/*  (c) copyright 1988 by the Vrije Universiteit, Amsterdam, The Netherlands.  See the copyright notice in the ACK home directory, in the file "Copyright".*//* $Header: /cvsup/minix/src/lib/ack/float/compact.fc,v 1.1 2005/10/10 15:27:42 beng Exp $ *//*	COMPACT EXTEND FORMAT INTO FLOAT OF PROPER SIZE*/# include "FP_bias.h"# include "FP_shift.h"# include "FP_trap.h"# include "FP_types.h"# include "get_put.h"voidcompact(f,to,size)EXTEND	*f;unsigned long	*to;int	size;{	int	error = 0;	if (size == sizeof(DOUBLE)) {	/*	 * COMPACT EXTENDED INTO DOUBLE	 */		DOUBLE *DBL = (DOUBLE *) (void *) to;		if ((f->m1|(f->m2 & DBL_ZERO)) == 0L)	{			zrf8(DBL);			return;		}		f->exp += DBL_BIAS;	/* restore proper bias	*/		if (f->exp > DBL_MAX)	{dbl_over:			trap(EFOVFL);			f->exp = DBL_MAX+1;			f->m1 = 0;			f->m2 = 0;			if (error++)				return;		}		else if (f->exp < DBL_MIN)	{			b64_rsft(&(f->mantissa));			if (f->exp < 0) {				b64_sft(&(f->mantissa), -f->exp);				f->exp = 0;			}			/* underflow ??? */		}					/* local CAST conversion		*/		/* because of special format shift only 10 bits */		/* bit shift mantissa 10 bits		*/		/* first align within words, then do store operation */		DBL->d[0] = f->m1 >> DBL_RUNPACK;   /* plus 22 == 32 */		DBL->d[1] = f->m2 >> DBL_RUNPACK;   /* plus 22 == 32 */		DBL->d[1] |= (f->m1 << DBL_LUNPACK); /* plus 10 == 32 */		/* if not exact then round to nearest	*/		/* on a tie, round to even */#ifdef EXCEPTION_INEXACT		if ((f->m2 & DBL_EXACT) != 0) {		    INEXACT();#endif		    if (((f->m2 & DBL_EXACT) > DBL_ROUNDUP)			|| ((f->m2 & DBL_EXACT) == DBL_ROUNDUP			    && (f->m2 & (DBL_ROUNDUP << 1)))) {			DBL->d[1]++;	/* rounding up	*/			if (DBL->d[1] == 0L) { /* carry out	*/			    DBL->d[0]++;			    if (f->exp == 0 && (DBL->d[0] & ~DBL_MASK)) {					f->exp++;				}			    if (DBL->d[0] & DBL_CARRYOUT) { /* carry out */				if (DBL->d[0] & 01)				    DBL->d[1] = CARRYBIT;				DBL->d[0] >>= 1;				f->exp++;			    }			}			/*	check for overflow			*/			if (f->exp > DBL_MAX)		    		goto dbl_over;		    }#ifdef EXCEPTION_INEXACT		}#endif		/*		 * STORE EXPONENT AND SIGN:		 *		 * 1) clear leading bits (B4-B15)		 * 2) shift and store exponent		 */		DBL->d[0] &= DBL_MASK;		DBL->d[0] |= 			((long) (f->exp << DBL_EXPSHIFT) << EXP_STORE);		if (f->sign)			DBL->d[0] |= CARRYBIT;		/*		 * STORE MANTISSA		 */#if FL_MSL_AT_LOW_ADDRESS		put4(DBL->d[0], (char *) &DBL->d[0]);		put4(DBL->d[1], (char *) &DBL->d[1]);#else		{ unsigned long l;		  put4(DBL->d[1], (char *) &l);		  put4(DBL->d[0], (char *) &DBL->d[1]);		  DBL->d[0] = l;		}#endif	}	else {		/*		 * COMPACT EXTENDED INTO FLOAT		 */		SINGLE	*SGL;		/* local CAST conversion		*/		SGL = (SINGLE *) (void *) to;		if ((f->m1 & SGL_ZERO) == 0L)	{			*SGL = 0L;			return;		}		f->exp += SGL_BIAS;	/* restore bias	*/		if (f->exp > SGL_MAX)	{sgl_over:			trap(EFOVFL);			f->exp = SGL_MAX+1;			f->m1 = 0L;			f->m2 = 0L;			if (error++)				return;		}		else if (f->exp < SGL_MIN)	{			b64_rsft(&(f->mantissa));			if (f->exp < 0) {				b64_sft(&(f->mantissa), -f->exp);				f->exp = 0;			}			/* underflow ??? */		}		/* shift mantissa and store	*/		*SGL = (f->m1 >> SGL_RUNPACK);		/* check for rounding to nearest	*/		/* on a tie, round to even		*/#ifdef EXCEPTION_INEXACT		if (f->m2 != 0 ||		    (f->m1 & SGL_EXACT) != 0L) {			INEXACT();#endif		        if (((f->m1 & SGL_EXACT) > SGL_ROUNDUP)			    || ((f->m1 & SGL_EXACT) == SGL_ROUNDUP			        && (f->m1 & (SGL_ROUNDUP << 1)))) {				(*SGL)++;				if (f->exp == 0 && (*SGL & ~SGL_MASK)) {					f->exp++;				}			/* check normal */				if (*SGL & SGL_CARRYOUT)	{					*SGL >>= 1;					f->exp++;				}				if (f->exp > SGL_MAX)					goto sgl_over;			}#ifdef EXCEPTION_INEXACT		}#endif		/*		 * STORE EXPONENT AND SIGN:		 *		 * 1) clear leading bit of fraction		 * 2) shift and store exponent		 */		*SGL &= SGL_MASK; /* B23-B31 are 0 */		*SGL |= ((long) (f->exp << SGL_EXPSHIFT) << EXP_STORE);		if (f->sign)			*SGL |= CARRYBIT;		/*		 * STORE MANTISSA		 */		put4(*SGL, (char *) &SGL);	}}