• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 
2 /*============================================================================
3 
4 This C source fragment is part of the SoftFloat IEC/IEEE Floating-point
5 Arithmetic Package, Release 2b.
6 
7 Written by John R. Hauser.  This work was made possible in part by the
8 International Computer Science Institute, located at Suite 600, 1947 Center
9 Street, Berkeley, California 94704.  Funding was partially provided by the
10 National Science Foundation under grant MIP-9311980.  The original version
11 of this code was written as part of a project to build a fixed-point vector
12 processor in collaboration with the University of California at Berkeley,
13 overseen by Profs. Nelson Morgan and John Wawrzynek.  More information
14 is available through the Web page `http://www.cs.berkeley.edu/~jhauser/
15 arithmetic/SoftFloat.html'.
16 
17 THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE.  Although reasonable effort has
18 been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT TIMES
19 RESULT IN INCORRECT BEHAVIOR.  USE OF THIS SOFTWARE IS RESTRICTED TO PERSONS
20 AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ALL LOSSES,
21 COSTS, OR OTHER PROBLEMS THEY INCUR DUE TO THE SOFTWARE, AND WHO FURTHERMORE
22 EFFECTIVELY INDEMNIFY JOHN HAUSER AND THE INTERNATIONAL COMPUTER SCIENCE
23 INSTITUTE (possibly via similar legal notice) AGAINST ALL LOSSES, COSTS, OR
24 OTHER PROBLEMS INCURRED BY THEIR CUSTOMERS AND CLIENTS DUE TO THE SOFTWARE.
25 
26 Derivative works are acceptable, even for commercial purposes, so long as
27 (1) the source code for the derivative work includes prominent notice that
28 the work is derivative, and (2) the source code includes prominent notice with
29 these four paragraphs for those parts of this code that are retained.
30 
31 =============================================================================*/
32 
33 /*----------------------------------------------------------------------------
34 | Shifts `a' right by the number of bits given in `count'.  If any nonzero
35 | bits are shifted off, they are ``jammed'' into the least significant bit of
36 | the result by setting the least significant bit to 1.  The value of `count'
37 | can be arbitrarily large; in particular, if `count' is greater than 32, the
38 | result will be either 0 or 1, depending on whether `a' is zero or nonzero.
39 | The result is stored in the location pointed to by `zPtr'.
40 *----------------------------------------------------------------------------*/
41 
shift32RightJamming(bits32 a,int16 count,bits32 * zPtr)42 INLINE void shift32RightJamming( bits32 a, int16 count, bits32 *zPtr )
43 {
44     bits32 z;
45 
46     if ( count == 0 ) {
47         z = a;
48     }
49     else if ( count < 32 ) {
50         z = ( a>>count ) | ( ( a<<( ( - count ) & 31 ) ) != 0 );
51     }
52     else {
53         z = ( a != 0 );
54     }
55     *zPtr = z;
56 
57 }
58 
59 /*----------------------------------------------------------------------------
60 | Shifts `a' right by the number of bits given in `count'.  If any nonzero
61 | bits are shifted off, they are ``jammed'' into the least significant bit of
62 | the result by setting the least significant bit to 1.  The value of `count'
63 | can be arbitrarily large; in particular, if `count' is greater than 64, the
64 | result will be either 0 or 1, depending on whether `a' is zero or nonzero.
65 | The result is stored in the location pointed to by `zPtr'.
66 *----------------------------------------------------------------------------*/
67 
shift64RightJamming(bits64 a,int16 count,bits64 * zPtr)68 INLINE void shift64RightJamming( bits64 a, int16 count, bits64 *zPtr )
69 {
70     bits64 z;
71 
72     if ( count == 0 ) {
73         z = a;
74     }
75     else if ( count < 64 ) {
76         z = ( a>>count ) | ( ( a<<( ( - count ) & 63 ) ) != 0 );
77     }
78     else {
79         z = ( a != 0 );
80     }
81     *zPtr = z;
82 
83 }
84 
85 /*----------------------------------------------------------------------------
86 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by 64
87 | _plus_ the number of bits given in `count'.  The shifted result is at most
88 | 64 nonzero bits; this is stored at the location pointed to by `z0Ptr'.  The
89 | bits shifted off form a second 64-bit result as follows:  The _last_ bit
90 | shifted off is the most-significant bit of the extra result, and the other
91 | 63 bits of the extra result are all zero if and only if _all_but_the_last_
92 | bits shifted off were all zero.  This extra result is stored in the location
93 | pointed to by `z1Ptr'.  The value of `count' can be arbitrarily large.
94 |     (This routine makes more sense if `a0' and `a1' are considered to form
95 | a fixed-point value with binary point between `a0' and `a1'.  This fixed-
96 | point value is shifted right by the number of bits given in `count', and
97 | the integer part of the result is returned at the location pointed to by
98 | `z0Ptr'.  The fractional part of the result may be slightly corrupted as
99 | described above, and is returned at the location pointed to by `z1Ptr'.)
100 *----------------------------------------------------------------------------*/
101 
102 INLINE void
shift64ExtraRightJamming(bits64 a0,bits64 a1,int16 count,bits64 * z0Ptr,bits64 * z1Ptr)103  shift64ExtraRightJamming(
104      bits64 a0, bits64 a1, int16 count, bits64 *z0Ptr, bits64 *z1Ptr )
105 {
106     bits64 z0, z1;
107     int8 negCount = ( - count ) & 63;
108 
109     if ( count == 0 ) {
110         z1 = a1;
111         z0 = a0;
112     }
113     else if ( count < 64 ) {
114         z1 = ( a0<<negCount ) | ( a1 != 0 );
115         z0 = a0>>count;
116     }
117     else {
118         if ( count == 64 ) {
119             z1 = a0 | ( a1 != 0 );
120         }
121         else {
122             z1 = ( ( a0 | a1 ) != 0 );
123         }
124         z0 = 0;
125     }
126     *z1Ptr = z1;
127     *z0Ptr = z0;
128 
129 }
130 
131 /*----------------------------------------------------------------------------
132 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the
133 | number of bits given in `count'.  Any bits shifted off are lost.  The value
134 | of `count' can be arbitrarily large; in particular, if `count' is greater
135 | than 128, the result will be 0.  The result is broken into two 64-bit pieces
136 | which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
137 *----------------------------------------------------------------------------*/
138 
139 INLINE void
shift128Right(bits64 a0,bits64 a1,int16 count,bits64 * z0Ptr,bits64 * z1Ptr)140  shift128Right(
141      bits64 a0, bits64 a1, int16 count, bits64 *z0Ptr, bits64 *z1Ptr )
142 {
143     bits64 z0, z1;
144     int8 negCount = ( - count ) & 63;
145 
146     if ( count == 0 ) {
147         z1 = a1;
148         z0 = a0;
149     }
150     else if ( count < 64 ) {
151         z1 = ( a0<<negCount ) | ( a1>>count );
152         z0 = a0>>count;
153     }
154     else {
155         z1 = ( count < 64 ) ? ( a0>>( count & 63 ) ) : 0;
156         z0 = 0;
157     }
158     *z1Ptr = z1;
159     *z0Ptr = z0;
160 
161 }
162 
163 /*----------------------------------------------------------------------------
164 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the
165 | number of bits given in `count'.  If any nonzero bits are shifted off, they
166 | are ``jammed'' into the least significant bit of the result by setting the
167 | least significant bit to 1.  The value of `count' can be arbitrarily large;
168 | in particular, if `count' is greater than 128, the result will be either
169 | 0 or 1, depending on whether the concatenation of `a0' and `a1' is zero or
170 | nonzero.  The result is broken into two 64-bit pieces which are stored at
171 | the locations pointed to by `z0Ptr' and `z1Ptr'.
172 *----------------------------------------------------------------------------*/
173 
174 INLINE void
shift128RightJamming(bits64 a0,bits64 a1,int16 count,bits64 * z0Ptr,bits64 * z1Ptr)175  shift128RightJamming(
176      bits64 a0, bits64 a1, int16 count, bits64 *z0Ptr, bits64 *z1Ptr )
177 {
178     bits64 z0, z1;
179     int8 negCount = ( - count ) & 63;
180 
181     if ( count == 0 ) {
182         z1 = a1;
183         z0 = a0;
184     }
185     else if ( count < 64 ) {
186         z1 = ( a0<<negCount ) | ( a1>>count ) | ( ( a1<<negCount ) != 0 );
187         z0 = a0>>count;
188     }
189     else {
190         if ( count == 64 ) {
191             z1 = a0 | ( a1 != 0 );
192         }
193         else if ( count < 128 ) {
194             z1 = ( a0>>( count & 63 ) ) | ( ( ( a0<<negCount ) | a1 ) != 0 );
195         }
196         else {
197             z1 = ( ( a0 | a1 ) != 0 );
198         }
199         z0 = 0;
200     }
201     *z1Ptr = z1;
202     *z0Ptr = z0;
203 
204 }
205 
206 /*----------------------------------------------------------------------------
207 | Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' right
208 | by 64 _plus_ the number of bits given in `count'.  The shifted result is
209 | at most 128 nonzero bits; these are broken into two 64-bit pieces which are
210 | stored at the locations pointed to by `z0Ptr' and `z1Ptr'.  The bits shifted
211 | off form a third 64-bit result as follows:  The _last_ bit shifted off is
212 | the most-significant bit of the extra result, and the other 63 bits of the
213 | extra result are all zero if and only if _all_but_the_last_ bits shifted off
214 | were all zero.  This extra result is stored in the location pointed to by
215 | `z2Ptr'.  The value of `count' can be arbitrarily large.
216 |     (This routine makes more sense if `a0', `a1', and `a2' are considered
217 | to form a fixed-point value with binary point between `a1' and `a2'.  This
218 | fixed-point value is shifted right by the number of bits given in `count',
219 | and the integer part of the result is returned at the locations pointed to
220 | by `z0Ptr' and `z1Ptr'.  The fractional part of the result may be slightly
221 | corrupted as described above, and is returned at the location pointed to by
222 | `z2Ptr'.)
223 *----------------------------------------------------------------------------*/
224 
225 INLINE void
shift128ExtraRightJamming(bits64 a0,bits64 a1,bits64 a2,int16 count,bits64 * z0Ptr,bits64 * z1Ptr,bits64 * z2Ptr)226  shift128ExtraRightJamming(
227      bits64 a0,
228      bits64 a1,
229      bits64 a2,
230      int16 count,
231      bits64 *z0Ptr,
232      bits64 *z1Ptr,
233      bits64 *z2Ptr
234  )
235 {
236     bits64 z0, z1, z2;
237     int8 negCount = ( - count ) & 63;
238 
239     if ( count == 0 ) {
240         z2 = a2;
241         z1 = a1;
242         z0 = a0;
243     }
244     else {
245         if ( count < 64 ) {
246             z2 = a1<<negCount;
247             z1 = ( a0<<negCount ) | ( a1>>count );
248             z0 = a0>>count;
249         }
250         else {
251             if ( count == 64 ) {
252                 z2 = a1;
253                 z1 = a0;
254             }
255             else {
256                 a2 |= a1;
257                 if ( count < 128 ) {
258                     z2 = a0<<negCount;
259                     z1 = a0>>( count & 63 );
260                 }
261                 else {
262                     z2 = ( count == 128 ) ? a0 : ( a0 != 0 );
263                     z1 = 0;
264                 }
265             }
266             z0 = 0;
267         }
268         z2 |= ( a2 != 0 );
269     }
270     *z2Ptr = z2;
271     *z1Ptr = z1;
272     *z0Ptr = z0;
273 
274 }
275 
276 /*----------------------------------------------------------------------------
277 | Shifts the 128-bit value formed by concatenating `a0' and `a1' left by the
278 | number of bits given in `count'.  Any bits shifted off are lost.  The value
279 | of `count' must be less than 64.  The result is broken into two 64-bit
280 | pieces which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
281 *----------------------------------------------------------------------------*/
282 
283 INLINE void
shortShift128Left(bits64 a0,bits64 a1,int16 count,bits64 * z0Ptr,bits64 * z1Ptr)284  shortShift128Left(
285      bits64 a0, bits64 a1, int16 count, bits64 *z0Ptr, bits64 *z1Ptr )
286 {
287 
288     *z1Ptr = a1<<count;
289     *z0Ptr =
290         ( count == 0 ) ? a0 : ( a0<<count ) | ( a1>>( ( - count ) & 63 ) );
291 
292 }
293 
294 /*----------------------------------------------------------------------------
295 | Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' left
296 | by the number of bits given in `count'.  Any bits shifted off are lost.
297 | The value of `count' must be less than 64.  The result is broken into three
298 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr',
299 | `z1Ptr', and `z2Ptr'.
300 *----------------------------------------------------------------------------*/
301 
302 INLINE void
shortShift192Left(bits64 a0,bits64 a1,bits64 a2,int16 count,bits64 * z0Ptr,bits64 * z1Ptr,bits64 * z2Ptr)303  shortShift192Left(
304      bits64 a0,
305      bits64 a1,
306      bits64 a2,
307      int16 count,
308      bits64 *z0Ptr,
309      bits64 *z1Ptr,
310      bits64 *z2Ptr
311  )
312 {
313     bits64 z0, z1, z2;
314     int8 negCount;
315 
316     z2 = a2<<count;
317     z1 = a1<<count;
318     z0 = a0<<count;
319     if ( 0 < count ) {
320         negCount = ( ( - count ) & 63 );
321         z1 |= a2>>negCount;
322         z0 |= a1>>negCount;
323     }
324     *z2Ptr = z2;
325     *z1Ptr = z1;
326     *z0Ptr = z0;
327 
328 }
329 
330 /*----------------------------------------------------------------------------
331 | Adds the 128-bit value formed by concatenating `a0' and `a1' to the 128-bit
332 | value formed by concatenating `b0' and `b1'.  Addition is modulo 2^128, so
333 | any carry out is lost.  The result is broken into two 64-bit pieces which
334 | are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
335 *----------------------------------------------------------------------------*/
336 
337 INLINE void
add128(bits64 a0,bits64 a1,bits64 b0,bits64 b1,bits64 * z0Ptr,bits64 * z1Ptr)338  add128(
339      bits64 a0, bits64 a1, bits64 b0, bits64 b1, bits64 *z0Ptr, bits64 *z1Ptr )
340 {
341     bits64 z1;
342 
343     z1 = a1 + b1;
344     *z1Ptr = z1;
345     *z0Ptr = a0 + b0 + ( z1 < a1 );
346 
347 }
348 
349 /*----------------------------------------------------------------------------
350 | Adds the 192-bit value formed by concatenating `a0', `a1', and `a2' to the
351 | 192-bit value formed by concatenating `b0', `b1', and `b2'.  Addition is
352 | modulo 2^192, so any carry out is lost.  The result is broken into three
353 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr',
354 | `z1Ptr', and `z2Ptr'.
355 *----------------------------------------------------------------------------*/
356 
357 INLINE void
add192(bits64 a0,bits64 a1,bits64 a2,bits64 b0,bits64 b1,bits64 b2,bits64 * z0Ptr,bits64 * z1Ptr,bits64 * z2Ptr)358  add192(
359      bits64 a0,
360      bits64 a1,
361      bits64 a2,
362      bits64 b0,
363      bits64 b1,
364      bits64 b2,
365      bits64 *z0Ptr,
366      bits64 *z1Ptr,
367      bits64 *z2Ptr
368  )
369 {
370     bits64 z0, z1, z2;
371     int8 carry0, carry1;
372 
373     z2 = a2 + b2;
374     carry1 = ( z2 < a2 );
375     z1 = a1 + b1;
376     carry0 = ( z1 < a1 );
377     z0 = a0 + b0;
378     z1 += carry1;
379     z0 += ( z1 < carry1 );
380     z0 += carry0;
381     *z2Ptr = z2;
382     *z1Ptr = z1;
383     *z0Ptr = z0;
384 
385 }
386 
387 /*----------------------------------------------------------------------------
388 | Subtracts the 128-bit value formed by concatenating `b0' and `b1' from the
389 | 128-bit value formed by concatenating `a0' and `a1'.  Subtraction is modulo
390 | 2^128, so any borrow out (carry out) is lost.  The result is broken into two
391 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr' and
392 | `z1Ptr'.
393 *----------------------------------------------------------------------------*/
394 
395 INLINE void
sub128(bits64 a0,bits64 a1,bits64 b0,bits64 b1,bits64 * z0Ptr,bits64 * z1Ptr)396  sub128(
397      bits64 a0, bits64 a1, bits64 b0, bits64 b1, bits64 *z0Ptr, bits64 *z1Ptr )
398 {
399 
400     *z1Ptr = a1 - b1;
401     *z0Ptr = a0 - b0 - ( a1 < b1 );
402 
403 }
404 
405 /*----------------------------------------------------------------------------
406 | Subtracts the 192-bit value formed by concatenating `b0', `b1', and `b2'
407 | from the 192-bit value formed by concatenating `a0', `a1', and `a2'.
408 | Subtraction is modulo 2^192, so any borrow out (carry out) is lost.  The
409 | result is broken into three 64-bit pieces which are stored at the locations
410 | pointed to by `z0Ptr', `z1Ptr', and `z2Ptr'.
411 *----------------------------------------------------------------------------*/
412 
413 INLINE void
sub192(bits64 a0,bits64 a1,bits64 a2,bits64 b0,bits64 b1,bits64 b2,bits64 * z0Ptr,bits64 * z1Ptr,bits64 * z2Ptr)414  sub192(
415      bits64 a0,
416      bits64 a1,
417      bits64 a2,
418      bits64 b0,
419      bits64 b1,
420      bits64 b2,
421      bits64 *z0Ptr,
422      bits64 *z1Ptr,
423      bits64 *z2Ptr
424  )
425 {
426     bits64 z0, z1, z2;
427     int8 borrow0, borrow1;
428 
429     z2 = a2 - b2;
430     borrow1 = ( a2 < b2 );
431     z1 = a1 - b1;
432     borrow0 = ( a1 < b1 );
433     z0 = a0 - b0;
434     z0 -= ( z1 < borrow1 );
435     z1 -= borrow1;
436     z0 -= borrow0;
437     *z2Ptr = z2;
438     *z1Ptr = z1;
439     *z0Ptr = z0;
440 
441 }
442 
443 /*----------------------------------------------------------------------------
444 | Multiplies `a' by `b' to obtain a 128-bit product.  The product is broken
445 | into two 64-bit pieces which are stored at the locations pointed to by
446 | `z0Ptr' and `z1Ptr'.
447 *----------------------------------------------------------------------------*/
448 
mul64To128(bits64 a,bits64 b,bits64 * z0Ptr,bits64 * z1Ptr)449 INLINE void mul64To128( bits64 a, bits64 b, bits64 *z0Ptr, bits64 *z1Ptr )
450 {
451     bits32 aHigh, aLow, bHigh, bLow;
452     bits64 z0, zMiddleA, zMiddleB, z1;
453 
454     aLow = a;
455     aHigh = a>>32;
456     bLow = b;
457     bHigh = b>>32;
458     z1 = ( (bits64) aLow ) * bLow;
459     zMiddleA = ( (bits64) aLow ) * bHigh;
460     zMiddleB = ( (bits64) aHigh ) * bLow;
461     z0 = ( (bits64) aHigh ) * bHigh;
462     zMiddleA += zMiddleB;
463     z0 += ( ( (bits64) ( zMiddleA < zMiddleB ) )<<32 ) + ( zMiddleA>>32 );
464     zMiddleA <<= 32;
465     z1 += zMiddleA;
466     z0 += ( z1 < zMiddleA );
467     *z1Ptr = z1;
468     *z0Ptr = z0;
469 
470 }
471 
472 /*----------------------------------------------------------------------------
473 | Multiplies the 128-bit value formed by concatenating `a0' and `a1' by
474 | `b' to obtain a 192-bit product.  The product is broken into three 64-bit
475 | pieces which are stored at the locations pointed to by `z0Ptr', `z1Ptr', and
476 | `z2Ptr'.
477 *----------------------------------------------------------------------------*/
478 
479 INLINE void
mul128By64To192(bits64 a0,bits64 a1,bits64 b,bits64 * z0Ptr,bits64 * z1Ptr,bits64 * z2Ptr)480  mul128By64To192(
481      bits64 a0,
482      bits64 a1,
483      bits64 b,
484      bits64 *z0Ptr,
485      bits64 *z1Ptr,
486      bits64 *z2Ptr
487  )
488 {
489     bits64 z0, z1, z2, more1;
490 
491     mul64To128( a1, b, &z1, &z2 );
492     mul64To128( a0, b, &z0, &more1 );
493     add128( z0, more1, 0, z1, &z0, &z1 );
494     *z2Ptr = z2;
495     *z1Ptr = z1;
496     *z0Ptr = z0;
497 
498 }
499 
500 /*----------------------------------------------------------------------------
501 | Multiplies the 128-bit value formed by concatenating `a0' and `a1' to the
502 | 128-bit value formed by concatenating `b0' and `b1' to obtain a 256-bit
503 | product.  The product is broken into four 64-bit pieces which are stored at
504 | the locations pointed to by `z0Ptr', `z1Ptr', `z2Ptr', and `z3Ptr'.
505 *----------------------------------------------------------------------------*/
506 
507 INLINE void
mul128To256(bits64 a0,bits64 a1,bits64 b0,bits64 b1,bits64 * z0Ptr,bits64 * z1Ptr,bits64 * z2Ptr,bits64 * z3Ptr)508  mul128To256(
509      bits64 a0,
510      bits64 a1,
511      bits64 b0,
512      bits64 b1,
513      bits64 *z0Ptr,
514      bits64 *z1Ptr,
515      bits64 *z2Ptr,
516      bits64 *z3Ptr
517  )
518 {
519     bits64 z0, z1, z2, z3;
520     bits64 more1, more2;
521 
522     mul64To128( a1, b1, &z2, &z3 );
523     mul64To128( a1, b0, &z1, &more2 );
524     add128( z1, more2, 0, z2, &z1, &z2 );
525     mul64To128( a0, b0, &z0, &more1 );
526     add128( z0, more1, 0, z1, &z0, &z1 );
527     mul64To128( a0, b1, &more1, &more2 );
528     add128( more1, more2, 0, z2, &more1, &z2 );
529     add128( z0, z1, 0, more1, &z0, &z1 );
530     *z3Ptr = z3;
531     *z2Ptr = z2;
532     *z1Ptr = z1;
533     *z0Ptr = z0;
534 
535 }
536 
537 /*----------------------------------------------------------------------------
538 | Returns an approximation to the 64-bit integer quotient obtained by dividing
539 | `b' into the 128-bit value formed by concatenating `a0' and `a1'.  The
540 | divisor `b' must be at least 2^63.  If q is the exact quotient truncated
541 | toward zero, the approximation returned lies between q and q + 2 inclusive.
542 | If the exact quotient q is larger than 64 bits, the maximum positive 64-bit
543 | unsigned integer is returned.
544 *----------------------------------------------------------------------------*/
545 
estimateDiv128To64(bits64 a0,bits64 a1,bits64 b)546 static bits64 estimateDiv128To64( bits64 a0, bits64 a1, bits64 b )
547 {
548     bits64 b0, b1;
549     bits64 rem0, rem1, term0, term1;
550     bits64 z;
551 
552     if ( b <= a0 ) return LIT64( 0xFFFFFFFFFFFFFFFF );
553     b0 = b>>32;
554     z = ( b0<<32 <= a0 ) ? LIT64( 0xFFFFFFFF00000000 ) : ( a0 / b0 )<<32;
555     mul64To128( b, z, &term0, &term1 );
556     sub128( a0, a1, term0, term1, &rem0, &rem1 );
557     while ( ( (sbits64) rem0 ) < 0 ) {
558         z -= LIT64( 0x100000000 );
559         b1 = b<<32;
560         add128( rem0, rem1, b0, b1, &rem0, &rem1 );
561     }
562     rem0 = ( rem0<<32 ) | ( rem1>>32 );
563     z |= ( b0<<32 <= rem0 ) ? 0xFFFFFFFF : rem0 / b0;
564     return z;
565 
566 }
567 
568 /*----------------------------------------------------------------------------
569 | Returns an approximation to the square root of the 32-bit significand given
570 | by `a'.  Considered as an integer, `a' must be at least 2^31.  If bit 0 of
571 | `aExp' (the least significant bit) is 1, the integer returned approximates
572 | 2^31*sqrt(`a'/2^31), where `a' is considered an integer.  If bit 0 of `aExp'
573 | is 0, the integer returned approximates 2^31*sqrt(`a'/2^30).  In either
574 | case, the approximation returned lies strictly within +/-2 of the exact
575 | value.
576 *----------------------------------------------------------------------------*/
577 
estimateSqrt32(int16 aExp,bits32 a)578 static bits32 estimateSqrt32( int16 aExp, bits32 a )
579 {
580     static const bits16 sqrtOddAdjustments[] = {
581         0x0004, 0x0022, 0x005D, 0x00B1, 0x011D, 0x019F, 0x0236, 0x02E0,
582         0x039C, 0x0468, 0x0545, 0x0631, 0x072B, 0x0832, 0x0946, 0x0A67
583     };
584     static const bits16 sqrtEvenAdjustments[] = {
585         0x0A2D, 0x08AF, 0x075A, 0x0629, 0x051A, 0x0429, 0x0356, 0x029E,
586         0x0200, 0x0179, 0x0109, 0x00AF, 0x0068, 0x0034, 0x0012, 0x0002
587     };
588     int8 index;
589     bits32 z;
590 
591     index = ( a>>27 ) & 15;
592     if ( aExp & 1 ) {
593         z = 0x4000 + ( a>>17 ) - sqrtOddAdjustments[ (int)index ];
594         z = ( ( a / z )<<14 ) + ( z<<15 );
595         a >>= 1;
596     }
597     else {
598         z = 0x8000 + ( a>>17 ) - sqrtEvenAdjustments[ (int)index ];
599         z = a / z + z;
600         z = ( 0x20000 <= z ) ? 0xFFFF8000 : ( z<<15 );
601         if ( z <= a ) return (bits32) ( ( (sbits32) a )>>1 );
602     }
603     return ( (bits32) ( ( ( (bits64) a )<<31 ) / z ) ) + ( z>>1 );
604 
605 }
606 
607 /*----------------------------------------------------------------------------
608 | Returns the number of leading 0 bits before the most-significant 1 bit of
609 | `a'.  If `a' is zero, 32 is returned.
610 *----------------------------------------------------------------------------*/
611 
countLeadingZeros32(bits32 a)612 static int8 countLeadingZeros32( bits32 a )
613 {
614     static const int8 countLeadingZerosHigh[] = {
615         8, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4,
616         3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
617         2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
618         2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
619         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
620         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
621         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
622         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
623         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
624         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
625         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
626         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
627         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
628         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
629         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
630         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
631     };
632     int8 shiftCount;
633 
634     shiftCount = 0;
635     if ( a < 0x10000 ) {
636         shiftCount += 16;
637         a <<= 16;
638     }
639     if ( a < 0x1000000 ) {
640         shiftCount += 8;
641         a <<= 8;
642     }
643     shiftCount += countLeadingZerosHigh[ a>>24 ];
644     return shiftCount;
645 
646 }
647 
648 /*----------------------------------------------------------------------------
649 | Returns the number of leading 0 bits before the most-significant 1 bit of
650 | `a'.  If `a' is zero, 64 is returned.
651 *----------------------------------------------------------------------------*/
652 
countLeadingZeros64(bits64 a)653 static int8 countLeadingZeros64( bits64 a )
654 {
655     int8 shiftCount;
656 
657     shiftCount = 0;
658     if ( a < ( (bits64) 1 )<<32 ) {
659         shiftCount += 32;
660     }
661     else {
662         a >>= 32;
663     }
664     shiftCount += countLeadingZeros32( a );
665     return shiftCount;
666 
667 }
668 
669 /*----------------------------------------------------------------------------
670 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1'
671 | is equal to the 128-bit value formed by concatenating `b0' and `b1'.
672 | Otherwise, returns 0.
673 *----------------------------------------------------------------------------*/
674 
eq128(bits64 a0,bits64 a1,bits64 b0,bits64 b1)675 INLINE flag eq128( bits64 a0, bits64 a1, bits64 b0, bits64 b1 )
676 {
677 
678     return ( a0 == b0 ) && ( a1 == b1 );
679 
680 }
681 
682 /*----------------------------------------------------------------------------
683 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less
684 | than or equal to the 128-bit value formed by concatenating `b0' and `b1'.
685 | Otherwise, returns 0.
686 *----------------------------------------------------------------------------*/
687 
le128(bits64 a0,bits64 a1,bits64 b0,bits64 b1)688 INLINE flag le128( bits64 a0, bits64 a1, bits64 b0, bits64 b1 )
689 {
690 
691     return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 <= b1 ) );
692 
693 }
694 
695 /*----------------------------------------------------------------------------
696 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less
697 | than the 128-bit value formed by concatenating `b0' and `b1'.  Otherwise,
698 | returns 0.
699 *----------------------------------------------------------------------------*/
700 
lt128(bits64 a0,bits64 a1,bits64 b0,bits64 b1)701 INLINE flag lt128( bits64 a0, bits64 a1, bits64 b0, bits64 b1 )
702 {
703 
704     return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 < b1 ) );
705 
706 }
707 
708 /*----------------------------------------------------------------------------
709 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is
710 | not equal to the 128-bit value formed by concatenating `b0' and `b1'.
711 | Otherwise, returns 0.
712 *----------------------------------------------------------------------------*/
713 
ne128(bits64 a0,bits64 a1,bits64 b0,bits64 b1)714 INLINE flag ne128( bits64 a0, bits64 a1, bits64 b0, bits64 b1 )
715 {
716 
717     return ( a0 != b0 ) || ( a1 != b1 );
718 
719 }
720