| blob | a9f7f01290672feded67ad7b4bd6f662668355c8 |
1 /*
2 * fp_util.S
3 *
4 * Copyright Roman Zippel, 1997. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, and the entire permission notice in its entirety,
11 * including the disclaimer of warranties.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. The name of the author may not be used to endorse or promote
16 * products derived from this software without specific prior
17 * written permission.
18 *
19 * ALTERNATIVELY, this product may be distributed under the terms of
20 * the GNU General Public License, in which case the provisions of the GPL are
21 * required INSTEAD OF the above restrictions. (This clause is
22 * necessary due to a potential bad interaction between the GPL and
23 * the restrictions contained in a BSD-style copyright.)
24 *
25 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
27 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
29 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
30 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
31 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
33 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
35 * OF THE POSSIBILITY OF SUCH DAMAGE.
36 */
38 #include <linux/config.h>
39 #include "fp_emu.h"
41 /*
42 * Here are lots of conversion and normalization functions mainly
43 * used by fp_scan.S
44 * Note that these functions are optimized for "normal" numbers,
45 * these are handled first and exit as fast as possible, this is
46 * especially important for fp_normalize_ext/fp_conv_ext2ext, as
47 * it's called very often.
48 * The register usage is optimized for fp_scan.S and which register
49 * is currently at that time unused, be careful if you want change
50 * something here. %d0 and %d1 is always usable, sometimes %d2 (or
51 * only the lower half) most function have to return the %a0
52 * unmodified, so that the caller can immediately reuse it.
53 */
55 .globl fp_ill, fp_end
57 | exits from fp_scan:
58 | illegal instruction
59 fp_ill:
60 printf ,"fp_illegal\n"
61 rts
62 | completed instruction
63 fp_end:
64 tst.l (TASK_MM-8,%a2)
65 jmi 1f
66 tst.l (TASK_MM-4,%a2)
67 jmi 1f
68 tst.l (TASK_MM,%a2)
69 jpl 2f
70 1: printf ,"oops:%p,%p,%p\n",3,%a2@(TASK_MM-8),%a2@(TASK_MM-4),%a2@(TASK_MM)
71 2: clr.l %d0
72 rts
74 .globl fp_conv_long2ext, fp_conv_single2ext
75 .globl fp_conv_double2ext, fp_conv_ext2ext
76 .globl fp_normalize_ext, fp_normalize_double
77 .globl fp_normalize_single, fp_normalize_single_fast
78 .globl fp_conv_ext2double, fp_conv_ext2single
79 .globl fp_conv_ext2long, fp_conv_ext2short
80 .globl fp_conv_ext2byte
81 .globl fp_finalrounding_single, fp_finalrounding_single_fast
82 .globl fp_finalrounding_double
83 .globl fp_finalrounding, fp_finaltest, fp_final
85 /*
86 * First several conversion functions from a source operand
87 * into the extended format. Note, that only fp_conv_ext2ext
88 * normalizes the number and is always called after the other
89 * conversion functions, which only move the information into
90 * fp_ext structure.
91 */
93 | fp_conv_long2ext:
94 |
95 | args: %d0 = source (32-bit long)
96 | %a0 = destination (ptr to struct fp_ext)
98 fp_conv_long2ext:
99 printf PCONV,"l2e: %p -> %p(",2,%d0,%a0
100 clr.l %d1 | sign defaults to zero
101 tst.l %d0
102 jeq fp_l2e_zero | is source zero?
103 jpl 1f | positive?
104 moveq #1,%d1
105 neg.l %d0
106 1: swap %d1
107 move.w #0x3fff+31,%d1
108 move.l %d1,(%a0)+ | set sign / exp
109 move.l %d0,(%a0)+ | set mantissa
110 clr.l (%a0)
111 subq.l #8,%a0 | restore %a0
112 printx PCONV,%a0@
113 printf PCONV,")\n"
114 rts
115 | source is zero
116 fp_l2e_zero:
117 clr.l (%a0)+
118 clr.l (%a0)+
119 clr.l (%a0)
120 subq.l #8,%a0
121 printx PCONV,%a0@
122 printf PCONV,")\n"
123 rts
125 | fp_conv_single2ext
126 | args: %d0 = source (single-precision fp value)
127 | %a0 = dest (struct fp_ext *)
129 fp_conv_single2ext:
130 printf PCONV,"s2e: %p -> %p(",2,%d0,%a0
131 move.l %d0,%d1
132 lsl.l #8,%d0 | shift mantissa
133 lsr.l #8,%d1 | exponent / sign
134 lsr.l #7,%d1
135 lsr.w #8,%d1
136 jeq fp_s2e_small | zero / denormal?
137 cmp.w #0xff,%d1 | NaN / Inf?
138 jeq fp_s2e_large
139 bset #31,%d0 | set explizit bit
140 add.w #0x3fff-0x7f,%d1 | re-bias the exponent.
141 9: move.l %d1,(%a0)+ | fp_ext.sign, fp_ext.exp
142 move.l %d0,(%a0)+ | high lword of fp_ext.mant
143 clr.l (%a0) | low lword = 0
144 subq.l #8,%a0
145 printx PCONV,%a0@
146 printf PCONV,")\n"
147 rts
148 | zeros and denormalized
149 fp_s2e_small:
150 | exponent is zero, so explizit bit is already zero too
151 tst.l %d0
152 jeq 9b
153 move.w #0x4000-0x7f,%d1
154 jra 9b
155 | infinities and NAN
156 fp_s2e_large:
157 bclr #31,%d0 | clear explizit bit
158 move.w #0x7fff,%d1
159 jra 9b
161 fp_conv_double2ext:
162 #ifdef FPU_EMU_DEBUG
163 getuser.l %a1@(0),%d0,fp_err_ua2,%a1
164 getuser.l %a1@(4),%d1,fp_err_ua2,%a1
165 printf PCONV,"d2e: %p%p -> %p(",3,%d0,%d1,%a0
166 #endif
167 getuser.l (%a1)+,%d0,fp_err_ua2,%a1
168 move.l %d0,%d1
169 lsl.l #8,%d0 | shift high mantissa
170 lsl.l #3,%d0
171 lsr.l #8,%d1 | exponent / sign
172 lsr.l #7,%d1
173 lsr.w #5,%d1
174 jeq fp_d2e_small | zero / denormal?
175 cmp.w #0x7ff,%d1 | NaN / Inf?
176 jeq fp_d2e_large
177 bset #31,%d0 | set explizit bit
178 add.w #0x3fff-0x3ff,%d1 | re-bias the exponent.
179 9: move.l %d1,(%a0)+ | fp_ext.sign, fp_ext.exp
180 move.l %d0,(%a0)+
181 getuser.l (%a1)+,%d0,fp_err_ua2,%a1
182 move.l %d0,%d1
183 lsl.l #8,%d0
184 lsl.l #3,%d0
185 move.l %d0,(%a0)
186 moveq #21,%d0
187 lsr.l %d0,%d1
188 or.l %d1,-(%a0)
189 subq.l #4,%a0
190 printx PCONV,%a0@
191 printf PCONV,")\n"
192 rts
193 | zeros and denormalized
194 fp_d2e_small:
195 | exponent is zero, so explizit bit is already zero too
196 tst.l %d0
197 jeq 9b
198 move.w #0x4000-0x3ff,%d1
199 jra 9b
200 | infinities and NAN
201 fp_d2e_large:
202 bclr #31,%d0 | clear explizit bit
203 move.w #0x7fff,%d1
204 jra 9b
206 | fp_conv_ext2ext:
207 | originally used to get longdouble from userspace, now it's
208 | called before arithmetic operations to make sure the number
209 | is normalized [maybe rename it?].
210 | args: %a0 = dest (struct fp_ext *)
211 | returns 0 in %d0 for a NaN, otherwise 1
213 fp_conv_ext2ext:
214 printf PCONV,"e2e: %p(",1,%a0
215 printx PCONV,%a0@
216 printf PCONV,"), "
217 move.l (%a0)+,%d0
218 cmp.w #0x7fff,%d0 | Inf / NaN?
219 jeq fp_e2e_large
220 move.l (%a0),%d0
221 jpl fp_e2e_small | zero / denorm?
222 | The high bit is set, so normalization is irrelevant.
223 fp_e2e_checkround:
224 subq.l #4,%a0
225 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
226 move.b (%a0),%d0
227 jne fp_e2e_round
228 #endif
229 printf PCONV,"%p(",1,%a0
230 printx PCONV,%a0@
231 printf PCONV,")\n"
232 moveq #1,%d0
233 rts
234 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
235 fp_e2e_round:
236 fp_set_sr FPSR_EXC_INEX2
237 clr.b (%a0)
238 move.w (FPD_RND,FPDATA),%d2
239 jne fp_e2e_roundother | %d2 == 0, round to nearest
240 tst.b %d0 | test guard bit
241 jpl 9f | zero is closer
242 btst #0,(11,%a0) | test lsb bit
243 jne fp_e2e_doroundup | round to infinity
244 lsl.b #1,%d0 | check low bits
245 jeq 9f | round to zero
246 fp_e2e_doroundup:
247 addq.l #1,(8,%a0)
248 jcc 9f
249 addq.l #1,(4,%a0)
250 jcc 9f
251 move.w #0x8000,(4,%a0)
252 addq.w #1,(2,%a0)
253 9: printf PNORM,"%p(",1,%a0
254 printx PNORM,%a0@
255 printf PNORM,")\n"
256 rts
257 fp_e2e_roundother:
258 subq.w #2,%d2
259 jcs 9b | %d2 < 2, round to zero
260 jhi 1f | %d2 > 2, round to +infinity
261 tst.b (1,%a0) | to -inf
262 jne fp_e2e_doroundup | negative, round to infinity
263 jra 9b | positive, round to zero
264 1: tst.b (1,%a0) | to +inf
265 jeq fp_e2e_doroundup | positive, round to infinity
266 jra 9b | negative, round to zero
267 #endif
268 | zeros and subnormals:
269 | try to normalize these anyway.
270 fp_e2e_small:
271 jne fp_e2e_small1 | high lword zero?
272 move.l (4,%a0),%d0
273 jne fp_e2e_small2
274 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
275 clr.l %d0
276 move.b (-4,%a0),%d0
277 jne fp_e2e_small3
278 #endif
279 | Genuine zero.
280 clr.w -(%a0)
281 subq.l #2,%a0
282 printf PNORM,"%p(",1,%a0
283 printx PNORM,%a0@
284 printf PNORM,")\n"
285 moveq #1,%d0
286 rts
287 | definitely subnormal, need to shift all 64 bits
288 fp_e2e_small1:
289 bfffo %d0{#0,#32},%d1
290 move.w -(%a0),%d2
291 sub.w %d1,%d2
292 jcc 1f
293 | Pathologically small, denormalize.
294 add.w %d2,%d1
295 clr.w %d2
296 1: move.w %d2,(%a0)+
297 move.w %d1,%d2
298 jeq fp_e2e_checkround
299 | fancy 64-bit double-shift begins here
300 lsl.l %d2,%d0
301 move.l %d0,(%a0)+
302 move.l (%a0),%d0
303 move.l %d0,%d1
304 lsl.l %d2,%d0
305 move.l %d0,(%a0)
306 neg.w %d2
307 and.w #0x1f,%d2
308 lsr.l %d2,%d1
309 or.l %d1,-(%a0)
310 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
311 fp_e2e_extra1:
312 clr.l %d0
313 move.b (-4,%a0),%d0
314 neg.w %d2
315 add.w #24,%d2
316 jcc 1f
317 clr.b (-4,%a0)
318 lsl.l %d2,%d0
319 or.l %d0,(4,%a0)
320 jra fp_e2e_checkround
321 1: addq.w #8,%d2
322 lsl.l %d2,%d0
323 move.b %d0,(-4,%a0)
324 lsr.l #8,%d0
325 or.l %d0,(4,%a0)
326 #endif
327 jra fp_e2e_checkround
328 | pathologically small subnormal
329 fp_e2e_small2:
330 bfffo %d0{#0,#32},%d1
331 add.w #32,%d1
332 move.w -(%a0),%d2
333 sub.w %d1,%d2
334 jcc 1f
335 | Beyond pathologically small, denormalize.
336 add.w %d2,%d1
337 clr.w %d2
338 1: move.w %d2,(%a0)+
339 ext.l %d1
340 jeq fp_e2e_checkround
341 clr.l (4,%a0)
342 sub.w #32,%d2
343 jcs 1f
344 lsl.l %d1,%d0 | lower lword needs only to be shifted
345 move.l %d0,(%a0) | into the higher lword
346 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
347 clr.l %d0
348 move.b (-4,%a0),%d0
349 clr.b (-4,%a0)
350 neg.w %d1
351 add.w #32,%d1
352 bfins %d0,(%a0){%d1,#8}
353 #endif
354 jra fp_e2e_checkround
355 1: neg.w %d1 | lower lword is splitted between
356 bfins %d0,(%a0){%d1,#32} | higher and lower lword
357 #ifndef CONFIG_M68KFPU_EMU_EXTRAPREC
358 jra fp_e2e_checkround
359 #else
360 move.w %d1,%d2
361 jra fp_e2e_extra1
362 | These are extremely small numbers, that will mostly end up as zero
363 | anyway, so this is only important for correct rounding.
364 fp_e2e_small3:
365 bfffo %d0{#24,#8},%d1
366 add.w #40,%d1
367 move.w -(%a0),%d2
368 sub.w %d1,%d2
369 jcc 1f
370 | Pathologically small, denormalize.
371 add.w %d2,%d1
372 clr.w %d2
373 1: move.w %d2,(%a0)+
374 ext.l %d1
375 jeq fp_e2e_checkround
376 cmp.w #8,%d1
377 jcs 2f
378 1: clr.b (-4,%a0)
379 sub.w #64,%d1
380 jcs 1f
381 add.w #24,%d1
382 lsl.l %d1,%d0
383 move.l %d0,(%a0)
384 jra fp_e2e_checkround
385 1: neg.w %d1
386 bfins %d0,(%a0){%d1,#8}
387 jra fp_e2e_checkround
388 2: lsl.l %d1,%d0
389 move.b %d0,(-4,%a0)
390 lsr.l #8,%d0
391 move.b %d0,(7,%a0)
392 jra fp_e2e_checkround
393 #endif
394 1: move.l %d0,%d1 | lower lword is splitted between
395 lsl.l %d2,%d0 | higher and lower lword
396 move.l %d0,(%a0)
397 move.l %d1,%d0
398 neg.w %d2
399 add.w #32,%d2
400 lsr.l %d2,%d0
401 move.l %d0,-(%a0)
402 jra fp_e2e_checkround
403 | Infinities and NaNs
404 fp_e2e_large:
405 move.l (%a0)+,%d0
406 jne 3f
407 1: tst.l (%a0)
408 jne 4f
409 moveq #1,%d0
410 2: subq.l #8,%a0
411 printf PCONV,"%p(",1,%a0
412 printx PCONV,%a0@
413 printf PCONV,")\n"
414 rts
415 | we have maybe a NaN, shift off the highest bit
416 3: lsl.l #1,%d0
417 jeq 1b
418 | we have a NaN, clear the return value
419 4: clrl %d0
420 jra 2b
423 /*
424 * Normalization functions. Call these on the output of general
425 * FP operators, and before any conversion into the destination
426 * formats. fp_normalize_ext has always to be called first, the
427 * following conversion functions expect an already normalized
428 * number.
429 */
431 | fp_normalize_ext:
432 | normalize an extended in extended (unpacked) format, basically
433 | it does the same as fp_conv_ext2ext, additionally it also does
434 | the necessary postprocessing checks.
435 | args: %a0 (struct fp_ext *)
436 | NOTE: it does _not_ modify %a0/%a1 and the upper word of %d2
438 fp_normalize_ext:
439 printf PNORM,"ne: %p(",1,%a0
440 printx PNORM,%a0@
441 printf PNORM,"), "
442 move.l (%a0)+,%d0
443 cmp.w #0x7fff,%d0 | Inf / NaN?
444 jeq fp_ne_large
445 move.l (%a0),%d0
446 jpl fp_ne_small | zero / denorm?
447 | The high bit is set, so normalization is irrelevant.
448 fp_ne_checkround:
449 subq.l #4,%a0
450 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
451 move.b (%a0),%d0
452 jne fp_ne_round
453 #endif
454 printf PNORM,"%p(",1,%a0
455 printx PNORM,%a0@
456 printf PNORM,")\n"
457 rts
458 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
459 fp_ne_round:
460 fp_set_sr FPSR_EXC_INEX2
461 clr.b (%a0)
462 move.w (FPD_RND,FPDATA),%d2
463 jne fp_ne_roundother | %d2 == 0, round to nearest
464 tst.b %d0 | test guard bit
465 jpl 9f | zero is closer
466 btst #0,(11,%a0) | test lsb bit
467 jne fp_ne_doroundup | round to infinity
468 lsl.b #1,%d0 | check low bits
469 jeq 9f | round to zero
470 fp_ne_doroundup:
471 addq.l #1,(8,%a0)
472 jcc 9f
473 addq.l #1,(4,%a0)
474 jcc 9f
475 addq.w #1,(2,%a0)
476 move.w #0x8000,(4,%a0)
477 9: printf PNORM,"%p(",1,%a0
478 printx PNORM,%a0@
479 printf PNORM,")\n"
480 rts
481 fp_ne_roundother:
482 subq.w #2,%d2
483 jcs 9b | %d2 < 2, round to zero
484 jhi 1f | %d2 > 2, round to +infinity
485 tst.b (1,%a0) | to -inf
486 jne fp_ne_doroundup | negative, round to infinity
487 jra 9b | positive, round to zero
488 1: tst.b (1,%a0) | to +inf
489 jeq fp_ne_doroundup | positive, round to infinity
490 jra 9b | negative, round to zero
491 #endif
492 | Zeros and subnormal numbers
493 | These are probably merely subnormal, rather than "denormalized"
494 | numbers, so we will try to make them normal again.
495 fp_ne_small:
496 jne fp_ne_small1 | high lword zero?
497 move.l (4,%a0),%d0
498 jne fp_ne_small2
499 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
500 clr.l %d0
501 move.b (-4,%a0),%d0
502 jne fp_ne_small3
503 #endif
504 | Genuine zero.
505 clr.w -(%a0)
506 subq.l #2,%a0
507 printf PNORM,"%p(",1,%a0
508 printx PNORM,%a0@
509 printf PNORM,")\n"
510 rts
511 | Subnormal.
512 fp_ne_small1:
513 bfffo %d0{#0,#32},%d1
514 move.w -(%a0),%d2
515 sub.w %d1,%d2
516 jcc 1f
517 | Pathologically small, denormalize.
518 add.w %d2,%d1
519 clr.w %d2
520 fp_set_sr FPSR_EXC_UNFL
521 1: move.w %d2,(%a0)+
522 move.w %d1,%d2
523 jeq fp_ne_checkround
524 | This is exactly the same 64-bit double shift as seen above.
525 lsl.l %d2,%d0
526 move.l %d0,(%a0)+
527 move.l (%a0),%d0
528 move.l %d0,%d1
529 lsl.l %d2,%d0
530 move.l %d0,(%a0)
531 neg.w %d2
532 and.w #0x1f,%d2
533 lsr.l %d2,%d1
534 or.l %d1,-(%a0)
535 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
536 fp_ne_extra1:
537 clr.l %d0
538 move.b (-4,%a0),%d0
539 neg.w %d2
540 add.w #24,%d2
541 jcc 1f
542 clr.b (-4,%a0)
543 lsl.l %d2,%d0
544 or.l %d0,(4,%a0)
545 jra fp_ne_checkround
546 1: addq.w #8,%d2
547 lsl.l %d2,%d0
548 move.b %d0,(-4,%a0)
549 lsr.l #8,%d0
550 or.l %d0,(4,%a0)
551 #endif
552 jra fp_ne_checkround
553 | May or may not be subnormal, if so, only 32 bits to shift.
554 fp_ne_small2:
555 bfffo %d0{#0,#32},%d1
556 add.w #32,%d1
557 move.w -(%a0),%d2
558 sub.w %d1,%d2
559 jcc 1f
560 | Beyond pathologically small, denormalize.
561 add.w %d2,%d1
562 clr.w %d2
563 fp_set_sr FPSR_EXC_UNFL
564 1: move.w %d2,(%a0)+
565 ext.l %d1
566 jeq fp_ne_checkround
567 clr.l (4,%a0)
568 sub.w #32,%d1
569 jcs 1f
570 lsl.l %d1,%d0 | lower lword needs only to be shifted
571 move.l %d0,(%a0) | into the higher lword
572 #ifdef CONFIG_M68KFPU_EMU_EXTRAPREC
573 clr.l %d0
574 move.b (-4,%a0),%d0
575 clr.b (-4,%a0)
576 neg.w %d1
577 add.w #32,%d1
578 bfins %d0,(%a0){%d1,#8}
579 #endif
580 jra fp_ne_checkround
581 1: neg.w %d1 | lower lword is splitted between
582 bfins %d0,(%a0){%d1,#32} | higher and lower lword
583 #ifndef CONFIG_M68KFPU_EMU_EXTRAPREC
584 jra fp_ne_checkround
585 #else
586 move.w %d1,%d2
587 jra fp_ne_extra1
588 | These are extremely small numbers, that will mostly end up as zero
589 | anyway, so this is only important for correct rounding.
590 fp_ne_small3:
591 bfffo %d0{#24,#8},%d1
592 add.w #40,%d1
593 move.w -(%a0),%d2
594 sub.w %d1,%d2
595 jcc 1f
596 | Pathologically small, denormalize.
597 add.w %d2,%d1
598 clr.w %d2
599 1: move.w %d2,(%a0)+
600 ext.l %d1
601 jeq fp_ne_checkround
602 cmp.w #8,%d1
603 jcs 2f
604 1: clr.b (-4,%a0)
605 sub.w #64,%d1
606 jcs 1f
607 add.w #24,%d1
608 lsl.l %d1,%d0
609 move.l %d0,(%a0)
610 jra fp_ne_checkround
611 1: neg.w %d1
612 bfins %d0,(%a0){%d1,#8}
613 jra fp_ne_checkround
614 2: lsl.l %d1,%d0
615 move.b %d0,(-4,%a0)
616 lsr.l #8,%d0
617 move.b %d0,(7,%a0)
618 jra fp_ne_checkround
619 #endif
620 | Infinities and NaNs, again, same as above.
621 fp_ne_large:
622 move.l (%a0)+,%d0
623 jne 3f
624 1: tst.l (%a0)
625 jne 4f
626 2: subq.l #8,%a0
627 printf PNORM,"%p(",1,%a0
628 printx PNORM,%a0@
629 printf PNORM,")\n"
630 rts
631 | we have maybe a NaN, shift off the highest bit
632 3: move.l %d0,%d1
633 lsl.l #1,%d1
634 jne 4f
635 clr.l (-4,%a0)
636 jra 1b
637 | we have a NaN, test if it is signaling
638 4: bset #30,%d0
639 jne 2b
640 fp_set_sr FPSR_EXC_SNAN
641 move.l %d0,(-4,%a0)
642 jra 2b
644 | these next two do rounding as per the IEEE standard.
645 | values for the rounding modes appear to be:
646 | 0: Round to nearest
647 | 1: Round to zero
648 | 2: Round to -Infinity
649 | 3: Round to +Infinity
650 | both functions expect that fp_normalize was already
651 | called (and extended argument is already normalized
652 | as far as possible), these are used if there is different
653 | rounding precision is selected and before converting
654 | into single/double
656 | fp_normalize_double:
657 | normalize an extended with double (52-bit) precision
658 | args: %a0 (struct fp_ext *)
660 fp_normalize_double:
661 printf PNORM,"nd: %p(",1,%a0
662 printx PNORM,%a0@
663 printf PNORM,"), "
664 move.l (%a0)+,%d2
665 tst.w %d2
666 jeq fp_nd_zero | zero / denormalized
667 cmp.w #0x7fff,%d2
668 jeq fp_nd_huge | NaN / infinitive.
669 sub.w #0x4000-0x3ff,%d2 | will the exponent fit?
670 jcs fp_nd_small | too small.
671 cmp.w #0x7fe,%d2
672 jcc fp_nd_large | too big.
673 addq.l #4,%a0
674 move.l (%a0),%d0 | low lword of mantissa
675 | now, round off the low 11 bits.
676 fp_nd_round:
677 moveq #21,%d1
678 lsl.l %d1,%d0 | keep 11 low bits.
679 jne fp_nd_checkround | Are they non-zero?
680 | nothing to do here
681 9: subq.l #8,%a0
682 printf PNORM,"%p(",1,%a0
683 printx PNORM,%a0@
684 printf PNORM,")\n"
685 rts
686 | Be careful with the X bit! It contains the lsb
687 | from the shift above, it is needed for round to nearest.
688 fp_nd_checkround:
689 fp_set_sr FPSR_EXC_INEX2 | INEX2 bit
690 and.w #0xf800,(2,%a0) | clear bits 0-10
691 move.w (FPD_RND,FPDATA),%d2 | rounding mode
692 jne 2f | %d2 == 0, round to nearest
693 tst.l %d0 | test guard bit
694 jpl 9b | zero is closer
695 | here we test the X bit by adding it to %d2
696 clr.w %d2 | first set z bit, addx only clears it
697 addx.w %d2,%d2 | test lsb bit
698 | IEEE754-specified "round to even" behaviour. If the guard
699 | bit is set, then the number is odd, so rounding works like
700 | in grade-school arithmetic (i.e. 1.5 rounds to 2.0)
701 | Otherwise, an equal distance rounds towards zero, so as not
702 | to produce an odd number. This is strange, but it is what
703 | the standard says.
704 jne fp_nd_doroundup | round to infinity
705 lsl.l #1,%d0 | check low bits
706 jeq 9b | round to zero
707 fp_nd_doroundup:
708 | round (the mantissa, that is) towards infinity
709 add.l #0x800,(%a0)
710 jcc 9b | no overflow, good.
711 addq.l #1,-(%a0) | extend to high lword
712 jcc 1f | no overflow, good.
713 | Yow! we have managed to overflow the mantissa. Since this
714 | only happens when %d1 was 0xfffff800, it is now zero, so
715 | reset the high bit, and increment the exponent.
716 move.w #0x8000,(%a0)
717 addq.w #1,-(%a0)
718 cmp.w #0x43ff,(%a0)+ | exponent now overflown?
719 jeq fp_nd_large | yes, so make it infinity.
720 1: subq.l #4,%a0
721 printf PNORM,"%p(",1,%a0
722 printx PNORM,%a0@
723 printf PNORM,")\n"
724 rts
725 2: subq.w #2,%d2
726 jcs 9b | %d2 < 2, round to zero
727 jhi 3f | %d2 > 2, round to +infinity
728 | Round to +Inf or -Inf. High word of %d2 contains the
729 | sign of the number, by the way.
730 swap %d2 | to -inf
731 tst.b %d2
732 jne fp_nd_doroundup | negative, round to infinity
733 jra 9b | positive, round to zero
734 3: swap %d2 | to +inf
735 tst.b %d2
736 jeq fp_nd_doroundup | positive, round to infinity
737 jra 9b | negative, round to zero
738 | Exponent underflow. Try to make a denormal, and set it to
739 | the smallest possible fraction if this fails.
740 fp_nd_small:
741 fp_set_sr FPSR_EXC_UNFL | set UNFL bit
742 move.w #0x3c01,(-2,%a0) | 2**-1022
743 neg.w %d2 | degree of underflow
744 cmp.w #32,%d2 | single or double shift?
745 jcc 1f
746 | Again, another 64-bit double shift.
747 move.l (%a0),%d0
748 move.l %d0,%d1
749 lsr.l %d2,%d0
750 move.l %d0,(%a0)+
751 move.l (%a0),%d0
752 lsr.l %d2,%d0
753 neg.w %d2
754 add.w #32,%d2
755 lsl.l %d2,%d1
756 or.l %d1,%d0
757 move.l (%a0),%d1
758 move.l %d0,(%a0)
759 | Check to see if we shifted off any significant bits
760 lsl.l %d2,%d1
761 jeq fp_nd_round | Nope, round.
762 bset #0,%d0 | Yes, so set the "sticky bit".
763 jra fp_nd_round | Now, round.
764 | Another 64-bit single shift and store
765 1: sub.w #32,%d2
766 cmp.w #32,%d2 | Do we really need to shift?
767 jcc 2f | No, the number is too small.
768 move.l (%a0),%d0
769 clr.l (%a0)+
770 move.l %d0,%d1
771 lsr.l %d2,%d0
772 neg.w %d2
773 add.w #32,%d2
774 | Again, check to see if we shifted off any significant bits.
775 tst.l (%a0)
776 jeq 1f
777 bset #0,%d0 | Sticky bit.
778 1: move.l %d0,(%a0)
779 lsl.l %d2,%d1
780 jeq fp_nd_round
781 bset #0,%d0
782 jra fp_nd_round
783 | Sorry, the number is just too small.
784 2: clr.l (%a0)+
785 clr.l (%a0)
786 moveq #1,%d0 | Smallest possible fraction,
787 jra fp_nd_round | round as desired.
788 | zero and denormalized
789 fp_nd_zero:
790 tst.l (%a0)+
791 jne 1f
792 tst.l (%a0)
793 jne 1f
794 subq.l #8,%a0
795 printf PNORM,"%p(",1,%a0
796 printx PNORM,%a0@
797 printf PNORM,")\n"
798 rts | zero. nothing to do.
799 | These are not merely subnormal numbers, but true denormals,
800 | i.e. pathologically small (exponent is 2**-16383) numbers.
801 | It is clearly impossible for even a normal extended number
802 | with that exponent to fit into double precision, so just
803 | write these ones off as "too darn small".
804 1: fp_set_sr FPSR_EXC_UNFL | Set UNFL bit
805 clr.l (%a0)
806 clr.l -(%a0)
807 move.w #0x3c01,-(%a0) | i.e. 2**-1022
808 addq.l #6,%a0
809 moveq #1,%d0
810 jra fp_nd_round | round.
811 | Exponent overflow. Just call it infinity.
812 fp_nd_large:
813 move.w #0x7ff,%d0
814 and.w (6,%a0),%d0
815 jeq 1f
816 fp_set_sr FPSR_EXC_INEX2
817 1: fp_set_sr FPSR_EXC_OVFL
818 move.w (FPD_RND,FPDATA),%d2
819 jne 3f | %d2 = 0 round to nearest
820 1: move.w #0x7fff,(-2,%a0)
821 clr.l (%a0)+
822 clr.l (%a0)
823 2: subq.l #8,%a0
824 printf PNORM,"%p(",1,%a0
825 printx PNORM,%a0@
826 printf PNORM,")\n"
827 rts
828 3: subq.w #2,%d2
829 jcs 5f | %d2 < 2, round to zero
830 jhi 4f | %d2 > 2, round to +infinity
831 tst.b (-3,%a0) | to -inf
832 jne 1b
833 jra 5f
834 4: tst.b (-3,%a0) | to +inf
835 jeq 1b
836 5: move.w #0x43fe,(-2,%a0)
837 moveq #-1,%d0
838 move.l %d0,(%a0)+
839 move.w #0xf800,%d0
840 move.l %d0,(%a0)
841 jra 2b
842 | Infinities or NaNs
843 fp_nd_huge:
844 subq.l #4,%a0
845 printf PNORM,"%p(",1,%a0
846 printx PNORM,%a0@
847 printf PNORM,")\n"
848 rts
850 | fp_normalize_single:
851 | normalize an extended with single (23-bit) precision
852 | args: %a0 (struct fp_ext *)
854 fp_normalize_single:
855 printf PNORM,"ns: %p(",1,%a0
856 printx PNORM,%a0@
857 printf PNORM,") "
858 addq.l #2,%a0
859 move.w (%a0)+,%d2
860 jeq fp_ns_zero | zero / denormalized
861 cmp.w #0x7fff,%d2
862 jeq fp_ns_huge | NaN / infinitive.
863 sub.w #0x4000-0x7f,%d2 | will the exponent fit?
864 jcs fp_ns_small | too small.
865 cmp.w #0xfe,%d2
866 jcc fp_ns_large | too big.
867 move.l (%a0)+,%d0 | get high lword of mantissa
868 fp_ns_round:
869 tst.l (%a0) | check the low lword
870 jeq 1f
871 | Set a sticky bit if it is non-zero. This should only
872 | affect the rounding in what would otherwise be equal-
873 | distance situations, which is what we want it to do.
874 bset #0,%d0
875 1: clr.l (%a0) | zap it from memory.
876 | now, round off the low 8 bits of the hi lword.
877 tst.b %d0 | 8 low bits.
878 jne fp_ns_checkround | Are they non-zero?
879 | nothing to do here
880 subq.l #8,%a0
881 printf PNORM,"%p(",1,%a0
882 printx PNORM,%a0@
883 printf PNORM,")\n"
884 rts
885 fp_ns_checkround:
886 fp_set_sr FPSR_EXC_INEX2 | INEX2 bit
887 clr.b -(%a0) | clear low byte of high lword
888 subq.l #3,%a0
889 move.w (FPD_RND,FPDATA),%d2 | rounding mode
890 jne 2f | %d2 == 0, round to nearest
891 tst.b %d0 | test guard bit
892 jpl 9f | zero is closer
893 btst #8,%d0 | test lsb bit
894 | round to even behaviour, see above.
895 jne fp_ns_doroundup | round to infinity
896 lsl.b #1,%d0 | check low bits
897 jeq 9f | round to zero
898 fp_ns_doroundup:
899 | round (the mantissa, that is) towards infinity
900 add.l #0x100,(%a0)
901 jcc 9f | no overflow, good.
902 | Overflow. This means that the %d1 was 0xffffff00, so it
903 | is now zero. We will set the mantissa to reflect this, and
904 | increment the exponent (checking for overflow there too)
905 move.w #0x8000,(%a0)
906 addq.w #1,-(%a0)
907 cmp.w #0x407f,(%a0)+ | exponent now overflown?
908 jeq fp_ns_large | yes, so make it infinity.
909 9: subq.l #4,%a0
910 printf PNORM,"%p(",1,%a0
911 printx PNORM,%a0@
912 printf PNORM,")\n"
913 rts
914 | check nondefault rounding modes
915 2: subq.w #2,%d2
916 jcs 9b | %d2 < 2, round to zero
917 jhi 3f | %d2 > 2, round to +infinity
918 tst.b (-3,%a0) | to -inf
919 jne fp_ns_doroundup | negative, round to infinity
920 jra 9b | positive, round to zero
921 3: tst.b (-3,%a0) | to +inf
922 jeq fp_ns_doroundup | positive, round to infinity
923 jra 9b | negative, round to zero
924 | Exponent underflow. Try to make a denormal, and set it to
925 | the smallest possible fraction if this fails.
926 fp_ns_small:
927 fp_set_sr FPSR_EXC_UNFL | set UNFL bit
928 move.w #0x3f81,(-2,%a0) | 2**-126
929 neg.w %d2 | degree of underflow
930 cmp.w #32,%d2 | single or double shift?
931 jcc 2f
932 | a 32-bit shift.
933 move.l (%a0),%d0
934 move.l %d0,%d1
935 lsr.l %d2,%d0
936 move.l %d0,(%a0)+
937 | Check to see if we shifted off any significant bits.
938 neg.w %d2
939 add.w #32,%d2
940 lsl.l %d2,%d1
941 jeq 1f
942 bset #0,%d0 | Sticky bit.
943 | Check the lower lword
944 1: tst.l (%a0)
945 jeq fp_ns_round
946 clr (%a0)
947 bset #0,%d0 | Sticky bit.
948 jra fp_ns_round
949 | Sorry, the number is just too small.
950 2: clr.l (%a0)+
951 clr.l (%a0)
952 moveq #1,%d0 | Smallest possible fraction,
953 jra fp_ns_round | round as desired.
954 | Exponent overflow. Just call it infinity.
955 fp_ns_large:
956 tst.b (3,%a0)
957 jeq 1f
958 fp_set_sr FPSR_EXC_INEX2
959 1: fp_set_sr FPSR_EXC_OVFL
960 move.w (FPD_RND,FPDATA),%d2
961 jne 3f | %d2 = 0 round to nearest
962 1: move.w #0x7fff,(-2,%a0)
963 clr.l (%a0)+
964 clr.l (%a0)
965 2: subq.l #8,%a0
966 printf PNORM,"%p(",1,%a0
967 printx PNORM,%a0@
968 printf PNORM,")\n"
969 rts
970 3: subq.w #2,%d2
971 jcs 5f | %d2 < 2, round to zero
972 jhi 4f | %d2 > 2, round to +infinity
973 tst.b (-3,%a0) | to -inf
974 jne 1b
975 jra 5f
976 4: tst.b (-3,%a0) | to +inf
977 jeq 1b
978 5: move.w #0x407e,(-2,%a0)
979 move.l #0xffffff00,(%a0)+
980 clr.l (%a0)
981 jra 2b
982 | zero and denormalized
983 fp_ns_zero:
984 tst.l (%a0)+
985 jne 1f
986 tst.l (%a0)
987 jne 1f
988 subq.l #8,%a0
989 printf PNORM,"%p(",1,%a0
990 printx PNORM,%a0@
991 printf PNORM,")\n"
992 rts | zero. nothing to do.
993 | These are not merely subnormal numbers, but true denormals,
994 | i.e. pathologically small (exponent is 2**-16383) numbers.
995 | It is clearly impossible for even a normal extended number
996 | with that exponent to fit into single precision, so just
997 | write these ones off as "too darn small".
998 1: fp_set_sr FPSR_EXC_UNFL | Set UNFL bit
999 clr.l (%a0)
1000 clr.l -(%a0)
1001 move.w #0x3f81,-(%a0) | i.e. 2**-126
1002 addq.l #6,%a0
1003 moveq #1,%d0
1004 jra fp_ns_round | round.
1005 | Infinities or NaNs
1006 fp_ns_huge:
1007 subq.l #4,%a0
1008 printf PNORM,"%p(",1,%a0
1009 printx PNORM,%a0@
1010 printf PNORM,")\n"
1011 rts
1013 | fp_normalize_single_fast:
1014 | normalize an extended with single (23-bit) precision
1015 | this is only used by fsgldiv/fsgdlmul, where the
1016 | operand is not completly normalized.
1017 | args: %a0 (struct fp_ext *)
1019 fp_normalize_single_fast:
1020 printf PNORM,"nsf: %p(",1,%a0
1021 printx PNORM,%a0@
1022 printf PNORM,") "
1023 addq.l #2,%a0
1024 move.w (%a0)+,%d2
1025 cmp.w #0x7fff,%d2
1026 jeq fp_nsf_huge | NaN / infinitive.
1027 move.l (%a0)+,%d0 | get high lword of mantissa
1028 fp_nsf_round:
1029 tst.l (%a0) | check the low lword
1030 jeq 1f
1031 | Set a sticky bit if it is non-zero. This should only
1032 | affect the rounding in what would otherwise be equal-
1033 | distance situations, which is what we want it to do.
1034 bset #0,%d0
1035 1: clr.l (%a0) | zap it from memory.
1036 | now, round off the low 8 bits of the hi lword.
1037 tst.b %d0 | 8 low bits.
1038 jne fp_nsf_checkround | Are they non-zero?
1039 | nothing to do here
1040 subq.l #8,%a0
1041 printf PNORM,"%p(",1,%a0
1042 printx PNORM,%a0@
1043 printf PNORM,")\n"
1044 rts
1045 fp_nsf_checkround:
1046 fp_set_sr FPSR_EXC_INEX2 | INEX2 bit
1047 clr.b -(%a0) | clear low byte of high lword
1048 subq.l #3,%a0
1049 move.w (FPD_RND,FPDATA),%d2 | rounding mode
1050 jne 2f | %d2 == 0, round to nearest
1051 tst.b %d0 | test guard bit
1052 jpl 9f | zero is closer
1053 btst #8,%d0 | test lsb bit
1054 | round to even behaviour, see above.
1055 jne fp_nsf_doroundup | round to infinity
1056 lsl.b #1,%d0 | check low bits
1057 jeq 9f | round to zero
1058 fp_nsf_doroundup:
1059 | round (the mantissa, that is) towards infinity
1060 add.l #0x100,(%a0)
1061 jcc 9f | no overflow, good.
1062 | Overflow. This means that the %d1 was 0xffffff00, so it
1063 | is now zero. We will set the mantissa to reflect this, and
1064 | increment the exponent (checking for overflow there too)
1065 move.w #0x8000,(%a0)
1066 addq.w #1,-(%a0)
1067 cmp.w #0x407f,(%a0)+ | exponent now overflown?
1068 jeq fp_nsf_large | yes, so make it infinity.
1069 9: subq.l #4,%a0
1070 printf PNORM,"%p(",1,%a0
1071 printx PNORM,%a0@
1072 printf PNORM,")\n"
1073 rts
1074 | check nondefault rounding modes
1075 2: subq.w #2,%d2
1076 jcs 9b | %d2 < 2, round to zero
1077 jhi 3f | %d2 > 2, round to +infinity
1078 tst.b (-3,%a0) | to -inf
1079 jne fp_nsf_doroundup | negative, round to infinity
1080 jra 9b | positive, round to zero
1081 3: tst.b (-3,%a0) | to +inf
1082 jeq fp_nsf_doroundup | positive, round to infinity
1083 jra 9b | negative, round to zero
1084 | Exponent overflow. Just call it infinity.
1085 fp_nsf_large:
1086 tst.b (3,%a0)
1087 jeq 1f
1088 fp_set_sr FPSR_EXC_INEX2
1089 1: fp_set_sr FPSR_EXC_OVFL
1090 move.w (FPD_RND,FPDATA),%d2
1091 jne 3f | %d2 = 0 round to nearest
1092 1: move.w #0x7fff,(-2,%a0)
1093 clr.l (%a0)+
1094 clr.l (%a0)
1095 2: subq.l #8,%a0
1096 printf PNORM,"%p(",1,%a0
1097 printx PNORM,%a0@
1098 printf PNORM,")\n"
1099 rts
1100 3: subq.w #2,%d2
1101 jcs 5f | %d2 < 2, round to zero
1102 jhi 4f | %d2 > 2, round to +infinity
1103 tst.b (-3,%a0) | to -inf
1104 jne 1b
1105 jra 5f
1106 4: tst.b (-3,%a0) | to +inf
1107 jeq 1b
1108 5: move.w #0x407e,(-2,%a0)
1109 move.l #0xffffff00,(%a0)+
1110 clr.l (%a0)
1111 jra 2b
1112 | Infinities or NaNs
1113 fp_nsf_huge:
1114 subq.l #4,%a0
1115 printf PNORM,"%p(",1,%a0
1116 printx PNORM,%a0@
1117 printf PNORM,")\n"
1118 rts
1120 | conv_ext2int (macro):
1121 | Generates a subroutine that converts an extended value to an
1122 | integer of a given size, again, with the appropriate type of
1123 | rounding.
1125 | Macro arguments:
1126 | s: size, as given in an assembly instruction.
1127 | b: number of bits in that size.
1129 | Subroutine arguments:
1130 | %a0: source (struct fp_ext *)
1132 | Returns the integer in %d0 (like it should)
1134 .macro conv_ext2int s,b
1135 .set inf,(1<<(\b-1))-1 | i.e. MAXINT
1136 printf PCONV,"e2i%d: %p(",2,#\b,%a0
1137 printx PCONV,%a0@
1138 printf PCONV,") "
1139 addq.l #2,%a0
1140 move.w (%a0)+,%d2 | exponent
1141 jeq fp_e2i_zero\b | zero / denorm (== 0, here)
1142 cmp.w #0x7fff,%d2
1143 jeq fp_e2i_huge\b | Inf / NaN
1144 sub.w #0x3ffe,%d2
1145 jcs fp_e2i_small\b
1146 cmp.w #\b,%d2
1147 jhi fp_e2i_large\b
1148 move.l (%a0),%d0
1149 move.l %d0,%d1
1150 lsl.l %d2,%d1
1151 jne fp_e2i_round\b
1152 tst.l (4,%a0)
1153 jne fp_e2i_round\b
1154 neg.w %d2
1155 add.w #32,%d2
1156 lsr.l %d2,%d0
1157 9: tst.w (-4,%a0)
1158 jne 1f
1159 tst.\s %d0
1160 jmi fp_e2i_large\b
1161 printf PCONV,"-> %p\n",1,%d0
1162 rts
1163 1: neg.\s %d0
1164 jeq 1f
1165 jpl fp_e2i_large\b
1166 1: printf PCONV,"-> %p\n",1,%d0
1167 rts
1168 fp_e2i_round\b:
1169 fp_set_sr FPSR_EXC_INEX2 | INEX2 bit
1170 neg.w %d2
1171 add.w #32,%d2
1172 .if \b>16
1173 jeq 5f
1174 .endif
1175 lsr.l %d2,%d0
1176 move.w (FPD_RND,FPDATA),%d2 | rounding mode
1177 jne 2f | %d2 == 0, round to nearest
1178 tst.l %d1 | test guard bit
1179 jpl 9b | zero is closer
1180 btst %d2,%d0 | test lsb bit (%d2 still 0)
1181 jne fp_e2i_doroundup\b
1182 lsl.l #1,%d1 | check low bits
1183 jne fp_e2i_doroundup\b
1184 tst.l (4,%a0)
1185 jeq 9b
1186 fp_e2i_doroundup\b:
1187 addq.l #1,%d0
1188 jra 9b
1189 | check nondefault rounding modes
1190 2: subq.w #2,%d2
1191 jcs 9b | %d2 < 2, round to zero
1192 jhi 3f | %d2 > 2, round to +infinity
1193 tst.w (-4,%a0) | to -inf
1194 jne fp_e2i_doroundup\b | negative, round to infinity
1195 jra 9b | positive, round to zero
1196 3: tst.w (-4,%a0) | to +inf
1197 jeq fp_e2i_doroundup\b | positive, round to infinity
1198 jra 9b | negative, round to zero
1199 | we are only want -2**127 get correctly rounded here,
1200 | since the guard bit is in the lower lword.
1201 | everything else ends up anyway as overflow.
1202 .if \b>16
1203 5: move.w (FPD_RND,FPDATA),%d2 | rounding mode
1204 jne 2b | %d2 == 0, round to nearest
1205 move.l (4,%a0),%d1 | test guard bit
1206 jpl 9b | zero is closer
1207 lsl.l #1,%d1 | check low bits
1208 jne fp_e2i_doroundup\b
1209 jra 9b
1210 .endif
1211 fp_e2i_zero\b:
1212 clr.l %d0
1213 tst.l (%a0)+
1214 jne 1f
1215 tst.l (%a0)
1216 jeq 3f
1217 1: subq.l #4,%a0
1218 fp_clr_sr FPSR_EXC_UNFL | fp_normalize_ext has set this bit
1219 fp_e2i_small\b:
1220 fp_set_sr FPSR_EXC_INEX2
1221 clr.l %d0
1222 move.w (FPD_RND,FPDATA),%d2 | rounding mode
1223 subq.w #2,%d2
1224 jcs 3f | %d2 < 2, round to nearest/zero
1225 jhi 2f | %d2 > 2, round to +infinity
1226 tst.w (-4,%a0) | to -inf
1227 jeq 3f
1228 subq.\s #1,%d0
1229 jra 3f
1230 2: tst.w (-4,%a0) | to +inf
1231 jne 3f
1232 addq.\s #1,%d0
1233 3: printf PCONV,"-> %p\n",1,%d0
1234 rts
1235 fp_e2i_large\b:
1236 fp_set_sr FPSR_EXC_OPERR
1237 move.\s #inf,%d0
1238 tst.w (-4,%a0)
1239 jeq 1f
1240 addq.\s #1,%d0
1241 1: printf PCONV,"-> %p\n",1,%d0
1242 rts
1243 fp_e2i_huge\b:
1244 move.\s (%a0),%d0
1245 tst.l (%a0)
1246 jne 1f
1247 tst.l (%a0)
1248 jeq fp_e2i_large\b
1249 | fp_normalize_ext has set this bit already
1250 | and made the number nonsignaling
1251 1: fp_tst_sr FPSR_EXC_SNAN
1252 jne 1f
1253 fp_set_sr FPSR_EXC_OPERR
1254 1: printf PCONV,"-> %p\n",1,%d0
1255 rts
1256 .endm
1258 fp_conv_ext2long:
1259 conv_ext2int l,32
1261 fp_conv_ext2short:
1262 conv_ext2int w,16
1264 fp_conv_ext2byte:
1265 conv_ext2int b,8
1267 fp_conv_ext2double:
1268 jsr fp_normalize_double
1269 printf PCONV,"e2d: %p(",1,%a0
1270 printx PCONV,%a0@
1271 printf PCONV,"), "
1272 move.l (%a0)+,%d2
1273 cmp.w #0x7fff,%d2
1274 jne 1f
1275 move.w #0x7ff,%d2
1276 move.l (%a0)+,%d0
1277 jra 2f
1278 1: sub.w #0x3fff-0x3ff,%d2
1279 move.l (%a0)+,%d0
1280 jmi 2f
1281 clr.w %d2
1282 2: lsl.w #5,%d2
1283 lsl.l #7,%d2
1284 lsl.l #8,%d2
1285 move.l %d0,%d1
1286 lsl.l #1,%d0
1287 lsr.l #4,%d0
1288 lsr.l #8,%d0
1289 or.l %d2,%d0
1290 putuser.l %d0,(%a1)+,fp_err_ua2,%a1
1291 moveq #21,%d0
1292 lsl.l %d0,%d1
1293 move.l (%a0),%d0
1294 lsr.l #4,%d0
1295 lsr.l #7,%d0
1296 or.l %d1,%d0
1297 putuser.l %d0,(%a1),fp_err_ua2,%a1
1298 #ifdef FPU_EMU_DEBUG
1299 getuser.l %a1@(-4),%d0,fp_err_ua2,%a1
1300 getuser.l %a1@(0),%d1,fp_err_ua2,%a1
1301 printf PCONV,"%p(%08x%08x)\n",3,%a1,%d0,%d1
1302 #endif
1303 rts
1305 fp_conv_ext2single:
1306 jsr fp_normalize_single
1307 printf PCONV,"e2s: %p(",1,%a0
1308 printx PCONV,%a0@
1309 printf PCONV,"), "
1310 move.l (%a0)+,%d1
1311 cmp.w #0x7fff,%d1
1312 jne 1f
1313 move.w #0xff,%d1
1314 move.l (%a0)+,%d0
1315 jra 2f
1316 1: sub.w #0x3fff-0x7f,%d1
1317 move.l (%a0)+,%d0
1318 jmi 2f
1319 clr.w %d1
1320 2: lsl.w #8,%d1
1321 lsl.l #7,%d1
1322 lsl.l #8,%d1
1323 bclr #31,%d0
1324 lsr.l #8,%d0
1325 or.l %d1,%d0
1326 printf PCONV,"%08x\n",1,%d0
1327 rts
1329 | special return addresses for instr that
1330 | encode the rounding precision in the opcode
1331 | (e.g. fsmove,fdmove)
1333 fp_finalrounding_single:
1334 addq.l #8,%sp
1335 jsr fp_normalize_ext
1336 jsr fp_normalize_single
1337 jra fp_finaltest
1339 fp_finalrounding_single_fast:
1340 addq.l #8,%sp
1341 jsr fp_normalize_ext
1342 jsr fp_normalize_single_fast
1343 jra fp_finaltest
1345 fp_finalrounding_double:
1346 addq.l #8,%sp
1347 jsr fp_normalize_ext
1348 jsr fp_normalize_double
1349 jra fp_finaltest
1351 | fp_finaltest:
1352 | set the emulated status register based on the outcome of an
1353 | emulated instruction.
1355 fp_finalrounding:
1356 addq.l #8,%sp
1357 | printf ,"f: %p\n",1,%a0
1358 jsr fp_normalize_ext
1359 move.w (FPD_PREC,FPDATA),%d0
1360 subq.w #1,%d0
1361 jcs fp_finaltest
1362 jne 1f
1363 jsr fp_normalize_single
1364 jra 2f
1365 1: jsr fp_normalize_double
1366 2:| printf ,"f: %p\n",1,%a0
1367 fp_finaltest:
1368 | First, we do some of the obvious tests for the exception
1369 | status byte and condition code bytes of fp_sr here, so that
1370 | they do not have to be handled individually by every
1371 | emulated instruction.
1372 clr.l %d0
1373 addq.l #1,%a0
1374 tst.b (%a0)+ | sign
1375 jeq 1f
1376 bset #FPSR_CC_NEG-24,%d0 | N bit
1377 1: cmp.w #0x7fff,(%a0)+ | exponent
1378 jeq 2f
1379 | test for zero
1380 moveq #FPSR_CC_Z-24,%d1
1381 tst.l (%a0)+
1382 jne 9f
1383 tst.l (%a0)
1384 jne 9f
1385 jra 8f
1386 | infinitiv and NAN
1387 2: moveq #FPSR_CC_NAN-24,%d1
1388 move.l (%a0)+,%d2
1389 lsl.l #1,%d2 | ignore high bit
1390 jne 8f
1391 tst.l (%a0)
1392 jne 8f
1393 moveq #FPSR_CC_INF-24,%d1
1394 8: bset %d1,%d0
1395 9: move.b %d0,(FPD_FPSR+0,FPDATA) | set condition test result
1396 | move instructions enter here
1397 | Here, we test things in the exception status byte, and set
1398 | other things in the accrued exception byte accordingly.
1399 | Emulated instructions can set various things in the former,
1400 | as defined in fp_emu.h.
1401 fp_final:
1402 move.l (FPD_FPSR,FPDATA),%d0
1403 #if 0
1404 btst #FPSR_EXC_SNAN,%d0 | EXC_SNAN
1405 jne 1f
1406 btst #FPSR_EXC_OPERR,%d0 | EXC_OPERR
1407 jeq 2f
1408 1: bset #FPSR_AEXC_IOP,%d0 | set IOP bit
1409 2: btst #FPSR_EXC_OVFL,%d0 | EXC_OVFL
1410 jeq 1f
1411 bset #FPSR_AEXC_OVFL,%d0 | set OVFL bit
1412 1: btst #FPSR_EXC_UNFL,%d0 | EXC_UNFL
1413 jeq 1f
1414 btst #FPSR_EXC_INEX2,%d0 | EXC_INEX2
1415 jeq 1f
1416 bset #FPSR_AEXC_UNFL,%d0 | set UNFL bit
1417 1: btst #FPSR_EXC_DZ,%d0 | EXC_INEX1
1418 jeq 1f
1419 bset #FPSR_AEXC_DZ,%d0 | set DZ bit
1420 1: btst #FPSR_EXC_OVFL,%d0 | EXC_OVFL
1421 jne 1f
1422 btst #FPSR_EXC_INEX2,%d0 | EXC_INEX2
1423 jne 1f
1424 btst #FPSR_EXC_INEX1,%d0 | EXC_INEX1
1425 jeq 2f
1426 1: bset #FPSR_AEXC_INEX,%d0 | set INEX bit
1427 2: move.l %d0,(FPD_FPSR,FPDATA)
1428 #else
1429 | same as above, greatly optimized, but untested (yet)
1430 move.l %d0,%d2
1431 lsr.l #5,%d0
1432 move.l %d0,%d1
1433 lsr.l #4,%d1
1434 or.l %d0,%d1
1435 and.b #0x08,%d1
1436 move.l %d2,%d0
1437 lsr.l #6,%d0
1438 or.l %d1,%d0
1439 move.l %d2,%d1
1440 lsr.l #4,%d1
1441 or.b #0xdf,%d1
1442 and.b %d1,%d0
1443 move.l %d2,%d1
1444 lsr.l #7,%d1
1445 and.b #0x80,%d1
1446 or.b %d1,%d0
1447 and.b #0xf8,%d0
1448 or.b %d0,%d2
1449 move.l %d2,(FPD_FPSR,FPDATA)
1450 #endif
1451 move.b (FPD_FPSR+2,FPDATA),%d0
1452 and.b (FPD_FPCR+2,FPDATA),%d0
1453 jeq 1f
1454 printf ,"send signal!!!\n"
1455 1: jra fp_end