aboutsummaryrefslogtreecommitdiffstats
path: root/arch/parisc/math-emu/dfmpy.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 18:20:36 -0400
committerLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 18:20:36 -0400
commit1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch)
tree0bba044c4ce775e45a88a51686b5d9f90697ea9d /arch/parisc/math-emu/dfmpy.c
Linux-2.6.12-rc2v2.6.12-rc2
Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
Diffstat (limited to 'arch/parisc/math-emu/dfmpy.c')
-rw-r--r--arch/parisc/math-emu/dfmpy.c394
1 files changed, 394 insertions, 0 deletions
diff --git a/arch/parisc/math-emu/dfmpy.c b/arch/parisc/math-emu/dfmpy.c
new file mode 100644
index 000000000000..4380f5a62ad1
--- /dev/null
+++ b/arch/parisc/math-emu/dfmpy.c
@@ -0,0 +1,394 @@
1/*
2 * Linux/PA-RISC Project (http://www.parisc-linux.org/)
3 *
4 * Floating-point emulation code
5 * Copyright (C) 2001 Hewlett-Packard (Paul Bame) <bame@debian.org>
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2, or (at your option)
10 * any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 */
21/*
22 * BEGIN_DESC
23 *
24 * File:
25 * @(#) pa/spmath/dfmpy.c $Revision: 1.1 $
26 *
27 * Purpose:
28 * Double Precision Floating-point Multiply
29 *
30 * External Interfaces:
31 * dbl_fmpy(srcptr1,srcptr2,dstptr,status)
32 *
33 * Internal Interfaces:
34 *
35 * Theory:
36 * <<please update with a overview of the operation of this file>>
37 *
38 * END_DESC
39*/
40
41
42#include "float.h"
43#include "dbl_float.h"
44
45/*
46 * Double Precision Floating-point Multiply
47 */
48
49int
50dbl_fmpy(
51 dbl_floating_point *srcptr1,
52 dbl_floating_point *srcptr2,
53 dbl_floating_point *dstptr,
54 unsigned int *status)
55{
56 register unsigned int opnd1p1, opnd1p2, opnd2p1, opnd2p2;
57 register unsigned int opnd3p1, opnd3p2, resultp1, resultp2;
58 register int dest_exponent, count;
59 register boolean inexact = FALSE, guardbit = FALSE, stickybit = FALSE;
60 boolean is_tiny;
61
62 Dbl_copyfromptr(srcptr1,opnd1p1,opnd1p2);
63 Dbl_copyfromptr(srcptr2,opnd2p1,opnd2p2);
64
65 /*
66 * set sign bit of result
67 */
68 if (Dbl_sign(opnd1p1) ^ Dbl_sign(opnd2p1))
69 Dbl_setnegativezerop1(resultp1);
70 else Dbl_setzerop1(resultp1);
71 /*
72 * check first operand for NaN's or infinity
73 */
74 if (Dbl_isinfinity_exponent(opnd1p1)) {
75 if (Dbl_iszero_mantissa(opnd1p1,opnd1p2)) {
76 if (Dbl_isnotnan(opnd2p1,opnd2p2)) {
77 if (Dbl_iszero_exponentmantissa(opnd2p1,opnd2p2)) {
78 /*
79 * invalid since operands are infinity
80 * and zero
81 */
82 if (Is_invalidtrap_enabled())
83 return(INVALIDEXCEPTION);
84 Set_invalidflag();
85 Dbl_makequietnan(resultp1,resultp2);
86 Dbl_copytoptr(resultp1,resultp2,dstptr);
87 return(NOEXCEPTION);
88 }
89 /*
90 * return infinity
91 */
92 Dbl_setinfinity_exponentmantissa(resultp1,resultp2);
93 Dbl_copytoptr(resultp1,resultp2,dstptr);
94 return(NOEXCEPTION);
95 }
96 }
97 else {
98 /*
99 * is NaN; signaling or quiet?
100 */
101 if (Dbl_isone_signaling(opnd1p1)) {
102 /* trap if INVALIDTRAP enabled */
103 if (Is_invalidtrap_enabled())
104 return(INVALIDEXCEPTION);
105 /* make NaN quiet */
106 Set_invalidflag();
107 Dbl_set_quiet(opnd1p1);
108 }
109 /*
110 * is second operand a signaling NaN?
111 */
112 else if (Dbl_is_signalingnan(opnd2p1)) {
113 /* trap if INVALIDTRAP enabled */
114 if (Is_invalidtrap_enabled())
115 return(INVALIDEXCEPTION);
116 /* make NaN quiet */
117 Set_invalidflag();
118 Dbl_set_quiet(opnd2p1);
119 Dbl_copytoptr(opnd2p1,opnd2p2,dstptr);
120 return(NOEXCEPTION);
121 }
122 /*
123 * return quiet NaN
124 */
125 Dbl_copytoptr(opnd1p1,opnd1p2,dstptr);
126 return(NOEXCEPTION);
127 }
128 }
129 /*
130 * check second operand for NaN's or infinity
131 */
132 if (Dbl_isinfinity_exponent(opnd2p1)) {
133 if (Dbl_iszero_mantissa(opnd2p1,opnd2p2)) {
134 if (Dbl_iszero_exponentmantissa(opnd1p1,opnd1p2)) {
135 /* invalid since operands are zero & infinity */
136 if (Is_invalidtrap_enabled())
137 return(INVALIDEXCEPTION);
138 Set_invalidflag();
139 Dbl_makequietnan(opnd2p1,opnd2p2);
140 Dbl_copytoptr(opnd2p1,opnd2p2,dstptr);
141 return(NOEXCEPTION);
142 }
143 /*
144 * return infinity
145 */
146 Dbl_setinfinity_exponentmantissa(resultp1,resultp2);
147 Dbl_copytoptr(resultp1,resultp2,dstptr);
148 return(NOEXCEPTION);
149 }
150 /*
151 * is NaN; signaling or quiet?
152 */
153 if (Dbl_isone_signaling(opnd2p1)) {
154 /* trap if INVALIDTRAP enabled */
155 if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION);
156 /* make NaN quiet */
157 Set_invalidflag();
158 Dbl_set_quiet(opnd2p1);
159 }
160 /*
161 * return quiet NaN
162 */
163 Dbl_copytoptr(opnd2p1,opnd2p2,dstptr);
164 return(NOEXCEPTION);
165 }
166 /*
167 * Generate exponent
168 */
169 dest_exponent = Dbl_exponent(opnd1p1) + Dbl_exponent(opnd2p1) -DBL_BIAS;
170
171 /*
172 * Generate mantissa
173 */
174 if (Dbl_isnotzero_exponent(opnd1p1)) {
175 /* set hidden bit */
176 Dbl_clear_signexponent_set_hidden(opnd1p1);
177 }
178 else {
179 /* check for zero */
180 if (Dbl_iszero_mantissa(opnd1p1,opnd1p2)) {
181 Dbl_setzero_exponentmantissa(resultp1,resultp2);
182 Dbl_copytoptr(resultp1,resultp2,dstptr);
183 return(NOEXCEPTION);
184 }
185 /* is denormalized, adjust exponent */
186 Dbl_clear_signexponent(opnd1p1);
187 Dbl_leftshiftby1(opnd1p1,opnd1p2);
188 Dbl_normalize(opnd1p1,opnd1p2,dest_exponent);
189 }
190 /* opnd2 needs to have hidden bit set with msb in hidden bit */
191 if (Dbl_isnotzero_exponent(opnd2p1)) {
192 Dbl_clear_signexponent_set_hidden(opnd2p1);
193 }
194 else {
195 /* check for zero */
196 if (Dbl_iszero_mantissa(opnd2p1,opnd2p2)) {
197 Dbl_setzero_exponentmantissa(resultp1,resultp2);
198 Dbl_copytoptr(resultp1,resultp2,dstptr);
199 return(NOEXCEPTION);
200 }
201 /* is denormalized; want to normalize */
202 Dbl_clear_signexponent(opnd2p1);
203 Dbl_leftshiftby1(opnd2p1,opnd2p2);
204 Dbl_normalize(opnd2p1,opnd2p2,dest_exponent);
205 }
206
207 /* Multiply two source mantissas together */
208
209 /* make room for guard bits */
210 Dbl_leftshiftby7(opnd2p1,opnd2p2);
211 Dbl_setzero(opnd3p1,opnd3p2);
212 /*
213 * Four bits at a time are inspected in each loop, and a
214 * simple shift and add multiply algorithm is used.
215 */
216 for (count=1;count<=DBL_P;count+=4) {
217 stickybit |= Dlow4p2(opnd3p2);
218 Dbl_rightshiftby4(opnd3p1,opnd3p2);
219 if (Dbit28p2(opnd1p2)) {
220 /* Twoword_add should be an ADDC followed by an ADD. */
221 Twoword_add(opnd3p1, opnd3p2, opnd2p1<<3 | opnd2p2>>29,
222 opnd2p2<<3);
223 }
224 if (Dbit29p2(opnd1p2)) {
225 Twoword_add(opnd3p1, opnd3p2, opnd2p1<<2 | opnd2p2>>30,
226 opnd2p2<<2);
227 }
228 if (Dbit30p2(opnd1p2)) {
229 Twoword_add(opnd3p1, opnd3p2, opnd2p1<<1 | opnd2p2>>31,
230 opnd2p2<<1);
231 }
232 if (Dbit31p2(opnd1p2)) {
233 Twoword_add(opnd3p1, opnd3p2, opnd2p1, opnd2p2);
234 }
235 Dbl_rightshiftby4(opnd1p1,opnd1p2);
236 }
237 if (Dbit3p1(opnd3p1)==0) {
238 Dbl_leftshiftby1(opnd3p1,opnd3p2);
239 }
240 else {
241 /* result mantissa >= 2. */
242 dest_exponent++;
243 }
244 /* check for denormalized result */
245 while (Dbit3p1(opnd3p1)==0) {
246 Dbl_leftshiftby1(opnd3p1,opnd3p2);
247 dest_exponent--;
248 }
249 /*
250 * check for guard, sticky and inexact bits
251 */
252 stickybit |= Dallp2(opnd3p2) << 25;
253 guardbit = (Dallp2(opnd3p2) << 24) >> 31;
254 inexact = guardbit | stickybit;
255
256 /* align result mantissa */
257 Dbl_rightshiftby8(opnd3p1,opnd3p2);
258
259 /*
260 * round result
261 */
262 if (inexact && (dest_exponent>0 || Is_underflowtrap_enabled())) {
263 Dbl_clear_signexponent(opnd3p1);
264 switch (Rounding_mode()) {
265 case ROUNDPLUS:
266 if (Dbl_iszero_sign(resultp1))
267 Dbl_increment(opnd3p1,opnd3p2);
268 break;
269 case ROUNDMINUS:
270 if (Dbl_isone_sign(resultp1))
271 Dbl_increment(opnd3p1,opnd3p2);
272 break;
273 case ROUNDNEAREST:
274 if (guardbit) {
275 if (stickybit || Dbl_isone_lowmantissap2(opnd3p2))
276 Dbl_increment(opnd3p1,opnd3p2);
277 }
278 }
279 if (Dbl_isone_hidden(opnd3p1)) dest_exponent++;
280 }
281 Dbl_set_mantissa(resultp1,resultp2,opnd3p1,opnd3p2);
282
283 /*
284 * Test for overflow
285 */
286 if (dest_exponent >= DBL_INFINITY_EXPONENT) {
287 /* trap if OVERFLOWTRAP enabled */
288 if (Is_overflowtrap_enabled()) {
289 /*
290 * Adjust bias of result
291 */
292 Dbl_setwrapped_exponent(resultp1,dest_exponent,ovfl);
293 Dbl_copytoptr(resultp1,resultp2,dstptr);
294 if (inexact)
295 if (Is_inexacttrap_enabled())
296 return (OVERFLOWEXCEPTION | INEXACTEXCEPTION);
297 else Set_inexactflag();
298 return (OVERFLOWEXCEPTION);
299 }
300 inexact = TRUE;
301 Set_overflowflag();
302 /* set result to infinity or largest number */
303 Dbl_setoverflow(resultp1,resultp2);
304 }
305 /*
306 * Test for underflow
307 */
308 else if (dest_exponent <= 0) {
309 /* trap if UNDERFLOWTRAP enabled */
310 if (Is_underflowtrap_enabled()) {
311 /*
312 * Adjust bias of result
313 */
314 Dbl_setwrapped_exponent(resultp1,dest_exponent,unfl);
315 Dbl_copytoptr(resultp1,resultp2,dstptr);
316 if (inexact)
317 if (Is_inexacttrap_enabled())
318 return (UNDERFLOWEXCEPTION | INEXACTEXCEPTION);
319 else Set_inexactflag();
320 return (UNDERFLOWEXCEPTION);
321 }
322
323 /* Determine if should set underflow flag */
324 is_tiny = TRUE;
325 if (dest_exponent == 0 && inexact) {
326 switch (Rounding_mode()) {
327 case ROUNDPLUS:
328 if (Dbl_iszero_sign(resultp1)) {
329 Dbl_increment(opnd3p1,opnd3p2);
330 if (Dbl_isone_hiddenoverflow(opnd3p1))
331 is_tiny = FALSE;
332 Dbl_decrement(opnd3p1,opnd3p2);
333 }
334 break;
335 case ROUNDMINUS:
336 if (Dbl_isone_sign(resultp1)) {
337 Dbl_increment(opnd3p1,opnd3p2);
338 if (Dbl_isone_hiddenoverflow(opnd3p1))
339 is_tiny = FALSE;
340 Dbl_decrement(opnd3p1,opnd3p2);
341 }
342 break;
343 case ROUNDNEAREST:
344 if (guardbit && (stickybit ||
345 Dbl_isone_lowmantissap2(opnd3p2))) {
346 Dbl_increment(opnd3p1,opnd3p2);
347 if (Dbl_isone_hiddenoverflow(opnd3p1))
348 is_tiny = FALSE;
349 Dbl_decrement(opnd3p1,opnd3p2);
350 }
351 break;
352 }
353 }
354
355 /*
356 * denormalize result or set to signed zero
357 */
358 stickybit = inexact;
359 Dbl_denormalize(opnd3p1,opnd3p2,dest_exponent,guardbit,
360 stickybit,inexact);
361
362 /* return zero or smallest number */
363 if (inexact) {
364 switch (Rounding_mode()) {
365 case ROUNDPLUS:
366 if (Dbl_iszero_sign(resultp1)) {
367 Dbl_increment(opnd3p1,opnd3p2);
368 }
369 break;
370 case ROUNDMINUS:
371 if (Dbl_isone_sign(resultp1)) {
372 Dbl_increment(opnd3p1,opnd3p2);
373 }
374 break;
375 case ROUNDNEAREST:
376 if (guardbit && (stickybit ||
377 Dbl_isone_lowmantissap2(opnd3p2))) {
378 Dbl_increment(opnd3p1,opnd3p2);
379 }
380 break;
381 }
382 if (is_tiny) Set_underflowflag();
383 }
384 Dbl_set_exponentmantissa(resultp1,resultp2,opnd3p1,opnd3p2);
385 }
386 else Dbl_set_exponent(resultp1,dest_exponent);
387 /* check for inexact */
388 Dbl_copytoptr(resultp1,resultp2,dstptr);
389 if (inexact) {
390 if (Is_inexacttrap_enabled()) return(INEXACTEXCEPTION);
391 else Set_inexactflag();
392 }
393 return(NOEXCEPTION);
394}