mirror of
https://github.com/deater/dos33fsprogs.git
synced 2024-09-08 11:54:48 +00:00
211 lines
4.5 KiB
ArmAsm
211 lines
4.5 KiB
ArmAsm
; Lo-res fire animation, size-optimized
|
|
|
|
; by deater (Vince Weaver) <vince@deater.net>
|
|
|
|
; based on code described here http://fabiensanglard.net/doom_fire_psx/
|
|
|
|
; 611 bytes at first
|
|
; 601 bytes -- strip out some unused code
|
|
; 592 bytes -- don't init screen
|
|
; 443 bytes -- remove more dead code
|
|
; 206 bytes -- no need to clear screen
|
|
; 193 bytes -- un-cycle exact the random16 code
|
|
; 189 bytes -- more optimization of random16 code
|
|
; 161 bytes -- move to 8-bit RNG
|
|
; 152 bytes -- reduce lookup to top half colors (had to remove brown)
|
|
; also changed maroon to pink
|
|
; 149 bytes -- use monitor GR
|
|
; 149 bytes -- load into zero page
|
|
; 140 bytes -- start using zero-page addressing
|
|
; 139 bytes -- rotate instead of mask for low bit
|
|
; 138 bytes -- bcs instead of jmp
|
|
; 137 bytes -- BIT nop trick to get rid of jump
|
|
; 135 bytes -- push/pull instead of saving to zero page
|
|
; 134 bytes -- replace half of lookup table with math
|
|
; 119 bytes -- replace that half of lookup table with better math
|
|
; 134 bytes -- replace other half of lookup table with math
|
|
; 132 bytes -- replace stack with zp loads
|
|
; 129 bytes -- use Y instead of X
|
|
; 125 bytes -- optimize low byte generationw with branch
|
|
; 113 bytes -- make get_row a subroutine
|
|
; 112 bytes -- regrettable change to the low-byte code
|
|
; 109 bytes -- replace BIT/OR in low calc with an ADC
|
|
; 107 bytes -- replace self-modifying load/store absolute with Y-indirect
|
|
; 106 bytes -- assume bit 8 is as random as bit 0
|
|
; 105 bytes -- qkumba points out that GR leaves BASL:BASH pointing at line 23
|
|
; so we can use Y-indirect of BASL to draw the bottom white line
|
|
|
|
|
|
; Zero Page
|
|
SEEDL = $4E
|
|
TEMP = $00
|
|
TEMPY = $01
|
|
OUTL = $02
|
|
OUTH = $03
|
|
INL = $04
|
|
INH = $05
|
|
BASL = $28
|
|
|
|
; 100 = $64
|
|
|
|
; Soft Switches
|
|
SET_GR = $C050 ; Enable graphics
|
|
FULLGR = $C052 ; Full screen, no text
|
|
LORES = $C056 ; Enable LORES graphics
|
|
|
|
; monitor routines
|
|
ASOFT_GR = $F390
|
|
MON_SETGR = $FB40
|
|
|
|
fire_demo:
|
|
|
|
; GR part
|
|
jsr MON_SETGR ; 3
|
|
bit FULLGR ; 3
|
|
;==========
|
|
; 6
|
|
|
|
|
|
; Setup white line on bottom
|
|
|
|
; GR leaves BASL pointing at hline 23
|
|
|
|
lda #$ff ; 2
|
|
ldy #39 ; 2
|
|
white_loop:
|
|
sta (BASL),Y ; hline 23 (46+47) ; 3
|
|
dey ; 1
|
|
bpl white_loop ; 2
|
|
;============
|
|
; 10
|
|
|
|
fire_loop:
|
|
|
|
ldx #22 ; 22 ; 2
|
|
|
|
yloop:
|
|
stx TEMPY ; txa/pha not any better ; 2
|
|
|
|
; setup the load/store addresses
|
|
; using Y-indirect is smaller than self-modifying code
|
|
|
|
jsr get_row ; 3
|
|
sty OUTL ; 2
|
|
sta OUTH ; 2
|
|
|
|
inx ; 1
|
|
|
|
jsr get_row ; 3
|
|
sty INL ; 2
|
|
sta INH ; 2
|
|
|
|
ldy #39 ; 2
|
|
xloop:
|
|
|
|
;=============================
|
|
; random8
|
|
;=============================
|
|
; 8-bit 6502 Random Number Generator
|
|
; Linear feedback shift register PRNG by White Flame
|
|
; http://codebase64.org/doku.php?id=base:small_fast_8-bit_prng
|
|
|
|
random8:
|
|
lda SEEDL ; 2
|
|
beq doEor ; 2
|
|
asl ; 1
|
|
beq noEor ; if the input was $80, skip the EOR ; 2
|
|
bcc noEor ; 2
|
|
doEor: eor #$1d ; 2
|
|
noEor: sta SEEDL ; 2
|
|
|
|
; end inlined RNG
|
|
|
|
bmi no_change ; assume bit 8 is as random as bit 0 ; 2
|
|
|
|
|
|
lda (INL),Y ; load value at row+1 ; 2
|
|
and #$7 ; mask off ; 2
|
|
tax ; 1
|
|
lda <(color_progression),X ; 2
|
|
|
|
.byte $2c ; BIT trick, nops out next instruction ; 1
|
|
no_change:
|
|
lda (INL),Y ; load value at row+1 ; 2
|
|
|
|
|
|
smc_sta:
|
|
sta (OUTL),Y ; 2
|
|
dey ; 1
|
|
bpl xloop ; 2
|
|
|
|
ldx TEMPY ; 2
|
|
|
|
dex ; 1
|
|
bpl yloop ; 2
|
|
|
|
bmi fire_loop ; 2
|
|
|
|
|
|
color_progression:
|
|
.byte $00 ; 8->0 ; 1000 0000
|
|
.byte $bb ; 9->11 ; 1001 1011
|
|
.byte $00 ; 10->0 ; 1010 0000
|
|
.byte $aa ; 11->10 ; 1011 1010
|
|
.byte $00 ; 12->0 ; 1100 0000
|
|
.byte $99 ; 13->9 ; 1101 1001
|
|
.byte $00 ; 14->0 ; 1110 0000
|
|
.byte $dd ; 15->13 ; 1111 1101
|
|
|
|
|
|
|
|
|
|
; Take row X and convert to address A:Y
|
|
get_row:
|
|
; get low byte
|
|
|
|
; 000X X00O
|
|
; lsr
|
|
; 0000 XX00 O
|
|
; adc #1
|
|
; 0000 XXOx
|
|
; lsr
|
|
; 0000 0XXO
|
|
; lsr
|
|
; 0000 00XX O
|
|
; ror
|
|
|
|
txa ; 1
|
|
and #$19 ; 2
|
|
lsr ; 1
|
|
adc #1 ; 2
|
|
lsr ; 1
|
|
lsr ; 1
|
|
tay ; 1
|
|
lda <low_offsets,Y ; 2
|
|
ror ; 1
|
|
tay ; 1
|
|
;===========
|
|
; 13
|
|
|
|
|
|
; high byte of store address
|
|
txa ; 1
|
|
and #$7 ; 2
|
|
lsr ; 1
|
|
ora #$4 ; 2
|
|
;===========
|
|
; 6
|
|
|
|
rts ; 1
|
|
|
|
|
|
; 14+6 = 20, as opposed to full lookup table which would be
|
|
; at least 24
|
|
; down to 13+3 at the expense of readability
|
|
|
|
; these are shifted left by one due to the algorithm
|
|
low_offsets:
|
|
.byte $00,$50,$a0
|
|
|
|
|