]>
git.proxmox.com Git - mirror_qemu.git/blob - include/qemu/host-utils.h
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
26 /* Portions of this work are licensed under the terms of the GNU GPL,
27 * version 2 or later. See the COPYING file in the top-level directory.
33 #include "qemu/bswap.h"
34 #include "qemu/int128.h"
37 static inline void mulu64(uint64_t *plow
, uint64_t *phigh
,
38 uint64_t a
, uint64_t b
)
40 __uint128_t r
= (__uint128_t
)a
* b
;
45 static inline void muls64(uint64_t *plow
, uint64_t *phigh
,
48 __int128_t r
= (__int128_t
)a
* b
;
53 /* compute with 96 bit intermediate result: (a*b)/c */
54 static inline uint64_t muldiv64(uint64_t a
, uint32_t b
, uint32_t c
)
56 return (__int128_t
)a
* b
/ c
;
59 static inline uint64_t muldiv64_round_up(uint64_t a
, uint32_t b
, uint32_t c
)
61 return ((__int128_t
)a
* b
+ c
- 1) / c
;
64 static inline uint64_t divu128(uint64_t *plow
, uint64_t *phigh
,
67 __uint128_t dividend
= ((__uint128_t
)*phigh
<< 64) | *plow
;
68 __uint128_t result
= dividend
/ divisor
;
71 *phigh
= result
>> 64;
72 return dividend
% divisor
;
75 static inline int64_t divs128(uint64_t *plow
, int64_t *phigh
,
78 __int128_t dividend
= ((__int128_t
)*phigh
<< 64) | *plow
;
79 __int128_t result
= dividend
/ divisor
;
82 *phigh
= result
>> 64;
83 return dividend
% divisor
;
86 void muls64(uint64_t *plow
, uint64_t *phigh
, int64_t a
, int64_t b
);
87 void mulu64(uint64_t *plow
, uint64_t *phigh
, uint64_t a
, uint64_t b
);
88 uint64_t divu128(uint64_t *plow
, uint64_t *phigh
, uint64_t divisor
);
89 int64_t divs128(uint64_t *plow
, int64_t *phigh
, int64_t divisor
);
91 static inline uint64_t muldiv64_rounding(uint64_t a
, uint32_t b
, uint32_t c
,
107 rl
= (uint64_t)u
.l
.low
* (uint64_t)b
;
111 rh
= (uint64_t)u
.l
.high
* (uint64_t)b
;
114 res
.l
.low
= (((rh
% c
) << 32) + (rl
& 0xffffffff)) / c
;
118 static inline uint64_t muldiv64(uint64_t a
, uint32_t b
, uint32_t c
)
120 return muldiv64_rounding(a
, b
, c
, false);
123 static inline uint64_t muldiv64_round_up(uint64_t a
, uint32_t b
, uint32_t c
)
125 return muldiv64_rounding(a
, b
, c
, true);
130 * clz8 - count leading zeros in a 8-bit value.
131 * @val: The value to search
133 * Returns 8 if the value is zero. Note that the GCC builtin is
134 * undefined if the value is zero.
136 * Note that the GCC builtin will upcast its argument to an `unsigned int`
137 * so this function subtracts off the number of prepended zeroes.
139 static inline int clz8(uint8_t val
)
141 return val
? __builtin_clz(val
) - 24 : 8;
145 * clz16 - count leading zeros in a 16-bit value.
146 * @val: The value to search
148 * Returns 16 if the value is zero. Note that the GCC builtin is
149 * undefined if the value is zero.
151 * Note that the GCC builtin will upcast its argument to an `unsigned int`
152 * so this function subtracts off the number of prepended zeroes.
154 static inline int clz16(uint16_t val
)
156 return val
? __builtin_clz(val
) - 16 : 16;
160 * clz32 - count leading zeros in a 32-bit value.
161 * @val: The value to search
163 * Returns 32 if the value is zero. Note that the GCC builtin is
164 * undefined if the value is zero.
166 static inline int clz32(uint32_t val
)
168 return val
? __builtin_clz(val
) : 32;
172 * clo32 - count leading ones in a 32-bit value.
173 * @val: The value to search
175 * Returns 32 if the value is -1.
177 static inline int clo32(uint32_t val
)
183 * clz64 - count leading zeros in a 64-bit value.
184 * @val: The value to search
186 * Returns 64 if the value is zero. Note that the GCC builtin is
187 * undefined if the value is zero.
189 static inline int clz64(uint64_t val
)
191 return val
? __builtin_clzll(val
) : 64;
195 * clo64 - count leading ones in a 64-bit value.
196 * @val: The value to search
198 * Returns 64 if the value is -1.
200 static inline int clo64(uint64_t val
)
206 * ctz8 - count trailing zeros in a 8-bit value.
207 * @val: The value to search
209 * Returns 8 if the value is zero. Note that the GCC builtin is
210 * undefined if the value is zero.
212 static inline int ctz8(uint8_t val
)
214 return val
? __builtin_ctz(val
) : 8;
218 * ctz16 - count trailing zeros in a 16-bit value.
219 * @val: The value to search
221 * Returns 16 if the value is zero. Note that the GCC builtin is
222 * undefined if the value is zero.
224 static inline int ctz16(uint16_t val
)
226 return val
? __builtin_ctz(val
) : 16;
230 * ctz32 - count trailing zeros in a 32-bit value.
231 * @val: The value to search
233 * Returns 32 if the value is zero. Note that the GCC builtin is
234 * undefined if the value is zero.
236 static inline int ctz32(uint32_t val
)
238 return val
? __builtin_ctz(val
) : 32;
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 return val
? __builtin_ctzll(val
) : 64;
265 * cto64 - count trailing ones in a 64-bit value.
266 * @val: The value to search
268 * Returns 64 if the value is -1.
270 static inline int cto64(uint64_t val
)
276 * clrsb32 - count leading redundant sign bits in a 32-bit value.
277 * @val: The value to search
279 * Returns the number of bits following the sign bit that are equal to it.
280 * No special cases; output range is [0-31].
282 static inline int clrsb32(uint32_t val
)
284 #if __has_builtin(__builtin_clrsb) || !defined(__clang__)
285 return __builtin_clrsb(val
);
287 return clz32(val
^ ((int32_t)val
>> 1)) - 1;
292 * clrsb64 - count leading redundant sign bits in a 64-bit value.
293 * @val: The value to search
295 * Returns the number of bits following the sign bit that are equal to it.
296 * No special cases; output range is [0-63].
298 static inline int clrsb64(uint64_t val
)
300 #if __has_builtin(__builtin_clrsbll) || !defined(__clang__)
301 return __builtin_clrsbll(val
);
303 return clz64(val
^ ((int64_t)val
>> 1)) - 1;
308 * ctpop8 - count the population of one bits in an 8-bit value.
309 * @val: The value to search
311 static inline int ctpop8(uint8_t val
)
313 return __builtin_popcount(val
);
317 * ctpop16 - count the population of one bits in a 16-bit value.
318 * @val: The value to search
320 static inline int ctpop16(uint16_t val
)
322 return __builtin_popcount(val
);
326 * ctpop32 - count the population of one bits in a 32-bit value.
327 * @val: The value to search
329 static inline int ctpop32(uint32_t val
)
331 return __builtin_popcount(val
);
335 * ctpop64 - count the population of one bits in a 64-bit value.
336 * @val: The value to search
338 static inline int ctpop64(uint64_t val
)
340 return __builtin_popcountll(val
);
344 * revbit8 - reverse the bits in an 8-bit value.
345 * @x: The value to modify.
347 static inline uint8_t revbit8(uint8_t x
)
349 #if __has_builtin(__builtin_bitreverse8)
350 return __builtin_bitreverse8(x
);
352 /* Assign the correct nibble position. */
353 x
= ((x
& 0xf0) >> 4)
355 /* Assign the correct bit position. */
356 x
= ((x
& 0x88) >> 3)
365 * revbit16 - reverse the bits in a 16-bit value.
366 * @x: The value to modify.
368 static inline uint16_t revbit16(uint16_t x
)
370 #if __has_builtin(__builtin_bitreverse16)
371 return __builtin_bitreverse16(x
);
373 /* Assign the correct byte position. */
375 /* Assign the correct nibble position. */
376 x
= ((x
& 0xf0f0) >> 4)
377 | ((x
& 0x0f0f) << 4);
378 /* Assign the correct bit position. */
379 x
= ((x
& 0x8888) >> 3)
380 | ((x
& 0x4444) >> 1)
381 | ((x
& 0x2222) << 1)
382 | ((x
& 0x1111) << 3);
388 * revbit32 - reverse the bits in a 32-bit value.
389 * @x: The value to modify.
391 static inline uint32_t revbit32(uint32_t x
)
393 #if __has_builtin(__builtin_bitreverse32)
394 return __builtin_bitreverse32(x
);
396 /* Assign the correct byte position. */
398 /* Assign the correct nibble position. */
399 x
= ((x
& 0xf0f0f0f0u
) >> 4)
400 | ((x
& 0x0f0f0f0fu
) << 4);
401 /* Assign the correct bit position. */
402 x
= ((x
& 0x88888888u
) >> 3)
403 | ((x
& 0x44444444u
) >> 1)
404 | ((x
& 0x22222222u
) << 1)
405 | ((x
& 0x11111111u
) << 3);
411 * revbit64 - reverse the bits in a 64-bit value.
412 * @x: The value to modify.
414 static inline uint64_t revbit64(uint64_t x
)
416 #if __has_builtin(__builtin_bitreverse64)
417 return __builtin_bitreverse64(x
);
419 /* Assign the correct byte position. */
421 /* Assign the correct nibble position. */
422 x
= ((x
& 0xf0f0f0f0f0f0f0f0ull
) >> 4)
423 | ((x
& 0x0f0f0f0f0f0f0f0full
) << 4);
424 /* Assign the correct bit position. */
425 x
= ((x
& 0x8888888888888888ull
) >> 3)
426 | ((x
& 0x4444444444444444ull
) >> 1)
427 | ((x
& 0x2222222222222222ull
) << 1)
428 | ((x
& 0x1111111111111111ull
) << 3);
434 * Return the absolute value of a 64-bit integer as an unsigned 64-bit value
436 static inline uint64_t uabs64(int64_t v
)
438 return v
< 0 ? -v
: v
;
442 * sadd32_overflow - addition with overflow indication
444 * @ret: Output for sum
446 * Computes *@ret = @x + @y, and returns true if and only if that
447 * value has been truncated.
449 static inline bool sadd32_overflow(int32_t x
, int32_t y
, int32_t *ret
)
451 return __builtin_add_overflow(x
, y
, ret
);
455 * sadd64_overflow - addition with overflow indication
457 * @ret: Output for sum
459 * Computes *@ret = @x + @y, and returns true if and only if that
460 * value has been truncated.
462 static inline bool sadd64_overflow(int64_t x
, int64_t y
, int64_t *ret
)
464 return __builtin_add_overflow(x
, y
, ret
);
468 * uadd32_overflow - addition with overflow indication
470 * @ret: Output for sum
472 * Computes *@ret = @x + @y, and returns true if and only if that
473 * value has been truncated.
475 static inline bool uadd32_overflow(uint32_t x
, uint32_t y
, uint32_t *ret
)
477 return __builtin_add_overflow(x
, y
, ret
);
481 * uadd64_overflow - addition with overflow indication
483 * @ret: Output for sum
485 * Computes *@ret = @x + @y, and returns true if and only if that
486 * value has been truncated.
488 static inline bool uadd64_overflow(uint64_t x
, uint64_t y
, uint64_t *ret
)
490 return __builtin_add_overflow(x
, y
, ret
);
494 * ssub32_overflow - subtraction with overflow indication
497 * @ret: Output for difference
499 * Computes *@ret = @x - @y, and returns true if and only if that
500 * value has been truncated.
502 static inline bool ssub32_overflow(int32_t x
, int32_t y
, int32_t *ret
)
504 return __builtin_sub_overflow(x
, y
, ret
);
508 * ssub64_overflow - subtraction with overflow indication
511 * @ret: Output for sum
513 * Computes *@ret = @x - @y, and returns true if and only if that
514 * value has been truncated.
516 static inline bool ssub64_overflow(int64_t x
, int64_t y
, int64_t *ret
)
518 return __builtin_sub_overflow(x
, y
, ret
);
522 * usub32_overflow - subtraction with overflow indication
525 * @ret: Output for sum
527 * Computes *@ret = @x - @y, and returns true if and only if that
528 * value has been truncated.
530 static inline bool usub32_overflow(uint32_t x
, uint32_t y
, uint32_t *ret
)
532 return __builtin_sub_overflow(x
, y
, ret
);
536 * usub64_overflow - subtraction with overflow indication
539 * @ret: Output for sum
541 * Computes *@ret = @x - @y, and returns true if and only if that
542 * value has been truncated.
544 static inline bool usub64_overflow(uint64_t x
, uint64_t y
, uint64_t *ret
)
546 return __builtin_sub_overflow(x
, y
, ret
);
550 * smul32_overflow - multiplication with overflow indication
551 * @x, @y: Input multipliers
552 * @ret: Output for product
554 * Computes *@ret = @x * @y, and returns true if and only if that
555 * value has been truncated.
557 static inline bool smul32_overflow(int32_t x
, int32_t y
, int32_t *ret
)
559 return __builtin_mul_overflow(x
, y
, ret
);
563 * smul64_overflow - multiplication with overflow indication
564 * @x, @y: Input multipliers
565 * @ret: Output for product
567 * Computes *@ret = @x * @y, and returns true if and only if that
568 * value has been truncated.
570 static inline bool smul64_overflow(int64_t x
, int64_t y
, int64_t *ret
)
572 return __builtin_mul_overflow(x
, y
, ret
);
576 * umul32_overflow - multiplication with overflow indication
577 * @x, @y: Input multipliers
578 * @ret: Output for product
580 * Computes *@ret = @x * @y, and returns true if and only if that
581 * value has been truncated.
583 static inline bool umul32_overflow(uint32_t x
, uint32_t y
, uint32_t *ret
)
585 return __builtin_mul_overflow(x
, y
, ret
);
589 * umul64_overflow - multiplication with overflow indication
590 * @x, @y: Input multipliers
591 * @ret: Output for product
593 * Computes *@ret = @x * @y, and returns true if and only if that
594 * value has been truncated.
596 static inline bool umul64_overflow(uint64_t x
, uint64_t y
, uint64_t *ret
)
598 return __builtin_mul_overflow(x
, y
, ret
);
602 * Unsigned 128x64 multiplication.
603 * Returns true if the result got truncated to 128 bits.
604 * Otherwise, returns false and the multiplication result via plow and phigh.
606 static inline bool mulu128(uint64_t *plow
, uint64_t *phigh
, uint64_t factor
)
608 #if defined(CONFIG_INT128)
611 __uint128_t f
= ((__uint128_t
)*phigh
<< 64) | *plow
;
612 res
= __builtin_mul_overflow(f
, factor
, &r
);
619 uint64_t dhi
= *phigh
;
620 uint64_t dlo
= *plow
;
625 mulu64(plow
, phigh
, dlo
, factor
);
629 mulu64(plow
, &ahi
, dlo
, factor
);
630 mulu64(&blo
, &bhi
, dhi
, factor
);
632 return uadd64_overflow(ahi
, blo
, phigh
) || bhi
!= 0;
637 * uadd64_carry - addition with carry-in and carry-out
639 * @pcarry: in-out carry value
641 * Computes @x + @y + *@pcarry, placing the carry-out back
642 * into *@pcarry and returning the 64-bit sum.
644 static inline uint64_t uadd64_carry(uint64_t x
, uint64_t y
, bool *pcarry
)
646 #if __has_builtin(__builtin_addcll)
647 unsigned long long c
= *pcarry
;
648 x
= __builtin_addcll(x
, y
, c
, &c
);
653 /* This is clang's internal expansion of __builtin_addc. */
654 c
= uadd64_overflow(x
, c
, &x
);
655 c
|= uadd64_overflow(x
, y
, &x
);
662 * usub64_borrow - subtraction with borrow-in and borrow-out
664 * @pborrow: in-out borrow value
666 * Computes @x - @y - *@pborrow, placing the borrow-out back
667 * into *@pborrow and returning the 64-bit sum.
669 static inline uint64_t usub64_borrow(uint64_t x
, uint64_t y
, bool *pborrow
)
671 #if __has_builtin(__builtin_subcll) && !defined(BUILTIN_SUBCLL_BROKEN)
672 unsigned long long b
= *pborrow
;
673 x
= __builtin_subcll(x
, y
, b
, &b
);
678 b
= usub64_overflow(x
, b
, &x
);
679 b
|= usub64_overflow(x
, y
, &x
);
685 /* Host type specific sizes of these routines. */
687 #if ULONG_MAX == UINT32_MAX
692 # define ctpopl ctpop32
693 # define revbitl revbit32
694 #elif ULONG_MAX == UINT64_MAX
699 # define ctpopl ctpop64
700 # define revbitl revbit64
702 # error Unknown sizeof long
705 static inline bool is_power_of_2(uint64_t value
)
711 return !(value
& (value
- 1));
715 * Return @value rounded down to the nearest power of two or zero.
717 static inline uint64_t pow2floor(uint64_t value
)
720 /* Avoid undefined shift by 64 */
723 return 0x8000000000000000ull
>> clz64(value
);
727 * Return @value rounded up to the nearest power of two modulo 2^64.
728 * This is *zero* for @value > 2^63, so be careful.
730 static inline uint64_t pow2ceil(uint64_t value
)
732 int n
= clz64(value
- 1);
736 * @value - 1 has no leading zeroes, thus @value - 1 >= 2^63
737 * Therefore, either @value == 0 or @value > 2^63.
738 * If it's 0, return 1, else return 0.
742 return 0x8000000000000000ull
>> (n
- 1);
745 static inline uint32_t pow2roundup32(uint32_t x
)
756 * urshift - 128-bit Unsigned Right Shift.
757 * @plow: in/out - lower 64-bit integer.
758 * @phigh: in/out - higher 64-bit integer.
759 * @shift: in - bytes to shift, between 0 and 127.
761 * Result is zero-extended and stored in plow/phigh, which are
762 * input/output variables. Shift values outside the range will
763 * be mod to 128. In other words, the caller is responsible to
764 * verify/assert both the shift range and plow/phigh pointers.
766 void urshift(uint64_t *plow
, uint64_t *phigh
, int32_t shift
);
769 * ulshift - 128-bit Unsigned Left Shift.
770 * @plow: in/out - lower 64-bit integer.
771 * @phigh: in/out - higher 64-bit integer.
772 * @shift: in - bytes to shift, between 0 and 127.
773 * @overflow: out - true if any 1-bit is shifted out.
775 * Result is zero-extended and stored in plow/phigh, which are
776 * input/output variables. Shift values outside the range will
777 * be mod to 128. In other words, the caller is responsible to
778 * verify/assert both the shift range and plow/phigh pointers.
780 void ulshift(uint64_t *plow
, uint64_t *phigh
, int32_t shift
, bool *overflow
);
782 /* From the GNU Multi Precision Library - longlong.h __udiv_qrnnd
783 * (https://gmplib.org/repo/gmp/file/tip/longlong.h)
785 * Licensed under the GPLv2/LGPLv3
787 static inline uint64_t udiv_qrnnd(uint64_t *r
, uint64_t n1
,
788 uint64_t n0
, uint64_t d
)
790 #if defined(__x86_64__)
792 asm("divq %4" : "=a"(q
), "=d"(*r
) : "0"(n0
), "1"(n1
), "rm"(d
));
794 #elif defined(__s390x__) && !defined(__clang__)
795 /* Need to use a TImode type to get an even register pair for DLGR. */
796 unsigned __int128 n
= (unsigned __int128
)n1
<< 64 | n0
;
797 asm("dlgr %0, %1" : "+r"(n
) : "r"(d
));
800 #elif defined(_ARCH_PPC64) && defined(_ARCH_PWR7)
801 /* From Power ISA 2.06, programming note for divdeu. */
802 uint64_t q1
, q2
, Q
, r1
, r2
, R
;
803 asm("divdeu %0,%2,%4; divdu %1,%3,%4"
804 : "=&r"(q1
), "=r"(q2
)
805 : "r"(n1
), "r"(n0
), "r"(d
));
806 r1
= -(q1
* d
); /* low part of (n1<<64) - (q1 * d) */
810 if (R
>= d
|| R
< r2
) { /* overflow implies R > d */
817 uint64_t d0
, d1
, q0
, q1
, r1
, r0
, m
;
825 r1
= (r1
<< 32) | (n0
>> 32);
841 r0
= (r0
<< 32) | (uint32_t)n0
;
855 return (q1
<< 32) | q0
;
859 Int128
divu256(Int128
*plow
, Int128
*phigh
, Int128 divisor
);
860 Int128
divs256(Int128
*plow
, Int128
*phigh
, Int128 divisor
);