xref: /linux/include/linux/math64.h (revision 621cde16e49b3ecf7d59a8106a20aaebfb4a59a9)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MATH64_H
3 #define _LINUX_MATH64_H
4 
5 #include <linux/types.h>
6 #include <linux/math.h>
7 #include <asm/div64.h>
8 #include <vdso/math64.h>
9 
10 #if BITS_PER_LONG == 64
11 
12 #define div64_long(x, y) div64_s64((x), (y))
13 #define div64_ul(x, y)   div64_u64((x), (y))
14 
15 /**
16  * div_u64_rem - unsigned 64bit divide with 32bit divisor with remainder
17  * @dividend: unsigned 64bit dividend
18  * @divisor: unsigned 32bit divisor
19  * @remainder: pointer to unsigned 32bit remainder
20  *
21  * Return: sets ``*remainder``, then returns dividend / divisor
22  *
23  * This is commonly provided by 32bit archs to provide an optimized 64bit
24  * divide.
25  */
div_u64_rem(u64 dividend,u32 divisor,u32 * remainder)26 static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder)
27 {
28 	*remainder = dividend % divisor;
29 	return dividend / divisor;
30 }
31 
32 /**
33  * div_s64_rem - signed 64bit divide with 32bit divisor with remainder
34  * @dividend: signed 64bit dividend
35  * @divisor: signed 32bit divisor
36  * @remainder: pointer to signed 32bit remainder
37  *
38  * Return: sets ``*remainder``, then returns dividend / divisor
39  */
div_s64_rem(s64 dividend,s32 divisor,s32 * remainder)40 static inline s64 div_s64_rem(s64 dividend, s32 divisor, s32 *remainder)
41 {
42 	*remainder = dividend % divisor;
43 	return dividend / divisor;
44 }
45 
46 /**
47  * div64_u64_rem - unsigned 64bit divide with 64bit divisor and remainder
48  * @dividend: unsigned 64bit dividend
49  * @divisor: unsigned 64bit divisor
50  * @remainder: pointer to unsigned 64bit remainder
51  *
52  * Return: sets ``*remainder``, then returns dividend / divisor
53  */
div64_u64_rem(u64 dividend,u64 divisor,u64 * remainder)54 static inline u64 div64_u64_rem(u64 dividend, u64 divisor, u64 *remainder)
55 {
56 	*remainder = dividend % divisor;
57 	return dividend / divisor;
58 }
59 
60 /**
61  * div64_u64 - unsigned 64bit divide with 64bit divisor
62  * @dividend: unsigned 64bit dividend
63  * @divisor: unsigned 64bit divisor
64  *
65  * Return: dividend / divisor
66  */
div64_u64(u64 dividend,u64 divisor)67 static inline u64 div64_u64(u64 dividend, u64 divisor)
68 {
69 	return dividend / divisor;
70 }
71 
72 /**
73  * div64_s64 - signed 64bit divide with 64bit divisor
74  * @dividend: signed 64bit dividend
75  * @divisor: signed 64bit divisor
76  *
77  * Return: dividend / divisor
78  */
div64_s64(s64 dividend,s64 divisor)79 static inline s64 div64_s64(s64 dividend, s64 divisor)
80 {
81 	return dividend / divisor;
82 }
83 
84 #elif BITS_PER_LONG == 32
85 
86 #define div64_long(x, y) div_s64((x), (y))
87 #define div64_ul(x, y)   div_u64((x), (y))
88 
89 #ifndef div_u64_rem
div_u64_rem(u64 dividend,u32 divisor,u32 * remainder)90 static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder)
91 {
92 	*remainder = do_div(dividend, divisor);
93 	return dividend;
94 }
95 #endif
96 
97 #ifndef div_s64_rem
98 extern s64 div_s64_rem(s64 dividend, s32 divisor, s32 *remainder);
99 #endif
100 
101 #ifndef div64_u64_rem
102 extern u64 div64_u64_rem(u64 dividend, u64 divisor, u64 *remainder);
103 #endif
104 
105 #ifndef div64_u64
106 extern u64 div64_u64(u64 dividend, u64 divisor);
107 #endif
108 
109 #ifndef div64_s64
110 extern s64 div64_s64(s64 dividend, s64 divisor);
111 #endif
112 
113 #endif /* BITS_PER_LONG */
114 
115 /**
116  * div_u64 - unsigned 64bit divide with 32bit divisor
117  * @dividend: unsigned 64bit dividend
118  * @divisor: unsigned 32bit divisor
119  *
120  * This is the most common 64bit divide and should be used if possible,
121  * as many 32bit archs can optimize this variant better than a full 64bit
122  * divide.
123  *
124  * Return: dividend / divisor
125  */
126 #ifndef div_u64
div_u64(u64 dividend,u32 divisor)127 static inline u64 div_u64(u64 dividend, u32 divisor)
128 {
129 	u32 remainder;
130 	return div_u64_rem(dividend, divisor, &remainder);
131 }
132 #endif
133 
134 /**
135  * div_s64 - signed 64bit divide with 32bit divisor
136  * @dividend: signed 64bit dividend
137  * @divisor: signed 32bit divisor
138  *
139  * Return: dividend / divisor
140  */
141 #ifndef div_s64
div_s64(s64 dividend,s32 divisor)142 static inline s64 div_s64(s64 dividend, s32 divisor)
143 {
144 	s32 remainder;
145 	return div_s64_rem(dividend, divisor, &remainder);
146 }
147 #endif
148 
149 u32 iter_div_u64_rem(u64 dividend, u32 divisor, u64 *remainder);
150 
151 #ifndef mul_u32_u32
152 /*
153  * Many a GCC version messes this up and generates a 64x64 mult :-(
154  */
mul_u32_u32(u32 a,u32 b)155 static inline u64 mul_u32_u32(u32 a, u32 b)
156 {
157 	return (u64)a * b;
158 }
159 #endif
160 
161 #if defined(CONFIG_ARCH_SUPPORTS_INT128) && defined(__SIZEOF_INT128__)
162 
163 #ifndef mul_u64_u32_shr
mul_u64_u32_shr(u64 a,u32 mul,unsigned int shift)164 static __always_inline u64 mul_u64_u32_shr(u64 a, u32 mul, unsigned int shift)
165 {
166 	return (u64)(((unsigned __int128)a * mul) >> shift);
167 }
168 #endif /* mul_u64_u32_shr */
169 
170 #ifndef mul_u64_u64_shr
mul_u64_u64_shr(u64 a,u64 mul,unsigned int shift)171 static __always_inline u64 mul_u64_u64_shr(u64 a, u64 mul, unsigned int shift)
172 {
173 	return (u64)(((unsigned __int128)a * mul) >> shift);
174 }
175 #endif /* mul_u64_u64_shr */
176 
177 #else
178 
179 #ifndef mul_u64_u32_shr
mul_u64_u32_shr(u64 a,u32 mul,unsigned int shift)180 static __always_inline u64 mul_u64_u32_shr(u64 a, u32 mul, unsigned int shift)
181 {
182 	u32 ah = a >> 32, al = a;
183 	u64 ret;
184 
185 	ret = mul_u32_u32(al, mul) >> shift;
186 	if (ah)
187 		ret += mul_u32_u32(ah, mul) << (32 - shift);
188 	return ret;
189 }
190 #endif /* mul_u64_u32_shr */
191 
192 #ifndef mul_u64_u64_shr
mul_u64_u64_shr(u64 a,u64 b,unsigned int shift)193 static inline u64 mul_u64_u64_shr(u64 a, u64 b, unsigned int shift)
194 {
195 	union {
196 		u64 ll;
197 		struct {
198 #ifdef __BIG_ENDIAN
199 			u32 high, low;
200 #else
201 			u32 low, high;
202 #endif
203 		} l;
204 	} rl, rm, rn, rh, a0, b0;
205 	u64 c;
206 
207 	a0.ll = a;
208 	b0.ll = b;
209 
210 	rl.ll = mul_u32_u32(a0.l.low, b0.l.low);
211 	rm.ll = mul_u32_u32(a0.l.low, b0.l.high);
212 	rn.ll = mul_u32_u32(a0.l.high, b0.l.low);
213 	rh.ll = mul_u32_u32(a0.l.high, b0.l.high);
214 
215 	/*
216 	 * Each of these lines computes a 64-bit intermediate result into "c",
217 	 * starting at bits 32-95.  The low 32-bits go into the result of the
218 	 * multiplication, the high 32-bits are carried into the next step.
219 	 */
220 	rl.l.high = c = (u64)rl.l.high + rm.l.low + rn.l.low;
221 	rh.l.low = c = (c >> 32) + rm.l.high + rn.l.high + rh.l.low;
222 	rh.l.high = (c >> 32) + rh.l.high;
223 
224 	/*
225 	 * The 128-bit result of the multiplication is in rl.ll and rh.ll,
226 	 * shift it right and throw away the high part of the result.
227 	 */
228 	if (shift == 0)
229 		return rl.ll;
230 	if (shift < 64)
231 		return (rl.ll >> shift) | (rh.ll << (64 - shift));
232 	return rh.ll >> (shift & 63);
233 }
234 #endif /* mul_u64_u64_shr */
235 
236 #endif
237 
238 #ifndef mul_s64_u64_shr
mul_s64_u64_shr(s64 a,u64 b,unsigned int shift)239 static inline u64 mul_s64_u64_shr(s64 a, u64 b, unsigned int shift)
240 {
241 	u64 ret;
242 
243 	/*
244 	 * Extract the sign before the multiplication and put it back
245 	 * afterwards if needed.
246 	 */
247 	ret = mul_u64_u64_shr(abs(a), b, shift);
248 
249 	if (a < 0)
250 		ret = -((s64) ret);
251 
252 	return ret;
253 }
254 #endif /* mul_s64_u64_shr */
255 
256 #ifndef mul_u64_u32_div
mul_u64_u32_div(u64 a,u32 mul,u32 divisor)257 static inline u64 mul_u64_u32_div(u64 a, u32 mul, u32 divisor)
258 {
259 	union {
260 		u64 ll;
261 		struct {
262 #ifdef __BIG_ENDIAN
263 			u32 high, low;
264 #else
265 			u32 low, high;
266 #endif
267 		} l;
268 	} u, rl, rh;
269 
270 	u.ll = a;
271 	rl.ll = mul_u32_u32(u.l.low, mul);
272 	rh.ll = mul_u32_u32(u.l.high, mul) + rl.l.high;
273 
274 	/* Bits 32-63 of the result will be in rh.l.low. */
275 	rl.l.high = do_div(rh.ll, divisor);
276 
277 	/* Bits 0-31 of the result will be in rl.l.low.	*/
278 	do_div(rl.ll, divisor);
279 
280 	rl.l.high = rh.l.low;
281 	return rl.ll;
282 }
283 #endif /* mul_u64_u32_div */
284 
285 u64 mul_u64_u64_div_u64(u64 a, u64 mul, u64 div);
286 
287 /**
288  * DIV64_U64_ROUND_UP - unsigned 64bit divide with 64bit divisor rounded up
289  * @ll: unsigned 64bit dividend
290  * @d: unsigned 64bit divisor
291  *
292  * Divide unsigned 64bit dividend by unsigned 64bit divisor
293  * and round up.
294  *
295  * Return: dividend / divisor rounded up
296  */
297 #define DIV64_U64_ROUND_UP(ll, d)	\
298 	({ u64 _tmp = (d); div64_u64((ll) + _tmp - 1, _tmp); })
299 
300 /**
301  * DIV64_U64_ROUND_CLOSEST - unsigned 64bit divide with 64bit divisor rounded to nearest integer
302  * @dividend: unsigned 64bit dividend
303  * @divisor: unsigned 64bit divisor
304  *
305  * Divide unsigned 64bit dividend by unsigned 64bit divisor
306  * and round to closest integer.
307  *
308  * Return: dividend / divisor rounded to nearest integer
309  */
310 #define DIV64_U64_ROUND_CLOSEST(dividend, divisor)	\
311 	({ u64 _tmp = (divisor); div64_u64((dividend) + _tmp / 2, _tmp); })
312 
313 /**
314  * DIV_U64_ROUND_CLOSEST - unsigned 64bit divide with 32bit divisor rounded to nearest integer
315  * @dividend: unsigned 64bit dividend
316  * @divisor: unsigned 32bit divisor
317  *
318  * Divide unsigned 64bit dividend by unsigned 32bit divisor
319  * and round to closest integer.
320  *
321  * Return: dividend / divisor rounded to nearest integer
322  */
323 #define DIV_U64_ROUND_CLOSEST(dividend, divisor)	\
324 	({ u32 _tmp = (divisor); div_u64((u64)(dividend) + _tmp / 2, _tmp); })
325 
326 /**
327  * DIV_S64_ROUND_CLOSEST - signed 64bit divide with 32bit divisor rounded to nearest integer
328  * @dividend: signed 64bit dividend
329  * @divisor: signed 32bit divisor
330  *
331  * Divide signed 64bit dividend by signed 32bit divisor
332  * and round to closest integer.
333  *
334  * Return: dividend / divisor rounded to nearest integer
335  */
336 #define DIV_S64_ROUND_CLOSEST(dividend, divisor)(	\
337 {							\
338 	s64 __x = (dividend);				\
339 	s32 __d = (divisor);				\
340 	((__x > 0) == (__d > 0)) ?			\
341 		div_s64((__x + (__d / 2)), __d) :	\
342 		div_s64((__x - (__d / 2)), __d);	\
343 }							\
344 )
345 #endif /* _LINUX_MATH64_H */
346