y) -> byte @A {
%asm {{
tay
lda sinr8._sinecosR8+45,y
rts
}}
}
asmsub rnd() -> ubyte @A {
%asm {{
jmp math.randbyte
}}
}
asmsub rndw() -> uword @AY {
%asm {{
jmp math.randword
}}
}
asmsub rndseed(uword seed1 @AY, uword seed2 @R0) clobbers(A,Y) {
; -- set new pseudo RNG's seed values. Defaults are: $00c2, $1137
%asm {{
sta math.randword.x1
sty math.randword.c1
lda cx16.r0L
sta math.randword.a1
lda cx16.r0H
sta math.randword.b1
rts
}}
}
2023-06-16 22:43:23 +00:00
2023-08-31 22:42:15 +00:00
asmsub log2(ubyte value @A) -> ubyte @Y {
%asm {{
2023-08-31 22:50:24 +00:00
sta P8ZP_SCRATCH_B1
lda #$80
2023-08-31 22:42:15 +00:00
ldy #7
- bit P8ZP_SCRATCH_B1
beq +
rts
+ dey
bne +
rts
2023-08-31 22:50:24 +00:00
+ lsr a
2023-08-31 22:42:15 +00:00
bne -
}}
}
asmsub log2w(uword value @AY) -> ubyte @Y {
%asm {{
sta P8ZP_SCRATCH_W1
sty P8ZP_SCRATCH_W1+1
lda #<$8000
sta cx16.r0
lda #>$8000
sta cx16.r0+1
ldy #15
- lda P8ZP_SCRATCH_W1
and cx16.r0
sta P8ZP_SCRATCH_B1
lda P8ZP_SCRATCH_W1+1
and cx16.r0+1
ora P8ZP_SCRATCH_B1
beq +
rts
+ dey
bne +
rts
+ lsr cx16.r0+1
ror cx16.r0
jmp -
}}
}
2023-06-16 22:43:23 +00:00
asmsub mul16_last_upper() -> uword @AY {
2023-09-28 01:18:49 +00:00
; This routine peeks into the internal 32 bits multiplication result buffer of the
; 16*16 bits multiplication routine, to fetch the upper 16 bits of the last calculation.
2023-09-28 01:18:49 +00:00
; Notes:
; - to avoid interference it's best to fetch and store this value immediately after the multiplication expression.
; for instance, simply printing a number may already result in new multiplication calls being performed
; - not all multiplications in the source code result in an actual multiplication call:
; some simpler multiplications will be optimized away into faster routines. These will not set the upper 16 bits at all!
%asm {{
lda multiply_words.result+2
ldy multiply_words.result+3
rts
}}
}
2023-06-17 13:36:53 +00:00
sub direction_sc(byte x1, byte y1, byte x2, byte y2) -> ubyte {
2023-06-16 22:43:23 +00:00
; From a pair of signed coordinates around the origin, calculate discrete direction between 0 and 23 into A.
cx16.r0L = 3 ; quadrant
cx16.r1sL = x2-x1 ; xdelta
if_neg {
cx16.r0L--
cx16.r1sL = -cx16.r1sL
}
cx16.r2sL = y2-y1 ; ydelta
if_neg {
cx16.r0L-=2
cx16.r2sL = -cx16.r2sL
}
2023-06-17 13:36:53 +00:00
return direction_qd(cx16.r0L, cx16.r1L, cx16.r2L)
2023-06-16 22:43:23 +00:00
}
2023-06-17 13:36:53 +00:00
sub direction(ubyte x1, ubyte y1, ubyte x2, ubyte y2) -> ubyte {
2023-06-16 22:43:23 +00:00
; From a pair of positive coordinates, calculate discrete direction between 0 and 23 into A.
cx16.r0L = 3 ; quadrant
if x2>=x1 {
cx16.r1L = x2-x1
} else {
cx16.r1L = x1-x2
cx16.r0L--
}
if y2>=y1 {
cx16.r2L = y2-y1
} else {
cx16.r2L = y1-y2
cx16.r0L -= 2
}
2023-06-17 13:36:53 +00:00
return direction_qd(cx16.r0L, cx16.r1L, cx16.r2L)
2023-06-16 22:43:23 +00:00
}
2023-06-17 13:36:53 +00:00
asmsub direction_qd(ubyte quadrant @A, ubyte xdelta @X, ubyte ydelta @Y) -> ubyte @A {
2023-06-16 22:43:23 +00:00
;Arctan https://github.com/dustmop/arctan24
; From a pair of X/Y deltas (both >=0), and quadrant 0-3, calculate discrete direction between 0 and 23 into A.
; .reg:a @in quadrant Number 0 to 3.
; .reg:x @in x_delta Delta for x direction.
; .reg:y @in y_delta Delta for y direction.
; Returns A as the direction (0-23).
%asm {{
x_delta = cx16.r0L
y_delta = cx16.r1L
quadrant = cx16.r2L
half_value = cx16.r3L
region_number = cx16.r4L
small = cx16.r5L
large = cx16.r5H
sta quadrant
sty y_delta
stx x_delta
cpx y_delta
bcs _XGreaterOrEqualY
_XLessY:
lda #16
sta region_number
stx small
sty large
bne _DetermineRegion
_XGreaterOrEqualY:
lda #0
sta region_number
stx large
sty small
_DetermineRegion:
; set A = small * 2.5
lda small
lsr a
sta half_value
lda small
asl a
bcs _SmallerQuotient
clc
adc half_value
bcs _SmallerQuotient
cmp large
bcc _LargerQuotient
; S * 2.5 > L
_SmallerQuotient:
; set A = S * 1.25
lsr half_value
lda small
clc
adc half_value
cmp large
bcc _Region1 ; if S * 1.25 < L then goto Region1 (L / S > 1.25)
bcs _Region0 ; (L / S < 1.25)
; S * 2.5 < L
_LargerQuotient:
; set A = S * 7.5
lda small
asl a
asl a
asl a
bcs _Region2
sec
sbc half_value
cmp large
bcc _Region3 ; if S * 7.5 < L then goto Region3 (L / S > 7.5)
jmp _Region2 ; (L / S < 7.5)
_Region0:
; L / S < 1.25. d=3,9,15,21
jmp _LookupResult
_Region1:
; 1.25 < L / S < 2.5. d=2,4,8,10,14,16,20,22
lda region_number
clc
adc #4
sta region_number
bpl _LookupResult
_Region2:
; 2.5 < L / S < 7.5. d=1,5,7,11,13,17,19,23
lda region_number
clc
adc #8
sta region_number
bpl _LookupResult
_Region3:
; 7.5 < L / S. d=0,6,12,18
lda region_number
clc
adc #12
sta region_number
_LookupResult:
lda quadrant
clc
adc region_number
tax
lda _quadrant_region_to_direction,x
rts
_quadrant_region_to_direction:
.byte 9, 3,15,21
.byte 10, 2,14,22
.byte 11, 1,13,23
.byte 12, 0,12, 0
.byte 9, 3,15,21
.byte 8, 4,16,20
.byte 7, 5,17,19
.byte 6, 6,18,18
}}
}
2023-06-18 11:05:36 +00:00
asmsub atan2(ubyte x1 @R0, ubyte y1 @R1, ubyte x2 @R2, ubyte y2 @R3) -> ubyte @A {
2023-06-16 22:43:23 +00:00
;; Calculate the angle, in a 256-degree circle, between two points into A.
;; The points (x1, y1) and (x2, y2) have to use *unsigned coordinates only* from the positive quadrant in the carthesian plane!
;; https://www.codebase64.org/doku.php?id=base:8bit_atan2_8-bit_angle
;; This uses 2 large lookup tables so uses a lot of memory but is super fast.
%asm {{
x1 = cx16.r0L
y1 = cx16.r1L
x2 = cx16.r2L
y2 = cx16.r3L
octant = cx16.r4L ;; temporary zeropage variable
lda x1
2023-06-17 13:36:53 +00:00
sec
2023-06-16 22:43:23 +00:00
sbc x2
bcs *+4
eor #$ff
tax
rol octant
lda y1
2023-06-17 13:36:53 +00:00
sec
2023-06-16 22:43:23 +00:00
sbc y2
bcs *+4
eor #$ff
tay
rol octant
lda log2_tab,x
2023-06-17 13:36:53 +00:00
sec
2023-06-16 22:43:23 +00:00
sbc log2_tab,y
bcc *+4
eor #$ff
tax
lda octant
rol a
and #%111
tay
lda atan_tab,x
eor octant_adjust,y
rts
octant_adjust
.byte %00111111 ;; x+,y+,|x|>|y|
.byte %00000000 ;; x+,y+,|x|<|y|
.byte %11000000 ;; x+,y-,|x|>|y|
.byte %11111111 ;; x+,y-,|x|<|y|
.byte %01000000 ;; x-,y+,|x|>|y|
.byte %01111111 ;; x-,y+,|x|<|y|
.byte %10111111 ;; x-,y-,|x|>|y|
.byte %10000000 ;; x-,y-,|x|<|y|
;;;;;;;; atan(2^(x/32))*128/pi ;;;;;;;;
atan_tab
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$00,$00,$00
.byte $00,$00,$00,$00,$00,$01,$01,$01
.byte $01,$01,$01,$01,$01,$01,$01,$01
.byte $01,$01,$01,$01,$01,$01,$01,$01
.byte $01,$01,$01,$01,$01,$01,$01,$01
.byte $01,$01,$01,$01,$01,$02,$02,$02
.byte $02,$02,$02,$02,$02,$02,$02,$02
.byte $02,$02,$02,$02,$02,$02,$02,$02
.byte $03,$03,$03,$03,$03,$03,$03,$03
.byte $03,$03,$03,$03,$03,$04,$04,$04
.byte $04,$04,$04,$04,$04,$04,$04,$04
.byte $05,$05,$05,$05,$05,$05,$05,$05
.byte $06,$06,$06,$06,$06,$06,$06,$06
.byte $07,$07,$07,$07,$07,$07,$08,$08
.byte $08,$08,$08,$08,$09,$09,$09,$09
.byte $09,$0a,$0a,$0a,$0a,$0b,$0b,$0b
.byte $0b,$0c,$0c,$0c,$0c,$0d,$0d,$0d
.byte $0d,$0e,$0e,$0e,$0e,$0f,$0f,$0f
.byte $10,$10,$10,$11,$11,$11,$12,$12
.byte $12,$13,$13,$13,$14,$14,$15,$15
.byte $15,$16,$16,$17,$17,$17,$18,$18
.byte $19,$19,$19,$1a,$1a,$1b,$1b,$1c
.byte $1c,$1c,$1d,$1d,$1e,$1e,$1f,$1f
;;;;;;;; log2(x)*32 ;;;;;;;;
log2_tab
.byte $00,$00,$20,$32,$40,$4a,$52,$59
.byte $60,$65,$6a,$6e,$72,$76,$79,$7d
.byte $80,$82,$85,$87,$8a,$8c,$8e,$90
.byte $92,$94,$96,$98,$99,$9b,$9d,$9e
.byte $a0,$a1,$a2,$a4,$a5,$a6,$a7,$a9
.byte $aa,$ab,$ac,$ad,$ae,$af,$b0,$b1
.byte $b2,$b3,$b4,$b5,$b6,$b7,$b8,$b9
.byte $b9,$ba,$bb,$bc,$bd,$bd,$be,$bf
.byte $c0,$c0,$c1,$c2,$c2,$c3,$c4,$c4
.byte $c5,$c6,$c6,$c7,$c7,$c8,$c9,$c9
.byte $ca,$ca,$cb,$cc,$cc,$cd,$cd,$ce
.byte $ce,$cf,$cf,$d0,$d0,$d1,$d1,$d2
.byte $d2,$d3,$d3,$d4,$d4,$d5,$d5,$d5
.byte $d6,$d6,$d7,$d7,$d8,$d8,$d9,$d9
.byte $d9,$da,$da,$db,$db,$db,$dc,$dc
.byte $dd,$dd,$dd,$de,$de,$de,$df,$df
.byte $df,$e0,$e0,$e1,$e1,$e1,$e2,$e2
.byte $e2,$e3,$e3,$e3,$e4,$e4,$e4,$e5
.byte $e5,$e5,$e6,$e6,$e6,$e7,$e7,$e7
.byte $e7,$e8,$e8,$e8,$e9,$e9,$e9,$ea
.byte $ea,$ea,$ea,$eb,$eb,$eb,$ec,$ec
.byte $ec,$ec,$ed,$ed,$ed,$ed,$ee,$ee
.byte $ee,$ee,$ef,$ef,$ef,$ef,$f0,$f0
.byte $f0,$f1,$f1,$f1,$f1,$f1,$f2,$f2
.byte $f2,$f2,$f3,$f3,$f3,$f3,$f4,$f4
.byte $f4,$f4,$f5,$f5,$f5,$f5,$f5,$f6
.byte $f6,$f6,$f6,$f7,$f7,$f7,$f7,$f7
.byte $f8,$f8,$f8,$f8,$f9,$f9,$f9,$f9
.byte $f9,$fa,$fa,$fa,$fa,$fa,$fb,$fb
.byte $fb,$fb,$fb,$fc,$fc,$fc,$fc,$fc
.byte $fd,$fd,$fd,$fd,$fd,$fd,$fe,$fe
.byte $fe,$fe,$fe,$ff,$ff,$ff,$ff,$ff
}}
}
2023-10-27 20:36:43 +00:00
asmsub diff(ubyte v1 @A, ubyte v2 @Y) -> ubyte @A {
; -- returns the (absolute) difference, or distance, between the two bytes
%asm {{
sty P8ZP_SCRATCH_REG
sec
sbc P8ZP_SCRATCH_REG
bcs +
eor #255
inc a
+ rts
}}
}
asmsub diffw(uword w1 @R0, uword w2 @AY) -> uword @AY {
; -- returns the (absolute) difference, or distance, between the two words
%asm {{
sec
sbc cx16.r0L
sta cx16.r0L
tya
sbc cx16.r0H
sta cx16.r0H
bcs +
eor #255
sta cx16.r0H
lda cx16.r0L
eor #255
inc a
sta cx16.r0L
bne +
inc cx16.r0H
+ lda cx16.r0L
ldy cx16.r0H
rts
}}
}
}