Commit | Line | Data |
---|---|---|
1da177e4 LT |
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/sfadd.c $Revision: 1.1 $ | |
26 | * | |
27 | * Purpose: | |
28 | * Single_add: add two single precision values. | |
29 | * | |
30 | * External Interfaces: | |
31 | * sgl_fadd(leftptr, rightptr, 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 "sgl_float.h" | |
44 | ||
45 | /* | |
46 | * Single_add: add two single precision values. | |
47 | */ | |
48 | int | |
49 | sgl_fadd( | |
50 | sgl_floating_point *leftptr, | |
51 | sgl_floating_point *rightptr, | |
52 | sgl_floating_point *dstptr, | |
53 | unsigned int *status) | |
54 | { | |
55 | register unsigned int left, right, result, extent; | |
56 | register unsigned int signless_upper_left, signless_upper_right, save; | |
57 | ||
58 | ||
59 | register int result_exponent, right_exponent, diff_exponent; | |
60 | register int sign_save, jumpsize; | |
61 | register boolean inexact = FALSE; | |
62 | register boolean underflowtrap; | |
63 | ||
64 | /* Create local copies of the numbers */ | |
65 | left = *leftptr; | |
66 | right = *rightptr; | |
67 | ||
68 | /* A zero "save" helps discover equal operands (for later), * | |
69 | * and is used in swapping operands (if needed). */ | |
70 | Sgl_xortointp1(left,right,/*to*/save); | |
71 | ||
72 | /* | |
73 | * check first operand for NaN's or infinity | |
74 | */ | |
75 | if ((result_exponent = Sgl_exponent(left)) == SGL_INFINITY_EXPONENT) | |
76 | { | |
77 | if (Sgl_iszero_mantissa(left)) | |
78 | { | |
79 | if (Sgl_isnotnan(right)) | |
80 | { | |
81 | if (Sgl_isinfinity(right) && save!=0) | |
82 | { | |
83 | /* | |
84 | * invalid since operands are opposite signed infinity's | |
85 | */ | |
86 | if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION); | |
87 | Set_invalidflag(); | |
88 | Sgl_makequietnan(result); | |
89 | *dstptr = result; | |
90 | return(NOEXCEPTION); | |
91 | } | |
92 | /* | |
93 | * return infinity | |
94 | */ | |
95 | *dstptr = left; | |
96 | return(NOEXCEPTION); | |
97 | } | |
98 | } | |
99 | else | |
100 | { | |
101 | /* | |
102 | * is NaN; signaling or quiet? | |
103 | */ | |
104 | if (Sgl_isone_signaling(left)) | |
105 | { | |
106 | /* trap if INVALIDTRAP enabled */ | |
107 | if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION); | |
108 | /* make NaN quiet */ | |
109 | Set_invalidflag(); | |
110 | Sgl_set_quiet(left); | |
111 | } | |
112 | /* | |
113 | * is second operand a signaling NaN? | |
114 | */ | |
115 | else if (Sgl_is_signalingnan(right)) | |
116 | { | |
117 | /* trap if INVALIDTRAP enabled */ | |
118 | if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION); | |
119 | /* make NaN quiet */ | |
120 | Set_invalidflag(); | |
121 | Sgl_set_quiet(right); | |
122 | *dstptr = right; | |
123 | return(NOEXCEPTION); | |
124 | } | |
125 | /* | |
126 | * return quiet NaN | |
127 | */ | |
128 | *dstptr = left; | |
129 | return(NOEXCEPTION); | |
130 | } | |
131 | } /* End left NaN or Infinity processing */ | |
132 | /* | |
133 | * check second operand for NaN's or infinity | |
134 | */ | |
135 | if (Sgl_isinfinity_exponent(right)) | |
136 | { | |
137 | if (Sgl_iszero_mantissa(right)) | |
138 | { | |
139 | /* return infinity */ | |
140 | *dstptr = right; | |
141 | return(NOEXCEPTION); | |
142 | } | |
143 | /* | |
144 | * is NaN; signaling or quiet? | |
145 | */ | |
146 | if (Sgl_isone_signaling(right)) | |
147 | { | |
148 | /* trap if INVALIDTRAP enabled */ | |
149 | if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION); | |
150 | /* make NaN quiet */ | |
151 | Set_invalidflag(); | |
152 | Sgl_set_quiet(right); | |
153 | } | |
154 | /* | |
155 | * return quiet NaN | |
156 | */ | |
157 | *dstptr = right; | |
158 | return(NOEXCEPTION); | |
159 | } /* End right NaN or Infinity processing */ | |
160 | ||
161 | /* Invariant: Must be dealing with finite numbers */ | |
162 | ||
163 | /* Compare operands by removing the sign */ | |
164 | Sgl_copytoint_exponentmantissa(left,signless_upper_left); | |
165 | Sgl_copytoint_exponentmantissa(right,signless_upper_right); | |
166 | ||
167 | /* sign difference selects add or sub operation. */ | |
168 | if(Sgl_ismagnitudeless(signless_upper_left,signless_upper_right)) | |
169 | { | |
170 | /* Set the left operand to the larger one by XOR swap * | |
171 | * First finish the first word using "save" */ | |
172 | Sgl_xorfromintp1(save,right,/*to*/right); | |
173 | Sgl_xorfromintp1(save,left,/*to*/left); | |
174 | result_exponent = Sgl_exponent(left); | |
175 | } | |
176 | /* Invariant: left is not smaller than right. */ | |
177 | ||
178 | if((right_exponent = Sgl_exponent(right)) == 0) | |
179 | { | |
180 | /* Denormalized operands. First look for zeroes */ | |
181 | if(Sgl_iszero_mantissa(right)) | |
182 | { | |
183 | /* right is zero */ | |
184 | if(Sgl_iszero_exponentmantissa(left)) | |
185 | { | |
186 | /* Both operands are zeros */ | |
187 | if(Is_rounding_mode(ROUNDMINUS)) | |
188 | { | |
189 | Sgl_or_signs(left,/*with*/right); | |
190 | } | |
191 | else | |
192 | { | |
193 | Sgl_and_signs(left,/*with*/right); | |
194 | } | |
195 | } | |
196 | else | |
197 | { | |
198 | /* Left is not a zero and must be the result. Trapped | |
199 | * underflows are signaled if left is denormalized. Result | |
200 | * is always exact. */ | |
201 | if( (result_exponent == 0) && Is_underflowtrap_enabled() ) | |
202 | { | |
203 | /* need to normalize results mantissa */ | |
204 | sign_save = Sgl_signextendedsign(left); | |
205 | Sgl_leftshiftby1(left); | |
206 | Sgl_normalize(left,result_exponent); | |
207 | Sgl_set_sign(left,/*using*/sign_save); | |
208 | Sgl_setwrapped_exponent(left,result_exponent,unfl); | |
209 | *dstptr = left; | |
210 | return(UNDERFLOWEXCEPTION); | |
211 | } | |
212 | } | |
213 | *dstptr = left; | |
214 | return(NOEXCEPTION); | |
215 | } | |
216 | ||
217 | /* Neither are zeroes */ | |
218 | Sgl_clear_sign(right); /* Exponent is already cleared */ | |
219 | if(result_exponent == 0 ) | |
220 | { | |
221 | /* Both operands are denormalized. The result must be exact | |
222 | * and is simply calculated. A sum could become normalized and a | |
223 | * difference could cancel to a true zero. */ | |
224 | if( (/*signed*/int) save < 0 ) | |
225 | { | |
226 | Sgl_subtract(left,/*minus*/right,/*into*/result); | |
227 | if(Sgl_iszero_mantissa(result)) | |
228 | { | |
229 | if(Is_rounding_mode(ROUNDMINUS)) | |
230 | { | |
231 | Sgl_setone_sign(result); | |
232 | } | |
233 | else | |
234 | { | |
235 | Sgl_setzero_sign(result); | |
236 | } | |
237 | *dstptr = result; | |
238 | return(NOEXCEPTION); | |
239 | } | |
240 | } | |
241 | else | |
242 | { | |
243 | Sgl_addition(left,right,/*into*/result); | |
244 | if(Sgl_isone_hidden(result)) | |
245 | { | |
246 | *dstptr = result; | |
247 | return(NOEXCEPTION); | |
248 | } | |
249 | } | |
250 | if(Is_underflowtrap_enabled()) | |
251 | { | |
252 | /* need to normalize result */ | |
253 | sign_save = Sgl_signextendedsign(result); | |
254 | Sgl_leftshiftby1(result); | |
255 | Sgl_normalize(result,result_exponent); | |
256 | Sgl_set_sign(result,/*using*/sign_save); | |
257 | Sgl_setwrapped_exponent(result,result_exponent,unfl); | |
258 | *dstptr = result; | |
259 | return(UNDERFLOWEXCEPTION); | |
260 | } | |
261 | *dstptr = result; | |
262 | return(NOEXCEPTION); | |
263 | } | |
264 | right_exponent = 1; /* Set exponent to reflect different bias | |
265 | * with denomalized numbers. */ | |
266 | } | |
267 | else | |
268 | { | |
269 | Sgl_clear_signexponent_set_hidden(right); | |
270 | } | |
271 | Sgl_clear_exponent_set_hidden(left); | |
272 | diff_exponent = result_exponent - right_exponent; | |
273 | ||
274 | /* | |
275 | * Special case alignment of operands that would force alignment | |
276 | * beyond the extent of the extension. A further optimization | |
277 | * could special case this but only reduces the path length for this | |
278 | * infrequent case. | |
279 | */ | |
280 | if(diff_exponent > SGL_THRESHOLD) | |
281 | { | |
282 | diff_exponent = SGL_THRESHOLD; | |
283 | } | |
284 | ||
285 | /* Align right operand by shifting to right */ | |
286 | Sgl_right_align(/*operand*/right,/*shifted by*/diff_exponent, | |
287 | /*and lower to*/extent); | |
288 | ||
289 | /* Treat sum and difference of the operands separately. */ | |
290 | if( (/*signed*/int) save < 0 ) | |
291 | { | |
292 | /* | |
293 | * Difference of the two operands. Their can be no overflow. A | |
294 | * borrow can occur out of the hidden bit and force a post | |
295 | * normalization phase. | |
296 | */ | |
297 | Sgl_subtract_withextension(left,/*minus*/right,/*with*/extent,/*into*/result); | |
298 | if(Sgl_iszero_hidden(result)) | |
299 | { | |
300 | /* Handle normalization */ | |
301 | /* A straight foward algorithm would now shift the result | |
302 | * and extension left until the hidden bit becomes one. Not | |
303 | * all of the extension bits need participate in the shift. | |
304 | * Only the two most significant bits (round and guard) are | |
305 | * needed. If only a single shift is needed then the guard | |
306 | * bit becomes a significant low order bit and the extension | |
307 | * must participate in the rounding. If more than a single | |
308 | * shift is needed, then all bits to the right of the guard | |
309 | * bit are zeros, and the guard bit may or may not be zero. */ | |
310 | sign_save = Sgl_signextendedsign(result); | |
311 | Sgl_leftshiftby1_withextent(result,extent,result); | |
312 | ||
313 | /* Need to check for a zero result. The sign and exponent | |
314 | * fields have already been zeroed. The more efficient test | |
315 | * of the full object can be used. | |
316 | */ | |
317 | if(Sgl_iszero(result)) | |
318 | /* Must have been "x-x" or "x+(-x)". */ | |
319 | { | |
320 | if(Is_rounding_mode(ROUNDMINUS)) Sgl_setone_sign(result); | |
321 | *dstptr = result; | |
322 | return(NOEXCEPTION); | |
323 | } | |
324 | result_exponent--; | |
325 | /* Look to see if normalization is finished. */ | |
326 | if(Sgl_isone_hidden(result)) | |
327 | { | |
328 | if(result_exponent==0) | |
329 | { | |
330 | /* Denormalized, exponent should be zero. Left operand * | |
331 | * was normalized, so extent (guard, round) was zero */ | |
332 | goto underflow; | |
333 | } | |
334 | else | |
335 | { | |
336 | /* No further normalization is needed. */ | |
337 | Sgl_set_sign(result,/*using*/sign_save); | |
338 | Ext_leftshiftby1(extent); | |
339 | goto round; | |
340 | } | |
341 | } | |
342 | ||
343 | /* Check for denormalized, exponent should be zero. Left * | |
344 | * operand was normalized, so extent (guard, round) was zero */ | |
345 | if(!(underflowtrap = Is_underflowtrap_enabled()) && | |
346 | result_exponent==0) goto underflow; | |
347 | ||
348 | /* Shift extension to complete one bit of normalization and | |
349 | * update exponent. */ | |
350 | Ext_leftshiftby1(extent); | |
351 | ||
352 | /* Discover first one bit to determine shift amount. Use a | |
353 | * modified binary search. We have already shifted the result | |
354 | * one position right and still not found a one so the remainder | |
355 | * of the extension must be zero and simplifies rounding. */ | |
356 | /* Scan bytes */ | |
357 | while(Sgl_iszero_hiddenhigh7mantissa(result)) | |
358 | { | |
359 | Sgl_leftshiftby8(result); | |
360 | if((result_exponent -= 8) <= 0 && !underflowtrap) | |
361 | goto underflow; | |
362 | } | |
363 | /* Now narrow it down to the nibble */ | |
364 | if(Sgl_iszero_hiddenhigh3mantissa(result)) | |
365 | { | |
366 | /* The lower nibble contains the normalizing one */ | |
367 | Sgl_leftshiftby4(result); | |
368 | if((result_exponent -= 4) <= 0 && !underflowtrap) | |
369 | goto underflow; | |
370 | } | |
371 | /* Select case were first bit is set (already normalized) | |
372 | * otherwise select the proper shift. */ | |
373 | if((jumpsize = Sgl_hiddenhigh3mantissa(result)) > 7) | |
374 | { | |
375 | /* Already normalized */ | |
376 | if(result_exponent <= 0) goto underflow; | |
377 | Sgl_set_sign(result,/*using*/sign_save); | |
378 | Sgl_set_exponent(result,/*using*/result_exponent); | |
379 | *dstptr = result; | |
380 | return(NOEXCEPTION); | |
381 | } | |
382 | Sgl_sethigh4bits(result,/*using*/sign_save); | |
383 | switch(jumpsize) | |
384 | { | |
385 | case 1: | |
386 | { | |
387 | Sgl_leftshiftby3(result); | |
388 | result_exponent -= 3; | |
389 | break; | |
390 | } | |
391 | case 2: | |
392 | case 3: | |
393 | { | |
394 | Sgl_leftshiftby2(result); | |
395 | result_exponent -= 2; | |
396 | break; | |
397 | } | |
398 | case 4: | |
399 | case 5: | |
400 | case 6: | |
401 | case 7: | |
402 | { | |
403 | Sgl_leftshiftby1(result); | |
404 | result_exponent -= 1; | |
405 | break; | |
406 | } | |
407 | } | |
408 | if(result_exponent > 0) | |
409 | { | |
410 | Sgl_set_exponent(result,/*using*/result_exponent); | |
411 | *dstptr = result; | |
412 | return(NOEXCEPTION); /* Sign bit is already set */ | |
413 | } | |
414 | /* Fixup potential underflows */ | |
415 | underflow: | |
416 | if(Is_underflowtrap_enabled()) | |
417 | { | |
418 | Sgl_set_sign(result,sign_save); | |
419 | Sgl_setwrapped_exponent(result,result_exponent,unfl); | |
420 | *dstptr = result; | |
421 | /* inexact = FALSE; */ | |
422 | return(UNDERFLOWEXCEPTION); | |
423 | } | |
424 | /* | |
425 | * Since we cannot get an inexact denormalized result, | |
426 | * we can now return. | |
427 | */ | |
428 | Sgl_right_align(result,/*by*/(1-result_exponent),extent); | |
429 | Sgl_clear_signexponent(result); | |
430 | Sgl_set_sign(result,sign_save); | |
431 | *dstptr = result; | |
432 | return(NOEXCEPTION); | |
433 | } /* end if(hidden...)... */ | |
434 | /* Fall through and round */ | |
435 | } /* end if(save < 0)... */ | |
436 | else | |
437 | { | |
438 | /* Add magnitudes */ | |
439 | Sgl_addition(left,right,/*to*/result); | |
440 | if(Sgl_isone_hiddenoverflow(result)) | |
441 | { | |
442 | /* Prenormalization required. */ | |
443 | Sgl_rightshiftby1_withextent(result,extent,extent); | |
444 | Sgl_arithrightshiftby1(result); | |
445 | result_exponent++; | |
446 | } /* end if hiddenoverflow... */ | |
447 | } /* end else ...add magnitudes... */ | |
448 | ||
449 | /* Round the result. If the extension is all zeros,then the result is | |
450 | * exact. Otherwise round in the correct direction. No underflow is | |
451 | * possible. If a postnormalization is necessary, then the mantissa is | |
452 | * all zeros so no shift is needed. */ | |
453 | round: | |
454 | if(Ext_isnotzero(extent)) | |
455 | { | |
456 | inexact = TRUE; | |
457 | switch(Rounding_mode()) | |
458 | { | |
459 | case ROUNDNEAREST: /* The default. */ | |
460 | if(Ext_isone_sign(extent)) | |
461 | { | |
462 | /* at least 1/2 ulp */ | |
463 | if(Ext_isnotzero_lower(extent) || | |
464 | Sgl_isone_lowmantissa(result)) | |
465 | { | |
466 | /* either exactly half way and odd or more than 1/2ulp */ | |
467 | Sgl_increment(result); | |
468 | } | |
469 | } | |
470 | break; | |
471 | ||
472 | case ROUNDPLUS: | |
473 | if(Sgl_iszero_sign(result)) | |
474 | { | |
475 | /* Round up positive results */ | |
476 | Sgl_increment(result); | |
477 | } | |
478 | break; | |
479 | ||
480 | case ROUNDMINUS: | |
481 | if(Sgl_isone_sign(result)) | |
482 | { | |
483 | /* Round down negative results */ | |
484 | Sgl_increment(result); | |
485 | } | |
486 | ||
487 | case ROUNDZERO:; | |
488 | /* truncate is simple */ | |
489 | } /* end switch... */ | |
490 | if(Sgl_isone_hiddenoverflow(result)) result_exponent++; | |
491 | } | |
492 | if(result_exponent == SGL_INFINITY_EXPONENT) | |
493 | { | |
494 | /* Overflow */ | |
495 | if(Is_overflowtrap_enabled()) | |
496 | { | |
497 | Sgl_setwrapped_exponent(result,result_exponent,ovfl); | |
498 | *dstptr = result; | |
499 | if (inexact) | |
500 | if (Is_inexacttrap_enabled()) | |
501 | return(OVERFLOWEXCEPTION | INEXACTEXCEPTION); | |
502 | else Set_inexactflag(); | |
503 | return(OVERFLOWEXCEPTION); | |
504 | } | |
505 | else | |
506 | { | |
507 | Set_overflowflag(); | |
508 | inexact = TRUE; | |
509 | Sgl_setoverflow(result); | |
510 | } | |
511 | } | |
512 | else Sgl_set_exponent(result,result_exponent); | |
513 | *dstptr = result; | |
514 | if(inexact) | |
515 | if(Is_inexacttrap_enabled()) return(INEXACTEXCEPTION); | |
516 | else Set_inexactflag(); | |
517 | return(NOEXCEPTION); | |
518 | } |