xref: /openbmc/qemu/include/fpu/softfloat-macros.h (revision 2b108085)
1 /*
2  * QEMU float support macros
3  *
4  * The code in this source file is derived from release 2a of the SoftFloat
5  * IEC/IEEE Floating-point Arithmetic Package. Those parts of the code (and
6  * some later contributions) are provided under that license, as detailed below.
7  * It has subsequently been modified by contributors to the QEMU Project,
8  * so some portions are provided under:
9  *  the SoftFloat-2a license
10  *  the BSD license
11  *  GPL-v2-or-later
12  *
13  * Any future contributions to this file after December 1st 2014 will be
14  * taken to be licensed under the Softfloat-2a license unless specifically
15  * indicated otherwise.
16  */
17 
18 /*
19 ===============================================================================
20 This C source fragment is part of the SoftFloat IEC/IEEE Floating-point
21 Arithmetic Package, Release 2a.
22 
23 Written by John R. Hauser.  This work was made possible in part by the
24 International Computer Science Institute, located at Suite 600, 1947 Center
25 Street, Berkeley, California 94704.  Funding was partially provided by the
26 National Science Foundation under grant MIP-9311980.  The original version
27 of this code was written as part of a project to build a fixed-point vector
28 processor in collaboration with the University of California at Berkeley,
29 overseen by Profs. Nelson Morgan and John Wawrzynek.  More information
30 is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/
31 arithmetic/SoftFloat.html'.
32 
33 THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE.  Although reasonable effort
34 has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT
35 TIMES RESULT IN INCORRECT BEHAVIOR.  USE OF THIS SOFTWARE IS RESTRICTED TO
36 PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY
37 AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE.
38 
39 Derivative works are acceptable, even for commercial purposes, so long as
40 (1) they include prominent notice that the work is derivative, and (2) they
41 include prominent notice akin to these four paragraphs for those parts of
42 this code that are retained.
43 
44 ===============================================================================
45 */
46 
47 /* BSD licensing:
48  * Copyright (c) 2006, Fabrice Bellard
49  * All rights reserved.
50  *
51  * Redistribution and use in source and binary forms, with or without
52  * modification, are permitted provided that the following conditions are met:
53  *
54  * 1. Redistributions of source code must retain the above copyright notice,
55  * this list of conditions and the following disclaimer.
56  *
57  * 2. Redistributions in binary form must reproduce the above copyright notice,
58  * this list of conditions and the following disclaimer in the documentation
59  * and/or other materials provided with the distribution.
60  *
61  * 3. Neither the name of the copyright holder nor the names of its contributors
62  * may be used to endorse or promote products derived from this software without
63  * specific prior written permission.
64  *
65  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
66  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
67  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
68  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
69  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
70  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
71  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
72  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
73  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
74  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
75  * THE POSSIBILITY OF SUCH DAMAGE.
76  */
77 
78 /* Portions of this work are licensed under the terms of the GNU GPL,
79  * version 2 or later. See the COPYING file in the top-level directory.
80  */
81 
82 /*----------------------------------------------------------------------------
83 | This macro tests for minimum version of the GNU C compiler.
84 *----------------------------------------------------------------------------*/
85 #if defined(__GNUC__) && defined(__GNUC_MINOR__)
86 # define SOFTFLOAT_GNUC_PREREQ(maj, min) \
87          ((__GNUC__ << 16) + __GNUC_MINOR__ >= ((maj) << 16) + (min))
88 #else
89 # define SOFTFLOAT_GNUC_PREREQ(maj, min) 0
90 #endif
91 
92 
93 /*----------------------------------------------------------------------------
94 | Shifts `a' right by the number of bits given in `count'.  If any nonzero
95 | bits are shifted off, they are ``jammed'' into the least significant bit of
96 | the result by setting the least significant bit to 1.  The value of `count'
97 | can be arbitrarily large; in particular, if `count' is greater than 32, the
98 | result will be either 0 or 1, depending on whether `a' is zero or nonzero.
99 | The result is stored in the location pointed to by `zPtr'.
100 *----------------------------------------------------------------------------*/
101 
102 static inline void shift32RightJamming(uint32_t a, int count, uint32_t *zPtr)
103 {
104     uint32_t z;
105 
106     if ( count == 0 ) {
107         z = a;
108     }
109     else if ( count < 32 ) {
110         z = ( a>>count ) | ( ( a<<( ( - count ) & 31 ) ) != 0 );
111     }
112     else {
113         z = ( a != 0 );
114     }
115     *zPtr = z;
116 
117 }
118 
119 /*----------------------------------------------------------------------------
120 | Shifts `a' right by the number of bits given in `count'.  If any nonzero
121 | bits are shifted off, they are ``jammed'' into the least significant bit of
122 | the result by setting the least significant bit to 1.  The value of `count'
123 | can be arbitrarily large; in particular, if `count' is greater than 64, the
124 | result will be either 0 or 1, depending on whether `a' is zero or nonzero.
125 | The result is stored in the location pointed to by `zPtr'.
126 *----------------------------------------------------------------------------*/
127 
128 static inline void shift64RightJamming(uint64_t a, int count, uint64_t *zPtr)
129 {
130     uint64_t z;
131 
132     if ( count == 0 ) {
133         z = a;
134     }
135     else if ( count < 64 ) {
136         z = ( a>>count ) | ( ( a<<( ( - count ) & 63 ) ) != 0 );
137     }
138     else {
139         z = ( a != 0 );
140     }
141     *zPtr = z;
142 
143 }
144 
145 /*----------------------------------------------------------------------------
146 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by 64
147 | _plus_ the number of bits given in `count'.  The shifted result is at most
148 | 64 nonzero bits; this is stored at the location pointed to by `z0Ptr'.  The
149 | bits shifted off form a second 64-bit result as follows:  The _last_ bit
150 | shifted off is the most-significant bit of the extra result, and the other
151 | 63 bits of the extra result are all zero if and only if _all_but_the_last_
152 | bits shifted off were all zero.  This extra result is stored in the location
153 | pointed to by `z1Ptr'.  The value of `count' can be arbitrarily large.
154 |     (This routine makes more sense if `a0' and `a1' are considered to form a
155 | fixed-point value with binary point between `a0' and `a1'.  This fixed-point
156 | value is shifted right by the number of bits given in `count', and the
157 | integer part of the result is returned at the location pointed to by
158 | `z0Ptr'.  The fractional part of the result may be slightly corrupted as
159 | described above, and is returned at the location pointed to by `z1Ptr'.)
160 *----------------------------------------------------------------------------*/
161 
162 static inline void
163  shift64ExtraRightJamming(
164      uint64_t a0, uint64_t a1, int count, uint64_t *z0Ptr, uint64_t *z1Ptr)
165 {
166     uint64_t z0, z1;
167     int8_t negCount = ( - count ) & 63;
168 
169     if ( count == 0 ) {
170         z1 = a1;
171         z0 = a0;
172     }
173     else if ( count < 64 ) {
174         z1 = ( a0<<negCount ) | ( a1 != 0 );
175         z0 = a0>>count;
176     }
177     else {
178         if ( count == 64 ) {
179             z1 = a0 | ( a1 != 0 );
180         }
181         else {
182             z1 = ( ( a0 | a1 ) != 0 );
183         }
184         z0 = 0;
185     }
186     *z1Ptr = z1;
187     *z0Ptr = z0;
188 
189 }
190 
191 /*----------------------------------------------------------------------------
192 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the
193 | number of bits given in `count'.  Any bits shifted off are lost.  The value
194 | of `count' can be arbitrarily large; in particular, if `count' is greater
195 | than 128, the result will be 0.  The result is broken into two 64-bit pieces
196 | which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
197 *----------------------------------------------------------------------------*/
198 
199 static inline void
200  shift128Right(
201      uint64_t a0, uint64_t a1, int count, uint64_t *z0Ptr, uint64_t *z1Ptr)
202 {
203     uint64_t z0, z1;
204     int8_t negCount = ( - count ) & 63;
205 
206     if ( count == 0 ) {
207         z1 = a1;
208         z0 = a0;
209     }
210     else if ( count < 64 ) {
211         z1 = ( a0<<negCount ) | ( a1>>count );
212         z0 = a0>>count;
213     }
214     else {
215         z1 = (count < 128) ? (a0 >> (count & 63)) : 0;
216         z0 = 0;
217     }
218     *z1Ptr = z1;
219     *z0Ptr = z0;
220 
221 }
222 
223 /*----------------------------------------------------------------------------
224 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the
225 | number of bits given in `count'.  If any nonzero bits are shifted off, they
226 | are ``jammed'' into the least significant bit of the result by setting the
227 | least significant bit to 1.  The value of `count' can be arbitrarily large;
228 | in particular, if `count' is greater than 128, the result will be either
229 | 0 or 1, depending on whether the concatenation of `a0' and `a1' is zero or
230 | nonzero.  The result is broken into two 64-bit pieces which are stored at
231 | the locations pointed to by `z0Ptr' and `z1Ptr'.
232 *----------------------------------------------------------------------------*/
233 
234 static inline void
235  shift128RightJamming(
236      uint64_t a0, uint64_t a1, int count, uint64_t *z0Ptr, uint64_t *z1Ptr)
237 {
238     uint64_t z0, z1;
239     int8_t negCount = ( - count ) & 63;
240 
241     if ( count == 0 ) {
242         z1 = a1;
243         z0 = a0;
244     }
245     else if ( count < 64 ) {
246         z1 = ( a0<<negCount ) | ( a1>>count ) | ( ( a1<<negCount ) != 0 );
247         z0 = a0>>count;
248     }
249     else {
250         if ( count == 64 ) {
251             z1 = a0 | ( a1 != 0 );
252         }
253         else if ( count < 128 ) {
254             z1 = ( a0>>( count & 63 ) ) | ( ( ( a0<<negCount ) | a1 ) != 0 );
255         }
256         else {
257             z1 = ( ( a0 | a1 ) != 0 );
258         }
259         z0 = 0;
260     }
261     *z1Ptr = z1;
262     *z0Ptr = z0;
263 
264 }
265 
266 /*----------------------------------------------------------------------------
267 | Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' right
268 | by 64 _plus_ the number of bits given in `count'.  The shifted result is
269 | at most 128 nonzero bits; these are broken into two 64-bit pieces which are
270 | stored at the locations pointed to by `z0Ptr' and `z1Ptr'.  The bits shifted
271 | off form a third 64-bit result as follows:  The _last_ bit shifted off is
272 | the most-significant bit of the extra result, and the other 63 bits of the
273 | extra result are all zero if and only if _all_but_the_last_ bits shifted off
274 | were all zero.  This extra result is stored in the location pointed to by
275 | `z2Ptr'.  The value of `count' can be arbitrarily large.
276 |     (This routine makes more sense if `a0', `a1', and `a2' are considered
277 | to form a fixed-point value with binary point between `a1' and `a2'.  This
278 | fixed-point value is shifted right by the number of bits given in `count',
279 | and the integer part of the result is returned at the locations pointed to
280 | by `z0Ptr' and `z1Ptr'.  The fractional part of the result may be slightly
281 | corrupted as described above, and is returned at the location pointed to by
282 | `z2Ptr'.)
283 *----------------------------------------------------------------------------*/
284 
285 static inline void
286  shift128ExtraRightJamming(
287      uint64_t a0,
288      uint64_t a1,
289      uint64_t a2,
290      int count,
291      uint64_t *z0Ptr,
292      uint64_t *z1Ptr,
293      uint64_t *z2Ptr
294  )
295 {
296     uint64_t z0, z1, z2;
297     int8_t negCount = ( - count ) & 63;
298 
299     if ( count == 0 ) {
300         z2 = a2;
301         z1 = a1;
302         z0 = a0;
303     }
304     else {
305         if ( count < 64 ) {
306             z2 = a1<<negCount;
307             z1 = ( a0<<negCount ) | ( a1>>count );
308             z0 = a0>>count;
309         }
310         else {
311             if ( count == 64 ) {
312                 z2 = a1;
313                 z1 = a0;
314             }
315             else {
316                 a2 |= a1;
317                 if ( count < 128 ) {
318                     z2 = a0<<negCount;
319                     z1 = a0>>( count & 63 );
320                 }
321                 else {
322                     z2 = ( count == 128 ) ? a0 : ( a0 != 0 );
323                     z1 = 0;
324                 }
325             }
326             z0 = 0;
327         }
328         z2 |= ( a2 != 0 );
329     }
330     *z2Ptr = z2;
331     *z1Ptr = z1;
332     *z0Ptr = z0;
333 
334 }
335 
336 /*----------------------------------------------------------------------------
337 | Shifts the 128-bit value formed by concatenating `a0' and `a1' left by the
338 | number of bits given in `count'.  Any bits shifted off are lost.  The value
339 | of `count' must be less than 64.  The result is broken into two 64-bit
340 | pieces which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
341 *----------------------------------------------------------------------------*/
342 
343 static inline void
344  shortShift128Left(
345      uint64_t a0, uint64_t a1, int count, uint64_t *z0Ptr, uint64_t *z1Ptr)
346 {
347 
348     *z1Ptr = a1<<count;
349     *z0Ptr =
350         ( count == 0 ) ? a0 : ( a0<<count ) | ( a1>>( ( - count ) & 63 ) );
351 
352 }
353 
354 /*----------------------------------------------------------------------------
355 | Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' left
356 | by the number of bits given in `count'.  Any bits shifted off are lost.
357 | The value of `count' must be less than 64.  The result is broken into three
358 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr',
359 | `z1Ptr', and `z2Ptr'.
360 *----------------------------------------------------------------------------*/
361 
362 static inline void
363  shortShift192Left(
364      uint64_t a0,
365      uint64_t a1,
366      uint64_t a2,
367      int count,
368      uint64_t *z0Ptr,
369      uint64_t *z1Ptr,
370      uint64_t *z2Ptr
371  )
372 {
373     uint64_t z0, z1, z2;
374     int8_t negCount;
375 
376     z2 = a2<<count;
377     z1 = a1<<count;
378     z0 = a0<<count;
379     if ( 0 < count ) {
380         negCount = ( ( - count ) & 63 );
381         z1 |= a2>>negCount;
382         z0 |= a1>>negCount;
383     }
384     *z2Ptr = z2;
385     *z1Ptr = z1;
386     *z0Ptr = z0;
387 
388 }
389 
390 /*----------------------------------------------------------------------------
391 | Adds the 128-bit value formed by concatenating `a0' and `a1' to the 128-bit
392 | value formed by concatenating `b0' and `b1'.  Addition is modulo 2^128, so
393 | any carry out is lost.  The result is broken into two 64-bit pieces which
394 | are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
395 *----------------------------------------------------------------------------*/
396 
397 static inline void
398  add128(
399      uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1, uint64_t *z0Ptr, uint64_t *z1Ptr )
400 {
401     uint64_t z1;
402 
403     z1 = a1 + b1;
404     *z1Ptr = z1;
405     *z0Ptr = a0 + b0 + ( z1 < a1 );
406 
407 }
408 
409 /*----------------------------------------------------------------------------
410 | Adds the 192-bit value formed by concatenating `a0', `a1', and `a2' to the
411 | 192-bit value formed by concatenating `b0', `b1', and `b2'.  Addition is
412 | modulo 2^192, so any carry out is lost.  The result is broken into three
413 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr',
414 | `z1Ptr', and `z2Ptr'.
415 *----------------------------------------------------------------------------*/
416 
417 static inline void
418  add192(
419      uint64_t a0,
420      uint64_t a1,
421      uint64_t a2,
422      uint64_t b0,
423      uint64_t b1,
424      uint64_t b2,
425      uint64_t *z0Ptr,
426      uint64_t *z1Ptr,
427      uint64_t *z2Ptr
428  )
429 {
430     uint64_t z0, z1, z2;
431     int8_t carry0, carry1;
432 
433     z2 = a2 + b2;
434     carry1 = ( z2 < a2 );
435     z1 = a1 + b1;
436     carry0 = ( z1 < a1 );
437     z0 = a0 + b0;
438     z1 += carry1;
439     z0 += ( z1 < carry1 );
440     z0 += carry0;
441     *z2Ptr = z2;
442     *z1Ptr = z1;
443     *z0Ptr = z0;
444 
445 }
446 
447 /*----------------------------------------------------------------------------
448 | Subtracts the 128-bit value formed by concatenating `b0' and `b1' from the
449 | 128-bit value formed by concatenating `a0' and `a1'.  Subtraction is modulo
450 | 2^128, so any borrow out (carry out) is lost.  The result is broken into two
451 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr' and
452 | `z1Ptr'.
453 *----------------------------------------------------------------------------*/
454 
455 static inline void
456  sub128(
457      uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1, uint64_t *z0Ptr, uint64_t *z1Ptr )
458 {
459 
460     *z1Ptr = a1 - b1;
461     *z0Ptr = a0 - b0 - ( a1 < b1 );
462 
463 }
464 
465 /*----------------------------------------------------------------------------
466 | Subtracts the 192-bit value formed by concatenating `b0', `b1', and `b2'
467 | from the 192-bit value formed by concatenating `a0', `a1', and `a2'.
468 | Subtraction is modulo 2^192, so any borrow out (carry out) is lost.  The
469 | result is broken into three 64-bit pieces which are stored at the locations
470 | pointed to by `z0Ptr', `z1Ptr', and `z2Ptr'.
471 *----------------------------------------------------------------------------*/
472 
473 static inline void
474  sub192(
475      uint64_t a0,
476      uint64_t a1,
477      uint64_t a2,
478      uint64_t b0,
479      uint64_t b1,
480      uint64_t b2,
481      uint64_t *z0Ptr,
482      uint64_t *z1Ptr,
483      uint64_t *z2Ptr
484  )
485 {
486     uint64_t z0, z1, z2;
487     int8_t borrow0, borrow1;
488 
489     z2 = a2 - b2;
490     borrow1 = ( a2 < b2 );
491     z1 = a1 - b1;
492     borrow0 = ( a1 < b1 );
493     z0 = a0 - b0;
494     z0 -= ( z1 < borrow1 );
495     z1 -= borrow1;
496     z0 -= borrow0;
497     *z2Ptr = z2;
498     *z1Ptr = z1;
499     *z0Ptr = z0;
500 
501 }
502 
503 /*----------------------------------------------------------------------------
504 | Multiplies `a' by `b' to obtain a 128-bit product.  The product is broken
505 | into two 64-bit pieces which are stored at the locations pointed to by
506 | `z0Ptr' and `z1Ptr'.
507 *----------------------------------------------------------------------------*/
508 
509 static inline void mul64To128( uint64_t a, uint64_t b, uint64_t *z0Ptr, uint64_t *z1Ptr )
510 {
511     uint32_t aHigh, aLow, bHigh, bLow;
512     uint64_t z0, zMiddleA, zMiddleB, z1;
513 
514     aLow = a;
515     aHigh = a>>32;
516     bLow = b;
517     bHigh = b>>32;
518     z1 = ( (uint64_t) aLow ) * bLow;
519     zMiddleA = ( (uint64_t) aLow ) * bHigh;
520     zMiddleB = ( (uint64_t) aHigh ) * bLow;
521     z0 = ( (uint64_t) aHigh ) * bHigh;
522     zMiddleA += zMiddleB;
523     z0 += ( ( (uint64_t) ( zMiddleA < zMiddleB ) )<<32 ) + ( zMiddleA>>32 );
524     zMiddleA <<= 32;
525     z1 += zMiddleA;
526     z0 += ( z1 < zMiddleA );
527     *z1Ptr = z1;
528     *z0Ptr = z0;
529 
530 }
531 
532 /*----------------------------------------------------------------------------
533 | Multiplies the 128-bit value formed by concatenating `a0' and `a1' by
534 | `b' to obtain a 192-bit product.  The product is broken into three 64-bit
535 | pieces which are stored at the locations pointed to by `z0Ptr', `z1Ptr', and
536 | `z2Ptr'.
537 *----------------------------------------------------------------------------*/
538 
539 static inline void
540  mul128By64To192(
541      uint64_t a0,
542      uint64_t a1,
543      uint64_t b,
544      uint64_t *z0Ptr,
545      uint64_t *z1Ptr,
546      uint64_t *z2Ptr
547  )
548 {
549     uint64_t z0, z1, z2, more1;
550 
551     mul64To128( a1, b, &z1, &z2 );
552     mul64To128( a0, b, &z0, &more1 );
553     add128( z0, more1, 0, z1, &z0, &z1 );
554     *z2Ptr = z2;
555     *z1Ptr = z1;
556     *z0Ptr = z0;
557 
558 }
559 
560 /*----------------------------------------------------------------------------
561 | Multiplies the 128-bit value formed by concatenating `a0' and `a1' to the
562 | 128-bit value formed by concatenating `b0' and `b1' to obtain a 256-bit
563 | product.  The product is broken into four 64-bit pieces which are stored at
564 | the locations pointed to by `z0Ptr', `z1Ptr', `z2Ptr', and `z3Ptr'.
565 *----------------------------------------------------------------------------*/
566 
567 static inline void
568  mul128To256(
569      uint64_t a0,
570      uint64_t a1,
571      uint64_t b0,
572      uint64_t b1,
573      uint64_t *z0Ptr,
574      uint64_t *z1Ptr,
575      uint64_t *z2Ptr,
576      uint64_t *z3Ptr
577  )
578 {
579     uint64_t z0, z1, z2, z3;
580     uint64_t more1, more2;
581 
582     mul64To128( a1, b1, &z2, &z3 );
583     mul64To128( a1, b0, &z1, &more2 );
584     add128( z1, more2, 0, z2, &z1, &z2 );
585     mul64To128( a0, b0, &z0, &more1 );
586     add128( z0, more1, 0, z1, &z0, &z1 );
587     mul64To128( a0, b1, &more1, &more2 );
588     add128( more1, more2, 0, z2, &more1, &z2 );
589     add128( z0, z1, 0, more1, &z0, &z1 );
590     *z3Ptr = z3;
591     *z2Ptr = z2;
592     *z1Ptr = z1;
593     *z0Ptr = z0;
594 
595 }
596 
597 /*----------------------------------------------------------------------------
598 | Returns an approximation to the 64-bit integer quotient obtained by dividing
599 | `b' into the 128-bit value formed by concatenating `a0' and `a1'.  The
600 | divisor `b' must be at least 2^63.  If q is the exact quotient truncated
601 | toward zero, the approximation returned lies between q and q + 2 inclusive.
602 | If the exact quotient q is larger than 64 bits, the maximum positive 64-bit
603 | unsigned integer is returned.
604 *----------------------------------------------------------------------------*/
605 
606 static inline uint64_t estimateDiv128To64(uint64_t a0, uint64_t a1, uint64_t b)
607 {
608     uint64_t b0, b1;
609     uint64_t rem0, rem1, term0, term1;
610     uint64_t z;
611 
612     if ( b <= a0 ) return LIT64( 0xFFFFFFFFFFFFFFFF );
613     b0 = b>>32;
614     z = ( b0<<32 <= a0 ) ? LIT64( 0xFFFFFFFF00000000 ) : ( a0 / b0 )<<32;
615     mul64To128( b, z, &term0, &term1 );
616     sub128( a0, a1, term0, term1, &rem0, &rem1 );
617     while ( ( (int64_t) rem0 ) < 0 ) {
618         z -= LIT64( 0x100000000 );
619         b1 = b<<32;
620         add128( rem0, rem1, b0, b1, &rem0, &rem1 );
621     }
622     rem0 = ( rem0<<32 ) | ( rem1>>32 );
623     z |= ( b0<<32 <= rem0 ) ? 0xFFFFFFFF : rem0 / b0;
624     return z;
625 
626 }
627 
628 /* From the GNU Multi Precision Library - longlong.h __udiv_qrnnd
629  * (https://gmplib.org/repo/gmp/file/tip/longlong.h)
630  *
631  * Licensed under the GPLv2/LGPLv3
632  */
633 static inline uint64_t div128To64(uint64_t n0, uint64_t n1, uint64_t d)
634 {
635     uint64_t d0, d1, q0, q1, r1, r0, m;
636 
637     d0 = (uint32_t)d;
638     d1 = d >> 32;
639 
640     r1 = n1 % d1;
641     q1 = n1 / d1;
642     m = q1 * d0;
643     r1 = (r1 << 32) | (n0 >> 32);
644     if (r1 < m) {
645         q1 -= 1;
646         r1 += d;
647         if (r1 >= d) {
648             if (r1 < m) {
649                 q1 -= 1;
650                 r1 += d;
651             }
652         }
653     }
654     r1 -= m;
655 
656     r0 = r1 % d1;
657     q0 = r1 / d1;
658     m = q0 * d0;
659     r0 = (r0 << 32) | (uint32_t)n0;
660     if (r0 < m) {
661         q0 -= 1;
662         r0 += d;
663         if (r0 >= d) {
664             if (r0 < m) {
665                 q0 -= 1;
666                 r0 += d;
667             }
668         }
669     }
670     r0 -= m;
671 
672     /* Return remainder in LSB */
673     return (q1 << 32) | q0 | (r0 != 0);
674 }
675 
676 /*----------------------------------------------------------------------------
677 | Returns an approximation to the square root of the 32-bit significand given
678 | by `a'.  Considered as an integer, `a' must be at least 2^31.  If bit 0 of
679 | `aExp' (the least significant bit) is 1, the integer returned approximates
680 | 2^31*sqrt(`a'/2^31), where `a' is considered an integer.  If bit 0 of `aExp'
681 | is 0, the integer returned approximates 2^31*sqrt(`a'/2^30).  In either
682 | case, the approximation returned lies strictly within +/-2 of the exact
683 | value.
684 *----------------------------------------------------------------------------*/
685 
686 static inline uint32_t estimateSqrt32(int aExp, uint32_t a)
687 {
688     static const uint16_t sqrtOddAdjustments[] = {
689         0x0004, 0x0022, 0x005D, 0x00B1, 0x011D, 0x019F, 0x0236, 0x02E0,
690         0x039C, 0x0468, 0x0545, 0x0631, 0x072B, 0x0832, 0x0946, 0x0A67
691     };
692     static const uint16_t sqrtEvenAdjustments[] = {
693         0x0A2D, 0x08AF, 0x075A, 0x0629, 0x051A, 0x0429, 0x0356, 0x029E,
694         0x0200, 0x0179, 0x0109, 0x00AF, 0x0068, 0x0034, 0x0012, 0x0002
695     };
696     int8_t index;
697     uint32_t z;
698 
699     index = ( a>>27 ) & 15;
700     if ( aExp & 1 ) {
701         z = 0x4000 + ( a>>17 ) - sqrtOddAdjustments[ (int)index ];
702         z = ( ( a / z )<<14 ) + ( z<<15 );
703         a >>= 1;
704     }
705     else {
706         z = 0x8000 + ( a>>17 ) - sqrtEvenAdjustments[ (int)index ];
707         z = a / z + z;
708         z = ( 0x20000 <= z ) ? 0xFFFF8000 : ( z<<15 );
709         if ( z <= a ) return (uint32_t) ( ( (int32_t) a )>>1 );
710     }
711     return ( (uint32_t) ( ( ( (uint64_t) a )<<31 ) / z ) ) + ( z>>1 );
712 
713 }
714 
715 /*----------------------------------------------------------------------------
716 | Returns the number of leading 0 bits before the most-significant 1 bit of
717 | `a'.  If `a' is zero, 32 is returned.
718 *----------------------------------------------------------------------------*/
719 
720 static inline int8_t countLeadingZeros32(uint32_t a)
721 {
722 #if SOFTFLOAT_GNUC_PREREQ(3, 4)
723     if (a) {
724         return __builtin_clz(a);
725     } else {
726         return 32;
727     }
728 #else
729     static const int8_t countLeadingZerosHigh[] = {
730         8, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4,
731         3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
732         2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
733         2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
734         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
735         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
736         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
737         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
738         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
739         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
740         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
741         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
742         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
743         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
744         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
745         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
746     };
747     int8_t shiftCount;
748 
749     shiftCount = 0;
750     if ( a < 0x10000 ) {
751         shiftCount += 16;
752         a <<= 16;
753     }
754     if ( a < 0x1000000 ) {
755         shiftCount += 8;
756         a <<= 8;
757     }
758     shiftCount += countLeadingZerosHigh[ a>>24 ];
759     return shiftCount;
760 #endif
761 }
762 
763 /*----------------------------------------------------------------------------
764 | Returns the number of leading 0 bits before the most-significant 1 bit of
765 | `a'.  If `a' is zero, 64 is returned.
766 *----------------------------------------------------------------------------*/
767 
768 static inline int8_t countLeadingZeros64(uint64_t a)
769 {
770 #if SOFTFLOAT_GNUC_PREREQ(3, 4)
771     if (a) {
772         return __builtin_clzll(a);
773     } else {
774         return 64;
775     }
776 #else
777     int8_t shiftCount;
778 
779     shiftCount = 0;
780     if ( a < ( (uint64_t) 1 )<<32 ) {
781         shiftCount += 32;
782     }
783     else {
784         a >>= 32;
785     }
786     shiftCount += countLeadingZeros32( a );
787     return shiftCount;
788 #endif
789 }
790 
791 /*----------------------------------------------------------------------------
792 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1'
793 | is equal to the 128-bit value formed by concatenating `b0' and `b1'.
794 | Otherwise, returns 0.
795 *----------------------------------------------------------------------------*/
796 
797 static inline flag eq128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 )
798 {
799 
800     return ( a0 == b0 ) && ( a1 == b1 );
801 
802 }
803 
804 /*----------------------------------------------------------------------------
805 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less
806 | than or equal to the 128-bit value formed by concatenating `b0' and `b1'.
807 | Otherwise, returns 0.
808 *----------------------------------------------------------------------------*/
809 
810 static inline flag le128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 )
811 {
812 
813     return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 <= b1 ) );
814 
815 }
816 
817 /*----------------------------------------------------------------------------
818 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less
819 | than the 128-bit value formed by concatenating `b0' and `b1'.  Otherwise,
820 | returns 0.
821 *----------------------------------------------------------------------------*/
822 
823 static inline flag lt128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 )
824 {
825 
826     return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 < b1 ) );
827 
828 }
829 
830 /*----------------------------------------------------------------------------
831 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is
832 | not equal to the 128-bit value formed by concatenating `b0' and `b1'.
833 | Otherwise, returns 0.
834 *----------------------------------------------------------------------------*/
835 
836 static inline flag ne128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 )
837 {
838 
839     return ( a0 != b0 ) || ( a1 != b1 );
840 
841 }
842