root / fpu / softfloat-macros.h @ 0e326109
History | View | Annotate | Download (24.2 kB)
1 | 8d725fac | Andreas Färber | /*
|
---|---|---|---|
2 | 8d725fac | Andreas Färber | * QEMU float support macros
|
3 | 8d725fac | Andreas Färber | *
|
4 | 8d725fac | Andreas Färber | * Derived from SoftFloat.
|
5 | 8d725fac | Andreas Färber | */
|
6 | 158142c2 | bellard | |
7 | 158142c2 | bellard | /*============================================================================
|
8 | 158142c2 | bellard | |
9 | 158142c2 | bellard | This C source fragment is part of the SoftFloat IEC/IEEE Floating-point
|
10 | 158142c2 | bellard | Arithmetic Package, Release 2b.
|
11 | 158142c2 | bellard | |
12 | 158142c2 | bellard | Written by John R. Hauser. This work was made possible in part by the
|
13 | 158142c2 | bellard | International Computer Science Institute, located at Suite 600, 1947 Center
|
14 | 158142c2 | bellard | Street, Berkeley, California 94704. Funding was partially provided by the
|
15 | 158142c2 | bellard | National Science Foundation under grant MIP-9311980. The original version
|
16 | 158142c2 | bellard | of this code was written as part of a project to build a fixed-point vector
|
17 | 158142c2 | bellard | processor in collaboration with the University of California at Berkeley,
|
18 | 158142c2 | bellard | overseen by Profs. Nelson Morgan and John Wawrzynek. More information
|
19 | 158142c2 | bellard | is available through the Web page `http://www.cs.berkeley.edu/~jhauser/
|
20 | 158142c2 | bellard | arithmetic/SoftFloat.html'.
|
21 | 158142c2 | bellard | |
22 | 158142c2 | bellard | THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort has
|
23 | 158142c2 | bellard | been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT TIMES
|
24 | 158142c2 | bellard | RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO PERSONS
|
25 | 158142c2 | bellard | AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ALL LOSSES,
|
26 | 158142c2 | bellard | COSTS, OR OTHER PROBLEMS THEY INCUR DUE TO THE SOFTWARE, AND WHO FURTHERMORE
|
27 | 158142c2 | bellard | EFFECTIVELY INDEMNIFY JOHN HAUSER AND THE INTERNATIONAL COMPUTER SCIENCE
|
28 | 158142c2 | bellard | INSTITUTE (possibly via similar legal notice) AGAINST ALL LOSSES, COSTS, OR
|
29 | 158142c2 | bellard | OTHER PROBLEMS INCURRED BY THEIR CUSTOMERS AND CLIENTS DUE TO THE SOFTWARE.
|
30 | 158142c2 | bellard | |
31 | 158142c2 | bellard | Derivative works are acceptable, even for commercial purposes, so long as
|
32 | 158142c2 | bellard | (1) the source code for the derivative work includes prominent notice that
|
33 | 158142c2 | bellard | the work is derivative, and (2) the source code includes prominent notice with
|
34 | 158142c2 | bellard | these four paragraphs for those parts of this code that are retained.
|
35 | 158142c2 | bellard | |
36 | 158142c2 | bellard | =============================================================================*/
|
37 | 158142c2 | bellard | |
38 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
39 | 158142c2 | bellard | | Shifts `a' right by the number of bits given in `count'. If any nonzero
|
40 | 158142c2 | bellard | | bits are shifted off, they are ``jammed'' into the least significant bit of
|
41 | 158142c2 | bellard | | the result by setting the least significant bit to 1. The value of `count'
|
42 | 158142c2 | bellard | | can be arbitrarily large; in particular, if `count' is greater than 32, the
|
43 | 158142c2 | bellard | | result will be either 0 or 1, depending on whether `a' is zero or nonzero.
|
44 | 158142c2 | bellard | | The result is stored in the location pointed to by `zPtr'.
|
45 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
46 | 158142c2 | bellard | |
47 | bb98fe42 | Andreas Färber | INLINE void shift32RightJamming( uint32_t a, int16 count, uint32_t *zPtr )
|
48 | 158142c2 | bellard | { |
49 | bb98fe42 | Andreas Färber | uint32_t z; |
50 | 158142c2 | bellard | |
51 | 158142c2 | bellard | if ( count == 0 ) { |
52 | 158142c2 | bellard | z = a; |
53 | 158142c2 | bellard | } |
54 | 158142c2 | bellard | else if ( count < 32 ) { |
55 | 158142c2 | bellard | z = ( a>>count ) | ( ( a<<( ( - count ) & 31 ) ) != 0 ); |
56 | 158142c2 | bellard | } |
57 | 158142c2 | bellard | else {
|
58 | 158142c2 | bellard | z = ( a != 0 );
|
59 | 158142c2 | bellard | } |
60 | 158142c2 | bellard | *zPtr = z; |
61 | 158142c2 | bellard | |
62 | 158142c2 | bellard | } |
63 | 158142c2 | bellard | |
64 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
65 | 158142c2 | bellard | | Shifts `a' right by the number of bits given in `count'. If any nonzero
|
66 | 158142c2 | bellard | | bits are shifted off, they are ``jammed'' into the least significant bit of
|
67 | 158142c2 | bellard | | the result by setting the least significant bit to 1. The value of `count'
|
68 | 158142c2 | bellard | | can be arbitrarily large; in particular, if `count' is greater than 64, the
|
69 | 158142c2 | bellard | | result will be either 0 or 1, depending on whether `a' is zero or nonzero.
|
70 | 158142c2 | bellard | | The result is stored in the location pointed to by `zPtr'.
|
71 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
72 | 158142c2 | bellard | |
73 | bb98fe42 | Andreas Färber | INLINE void shift64RightJamming( uint64_t a, int16 count, uint64_t *zPtr )
|
74 | 158142c2 | bellard | { |
75 | bb98fe42 | Andreas Färber | uint64_t z; |
76 | 158142c2 | bellard | |
77 | 158142c2 | bellard | if ( count == 0 ) { |
78 | 158142c2 | bellard | z = a; |
79 | 158142c2 | bellard | } |
80 | 158142c2 | bellard | else if ( count < 64 ) { |
81 | 158142c2 | bellard | z = ( a>>count ) | ( ( a<<( ( - count ) & 63 ) ) != 0 ); |
82 | 158142c2 | bellard | } |
83 | 158142c2 | bellard | else {
|
84 | 158142c2 | bellard | z = ( a != 0 );
|
85 | 158142c2 | bellard | } |
86 | 158142c2 | bellard | *zPtr = z; |
87 | 158142c2 | bellard | |
88 | 158142c2 | bellard | } |
89 | 158142c2 | bellard | |
90 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
91 | 158142c2 | bellard | | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by 64
|
92 | 158142c2 | bellard | | _plus_ the number of bits given in `count'. The shifted result is at most
|
93 | 158142c2 | bellard | | 64 nonzero bits; this is stored at the location pointed to by `z0Ptr'. The
|
94 | 158142c2 | bellard | | bits shifted off form a second 64-bit result as follows: The _last_ bit
|
95 | 158142c2 | bellard | | shifted off is the most-significant bit of the extra result, and the other
|
96 | 158142c2 | bellard | | 63 bits of the extra result are all zero if and only if _all_but_the_last_
|
97 | 158142c2 | bellard | | bits shifted off were all zero. This extra result is stored in the location
|
98 | 158142c2 | bellard | | pointed to by `z1Ptr'. The value of `count' can be arbitrarily large.
|
99 | 158142c2 | bellard | | (This routine makes more sense if `a0' and `a1' are considered to form
|
100 | 158142c2 | bellard | | a fixed-point value with binary point between `a0' and `a1'. This fixed-
|
101 | 158142c2 | bellard | | point value is shifted right by the number of bits given in `count', and
|
102 | 158142c2 | bellard | | the integer part of the result is returned at the location pointed to by
|
103 | 158142c2 | bellard | | `z0Ptr'. The fractional part of the result may be slightly corrupted as
|
104 | 158142c2 | bellard | | described above, and is returned at the location pointed to by `z1Ptr'.)
|
105 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
106 | 158142c2 | bellard | |
107 | 158142c2 | bellard | INLINE void
|
108 | 158142c2 | bellard | shift64ExtraRightJamming( |
109 | bb98fe42 | Andreas Färber | uint64_t a0, uint64_t a1, int16 count, uint64_t *z0Ptr, uint64_t *z1Ptr ) |
110 | 158142c2 | bellard | { |
111 | bb98fe42 | Andreas Färber | uint64_t z0, z1; |
112 | 158142c2 | bellard | int8 negCount = ( - count ) & 63;
|
113 | 158142c2 | bellard | |
114 | 158142c2 | bellard | if ( count == 0 ) { |
115 | 158142c2 | bellard | z1 = a1; |
116 | 158142c2 | bellard | z0 = a0; |
117 | 158142c2 | bellard | } |
118 | 158142c2 | bellard | else if ( count < 64 ) { |
119 | 158142c2 | bellard | z1 = ( a0<<negCount ) | ( a1 != 0 );
|
120 | 158142c2 | bellard | z0 = a0>>count; |
121 | 158142c2 | bellard | } |
122 | 158142c2 | bellard | else {
|
123 | 158142c2 | bellard | if ( count == 64 ) { |
124 | 158142c2 | bellard | z1 = a0 | ( a1 != 0 );
|
125 | 158142c2 | bellard | } |
126 | 158142c2 | bellard | else {
|
127 | 158142c2 | bellard | z1 = ( ( a0 | a1 ) != 0 );
|
128 | 158142c2 | bellard | } |
129 | 158142c2 | bellard | z0 = 0;
|
130 | 158142c2 | bellard | } |
131 | 158142c2 | bellard | *z1Ptr = z1; |
132 | 158142c2 | bellard | *z0Ptr = z0; |
133 | 158142c2 | bellard | |
134 | 158142c2 | bellard | } |
135 | 158142c2 | bellard | |
136 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
137 | 158142c2 | bellard | | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the
|
138 | 158142c2 | bellard | | number of bits given in `count'. Any bits shifted off are lost. The value
|
139 | 158142c2 | bellard | | of `count' can be arbitrarily large; in particular, if `count' is greater
|
140 | 158142c2 | bellard | | than 128, the result will be 0. The result is broken into two 64-bit pieces
|
141 | 158142c2 | bellard | | which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
|
142 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
143 | 158142c2 | bellard | |
144 | 158142c2 | bellard | INLINE void
|
145 | 158142c2 | bellard | shift128Right( |
146 | bb98fe42 | Andreas Färber | uint64_t a0, uint64_t a1, int16 count, uint64_t *z0Ptr, uint64_t *z1Ptr ) |
147 | 158142c2 | bellard | { |
148 | bb98fe42 | Andreas Färber | uint64_t z0, z1; |
149 | 158142c2 | bellard | int8 negCount = ( - count ) & 63;
|
150 | 158142c2 | bellard | |
151 | 158142c2 | bellard | if ( count == 0 ) { |
152 | 158142c2 | bellard | z1 = a1; |
153 | 158142c2 | bellard | z0 = a0; |
154 | 158142c2 | bellard | } |
155 | 158142c2 | bellard | else if ( count < 64 ) { |
156 | 158142c2 | bellard | z1 = ( a0<<negCount ) | ( a1>>count ); |
157 | 158142c2 | bellard | z0 = a0>>count; |
158 | 158142c2 | bellard | } |
159 | 158142c2 | bellard | else {
|
160 | 158142c2 | bellard | z1 = ( count < 64 ) ? ( a0>>( count & 63 ) ) : 0; |
161 | 158142c2 | bellard | z0 = 0;
|
162 | 158142c2 | bellard | } |
163 | 158142c2 | bellard | *z1Ptr = z1; |
164 | 158142c2 | bellard | *z0Ptr = z0; |
165 | 158142c2 | bellard | |
166 | 158142c2 | bellard | } |
167 | 158142c2 | bellard | |
168 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
169 | 158142c2 | bellard | | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the
|
170 | 158142c2 | bellard | | number of bits given in `count'. If any nonzero bits are shifted off, they
|
171 | 158142c2 | bellard | | are ``jammed'' into the least significant bit of the result by setting the
|
172 | 158142c2 | bellard | | least significant bit to 1. The value of `count' can be arbitrarily large;
|
173 | 158142c2 | bellard | | in particular, if `count' is greater than 128, the result will be either
|
174 | 158142c2 | bellard | | 0 or 1, depending on whether the concatenation of `a0' and `a1' is zero or
|
175 | 158142c2 | bellard | | nonzero. The result is broken into two 64-bit pieces which are stored at
|
176 | 158142c2 | bellard | | the locations pointed to by `z0Ptr' and `z1Ptr'.
|
177 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
178 | 158142c2 | bellard | |
179 | 158142c2 | bellard | INLINE void
|
180 | 158142c2 | bellard | shift128RightJamming( |
181 | bb98fe42 | Andreas Färber | uint64_t a0, uint64_t a1, int16 count, uint64_t *z0Ptr, uint64_t *z1Ptr ) |
182 | 158142c2 | bellard | { |
183 | bb98fe42 | Andreas Färber | uint64_t z0, z1; |
184 | 158142c2 | bellard | int8 negCount = ( - count ) & 63;
|
185 | 158142c2 | bellard | |
186 | 158142c2 | bellard | if ( count == 0 ) { |
187 | 158142c2 | bellard | z1 = a1; |
188 | 158142c2 | bellard | z0 = a0; |
189 | 158142c2 | bellard | } |
190 | 158142c2 | bellard | else if ( count < 64 ) { |
191 | 158142c2 | bellard | z1 = ( a0<<negCount ) | ( a1>>count ) | ( ( a1<<negCount ) != 0 );
|
192 | 158142c2 | bellard | z0 = a0>>count; |
193 | 158142c2 | bellard | } |
194 | 158142c2 | bellard | else {
|
195 | 158142c2 | bellard | if ( count == 64 ) { |
196 | 158142c2 | bellard | z1 = a0 | ( a1 != 0 );
|
197 | 158142c2 | bellard | } |
198 | 158142c2 | bellard | else if ( count < 128 ) { |
199 | 158142c2 | bellard | z1 = ( a0>>( count & 63 ) ) | ( ( ( a0<<negCount ) | a1 ) != 0 ); |
200 | 158142c2 | bellard | } |
201 | 158142c2 | bellard | else {
|
202 | 158142c2 | bellard | z1 = ( ( a0 | a1 ) != 0 );
|
203 | 158142c2 | bellard | } |
204 | 158142c2 | bellard | z0 = 0;
|
205 | 158142c2 | bellard | } |
206 | 158142c2 | bellard | *z1Ptr = z1; |
207 | 158142c2 | bellard | *z0Ptr = z0; |
208 | 158142c2 | bellard | |
209 | 158142c2 | bellard | } |
210 | 158142c2 | bellard | |
211 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
212 | 158142c2 | bellard | | Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' right
|
213 | 158142c2 | bellard | | by 64 _plus_ the number of bits given in `count'. The shifted result is
|
214 | 158142c2 | bellard | | at most 128 nonzero bits; these are broken into two 64-bit pieces which are
|
215 | 158142c2 | bellard | | stored at the locations pointed to by `z0Ptr' and `z1Ptr'. The bits shifted
|
216 | 158142c2 | bellard | | off form a third 64-bit result as follows: The _last_ bit shifted off is
|
217 | 158142c2 | bellard | | the most-significant bit of the extra result, and the other 63 bits of the
|
218 | 158142c2 | bellard | | extra result are all zero if and only if _all_but_the_last_ bits shifted off
|
219 | 158142c2 | bellard | | were all zero. This extra result is stored in the location pointed to by
|
220 | 158142c2 | bellard | | `z2Ptr'. The value of `count' can be arbitrarily large.
|
221 | 158142c2 | bellard | | (This routine makes more sense if `a0', `a1', and `a2' are considered
|
222 | 158142c2 | bellard | | to form a fixed-point value with binary point between `a1' and `a2'. This
|
223 | 158142c2 | bellard | | fixed-point value is shifted right by the number of bits given in `count',
|
224 | 158142c2 | bellard | | and the integer part of the result is returned at the locations pointed to
|
225 | 158142c2 | bellard | | by `z0Ptr' and `z1Ptr'. The fractional part of the result may be slightly
|
226 | 158142c2 | bellard | | corrupted as described above, and is returned at the location pointed to by
|
227 | 158142c2 | bellard | | `z2Ptr'.)
|
228 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
229 | 158142c2 | bellard | |
230 | 158142c2 | bellard | INLINE void
|
231 | 158142c2 | bellard | shift128ExtraRightJamming( |
232 | bb98fe42 | Andreas Färber | uint64_t a0, |
233 | bb98fe42 | Andreas Färber | uint64_t a1, |
234 | bb98fe42 | Andreas Färber | uint64_t a2, |
235 | 158142c2 | bellard | int16 count, |
236 | bb98fe42 | Andreas Färber | uint64_t *z0Ptr, |
237 | bb98fe42 | Andreas Färber | uint64_t *z1Ptr, |
238 | bb98fe42 | Andreas Färber | uint64_t *z2Ptr |
239 | 158142c2 | bellard | ) |
240 | 158142c2 | bellard | { |
241 | bb98fe42 | Andreas Färber | uint64_t z0, z1, z2; |
242 | 158142c2 | bellard | int8 negCount = ( - count ) & 63;
|
243 | 158142c2 | bellard | |
244 | 158142c2 | bellard | if ( count == 0 ) { |
245 | 158142c2 | bellard | z2 = a2; |
246 | 158142c2 | bellard | z1 = a1; |
247 | 158142c2 | bellard | z0 = a0; |
248 | 158142c2 | bellard | } |
249 | 158142c2 | bellard | else {
|
250 | 158142c2 | bellard | if ( count < 64 ) { |
251 | 158142c2 | bellard | z2 = a1<<negCount; |
252 | 158142c2 | bellard | z1 = ( a0<<negCount ) | ( a1>>count ); |
253 | 158142c2 | bellard | z0 = a0>>count; |
254 | 158142c2 | bellard | } |
255 | 158142c2 | bellard | else {
|
256 | 158142c2 | bellard | if ( count == 64 ) { |
257 | 158142c2 | bellard | z2 = a1; |
258 | 158142c2 | bellard | z1 = a0; |
259 | 158142c2 | bellard | } |
260 | 158142c2 | bellard | else {
|
261 | 158142c2 | bellard | a2 |= a1; |
262 | 158142c2 | bellard | if ( count < 128 ) { |
263 | 158142c2 | bellard | z2 = a0<<negCount; |
264 | 158142c2 | bellard | z1 = a0>>( count & 63 );
|
265 | 158142c2 | bellard | } |
266 | 158142c2 | bellard | else {
|
267 | 158142c2 | bellard | z2 = ( count == 128 ) ? a0 : ( a0 != 0 ); |
268 | 158142c2 | bellard | z1 = 0;
|
269 | 158142c2 | bellard | } |
270 | 158142c2 | bellard | } |
271 | 158142c2 | bellard | z0 = 0;
|
272 | 158142c2 | bellard | } |
273 | 158142c2 | bellard | z2 |= ( a2 != 0 );
|
274 | 158142c2 | bellard | } |
275 | 158142c2 | bellard | *z2Ptr = z2; |
276 | 158142c2 | bellard | *z1Ptr = z1; |
277 | 158142c2 | bellard | *z0Ptr = z0; |
278 | 158142c2 | bellard | |
279 | 158142c2 | bellard | } |
280 | 158142c2 | bellard | |
281 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
282 | 158142c2 | bellard | | Shifts the 128-bit value formed by concatenating `a0' and `a1' left by the
|
283 | 158142c2 | bellard | | number of bits given in `count'. Any bits shifted off are lost. The value
|
284 | 158142c2 | bellard | | of `count' must be less than 64. The result is broken into two 64-bit
|
285 | 158142c2 | bellard | | pieces which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
|
286 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
287 | 158142c2 | bellard | |
288 | 158142c2 | bellard | INLINE void
|
289 | 158142c2 | bellard | shortShift128Left( |
290 | bb98fe42 | Andreas Färber | uint64_t a0, uint64_t a1, int16 count, uint64_t *z0Ptr, uint64_t *z1Ptr ) |
291 | 158142c2 | bellard | { |
292 | 158142c2 | bellard | |
293 | 158142c2 | bellard | *z1Ptr = a1<<count; |
294 | 158142c2 | bellard | *z0Ptr = |
295 | 158142c2 | bellard | ( count == 0 ) ? a0 : ( a0<<count ) | ( a1>>( ( - count ) & 63 ) ); |
296 | 158142c2 | bellard | |
297 | 158142c2 | bellard | } |
298 | 158142c2 | bellard | |
299 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
300 | 158142c2 | bellard | | Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' left
|
301 | 158142c2 | bellard | | by the number of bits given in `count'. Any bits shifted off are lost.
|
302 | 158142c2 | bellard | | The value of `count' must be less than 64. The result is broken into three
|
303 | 158142c2 | bellard | | 64-bit pieces which are stored at the locations pointed to by `z0Ptr',
|
304 | 158142c2 | bellard | | `z1Ptr', and `z2Ptr'.
|
305 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
306 | 158142c2 | bellard | |
307 | 158142c2 | bellard | INLINE void
|
308 | 158142c2 | bellard | shortShift192Left( |
309 | bb98fe42 | Andreas Färber | uint64_t a0, |
310 | bb98fe42 | Andreas Färber | uint64_t a1, |
311 | bb98fe42 | Andreas Färber | uint64_t a2, |
312 | 158142c2 | bellard | int16 count, |
313 | bb98fe42 | Andreas Färber | uint64_t *z0Ptr, |
314 | bb98fe42 | Andreas Färber | uint64_t *z1Ptr, |
315 | bb98fe42 | Andreas Färber | uint64_t *z2Ptr |
316 | 158142c2 | bellard | ) |
317 | 158142c2 | bellard | { |
318 | bb98fe42 | Andreas Färber | uint64_t z0, z1, z2; |
319 | 158142c2 | bellard | int8 negCount; |
320 | 158142c2 | bellard | |
321 | 158142c2 | bellard | z2 = a2<<count; |
322 | 158142c2 | bellard | z1 = a1<<count; |
323 | 158142c2 | bellard | z0 = a0<<count; |
324 | 158142c2 | bellard | if ( 0 < count ) { |
325 | 158142c2 | bellard | negCount = ( ( - count ) & 63 );
|
326 | 158142c2 | bellard | z1 |= a2>>negCount; |
327 | 158142c2 | bellard | z0 |= a1>>negCount; |
328 | 158142c2 | bellard | } |
329 | 158142c2 | bellard | *z2Ptr = z2; |
330 | 158142c2 | bellard | *z1Ptr = z1; |
331 | 158142c2 | bellard | *z0Ptr = z0; |
332 | 158142c2 | bellard | |
333 | 158142c2 | bellard | } |
334 | 158142c2 | bellard | |
335 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
336 | 158142c2 | bellard | | Adds the 128-bit value formed by concatenating `a0' and `a1' to the 128-bit
|
337 | 158142c2 | bellard | | value formed by concatenating `b0' and `b1'. Addition is modulo 2^128, so
|
338 | 158142c2 | bellard | | any carry out is lost. The result is broken into two 64-bit pieces which
|
339 | 158142c2 | bellard | | are stored at the locations pointed to by `z0Ptr' and `z1Ptr'.
|
340 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
341 | 158142c2 | bellard | |
342 | 158142c2 | bellard | INLINE void
|
343 | 158142c2 | bellard | add128( |
344 | bb98fe42 | Andreas Färber | uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1, uint64_t *z0Ptr, uint64_t *z1Ptr ) |
345 | 158142c2 | bellard | { |
346 | bb98fe42 | Andreas Färber | uint64_t z1; |
347 | 158142c2 | bellard | |
348 | 158142c2 | bellard | z1 = a1 + b1; |
349 | 158142c2 | bellard | *z1Ptr = z1; |
350 | 158142c2 | bellard | *z0Ptr = a0 + b0 + ( z1 < a1 ); |
351 | 158142c2 | bellard | |
352 | 158142c2 | bellard | } |
353 | 158142c2 | bellard | |
354 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
355 | 158142c2 | bellard | | Adds the 192-bit value formed by concatenating `a0', `a1', and `a2' to the
|
356 | 158142c2 | bellard | | 192-bit value formed by concatenating `b0', `b1', and `b2'. Addition is
|
357 | 158142c2 | bellard | | modulo 2^192, so any carry out is lost. The result is broken into three
|
358 | 158142c2 | bellard | | 64-bit pieces which are stored at the locations pointed to by `z0Ptr',
|
359 | 158142c2 | bellard | | `z1Ptr', and `z2Ptr'.
|
360 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
361 | 158142c2 | bellard | |
362 | 158142c2 | bellard | INLINE void
|
363 | 158142c2 | bellard | add192( |
364 | bb98fe42 | Andreas Färber | uint64_t a0, |
365 | bb98fe42 | Andreas Färber | uint64_t a1, |
366 | bb98fe42 | Andreas Färber | uint64_t a2, |
367 | bb98fe42 | Andreas Färber | uint64_t b0, |
368 | bb98fe42 | Andreas Färber | uint64_t b1, |
369 | bb98fe42 | Andreas Färber | uint64_t b2, |
370 | bb98fe42 | Andreas Färber | uint64_t *z0Ptr, |
371 | bb98fe42 | Andreas Färber | uint64_t *z1Ptr, |
372 | bb98fe42 | Andreas Färber | uint64_t *z2Ptr |
373 | 158142c2 | bellard | ) |
374 | 158142c2 | bellard | { |
375 | bb98fe42 | Andreas Färber | uint64_t z0, z1, z2; |
376 | 158142c2 | bellard | int8 carry0, carry1; |
377 | 158142c2 | bellard | |
378 | 158142c2 | bellard | z2 = a2 + b2; |
379 | 158142c2 | bellard | carry1 = ( z2 < a2 ); |
380 | 158142c2 | bellard | z1 = a1 + b1; |
381 | 158142c2 | bellard | carry0 = ( z1 < a1 ); |
382 | 158142c2 | bellard | z0 = a0 + b0; |
383 | 158142c2 | bellard | z1 += carry1; |
384 | 158142c2 | bellard | z0 += ( z1 < carry1 ); |
385 | 158142c2 | bellard | z0 += carry0; |
386 | 158142c2 | bellard | *z2Ptr = z2; |
387 | 158142c2 | bellard | *z1Ptr = z1; |
388 | 158142c2 | bellard | *z0Ptr = z0; |
389 | 158142c2 | bellard | |
390 | 158142c2 | bellard | } |
391 | 158142c2 | bellard | |
392 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
393 | 158142c2 | bellard | | Subtracts the 128-bit value formed by concatenating `b0' and `b1' from the
|
394 | 158142c2 | bellard | | 128-bit value formed by concatenating `a0' and `a1'. Subtraction is modulo
|
395 | 158142c2 | bellard | | 2^128, so any borrow out (carry out) is lost. The result is broken into two
|
396 | 158142c2 | bellard | | 64-bit pieces which are stored at the locations pointed to by `z0Ptr' and
|
397 | 158142c2 | bellard | | `z1Ptr'.
|
398 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
399 | 158142c2 | bellard | |
400 | 158142c2 | bellard | INLINE void
|
401 | 158142c2 | bellard | sub128( |
402 | bb98fe42 | Andreas Färber | uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1, uint64_t *z0Ptr, uint64_t *z1Ptr ) |
403 | 158142c2 | bellard | { |
404 | 158142c2 | bellard | |
405 | 158142c2 | bellard | *z1Ptr = a1 - b1; |
406 | 158142c2 | bellard | *z0Ptr = a0 - b0 - ( a1 < b1 ); |
407 | 158142c2 | bellard | |
408 | 158142c2 | bellard | } |
409 | 158142c2 | bellard | |
410 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
411 | 158142c2 | bellard | | Subtracts the 192-bit value formed by concatenating `b0', `b1', and `b2'
|
412 | 158142c2 | bellard | | from the 192-bit value formed by concatenating `a0', `a1', and `a2'.
|
413 | 158142c2 | bellard | | Subtraction is modulo 2^192, so any borrow out (carry out) is lost. The
|
414 | 158142c2 | bellard | | result is broken into three 64-bit pieces which are stored at the locations
|
415 | 158142c2 | bellard | | pointed to by `z0Ptr', `z1Ptr', and `z2Ptr'.
|
416 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
417 | 158142c2 | bellard | |
418 | 158142c2 | bellard | INLINE void
|
419 | 158142c2 | bellard | sub192( |
420 | bb98fe42 | Andreas Färber | uint64_t a0, |
421 | bb98fe42 | Andreas Färber | uint64_t a1, |
422 | bb98fe42 | Andreas Färber | uint64_t a2, |
423 | bb98fe42 | Andreas Färber | uint64_t b0, |
424 | bb98fe42 | Andreas Färber | uint64_t b1, |
425 | bb98fe42 | Andreas Färber | uint64_t b2, |
426 | bb98fe42 | Andreas Färber | uint64_t *z0Ptr, |
427 | bb98fe42 | Andreas Färber | uint64_t *z1Ptr, |
428 | bb98fe42 | Andreas Färber | uint64_t *z2Ptr |
429 | 158142c2 | bellard | ) |
430 | 158142c2 | bellard | { |
431 | bb98fe42 | Andreas Färber | uint64_t z0, z1, z2; |
432 | 158142c2 | bellard | int8 borrow0, borrow1; |
433 | 158142c2 | bellard | |
434 | 158142c2 | bellard | z2 = a2 - b2; |
435 | 158142c2 | bellard | borrow1 = ( a2 < b2 ); |
436 | 158142c2 | bellard | z1 = a1 - b1; |
437 | 158142c2 | bellard | borrow0 = ( a1 < b1 ); |
438 | 158142c2 | bellard | z0 = a0 - b0; |
439 | 158142c2 | bellard | z0 -= ( z1 < borrow1 ); |
440 | 158142c2 | bellard | z1 -= borrow1; |
441 | 158142c2 | bellard | z0 -= borrow0; |
442 | 158142c2 | bellard | *z2Ptr = z2; |
443 | 158142c2 | bellard | *z1Ptr = z1; |
444 | 158142c2 | bellard | *z0Ptr = z0; |
445 | 158142c2 | bellard | |
446 | 158142c2 | bellard | } |
447 | 158142c2 | bellard | |
448 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
449 | 158142c2 | bellard | | Multiplies `a' by `b' to obtain a 128-bit product. The product is broken
|
450 | 158142c2 | bellard | | into two 64-bit pieces which are stored at the locations pointed to by
|
451 | 158142c2 | bellard | | `z0Ptr' and `z1Ptr'.
|
452 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
453 | 158142c2 | bellard | |
454 | bb98fe42 | Andreas Färber | INLINE void mul64To128( uint64_t a, uint64_t b, uint64_t *z0Ptr, uint64_t *z1Ptr )
|
455 | 158142c2 | bellard | { |
456 | bb98fe42 | Andreas Färber | uint32_t aHigh, aLow, bHigh, bLow; |
457 | bb98fe42 | Andreas Färber | uint64_t z0, zMiddleA, zMiddleB, z1; |
458 | 158142c2 | bellard | |
459 | 158142c2 | bellard | aLow = a; |
460 | 158142c2 | bellard | aHigh = a>>32;
|
461 | 158142c2 | bellard | bLow = b; |
462 | 158142c2 | bellard | bHigh = b>>32;
|
463 | bb98fe42 | Andreas Färber | z1 = ( (uint64_t) aLow ) * bLow; |
464 | bb98fe42 | Andreas Färber | zMiddleA = ( (uint64_t) aLow ) * bHigh; |
465 | bb98fe42 | Andreas Färber | zMiddleB = ( (uint64_t) aHigh ) * bLow; |
466 | bb98fe42 | Andreas Färber | z0 = ( (uint64_t) aHigh ) * bHigh; |
467 | 158142c2 | bellard | zMiddleA += zMiddleB; |
468 | bb98fe42 | Andreas Färber | z0 += ( ( (uint64_t) ( zMiddleA < zMiddleB ) )<<32 ) + ( zMiddleA>>32 ); |
469 | 158142c2 | bellard | zMiddleA <<= 32;
|
470 | 158142c2 | bellard | z1 += zMiddleA; |
471 | 158142c2 | bellard | z0 += ( z1 < zMiddleA ); |
472 | 158142c2 | bellard | *z1Ptr = z1; |
473 | 158142c2 | bellard | *z0Ptr = z0; |
474 | 158142c2 | bellard | |
475 | 158142c2 | bellard | } |
476 | 158142c2 | bellard | |
477 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
478 | 158142c2 | bellard | | Multiplies the 128-bit value formed by concatenating `a0' and `a1' by
|
479 | 158142c2 | bellard | | `b' to obtain a 192-bit product. The product is broken into three 64-bit
|
480 | 158142c2 | bellard | | pieces which are stored at the locations pointed to by `z0Ptr', `z1Ptr', and
|
481 | 158142c2 | bellard | | `z2Ptr'.
|
482 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
483 | 158142c2 | bellard | |
484 | 158142c2 | bellard | INLINE void
|
485 | 158142c2 | bellard | mul128By64To192( |
486 | bb98fe42 | Andreas Färber | uint64_t a0, |
487 | bb98fe42 | Andreas Färber | uint64_t a1, |
488 | bb98fe42 | Andreas Färber | uint64_t b, |
489 | bb98fe42 | Andreas Färber | uint64_t *z0Ptr, |
490 | bb98fe42 | Andreas Färber | uint64_t *z1Ptr, |
491 | bb98fe42 | Andreas Färber | uint64_t *z2Ptr |
492 | 158142c2 | bellard | ) |
493 | 158142c2 | bellard | { |
494 | bb98fe42 | Andreas Färber | uint64_t z0, z1, z2, more1; |
495 | 158142c2 | bellard | |
496 | 158142c2 | bellard | mul64To128( a1, b, &z1, &z2 ); |
497 | 158142c2 | bellard | mul64To128( a0, b, &z0, &more1 ); |
498 | 158142c2 | bellard | add128( z0, more1, 0, z1, &z0, &z1 );
|
499 | 158142c2 | bellard | *z2Ptr = z2; |
500 | 158142c2 | bellard | *z1Ptr = z1; |
501 | 158142c2 | bellard | *z0Ptr = z0; |
502 | 158142c2 | bellard | |
503 | 158142c2 | bellard | } |
504 | 158142c2 | bellard | |
505 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
506 | 158142c2 | bellard | | Multiplies the 128-bit value formed by concatenating `a0' and `a1' to the
|
507 | 158142c2 | bellard | | 128-bit value formed by concatenating `b0' and `b1' to obtain a 256-bit
|
508 | 158142c2 | bellard | | product. The product is broken into four 64-bit pieces which are stored at
|
509 | 158142c2 | bellard | | the locations pointed to by `z0Ptr', `z1Ptr', `z2Ptr', and `z3Ptr'.
|
510 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
511 | 158142c2 | bellard | |
512 | 158142c2 | bellard | INLINE void
|
513 | 158142c2 | bellard | mul128To256( |
514 | bb98fe42 | Andreas Färber | uint64_t a0, |
515 | bb98fe42 | Andreas Färber | uint64_t a1, |
516 | bb98fe42 | Andreas Färber | uint64_t b0, |
517 | bb98fe42 | Andreas Färber | uint64_t b1, |
518 | bb98fe42 | Andreas Färber | uint64_t *z0Ptr, |
519 | bb98fe42 | Andreas Färber | uint64_t *z1Ptr, |
520 | bb98fe42 | Andreas Färber | uint64_t *z2Ptr, |
521 | bb98fe42 | Andreas Färber | uint64_t *z3Ptr |
522 | 158142c2 | bellard | ) |
523 | 158142c2 | bellard | { |
524 | bb98fe42 | Andreas Färber | uint64_t z0, z1, z2, z3; |
525 | bb98fe42 | Andreas Färber | uint64_t more1, more2; |
526 | 158142c2 | bellard | |
527 | 158142c2 | bellard | mul64To128( a1, b1, &z2, &z3 ); |
528 | 158142c2 | bellard | mul64To128( a1, b0, &z1, &more2 ); |
529 | 158142c2 | bellard | add128( z1, more2, 0, z2, &z1, &z2 );
|
530 | 158142c2 | bellard | mul64To128( a0, b0, &z0, &more1 ); |
531 | 158142c2 | bellard | add128( z0, more1, 0, z1, &z0, &z1 );
|
532 | 158142c2 | bellard | mul64To128( a0, b1, &more1, &more2 ); |
533 | 158142c2 | bellard | add128( more1, more2, 0, z2, &more1, &z2 );
|
534 | 158142c2 | bellard | add128( z0, z1, 0, more1, &z0, &z1 );
|
535 | 158142c2 | bellard | *z3Ptr = z3; |
536 | 158142c2 | bellard | *z2Ptr = z2; |
537 | 158142c2 | bellard | *z1Ptr = z1; |
538 | 158142c2 | bellard | *z0Ptr = z0; |
539 | 158142c2 | bellard | |
540 | 158142c2 | bellard | } |
541 | 158142c2 | bellard | |
542 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
543 | 158142c2 | bellard | | Returns an approximation to the 64-bit integer quotient obtained by dividing
|
544 | 158142c2 | bellard | | `b' into the 128-bit value formed by concatenating `a0' and `a1'. The
|
545 | 158142c2 | bellard | | divisor `b' must be at least 2^63. If q is the exact quotient truncated
|
546 | 158142c2 | bellard | | toward zero, the approximation returned lies between q and q + 2 inclusive.
|
547 | 158142c2 | bellard | | If the exact quotient q is larger than 64 bits, the maximum positive 64-bit
|
548 | 158142c2 | bellard | | unsigned integer is returned.
|
549 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
550 | 158142c2 | bellard | |
551 | bb98fe42 | Andreas Färber | static uint64_t estimateDiv128To64( uint64_t a0, uint64_t a1, uint64_t b )
|
552 | 158142c2 | bellard | { |
553 | bb98fe42 | Andreas Färber | uint64_t b0, b1; |
554 | bb98fe42 | Andreas Färber | uint64_t rem0, rem1, term0, term1; |
555 | bb98fe42 | Andreas Färber | uint64_t z; |
556 | 158142c2 | bellard | |
557 | 158142c2 | bellard | if ( b <= a0 ) return LIT64( 0xFFFFFFFFFFFFFFFF ); |
558 | 158142c2 | bellard | b0 = b>>32;
|
559 | 158142c2 | bellard | z = ( b0<<32 <= a0 ) ? LIT64( 0xFFFFFFFF00000000 ) : ( a0 / b0 )<<32; |
560 | 158142c2 | bellard | mul64To128( b, z, &term0, &term1 ); |
561 | 158142c2 | bellard | sub128( a0, a1, term0, term1, &rem0, &rem1 ); |
562 | bb98fe42 | Andreas Färber | while ( ( (int64_t) rem0 ) < 0 ) { |
563 | 158142c2 | bellard | z -= LIT64( 0x100000000 );
|
564 | 158142c2 | bellard | b1 = b<<32;
|
565 | 158142c2 | bellard | add128( rem0, rem1, b0, b1, &rem0, &rem1 ); |
566 | 158142c2 | bellard | } |
567 | 158142c2 | bellard | rem0 = ( rem0<<32 ) | ( rem1>>32 ); |
568 | 158142c2 | bellard | z |= ( b0<<32 <= rem0 ) ? 0xFFFFFFFF : rem0 / b0; |
569 | 158142c2 | bellard | return z;
|
570 | 158142c2 | bellard | |
571 | 158142c2 | bellard | } |
572 | 158142c2 | bellard | |
573 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
574 | 158142c2 | bellard | | Returns an approximation to the square root of the 32-bit significand given
|
575 | 158142c2 | bellard | | by `a'. Considered as an integer, `a' must be at least 2^31. If bit 0 of
|
576 | 158142c2 | bellard | | `aExp' (the least significant bit) is 1, the integer returned approximates
|
577 | 158142c2 | bellard | | 2^31*sqrt(`a'/2^31), where `a' is considered an integer. If bit 0 of `aExp'
|
578 | 158142c2 | bellard | | is 0, the integer returned approximates 2^31*sqrt(`a'/2^30). In either
|
579 | 158142c2 | bellard | | case, the approximation returned lies strictly within +/-2 of the exact
|
580 | 158142c2 | bellard | | value.
|
581 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
582 | 158142c2 | bellard | |
583 | bb98fe42 | Andreas Färber | static uint32_t estimateSqrt32( int16 aExp, uint32_t a )
|
584 | 158142c2 | bellard | { |
585 | bb98fe42 | Andreas Färber | static const uint16_t sqrtOddAdjustments[] = { |
586 | 158142c2 | bellard | 0x0004, 0x0022, 0x005D, 0x00B1, 0x011D, 0x019F, 0x0236, 0x02E0, |
587 | 158142c2 | bellard | 0x039C, 0x0468, 0x0545, 0x0631, 0x072B, 0x0832, 0x0946, 0x0A67 |
588 | 158142c2 | bellard | }; |
589 | bb98fe42 | Andreas Färber | static const uint16_t sqrtEvenAdjustments[] = { |
590 | 158142c2 | bellard | 0x0A2D, 0x08AF, 0x075A, 0x0629, 0x051A, 0x0429, 0x0356, 0x029E, |
591 | 158142c2 | bellard | 0x0200, 0x0179, 0x0109, 0x00AF, 0x0068, 0x0034, 0x0012, 0x0002 |
592 | 158142c2 | bellard | }; |
593 | 158142c2 | bellard | int8 index; |
594 | bb98fe42 | Andreas Färber | uint32_t z; |
595 | 158142c2 | bellard | |
596 | 158142c2 | bellard | index = ( a>>27 ) & 15; |
597 | 158142c2 | bellard | if ( aExp & 1 ) { |
598 | 3f4cb3d3 | blueswir1 | z = 0x4000 + ( a>>17 ) - sqrtOddAdjustments[ (int)index ]; |
599 | 158142c2 | bellard | z = ( ( a / z )<<14 ) + ( z<<15 ); |
600 | 158142c2 | bellard | a >>= 1;
|
601 | 158142c2 | bellard | } |
602 | 158142c2 | bellard | else {
|
603 | 3f4cb3d3 | blueswir1 | z = 0x8000 + ( a>>17 ) - sqrtEvenAdjustments[ (int)index ]; |
604 | 158142c2 | bellard | z = a / z + z; |
605 | 158142c2 | bellard | z = ( 0x20000 <= z ) ? 0xFFFF8000 : ( z<<15 ); |
606 | bb98fe42 | Andreas Färber | if ( z <= a ) return (uint32_t) ( ( (int32_t) a )>>1 ); |
607 | 158142c2 | bellard | } |
608 | bb98fe42 | Andreas Färber | return ( (uint32_t) ( ( ( (uint64_t) a )<<31 ) / z ) ) + ( z>>1 ); |
609 | 158142c2 | bellard | |
610 | 158142c2 | bellard | } |
611 | 158142c2 | bellard | |
612 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
613 | 158142c2 | bellard | | Returns the number of leading 0 bits before the most-significant 1 bit of
|
614 | 158142c2 | bellard | | `a'. If `a' is zero, 32 is returned.
|
615 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
616 | 158142c2 | bellard | |
617 | bb98fe42 | Andreas Färber | static int8 countLeadingZeros32( uint32_t a )
|
618 | 158142c2 | bellard | { |
619 | 158142c2 | bellard | static const int8 countLeadingZerosHigh[] = { |
620 | 158142c2 | bellard | 8, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, |
621 | 158142c2 | bellard | 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, |
622 | 158142c2 | bellard | 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, |
623 | 158142c2 | bellard | 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, |
624 | 158142c2 | bellard | 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, |
625 | 158142c2 | bellard | 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, |
626 | 158142c2 | bellard | 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, |
627 | 158142c2 | bellard | 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, |
628 | 158142c2 | bellard | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
629 | 158142c2 | bellard | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
630 | 158142c2 | bellard | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
631 | 158142c2 | bellard | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
632 | 158142c2 | bellard | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
633 | 158142c2 | bellard | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
634 | 158142c2 | bellard | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
635 | 158142c2 | bellard | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 |
636 | 158142c2 | bellard | }; |
637 | 158142c2 | bellard | int8 shiftCount; |
638 | 158142c2 | bellard | |
639 | 158142c2 | bellard | shiftCount = 0;
|
640 | 158142c2 | bellard | if ( a < 0x10000 ) { |
641 | 158142c2 | bellard | shiftCount += 16;
|
642 | 158142c2 | bellard | a <<= 16;
|
643 | 158142c2 | bellard | } |
644 | 158142c2 | bellard | if ( a < 0x1000000 ) { |
645 | 158142c2 | bellard | shiftCount += 8;
|
646 | 158142c2 | bellard | a <<= 8;
|
647 | 158142c2 | bellard | } |
648 | 158142c2 | bellard | shiftCount += countLeadingZerosHigh[ a>>24 ];
|
649 | 158142c2 | bellard | return shiftCount;
|
650 | 158142c2 | bellard | |
651 | 158142c2 | bellard | } |
652 | 158142c2 | bellard | |
653 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
654 | 158142c2 | bellard | | Returns the number of leading 0 bits before the most-significant 1 bit of
|
655 | 158142c2 | bellard | | `a'. If `a' is zero, 64 is returned.
|
656 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
657 | 158142c2 | bellard | |
658 | bb98fe42 | Andreas Färber | static int8 countLeadingZeros64( uint64_t a )
|
659 | 158142c2 | bellard | { |
660 | 158142c2 | bellard | int8 shiftCount; |
661 | 158142c2 | bellard | |
662 | 158142c2 | bellard | shiftCount = 0;
|
663 | bb98fe42 | Andreas Färber | if ( a < ( (uint64_t) 1 )<<32 ) { |
664 | 158142c2 | bellard | shiftCount += 32;
|
665 | 158142c2 | bellard | } |
666 | 158142c2 | bellard | else {
|
667 | 158142c2 | bellard | a >>= 32;
|
668 | 158142c2 | bellard | } |
669 | 158142c2 | bellard | shiftCount += countLeadingZeros32( a ); |
670 | 158142c2 | bellard | return shiftCount;
|
671 | 158142c2 | bellard | |
672 | 158142c2 | bellard | } |
673 | 158142c2 | bellard | |
674 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
675 | 158142c2 | bellard | | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1'
|
676 | 158142c2 | bellard | | is equal to the 128-bit value formed by concatenating `b0' and `b1'.
|
677 | 158142c2 | bellard | | Otherwise, returns 0.
|
678 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
679 | 158142c2 | bellard | |
680 | bb98fe42 | Andreas Färber | INLINE flag eq128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 ) |
681 | 158142c2 | bellard | { |
682 | 158142c2 | bellard | |
683 | 158142c2 | bellard | return ( a0 == b0 ) && ( a1 == b1 );
|
684 | 158142c2 | bellard | |
685 | 158142c2 | bellard | } |
686 | 158142c2 | bellard | |
687 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
688 | 158142c2 | bellard | | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less
|
689 | 158142c2 | bellard | | than or equal to the 128-bit value formed by concatenating `b0' and `b1'.
|
690 | 158142c2 | bellard | | Otherwise, returns 0.
|
691 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
692 | 158142c2 | bellard | |
693 | bb98fe42 | Andreas Färber | INLINE flag le128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 ) |
694 | 158142c2 | bellard | { |
695 | 158142c2 | bellard | |
696 | 158142c2 | bellard | return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 <= b1 ) );
|
697 | 158142c2 | bellard | |
698 | 158142c2 | bellard | } |
699 | 158142c2 | bellard | |
700 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
701 | 158142c2 | bellard | | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less
|
702 | 158142c2 | bellard | | than the 128-bit value formed by concatenating `b0' and `b1'. Otherwise,
|
703 | 158142c2 | bellard | | returns 0.
|
704 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
705 | 158142c2 | bellard | |
706 | bb98fe42 | Andreas Färber | INLINE flag lt128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 ) |
707 | 158142c2 | bellard | { |
708 | 158142c2 | bellard | |
709 | 158142c2 | bellard | return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 < b1 ) );
|
710 | 158142c2 | bellard | |
711 | 158142c2 | bellard | } |
712 | 158142c2 | bellard | |
713 | 158142c2 | bellard | /*----------------------------------------------------------------------------
|
714 | 158142c2 | bellard | | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is
|
715 | 158142c2 | bellard | | not equal to the 128-bit value formed by concatenating `b0' and `b1'.
|
716 | 158142c2 | bellard | | Otherwise, returns 0.
|
717 | 158142c2 | bellard | *----------------------------------------------------------------------------*/
|
718 | 158142c2 | bellard | |
719 | bb98fe42 | Andreas Färber | INLINE flag ne128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 ) |
720 | 158142c2 | bellard | { |
721 | 158142c2 | bellard | |
722 | 158142c2 | bellard | return ( a0 != b0 ) || ( a1 != b1 );
|
723 | 158142c2 | bellard | |
724 | 158142c2 | bellard | } |