ZX Spectrum Next - ULA double buffering example v2 (advanced techniques: DMA + custom IM1 interrupt)

flipULA2.asm

    DEFINE USE_DMA_TO_CLEAR_SCREEN  ; comment out to get LDIR clear version
    DEFINE USE_DOUBLE_BUFFERING     ; comment out to see single-buffer redraw issues

    DEVICE ZXSPECTRUMNEXT

BORDER  MACRO color?
            ld  a,color?
            out (254),a
        ENDM

    ORG $8000   ; this example does map Bank5/Bank7 ULA to $4000 (to use "pixelad" easily)
start:
    nextreg $07,3           ; ensure 28MHz
    ; init zxnDMA to known state and with some values partially pre-loaded
    ld      hl,dmaInitData
    ld      bc,(dmaInitSize<<8) + $6B
    otir
    ; init IM1 interrupt mode with RAM mapped into ROM area to provide our custom handler
    nextreg $50,rst38page   ; map 8ki page with IM1 handler to $0000 ROM area
    im      1               ; make sure IM1 interrupt mode is used
    nextreg $22,%00000'110  ; disable ULA interrupt, enable video-line interrupt, line.MSB=0
    nextreg $23,0           ; line.LSB=0 ("moves" IM1 interrupt about ~100T ahead of pixel[0,0])
    ei
mainLoop:
    halt                    ; wait for custom interrupt (in IM1 mode mapped in ROM area)
    call    ClearUlaBuffer  ; clear screen (make it blink without double-buffering)
    call    drawDots
    jr      mainLoop

ClearUlaBuffer:
    BORDER  1               ; blue border to "time" DMA ULA clearing
    IFDEF USE_DMA_TO_CLEAR_SCREEN
        xor a
        ld  (dmaSourceByte),a
        ld  hl,dmaClearPixels
        ld  bc,(dmaClearPixelsSize<<8) + $6B
        otir
        ld  a,$05           ; black paper, cyan ink
        ld  (dmaSourceByte),a
        ld  hl,dmaClearAttr
        ld  bc,(dmaClearAttrSize<<8) + $6B
        otir
    ELSE
        ld  hl,$4000
        ld  de,$4001
        ld  bc,$1800
        ld  (hl),l
        ldir
        ld  (hl),$05        ; black paper, cyan ink
        ld  bc,$02FF
        ldir
    ENDIF
    BORDER 0
    ret

drawDots:
    ; adjust position for next frame
    ld      de,(DotPos)
    inc     e               ; ++X
    ld      (DotPos),de
    BORDER  2               ; red border to "time" dot drawings
    ; draw dot on every second line
    ld      b,192/2
dotLoop:
    pixelad
    setae
    ld      (hl),a
    inc     d
    inc     d
    djnz    dotLoop
    BORDER  0
    ret

dmaSourceByte:
    db      0

ULABank:
    db  10          ; holds current ULA screen in use
DotPos:
    db  0, 0        ; X=0, Y=0

dmaInitData:
    hex C3 C3 C3 C3 C3 C3       ; reset DMA from any possible state (6x reset command)
    db  %10'0'0'0010            ; WR5 = stop on end of block, /CE only
    db  %1'0'000000             ; WR3 = all disabled (important only for real DMA chips)
    db  %0'01'11'1'01           ; WR0 = A->B transfer
    dw  dmaSourceByte           ; + port address A (byte to fill memory with)
    db  0                       ; + size.LO = always zero
    db  %0'1'10'0'100           ; WR1 = A address fixed, memory
    db  2                       ; + custom 2T timing
    db  %0'1'01'0'000           ; WR2 = B address ++, memory
    db  2                       ; + custom 2T timing
    db  %1'01'0'01'01           ; WR4 = continuous mode
    db  0                       ; + port address B.LO = always zero
dmaInitSize: EQU $ - dmaInitData

dmaClearPixels:
    db  %1'01'0'10'01, $40      ; WR4: addresB.HI = $40 (no other change)
    db  %0'10'00'1'01, $18      ; WR0: size.HI = $18 (no other change)
    hex CF 87                   ; LOAD + ENABLE (transfer is executed)
dmaClearPixelsSize: EQU $ - dmaClearPixels

dmaClearAttr:
    db  %1'01'0'10'01, $58      ; WR4: addresB.HI = $58 (no other change)
    db  %0'10'00'1'01, $03      ; WR0: size.HI = $03 (no other change)
    hex CF 87                   ; LOAD + ENABLE (transfer is executed)
dmaClearAttrSize: EQU $ - dmaClearAttr

    ; create new custom interrupt handler routine at $0038
    ORG $A038       ; put it into next page after main code at $8000 at +$38 offset
    DISP $0038      ; but assemble it as if destined for address $0038
rst38:
rst38page:  EQU $$$$    ; remember "physical" 8ki page where this code is landing (it's 5)
    IFDEF USE_DOUBLE_BUFFERING  ; if double-buffering is OFF, then nothing to do here
        push    af
        BORDER  5       ; cyan border to "time" IM1 handler
        ; switch currently displayed ULA (ASAP to make it switch before first line will start)
        ld a,(ULABank)  ; Get screen to display this frame
        ; if A==14, then do NR_69=64 (display Bank7), if A==10: NR_69=0 (display Bank5)
        ; its %0000'1110 vs %0000'1010 in binary, so extract bit2 and move it to bit6
        and %0000'0100  ; $04 from A=14, $00 from A=10
        swapnib         ; bit6 set from bit2
        nextreg $69,a   ; Select Timex/ULA screen to show
        ; switch the active drawing buffer at $4000 to make next drawing into new buffer
        ld a,(ULABank)
        xor 10^14       ; alternate between 10 and 14
        ld (ULABank),a
        nextreg $52,a   ; map the new "backbuffer" to $4000 (for next drawing)
        ; exit interrupt handler
        BORDER 0
        pop af
    ENDIF
    ei
    reti
    ENDT                ; end the DISP block

    SAVENEX OPEN "flipULA2.nex", start, $BF00 : SAVENEX AUTO : SAVENEX CLOSE