]> git.ipfire.org Git - thirdparty/gcc.git/blame - libgcc/config/arc/lib1funcs.S
projects / thirdparty / gcc.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Update copyright years.
[thirdparty/gcc.git] / libgcc / config / arc / lib1funcs.S
CommitLineData
d38a64b4
JR		 1; libgcc1 routines for Synopsys DesignWare ARC cpu.
2
a5544970 3/* Copyright (C) 1995-2019 Free Software Foundation, Inc.
d38a64b4
JR		 4 Contributor: Joern Rennecke <joern.rennecke@embecosm.com>
5 on behalf of Synopsys Inc.
6
7This file is part of GCC.
8
9GCC is free software; you can redistribute it and/or modify it under
10the terms of the GNU General Public License as published by the Free
11Software Foundation; either version 3, or (at your option) any later
12version.
13
14GCC is distributed in the hope that it will be useful, but WITHOUT ANY
15WARRANTY; without even the implied warranty of MERCHANTABILITY or
16FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
17for more details.
18
19Under Section 7 of GPL version 3, you are granted additional
20permissions described in the GCC Runtime Library Exception, version
213.1, as published by the Free Software Foundation.
22
23You should have received a copy of the GNU General Public License and
24a copy of the GCC Runtime Library Exception along with this program;
25see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
26<http://www.gnu.org/licenses/>. */
27
28/* As a special exception, if you link this library with other files,
29 some of which are compiled with GCC, to produce an executable,
30 this library does not by itself cause the resulting executable
31 to be covered by the GNU General Public License.
32 This exception does not however invalidate any other reasons why
33 the executable file might be covered by the GNU General Public License. */
34
48c842ab 35
d38a64b4 36 /* ANSI concatenation macros. */
48c842ab 37
d38a64b4
JR		 38 #define CONCAT1(a, b) CONCAT2(a, b)
39 #define CONCAT2(a, b) a ## b
48c842ab 40
d38a64b4 41 /* Use the right prefix for global labels. */
48c842ab 42
d38a64b4 43 #define SYM(x) CONCAT1 (__USER_LABEL_PREFIX__, x)
48c842ab 44
d38a64b4
JR		 45#ifndef WORKING_ASSEMBLER
46#define abs_l abs
47#define asl_l asl
48#define mov_l mov
49#endif
48c842ab 50
d38a64b4
JR		 51#define FUNC(X) .type SYM(X),@function
52#define HIDDEN_FUNC(X) FUNC(X)` .hidden X
53#define ENDFUNC0(X) .Lfe_##X: .size X,.Lfe_##X-X
54#define ENDFUNC(X) ENDFUNC0(X)
55
48c842ab
CZ		 56
57
d38a64b4
JR		 58#ifdef L_mulsi3
59 .section .text
60 .align 4
61
62 .global SYM(__mulsi3)
63SYM(__mulsi3):
64
65/* This the simple version.
66
48c842ab 67 while (a)
d38a64b4
JR		 68 {
69 if (a & 1)
48c842ab 70 r += b;
d38a64b4
JR		 71 a >>= 1;
72 b <<= 1;
73 }
74*/
75
76#if defined (__ARC_MUL64__)
77 FUNC(__mulsi3)
78 mulu64 r0,r1
79 j_s.d [blink]
80 mov_s r0,mlo
81 ENDFUNC(__mulsi3)
d64af69f 82#elif defined (__ARC_MPY__)
d38a64b4
JR		 83 HIDDEN_FUNC(__mulsi3)
84 mpyu r0,r0,r1
85 nop_s
86 j_s [blink]
87 ENDFUNC(__mulsi3)
88#elif defined (__ARC_NORM__)
89 FUNC(__mulsi3)
90 norm.f r2,r0
91 rsub lp_count,r2,31
92 mov.mi lp_count,32
93 mov_s r2,r0
94 mov_s r0,0
95 lpnz @.Lend ; loop is aligned
96 lsr.f r2,r2
97 add.cs r0,r0,r1
98 add_s r1,r1,r1
99.Lend: j_s [blink]
100 ENDFUNC(__mulsi3)
d64af69f 101#elif !defined (__OPTIMIZE_SIZE__) && defined (__ARC_BARREL_SHIFTER__)
d38a64b4
JR		 102 /* Up to 3.5 times faster than the simpler code below, but larger. */
103 FUNC(__mulsi3)
104 ror.f r2,r0,4
105 mov_s r0,0
106 add3.mi r0,r0,r1
107 asl.f r2,r2,2
108 add2.cs r0,r0,r1
109 jeq_s [blink]
110.Loop:
111 add1.mi r0,r0,r1
112 asl.f r2,r2,2
113 add.cs r0,r0,r1
114 asl_s r1,r1,4
115 ror.f r2,r2,8
116 add3.mi r0,r0,r1
117 asl.f r2,r2,2
118 bne.d .Loop
119 add2.cs r0,r0,r1
120 j_s [blink]
121 ENDFUNC(__mulsi3)
122#elif !defined (__OPTIMIZE_SIZE__) /* __ARC601__ */
123 FUNC(__mulsi3)
124 lsr.f r2,r0
125 mov_s r0,0
126 mov_s r3,0
127 add.cs r0,r0,r1
128.Loop:
129 lsr.f r2,r2
130 add1.cs r0,r0,r1
131 lsr.f r2,r2
132 add2.cs r0,r0,r1
133 lsr.f r2,r2
134 add3.cs r0,r0,r1
48c842ab 135 bne.d .Loop
d38a64b4
JR		 136 add3 r1,r3,r1
137 j_s [blink]
138 ENDFUNC(__mulsi3)
139#else
140/********************************************************/
141 FUNC(__mulsi3)
142 mov_s r2,0 ; Accumulate result here.
143.Lloop:
144 bbit0 r0,0,@.Ly
145 add_s r2,r2,r1 ; r += b
48c842ab 146.Ly:
d38a64b4 147 lsr_s r0,r0 ; a >>= 1
48c842ab
CZ		 148 asl_s r1,r1 ; b <<= 1
149 brne_s r0,0,@.Lloop
d38a64b4
JR		 150.Ldone:
151 j_s.d [blink]
152 mov_s r0,r2
153 ENDFUNC(__mulsi3)
154/********************************************************/
155#endif
48c842ab 156
d38a64b4
JR		 157#endif /* L_mulsi3 */
158
159#ifdef L_umulsidi3
160 .section .text
161 .align 4
162
163 .global SYM(__umulsidi3)
164SYM(__umulsidi3):
165 HIDDEN_FUNC(__umulsidi3)
166/* We need ARC700 /ARC_MUL64 definitions of __umulsidi3 / __umulsi3_highpart
167 in case some code has been compiled without multiply support enabled,
168 but linked with the multiply-support enabled libraries.
169 For ARC601 (i.e. without a barrel shifter), we also use umuldisi3 as our
170 umulsi3_highpart implementation; the use of the latter label doesn't
171 actually benefit ARC601 platforms, but is useful when ARC601 code is linked
172 against other libraries. */
d64af69f
CZ		 173#if defined (__ARC_MPY__) || defined (__ARC_MUL64__) \
174 || !defined (__ARC_BARREL_SHIFTER__)
d38a64b4
JR		 175 .global SYM(__umulsi3_highpart)
176SYM(__umulsi3_highpart):
177 HIDDEN_FUNC(__umulsi3_highpart)
178#endif
179
180/* This the simple version.
181
48c842ab 182 while (a)
d38a64b4
JR		 183 {
184 if (a & 1)
48c842ab 185 r += b;
d38a64b4
JR		 186 a >>= 1;
187 b <<= 1;
188 }
189*/
190#include "ieee-754/arc-ieee-754.h"
191
d64af69f 192#ifdef __ARC_MPY__
d38a64b4
JR		 193 mov_s r12,DBL0L
194 mpyu DBL0L,r12,DBL0H
195 j_s.d [blink]
d64af69f 196 MPYHU DBL0H,r12,DBL0H
d38a64b4
JR		 197#elif defined (__ARC_MUL64__)
198/* Likewise for __ARC_MUL64__ */
199 mulu64 r0,r1
200 mov_s DBL0L,mlo
201 j_s.d [blink]
202 mov_s DBL0H,mhi
d64af69f 203#else /* !__ARC_MPY__ && !__ARC_MUL64__ */
d38a64b4
JR		 204/* Although it might look tempting to extend this to handle muldi3,
205 using mulsi3 twice with 2.25 cycles per 32 bit add is faster
206 than one loop with 3 or four cycles per 32 bit add. */
207 asl.f r12,0 ; Top part of b.
208 mov_s r2,0 ; Accumulate result here.
209 bbit1.d r0,0,@.Ladd
210 mov_s r3,0
211.Llooptst:
212 rlc r12,r12
213 breq r0,0,@.Ldone ; while (a)
214.Lloop:
215 asl.f r1,r1 ; b <<= 1
216 bbit0.d r0,1,@.Llooptst
217 lsr r0,r0 ; a >>= 1
218 rlc r12,r12
219.Ladd:
220 add.f r3,r3,r1 ; r += b
221 brne.d r0,0,@.Lloop ; while (a);
222 adc r2,r2,r12
223.Ldone:
224 mov_s DBL0L,r3
225 j_s.d [blink]
226 mov DBL0H,r2
d64af69f 227#endif /* !__ARC_MPY__*/
d38a64b4 228 ENDFUNC(__umulsidi3)
d64af69f
CZ		 229#if defined (__ARC_MPY__) || defined (__ARC_MUL64__) \
230 || !defined (__ARC_BARREL_SHIFTER__)
d38a64b4
JR		 231 ENDFUNC(__umulsi3_highpart)
232#endif
233#endif /* L_umulsidi3 */
234
8180cde0
CZ		 235#ifdef L_muldi3
236 .section .text
237 .align 4
238 .global SYM(__muldi3)
239SYM(__muldi3):
240#ifdef __LITTLE_ENDIAN__
241 push_s blink
242 mov_s r4,r3 ;4
243 mov_s r5,r2 ;4
244 mov_s r9,r0 ;4
245 mov_s r8,r1 ;4
246 bl.d @__umulsidi3
247 mov_s r1,r2 ;4
248 mov_s r6,r0 ;4
249 mov_s r7,r1 ;4
250 mov_s r0,r9 ;4
251 bl.d @__mulsi3
252 mov_s r1,r4 ;4
253 mov_s r4,r0 ;4
254 mov_s r1,r8 ;4
255 bl.d @__mulsi3
256 mov_s r0,r5 ;4
257 pop_s blink
258 add_s r0,r0,r4 ;2
259 add r1,r0,r7
260 j_s.d [blink]
261 mov_s r0,r6 ;4
262#else
263 push_s blink
264 mov_s r5,r3
265 mov_s r9,r2
266 mov_s r4,r1
267 mov_s r8,r0
268 mov_s r0,r1
269 bl.d @__umulsidi3
270 mov_s r1,r3
271 mov_s r7,r0
272 mov_s r6,r1
273 mov_s r0,r4
274 bl.d @__mulsi3
275 mov_s r1,r9
276 mov_s r4,r0
277 mov_s r1,r8
278 bl.d @__mulsi3
279 mov_s r0,r5
280 pop_s blink
281 add_s r0,r0,r4
282 add_s r0,r0,r7
283 j_s.d [blink]
284 mov_s r1,r6
285#endif /* __LITTLE_ENDIAN__ */
286ENDFUNC(__muldi3)
287#endif /* L_muldi3 */
288
d38a64b4
JR		 289#ifdef L_umulsi3_highpart
290#include "ieee-754/arc-ieee-754.h"
291/* For use without a barrel shifter, and for ARC700 / ARC_MUL64, the
292 mulsidi3 algorithms above look better, so for these, there is an
293 extra label up there. */
d64af69f
CZ		 294#if !defined (__ARC_MPY__) && !defined (__ARC_MUL64__) \
295 && defined (__ARC_BARREL_SHIFTER__)
d38a64b4
JR		 296 .global SYM(__umulsi3_highpart)
297SYM(__umulsi3_highpart):
298 HIDDEN_FUNC(__umulsi3_highpart)
299 mov_s r2,0
300 mov_s r3,32
301.Loop:
302 lsr.f r0,r0
303 add.cs.f r2,r2,r1
304 sub_s r3,r3,1
305 brne.d r0,0,.Loop
306 rrc r2,r2
307 j_s.d [blink]
308/* Make the result register peephole-compatible with mulsidi3. */
309 lsr DBL0H,r2,r3
310 ENDFUNC(__umulsi3_highpart)
d64af69f 311#endif /* !__ARC_MPY__ && __ARC_BARREL_SHIFTER__ */
d38a64b4
JR		 312#endif /* L_umulsi3_highpart */
313
314#ifdef L_divmod_tools
315
316; Utilities used by all routines.
317
318 .section .text
319
320/*
321unsigned long
322udivmodsi4(int modwanted, unsigned long num, unsigned long den)
323{
324 unsigned long bit = 1;
325 unsigned long res = 0;
326
327 while (den < num && bit && !(den & (1L<<31)))
328 {
329 den <<=1;
330 bit <<=1;
331 }
332 while (bit)
333 {
334 if (num >= den)
335 {
336 num -= den;
337 res |= bit;
338 }
339 bit >>=1;
340 den >>=1;
341 }
342 if (modwanted) return num;
343 return res;
344}
345*/
346
347; inputs: r0 = numerator, r1 = denominator
348; outputs: r0 = quotient, r1 = remainder, r2/r3 trashed
349
350 .balign 4
351 .global SYM(__udivmodsi4)
352 FUNC(__udivmodsi4)
353SYM(__udivmodsi4):
354
d64af69f 355#if defined (__ARC_EA__)
d38a64b4
JR		 356/* Normalize divisor and divident, and then use the appropriate number of
357 divaw (the number of result bits, or one more) to produce the result.
358 There are some special conditions that need to be tested:
359 - We can only directly normalize unsigned numbers that fit in 31 bit. For
360 the divisor, we test early on that it is not 'negative'.
361 - divaw can't corrrectly process a divident that is larger than the divisor.
362 We handle this be checking that the divident prior to normalization is
363 not larger than the normalized divisor. As we then already know then
364 that the divisor fits 31 bit, this check also makes sure that the
365 divident fits.
366 - ordinary normalization of the divident could make it larger than the
367 normalized divisor, which again would be unsuitable for divaw.
368 Thus, we want to shift left the divident by one less, except that we
369 want to leave it alone if it is already 31 bit. To this end, we
370 double the input to norm with adds.
371 - If the divident has less bits than the divisor, that would leave us
372 with a negative number of divaw to execute. Although we could use a
373 conditional loop to avoid excess divaw, and then the quotient could
374 be extracted correctly as there'd be more than enough zero bits, the
375 remainder would be shifted left too far, requiring a conditional shift
376 right. The cost of that shift and the possible mispredict on the
377 conditional loop cost as much as putting in an early check for a zero
378 result. */
379 bmsk r3,r0,29
380 brne.d r3,r0,.Large_dividend
381 norm.f r2,r1
382 brlo r0,r1,.Lret0
383 norm r3,r0
384 asl_s r1,r1,r2
385 sub_s r3,r3,1
386 asl_l r0,r0,r3 ; not short to keep loop aligned
387 sub lp_count,r2,r3
388 lp .Ldiv_end
389 divaw r0,r0,r1
390.Ldiv_end:sub_s r3,r2,1
391 lsr r1,r0,r2
392 j_s.d [blink]
393 bmsk r0,r0,r3
394
395 .balign 4
396.Large_dividend:
397 bmi .Ltrivial
398 asl_s r1,r1,r2
399 mov_s r3,0
400 sub1.f r4,r0,r1
401 mov.lo r4,r0
402 mov.hs r3,2
403 cmp r4,r1
404 sub.hs r4,r4,r1
405 add.hs r3,r3,1
406 mov.f lp_count,r2
407 lpne .Ldiv_end2
408 divaw r4,r4,r1
409.Ldiv_end2:asl r0,r3,r2
410 lsr r1,r4,r2
411 sub_s r2,r2,1
412 bmsk r4,r4,r2
413 j_s.d [blink]
414 or.ne r0,r0,r4
415
416.Lret0:
417 mov_s r1,r0
418 j_s.d [blink]
419 mov_l r0,0
420 .balign 4
421.Ltrivial:
422 sub.f r1,r0,r1
423 mov.c r1,r0
424 mov_s r0,1
425 j_s.d [blink]
426 mov.c r0,0
048c6a9a 427#elif !defined (__OPTIMIZE_SIZE__) && !defined (__ARC_RF16__)
d64af69f 428#if defined (__ARC_NORM__) && defined (__ARC_BARREL_SHIFTER__)
d38a64b4
JR		 429 lsr_s r2,r0
430 brhs.d r1,r2,.Lret0_3
431 norm r2,r2
432 norm r3,r1
433 sub_s r3,r3,r2
434 asl_s r1,r1,r3
435 sub1.f 0,r0,r1
436 lsr.cs r1,r1,1
437 sbc r2,r3,0
438 sub1 r0,r0,r1
439 cmp_s r0,r1
440 mov.f lp_count,r2
441#else /* ! __ARC_NORM__ */
442 lsr_s r2,r0
443 brhs.d r1,r2,.Lret0_3
444 mov lp_count,32
445.Lloop1:
446 asl_s r1,r1 ; den <<= 1
447 brls.d r1,r2,@.Lloop1
448 sub lp_count,lp_count,1
449 sub_s r0,r0,r1
450 lsr_s r1,r1
451 cmp_s r0,r1
452 xor.f r2,lp_count,31
d64af69f 453#if !defined (__ARCEM__) && !defined (__ARCHS__)
d38a64b4 454 mov_s lp_count,r2
c0ab1970
CZ		 455#else
456 mov lp_count,r2
457 nop_s
d64af69f 458#endif /* !__ARCEM__ && !__ARCHS__ */
d38a64b4
JR		 459#endif /* !__ARC_NORM__ */
460 sub.cc r0,r0,r1
461 mov_s r3,3
462 sbc r3,r3,0
d64af69f 463#if defined (__ARC_BARREL_SHIFTER__)
d38a64b4
JR		 464 asl_s r3,r3,r2
465 rsub r1,r1,1
466 lpne @.Lloop2_end
467 add1.f r0,r1,r0
468 sub.cc r0,r0,r1
469.Lloop2_end:
470 lsr r1,r0,r2
471#else
472 rsub r1,r1,1
473 lpne @.Lloop2_end
474 asl_s r3,r3
475 add1.f r0,r1,r0
476 sub.cc r0,r0,r1
477.Lloop2_end:
478 lsr_s r1,r0
479 lsr.f lp_count,r2
480 mov.cc r1,r0
481 lpnz 1f
482 lsr_s r1,r1
483 lsr_s r1,r1
4841:
485#endif
486 bmsk r0,r0,r2
487 bclr r0,r0,r2
488 j_s.d [blink]
489 or_s r0,r0,r3
490.Lret0_3:
491#if 0 /* Slightly shorter, but slower. */
492 lp .Loop3_end
493 brhi.d r1,r0,.Loop3_end
494 sub_s r0,r0,r1
495.Loop3_end
496 add_s r1,r1,r0
497 j_s.d [blink]
498 rsub r0,lp_count,32-1
499#else
500 mov_s r4,r1
501 sub.f r1,r0,r1
502 sbc r0,r0,r0
503 sub.cc.f r1,r1,r4
504 sbc r0,r0,0
505 sub.cc.f r1,r1,r4
506 sbc r0,r0,-3
507 j_s.d [blink]
508 add.cs r1,r1,r4
509#endif
510#else /* Arctangent-A5 */
511 breq_s r1,0,@.Ldivmodend
512 mov_s r2,1 ; bit = 1
513 mov_s r3,0 ; res = 0
514.Lloop1:
48c842ab 515 brhs r1,r0,@.Lloop2
d38a64b4
JR		 516 bbit1 r1,31,@.Lloop2
517 asl_s r1,r1 ; den <<= 1
518 b.d @.Lloop1
519 asl_s r2,r2 ; bit <<= 1
520.Lloop2:
48c842ab 521 brlo r0,r1,@.Lshiftdown
d38a64b4
JR		 522 sub_s r0,r0,r1 ; num -= den
523 or_s r3,r3,r2 ; res |= bit
524.Lshiftdown:
525 lsr_s r2,r2 ; bit >>= 1
48c842ab 526 lsr_s r1,r1 ; den >>= 1
d38a64b4
JR		 527 brne_s r2,0,@.Lloop2
528.Ldivmodend:
529 mov_s r1,r0 ; r1 = mod
530 j.d [blink]
531 mov_s r0,r3 ; r0 = res
532/******************************************************/
533#endif
534 ENDFUNC(__udivmodsi4)
535
536#endif
537
538#ifdef L_udivsi3
539 .section .text
540 .align 4
541
542 .global SYM(__udivsi3)
543 FUNC(__udivsi3)
544SYM(__udivsi3):
545 b @SYM(__udivmodsi4)
546 ENDFUNC(__udivsi3)
547#if 0 /* interferes with linux loader */
548 .section .__arc_profile_forward, "a"
549 .long SYM(__udivsi3)
550 .long SYM(__udivmodsi4)
551 .long 65536
552#endif
553
554#endif /* L_udivsi3 */
555
556#ifdef L_divsi3
557 .section .text
558 .align 4
559
560 .global SYM(__divsi3)
561 FUNC(__divsi3)
562
d64af69f 563#ifndef __ARC_EA__
d38a64b4
JR		 564SYM(__divsi3):
565 /* A5 / ARC60? */
048c6a9a
CZ		 566 mov r12,blink
567 xor r11,r0,r1
d38a64b4
JR		 568 abs_s r0,r0
569 bl.d @SYM(__udivmodsi4)
048c6a9a
CZ		 570 abs_s r1,r1
571 tst r11,r11
572 j.d [r12]
573 neg.mi r0,r0
d64af69f 574#else /* !ifndef __ARC_EA__ */
d38a64b4
JR		 575 ;; We can use the abs, norm, divaw and mpy instructions for ARC700
576#define MULDIV
577#ifdef MULDIV
578/* This table has been generated by divtab-arc700.c. */
579/* 1/512 .. 1/256, normalized. There is a leading 1 in bit 31.
580 For powers of two, we list unnormalized numbers instead. The values
581 for powers of 2 are loaded, but not used. The value for 1 is actually
582 the first instruction after .Lmuldiv. */
583 .balign 4
584.Ldivtab:
585
586 .long 0x1000000
587 .long 0x80808081
588 .long 0x81020409
589 .long 0x81848DA9
590 .long 0x82082083
591 .long 0x828CBFBF
592 .long 0x83126E98
593 .long 0x83993053
594 .long 0x84210843
595 .long 0x84A9F9C9
596 .long 0x85340854
597 .long 0x85BF3762
598 .long 0x864B8A7E
599 .long 0x86D90545
600 .long 0x8767AB60
601 .long 0x87F78088
602 .long 0x88888889
603 .long 0x891AC73B
604 .long 0x89AE408A
605 .long 0x8A42F871
606 .long 0x8AD8F2FC
607 .long 0x8B70344B
608 .long 0x8C08C08D
609 .long 0x8CA29C05
610 .long 0x8D3DCB09
611 .long 0x8DDA5203
612 .long 0x8E78356E
613 .long 0x8F1779DA
614 .long 0x8FB823EF
615 .long 0x905A3864
616 .long 0x90FDBC0A
617 .long 0x91A2B3C5
618 .long 0x92492493
619 .long 0x92F11385
620 .long 0x939A85C5
621 .long 0x94458095
622 .long 0x94F20950
623 .long 0x95A02569
624 .long 0x964FDA6D
625 .long 0x97012E03
626 .long 0x97B425EE
627 .long 0x9868C80A
628 .long 0x991F1A52
629 .long 0x99D722DB
630 .long 0x9A90E7DA
631 .long 0x9B4C6F9F
632 .long 0x9C09C09D
633 .long 0x9CC8E161
634 .long 0x9D89D89E
635 .long 0x9E4CAD24
636 .long 0x9F1165E8
637 .long 0x9FD809FE
638 .long 0xA0A0A0A1
639 .long 0xA16B312F
640 .long 0xA237C32C
641 .long 0xA3065E40
642 .long 0xA3D70A3E
643 .long 0xA4A9CF1E
644 .long 0xA57EB503
645 .long 0xA655C43A
646 .long 0xA72F053A
647 .long 0xA80A80A9
648 .long 0xA8E83F58
649 .long 0xA9C84A48
650 .long 0xAAAAAAAB
651 .long 0xAB8F69E3
652 .long 0xAC769185
653 .long 0xAD602B59
654 .long 0xAE4C415D
655 .long 0xAF3ADDC7
656 .long 0xB02C0B03
657 .long 0xB11FD3B9
658 .long 0xB21642C9
659 .long 0xB30F6353
660 .long 0xB40B40B5
661 .long 0xB509E68B
662 .long 0xB60B60B7
663 .long 0xB70FBB5B
664 .long 0xB81702E1
665 .long 0xB92143FB
666 .long 0xBA2E8BA3
667 .long 0xBB3EE722
668 .long 0xBC52640C
669 .long 0xBD691048
670 .long 0xBE82FA0C
671 .long 0xBFA02FE9
672 .long 0xC0C0C0C1
673 .long 0xC1E4BBD6
674 .long 0xC30C30C4
675 .long 0xC4372F86
676 .long 0xC565C87C
677 .long 0xC6980C6A
678 .long 0xC7CE0C7D
679 .long 0xC907DA4F
680 .long 0xCA4587E7
681 .long 0xCB8727C1
682 .long 0xCCCCCCCD
683 .long 0xCE168A78
684 .long 0xCF6474A9
685 .long 0xD0B69FCC
686 .long 0xD20D20D3
687 .long 0xD3680D37
688 .long 0xD4C77B04
689 .long 0xD62B80D7
690 .long 0xD79435E6
691 .long 0xD901B204
692 .long 0xDA740DA8
693 .long 0xDBEB61EF
694 .long 0xDD67C8A7
695 .long 0xDEE95C4D
696 .long 0xE070381D
697 .long 0xE1FC780F
698 .long 0xE38E38E4
699 .long 0xE525982B
700 .long 0xE6C2B449
701 .long 0xE865AC7C
702 .long 0xEA0EA0EB
703 .long 0xEBBDB2A6
704 .long 0xED7303B6
705 .long 0xEF2EB720
706 .long 0xF0F0F0F1
707 .long 0xF2B9D649
708 .long 0xF4898D60
709 .long 0xF6603D99
710 .long 0xF83E0F84
711 .long 0xFA232CF3
712 .long 0xFC0FC0FD
713 .long 0xFE03F810
714 .long 0x2000000
715 .long 0x81020409
716 .long 0x82082083
717 .long 0x83126E98
718 .long 0x84210843
719 .long 0x85340854
720 .long 0x864B8A7E
721 .long 0x8767AB60
722 .long 0x88888889
723 .long 0x89AE408A
724 .long 0x8AD8F2FC
725 .long 0x8C08C08D
726 .long 0x8D3DCB09
727 .long 0x8E78356E
728 .long 0x8FB823EF
729 .long 0x90FDBC0A
730 .long 0x92492493
731 .long 0x939A85C5
732 .long 0x94F20950
733 .long 0x964FDA6D
734 .long 0x97B425EE
735 .long 0x991F1A52
736 .long 0x9A90E7DA
737 .long 0x9C09C09D
738 .long 0x9D89D89E
739 .long 0x9F1165E8
740 .long 0xA0A0A0A1
741 .long 0xA237C32C
742 .long 0xA3D70A3E
743 .long 0xA57EB503
744 .long 0xA72F053A
745 .long 0xA8E83F58
746 .long 0xAAAAAAAB
747 .long 0xAC769185
748 .long 0xAE4C415D
749 .long 0xB02C0B03
750 .long 0xB21642C9
751 .long 0xB40B40B5
752 .long 0xB60B60B7
753 .long 0xB81702E1
754 .long 0xBA2E8BA3
755 .long 0xBC52640C
756 .long 0xBE82FA0C
757 .long 0xC0C0C0C1
758 .long 0xC30C30C4
759 .long 0xC565C87C
760 .long 0xC7CE0C7D
761 .long 0xCA4587E7
762 .long 0xCCCCCCCD
763 .long 0xCF6474A9
764 .long 0xD20D20D3
765 .long 0xD4C77B04
766 .long 0xD79435E6
767 .long 0xDA740DA8
768 .long 0xDD67C8A7
769 .long 0xE070381D
770 .long 0xE38E38E4
771 .long 0xE6C2B449
772 .long 0xEA0EA0EB
773 .long 0xED7303B6
774 .long 0xF0F0F0F1
775 .long 0xF4898D60
776 .long 0xF83E0F84
777 .long 0xFC0FC0FD
778 .long 0x4000000
779 .long 0x82082083
780 .long 0x84210843
781 .long 0x864B8A7E
782 .long 0x88888889
783 .long 0x8AD8F2FC
784 .long 0x8D3DCB09
785 .long 0x8FB823EF
786 .long 0x92492493
787 .long 0x94F20950
788 .long 0x97B425EE
789 .long 0x9A90E7DA
790 .long 0x9D89D89E
791 .long 0xA0A0A0A1
792 .long 0xA3D70A3E
793 .long 0xA72F053A
794 .long 0xAAAAAAAB
795 .long 0xAE4C415D
796 .long 0xB21642C9
797 .long 0xB60B60B7
798 .long 0xBA2E8BA3
799 .long 0xBE82FA0C
800 .long 0xC30C30C4
801 .long 0xC7CE0C7D
802 .long 0xCCCCCCCD
803 .long 0xD20D20D3
804 .long 0xD79435E6
805 .long 0xDD67C8A7
806 .long 0xE38E38E4
807 .long 0xEA0EA0EB
808 .long 0xF0F0F0F1
809 .long 0xF83E0F84
810 .long 0x8000000
811 .long 0x84210843
812 .long 0x88888889
813 .long 0x8D3DCB09
814 .long 0x92492493
815 .long 0x97B425EE
816 .long 0x9D89D89E
817 .long 0xA3D70A3E
818 .long 0xAAAAAAAB
819 .long 0xB21642C9
820 .long 0xBA2E8BA3
821 .long 0xC30C30C4
822 .long 0xCCCCCCCD
823 .long 0xD79435E6
824 .long 0xE38E38E4
825 .long 0xF0F0F0F1
826 .long 0x10000000
827 .long 0x88888889
828 .long 0x92492493
829 .long 0x9D89D89E
830 .long 0xAAAAAAAB
831 .long 0xBA2E8BA3
832 .long 0xCCCCCCCD
833 .long 0xE38E38E4
834 .long 0x20000000
835 .long 0x92492493
836 .long 0xAAAAAAAB
837 .long 0xCCCCCCCD
838 .long 0x40000000
839 .long 0xAAAAAAAB
840 .long 0x80000000
841__muldiv:
842 neg r4,r2
843 ld.as r5,[pcl,r4]
844 abs_s r12,r0
48c842ab
CZ		 845 bic.f 0,r2,r4
846 mpyhu.ne r12,r12,r5
d38a64b4
JR		 847 norm r3,r2
848 xor.f 0,r0,r1
48c842ab
CZ		 849 ; write port allocation stall
850 rsub r3,r3,30
851 lsr r0,r12,r3
852 j_s.d [blink]
853 neg.mi r0,r0
d38a64b4
JR		 854
855 .balign 4
856SYM(__divsi3):
857 norm r3,r1
858 abs_s r2,r1
859 brhs r3,23,__muldiv
860 norm r4,r0
861 abs_l r12,r0
862 brhs r4,r3,.Lonebit
863 asl_s r2,r2,r3
864 asl r12,r12,r4
865 sub lp_count,r3,r4
866 sub.f r12,r12,r2
867 brge.d r12,r2,.Lsbit
868 sub r4,r3,r4
869 add.lo r12,r12,r2
870 lp .Ldivend
871.Ldivstart:divaw r12,r12,r2
872.Ldivend:xor_s r1,r1,r0
873 sub r0,r4,1
874 bmsk r0,r12,r0
875 bset.hs r0,r0,r4
876 tst_s r1,r1
877 j_s.d [blink]
878 neg.mi r0,r0
879.Lonebit:
880 xor_s r1,r1,r0
881 asr_s r1,r1,31
882 sub1.f 0,r12,r2 ; special case: -2**(n+1) / 2**n
883 or r0,r1,1
884 add.eq r0,r0,r0
885 cmp_s r12,r2
886 j_s.d [blink]
887 mov.lo r0,0
888.Lsbit:
889 ; Need to handle special cases involving negative powers of two:
890 ; r12,r2 are normalized dividend / divisor;
891 ; divide anything by 0x80000000, or divide 0x80000000 by 0x40000000
892 add_s r12,r12,r2
893 xor_s r1,r1,r0
894 rsub r4,r4,-1
895 ror r0,r12,r4
896 tst_s r2,r2
897 bmsk r0,r0,r3
898 add.pl r0,r0,r0
899 tst_s r1,r1
900 j_s.d [blink]
901 neg.mi r0,r0
902#else /* !MULDIV */
903/* This version requires that divaw works with a divisor of 0x80000000U */
904 abs_s r2,r1
905 norm r4,r0
906 neg_s r3,r2
907 norm r3,r3
908 abs_s r12,r0
909 brhs r4,r3,.Lonebit
910 asl_s r2,r2,r3
911 asl r12,r12,r4
912 sub lp_count,r3,r4
913 cmp_s r12,r2
914 sub.hs r12,r12,r2
915 lp .Ldivend
916.Ldivstart:divaw r12,r12,r2
917.Ldivend:xor_s r1,r1,r0
918 sub_s r0,r3,1
919 bmsk r0,r12,r0
920 bset.hs r0,r0,r3
921 tst_s r1,r1
922 j_s.d [blink]
923 negmi r0,r0
924.Lonebit:
925 xor_s r1,r1,r0
926 asr_s r1,r1,31
927 cmp_s r12,r2
928 mov_s r0,0
929 j_s.d [blink]
930 orhs r0,r1,1
931#endif /* MULDIV */
932
933#endif /* ifndef __ARC700__ */
934 ENDFUNC(__divsi3)
935
48c842ab 936
d38a64b4
JR		 937#endif /* L_divsi3 */
938
939#ifdef L_umodsi3
940 .section .text
941 .align 4
942
943 .global SYM(__umodsi3)
944 FUNC(__umodsi3)
945SYM(__umodsi3):
946 mov r7,blink
947 bl.nd @SYM(__udivmodsi4)
948 j.d [r7]
949 mov r0,r1
950 ENDFUNC(__umodsi3)
951#if 0 /* interferes with linux loader */
952 .section .__arc_profile_forward, "a"
953 .long SYM(__umodsi3)
954 .long SYM(__udivmodsi4)
955 .long 65536
956#endif
957
958#endif /* L_umodsi3 */
959
960#ifdef L_modsi3
961 .section .text
962 .align 4
963
964 .global SYM (__modsi3)
965 FUNC(__modsi3)
966SYM(__modsi3):
d64af69f 967#ifndef __ARC_EA__
d38a64b4
JR		 968 /* A5 / ARC60? */
969 mov_s r12,blink
048c6a9a 970 mov_s r11,r0
d38a64b4
JR		 971 abs_s r0,r0
972 bl.d @SYM(__udivmodsi4)
048c6a9a
CZ		 973 abs_s r1,r1
974 tst r11,r11
d38a64b4
JR		 975 neg_s r0,r1
976 j_s.d [r12]
048c6a9a 977 mov.pl r0,r1
d64af69f 978#else /* __ARC_EA__ */
d38a64b4
JR		 979 abs_s r2,r1
980 norm.f r4,r0
981 neg r5,r2
982 norm r3,r5
983 abs_l r12,r0
984 brhs r4,r3,.Lonebit
985 asl_s r2,r2,r3
986 asl r12,r12,r4
987 sub lp_count,r3,r4
988 cmp_s r12,r2
989 sub.hs r12,r12,r2
990 tst_s r0,r0
991 lp .Ldivend
992.Ldivstart:divaw r12,r12,r2
993.Ldivend:
994 lsr r0,r12,r3
995 j_s.d [blink]
996 neg.mi r0,r0
997 .balign 4
998.Lonebit:neg.pl r5,r5
999 cmp_s r12,r2
1000 j_s.d [blink]
1001 sub.hs r0,r0,r5
d64af69f 1002#endif /* !__ARC_EA__ */
d38a64b4
JR		 1003 ENDFUNC(__modsi3)
1004
1005#endif /* L_modsi3 */
1006
1007#ifdef L_clzsi2
1008 .section .text
1009 .align 4
1010 .global SYM (__clzsi2)
48c842ab 1011SYM(__clzsi2):
d38a64b4
JR		 1012#ifdef __ARC_NORM__
1013 HIDDEN_FUNC(__clzsi2)
1014 norm.f r0,r0
1015 mov.n r0,0
1016 j_s.d [blink]
1017 add.pl r0,r0,1
1018 ENDFUNC(__clzsi2)
d64af69f 1019#elif !defined (__ARC_BARREL_SHIFTER__)
d38a64b4
JR		 1020 FUNC(__clzsi2)
1021 mov lp_count,10
1022 mov_l r1,0
1023 bset r2,r1,29
1024 lp .Loop_end
1025 brhs r0,r2,.Loop_end
1026 add3 r0,r1,r0
1027.Loop_end:
1028 asl.f 0,r0
1029 sub2 r0,lp_count,lp_count
1030 sub.cs.f r0,r0,1
1031 add r0,r0,31
1032 j_s.d [blink]
1033 add.pl r0,r0,1
1034 ENDFUNC(__clzsi2)
1035#else
1036 FUNC(__clzsi2)
1037 asl.f 0,r0,2
1038 mov r1,-1
1039.Lcheck:
1040 bbit1.d r0,31,.Ldone
1041 asl.pl r0,r0,3
1042 bcs.d .Ldone_1
1043 add_s r1,r1,3
1044 bpnz.d .Lcheck
1045 asl.f 0,r0,2
1046 mov_s r0,32
1047 j_s.d [blink]
1048 mov.ne r0,r1
1049.Ldone:
1050 j_s.d [blink]
1051 add_s r0,r1,1
1052.Ldone_1:
1053 j_s.d [blink]
1054 sub_s r0,r1,1
1055 ENDFUNC(__clzsi2)
1056#endif
1057#endif /* L_clzsi2 */
1058 .section .text
1059
1060
1061;;; MILLICODE THUNK LIB ;***************
48c842ab 1062
d38a64b4
JR		 1063;;; .macro push_regs from, to, offset
1064;;; st_s "\from", [sp, \offset]
1065;;; .if \to-\from
1066;;; push_regs "(\from+1)", \to, "(\offset+4)"
1067;;; .endif
1068;;; .endm
1069;;; push_regs 13, 18, 0
1070;;;
1071
1072;;;; .macro sum from, to, three
1073;;;; .long \from
1074;;;; .long \three
1075;;;; .local regno
1076;;;; .set regno, \from+1
1077;;;; .set shift, 32
1078;;;; .set shift, shift - 1
48c842ab 1079;;;; # st_s %shift @3 lsl #shift
d38a64b4
JR		 1080;;;; .if \to-\from
1081;;;; sum "(\from+1)", \to, "(\three)"
48c842ab 1082;;;; .endif
d38a64b4 1083;;;; .endm
48c842ab 1084;;;;
d38a64b4 1085;;;; SUM 0,5, 9
48c842ab
CZ		 1086;;;;
1087; .altmacro
d38a64b4
JR		 1088;; .macro push_regs from=0, to=3, offset
1089;; st_s r\from, [sp, \offset]
1090;; .if \to-\from
1091;; push_regs "\from+1 ",\to,"(\offset+4)"
1092;; .endif
1093;; .endm
48c842ab 1094;;
d38a64b4
JR		 1095;; .macro expand_to_push from=13, to
1096;; ; .section .text
1097;; ; .align 4
1098;; ; .global st_
1099;; ; .type foo,
1100;; st_13_to_25:
1101;; ; push_regs \from, \to, 0
48c842ab 1102;; push_regs 0,3 ;
d38a64b4 1103;; .endm
48c842ab 1104;;
d38a64b4 1105;; expand_to_push 13,18
48c842ab 1106;;
d38a64b4
JR		 1107;#endif
1108
1109#ifdef L_millicodethunk_st
1110 .section .text
1111 .align 4
1112 .global SYM(__st_r13_to_r15)
1113 .global SYM(__st_r13_to_r16)
1114 .global SYM(__st_r13_to_r17)
1115 .global SYM(__st_r13_to_r18)
1116 .global SYM(__st_r13_to_r19)
1117 .global SYM(__st_r13_to_r20)
1118 .global SYM(__st_r13_to_r21)
1119 .global SYM(__st_r13_to_r22)
1120 .global SYM(__st_r13_to_r23)
1121 .global SYM(__st_r13_to_r24)
1122 .global SYM(__st_r13_to_r25)
1123 HIDDEN_FUNC(__st_r13_to_r15)
1124 HIDDEN_FUNC(__st_r13_to_r16)
1125 HIDDEN_FUNC(__st_r13_to_r17)
1126 HIDDEN_FUNC(__st_r13_to_r18)
1127 HIDDEN_FUNC(__st_r13_to_r19)
1128 HIDDEN_FUNC(__st_r13_to_r20)
1129 HIDDEN_FUNC(__st_r13_to_r21)
1130 HIDDEN_FUNC(__st_r13_to_r22)
1131 HIDDEN_FUNC(__st_r13_to_r23)
1132 HIDDEN_FUNC(__st_r13_to_r24)
1133 HIDDEN_FUNC(__st_r13_to_r25)
1134 .align 4
1135SYM(__st_r13_to_r25):
1136 st r25, [sp,48]
48c842ab 1137SYM(__st_r13_to_r24):
d38a64b4 1138 st r24, [sp,44]
48c842ab 1139SYM(__st_r13_to_r23):
d38a64b4 1140 st r23, [sp,40]
48c842ab 1141SYM(__st_r13_to_r22):
d38a64b4 1142 st r22, [sp,36]
48c842ab 1143SYM(__st_r13_to_r21):
d38a64b4 1144 st r21, [sp,32]
48c842ab
CZ		 1145SYM(__st_r13_to_r20):
1146 st r20, [sp,28]
1147SYM(__st_r13_to_r19):
d38a64b4 1148 st r19, [sp,24]
48c842ab 1149SYM(__st_r13_to_r18):
d38a64b4 1150 st r18, [sp,20]
48c842ab 1151SYM(__st_r13_to_r17):
d38a64b4 1152 st r17, [sp,16]
48c842ab 1153SYM(__st_r13_to_r16):
d38a64b4 1154 st r16, [sp,12]
48c842ab 1155SYM(__st_r13_to_r15):
d38a64b4
JR		 1156#ifdef __ARC700__
1157 st r15, [sp,8] ; minimum function size to avoid stall: 6 bytes.
1158#else
1159 st_s r15, [sp,8]
1160#endif
1161 st_s r14, [sp,4]
1162 j_s.d [%blink]
48c842ab 1163 st_s r13, [sp,0]
d38a64b4
JR		 1164 ENDFUNC(__st_r13_to_r15)
1165 ENDFUNC(__st_r13_to_r16)
1166 ENDFUNC(__st_r13_to_r17)
1167 ENDFUNC(__st_r13_to_r18)
1168 ENDFUNC(__st_r13_to_r19)
1169 ENDFUNC(__st_r13_to_r20)
1170 ENDFUNC(__st_r13_to_r21)
1171 ENDFUNC(__st_r13_to_r22)
1172 ENDFUNC(__st_r13_to_r23)
1173 ENDFUNC(__st_r13_to_r24)
1174 ENDFUNC(__st_r13_to_r25)
1175#endif /* L_millicodethunk_st */
1176
1177
1178#ifdef L_millicodethunk_ld
1179 .section .text
1180 .align 4
48c842ab 1181; ==================================
d38a64b4
JR		 1182; the loads
1183
1184 .global SYM(__ld_r13_to_r15)
1185 .global SYM(__ld_r13_to_r16)
1186 .global SYM(__ld_r13_to_r17)
1187 .global SYM(__ld_r13_to_r18)
1188 .global SYM(__ld_r13_to_r19)
1189 .global SYM(__ld_r13_to_r20)
1190 .global SYM(__ld_r13_to_r21)
1191 .global SYM(__ld_r13_to_r22)
1192 .global SYM(__ld_r13_to_r23)
1193 .global SYM(__ld_r13_to_r24)
1194 .global SYM(__ld_r13_to_r25)
1195 HIDDEN_FUNC(__ld_r13_to_r15)
1196 HIDDEN_FUNC(__ld_r13_to_r16)
1197 HIDDEN_FUNC(__ld_r13_to_r17)
1198 HIDDEN_FUNC(__ld_r13_to_r18)
1199 HIDDEN_FUNC(__ld_r13_to_r19)
1200 HIDDEN_FUNC(__ld_r13_to_r20)
1201 HIDDEN_FUNC(__ld_r13_to_r21)
1202 HIDDEN_FUNC(__ld_r13_to_r22)
1203 HIDDEN_FUNC(__ld_r13_to_r23)
1204 HIDDEN_FUNC(__ld_r13_to_r24)
1205 HIDDEN_FUNC(__ld_r13_to_r25)
1206SYM(__ld_r13_to_r25):
1207 ld r25, [sp,48]
1208SYM(__ld_r13_to_r24):
1209 ld r24, [sp,44]
1210SYM(__ld_r13_to_r23):
1211 ld r23, [sp,40]
1212SYM(__ld_r13_to_r22):
1213 ld r22, [sp,36]
1214SYM(__ld_r13_to_r21):
1215 ld r21, [sp,32]
1216SYM(__ld_r13_to_r20):
48c842ab 1217 ld r20, [sp,28]
d38a64b4
JR		 1218SYM(__ld_r13_to_r19):
1219 ld r19, [sp,24]
1220SYM(__ld_r13_to_r18):
1221 ld r18, [sp,20]
1222SYM(__ld_r13_to_r17):
1223 ld r17, [sp,16]
1224SYM(__ld_r13_to_r16):
1225 ld r16, [sp,12]
1226SYM(__ld_r13_to_r15):
1227#ifdef __ARC700__
1228 ld r15, [sp,8] ; minimum function size to avoid stall: 6 bytes.
1229#else
1230 ld_s r15, [sp,8]
1231#endif
1232 ld_s r14, [sp,4]
1233 j_s.d [%blink]
1234 ld_s r13, [sp,0]
1235 ENDFUNC(__ld_r13_to_r15)
1236 ENDFUNC(__ld_r13_to_r16)
1237 ENDFUNC(__ld_r13_to_r17)
1238 ENDFUNC(__ld_r13_to_r18)
1239 ENDFUNC(__ld_r13_to_r19)
1240 ENDFUNC(__ld_r13_to_r20)
1241 ENDFUNC(__ld_r13_to_r21)
1242 ENDFUNC(__ld_r13_to_r22)
1243 ENDFUNC(__ld_r13_to_r23)
1244 ENDFUNC(__ld_r13_to_r24)
1245 ENDFUNC(__ld_r13_to_r25)
1246
1247#endif /* L_millicodethunk_ld */
1248#ifdef L_millicodethunk_ret
1249 .global SYM(__ld_r13_to_r14_ret)
1250 .global SYM(__ld_r13_to_r15_ret)
1251 .global SYM(__ld_r13_to_r16_ret)
1252 .global SYM(__ld_r13_to_r17_ret)
1253 .global SYM(__ld_r13_to_r18_ret)
1254 .global SYM(__ld_r13_to_r19_ret)
1255 .global SYM(__ld_r13_to_r20_ret)
1256 .global SYM(__ld_r13_to_r21_ret)
1257 .global SYM(__ld_r13_to_r22_ret)
1258 .global SYM(__ld_r13_to_r23_ret)
1259 .global SYM(__ld_r13_to_r24_ret)
1260 .global SYM(__ld_r13_to_r25_ret)
1261 HIDDEN_FUNC(__ld_r13_to_r14_ret)
1262 HIDDEN_FUNC(__ld_r13_to_r15_ret)
1263 HIDDEN_FUNC(__ld_r13_to_r16_ret)
1264 HIDDEN_FUNC(__ld_r13_to_r17_ret)
1265 HIDDEN_FUNC(__ld_r13_to_r18_ret)
1266 HIDDEN_FUNC(__ld_r13_to_r19_ret)
1267 HIDDEN_FUNC(__ld_r13_to_r20_ret)
1268 HIDDEN_FUNC(__ld_r13_to_r21_ret)
1269 HIDDEN_FUNC(__ld_r13_to_r22_ret)
1270 HIDDEN_FUNC(__ld_r13_to_r23_ret)
1271 HIDDEN_FUNC(__ld_r13_to_r24_ret)
1272 HIDDEN_FUNC(__ld_r13_to_r25_ret)
1273 .section .text
1274 .align 4
1275SYM(__ld_r13_to_r25_ret):
1276 ld r25, [sp,48]
1277SYM(__ld_r13_to_r24_ret):
1278 ld r24, [sp,44]
1279SYM(__ld_r13_to_r23_ret):
1280 ld r23, [sp,40]
1281SYM(__ld_r13_to_r22_ret):
1282 ld r22, [sp,36]
1283SYM(__ld_r13_to_r21_ret):
1284 ld r21, [sp,32]
1285SYM(__ld_r13_to_r20_ret):
48c842ab 1286 ld r20, [sp,28]
d38a64b4
JR		 1287SYM(__ld_r13_to_r19_ret):
1288 ld r19, [sp,24]
1289SYM(__ld_r13_to_r18_ret):
1290 ld r18, [sp,20]
1291SYM(__ld_r13_to_r17_ret):
1292 ld r17, [sp,16]
1293SYM(__ld_r13_to_r16_ret):
1294 ld r16, [sp,12]
1295SYM(__ld_r13_to_r15_ret):
1296 ld r15, [sp,8]
1297SYM(__ld_r13_to_r14_ret):
1298 ld blink,[sp,r12]
1299 ld_s r14, [sp,4]
1300 ld.ab r13, [sp,r12]
1301 j_s.d [%blink]
1302 add_s sp,sp,4
1303 ENDFUNC(__ld_r13_to_r14_ret)
1304 ENDFUNC(__ld_r13_to_r15_ret)
1305 ENDFUNC(__ld_r13_to_r16_ret)
1306 ENDFUNC(__ld_r13_to_r17_ret)
1307 ENDFUNC(__ld_r13_to_r18_ret)
1308 ENDFUNC(__ld_r13_to_r19_ret)
1309 ENDFUNC(__ld_r13_to_r20_ret)
1310 ENDFUNC(__ld_r13_to_r21_ret)
1311 ENDFUNC(__ld_r13_to_r22_ret)
1312 ENDFUNC(__ld_r13_to_r23_ret)
1313 ENDFUNC(__ld_r13_to_r24_ret)
1314 ENDFUNC(__ld_r13_to_r25_ret)
1315
1316#endif /* L_millicodethunk_ret */
1317
48c842ab
CZ		 1318#define ARC_OPTFPE (defined (__ARC700__) || defined (__ARC_FPX_QUARK__))
1319
1320#if ARC_OPTFPE
d38a64b4
JR		 1321#ifdef L_adddf3
1322#ifdef __ARC_NORM__
1323#include "ieee-754/adddf3.S"
1324#endif
1325#endif
1326
1327#ifdef L_muldf3
48c842ab 1328#ifdef __ARC_MPY__
d38a64b4
JR		 1329#include "ieee-754/muldf3.S"
1330#elif defined (__ARC_NORM__) && defined(__ARC_MUL64__)
1331#include "ieee-754/arc600-mul64/muldf3.S"
1332#elif defined (__ARC_NORM__) && defined(__ARC_MUL32BY16__)
1333#include "ieee-754/arc600-dsp/muldf3.S"
1334#endif
1335#endif
1336
1337#ifdef L_addsf3
1338#ifdef __ARC_NORM__
1339#include "ieee-754/addsf3.S"
1340#endif
1341#endif
1342
1343#ifdef L_mulsf3
48c842ab 1344#ifdef __ARC_MPY__
d38a64b4
JR		 1345#include "ieee-754/mulsf3.S"
1346#elif defined (__ARC_NORM__) && defined(__ARC_MUL64__)
1347#include "ieee-754/arc600-mul64/mulsf3.S"
1348#elif defined (__ARC_NORM__) && defined(__ARC_MUL32BY16__)
1349#include "ieee-754/arc600-dsp/mulsf3.S"
1350#elif defined (__ARC_NORM__)
1351#include "ieee-754/arc600/mulsf3.S"
1352#endif
1353#endif
1354
1355#ifdef L_divdf3
48c842ab 1356#ifdef __ARC_MPY__
d38a64b4
JR		 1357#include "ieee-754/divdf3.S"
1358#elif defined (__ARC_NORM__) && defined(__ARC_MUL64__)
1359#include "ieee-754/arc600-mul64/divdf3.S"
1360#elif defined (__ARC_NORM__) && defined(__ARC_MUL32BY16__)
1361#include "ieee-754/arc600-dsp/divdf3.S"
1362#endif
1363#endif
1364
1365#ifdef L_divsf3
48c842ab 1366#ifdef __ARC_MPY__
d38a64b4
JR		 1367#include "ieee-754/divsf3-stdmul.S"
1368#elif defined (__ARC_NORM__) && defined(__ARC_MUL64__)
1369#include "ieee-754/arc600-mul64/divsf3.S"
1370#elif defined (__ARC_NORM__) && defined(__ARC_MUL32BY16__)
1371#include "ieee-754/arc600-dsp/divsf3.S"
1372#elif defined (__ARC_NORM__)
1373#include "ieee-754/arc600/divsf3.S"
1374#endif
1375#endif
1376
1377#ifdef L_extendsfdf2
1378#ifdef __ARC_NORM__
1379#include "ieee-754/extendsfdf2.S"
1380#endif
1381#endif
1382
1383#ifdef L_truncdfsf2
1384#ifdef __ARC_NORM__
1385#include "ieee-754/truncdfsf2.S"
1386#endif
1387#endif
1388
1389#ifdef L_floatsidf
1390#ifdef __ARC_NORM__
1391#include "ieee-754/floatsidf.S"
1392#endif
1393#endif
1394
1395#ifdef L_floatsisf
1396#ifdef __ARC_NORM__
1397#include "ieee-754/floatsisf.S"
1398#endif
1399#endif
1400
1401#ifdef L_floatunsidf
1402#ifdef __ARC_NORM__
1403#include "ieee-754/floatunsidf.S"
1404#endif
1405#endif
1406
1407#ifdef L_fixdfsi
1408#ifdef __ARC_NORM__
1409#include "ieee-754/fixdfsi.S"
1410#endif
1411#endif
1412
1413#ifdef L_fixsfsi
1414#ifdef __ARC_NORM__
1415#include "ieee-754/fixsfsi.S"
1416#endif
1417#endif
1418
1419#ifdef L_fixunsdfsi
1420#ifdef __ARC_NORM__
1421#include "ieee-754/fixunsdfsi.S"
1422#endif
1423#endif
1424
1425#ifdef L_eqdf2
1426#ifdef __ARC_NORM__
1427#include "ieee-754/eqdf2.S"
1428#endif
1429#endif
1430
1431#ifdef L_eqsf2
1432#ifdef __ARC_NORM__
1433#include "ieee-754/eqsf2.S"
1434#endif
1435#endif
1436
1437#ifdef L_gtdf2
1438#ifdef __ARC_NORM__
1439#include "ieee-754/gtdf2.S"
1440#endif
1441#endif
1442
1443#ifdef L_gtsf2
1444#ifdef __ARC_NORM__
1445#include "ieee-754/gtsf2.S"
1446#endif
1447#endif
1448
1449#ifdef L_gedf2
1450#ifdef __ARC_NORM__
1451#include "ieee-754/gedf2.S"
1452#endif
1453#endif
1454
1455#ifdef L_gesf2
1456#ifdef __ARC_NORM__
1457#include "ieee-754/gesf2.S"
1458#endif
1459#endif
1460
1461#ifdef L_uneqdf2
1462#ifdef __ARC_NORM__
1463#include "ieee-754/uneqdf2.S"
1464#endif
1465#endif
1466
1467#ifdef L_uneqsf2
1468#ifdef __ARC_NORM__
1469#include "ieee-754/uneqsf2.S"
1470#endif
1471#endif
1472
1473#ifdef L_orddf2
1474#ifdef __ARC_NORM__
1475#include "ieee-754/orddf2.S"
1476#endif
1477#endif
1478
1479#ifdef L_ordsf2
1480#ifdef __ARC_NORM__
1481#include "ieee-754/ordsf2.S"
1482#endif
1483#endif
48c842ab 1484#endif /* ARC_OPTFPE */
Mirror of https://gcc.gnu.org/git/gcc.git