ACE blitCopy

/**
 * Blit without mask - BltBitMap equivalent, but less safe
 * Channels:
 *  A: mask const, read disabled
 *  B: src  read
 *  C: dest read
 *  D: dest write
 * Descending mode is used under 2 cases:
 *  - Blit needs shifting to left with previous data coming from right (ubSrcDelta > ubDstDelta)
 *  - Ascending right mask shifted more than 16 bits
 * Source and destination regions should not overlap.
 * Function is slightly slower (~0.5 - 1.5ms) than bltBitMap:
 *  - Pre-loop calculations take ~50us on ASC mode, ~80us on DESC mode
 *  - Rewriting to assembly could speed things up a bit
 */
BYTE _blitCopy(
    tBitMap *pSrc, WORD wSrcX, WORD wSrcY,
    tBitMap *pDst, WORD wDstX, WORD wDstY, WORD wWidth, WORD wHeight,
    UBYTE ubMinterm, UBYTE ubMask, UWORD uwLine, char *szFile
) {
    if(!blitCheck(pSrc, wSrcX, wSrcY, pDst, wDstX, wDstY, wWidth, wHeight, uwLine, szFile))
        return 0;
    // Helper vars
    UWORD uwBlitWords, uwBlitWidth, uwSrcOffs, uwDstOffs;
    UBYTE ubShift, ubSrcDelta, ubDstDelta, ubWidthDelta, ubMaskFShift, ubMaskLShift, ubPlane;
    // Blitter register values
    UWORD uwBltCon0, uwBltCon1, uwFirstMask, uwLastMask;
    WORD wSrcModulo, wDstModulo;

    ubSrcDelta = wSrcX & 0xF;
    ubDstDelta = wDstX & 0xF;
    ubWidthDelta = (ubSrcDelta + wWidth) & 0xF;

    if(ubSrcDelta > ubDstDelta || ((wWidth+ubDstDelta+15) & 0xFFF0)-(wWidth+ubSrcDelta) > 16) {
        uwBlitWidth = (wWidth+(ubSrcDelta>ubDstDelta?ubSrcDelta:ubDstDelta)+15) & 0xFFF0;
        uwBlitWords = uwBlitWidth >> 4;

        ubMaskFShift = ((ubWidthDelta+15)&0xF0)-ubWidthDelta;
        ubMaskLShift = uwBlitWidth - (wWidth+ubMaskFShift);
        uwFirstMask = 0xFFFF << ubMaskFShift;
        uwLastMask = 0xFFFF >> ubMaskLShift;
        if(ubMaskLShift > 16) // Fix for 2-word blits
            uwFirstMask &= 0xFFFF >> (ubMaskLShift-16);

        ubShift = uwBlitWidth - (ubDstDelta+wWidth+ubMaskFShift);
        uwBltCon1 = ubShift << BSHIFTSHIFT | BLITREVERSE;

        uwSrcOffs = pSrc->BytesPerRow * (wSrcY+wHeight-1) + ((wSrcX+wWidth+ubMaskFShift-1)>>3);
        uwDstOffs = pDst->BytesPerRow * (wDstY+wHeight-1) + ((wDstX+wWidth+ubMaskFShift-1) >> 3);
    }
    else {
        uwBlitWidth = (wWidth+ubDstDelta+15) & 0xFFF0;
        uwBlitWords = uwBlitWidth >> 4;

        ubMaskFShift = ubSrcDelta;
        ubMaskLShift = uwBlitWidth-(wWidth+ubSrcDelta);

        uwFirstMask = 0xFFFF >> ubMaskFShift;
        uwLastMask = 0xFFFF << ubMaskLShift;

        ubShift = ubDstDelta-ubSrcDelta;
        uwBltCon1 = ubShift << BSHIFTSHIFT;

        uwSrcOffs = pSrc->BytesPerRow * wSrcY + (wSrcX>>3);
        uwDstOffs = pDst->BytesPerRow * wDstY + (wDstX>>3);
    }

    uwBltCon0 = (ubShift << ASHIFTSHIFT) | USEB|USEC|USED | ubMinterm;
    wSrcModulo = pSrc->BytesPerRow - (uwBlitWords<<1);
    wDstModulo = pDst->BytesPerRow - (uwBlitWords<<1);

    ubMask &= 0xFF >> (8- (pSrc->Depth < pDst->Depth? pSrc->Depth: pDst->Depth));
    ubPlane = 0;
    while(ubMask) {
        if(ubMask & 1)
            g_sBlitManager.pBlitterSetFn(
                uwBltCon0, uwBltCon1,                  // bltconX
                uwFirstMask, uwLastMask,               // bltaXwm
                0, wSrcModulo, wDstModulo, wDstModulo, // bltXmod
                0,                                     // bltapt
                pSrc->Planes[ubPlane] + uwSrcOffs,     // bltbpt
                pDst->Planes[ubPlane] + uwDstOffs,     // bltcpt
                pDst->Planes[ubPlane] + uwDstOffs,     // bltdpt
                0xFFFF, 0, 0,                          // bltXdat
                (wHeight << 6) | uwBlitWords           // bltsize
            );
        ubMask >>= 1;
        ++ubPlane;
    }
    return 1;
}