Compare commits

..

No commits in common. "main" and "main" have entirely different histories.
main ... main

3 changed files with 182 additions and 448 deletions

612
mandel.s
View file

@ -1,16 +1,16 @@
; Our zero-page vars
ox = $80 ; fixed6.26: center point x
oy = $84 ; fixed6.26: center point y
cx = $88 ; fixed6.26: c_x
cy = $8c ; fixed6.26: c_y
ox = $80 ; fixed8.24: center point x
oy = $84 ; fixed8.24: center point y
cx = $88 ; fixed8.24: c_x
cy = $8c ; fixed8.24: c_y
zx = $90 ; fixed6.26: z_x
zy = $94 ; fixed6.26: z_y
zx_2 = $98 ; fixed6.26: z_x^2
zy_2 = $9c ; fixed6.26: z_y^2
zx = $90 ; fixed8.24: z_x
zy = $94 ; fixed8.24: z_y
zx_2 = $98 ; fixed8.24: z_x^2
zy_2 = $9c ; fixed8.24: z_y^2
zx_zy = $a0 ; fixed6.26: z_x * z_y
dist = $a4 ; fixed6.26: z_x^2 + z_y^2
zx_zy = $a0 ; fixed8.24: z_x * z_y
dist = $a4 ; fixed8.24: z_x^2 + z_y^2
sx = $a8 ; i16: screen pixel x
sy = $aa ; i16: screen pixel y
z_buffer_active = $ac ; boolean: 1 if we triggered the lake, 0 if not
@ -31,12 +31,10 @@ chroma_offset = $bb ; u8
palette_ticks = $bc ; u8
chroma_ticks = $bd ; u8
count_frames = $be ; u8
; free space $bf
count_pixels = $bf ; u8
count_iters = $c0 ; u16
text_col = $c2 ; u8
text_row = $c3 ; u8
; free space c4-cb
total_pixels = $c0 ; float48
total_ms = $c6 ; float48
temp = $cc ; u16
temp2 = $ce ; u16
@ -61,12 +59,10 @@ LBUFF = $0580 ; result buffer for FASC routine
; FP ROM routine vectors
FASC = $D8E6 ; FLOATING POINT TO ASCII (output in INBUFF, last char has high bit set)
IFP = $D9AA ; INTEGER TO FLOATING POINT CONVERSION (FR0:u16 -> FR0:float48)
FPI = $D9D2 ; floating point to integer
FADD = $DA66 ; ADDITION (FR0 += FR1)
FSUB = $DA60 ; SUBTRACTION (FR0 -= FR1)
FMUL = $DADB ; MULTIPLICATION (FR0 *= FR1)
FDIV = $DB28 ; DIVISION (FR0 /= FR1)
ZFR0 = $DA44 ; clear FR0
ZF1 = $DA46 ; CLEAR ZERO PAGE FLOATING POINT NUMBER (XX)
FLD0R = $DD89 ; LOAD FR0 WITH FLOATING POINT NUMBER (YYXX)
FLD1R = $DD98 ; LOAD FR1 WITH FLOATING POINT NUMBER (YYXX)
@ -80,7 +76,7 @@ framebuffer_bottom = $b000
display_list = $bf00
framebuffer_end = $c000
height = 176
height = 184
half_height = height >> 1
width = 160
half_width = width >> 1
@ -144,52 +140,25 @@ strings:
str_self:
.byte "MANDEL-6502"
str_self_end:
.byte 0
str_speed:
.byte "us/iter: "
.byte " ms/px"
str_speed_end:
.byte 0
str_run:
.byte " RUN"
str_run_end:
.byte 0
str_done:
.byte "DONE"
str_done_end:
.byte 0
str_padding:
.byte " "
str_padding_end:
.byte 0
str_self_len = str_self_end - str_self
str_speed_len = str_speed_end - str_speed
str_run_len = str_run_end - str_run
str_done_len = str_done_end - str_done
str_padding_len = str_padding_end - str_padding
speed_precision = 6
speed_start = 40 - str_done_len - str_speed_len - str_padding_len - 1
speed_start = 40 - str_done_len - str_speed_len - speed_precision - 1
speed_len = 14 + str_speed_len
col_x = 1
str_x:
.byte "X:"
.byte 0
str_x_len = 2
str_x_space = 12
str_x_padding = 2
col_y = col_x + str_x_len + str_x_space + str_x_padding
str_y:
.byte "Y:"
.byte 0
str_y_len = 2
str_y_space = 12
str_y_padding = 2
col_zoom = col_y + str_y_len + str_y_space + str_y_padding
str_zoom:
.byte "ZOOM:"
.byte 0
str_zoom_len = 5
char_map:
; Map ATASCII string values to framebuffer font entries
@ -220,49 +189,20 @@ aspect:
;
; 184h is the equiv of 220.8h at square pixels
; 320 / 220.8 = 1.45 display aspect ratio
aspect_x: ; fixed3.13 5/4
.word 5 << (13 - 2)
aspect_x: ; fixed4.16 5/4
.word 5 << (12 - 2)
aspect_y: ; fixed3.13 3/4
.word 3 << (13 - 2)
aspect_y: ; fixed4.16 3/4
.word 3 << (12 - 2)
fixed3_13_as_float: ; float48
; 1 << 13
; 8192
; 81 92 . 00 00 00
.byte 65 ; exponent/sign - +1 byte
.byte $81
.byte $92
.byte $00
.byte $00
.byte $00
sec_per_frame: ; float48 00 . 01 66 66 66 67
.byte 63 ; exponent/sign - -1 bytes
.byte $01 ; BCD digits
ms_per_frame: ; float48 16.66666667
.byte 64 ; exponent/sign
.byte $16 ; BCD digits
.byte $66
.byte $66
.byte $66
.byte $67
us_per_sec: ; float48 1e9 01 00 0,0 00 . 00
.byte 67 ; exponent/sign +3 bytes
.byte $01 ; BCD digits
.byte $00
.byte $00
.byte $00
.byte $00
total_iters: ; float48
.repeat 6
.byte 0
.endrepeat
total_sec: ; float48
.repeat 6
.byte 0
.endrepeat
display_list_start:
; 24 lines overscan
.repeat 3
@ -286,10 +226,6 @@ display_list_start:
.byte $0e
.endrep
; 8 scan lines, 1 row of 40-column text
.byte $42
.addr textbuffer + 40
.byte $41 ; jump and blank
.addr display_list
display_list_end:
@ -298,9 +234,9 @@ display_list_len = display_list_end - display_list_start
color_map:
.byte 0
.repeat 85
.byte %01010101
.byte %10101010
.byte %11111111
.byte 1
.byte 2
.byte 3
.endrepeat
@ -349,34 +285,23 @@ fill_masks:
.byte %00000001
.byte %00000000
pixel_masks:
.byte %11111111
.byte %11110000
.byte %11000000
viewport_zoom:
.byte 0
.byte 5
.byte 7
.byte 5
.byte 7
.byte 7
.byte 1
.byte 6
.byte 8
.byte 6
viewport_ox:
.dword ($00000000 & $3fffffff) << 2
.dword ($ff110000 & $3fffffff) << 2
.dword ($ff110000 & $3fffffff) << 2
.dword ($fe400000 & $3fffffff) << 2
.dword ($fe3b0000 & $3fffffff) << 2
.dword $fd220000
.dword $00000000
.dword $ff110000
.dword $ff110000
.dword $fe400000
viewport_oy:
.dword ($00000000 & $3fffffff) << 2
.dword ($ffb60000 & $3fffffff) << 2
.dword ($ffbe0000 & $3fffffff) << 2
.dword ($00000000 & $3fffffff) << 2
.dword ($fffe0000 & $3fffffff) << 2
.dword $ff000000
.dword $00000000
.dword $ffb60000
.dword $ffbe0000
.dword $00000000
; 2 + 9 * byte cycles
.macro add bytes, dest, arg1, arg2
@ -528,6 +453,20 @@ viewport_oy:
sta dest + 1
.endmacro
; input: arg as u8
; input/output: dest as u16
; clobbers a, x
.macro sqr8_add16 dest, arg
ldx arg
clc
lda sqr_lobyte,x
adc dest
sta dest
lda sqr_hibyte,x
adc dest + 1
sta dest + 1
.endmacro
.segment "TABLES"
; lookup table for top byte -> PORTB value for bank-switch
.align 256
@ -810,8 +749,9 @@ inner_loop:
; h1*h2*256*256 + h1*l2*256 + h2*l1*256 + l1*l2
imul8 result, arg1, arg2, xe
imul8 result + 2, arg1 + 1, arg2 + 1, xe
lda #0
sta result + 2
sta result + 3
imul8 inter, arg1 + 1, arg2, xe
add16 result + 1, result + 1, inter
@ -821,6 +761,9 @@ inner_loop:
add16 result + 1, result + 1, inter
add_carry result + 3
imul8 inter, arg1 + 1, arg2 + 1, xe
add16 result + 2, result + 2, inter
; In case of negative inputs, adjust high word
; https://stackoverflow.com/a/28827013
lda arg1 + 1
@ -853,8 +796,9 @@ arg2_pos:
; h*h*256*256 + h*l*256 + h*l*256 + l*l
sqr8 result, arg
sqr8 result + 2, arg + 1
lda #0
sta result + 2
sta result + 3
imul8 inter, arg + 1, arg, xe
add16 result + 1, result + 1, inter
@ -862,6 +806,8 @@ arg2_pos:
add16 result + 1, result + 1, inter
add_carry result + 3
sqr8_add16 result + 2, arg + 1
rts ; 6 cyc
.endscope
.endmacro
@ -927,72 +873,10 @@ next:
.endmacro
; input in FR0, 16 bits signed 3.13 fixed
; output in FR0, Atari float
; clobbers a, x, y, FR0, FR1
.proc fixed3_13_to_float
ldx #.lobyte(fixed3_13_as_float)
ldy #.hibyte(fixed3_13_as_float)
jsr FLD1R
; check sign bit! conversion routine is for unsigned
lda FR0 + 1
bpl positive
negative:
neg16 FR0
jsr IFP
; set float sign bit
lda FR0
ora #$80
sta FR0
jmp common
positive:
jsr IFP
common:
jsr FDIV
rts
.endproc
; input in FR0, Atari float
; output in FR0, 16 bits signed 3.13 fixed
; clobbers a, x, y, FR0, FR1
.proc float_to_fixed3_13
ldx #.lobyte(fixed3_13_as_float)
ldy #.hibyte(fixed3_13_as_float)
jsr FLD1R
jsr FMUL
; check sign bit! conversion routine is for unsigned
lda FR0
bcc positive
negative:
; clearfloat sign bit
lda FR0
eor #$80
sta FR0
jsr FPI
neg16 FR0
jmp common
positive:
jsr FPI
common:
rts
.endproc
.proc mandelbrot
; input:
; cx: position scaled to 6.26 fixed point - -32..+31.9
; cy: position scaled to 6.26
; cx: position scaled to 8.24 fixed point - -128..+127.9
; cy: position scaled to 8.24
;
; output:
; iter: iteration count at escape or 0
@ -1043,11 +927,6 @@ common:
sta z_buffer_end
loop:
inc count_iters
bne low_iters
inc count_iters + 1
low_iters:
; iter++ & max-iters break
inc iter
bne keep_going
@ -1055,7 +934,7 @@ low_iters:
keep_going:
.macro quick_exit arg, max
; arg: fixed6.26
; arg: fixed8.24
; max: integer
.local positive
.local negative
@ -1068,12 +947,12 @@ keep_going:
bmi negative
positive:
cmp #(max << 2)
cmp #max
bmi all_done ; 'less than'
jmp exit_path
negative:
cmp #(256 - (max << 2))
cmp #(256 - max)
beq first_equal ; 'equal' on first byte
bpl all_done ; 'greater than'
@ -1093,7 +972,7 @@ keep_going:
all_done:
.endmacro
; 6.26: (-32 .. 31.9)
; 8.24: (-128 .. 127.9)
; zx = zx_2 - zy_2 + cx
sub32 zx, zx_2, zy_2
add32 zx, zx, cx
@ -1104,9 +983,9 @@ keep_going:
add32 zy, zy, cy
quick_exit zy, 2
; convert 6.26 -> 3.13: (-4 .. +3.9)
shift_round_16 zx, 3
shift_round_16 zy, 3
; convert 8.24 -> 4.12: (-8 .. +7.9)
shift_round_16 zx, 4
shift_round_16 zy, 4
; zx_2 = zx * zx
sqr16 zx_2, zx + 2
@ -1228,9 +1107,9 @@ enough:
.endmacro
.macro zoom_factor dest, src, aspect
; output: dest: fixed6.26
; input: src: fixed3.13
; aspect: fixed3.13
; output: dest: fixed8.24
; input: src: fixed4.12
; aspect: fixed4.12
; clobbers A, X, flags, etc
copy16 dest, src
scale_zoom dest
@ -1248,11 +1127,8 @@ enough:
; iter -> color
ldx iter
lda color_map,x
ldx fill_level
and pixel_masks,x
sta pixel_color
lda pixel_masks,x
eor #$ff
lda #(255 - 3)
sta pixel_mask
; sy -> line base address in temp
@ -1301,23 +1177,22 @@ point:
; pixel_mask <<= pixel_shift (shifting in ones)
and #3
sta pixel_shift
lda #3
sec
sbc pixel_shift
tax
shift_loop:
beq shift_done
lsr pixel_color
lsr pixel_color
asl pixel_color
asl pixel_color
sec
ror pixel_mask
rol pixel_mask
sec
ror pixel_mask
rol pixel_mask
dex
jmp shift_loop
shift_done:
ldy fill_level
ldx fill_masks,y
inx
; pixel_offset = temp >> 2
lda temp
lsr a
@ -1325,94 +1200,48 @@ shift_done:
sta pixel_offset
tay
draw_pixel:
; read, mask, or, write
lda (pixel_ptr),y
and pixel_mask
ora pixel_color
sta (pixel_ptr),y
dex
beq done
clc
lda #40
adc pixel_ptr
sta pixel_ptr
lda #0
adc pixel_ptr + 1
sta pixel_ptr + 1
jmp draw_pixel
done:
rts
.endproc
; in/out: column in text_col
; in: row in text_row
; in: pointer to string in INBUFF
; clobbers x/y/a/temp
.proc draw_string
drawptr = temp
strptr = INBUFF
clc
lda #.lobyte(textbuffer)
adc text_col
sta temp
lda #.hibyte(textbuffer)
adc #0
sta temp + 1
ldx text_row
beq done_rows
continue_rows:
clc
lda temp
adc #40
sta temp
lda temp + 1
adc #0
sta temp + 1
dex
bne continue_rows
done_rows:
ldy #0
.macro draw_text_indirect col, len, strptr
; clobbers A, X
.local loop
.local done
ldx #0
loop:
lda (strptr),y
; if char's null, terminate c-style
cpx #len
beq done
; save the char for terminator check
pha
; strip the high bit (terminator)
and #$7f
tax
lda char_map,x
sta (drawptr),y
iny
pla
; _last_ char has high bit set in atari rom routines
bmi done
txa
tay
lda (strptr),y
tay
lda char_map,y
sta textbuffer + col,x
inx
jmp loop
done:
; move the text column pointer
tya
clc
adc text_col
sta text_col
.endmacro
rts
.endproc
.macro draw_string_const str
lda #.lobyte(str)
sta INBUFF
lda #.hibyte(str)
sta INBUFF + 1
jsr draw_string
.macro draw_text col, len, cstr
; clobbers A, X
.local loop
.local done
ldx #0
loop:
cpx #len
beq done
ldy cstr,x
lda char_map,y
sta textbuffer + col,x
inx
jmp loop
done:
.endmacro
.proc vblank_handler
@ -1538,7 +1367,7 @@ skip_char:
plus:
lda zoom
cmp #7
cmp #8
bpl skip_char
inc zoom
jmp done
@ -1549,20 +1378,16 @@ minus:
dec zoom
jmp done
up:
add32 oy, oy, temp
jsr display_coords
sub32 oy, oy, temp
jmp done
down:
sub32 oy, oy, temp
jsr display_coords
add32 oy, oy, temp
jmp done
left:
sub32 ox, ox, temp
jsr display_coords
jmp done
right:
add32 ox, ox, temp
jsr display_coords
jmp done
number_keys:
@ -1574,10 +1399,6 @@ number_keys:
beq three
cpy #KEY_4
beq four
cpy #KEY_5
beq five
cpy #KEY_6
beq six
jmp skip_char
one:
@ -1591,12 +1412,6 @@ three:
jmp load_key_viewport
four:
ldx #3
jmp load_key_viewport
five:
ldx #4
jmp load_key_viewport
six:
ldx #5
; fall through
load_key_viewport:
jsr load_viewport
@ -1632,63 +1447,12 @@ zero_byte_loop:
.proc status_bar
; Status bar
lda #0
sta text_col
lda #0
sta text_row
draw_string_const str_self
lda #(40 - str_run_len)
sta text_col
draw_string_const str_run
draw_text 0, str_self_len, str_self
draw_text 40 - str_run_len, str_run_len, str_run
rts
.endproc
.proc display_coords
lda #1
sta text_row
lda #col_x
sta text_col
draw_string_const str_x
copy32 FR0, ox
shift_round_16 FR0, 3
copy16 FR0, FR0 + 2
jsr fixed3_13_to_float
jsr FASC
jsr draw_string
lda #col_y
sta text_col
draw_string_const str_y
copy32 FR0, oy
shift_round_16 FR0, 3
copy16 FR0, FR0 + 2
jsr fixed3_13_to_float
jsr FASC
jsr draw_string
lda #col_zoom
sta text_col
draw_string_const str_zoom
lda zoom
clc
adc #0
sta FR0
lda #0
sta FR0 + 1
jsr IFP
jsr FASC
jsr draw_string
rts
.endproc
; input: viewport selector in x
; clobbers: a, x
.proc load_viewport
@ -1740,7 +1504,6 @@ zero_byte_loop:
sta DMACTL
jsr clear_screen
jsr display_coords
; Copy the display list into properly aligned memory
; Can't cross 1024-byte boundaries :D
@ -1779,24 +1542,19 @@ copy_byte_loop:
jsr SETVBV
main_loop:
; count_frames = 0; count_iters = 0
; count_frames = 0; count_pixels = 0
lda #0
sta count_frames
sta count_iters
sta count_iters + 1
sta count_pixels
; total_sec = 0.0; total_iters = 0.0
jsr ZFR0
ldx #.lobyte(total_sec)
ldy #.hibyte(total_sec)
jsr FST0R
ldx #.lobyte(total_iters)
ldy #.hibyte(total_iters)
jsr FST0R
; total_ms = 0.0; total_pixels = 0.0
ldx #total_ms
jsr ZF1
ldx #total_pixels
jsr ZF1
jsr clear_screen
jsr status_bar
jsr display_coords
lda #0
sta fill_level
@ -1854,7 +1612,6 @@ not_skipped_mask:
zoom_factor cx, sx, aspect_x
add32 cx, cx, ox
zoom_factor cy, sy, aspect_y
neg32 cy
add32 cy, cy, oy
jsr mandelbrot
jsr pset
@ -1866,32 +1623,38 @@ not_skipped_mask:
no_key:
; check if we should update the counters
;
; count_pixels >= width? update!
inc count_pixels
lda count_pixels
cmp #width
bmi update_status
; count_frames >= 120? update!
lda count_frames
cmp #120 ; >= 2 seconds
bpl update_status
jmp skipped
bmi skipped
update_status:
; FR0 = (float)count_iters & clear count_iters
copy16 FR0, count_iters
jsr IFP
; FR0 = (float)count_pixels & clear count_pixels
lda count_pixels
sta FR0
lda #0
sta count_iters
sta count_iters + 1
sta FR0 + 1
sta count_pixels
jsr IFP
; FR1 = total_iters
ldx #.lobyte(total_iters)
ldy #.hibyte(total_iters)
; FR1 = total_pixels
ldx #.lobyte(total_pixels)
ldy #.hibyte(total_pixels)
jsr FLD1R
; FR0 += FR1
jsr FADD
; total_iters = FR0
ldx #.lobyte(total_iters)
ldy #.hibyte(total_iters)
; total_pixels = FR0
ldx #.lobyte(total_pixels)
ldy #.hibyte(total_pixels)
jsr FST0R
@ -1904,66 +1667,44 @@ update_status:
sta count_frames
jsr IFP
; FR0 *= sec_per_frame
ldx #.lobyte(sec_per_frame)
ldy #.hibyte(sec_per_frame)
; FR0 *= ms_per_frame
ldx #.lobyte(ms_per_frame)
ldy #.hibyte(ms_per_frame)
jsr FLD1R
jsr FMUL
; FR0 += total_sec
ldx #.lobyte(total_sec)
ldy #.hibyte(total_sec)
; FR0 += total_ms
ldx #total_ms
ldy #0
jsr FLD1R
jsr FADD
; total_sec = FR0
ldx #.lobyte(total_sec)
ldy #.hibyte(total_sec)
; total_ms = FR0
ldx #total_ms
ldy #0
jsr FST0R
; FR0 /= total_iters
ldx #.lobyte(total_iters)
ldy #.hibyte(total_iters)
; FR0 /= total_pixels
ldx #total_pixels
ldy #0
jsr FLD1R
jsr FDIV
; FR0 *= us_per_sec
ldx #.lobyte(us_per_sec)
ldy #.hibyte(us_per_sec)
jsr FLD1R
jsr FMUL
; round (down) to integer
jsr FPI
clc
jsr IFP
lda #speed_start
sta text_col
lda #0
sta text_row
draw_string_const str_speed
lda text_col
pha
draw_string_const str_padding
pla
sta text_col
; convert to ASCII in INBUFF and print
; convert to ASCII in INBUFF
jsr FASC
jsr draw_string
; print the first 6 digits
draw_text_indirect speed_start, speed_precision, INBUFF
draw_text speed_start + speed_precision, str_speed_len, str_speed
skipped:
; sx += fill_level[fill_masks] + 1
ldx fill_level
lda fill_masks,x
clc
adc #1 ; will never carry
adc sx
lda sx
adc #1
sta sx
lda #0
adc sx + 1
lda sx + 1
adc #0
sta sx + 1
lda sx
@ -1973,15 +1714,12 @@ skipped:
loop_sx_done:
; sy += fill_level[fill_masks] + 1
ldx fill_level
lda fill_masks,x
clc
adc #1 ; will never carry
adc sy
lda sy
adc #1
sta sy
lda #0
adc sy + 1
lda sy + 1
adc #0
sta sy + 1
lda sy
@ -2000,13 +1738,7 @@ fill_loop_done:
loop:
; finished
lda #(40 - str_done_len)
sta text_col
lda #0
sta text_row
draw_string_const str_done
draw_text 40 - str_done_len, str_done_len, str_done
jsr keycheck
beq loop
jmp main_loop

View file

@ -18,7 +18,7 @@ Enjoy! I'll probably work on this off and on for the next few weeks until I've g
## Current state
Basic rendering is functional, with interactive zoom/pan (+/-/arrows) and 6 preset viewports via the number keys.
Basic rendering is functional, with interactive zoom/pan (+/-/arrows) and 4 preset viewports via the number keys.
The 16-bit signed integer multiplication takes two 16-bit inputs and emits one 32-bit output in the zero page, using the Atari OS ROM's floating point registers as workspaces. Inputs are clobbered.
@ -27,7 +27,7 @@ The 16-bit signed integer multiplication takes two 16-bit inputs and emits one 3
* when expanded RAM is available as on 130XE, a 64KB 8-bit multiplication table accelerates the remaining multiplications
* without expanded RAM, a table of half-squares is used to implement the algorithm from https://everything2.com/title/Fast+6502+multiplication
The mandelbrot calculations are done using 3.13-precision fixed point numbers with 6.26-precision intermediates.
The mandelbrot calculations are done using 4.12-precision fixed point numbers with 8.24-precision intermediates. It may be possible to squish this down to 3.13/6.26.
Iterations are capped at 255.

14
todo.md
View file

@ -1,17 +1,19 @@
things to try:
* fix status bar to show elapsed time, per-iter time, per-pixel iter count
* 'turbo' mode disabling graphics in full or part
* skip add on the top-byte multiply in sqr8/mul8
* should save a few cycles, suggestion by jamey
* patch the entire expanded-ram imul8xe on top of imul8 to avoid the 3-cycle thunk penalty :D
* maybe clean up the load/layout of the big mul table
* consider alternate lookup tables in the top 16KB under ROM
* try 3.13 fixed point instead of 4.12 for more precision
* can we get away without the extra bit?
* since exit compare space would be 6.26 i think so
* y-axis mirror optimization
* 'wide pixels' 2x and 4x for a fuller initial image in the tiered rendering
* maybe redo tiering to just 4x4, 2x2, 1x1?
* extract viewport for display & re-input via keyboard
* fujinet screenshot/viewport uploader