maskbits.h

远程桌面连接工具

#ifdef __GNUC__ /* XXX don't want for Alpha? */
#ifdef vax
#define FASTGETBITS(psrc,x,w,dst) \
    __asm ("extzv %1,%2,%3,%0" \
	 : "=g" (dst) \
	 : "g" (x), "g" (w), "m" (*(char *)(psrc)))
#define getbits(psrc,x,w,dst) FASTGETBITS(psrc,x,w,dst)

#define FASTPUTBITS(src, x, w, pdst) \
    __asm ("insv %3,%1,%2,%0" \
	 : "=m" (*(char *)(pdst)) \
	 : "g" (x), "g" (w), "g" (src))
#define putbits(src, x, w, pdst) FASTPUTBITS(src, x, w, pdst)
#endif /* vax */
#ifdef mc68020
#define FASTGETBITS(psrc, x, w, dst) \
    __asm ("bfextu %3{%1:%2},%0" \
    : "=d" (dst) : "di" (x), "di" (w), "o" (*(char *)(psrc)))

#define getbits(psrc,x,w,dst) \
{ \
    FASTGETBITS(psrc, x, w, dst);\
    dst = SHL(dst,(32-(w))); \
}

#define FASTPUTBITS(src, x, w, pdst) \
    __asm ("bfins %3,%0{%1:%2}" \
	 : "=o" (*(char *)(pdst)) \
	 : "di" (x), "di" (w), "d" (src), "0" (*(char *) (pdst)))

#define putbits(src, x, w, pdst) FASTPUTBITS(SHR((src),32-(w)), x, w, pdst)

#endif /* mc68020 */
#endif /* __GNUC__ */

/*  The following flag is used to override a bugfix for sun 3/60+CG4 machines,
 */

/*  We don't need to be careful about this unless we're dealing with sun3's 
 *  We will default its usage for those who do not know anything, but will
 *  override its effect if the machine doesn't look like a sun3 
 */
#if !defined(mc68020) || !defined(sun)
#define NO_3_60_CG4
#endif

/* This is gross.  We want to #define u_putbits as something which can be used
 * in the case of the 3/60+CG4, but if we use /bin/cc or are on another
 * machine type, we want nothing to do with u_putbits.  What a hastle.  Here
 * I used slo_putbits as something which either u_putbits or putbits could be
 * defined as.
 *
 * putbits gets it iff it is not already defined with FASTPUTBITS above.
 * u_putbits gets it if we have FASTPUTBITS (putbits) from above and have not
 * 	overridden the NO_3_60_CG4 flag.
 */

#define slo_putbits(src, x, w, pdst) \
{ \
    register int n = (x)+(w)-PPW; \
    \
    if (n <= 0) \
    { \
	register PixelType tmpmask; \
	maskpartialbits((x), (w), tmpmask); \
	*(pdst) = (*(pdst) & ~tmpmask) | \
		(SCRRIGHT(src, x) & tmpmask); \
    } \
    else \
    { \
	*(pdst) = (*(pdst) & endtab[x]) | (SCRRIGHT((src), x)); \
	(pdst)[1] = ((pdst)[1] & starttab[n]) | \
		(SCRLEFT(src, PPW-(x)) & endtab[n]); \
    } \
}

#if defined(putbits) && !defined(NO_3_60_CG4)
#define u_putbits(src, x, w, pdst) slo_putbits(src, x, w, pdst)
#else
#define u_putbits(src, x, w, pdst) putbits(src, x, w, pdst)
#endif

#if !defined(putbits) 
#define putbits(src, x, w, pdst) slo_putbits(src, x, w, pdst)
#endif

/* Now if we have not gotten any really good bitfield macros, try some
 * moderately fast macros.  Alas, I don't know how to do asm instructions
 * without gcc.
 */

#ifndef getbits
#define getbits(psrc, x, w, dst) \
{ \
    dst = SCRLEFT(*(psrc), (x)); \
    if ( ((x) + (w)) > PPW) \
	dst |= (SCRRIGHT(*((psrc)+1), PPW-(x))); \
}
#endif

/*  We have to special-case putbitsrop because of 3/60+CG4 combos
 */

#define u_putbitsrop(src, x, w, pdst, rop) \
{\
	register PixelType t1, t2; \
	register int n = (x)+(w)-PPW; \
	\
	t1 = SCRRIGHT((src), (x)); \
	DoRop(t2, rop, t1, *(pdst)); \
	\
    if (n <= 0) \
    { \
	register PixelType tmpmask; \
	\
	maskpartialbits((x), (w), tmpmask); \
	*(pdst) = (*(pdst) & ~tmpmask) | (t2 & tmpmask); \
    } \
    else \
    { \
	int m = PPW-(x); \
	*(pdst) = (*(pdst) & endtab[x]) | (t2 & starttab[x]); \
	t1 = SCRLEFT((src), m); \
	DoRop(t2, rop, t1, (pdst)[1]); \
	(pdst)[1] = ((pdst)[1] & starttab[n]) | (t2 & endtab[n]); \
    } \
}

/* If our getbits and putbits are FAST enough,
 * do this brute force, it's faster
 */

#if defined(FASTPUTBITS) && defined(FASTGETBITS) && defined(NO_3_60_CG4)
#if (BITMAP_BIT_ORDER == MSBFirst)
#define putbitsrop(src, x, w, pdst, rop) \
{ \
  register PixelType _tmp, _tmp2; \
  FASTGETBITS(pdst, x, w, _tmp); \
  _tmp2 = SCRRIGHT(src, PPW-(w)); \
  DoRop(_tmp, rop, _tmp2, _tmp) \
  FASTPUTBITS(_tmp, x, w, pdst); \
}
#define putbitsrrop(src, x, w, pdst, rop) \
{ \
  register PixelType _tmp, _tmp2; \
 \
  FASTGETBITS(pdst, x, w, _tmp); \
  _tmp2 = SCRRIGHT(src, PPW-(w)); \
  _tmp= DoRRop(rop, _tmp2, _tmp); \
  FASTPUTBITS(_tmp, x, w, pdst); \
}
#undef u_putbitsrop
#else
#define putbitsrop(src, x, w, pdst, rop) \
{ \
  register PixelType _tmp; \
  FASTGETBITS(pdst, x, w, _tmp); \
  DoRop(_tmp, rop, src, _tmp) \
  FASTPUTBITS(_tmp, x, w, pdst); \
}
#define putbitsrrop(src, x, w, pdst, rop) \
{ \
  register PixelType _tmp; \
 \
  FASTGETBITS(pdst, x, w, _tmp); \
  _tmp= DoRRop(rop, src, _tmp); \
  FASTPUTBITS(_tmp, x, w, pdst); \
}
#undef u_putbitsrop
#endif
#endif

#ifndef putbitsrop
#define putbitsrop(src, x, w, pdst, rop)  u_putbitsrop(src, x, w, pdst, rop)
#endif 

#ifndef putbitsrrop
#define putbitsrrop(src, x, w, pdst, rop) \
{\
	register PixelType t1, t2; \
	register int n = (x)+(w)-PPW; \
	\
	t1 = SCRRIGHT((src), (x)); \
	t2 = DoRRop(rop, t1, *(pdst)); \
	\
    if (n <= 0) \
    { \
	register PixelType tmpmask; \
	\
	maskpartialbits((x), (w), tmpmask); \
	*(pdst) = (*(pdst) & ~tmpmask) | (t2 & tmpmask); \
    } \
    else \
    { \
	int m = PPW-(x); \
	*(pdst) = (*(pdst) & endtab[x]) | (t2 & starttab[x]); \
	t1 = SCRLEFT((src), m); \
	t2 = DoRRop(rop, t1, (pdst)[1]); \
	(pdst)[1] = ((pdst)[1] & starttab[n]) | (t2 & endtab[n]); \
    } \
}
#endif

#if GETLEFTBITS_ALIGNMENT == 1
#define getleftbits(psrc, w, dst)	dst = *((CARD32 *) psrc)
#endif /* GETLEFTBITS_ALIGNMENT == 1 */

#if GETLEFTBITS_ALIGNMENT == 2
#define getleftbits(psrc, w, dst) \
    { \
	if ( ((int)(psrc)) & 0x01 ) \
		getbits( ((CARD32 *)(((char *)(psrc))-1)), 8, (w), (dst) ); \
	else \
		getbits(psrc, 0, w, dst); \
    }
#endif /* GETLEFTBITS_ALIGNMENT == 2 */

#if GETLEFTBITS_ALIGNMENT == 4
#define getleftbits(psrc, w, dst) \
    { \
	int off, off_b; \
	off_b = (off = ( ((int)(psrc)) & 0x03)) << 3; \
	getbits( \
		(CARD32 *)( ((char *)(psrc)) - off), \
		(off_b), (w), (dst) \
	       ); \
    }
#endif /* GETLEFTBITS_ALIGNMENT == 4 */


#define getshiftedleftbits(psrc, offset, w, dst) \
	getleftbits((psrc), (w), (dst)); \
	dst = SCRLEFT((dst), (offset));

/* FASTGETBITS and FASTPUTBITS are not necessarily correct implementations of
 * getbits and putbits, but they work if used together.
 *
 * On a MSBFirst machine, a cpu bitfield extract instruction (like bfextu)
 * could normally assign its result to a long word register in the screen
 * right position.  This saves canceling register shifts by not fighting the
 * natural cpu byte order.
 *
 * Unfortunately, these fail on a 3/60+CG4 and cannot be used unmodified. Sigh.
 */
#if defined(FASTGETBITS) && defined(FASTPUTBITS)
#ifdef NO_3_60_CG4
#define u_FASTPUT(aa, bb, cc, dd)  FASTPUTBITS(aa, bb, cc, dd)
#else
#define u_FASTPUT(aa, bb, cc, dd)  u_putbits(SCRLEFT(aa, PPW-(cc)), bb, cc, dd)
#endif

#define getandputbits(psrc, srcbit, dstbit, width, pdst) \
{ \
    register PixelType _tmpbits; \
    FASTGETBITS(psrc, srcbit, width, _tmpbits); \
    u_FASTPUT(_tmpbits, dstbit, width, pdst); \
}

#define getandputrop(psrc, srcbit, dstbit, width, pdst, rop) \
{ \
  register PixelType _tmpsrc, _tmpdst; \
  FASTGETBITS(pdst, dstbit, width, _tmpdst); \
  FASTGETBITS(psrc, srcbit, width, _tmpsrc); \
  DoRop(_tmpdst, rop, _tmpsrc, _tmpdst); \
  u_FASTPUT(_tmpdst, dstbit, width, pdst); \
}

#define getandputrrop(psrc, srcbit, dstbit, width, pdst, rop) \
{ \
  register PixelType _tmpsrc, _tmpdst; \
  FASTGETBITS(pdst, dstbit, width, _tmpdst); \
  FASTGETBITS(psrc, srcbit, width, _tmpsrc); \
  _tmpdst = DoRRop(rop, _tmpsrc, _tmpdst); \
  u_FASTPUT(_tmpdst, dstbit, width, pdst); \
}

#define getandputbits0(psrc, srcbit, width, pdst) \
	getandputbits(psrc, srcbit, 0, width, pdst)

#define getandputrop0(psrc, srcbit, width, pdst, rop) \
    	getandputrop(psrc, srcbit, 0, width, pdst, rop)

#define getandputrrop0(psrc, srcbit, width, pdst, rop) \
    	getandputrrop(psrc, srcbit, 0, width, pdst, rop)


#else /* Slow poke */

/* pairs of getbits/putbits happen frequently. Some of the code can
 * be shared or avoided in a few specific instances.  It gets us a
 * small advantage, so we do it.  The getandput...0 macros are the only ones
 * which speed things here.  The others are here for compatibility w/the above
 * FAST ones
 */

#define getandputbits(psrc, srcbit, dstbit, width, pdst) \
{ \
    register PixelType _tmpbits; \
    getbits(psrc, srcbit, width, _tmpbits); \
    putbits(_tmpbits, dstbit, width, pdst); \
}

#define getandputrop(psrc, srcbit, dstbit, width, pdst, rop) \
{ \
    register PixelType _tmpbits; \
    getbits(psrc, srcbit, width, _tmpbits) \
    putbitsrop(_tmpbits, dstbit, width, pdst, rop) \
}

#define getandputrrop(psrc, srcbit, dstbit, width, pdst, rop) \
{ \
    register PixelType _tmpbits; \
    getbits(psrc, srcbit, width, _tmpbits) \
    putbitsrrop(_tmpbits, dstbit, width, pdst, rop) \
}


#define getandputbits0(psrc, sbindex, width, pdst) \
{			/* unroll the whole damn thing to see how it * behaves */ \
    register int          _flag = PPW - (sbindex); \
    register PixelType _src; \
 \
    _src = SCRLEFT (*(psrc), (sbindex)); \
    if ((width) > _flag) \
	_src |=  SCRRIGHT (*((psrc) + 1), _flag); \
 \
    *(pdst) = (*(pdst) & starttab[(width)]) | (_src & endtab[(width)]); \
}


#define getandputrop0(psrc, sbindex, width, pdst, rop) \
{			\
    register int          _flag = PPW - (sbindex); \
    register PixelType _src; \
 \
    _src = SCRLEFT (*(psrc), (sbindex)); \
    if ((width) > _flag) \
	_src |=  SCRRIGHT (*((psrc) + 1), _flag); \
    DoRop(_src, rop, _src, *(pdst)); \
 \
    *(pdst) = (*(pdst) & starttab[(width)]) | (_src & endtab[(width)]); \
}

#define getandputrrop0(psrc, sbindex, width, pdst, rop) \
{ \
    int             _flag = PPW - (sbindex); \
    register PixelType _src; \
 \
    _src = SCRLEFT (*(psrc), (sbindex)); \
    if ((width) > _flag) \
	_src |=  SCRRIGHT (*((psrc) + 1), _flag); \
    _src = DoRRop(rop, _src, *(pdst)); \
 \
    *(pdst) = (*(pdst) & starttab[(width)]) | (_src & endtab[(width)]); \
}

#endif  /* FASTGETBITS && FASTPUTBITS */

#endif /* MFB_CONSTS_ONLY */