*(ushort *) vbuff = (ushort)(pdc->pencolor);                *(vbuff + 2) = (BYTE)(pdc->pencolor >> 16);                vbuff += 3;                *(ushort *) vbuff = (ushort)(pdc->brushcolor);                *(vbuff + 2) = (BYTE)(pdc->brushcolor >> 16);                vbuff += 3;            }            break;        case 4:            for (i = 0; i < size; i += 8) {                *(uint *) vbuff = (uint)(pdc->pencolor);                vbuff += 4;                *(uint *) vbuff = (uint)(pdc->brushcolor);                vbuff += 4;            }            break;    }    PutSavedBoxOnDC (hdc, x, y, H_V ? w_h : 1, H_V ? 1 : w_h, bitmap);}/*  * This function performs a fast box scaling. * This is a DDA-based algorithm; Iteration over target bitmap. * * This function comes from SVGALib, Copyright 1993 Harm Hanemaayer *//* We use the 32-bit to 64-bit multiply and 64-bit to 32-bit divide of the * 386 (which gcc doesn't know well enough) to efficiently perform integer * scaling without having to worry about overflows. */#if defined(__GNUC__) && defined(i386)static inline int muldiv64(int m1, int m2, int d){    /* int32 * int32 -> int64 / int32 -> int32 */    int result;    int dummy;    __asm__ __volatile__ (               "imull %%edx\n\t"               "idivl %4\n\t"  :               "=a"(result), "=d"(dummy)        /* out */  :               "0"(m1), "1"(m2), "g"(d)                /* in */               /***rjr***:               "ax", "dx"***/        /* mod */        );    return result;}#elsestatic inline int muldiv64 (int m1, int m2, int d){    long long int mul = (long long int) m1 * m2;    return (int) (mul / d);}#endif/* Init a bitmap as a normal bitmap  */BOOL GUIAPI InitBitmap (HDC hdc, Uint32 w, Uint32 h, Uint32 pitch, BYTE* bits, PBITMAP bmp){    int depth = GetGDCapability (hdc, GDCAP_DEPTH);    int bpp = GetGDCapability (hdc, GDCAP_BPP);    if (w == 0 || h == 0)        return FALSE;    if (bits && pitch) {        if (bpp * w > pitch)            return FALSE;        bmp->bmBits = bits;        bmp->bmPitch = pitch;    }    else if (!bits) {        bmpComputePitch (depth, w, &bmp->bmPitch, TRUE);        if (!(bmp->bmBits = malloc (bmp->bmPitch * h)))            return FALSE;    }    else /* bits is not zero, but pitch is zero */        return FALSE;    bmp->bmType = BMP_TYPE_NORMAL;    bmp->bmWidth = w;    bmp->bmHeight = h;    bmp->bmBitsPerPixel = depth;    bmp->bmBytesPerPixel = bpp;    return TRUE;}BOOL ScaleBitmap (BITMAP *dst, const BITMAP *src){    BYTE *dp1 = src->bmBits;    BYTE *dp2 = dst->bmBits;    int xfactor;    int yfactor;    if (dst->bmWidth == 0 || dst->bmHeight == 0)        return TRUE;    if (dst->bmBytesPerPixel != src->bmBytesPerPixel)        return FALSE;    xfactor = muldiv64 (src->bmWidth, 65536, dst->bmWidth);         /* scaled by 65536 */    yfactor = muldiv64 (src->bmHeight, 65536, dst->bmHeight);       /* scaled by 65536 */    switch (dst->bmBytesPerPixel) {    case 1:        {            int y, sy;            sy = 0;            for (y = 0; y < dst->bmHeight;) {                int sx = 0;                BYTE *dp2old = dp2;                int x;                x = 0;                while (x < dst->bmWidth - 8) {#if 0 // defined(__GNUC__) && defined(i386)                    /* This saves just a couple of cycles per */                    /* pixel on a 486, but I couldn't resist. */                    __asm__ __volatile__("movl %4, %%eax\n\t"                                         "shrl $16, %%eax\n\t"                                         "addl %5, %4\n\t"                                         "movb (%3, %%eax), %%al\n\t"                                         "movb %%al, (%1, %2)\n\t"                                         "movl %4, %%eax\n\t"                                         "shrl $16, %%eax\n\t"                                         "addl %5, %4\n\t"                                         "movb (%3, %%eax), %%al\n\t"                                         "movb %%al, 1 (%1, %2)\n\t"                                         "movl %4, %%eax\n\t"                                         "shrl $16, %%eax\n\t"                                         "addl %5, %4\n\t"                                         "movb (%3, %%eax), %%al\n\t"                                         "movb %%al, 2 (%1, %2)\n\t"                                         "movl %4, %%eax\n\t"                                         "shrl $16, %%eax\n\t"                                         "addl %5, %4\n\t"                                         "movb (%3, %%eax), %%al\n\t"                                         "movb %%al, 3 (%1, %2)\n\t"                                         "movl %4, %%eax\n\t"                                         "shrl $16, %%eax\n\t"                                         "addl %5, %4\n\t"                                         "movb (%3, %%eax), %%al\n\t"                                         "movb %%al, 4 (%1, %2)\n\t"                                         "movl %4, %%eax\n\t"                                         "shrl $16, %%eax\n\t"                                         "addl %5, %4\n\t"                                         "movb (%3, %%eax), %%al\n\t"                                         "movb %%al, 5 (%1, %2)\n\t"                                         "movl %4, %%eax\n\t"                                         "shrl $16, %%eax\n\t"                                         "addl %5, %4\n\t"                                         "movb (%3, %%eax), %%al\n\t"                                         "movb %%al, 6 (%1, %2)\n\t"                                         "movl %4, %%eax\n\t"                                         "shrl $16, %%eax\n\t"                                         "addl %5, %4\n\t"                                         "movb (%3, %%eax), %%al\n\t"                                         "movb %%al, 7 (%1, %2)\n\t"                                         :        /* output */                                         :        /* input */                                         "ax"(0), "r"(dp2), "r"(x), "r"(dp1),                                         "r"(sx), "r"(xfactor)                                         :"ax", "4"                    );#else                    *(dp2 + x) = *(dp1 + (sx >> 16));                    sx += xfactor;                    *(dp2 + x + 1) = *(dp1 + (sx >> 16));                    sx += xfactor;                    *(dp2 + x + 2) = *(dp1 + (sx >> 16));                    sx += xfactor;                    *(dp2 + x + 3) = *(dp1 + (sx >> 16));                    sx += xfactor;                    *(dp2 + x + 4) = *(dp1 + (sx >> 16));                    sx += xfactor;                    *(dp2 + x + 5) = *(dp1 + (sx >> 16));                    sx += xfactor;                    *(dp2 + x + 6) = *(dp1 + (sx >> 16));                    sx += xfactor;                    *(dp2 + x + 7) = *(dp1 + (sx >> 16));                    sx += xfactor;#endif                    x += 8;                }                while (x < dst->bmWidth) {                    *(dp2 + x) = *(dp1 + (sx >> 16));                    sx += xfactor;                    x++;                }                dp2 += dst->bmPitch;                y++;                while (y < dst->bmHeight) {                    int l;                    int syint = sy >> 16;                    sy += yfactor;                    if ((sy >> 16) != syint)                        break;                    /* Copy identical lines. */                    l = dp2 - dp2old;                    memcpy(dp2, dp2old, l);                    dp2old = dp2;                    dp2 += l;                    y++;                }                dp1 = src->bmBits + (sy >> 16) * src->bmPitch;            }        }    break;    case 2:        {            int y, sy;            sy = 0;            for (y = 0; y < dst->bmHeight;) {                int sx = 0;                BYTE *dp2old = dp2;                int x;                x = 0;                /* This can be greatly optimized with loop */                /* unrolling; omitted to save space. */                while (x < dst->bmWidth) {                    *(unsigned short *) (dp2 + x * 2) =                        *(unsigned short *) (dp1 + (sx >> 16) * 2);                    sx += xfactor;                    x++;                }                dp2 += dst->bmPitch;                y++;                while (y < dst->bmHeight) {                    int l;                    int syint = sy >> 16;                    sy += yfactor;                    if ((sy >> 16) != syint)                        break;                    /* Copy identical lines. */                    l = dp2 - dp2old;                    memcpy(dp2, dp2old, l);                    dp2old = dp2;                    dp2 += l;                    y++;                }                dp1 = src->bmBits + (sy >> 16) * src->bmPitch;            }        }    break;    case 3:        {            int y, sy;            sy = 0;            for (y = 0; y < dst->bmHeight;) {                int sx = 0;                BYTE *dp2old = dp2;                int x;                x = 0;                /* This can be greatly optimized with loop */                /* unrolling; omitted to save space. */                while (x < dst->bmWidth) {                    *(unsigned short *) (dp2 + x * 3) =                        *(unsigned short *) (dp1 + (sx >> 16) * 3);                    *(unsigned char *) (dp2 + x * 3 + 2) =                        *(unsigned char *) (dp1 + (sx >> 16) * 3 + 2);                    sx += xfactor;                    x++;                }                dp2 += dst->bmPitch;                y++;                while (y < dst->bmHeight) {                    int l;                    int syint = sy >> 16;                    sy += yfactor;                    if ((sy >> 16) != syint)                        break;                    /* Copy identical lines. */                    l = dp2 - dp2old;                    memcpy(dp2, dp2old, l);                    dp2old = dp2;                    dp2 += l;                    y++;                }                dp1 = src->bmBits + (sy >> 16) * src->bmPitch;            }        }    break;    case 4:        {            int y, sy;            sy = 0;            for (y = 0; y < dst->bmHeight;) {                int sx = 0;                BYTE *dp2old = dp2;                int x;                x = 0;                /* This can be greatly optimized with loop */                /* unrolling; omitted to save space. */                while (x < dst->bmWidth) {                    *(unsigned *) (dp2 + x * 4) =                        *(unsigned *) (dp1 + (sx >> 16) * 4);                    sx += xfactor;                    x++;                }                dp2 += dst->bmPitch;                y++;                while (y < dst->bmHeight) {                    int l;                    int syint = sy >> 16;                    sy += yfactor;                    if ((sy >> 16) != syint)                        break;                    /* Copy identical lines. */                    l = dp2 - dp2old;                    memcpy(dp2, dp2old, l);                    dp2old = dp2;                    dp2 += l;                    y++;                }                dp1 = src->bmBits + (sy >> 16) * src->bmPitch;            }        }    break;    }    return TRUE;}gal_pixel GUIAPI GetPixelInBitmap (const BITMAP* bmp, int x, int y){    BYTE* dst;    if (x < 0 || y < 0 || x >= bmp->bmWidth || y >= bmp->bmHeight)        return 0;    dst = bmp->bmBits + y * bmp->bmPitch + x * bmp->bmBytesPerPixel;    return _mem_get_pixel (dst, bmp->bmBytesPerPixel);}BOOL GUIAPI SetPixelInBitmap (const BITMAP* bmp, int x, int y, gal_pixel pixel){    BYTE* dst;    if (x < 0 || y < 0 || x >= bmp->bmWidth || y >= bmp->bmHeight)        return FALSE;    dst = bmp->bmBits + y * bmp->bmPitch + x * bmp->bmBytesPerPixel;    _mem_set_pixel (dst, bmp->bmBytesPerPixel, pixel);    return TRUE;}