优化GameBoy Z80中的位操作算法

Rem*_*emy 13 assembly bit-manipulation z80 gameboy

不是一个家庭作业问题,而是我正在开发的游戏.

我有两种16位RGB颜色,并希望根据其他六个四位数量改变它们的六个通道.算法简单但乏味; 我正在寻找一种方法来通过一次做更多有用的工作来优化它.

高级概述:

  • hl指向四个颜色字节.[hl] = %gggrrrrr,[hl+1] = %0bbbbbgg,[hl+2] = %GGGRRRRR,和[hl+3] = %0BBBBBGG.(这是两种颜色,rgbRGB.)
  • bc指向三个delta字节.[bc] = %hhhhaaaa,[bc+1] = %ddddssss[bc+2] = %ppppqqqq.(这是6个Delta值h,a,d,s,p,和q).
  • 因此,有六个5位颜色通道值和六个4位增量值.我想将每个颜色通道C与一个delta值D配对,并且像这样改变C:C' = C +(D&%11) - ((D&%1100)>> 2),但保持C在其5之内-bit bounds [0,31].我实际上并不关心它们是如何配对的:任何方便的一对一配对都很好.如果C +((D&%1100)>> 2) - (D&%11)以某种方式允许更优雅的算法,我会好的.

如果我隔离颜色通道Ç在寄存器d和一个增量值d在寄存器e,那么该例程将做该对的变化:

VaryColorChannelByDV:
; d = color, e = DV
; a <- d + (e & %11) - (e >> 2), clamped to [0, 31]
    ld a, e
    and %11   ; a <- (e & %11)
    add d   ; a <- d + (e & %11)
    srl e
    srl e   ; e <- e >> 2
    sub e   ; a <- d + (e & %11) - (e >> 2)
    jr c, .zero   ; a < 0, clamp to 0
    cp 32
    ret c   ; 0 <= a < 32
    ld a, 31   ; a >= 32, clamp to 31
    ret
.zero
    xor a
    ret
Run Code Online (Sandbox Code Playgroud)

到目前为止,我有一个通用例程,可以将任何DV应用于任何颜色通道; 然后是三个隔离红色,绿色或蓝色通道并将特定DV应用于它们的例程; 最后是一个主程序,它选出六个DV并用它们调用适当的通道修改程序.这"足够好",但我确信还有改进的余地.执行速度似乎不是问题,但我想减少代码大小(当然,删除冗余指令也会提高速度).是否有任何asm位操作技巧会有所帮助?

这是完整的代码:

GetColorChannelVariedByDV:
; d = color, e = DV
; a <- d + (e & %11) - (e & %1100 >> 2), clamped to [0, 31]
    ld a, e
    and %11
    add d
    srl e
    srl e
    sub e
    jr c, .zero
    cp 32
    ret c
    ld a, 31
    ret
.zero
    xor a
    ret

VaryRedByDV:
;;; e = DV
;;; [hl+0] = gggr:rrrr
;;; [hl+1] = 0bbb:bbgg
; store red in d
    ld a, [hl]
    and %00011111
    ld d, a
; vary d according to e
    call GetColorChannelVariedByDV
; store a back in red
    ld d, a
    ld a, [hl]
    and %11100000
    or d
    ld [hl], a
    ret

VaryGreenByDV:
;;; e = DV
;;; [hl+0] = gggr:rrrr
;;; [hl+1] = 0bbb:bbgg
; store green in d
    ld a, [hli]
    and %11100000
    srl a
    swap a
    ld d, a ; d = 00000ggg
    ld a, [hld]
    and %00000011
    swap a
    srl a
    or d
    ld d, a
; vary d according to e
    call GetColorChannelVariedByDV
; store a back in green
    sla a
    swap a
    ld d, a
    and %11100000
    ld e, a
    ld a, d
    and %00000011
    ld d, a
    ld a, [hl]
    and %00011111
    or e
    ld [hli], a
    ld a, [hl]
    and %11111100
    or d
    ld [hld], a
    ret

VaryBlueByDV:
;;; e = DV
;;; [hl+0] = gggr:rrrr
;;; [hl+1] = 0bbb:bbgg
; store blue in d
    inc hl
    ld a, [hl]
    and %01111100
    srl a
    srl a
    ld d, a
; vary d according to e
    call GetColorChannelVariedByDV
; store a back in blue
    ld d, a
    sla d
    sla d
    ld a, [hl]
    and %10000011
    or d
    ld [hl], a
    dec hl
    ret

VaryColorsByDVs::
; hl = colors
; [hl+0] = gggr:rrrr
; [hl+1] = 0bbb:bbgg
; [hl+2] = GGGR:RRRR
; [hl+3] = 0BBB:BBGG
; bc = DVs
; [bc+0] = hhhh:aaaa
; [bc+1] = dddd:ssss
; [bc+2] = pppp:qqqq

;;; LiteRed ~ hDV, aka, rrrrr ~ hhhh
; store hDV in e
    ld a, [bc]
    swap a
    and %1111
    ld e, a
; vary LiteRed by e
    call VaryRedByDV

;;; LiteGrn ~ aDV, aka, ggggg ~ aaaa
; store aDV in e
    ld a, [bc]
    and %1111
    ld e, a
; vary LiteGrn by e
    call VaryGreenByDV

;;; move from h/a DV to d/s DV
    inc bc

;;; LiteBlu ~ dDV, aka, bbbbb ~ dddd
; store dDV in e
    ld a, [bc]
    swap a
    and %1111
    ld e, a
; vary LiteBlu by e
    call VaryBlueByDV

;;; Move from Lite color to Dark color
    inc hl
    inc hl

;;; DarkRed ~ sDV, aka, RRRRR ~ ssss
; store sDV in e
    ld a, [bc]
    and %1111
    ld e, a
; vary DarkRed by e
    call VaryRedByDV

;;; move from d/s DV to p/q DV
    inc bc

;;; DarkGrn ~ pDV, aka, GGGGG ~ pppp
; store pDV in e
    ld a, [bc]
    swap a
    and %1111
    ld e, a
; vary DarkGrn by e
    call VaryGreenByDV

;;; DarkBlu ~ qDV, aka, BBBBB ~ qqqq
; store qDV in e
    ld a, [bc]
    and %1111
    ld e, a
; vary DarkBlu by e
    call VaryBlueByDV

    ret
Run Code Online (Sandbox Code Playgroud)

jac*_*bly 9

我现在能想出的最小值是57个字节:

VaryColorsByDVs::
; hl = colors
; [hl+0] = gggr:rrrr
; [hl+1] = 0bbb:bbgg
; [hl+2] = GGGR:RRRR
; [hl+3] = 0BBB:BBGG
; bc = DVs
; [bc+0] = hhhh:aaaa
; [bc+1] = dddd:ssss
; [bc+2] = pppp:qqqq
    ld a, 2 ; -floor($100/3)*6 mod $100
.next:
    sla [hl]
    inc hl
    rl [hl]
.loop:
    push af
    rrca
    ld a, [bc]
    jr nc, .skip
    swap a
    inc bc
.skip:
    rlca
    ld d, a
    and %00011000
    ld e, a
    ld a, d
    rlca
    rlca
    and %00011000
    add a, [hl]
    jr nc, .noOverflow
    or %11111000
.noOverflow:
    sub e
    jr nc, .noUnderflow
    and %00000111
.noUnderflow:
    dec hl
    ld de, 5
.rotate:
    add a, a
    rl [hl]
    adc a, d
    dec e
    jr nz, .rotate
    inc hl
    ld [hl], a
    pop af
    add a, 85 ; floor($100/3)
    jr nc, .loop
    ret z
    inc hl
    jr .next
Run Code Online (Sandbox Code Playgroud)

修复Ped7g的注释只需要4个字节,总共61个字节:

VaryColorsByDVs::
; hl = colors
; [hl+0] = gggr:rrrr
; [hl+1] = 0bbb:bbgg
; [hl+2] = GGGR:RRRR
; [hl+3] = 0BBB:BBGG
; bc = DVs
; [bc+0] = hhhh:aaaa
; [bc+1] = dddd:ssss
; [bc+2] = pppp:qqqq
    ld a, 2 ; -floor($100/3)*6 mod $100
.next:
    sla [hl]
    inc hl
    rl [hl]
.loop:
    push af
    rrca
    ld a, [bc]
    jr nc, .skip
    swap a
    inc bc
.skip:
    ld d, a
    and %00001100
    ld e, a
    ld a, d
    rlca
    rlca
    and %00001100
    sub e
    add a, a
    jr nc, .positive
.negative:
    add a, [hl]
    jr c, .continue
    and %00000111
    db $38 ; jr c,
.positive:
    add a, [hl]
    jr nc, .continue
    or %11111000
.continue:
    dec hl
    ld de, 5
.rotate:
    add a, a
    rl [hl]
    adc a, d
    dec e
    jr nz, .rotate
    inc hl
    ld [hl], a
    pop af
    add a, 85 ; floor($100/3)
    jr nc, .loop
    ret z
    inc hl
    jr .next
Run Code Online (Sandbox Code Playgroud)