1#ifndef _LINUX_MATH64_H
2#define _LINUX_MATH64_H
3
4#include <div64.h>
5#include <linux/bitops.h>
6#include <linux/types.h>
7
8#if BITS_PER_LONG == 64
9
10#define div64_long(x, y) div64_s64((x), (y))
11#define div64_ul(x, y)   div64_u64((x), (y))
12
13/**
14 * div_u64_rem - unsigned 64bit divide with 32bit divisor with remainder
15 *
16 * This is commonly provided by 32bit archs to provide an optimized 64bit
17 * divide.
18 */
19static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder)
20{
21	*remainder = dividend % divisor;
22	return dividend / divisor;
23}
24
25/**
26 * div_s64_rem - signed 64bit divide with 32bit divisor with remainder
27 */
28static inline s64 div_s64_rem(s64 dividend, s32 divisor, s32 *remainder)
29{
30	*remainder = dividend % divisor;
31	return dividend / divisor;
32}
33
34/**
35 * div64_u64_rem - unsigned 64bit divide with 64bit divisor and remainder
36 */
37static inline u64 div64_u64_rem(u64 dividend, u64 divisor, u64 *remainder)
38{
39	*remainder = dividend % divisor;
40	return dividend / divisor;
41}
42
43/**
44 * div64_u64 - unsigned 64bit divide with 64bit divisor
45 */
46static inline u64 div64_u64(u64 dividend, u64 divisor)
47{
48	return dividend / divisor;
49}
50
51#define DIV64_U64_ROUND_UP(ll, d)	\
52	({ u64 _tmp = (d); div64_u64((ll) + _tmp - 1, _tmp); })
53
54/**
55 * div64_s64 - signed 64bit divide with 64bit divisor
56 */
57static inline s64 div64_s64(s64 dividend, s64 divisor)
58{
59	return dividend / divisor;
60}
61
62#elif BITS_PER_LONG == 32
63
64#define div64_long(x, y) div_s64((x), (y))
65#define div64_ul(x, y)   div_u64((x), (y))
66
67#ifndef div_u64_rem
68static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder)
69{
70	*remainder = do_div(dividend, divisor);
71	return dividend;
72}
73#endif
74
75#ifndef div_s64_rem
76extern s64 div_s64_rem(s64 dividend, s32 divisor, s32 *remainder);
77#endif
78
79#ifndef div64_u64_rem
80extern u64 div64_u64_rem(u64 dividend, u64 divisor, u64 *remainder);
81#endif
82
83#ifndef div64_u64
84extern u64 div64_u64(u64 dividend, u64 divisor);
85#endif
86
87#ifndef div64_s64
88extern s64 div64_s64(s64 dividend, s64 divisor);
89#endif
90
91#endif /* BITS_PER_LONG */
92
93/**
94 * div_u64 - unsigned 64bit divide with 32bit divisor
95 *
96 * This is the most common 64bit divide and should be used if possible,
97 * as many 32bit archs can optimize this variant better than a full 64bit
98 * divide.
99 */
100#ifndef div_u64
101static inline u64 div_u64(u64 dividend, u32 divisor)
102{
103	u32 remainder;
104	return div_u64_rem(dividend, divisor, &remainder);
105}
106#endif
107
108/**
109 * div_s64 - signed 64bit divide with 32bit divisor
110 */
111#ifndef div_s64
112static inline s64 div_s64(s64 dividend, s32 divisor)
113{
114	s32 remainder;
115	return div_s64_rem(dividend, divisor, &remainder);
116}
117#endif
118
119u32 iter_div_u64_rem(u64 dividend, u32 divisor, u64 *remainder);
120
121static __always_inline u32
122__iter_div_u64_rem(u64 dividend, u32 divisor, u64 *remainder)
123{
124	u32 ret = 0;
125
126	while (dividend >= divisor) {
127		/* The following asm() prevents the compiler from
128		   optimising this loop into a modulo operation.  */
129		asm("" : "+rm"(dividend));
130
131		dividend -= divisor;
132		ret++;
133	}
134
135	*remainder = dividend;
136
137	return ret;
138}
139
140#ifndef mul_u32_u32
141/*
142 * Many a GCC version messes this up and generates a 64x64 mult :-(
143 */
144static inline u64 mul_u32_u32(u32 a, u32 b)
145{
146	return (u64)a * b;
147}
148#endif
149
150#if defined(CONFIG_ARCH_SUPPORTS_INT128) && defined(__SIZEOF_INT128__)
151
152#ifndef mul_u64_u32_shr
153static inline u64 mul_u64_u32_shr(u64 a, u32 mul, unsigned int shift)
154{
155	return (u64)(((unsigned __int128)a * mul) >> shift);
156}
157#endif /* mul_u64_u32_shr */
158
159#ifndef mul_u64_u64_shr
160static inline u64 mul_u64_u64_shr(u64 a, u64 mul, unsigned int shift)
161{
162	return (u64)(((unsigned __int128)a * mul) >> shift);
163}
164#endif /* mul_u64_u64_shr */
165
166#else
167
168#ifndef mul_u64_u32_shr
169static inline u64 mul_u64_u32_shr(u64 a, u32 mul, unsigned int shift)
170{
171	u32 ah, al;
172	u64 ret;
173
174	al = a;
175	ah = a >> 32;
176
177	ret = mul_u32_u32(al, mul) >> shift;
178	if (ah)
179		ret += mul_u32_u32(ah, mul) << (32 - shift);
180
181	return ret;
182}
183#endif /* mul_u64_u32_shr */
184
185#ifndef mul_u64_u64_shr
186static inline u64 mul_u64_u64_shr(u64 a, u64 b, unsigned int shift)
187{
188	union {
189		u64 ll;
190		struct {
191#ifdef __BIG_ENDIAN
192			u32 high, low;
193#else
194			u32 low, high;
195#endif
196		} l;
197	} rl, rm, rn, rh, a0, b0;
198	u64 c;
199
200	a0.ll = a;
201	b0.ll = b;
202
203	rl.ll = mul_u32_u32(a0.l.low, b0.l.low);
204	rm.ll = mul_u32_u32(a0.l.low, b0.l.high);
205	rn.ll = mul_u32_u32(a0.l.high, b0.l.low);
206	rh.ll = mul_u32_u32(a0.l.high, b0.l.high);
207
208	/*
209	 * Each of these lines computes a 64-bit intermediate result into "c",
210	 * starting at bits 32-95.  The low 32-bits go into the result of the
211	 * multiplication, the high 32-bits are carried into the next step.
212	 */
213	rl.l.high = c = (u64)rl.l.high + rm.l.low + rn.l.low;
214	rh.l.low = c = (c >> 32) + rm.l.high + rn.l.high + rh.l.low;
215	rh.l.high = (c >> 32) + rh.l.high;
216
217	/*
218	 * The 128-bit result of the multiplication is in rl.ll and rh.ll,
219	 * shift it right and throw away the high part of the result.
220	 */
221	if (shift == 0)
222		return rl.ll;
223	if (shift < 64)
224		return (rl.ll >> shift) | (rh.ll << (64 - shift));
225	return rh.ll >> (shift & 63);
226}
227#endif /* mul_u64_u64_shr */
228
229#endif
230
231#ifndef mul_u64_u32_div
232static inline u64 mul_u64_u32_div(u64 a, u32 mul, u32 divisor)
233{
234	union {
235		u64 ll;
236		struct {
237#ifdef __BIG_ENDIAN
238			u32 high, low;
239#else
240			u32 low, high;
241#endif
242		} l;
243	} u, rl, rh;
244
245	u.ll = a;
246	rl.ll = mul_u32_u32(u.l.low, mul);
247	rh.ll = mul_u32_u32(u.l.high, mul) + rl.l.high;
248
249	/* Bits 32-63 of the result will be in rh.l.low. */
250	rl.l.high = do_div(rh.ll, divisor);
251
252	/* Bits 0-31 of the result will be in rl.l.low.	*/
253	do_div(rl.ll, divisor);
254
255	rl.l.high = rh.l.low;
256	return rl.ll;
257}
258#endif /* mul_u64_u32_div */
259
260#endif /* _LINUX_MATH64_H */
261