2 * Utility compute operations used by translated code.
4 * Copyright (c) 2007 Thiemo Seufer
5 * Copyright (c) 2007 Jocelyn Mayer
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
29 #include "qemu/bswap.h"
32 static inline void mulu64(uint64_t *plow, uint64_t *phigh,
33 uint64_t a, uint64_t b)
35 __uint128_t r = (__uint128_t)a * b;
40 static inline void muls64(uint64_t *plow, uint64_t *phigh,
43 __int128_t r = (__int128_t)a * b;
48 /* compute with 96 bit intermediate result: (a*b)/c */
49 static inline uint64_t muldiv64(uint64_t a, uint32_t b, uint32_t c)
51 return (__int128_t)a * b / c;
54 static inline int divu128(uint64_t *plow, uint64_t *phigh, uint64_t divisor)
59 __uint128_t dividend = ((__uint128_t)*phigh << 64) | *plow;
60 __uint128_t result = dividend / divisor;
62 *phigh = dividend % divisor;
63 return result > UINT64_MAX;
67 static inline int divs128(int64_t *plow, int64_t *phigh, int64_t divisor)
72 __int128_t dividend = ((__int128_t)*phigh << 64) | *plow;
73 __int128_t result = dividend / divisor;
75 *phigh = dividend % divisor;
76 return result != *plow;
80 void muls64(uint64_t *phigh, uint64_t *plow, int64_t a, int64_t b);
81 void mulu64(uint64_t *phigh, uint64_t *plow, uint64_t a, uint64_t b);
82 int divu128(uint64_t *plow, uint64_t *phigh, uint64_t divisor);
83 int divs128(int64_t *plow, int64_t *phigh, int64_t divisor);
85 static inline uint64_t muldiv64(uint64_t a, uint32_t b, uint32_t c)
90 #ifdef HOST_WORDS_BIGENDIAN
100 rl = (uint64_t)u.l.low * (uint64_t)b;
101 rh = (uint64_t)u.l.high * (uint64_t)b;
104 res.l.low = (((rh % c) << 32) + (rl & 0xffffffff)) / c;
110 * clz32 - count leading zeros in a 32-bit value.
111 * @val: The value to search
113 * Returns 32 if the value is zero. Note that the GCC builtin is
114 * undefined if the value is zero.
116 static inline int clz32(uint32_t val)
118 #if QEMU_GNUC_PREREQ(3, 4)
119 return val ? __builtin_clz(val) : 32;
121 /* Binary search for the leading one bit. */
124 if (!(val & 0xFFFF0000U)) {
128 if (!(val & 0xFF000000U)) {
132 if (!(val & 0xF0000000U)) {
136 if (!(val & 0xC0000000U)) {
140 if (!(val & 0x80000000U)) {
144 if (!(val & 0x80000000U)) {
152 * clo32 - count leading ones in a 32-bit value.
153 * @val: The value to search
155 * Returns 32 if the value is -1.
157 static inline int clo32(uint32_t val)
163 * clz64 - count leading zeros in a 64-bit value.
164 * @val: The value to search
166 * Returns 64 if the value is zero. Note that the GCC builtin is
167 * undefined if the value is zero.
169 static inline int clz64(uint64_t val)
171 #if QEMU_GNUC_PREREQ(3, 4)
172 return val ? __builtin_clzll(val) : 64;
182 return cnt + clz32(val);
187 * clo64 - count leading ones in a 64-bit value.
188 * @val: The value to search
190 * Returns 64 if the value is -1.
192 static inline int clo64(uint64_t val)
198 * ctz32 - count trailing zeros in a 32-bit value.
199 * @val: The value to search
201 * Returns 32 if the value is zero. Note that the GCC builtin is
202 * undefined if the value is zero.
204 static inline int ctz32(uint32_t val)
206 #if QEMU_GNUC_PREREQ(3, 4)
207 return val ? __builtin_ctz(val) : 32;
209 /* Binary search for the trailing one bit. */
213 if (!(val & 0x0000FFFFUL)) {
217 if (!(val & 0x000000FFUL)) {
221 if (!(val & 0x0000000FUL)) {
225 if (!(val & 0x00000003UL)) {
229 if (!(val & 0x00000001UL)) {
233 if (!(val & 0x00000001UL)) {
242 * cto32 - count trailing ones in a 32-bit value.
243 * @val: The value to search
245 * Returns 32 if the value is -1.
247 static inline int cto32(uint32_t val)
253 * ctz64 - count trailing zeros in a 64-bit value.
254 * @val: The value to search
256 * Returns 64 if the value is zero. Note that the GCC builtin is
257 * undefined if the value is zero.
259 static inline int ctz64(uint64_t val)
261 #if QEMU_GNUC_PREREQ(3, 4)
262 return val ? __builtin_ctzll(val) : 64;
267 if (!((uint32_t)val)) {
272 return cnt + ctz32(val);
277 * cto64 - count trailing ones in a 64-bit value.
278 * @val: The value to search
280 * Returns 64 if the value is -1.
282 static inline int cto64(uint64_t val)
288 * clrsb32 - count leading redundant sign bits in a 32-bit value.
289 * @val: The value to search
291 * Returns the number of bits following the sign bit that are equal to it.
292 * No special cases; output range is [0-31].
294 static inline int clrsb32(uint32_t val)
296 #if QEMU_GNUC_PREREQ(4, 7)
297 return __builtin_clrsb(val);
299 return clz32(val ^ ((int32_t)val >> 1)) - 1;
304 * clrsb64 - count leading redundant sign bits in a 64-bit value.
305 * @val: The value to search
307 * Returns the number of bits following the sign bit that are equal to it.
308 * No special cases; output range is [0-63].
310 static inline int clrsb64(uint64_t val)
312 #if QEMU_GNUC_PREREQ(4, 7)
313 return __builtin_clrsbll(val);
315 return clz64(val ^ ((int64_t)val >> 1)) - 1;
320 * ctpop8 - count the population of one bits in an 8-bit value.
321 * @val: The value to search
323 static inline int ctpop8(uint8_t val)
325 #if QEMU_GNUC_PREREQ(3, 4)
326 return __builtin_popcount(val);
328 val = (val & 0x55) + ((val >> 1) & 0x55);
329 val = (val & 0x33) + ((val >> 2) & 0x33);
330 val = (val & 0x0f) + ((val >> 4) & 0x0f);
337 * ctpop16 - count the population of one bits in a 16-bit value.
338 * @val: The value to search
340 static inline int ctpop16(uint16_t val)
342 #if QEMU_GNUC_PREREQ(3, 4)
343 return __builtin_popcount(val);
345 val = (val & 0x5555) + ((val >> 1) & 0x5555);
346 val = (val & 0x3333) + ((val >> 2) & 0x3333);
347 val = (val & 0x0f0f) + ((val >> 4) & 0x0f0f);
348 val = (val & 0x00ff) + ((val >> 8) & 0x00ff);
355 * ctpop32 - count the population of one bits in a 32-bit value.
356 * @val: The value to search
358 static inline int ctpop32(uint32_t val)
360 #if QEMU_GNUC_PREREQ(3, 4)
361 return __builtin_popcount(val);
363 val = (val & 0x55555555) + ((val >> 1) & 0x55555555);
364 val = (val & 0x33333333) + ((val >> 2) & 0x33333333);
365 val = (val & 0x0f0f0f0f) + ((val >> 4) & 0x0f0f0f0f);
366 val = (val & 0x00ff00ff) + ((val >> 8) & 0x00ff00ff);
367 val = (val & 0x0000ffff) + ((val >> 16) & 0x0000ffff);
374 * ctpop64 - count the population of one bits in a 64-bit value.
375 * @val: The value to search
377 static inline int ctpop64(uint64_t val)
379 #if QEMU_GNUC_PREREQ(3, 4)
380 return __builtin_popcountll(val);
382 val = (val & 0x5555555555555555ULL) + ((val >> 1) & 0x5555555555555555ULL);
383 val = (val & 0x3333333333333333ULL) + ((val >> 2) & 0x3333333333333333ULL);
384 val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) & 0x0f0f0f0f0f0f0f0fULL);
385 val = (val & 0x00ff00ff00ff00ffULL) + ((val >> 8) & 0x00ff00ff00ff00ffULL);
386 val = (val & 0x0000ffff0000ffffULL) + ((val >> 16) & 0x0000ffff0000ffffULL);
387 val = (val & 0x00000000ffffffffULL) + ((val >> 32) & 0x00000000ffffffffULL);
394 * revbit8 - reverse the bits in an 8-bit value.
395 * @x: The value to modify.
397 static inline uint8_t revbit8(uint8_t x)
399 /* Assign the correct nibble position. */
400 x = ((x & 0xf0) >> 4)
402 /* Assign the correct bit position. */
403 x = ((x & 0x88) >> 3)
411 * revbit16 - reverse the bits in a 16-bit value.
412 * @x: The value to modify.
414 static inline uint16_t revbit16(uint16_t x)
416 /* Assign the correct byte position. */
418 /* Assign the correct nibble position. */
419 x = ((x & 0xf0f0) >> 4)
420 | ((x & 0x0f0f) << 4);
421 /* Assign the correct bit position. */
422 x = ((x & 0x8888) >> 3)
423 | ((x & 0x4444) >> 1)
424 | ((x & 0x2222) << 1)
425 | ((x & 0x1111) << 3);
430 * revbit32 - reverse the bits in a 32-bit value.
431 * @x: The value to modify.
433 static inline uint32_t revbit32(uint32_t x)
435 /* Assign the correct byte position. */
437 /* Assign the correct nibble position. */
438 x = ((x & 0xf0f0f0f0u) >> 4)
439 | ((x & 0x0f0f0f0fu) << 4);
440 /* Assign the correct bit position. */
441 x = ((x & 0x88888888u) >> 3)
442 | ((x & 0x44444444u) >> 1)
443 | ((x & 0x22222222u) << 1)
444 | ((x & 0x11111111u) << 3);
449 * revbit64 - reverse the bits in a 64-bit value.
450 * @x: The value to modify.
452 static inline uint64_t revbit64(uint64_t x)
454 /* Assign the correct byte position. */
456 /* Assign the correct nibble position. */
457 x = ((x & 0xf0f0f0f0f0f0f0f0ull) >> 4)
458 | ((x & 0x0f0f0f0f0f0f0f0full) << 4);
459 /* Assign the correct bit position. */
460 x = ((x & 0x8888888888888888ull) >> 3)
461 | ((x & 0x4444444444444444ull) >> 1)
462 | ((x & 0x2222222222222222ull) << 1)
463 | ((x & 0x1111111111111111ull) << 3);
467 /* Host type specific sizes of these routines. */
469 #if ULONG_MAX == UINT32_MAX
474 # define ctpopl ctpop32
475 # define revbitl revbit32
476 #elif ULONG_MAX == UINT64_MAX
481 # define ctpopl ctpop64
482 # define revbitl revbit64
484 # error Unknown sizeof long
487 static inline bool is_power_of_2(uint64_t value)
493 return !(value & (value - 1));
496 /* round down to the nearest power of 2*/
497 static inline int64_t pow2floor(int64_t value)
499 if (!is_power_of_2(value)) {
500 value = 0x8000000000000000ULL >> clz64(value);
505 /* round up to the nearest power of 2 (0 if overflow) */
506 static inline uint64_t pow2ceil(uint64_t value)
508 uint8_t nlz = clz64(value);
510 if (is_power_of_2(value)) {
516 return 1ULL << (64 - nlz);