refactoring and start on squares
This commit is contained in:
parent
8ad996981a
commit
f903272335
2 changed files with 143 additions and 153 deletions
284
mandel.s
284
mandel.s
|
|
@ -374,65 +374,13 @@ viewport_oy:
|
|||
copy16 dest, FR2 + 2 ; 12 cyc
|
||||
.endmacro
|
||||
|
||||
; Adapted from https://everything2.com/title/Fast+6502+multiplication
|
||||
.macro imul8 dest, arg1, arg2
|
||||
.local under256
|
||||
.local next
|
||||
.local small_product
|
||||
; circa 92 cycles? this doesn't seem right
|
||||
; 81-92 cycles
|
||||
.scope
|
||||
mul_factor_a = arg1
|
||||
mul_factor_x = arg2
|
||||
mul_product_lo = dest
|
||||
mul_product_hi = dest + 1
|
||||
|
||||
lda mul_factor_a ; 3 cyc
|
||||
|
||||
; (a + x)^2/2
|
||||
clc ; 2 cyc
|
||||
adc mul_factor_x ; 3 cyc
|
||||
tax ; 2 cyc
|
||||
bcc under256 ; 2 cyc
|
||||
lda mul_hibyte512,x ; 4 cyc
|
||||
bcs next ; 2 cyc
|
||||
under256:
|
||||
lda mul_hibyte256,x ; 4 cyc
|
||||
sec ; 2 cyc
|
||||
next:
|
||||
sta mul_product_hi ; 3 cyc
|
||||
lda mul_lobyte256,x ; 4 cyc
|
||||
|
||||
; - a^2/2
|
||||
ldx mul_factor_a ; 3 cyc
|
||||
sbc mul_lobyte256,x ; 4 cyc
|
||||
sta mul_product_lo ; 3 cyc
|
||||
lda mul_product_hi ; 3 cyc
|
||||
sbc mul_hibyte256,x ; 4 cyc
|
||||
sta mul_product_hi ; 3 cyc
|
||||
|
||||
; + x & a & 1:
|
||||
; (this is a kludge to correct a
|
||||
; roundoff error that makes odd * odd too low)
|
||||
ldx mul_factor_x ; 3 cyc
|
||||
txa ; 2 cyc
|
||||
and mul_factor_a ; 3 cyc
|
||||
and #1 ; 2 cyc
|
||||
|
||||
clc ; 2 cyc
|
||||
adc mul_product_lo ; 3 cyc
|
||||
bcc small_product ; 2 cyc
|
||||
inc mul_product_hi ; 5 cyc
|
||||
|
||||
; - x^2/2
|
||||
small_product:
|
||||
sec ; 2 cyc
|
||||
sbc mul_lobyte256,x ; 4 cyc
|
||||
sta mul_product_lo ; 3 cyc
|
||||
lda mul_product_hi ; 3 cyc
|
||||
sbc mul_hibyte256,x ; 4 cyc
|
||||
sta mul_product_hi ; 3 cyc
|
||||
.endscope
|
||||
; clobbers a, x
|
||||
.macro sqr8 dest, arg
|
||||
ldx arg
|
||||
lda sqr_lobyte,x
|
||||
sta dest
|
||||
lda sqr_hibyte,x
|
||||
sta dest + 1
|
||||
.endmacro
|
||||
|
||||
; lookup table for top byte -> PORTB value for bank-switch
|
||||
|
|
@ -447,64 +395,121 @@ bank_switch_table:
|
|||
sta PORTB
|
||||
.endmacro
|
||||
|
||||
.macro imul8 dest, arg1, arg2, xe
|
||||
.if xe
|
||||
; using 64KB lookup table
|
||||
; 58-77 cycles
|
||||
; clobbers x, y, dest to dest + 3
|
||||
.scope
|
||||
output = dest
|
||||
ptr = dest + 2 ; scratch space assumed
|
||||
|
||||
; 58-77 cycles
|
||||
; clobbers x, y, dest to dest + 3
|
||||
.macro imul8xe dest, arg1, arg2
|
||||
.local done
|
||||
.local output
|
||||
.local ptr
|
||||
|
||||
output = dest
|
||||
ptr = dest + 2 ; scratch space assumed
|
||||
|
||||
; bottom 14 bits except the LSB are the per-bank table index
|
||||
; add $4000 for the bank pointer
|
||||
lda arg1 ; 3 cyc
|
||||
and #$fe ; 2 cyc
|
||||
sta ptr ; 3 cyc
|
||||
lda arg2 ; 3 cyc
|
||||
and #$3f ; 2 cyc
|
||||
clc ; 2 cyc
|
||||
adc #$40 ; 2 cyc
|
||||
sta ptr + 1 ; 3 cyc
|
||||
|
||||
; top 2 bits are the table bank selector
|
||||
ldx arg2 ; 3 cyc
|
||||
lda bank_switch_table,x ; 4 cyc
|
||||
sta PORTB ; 4 cyc
|
||||
; bottom 14 bits except the LSB are the per-bank table index
|
||||
; add $4000 for the bank pointer
|
||||
lda arg1 ; 3 cyc
|
||||
and #$fe ; 2 cyc
|
||||
sta ptr ; 3 cyc
|
||||
lda arg2 ; 3 cyc
|
||||
and #$3f ; 2 cyc
|
||||
clc ; 2 cyc
|
||||
adc #$40 ; 2 cyc
|
||||
sta ptr + 1 ; 3 cyc
|
||||
|
||||
; top 2 bits are the table bank selector
|
||||
ldx arg2 ; 3 cyc
|
||||
lda bank_switch_table,x ; 4 cyc
|
||||
sta PORTB ; 4 cyc
|
||||
|
||||
|
||||
; copy the entry into output
|
||||
ldy #0 ; 2 cyc
|
||||
lda (ptr),y ; 5 cyc
|
||||
sta output ; 3 cyc
|
||||
iny ; 2 cyc
|
||||
lda (ptr),y ; 5 cyc
|
||||
sta output+1 ; 3 cyc
|
||||
; copy the entry into output
|
||||
ldy #0 ; 2 cyc
|
||||
lda (ptr),y ; 5 cyc
|
||||
sta output ; 3 cyc
|
||||
iny ; 2 cyc
|
||||
lda (ptr),y ; 5 cyc
|
||||
sta output+1 ; 3 cyc
|
||||
|
||||
; note: we are not restoring memory to save 6 cycles!
|
||||
; this means those 16kb have to be switched back to base RAM
|
||||
; if we need to use them anywhere else
|
||||
;;; restore memory
|
||||
;;lda #$81 ; 2 cyc - disabled
|
||||
;;sta PORTB ; 4 cyc - disabled
|
||||
; note: we are not restoring memory to save 6 cycles!
|
||||
; this means those 16kb have to be switched back to base RAM
|
||||
; if we need to use them anywhere else
|
||||
;;; restore memory
|
||||
;;lda #$81 ; 2 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
|
||||
; check that 1 bit we skipped to fit into space
|
||||
lda arg1 ; 3 cyc
|
||||
and #1 ; 2 cyc
|
||||
beq done ; 2 cyc
|
||||
|
||||
; add the second param one last time for the skipped bit
|
||||
clc ; 2 cyc
|
||||
lda arg2 ; 3 cyc
|
||||
adc output ; 3 cyc
|
||||
sta output ; 3 cyc
|
||||
lda #0 ; 2 cyc
|
||||
adc output+1 ; 3 cyc
|
||||
sta output+1 ; 3 cyc
|
||||
; add the second param one last time for the skipped bit
|
||||
clc ; 2 cyc
|
||||
lda arg2 ; 3 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
|
||||
.else
|
||||
; Using base 48k RAM compatibility mode
|
||||
; Small table of half squares
|
||||
; Adapted from https://everything2.com/title/Fast+6502+multiplication
|
||||
; 81-92 cycles
|
||||
.scope
|
||||
mul_factor_a = arg1
|
||||
mul_factor_x = arg2
|
||||
mul_product_lo = dest
|
||||
mul_product_hi = dest + 1
|
||||
|
||||
lda mul_factor_a ; 3 cyc
|
||||
|
||||
; (a + x)^2/2
|
||||
clc ; 2 cyc
|
||||
adc mul_factor_x ; 3 cyc
|
||||
tax ; 2 cyc
|
||||
bcc under256 ; 2 cyc
|
||||
lda mul_hibyte512,x ; 4 cyc
|
||||
bcs next ; 2 cyc
|
||||
under256:
|
||||
lda mul_hibyte256,x ; 4 cyc
|
||||
sec ; 2 cyc
|
||||
next:
|
||||
sta mul_product_hi ; 3 cyc
|
||||
lda mul_lobyte256,x ; 4 cyc
|
||||
|
||||
; - a^2/2
|
||||
ldx mul_factor_a ; 3 cyc
|
||||
sbc mul_lobyte256,x ; 4 cyc
|
||||
sta mul_product_lo ; 3 cyc
|
||||
lda mul_product_hi ; 3 cyc
|
||||
sbc mul_hibyte256,x ; 4 cyc
|
||||
sta mul_product_hi ; 3 cyc
|
||||
|
||||
; + x & a & 1:
|
||||
; (this is a kludge to correct a
|
||||
; roundoff error that makes odd * odd too low)
|
||||
ldx mul_factor_x ; 3 cyc
|
||||
txa ; 2 cyc
|
||||
and mul_factor_a ; 3 cyc
|
||||
and #1 ; 2 cyc
|
||||
|
||||
clc ; 2 cyc
|
||||
adc mul_product_lo ; 3 cyc
|
||||
bcc small_product ; 2 cyc
|
||||
inc mul_product_hi ; 5 cyc
|
||||
|
||||
; - x^2/2
|
||||
small_product:
|
||||
sec ; 2 cyc
|
||||
sbc mul_lobyte256,x ; 4 cyc
|
||||
sta mul_product_lo ; 3 cyc
|
||||
lda mul_product_hi ; 3 cyc
|
||||
sbc mul_hibyte256,x ; 4 cyc
|
||||
sta mul_product_hi ; 3 cyc
|
||||
.endscope
|
||||
.endif
|
||||
.endmacro
|
||||
|
||||
.proc imul8xe_init
|
||||
|
|
@ -632,7 +637,13 @@ inner_loop:
|
|||
|
||||
.endproc
|
||||
|
||||
.proc imul16_func
|
||||
.macro imul16_impl xe
|
||||
.local arg1
|
||||
.local arg2
|
||||
.local result
|
||||
.local inter
|
||||
.local arg1_pos
|
||||
.local arg2_pos
|
||||
arg1 = FR0 ; 16-bit arg (clobbered)
|
||||
arg2 = FR1 ; 16-bit arg (clobbered)
|
||||
result = FR2 ; 32-bit result
|
||||
|
|
@ -643,20 +654,20 @@ inner_loop:
|
|||
; h1*256*(h2*256 + l2) + l1*(h2*256 + l2)
|
||||
; h1*h2*256*256 + h1*l2*256 + h2*l1*256 + l1*l2
|
||||
|
||||
imul8 result, arg1, arg2
|
||||
imul8 result, arg1, arg2, xe
|
||||
lda #0
|
||||
sta result + 2
|
||||
sta result + 3
|
||||
|
||||
imul8 inter, arg1 + 1, arg2
|
||||
imul8 inter, arg1 + 1, arg2, xe
|
||||
add16 result + 1, result + 1, inter
|
||||
add_carry result + 3
|
||||
|
||||
imul8 inter, arg1, arg2 + 1
|
||||
imul8 inter, arg1, arg2 + 1, xe
|
||||
add16 result + 1, result + 1, inter
|
||||
add_carry result + 3
|
||||
|
||||
imul8 inter, arg1 + 1, arg2 + 1
|
||||
imul8 inter, arg1 + 1, arg2 + 1, xe
|
||||
add16 result + 2, result + 2, inter
|
||||
|
||||
; In case of negative inputs, adjust high word
|
||||
|
|
@ -671,47 +682,14 @@ arg1_pos:
|
|||
arg2_pos:
|
||||
|
||||
rts ; 6 cyc
|
||||
.endmacro
|
||||
|
||||
.proc imul16_func
|
||||
imul16_impl 0
|
||||
.endproc
|
||||
|
||||
.proc imul16xe_func
|
||||
arg1 = FR0 ; 16-bit arg (clobbered)
|
||||
arg2 = FR1 ; 16-bit arg (clobbered)
|
||||
result = FR2 ; 32-bit result
|
||||
inter = temp2
|
||||
|
||||
; h1l1 * h2l2
|
||||
; (h1*256 + 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
|
||||
|
||||
imul8xe result, arg1, arg2
|
||||
lda #0
|
||||
sta result + 2
|
||||
sta result + 3
|
||||
|
||||
imul8xe inter, arg1 + 1, arg2
|
||||
add16 result + 1, result + 1, inter
|
||||
add_carry result + 3
|
||||
|
||||
imul8xe inter, arg1, arg2 + 1
|
||||
add16 result + 1, result + 1, inter
|
||||
add_carry result + 3
|
||||
|
||||
imul8xe inter, arg1 + 1, arg2 + 1
|
||||
add16 result + 2, result + 2, inter
|
||||
|
||||
; In case of negative inputs, adjust high word
|
||||
; https://stackoverflow.com/a/28827013
|
||||
lda arg1 + 1
|
||||
bpl arg1_pos
|
||||
sub16 result + 2, result + 2, arg2
|
||||
arg1_pos:
|
||||
lda arg2 + 1
|
||||
bpl arg2_pos
|
||||
sub16 result + 2, result + 2, arg1
|
||||
arg2_pos:
|
||||
|
||||
rts ; 6 cyc
|
||||
imul16_impl 1
|
||||
.endproc
|
||||
|
||||
.macro round16 arg
|
||||
|
|
|
|||
12
tables.js
12
tables.js
|
|
@ -22,7 +22,10 @@ console.log(
|
|||
.export mul_lobyte256
|
||||
.export mul_hibyte256
|
||||
.export mul_hibyte512
|
||||
.export sqr_lobyte
|
||||
.export sqr_hibyte
|
||||
|
||||
; (i * i + 1) / 2 for the multiplier
|
||||
.align 256
|
||||
mul_lobyte256:
|
||||
${db((i) => squares[i] & 0xff)}
|
||||
|
|
@ -35,4 +38,13 @@ ${db((i) => (squares[i] >> 8) & 0xff)}
|
|||
mul_hibyte512:
|
||||
${db((i) => (squares[i + 256] >> 8) & 0xff)}
|
||||
|
||||
; (i * i) for the plain squares
|
||||
.align 256
|
||||
sqr_lobyte:
|
||||
${db((i) => (i * i) & 0xff)}
|
||||
|
||||
.align 256
|
||||
sqr_hibyte:
|
||||
${db((i) => ((i * i) >> 8) & 0xff)}
|
||||
|
||||
`);
|
||||
|
|
|
|||
Loading…
Reference in a new issue