Compare commits

..
6 changed files with 285 additions and 649 deletions

View file

@ -2,11 +2,8 @@
all : mandel.xex all : mandel.xex
mandel.xex : mandel.o mandel-core.o tables.o atari-xex.cfg mandel.xex : mandel.o tables.o atari-asm-xex.cfg
ld65 -C ./atari-xex.cfg --mapfile mandel.map -o $@ mandel.o mandel-core.o tables.o atari.lib ld65 -C ./atari-asm-xex.cfg -o $@ mandel.o tables.o
mandel.s : mandel.c mandel.h
cc65 -o $@ mandel.c
%.o : %.s %.o : %.s
ca65 -o $@ $< ca65 -o $@ $<
@ -16,7 +13,6 @@ tables.s : tables.js
clean : clean :
rm -f tables.s rm -f tables.s
rm -f mandel.s
rm -f *.o rm -f *.o
rm -f *.xex rm -f *.xex
rm -f mandel.map

View file

@ -1,69 +0,0 @@
# Sample linker configuration for C programs using the Atari binary file support.
# Use with: cl65 -tatari -Catari-xex.cfg prog.c -o prog.xex
FEATURES {
STARTADDRESS: default = $8000;
}
SYMBOLS {
__SYSTEM_CHECK__: type = import; # force inclusion of "system check" load chunk
__STACKSIZE__: type = weak, value = $0800; # 2k stack
__STARTADDRESS__: type = export, value = %S;
__RESERVED_MEMORY__: type = weak, value = $0000;
__SYSCHKHDR__: type = export, value = 0; # Disable system check header
__SYSCHKTRL__: type = export, value = 0; # Disable system check trailer
__TABLESEG_START__: type = weak, value = $2E00 + $0300;
__TABLESEG_SIZE__: type = weak, value = 6 * $100;
__BANKSY_START__: type = weak, value = $4000;
__BANKSY_SIZE__: type = weak, value = $4000;
__FRAMEBUFFER_START__: type = weak, value = $A000;
}
MEMORY {
# Note -- $80 and $81 (LOMEM) appear to be reserved in ZP.
ZP: file = "", define = yes, start = $0082, size = $007E;
# "system check" load chunk
SYSCHKCHNK: file = %O, start = $2E00, size = $0300;
# Note $a000-$bfff is against the BASIC cartridge, may require booting with OPTION.
TABLES: file = %O, define = yes, start = __TABLESEG_START__, size = __TABLESEG_SIZE__;
# We reserve $4000-7fff for the bank-switch window.
# In theory we could keep data and code here that we only use on 48k/64k systems.
BANKSWITCH: file = "", define = yes, start = __BANKSY_START__, size = __BANKSY_SIZE__;
# "main program" load chunk
MAIN: file = %O, define = yes, start = %S, size = __FRAMEBUFFER_START__ - __STACKSIZE__ - __RESERVED_MEMORY__ - %S;
}
FILES {
%O: format = atari;
}
FORMATS {
atari: runad = start,
initad = SYSCHKCHNK: __SYSTEM_CHECK__;
}
SEGMENTS {
ZEROPAGE: load = ZP, type = zp;
EXTZP: load = ZP, type = zp, optional = yes;
SYSCHK: load = SYSCHKCHNK, type = rw, define = yes, optional = yes;
TABLES: load = TABLES, type = ro, optional = yes, align = 256;
BANKSWICH: load = BANKSWITCH, type = ro, optional = yes;
STARTUP: load = MAIN, type = ro, define = yes;
LOWBSS: load = MAIN, type = rw, optional = yes; # not zero initialized
LOWCODE: load = MAIN, type = ro, define = yes, optional = yes;
ONCE: load = MAIN, type = ro, optional = yes;
CODE: load = MAIN, type = ro, define = yes;
RODATA: load = MAIN, type = ro;
DATA: load = MAIN, type = rw;
INIT: load = MAIN, type = rw, optional = yes;
BSS: load = MAIN, type = bss, define = yes;
}
FEATURES {
CONDES: type = constructor,
label = __CONSTRUCTOR_TABLE__,
count = __CONSTRUCTOR_COUNT__,
segment = ONCE;
CONDES: type = destructor,
label = __DESTRUCTOR_TABLE__,
count = __DESTRUCTOR_COUNT__,
segment = RODATA;
CONDES: type = interruptor,
label = __INTERRUPTOR_TABLE__,
count = __INTERRUPTOR_COUNT__,
segment = RODATA,
import = __CALLIRQ__;
}

124
bitmul.s
View file

@ -1,124 +0,0 @@
; the old 16-bit bit-and-shift multiplier
; copied from old code, won't compile as-is
; inner loop for imul16
; bitnum < 8: 25 or 41 cycles
; bitnum >= 8: 30 or 46 cycles
.macro bitmul16 arg1, arg2, result, bitnum
.local zero
.local one
.local next
; does 16-bit adds
; arg1 and arg2 are treated as unsigned
; negative signed inputs must be flipped first
; 7 cycles up to the branch
; check if arg1 has 0 or 1 bit in this place
; 5 cycles either way
.if bitnum < 8
lda arg1 ; 3 cyc
and #(1 << (bitnum)) ; 2 cyc
.else
lda arg1 + 1 ; 3 cyc
and #(1 << ((bitnum) - 8)) ; 2 cyc
.endif
bne one ; 2 cyc
zero: ; 18 cyc, 23 cyc
lsr result + 3 ; 5 cyc
jmp next ; 3 cyc
one: ; 32 cyc, 37 cyc
; 16-bit add on the top bits
clc ; 2 cyc
lda result + 2 ; 3 cyc
adc arg2 ; 3 cyc
sta result + 2 ; 3 cyc
lda result + 3 ; 3 cyc
adc arg2 + 1 ; 3 cyc
ror a ; 2 cyc - get a jump on the shift
sta result + 3 ; 3 cyc
next:
ror result + 2 ; 5 cyc
ror result + 1 ; 5 cyc
.if bitnum >= 8
; we can save 5 cycles * 8 bits = 40 cycles total by skipping this byte
; when it's all uninitialized data
ror result ; 5 cyc
.endif
.endmacro
; 5 to 25 cycles
.macro check_sign arg
; Check sign bit and flip argument to postive,
; keeping a count of sign bits in the X register.
.local positive
lda arg + 1 ; 3 cyc
bpl positive ; 2 cyc
neg16 arg ; 18 cyc
inx ; 2 cyc
positive:
.endmacro
; 518 - 828 cyc
.macro imul16 dest, arg1, arg2
copy16 FR0, arg1 ; 12 cyc
copy16 FR1, arg2 ; 12 cyc
jsr imul16_func ; 470-780 cyc
copy32 dest, FR2 ; 24 cyc
.endmacro
.macro shift_round_16 arg, shift
.repeat shift
shl32 arg
.endrepeat
round16 arg
.endmacro
.macro imul16_round dest, arg1, arg2, shift
copy16 FR0, arg1 ; 12 cyc
copy16 FR1, arg2 ; 12 cyc
jsr imul16_func ; 470-780 cyc
shift_round_16 FR2, shift
copy16 dest, FR2 + 2 ; 12 cyc
.endmacro
; min 470 cycles
; max 780 cycles
.proc imul16_func
arg1 = FR0 ; 16-bit arg (clobbered)
arg2 = FR1 ; 16-bit arg (clobbered)
result = FR2 ; 32-bit result
ldx #0 ; 2 cyc
; counts the number of sign bits in X
check_sign arg1 ; 5 to 25 cyc
check_sign arg2 ; 5 to 25 cyc
; zero out the 32-bit temp's top 16 bits
lda #0 ; 2 cyc
sta result + 2 ; 3 cyc
sta result + 3 ; 3 cyc
; the bottom two bytes will get cleared by the shifts
; unrolled loop for maximum speed, at the cost
; of a larger routine
; 440 to 696 cycles
.repeat 16, bitnum
; bitnum < 8: 25 or 41 cycles
; bitnum >= 8: 30 or 46 cycles
bitmul16 arg1, arg2, result, bitnum
.endrepeat
; In case of mixed input signs, return a negative result.
cpx #1 ; 2 cyc
bne positive_result ; 2 cyc
neg32 result ; 34 cyc
positive_result:
rts ; 6 cyc
.endproc

View file

@ -1,15 +0,0 @@
/**
* The UI and I/O wrapper for the Mandelbrot runner, in C.
*
* For the moment *all* logic is in mandel-core.s, I'm just
* trying to get this to run within a cc65 environment.
* Eventually just the inner loop fun will live in there.
*/
#include <stdlib.h>
#include <stdio.h>
#include "mandel.h"
void main(void) {
mandel_start();
}

View file

@ -1,4 +0,0 @@
#include <inttypes.h>
// From mandel-core.s:
extern void mandel_start(void);

View file

@ -1,45 +1,44 @@
.zeropage ; 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: .res 4 ; fixed6.26: center point x zx = $90 ; fixed6.26: z_x
oy: .res 4 ; fixed6.26: center point y zy = $94 ; fixed6.26: z_y
cx: .res 4 ; fixed6.26: c_x zx_2 = $98 ; fixed6.26: z_x^2
cy: .res 4 ; fixed6.26: c_y zy_2 = $9c ; fixed6.26: z_y^2
zx: .res 4 ; fixed6.26: z_x zx_zy = $a0 ; fixed6.26: z_x * z_y
zy: .res 4 ; fixed6.26: z_y dist = $a4 ; fixed6.26: z_x^2 + z_y^2
zx_2: .res 4 ; fixed6.26: z_x^2 sx = $a8 ; i16: screen pixel x
zy_2: .res 4 ; fixed6.26: z_y^2 sy = $aa ; i16: screen pixel y
z_buffer_active = $ac ; boolean: 1 if we triggered the lake, 0 if not
z_buffer_start = $ad ; u8: index into z_buffer
z_buffer_end = $ae ; u8: index into z_buffer
iter = $af ; u8: iteration count
zx_zy: .res 4 ; fixed6.26: z_x * z_y ptr = $b0 ; u16
dist: .res 4 ; fixed6.26: z_x^2 + z_y^2 pixel_ptr = $b2 ; u16
zoom = $b4 ; u8: zoom shift level
fill_level = $b5 ; u8
pixel_color = $b6 ; u8
pixel_mask = $b7 ; u8
pixel_shift = $b8 ; u8
pixel_offset = $b9 ; u8
palette_offset = $ba ; u8
chroma_offset = $bb ; u8
palette_ticks = $bc ; u8
chroma_ticks = $bd ; u8
count_frames = $be ; u8
; free space $bf
z_buffer_active: .res 1 ; boolean: 1 if we triggered the lake, 0 if not count_iters = $c0 ; u16
z_buffer_start: .res 1 ; u8: index into z_buffer text_col = $c2 ; u8
z_buffer_end: .res 1 ; u8: index into z_buffer text_row = $c3 ; u8
iter: .res 1 ; u8: iteration count ; free space c4-cb
ptr: .res 2 ; u16 temp = $cc ; u16
pixel_ptr: .res 2 ; u16 temp2 = $ce ; u16
temp: .res 2 ; u16
temp2: .res 2 ; u16
.data
; can move to .data
sx: .res 2 ; i16: screen pixel x
sy: .res 2 ; i16: screen pixel y
zoom: .res 1 ; u8: zoom shift level
fill_level: .res 1 ; u8
pixel_color: .res 1 ; u8
pixel_mask: .res 1 ; u8
pixel_shift: .res 1 ; u8
pixel_offset: .res 1 ; u8
palette_offset: .res 1 ; u8
chroma_offset: .res 1 ; u8
palette_ticks: .res 1 ; u8
chroma_ticks: .res 1 ; u8
count_frames: .res 1 ; u8
count_iters: .res 2 ; u16
text_col: .res 1 ; u8
text_row: .res 1 ; u8
palette_delay = 23 palette_delay = 23
chroma_delay = 137 chroma_delay = 137
@ -127,12 +126,6 @@ KEY_7 = 51
KEY_8 = 53 KEY_8 = 53
KEY_9 = 48 KEY_9 = 48
KEY_0 = 50 KEY_0 = 50
KEY_PERIOD = 34
KEY_E = 42
KEY_X = 22
KEY_Y = 43
.data
.struct float48 .struct float48
exponent .byte exponent .byte
@ -145,6 +138,7 @@ KEY_Y = 43
.import sqr_lobyte .import sqr_lobyte
.import sqr_hibyte .import sqr_hibyte
.data
strings: strings:
str_self: str_self:
@ -363,7 +357,7 @@ z_buffer:
.word 0 .word 0
.endrepeat .endrepeat
.export _mandel_start .export start
;max_fill_level = 6 ;max_fill_level = 6
max_fill_level = 3 max_fill_level = 3
@ -409,13 +403,6 @@ elapsed_work:
elapsed_digit: elapsed_digit:
.byte 0 .byte 0
input_col:
.byte 0
input_row:
.byte 0
input_max:
.byte 0
; 2 + 9 * byte cycles ; 2 + 9 * byte cycles
.macro add bytes, dest, arg1, arg2 .macro add bytes, dest, arg1, arg2
clc ; 2 cyc clc ; 2 cyc
@ -436,13 +423,11 @@ input_max:
add 4, dest, arg1, arg2 add 4, dest, arg1, arg2
.endmacro .endmacro
; 2-7 cycles ; 8 cycles
.macro add_carry dest .macro add_carry dest
.scope lda dest ; 3 cyc
bcc after ; 2 cyc adc #0 ; 2 cyc
inc dest ; 5 cyc sta dest ; 3 cyc
after:
.endscope
.endmacro .endmacro
; 2 + 9 * byte cycles ; 2 + 9 * byte cycles
@ -465,7 +450,7 @@ input_max:
sub 4, dest, arg1, arg2 sub 4, dest, arg1, arg2
.endmacro .endmacro
; 3 + 5 * (bytes - 1) cycles ; 3 + 5 * bytes cycles
.macro shl bytes, arg .macro shl bytes, arg
asl arg ; 3 cyc asl arg ; 3 cyc
.repeat bytes-1, i .repeat bytes-1, i
@ -473,23 +458,22 @@ input_max:
.endrepeat .endrepeat
.endmacro .endmacro
; 8 cycles ; 13 cycles
.macro shl16 arg .macro shl16 arg
shl 2, arg shl 2, arg
.endmacro .endmacro
; 13 cycles ; 18 cycles
.macro shl24 arg .macro shl24 arg
shl 3, arg shl 3, arg
.endmacro .endmacro
; 18 cycles ; 23 cycles
.macro shl32 arg .macro shl32 arg
shl 4, arg shl 4, arg
.endmacro .endmacro
; 6 * bytes cycles ; 6 * bytes cycles
; 4 * bytes bytes
.macro copy bytes, dest, arg .macro copy bytes, dest, arg
.repeat bytes, byte ; 6 * bytes cycles .repeat bytes, byte ; 6 * bytes cycles
lda arg + byte ; 3 cyc lda arg + byte ; 3 cyc
@ -498,7 +482,6 @@ input_max:
.endmacro .endmacro
; 12 cycles ; 12 cycles
; 8 bytes
.macro copy16 dest, arg .macro copy16 dest, arg
copy 2, dest, arg copy 2, dest, arg
.endmacro .endmacro
@ -533,25 +516,11 @@ input_max:
neg 4, arg neg 4, arg
.endmacro .endmacro
.macro copy_neg bytes, dest, arg ; 11-27 + 23 * shift cycles
sec ; 2 cyc ; 103-119 cycles for shift=4
.repeat bytes, byte ; 8 * byte cycles
lda #00 ; 2 cyc
sbc arg + byte ; 3 cyc
sta dest + byte ; 3 cyc
.endrepeat
.endmacro
; 18 cycles
.macro copy_neg16 dest, arg
copy_neg 2, dest, arg
.endmacro
; 11-27 + 18 * shift cycles
; 65-81 cycles for shift=3
.macro shift_round_16 arg, shift .macro shift_round_16 arg, shift
.repeat shift .repeat shift
shl32 arg ; 18 cycles shl32 arg ; 23 cycles
.endrepeat .endrepeat
round16 arg ; 11-27 cycles round16 arg ; 11-27 cycles
.endmacro .endmacro
@ -567,22 +536,21 @@ input_max:
; input: arg as fixed4.12 ; input: arg as fixed4.12
; output: dest as fixed8.24 ; output: dest as fixed8.24
;.macro sqr16 dest, arg .macro sqr16 dest, arg
; copy16 FR0, arg ; 12 cyc copy16 FR0, arg ; 12 cyc
; jsr sqr16_func ; ? cyc jsr sqr16_func ; ? cyc
; copy32 dest, FR2 ; 24 cyc copy32 dest, FR2 ; 24 cyc
;.endmacro .endmacro
; input: arg as u8 ; input: arg as u8
; output: dest as u16 ; output: dest as u16
; clobbers a, x ; clobbers a, x
; 19 cyc
.macro sqr8 dest, arg .macro sqr8 dest, arg
ldx arg ; 3 cyc ldx arg
lda sqr_lobyte,x ; 5 cyc lda sqr_lobyte,x
sta dest ; 3 cyc sta dest
lda sqr_hibyte,x ; 5 cyc lda sqr_hibyte,x
sta dest + 1 ; 3 cyc sta dest + 1
.endmacro .endmacro
.segment "TABLES" .segment "TABLES"
@ -603,57 +571,32 @@ bank_switch_table:
.macro imul8 dest, arg1, arg2, xe .macro imul8 dest, arg1, arg2, xe
.if xe .if xe
; using 64KB lookup table ; using 64KB lookup table
; 50-55 cycles ; 51-70 cycles
; min: 11 + 9 + 9 + 21 = 50
; max: 11 + 9 + 9 + 26 = 55
; clobbers x, y, dest, ptr ; clobbers x, y, dest, ptr
.scope .scope
output = dest output = dest
; top 2 bits are the table bank selector ; top 2 bits are the table bank selector
; note we keep arg2 in X after this lookup as we need it later!
ldx arg2 ; 3 cyc ldx arg2 ; 3 cyc
lda bank_switch_table,x ; 4 cyc lda bank_switch_table,x ; 4 cyc
sta PORTB ; 4 cyc sta PORTB ; 4 cyc
; bottom 14 bits except the LSB are the per-bank table index ; bottom 14 bits except the LSB are the per-bank table index
; add $4000 for the bank pointer ; add $4000 for the bank pointer
; note the low byte of ptr is assumed to remain at 0! do not clobber it!
txa ; 2 cyc txa ; 2 cyc
and #$3f ; 2 cyc and #$3f ; 2 cyc
ora #$40 ; 2 cyc ora #$40 ; 2 cyc
sta ptr + 1 ; 3 cyc sta ptr + 1 ; 3 cyc
; mask off the low bit for the index, as the table contains
; entries for every other arg1 value...
lda arg1 ; 3 cyc
tay ; 2 cyc
and #1 ; 2 cyc
bne odd ; 2 cyc
even:
; copy the entry into output ; copy the entry into output
lda (ptr),y ; 5 cyc lda arg1 ; 3 cyc
sta output ; 3 cyc and #$fe ; 2 cyc
iny ; 2 cyc tay ; 2 cyc
lda (ptr),y ; 5 cyc lda (ptr),y ; 5 cyc
sta output + 1 ; 3 cyc sta output ; 3 cyc
jmp done ; 3 cyc iny ; 2 cyc
lda (ptr),y ; 5 cyc
odd: sta output+1 ; 3 cyc
; skip the LSB for the index
dey ; 2 cyc
; add arg2 one last time for the skipped bit
txa ; 2 cyc
clc ; 2 cyc
adc (ptr), y ; 5 cyc
sta output ; 3 cyc
lda #0 ; 2 cyc
iny ; 2 cyc
adc (ptr),y ; 5 cyc
sta output + 1 ; 3 cyc
; note: we are not restoring memory to save 6 cycles! ; note: we are not restoring memory to save 6 cycles!
; this means those 16kb have to be switched back to base RAM ; this means those 16kb have to be switched back to base RAM
@ -662,6 +605,20 @@ bank_switch_table:
;;lda #$81 ; 2 cyc - disabled ;;lda #$81 ; 2 cyc - disabled
;;sta PORTB ; 4 cyc - disabled ;;sta PORTB ; 4 cyc - disabled
; check that 1 bit we skipped to fit into space
lda arg1 ; 3 cyc
and #1 ; 2 cyc
beq done ; 2 cyc
; add arg2 one last time for the skipped bit
clc ; 2 cyc
txa ; 2 cyc
adc output ; 3 cyc
sta output ; 3 cyc
lda #0 ; 2 cyc
adc output+1 ; 3 cyc
sta output+1 ; 3 cyc
done: done:
.endscope .endscope
.else .else
@ -724,6 +681,71 @@ bank_switch_table:
.endif .endif
.endmacro .endmacro
.proc imul8xe_init
bank_switch 0
lda #0
sta EXTENDED_RAM
bank_switch 1
lda #1
sta EXTENDED_RAM
bank_switch 0
lda EXTENDED_RAM
beq init
; no bank switching available, we just overwrite the value in base ram
rts
init:
; patch imul16_func into a forwarding thunk to imul16xe_func
lda #$4c ; 'jmp' opcode
sta imul16_func
lda #.lobyte(imul16xe_func)
sta imul16_func + 1
lda #.hibyte(imul16xe_func)
sta imul16_func + 2
; ditto for sqr16_func -> sqr16xe_func
lda #$4c ; 'jmp' opcode
sta sqr16_func
lda #.lobyte(sqr16xe_func)
sta sqr16_func + 1
lda #.hibyte(sqr16xe_func)
sta sqr16_func + 2
; create the lookup table
; go through the input set, in four 16KB chunks
arg1 = FR1
arg2 = FR2
result = FR0
lda #$00
sta arg1
sta arg2
sta ptr
lda #$40
sta ptr + 1
; $00 * $00 -> $3f * $ff
bank_switch 0
jsr imul8xe_init_section
; $40 * $00 -> $7f * $ff
bank_switch 1
jsr imul8xe_init_section
; $80 * $00 -> $bf * $ff
bank_switch 2
jsr imul8xe_init_section
; $c0 * $00 -> $ff * $ff
bank_switch 3
jsr imul8xe_init_section
rts
.endproc
; Initialize a 16 KB chunk of the table ; Initialize a 16 KB chunk of the table
; input: multipliers in temp ; input: multipliers in temp
@ -795,128 +817,95 @@ inner_loop:
.endproc .endproc
; min base mem: 81 * 4 + 22 * 2 + 10 = 378 cyc .macro imul16_impl xe
; max base mem: 92 * 4 + 27 * 2 + 50 = 472 cyc .local arg1
; min ext-mem: 50 * 4 + 22 * 2 + 10 = 254 cyc .local arg2
; max ext-mem: 55 * 4 + 27 * 2 + 50 = 324 cyc .local result
.macro imul16_impl result, arg1, arg2, xe .local inter
.scope .local arg1_pos
;arg1 ; 16-bit arg .local arg2_pos
;arg2 ; 16-bit arg arg1 = FR0 ; 16-bit arg (clobbered)
;result ; 32-bit result (output) arg2 = FR1 ; 16-bit arg (clobbered)
inter = temp2 ; 16-bit temporary (clobbered) result = FR2 ; 32-bit result
inter = temp2
; h1l1 * h2l2 ; h1l1 * h2l2
; (h1*256 + l1) * (h2*256 + l2) ; (h1*256 + l1) * (h2*256 + l2)
; h1*256*(h2*256 + l2) + l1*(h2*256 + l2) ; h1*256*(h2*256 + l2) + l1*(h2*256 + l2)
; h1*h2*256*256 + h1*l2*256 + h2*l1*256 + l1*l2 ; h1*h2*256*256 + h1*l2*256 + h2*l1*256 + l1*l2
; l1 * l2 imul8 result, arg1, arg2, xe
imul8 result, arg1, arg2, xe ; 81-92 or 50-55
; 256 * 256 * h1 * h2 imul8 result + 2, arg1 + 1, arg2 + 1, xe
imul8 result + 2, arg1 + 1, arg2 + 1, xe ; 81-92 or 50-55
; 256 * h1 * l2 imul8 inter, arg1 + 1, arg2, xe
imul8 inter, arg1 + 1, arg2, xe ; 81-92 or 50-55 add16 result + 1, result + 1, inter
add16 result + 1, result + 1, inter ; 20 cyc add_carry result + 3
add_carry result + 3 ; 2-7 cyc
; 256 * l1 * h2 imul8 inter, arg1, arg2 + 1, xe
imul8 inter, arg1, arg2 + 1, xe ; 81-92 or 50-55 add16 result + 1, result + 1, inter
add16 result + 1, result + 1, inter ; 20 cyc add_carry result + 3
add_carry result + 3 ; 2-7 cyc
; In case of negative inputs, adjust high word ; In case of negative inputs, adjust high word
; https://stackoverflow.com/a/28827013 ; https://stackoverflow.com/a/28827013
; 10-50 cycles lda arg1 + 1
lda arg1 + 1 ; 3 cyc bpl arg1_pos
bpl arg1_pos ; 2 cyc sub16 result + 2, result + 2, arg2
sub16 result + 2, result + 2, arg2 ; 20 cyc arg1_pos:
arg1_pos: lda arg2 + 1
bpl arg2_pos
sub16 result + 2, result + 2, arg1
arg2_pos:
lda arg2 + 1 ; 3 cyc rts ; 6 cyc
bpl arg2_pos ; 2 cyc
sub16 result + 2, result + 2, arg1 ; 20 cyc
arg2_pos:
.endscope
.endmacro .endmacro
.macro sqr16_impl xe
; min base: 19 * 2 + 81 + 22 * 2 = 163
; max base: 19 * 2 + 92 + 27 * 2 = 184
; min ext: 19 * 2 + 50 + 22 * 2 = 132
; max ext: 19 * 2 + 55 + 27 * 2 = 147
.macro sqr16_impl_inner result, arg, xe
.scope .scope
inter = temp2 arg = FR0 ; 16-bit arg (clobbered)
result = FR2 ; 32-bit result
;inter = temp2
inter = FR1
lda arg + 1
bpl arg_pos
neg16 arg
arg_pos:
; hl * hl ; hl * hl
; (h*256 + l) * (h*256 + l) ; (h*256 + l) * (h*256 + l)
; h*256*(h*256 + l) + l*(h*256 + l) ; h*256*(h*256 + l) + l*(h*256 + l)
; h*h*256*256 + h*l*256 + h*l*256 + l*l ; h*h*256*256 + h*l*256 + h*l*256 + l*l
; l * l sqr8 result, arg
sqr8 result, arg ; 19 cyc
; 256 * 256 * hl * hl sqr8 result + 2, arg + 1
sqr8 result + 2, arg + 1 ; 19 cyc
; 2 * h * l * 256 imul8 inter, arg + 1, arg, xe
imul8 inter, arg + 1, arg, xe ; 81-92 / 50-55 cyc add16 result + 1, result + 1, inter
add16 result + 1, result + 1, inter ; 20 cyc add_carry result + 3
add_carry result + 3 ; 2-7 cyc add16 result + 1, result + 1, inter
add16 result + 1, result + 1, inter ; 20 cyc add_carry result + 3
add_carry result + 3 ; 2-7 cyc
.endscope
.endmacro
rts ; 6 cyc
; min base ram: 5 + 163 + 3 = 171
; max base ram: 7 + 184 + 18 = 209
; min ext ram: 5 + 132 + 3 = 140
; max ext ram: 7 + 147 + 18 = 172
.macro sqr16_impl result, arg, xe
.scope
; arg ; 16-bit arg
negated = FR1
; result ; 32-bit result
; 5-7 cycles
lda arg + 1 ; 3 cyc
bpl arg_pos ; 2 cyc
jmp arg_neg ; 3 cyc
arg_pos:
sqr16_impl_inner result, arg, xe ; 163-184 / 132-147 cyc
jmp done ; 3 cyc
arg_neg:
copy_neg16 negated, arg ; 18 cyc
sqr16_impl_inner result, negated, xe ; 163-184 / 132-147 cyc
done:
.endscope .endscope
.endmacro .endmacro
.proc imul16_func .proc imul16_func
imul16_impl FR2, FR0, FR1, 0 imul16_impl 0
rts
.endproc .endproc
.proc imul16xe_func .proc imul16xe_func
imul16_impl FR2, FR0, FR1, 1 imul16_impl 1
rts
.endproc .endproc
;.proc sqr16_func .proc sqr16_func
; sqr16_impl FR2, FR0, 0 sqr16_impl 0
; rts .endproc
;.endproc
;.proc sqr16xe_func .proc sqr16xe_func
; sqr16_impl FR2, FR0, 1 sqr16_impl 1
; rts .endproc
;.endproc
; 11-27 cycles ; 11-27 cycles
.macro round16 arg .macro round16 arg
@ -994,17 +983,6 @@ common:
.endproc .endproc
; rounds to 16-bit first!
; input in FR0, 32 bits signed 6.26 fixed
; output in FR0, Atari float
; clobbers a, x, y, FR0, FR1
.proc fixed6_26_to_float
shift_round_16 FR0, 3
copy16 FR0, FR0 + 2
jsr fixed3_13_to_float
rts
.endproc
; input in FR0, Atari float ; input in FR0, Atari float
; output in FR0, 16 bits signed 3.13 fixed ; output in FR0, 16 bits signed 3.13 fixed
; clobbers a, x, y, FR0, FR1 ; clobbers a, x, y, FR0, FR1
@ -1051,6 +1029,14 @@ common:
; zx_zy = 0 ; zx_zy = 0
; dist = 0 ; dist = 0
; iter = 0 ; iter = 0
; lda #00
; ldx #(iter - zx + 1)
;initloop:
; sta zx - 1,x
; dex
; bne initloop
; sta z_buffer_start
; sta z_buffer_end
lda #00 lda #00
sta zx sta zx
@ -1148,11 +1134,13 @@ keep_going:
shift_round_16 zy, 3 shift_round_16 zy, 3
; zx_2 = zx * zx ; zx_2 = zx * zx
sqr16 zx_2, zx + 2
; zy_2 = zy * zy ; zy_2 = zy * zy
sqr16 zy_2, zy + 2
; zx_zy = zx * zy ; zx_zy = zx * zy
fixup_mandelbrot_hotspot: imul16 zx_zy, zx + 2, zy + 2
jmp mandelbrot_hotspot
after_mandelbrot_hotspot:
; dist = zx_2 + zy_2 ; dist = zx_2 + zy_2
add32 dist, zx_2, zy_2 add32 dist, zx_2, zy_2
@ -1248,27 +1236,6 @@ next:
.endproc .endproc
.macro mandelbrot_hotspot_impl xe
; zx_2 = zx * zx
sqr16_impl zx_2, zx + 2, xe
; zy_2 = zy * zy
sqr16_impl zy_2, zy + 2, xe
; zx_zy = zx * zy
imul16_impl zx_zy, zx + 2, zy + 2, xe
jmp mandelbrot::after_mandelbrot_hotspot
.endmacro
.proc mandelbrot_hotspot
mandelbrot_hotspot_impl 0
.endproc
.proc mandelbrot_hotspot_xe
mandelbrot_hotspot_impl 1
.endproc
.macro scale_zoom dest .macro scale_zoom dest
; clobbers X, flags ; clobbers X, flags
.local cont .local cont
@ -1636,7 +1603,7 @@ number_keys:
beq five beq five
cpy #KEY_6 cpy #KEY_6
beq six beq six
jmp letter_keys jmp skip_char
one: one:
ldx #0 ldx #0
@ -1655,21 +1622,7 @@ five:
jmp load_key_viewport jmp load_key_viewport
six: six:
ldx #5 ldx #5
jmp load_key_viewport ; fall through
letter_keys:
cpy #KEY_X
bne not_x
jsr input_x
jmp done
not_x:
cpy #KEY_Y
bne not_y
jsr input_y
jmp done
not_y:
jmp skip_char
load_key_viewport: load_key_viewport:
jsr load_viewport jsr load_viewport
; fall through ; fall through
@ -1679,23 +1632,6 @@ done:
.endproc .endproc
.proc input_x
ldx #col_x
ldy #1
jsr input_number
rts
.endproc
.proc input_y
rts
.endproc
.proc input_number
rts
.endproc
.proc clear_screen .proc clear_screen
; zero the range from framebuffer_top to display_list ; zero the range from framebuffer_top to display_list
lda #.lobyte(framebuffer_top) lda #.lobyte(framebuffer_top)
@ -1743,7 +1679,9 @@ zero_byte_loop:
draw_string_const str_x draw_string_const str_x
copy32 FR0, ox copy32 FR0, ox
jsr fixed6_26_to_float shift_round_16 FR0, 3
copy16 FR0, FR0 + 2
jsr fixed3_13_to_float
jsr FASC jsr FASC
jsr draw_string jsr draw_string
@ -1752,7 +1690,9 @@ zero_byte_loop:
draw_string_const str_y draw_string_const str_y
copy32 FR0, oy copy32 FR0, oy
jsr fixed6_26_to_float shift_round_16 FR0, 3
copy16 FR0, FR0 + 2
jsr fixed3_13_to_float
jsr FASC jsr FASC
jsr draw_string jsr draw_string
@ -1812,7 +1752,7 @@ zero_byte_loop:
rts rts
.endproc .endproc
.proc _mandel_start .proc start
jsr imul8xe_init jsr imul8xe_init
@ -2051,25 +1991,51 @@ update_status:
lda FR0 + 1 lda FR0 + 1
sta elapsed_work + 1 sta elapsed_work + 1
draw_string_const str_space ;jsr IFP
;jsr FASC
;jsr draw_string
.macro do_countdown divisor, digits .macro countdown divisor, digits
ldx #.lobyte(divisor) .scope
ldy #.hibyte(divisor) ; count the hours
lda #.lobyte(digits) ldx #0
sta INBUFF countdown_loop:
lda #.hibyte(digits) lda elapsed_work + 1
sta INBUFF + 1 cmp #.hibyte(divisor)
jsr countdown bcc countdown_done
lda elapsed_work
cmp #.lobyte(divisor)
bcc countdown_done
sec
lda elapsed_work
sbc #.lobyte(divisor)
sta elapsed_work
lda elapsed_work + 1
sbc #.hibyte(divisor)
sta elapsed_work + 1
inx
jmp countdown_loop
countdown_done:
lda digits,x
eor #$80
sta elapsed_digit
lda #.lobyte(elapsed_digit)
sta INBUFF
lda #.hibyte(elapsed_digit)
sta INBUFF + 1
jsr draw_string
.endscope
.endmacro .endmacro
do_countdown 36000, digits_space
do_countdown 3600, digits_zero draw_string_const str_space
countdown 36000, digits_space
countdown 3600, digits_zero
draw_string_const str_h draw_string_const str_h
do_countdown 600, digits_zero countdown 600, digits_zero
do_countdown 60, digits_zero countdown 60, digits_zero
draw_string_const str_m draw_string_const str_m
do_countdown 10, digits_zero countdown 10, digits_zero
do_countdown 1, digits_zero countdown 1, digits_zero
draw_string_const str_s draw_string_const str_s
skipped: skipped:
@ -2131,117 +2097,3 @@ loop:
jmp main_loop jmp main_loop
.endproc .endproc
; digit string in INBUFF
; divisor X/Y
; clobbers temp, calls draw_string
.proc countdown
divisor = temp
stx divisor
sty divisor + 1
; count the hours
ldy #0
countdown_loop:
lda elapsed_work + 1
cmp divisor + 1
beq countdown_lobyte
bcc countdown_done
bcs countdown_inc
countdown_lobyte:
lda elapsed_work
cmp divisor
bcc countdown_done
countdown_inc:
sec
lda elapsed_work
sbc divisor
sta elapsed_work
lda elapsed_work + 1
sbc divisor + 1
sta elapsed_work + 1
iny
jmp countdown_loop
countdown_done:
lda (INBUFF),y
eor #$80
sta elapsed_digit
lda #.lobyte(elapsed_digit)
sta INBUFF
lda #.hibyte(elapsed_digit)
sta INBUFF + 1
jsr draw_string
rts
.endproc
.proc imul8xe_init
bank_switch 0
lda #0
sta EXTENDED_RAM
bank_switch 1
lda #1
sta EXTENDED_RAM
bank_switch 0
lda EXTENDED_RAM
beq init
; no bank switching available, we just overwrite the value in base ram
rts
init:
; patch imul16_func into a forwarding thunk to imul16xe_func
lda #$4c ; 'jmp' opcode
sta imul16_func
lda #.lobyte(imul16xe_func)
sta imul16_func + 1
lda #.hibyte(imul16xe_func)
sta imul16_func + 2
; ditto for sqr16_func -> sqr16xe_func
;lda #$4c ; 'jmp' opcode
;sta sqr16_func
;lda #.lobyte(sqr16xe_func)
;sta sqr16_func + 1
;lda #.hibyte(sqr16xe_func)
;sta sqr16_func + 2
lda #.lobyte(mandelbrot_hotspot_xe)
sta mandelbrot::fixup_mandelbrot_hotspot + 1
lda #.hibyte(mandelbrot_hotspot_xe)
sta mandelbrot::fixup_mandelbrot_hotspot + 2
; create the lookup table
; go through the input set, in four 16KB chunks
arg1 = FR1
arg2 = FR2
result = FR0
lda #$00
sta arg1
sta arg2
sta ptr
lda #$40
sta ptr + 1
; $00 * $00 -> $3f * $ff
bank_switch 0
jsr imul8xe_init_section
; $40 * $00 -> $7f * $ff
bank_switch 1
jsr imul8xe_init_section
; $80 * $00 -> $bf * $ff
bank_switch 2
jsr imul8xe_init_section
; $c0 * $00 -> $ff * $ff
bank_switch 3
jsr imul8xe_init_section
rts
.endproc