xref: /freebsd/sys/powerpc/fpu/fpu_implode.c (revision 4b9d6057)
1 /*	$NetBSD: fpu_implode.c,v 1.6 2005/12/11 12:18:42 christos Exp $ */
2 
3 /*-
4  * SPDX-License-Identifier: BSD-3-Clause
5  *
6  * Copyright (c) 1992, 1993
7  *	The Regents of the University of California.  All rights reserved.
8  *
9  * This software was developed by the Computer Systems Engineering group
10  * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
11  * contributed to Berkeley.
12  *
13  * All advertising materials mentioning features or use of this software
14  * must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Lawrence Berkeley Laboratory.
17  *
18  * Redistribution and use in source and binary forms, with or without
19  * modification, are permitted provided that the following conditions
20  * are met:
21  * 1. Redistributions of source code must retain the above copyright
22  *    notice, this list of conditions and the following disclaimer.
23  * 2. Redistributions in binary form must reproduce the above copyright
24  *    notice, this list of conditions and the following disclaimer in the
25  *    documentation and/or other materials provided with the distribution.
26  * 3. Neither the name of the University nor the names of its contributors
27  *    may be used to endorse or promote products derived from this software
28  *    without specific prior written permission.
29  *
30  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
31  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
32  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
33  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
34  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
35  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
36  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
37  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
38  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
39  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
40  * SUCH DAMAGE.
41  */
42 
43 /*
44  * FPU subroutines: `implode' internal format numbers into the machine's
45  * `packed binary' format.
46  */
47 
48 #include <sys/types.h>
49 #include <sys/systm.h>
50 
51 #include <machine/fpu.h>
52 #include <machine/ieee.h>
53 #include <machine/ieeefp.h>
54 
55 #include <powerpc/fpu/fpu_arith.h>
56 #include <powerpc/fpu/fpu_emu.h>
57 #include <powerpc/fpu/fpu_extern.h>
58 #include <powerpc/fpu/fpu_instr.h>
59 
60 static int round(struct fpemu *, struct fpn *);
61 static int toinf(struct fpemu *, int);
62 
63 /*
64  * Round a number (algorithm from Motorola MC68882 manual, modified for
65  * our internal format).  Set inexact exception if rounding is required.
66  * Return true iff we rounded up.
67  *
68  * After rounding, we discard the guard and round bits by shifting right
69  * 2 bits (a la fpu_shr(), but we do not bother with fp->fp_sticky).
70  * This saves effort later.
71  *
72  * Note that we may leave the value 2.0 in fp->fp_mant; it is the caller's
73  * responsibility to fix this if necessary.
74  */
75 static int
76 round(struct fpemu *fe, struct fpn *fp)
77 {
78 	u_int m0, m1, m2, m3;
79 	int gr, s;
80 	FPU_DECL_CARRY;
81 
82 	m0 = fp->fp_mant[0];
83 	m1 = fp->fp_mant[1];
84 	m2 = fp->fp_mant[2];
85 	m3 = fp->fp_mant[3];
86 	gr = m3 & 3;
87 	s = fp->fp_sticky;
88 
89 	/* mant >>= FP_NG */
90 	m3 = (m3 >> FP_NG) | (m2 << (32 - FP_NG));
91 	m2 = (m2 >> FP_NG) | (m1 << (32 - FP_NG));
92 	m1 = (m1 >> FP_NG) | (m0 << (32 - FP_NG));
93 	m0 >>= FP_NG;
94 
95 	if ((gr | s) == 0)	/* result is exact: no rounding needed */
96 		goto rounddown;
97 
98 	fe->fe_cx |= FPSCR_XX|FPSCR_FI;	/* inexact */
99 
100 	/* Go to rounddown to round down; break to round up. */
101 	switch ((fe->fe_fpscr) & FPSCR_RN) {
102 	case FP_RN:
103 	default:
104 		/*
105 		 * Round only if guard is set (gr & 2).  If guard is set,
106 		 * but round & sticky both clear, then we want to round
107 		 * but have a tie, so round to even, i.e., add 1 iff odd.
108 		 */
109 		if ((gr & 2) == 0)
110 			goto rounddown;
111 		if ((gr & 1) || fp->fp_sticky || (m3 & 1))
112 			break;
113 		goto rounddown;
114 
115 	case FP_RZ:
116 		/* Round towards zero, i.e., down. */
117 		goto rounddown;
118 
119 	case FP_RM:
120 		/* Round towards -Inf: up if negative, down if positive. */
121 		if (fp->fp_sign)
122 			break;
123 		goto rounddown;
124 
125 	case FP_RP:
126 		/* Round towards +Inf: up if positive, down otherwise. */
127 		if (!fp->fp_sign)
128 			break;
129 		goto rounddown;
130 	}
131 
132 	/* Bump low bit of mantissa, with carry. */
133 	fe->fe_cx |= FPSCR_FR;
134 
135 	FPU_ADDS(m3, m3, 1);
136 	FPU_ADDCS(m2, m2, 0);
137 	FPU_ADDCS(m1, m1, 0);
138 	FPU_ADDC(m0, m0, 0);
139 	fp->fp_mant[0] = m0;
140 	fp->fp_mant[1] = m1;
141 	fp->fp_mant[2] = m2;
142 	fp->fp_mant[3] = m3;
143 	return (1);
144 
145 rounddown:
146 	fp->fp_mant[0] = m0;
147 	fp->fp_mant[1] = m1;
148 	fp->fp_mant[2] = m2;
149 	fp->fp_mant[3] = m3;
150 	return (0);
151 }
152 
153 /*
154  * For overflow: return true if overflow is to go to +/-Inf, according
155  * to the sign of the overflowing result.  If false, overflow is to go
156  * to the largest magnitude value instead.
157  */
158 static int
159 toinf(struct fpemu *fe, int sign)
160 {
161 	int inf;
162 
163 	/* look at rounding direction */
164 	switch ((fe->fe_fpscr) & FPSCR_RN) {
165 	default:
166 	case FP_RN:		/* the nearest value is always Inf */
167 		inf = 1;
168 		break;
169 
170 	case FP_RZ:		/* toward 0 => never towards Inf */
171 		inf = 0;
172 		break;
173 
174 	case FP_RP:		/* toward +Inf iff positive */
175 		inf = sign == 0;
176 		break;
177 
178 	case FP_RM:		/* toward -Inf iff negative */
179 		inf = sign;
180 		break;
181 	}
182 	if (inf)
183 		fe->fe_cx |= FPSCR_OX;
184 	return (inf);
185 }
186 
187 /*
188  * fpn -> int (int value returned as return value).
189  *
190  * N.B.: this conversion always rounds towards zero (this is a peculiarity
191  * of the SPARC instruction set).
192  */
193 u_int
194 fpu_ftoi(struct fpemu *fe, struct fpn *fp)
195 {
196 	u_int i;
197 	int sign, exp;
198 
199 	sign = fp->fp_sign;
200 	switch (fp->fp_class) {
201 	case FPC_ZERO:
202 		return (0);
203 
204 	case FPC_NUM:
205 		/*
206 		 * If exp >= 2^32, overflow.  Otherwise shift value right
207 		 * into last mantissa word (this will not exceed 0xffffffff),
208 		 * shifting any guard and round bits out into the sticky
209 		 * bit.  Then ``round'' towards zero, i.e., just set an
210 		 * inexact exception if sticky is set (see round()).
211 		 * If the result is > 0x80000000, or is positive and equals
212 		 * 0x80000000, overflow; otherwise the last fraction word
213 		 * is the result.
214 		 */
215 		if ((exp = fp->fp_exp) >= 32)
216 			break;
217 		/* NB: the following includes exp < 0 cases */
218 		if (fpu_shr(fp, FP_NMANT - 1 - exp) != 0)
219 			fe->fe_cx |= FPSCR_UX;
220 		i = fp->fp_mant[3];
221 		if (i >= ((u_int)0x80000000 + sign))
222 			break;
223 		return (sign ? -i : i);
224 
225 	default:		/* Inf, qNaN, sNaN */
226 		break;
227 	}
228 	/* overflow: replace any inexact exception with invalid */
229 	fe->fe_cx |= FPSCR_VXCVI;
230 	return (0x7fffffff + sign);
231 }
232 
233 /*
234  * fpn -> extended int (high bits of int value returned as return value).
235  *
236  * N.B.: this conversion always rounds towards zero (this is a peculiarity
237  * of the SPARC instruction set).
238  */
239 u_int
240 fpu_ftox(struct fpemu *fe, struct fpn *fp, u_int *res)
241 {
242 	u_int64_t i;
243 	int sign, exp;
244 
245 	sign = fp->fp_sign;
246 	switch (fp->fp_class) {
247 	case FPC_ZERO:
248 		res[1] = 0;
249 		return (0);
250 
251 	case FPC_NUM:
252 		/*
253 		 * If exp >= 2^64, overflow.  Otherwise shift value right
254 		 * into last mantissa word (this will not exceed 0xffffffffffffffff),
255 		 * shifting any guard and round bits out into the sticky
256 		 * bit.  Then ``round'' towards zero, i.e., just set an
257 		 * inexact exception if sticky is set (see round()).
258 		 * If the result is > 0x8000000000000000, or is positive and equals
259 		 * 0x8000000000000000, overflow; otherwise the last fraction word
260 		 * is the result.
261 		 */
262 		if ((exp = fp->fp_exp) >= 64)
263 			break;
264 		/* NB: the following includes exp < 0 cases */
265 		if (fpu_shr(fp, FP_NMANT - 1 - exp) != 0)
266 			fe->fe_cx |= FPSCR_UX;
267 		i = ((u_int64_t)fp->fp_mant[2]<<32)|fp->fp_mant[3];
268 		if (i >= ((u_int64_t)0x8000000000000000LL + sign))
269 			break;
270 		return (sign ? -i : i);
271 
272 	default:		/* Inf, qNaN, sNaN */
273 		break;
274 	}
275 	/* overflow: replace any inexact exception with invalid */
276 	fe->fe_cx |= FPSCR_VXCVI;
277 	return (0x7fffffffffffffffLL + sign);
278 }
279 
280 /*
281  * fpn -> single (32 bit single returned as return value).
282  * We assume <= 29 bits in a single-precision fraction (1.f part).
283  */
284 u_int
285 fpu_ftos(struct fpemu *fe, struct fpn *fp)
286 {
287 	u_int sign = fp->fp_sign << 31;
288 	int exp;
289 
290 #define	SNG_EXP(e)	((e) << SNG_FRACBITS)	/* makes e an exponent */
291 #define	SNG_MASK	(SNG_EXP(1) - 1)	/* mask for fraction */
292 
293 	/* Take care of non-numbers first. */
294 	if (ISNAN(fp)) {
295 		/*
296 		 * Preserve upper bits of NaN, per SPARC V8 appendix N.
297 		 * Note that fp->fp_mant[0] has the quiet bit set,
298 		 * even if it is classified as a signalling NaN.
299 		 */
300 		(void) fpu_shr(fp, FP_NMANT - 1 - SNG_FRACBITS);
301 		exp = SNG_EXP_INFNAN;
302 		goto done;
303 	}
304 	if (ISINF(fp))
305 		return (sign | SNG_EXP(SNG_EXP_INFNAN));
306 	if (ISZERO(fp))
307 		return (sign);
308 
309 	/*
310 	 * Normals (including subnormals).  Drop all the fraction bits
311 	 * (including the explicit ``implied'' 1 bit) down into the
312 	 * single-precision range.  If the number is subnormal, move
313 	 * the ``implied'' 1 into the explicit range as well, and shift
314 	 * right to introduce leading zeroes.  Rounding then acts
315 	 * differently for normals and subnormals: the largest subnormal
316 	 * may round to the smallest normal (1.0 x 2^minexp), or may
317 	 * remain subnormal.  In the latter case, signal an underflow
318 	 * if the result was inexact or if underflow traps are enabled.
319 	 *
320 	 * Rounding a normal, on the other hand, always produces another
321 	 * normal (although either way the result might be too big for
322 	 * single precision, and cause an overflow).  If rounding a
323 	 * normal produces 2.0 in the fraction, we need not adjust that
324 	 * fraction at all, since both 1.0 and 2.0 are zero under the
325 	 * fraction mask.
326 	 *
327 	 * Note that the guard and round bits vanish from the number after
328 	 * rounding.
329 	 */
330 	if ((exp = fp->fp_exp + SNG_EXP_BIAS) <= 0) {	/* subnormal */
331 		/* -NG for g,r; -SNG_FRACBITS-exp for fraction */
332 		(void) fpu_shr(fp, FP_NMANT - FP_NG - SNG_FRACBITS - exp);
333 		if (round(fe, fp) && fp->fp_mant[3] == SNG_EXP(1))
334 			return (sign | SNG_EXP(1) | 0);
335 		if ((fe->fe_cx & FPSCR_FI) ||
336 		    (fe->fe_fpscr & FPSCR_UX))
337 			fe->fe_cx |= FPSCR_UX;
338 		return (sign | SNG_EXP(0) | fp->fp_mant[3]);
339 	}
340 	/* -FP_NG for g,r; -1 for implied 1; -SNG_FRACBITS for fraction */
341 	(void) fpu_shr(fp, FP_NMANT - FP_NG - 1 - SNG_FRACBITS);
342 #ifdef DIAGNOSTIC
343 	if ((fp->fp_mant[3] & SNG_EXP(1 << FP_NG)) == 0)
344 		panic("fpu_ftos");
345 #endif
346 	if (round(fe, fp) && fp->fp_mant[3] == SNG_EXP(2))
347 		exp++;
348 	if (exp >= SNG_EXP_INFNAN) {
349 		/* overflow to inf or to max single */
350 		if (toinf(fe, sign))
351 			return (sign | SNG_EXP(SNG_EXP_INFNAN));
352 		return (sign | SNG_EXP(SNG_EXP_INFNAN - 1) | SNG_MASK);
353 	}
354 done:
355 	/* phew, made it */
356 	return (sign | SNG_EXP(exp) | (fp->fp_mant[3] & SNG_MASK));
357 }
358 
359 /*
360  * fpn -> double (32 bit high-order result returned; 32-bit low order result
361  * left in res[1]).  Assumes <= 61 bits in double precision fraction.
362  *
363  * This code mimics fpu_ftos; see it for comments.
364  */
365 u_int
366 fpu_ftod(struct fpemu *fe, struct fpn *fp, u_int *res)
367 {
368 	u_int sign = fp->fp_sign << 31;
369 	int exp;
370 
371 #define	DBL_EXP(e)	((e) << (DBL_FRACBITS & 31))
372 #define	DBL_MASK	(DBL_EXP(1) - 1)
373 
374 	if (ISNAN(fp)) {
375 		(void) fpu_shr(fp, FP_NMANT - 1 - DBL_FRACBITS);
376 		exp = DBL_EXP_INFNAN;
377 		goto done;
378 	}
379 	if (ISINF(fp)) {
380 		sign |= DBL_EXP(DBL_EXP_INFNAN);
381 		goto zero;
382 	}
383 	if (ISZERO(fp)) {
384 zero:		res[1] = 0;
385 		return (sign);
386 	}
387 
388 	if ((exp = fp->fp_exp + DBL_EXP_BIAS) <= 0) {
389 		(void) fpu_shr(fp, FP_NMANT - FP_NG - DBL_FRACBITS - exp);
390 		if (round(fe, fp) && fp->fp_mant[2] == DBL_EXP(1)) {
391 			res[1] = 0;
392 			return (sign | DBL_EXP(1) | 0);
393 		}
394 		if ((fe->fe_cx & FPSCR_FI) ||
395 		    (fe->fe_fpscr & FPSCR_UX))
396 			fe->fe_cx |= FPSCR_UX;
397 		exp = 0;
398 		goto done;
399 	}
400 	(void) fpu_shr(fp, FP_NMANT - FP_NG - 1 - DBL_FRACBITS);
401 	if (round(fe, fp) && fp->fp_mant[2] == DBL_EXP(2))
402 		exp++;
403 	if (exp >= DBL_EXP_INFNAN) {
404 		fe->fe_cx |= FPSCR_OX | FPSCR_UX;
405 		if (toinf(fe, sign)) {
406 			res[1] = 0;
407 			return (sign | DBL_EXP(DBL_EXP_INFNAN) | 0);
408 		}
409 		res[1] = ~0;
410 		return (sign | DBL_EXP(DBL_EXP_INFNAN) | DBL_MASK);
411 	}
412 done:
413 	res[1] = fp->fp_mant[3];
414 	return (sign | DBL_EXP(exp) | (fp->fp_mant[2] & DBL_MASK));
415 }
416 
417 /*
418  * Implode an fpn, writing the result into the given space.
419  */
420 void
421 fpu_implode(struct fpemu *fe, struct fpn *fp, int type, u_int *space)
422 {
423 
424 	switch (type) {
425 	case FTYPE_LNG:
426 		space[0] = fpu_ftox(fe, fp, space);
427 		DPRINTF(FPE_REG, ("fpu_implode: long %x %x\n",
428 			space[0], space[1]));
429 		break;
430 
431 	case FTYPE_INT:
432 		space[0] = 0;
433 		space[1] = fpu_ftoi(fe, fp);
434 		DPRINTF(FPE_REG, ("fpu_implode: int %x\n",
435 			space[1]));
436 		break;
437 
438 	case FTYPE_SNG:
439 		space[0] = fpu_ftos(fe, fp);
440 		DPRINTF(FPE_REG, ("fpu_implode: single %x\n",
441 			space[0]));
442 		break;
443 
444 	case FTYPE_DBL:
445 		space[0] = fpu_ftod(fe, fp, space);
446 		DPRINTF(FPE_REG, ("fpu_implode: double %x %x\n",
447 			space[0], space[1]));
448 		break;		break;
449 
450 	default:
451 		panic("fpu_implode: invalid type %d", type);
452 	}
453 }
454