1 #ifndef _LINUX_MATH64_H
2 #define _LINUX_MATH64_H
5 #include <linux/bitops.h>
6 #include <linux/types.h>
8 #if BITS_PER_LONG == 64
10 #define div64_long(x, y) div64_s64((x), (y))
11 #define div64_ul(x, y) div64_u64((x), (y))
14 * div_u64_rem - unsigned 64bit divide with 32bit divisor with remainder
16 * This is commonly provided by 32bit archs to provide an optimized 64bit
19 static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder)
21 *remainder = dividend % divisor;
22 return dividend / divisor;
26 * div_s64_rem - signed 64bit divide with 32bit divisor with remainder
28 static inline s64 div_s64_rem(s64 dividend, s32 divisor, s32 *remainder)
30 *remainder = dividend % divisor;
31 return dividend / divisor;
35 * div64_u64_rem - unsigned 64bit divide with 64bit divisor and remainder
37 static inline u64 div64_u64_rem(u64 dividend, u64 divisor, u64 *remainder)
39 *remainder = dividend % divisor;
40 return dividend / divisor;
44 * div64_u64 - unsigned 64bit divide with 64bit divisor
46 static inline u64 div64_u64(u64 dividend, u64 divisor)
48 return dividend / divisor;
52 * div64_s64 - signed 64bit divide with 64bit divisor
54 static inline s64 div64_s64(s64 dividend, s64 divisor)
56 return dividend / divisor;
59 #elif BITS_PER_LONG == 32
61 #define div64_long(x, y) div_s64((x), (y))
62 #define div64_ul(x, y) div_u64((x), (y))
65 static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder)
67 *remainder = do_div(dividend, divisor);
73 extern s64 div_s64_rem(s64 dividend, s32 divisor, s32 *remainder);
77 extern u64 div64_u64_rem(u64 dividend, u64 divisor, u64 *remainder);
81 extern u64 div64_u64(u64 dividend, u64 divisor);
85 extern s64 div64_s64(s64 dividend, s64 divisor);
88 #endif /* BITS_PER_LONG */
91 * div_u64 - unsigned 64bit divide with 32bit divisor
93 * This is the most common 64bit divide and should be used if possible,
94 * as many 32bit archs can optimize this variant better than a full 64bit
98 static inline u64 div_u64(u64 dividend, u32 divisor)
101 return div_u64_rem(dividend, divisor, &remainder);
106 * div_s64 - signed 64bit divide with 32bit divisor
109 static inline s64 div_s64(s64 dividend, s32 divisor)
112 return div_s64_rem(dividend, divisor, &remainder);
116 u32 iter_div_u64_rem(u64 dividend, u32 divisor, u64 *remainder);
118 static __always_inline u32
119 __iter_div_u64_rem(u64 dividend, u32 divisor, u64 *remainder)
123 while (dividend >= divisor) {
124 /* The following asm() prevents the compiler from
125 optimising this loop into a modulo operation. */
126 asm("" : "+rm"(dividend));
132 *remainder = dividend;
139 * Many a GCC version messes this up and generates a 64x64 mult :-(
141 static inline u64 mul_u32_u32(u32 a, u32 b)
147 #if defined(CONFIG_ARCH_SUPPORTS_INT128) && defined(__SIZEOF_INT128__)
149 #ifndef mul_u64_u32_shr
150 static inline u64 mul_u64_u32_shr(u64 a, u32 mul, unsigned int shift)
152 return (u64)(((unsigned __int128)a * mul) >> shift);
154 #endif /* mul_u64_u32_shr */
156 #ifndef mul_u64_u64_shr
157 static inline u64 mul_u64_u64_shr(u64 a, u64 mul, unsigned int shift)
159 return (u64)(((unsigned __int128)a * mul) >> shift);
161 #endif /* mul_u64_u64_shr */
165 #ifndef mul_u64_u32_shr
166 static inline u64 mul_u64_u32_shr(u64 a, u32 mul, unsigned int shift)
174 ret = mul_u32_u32(al, mul) >> shift;
176 ret += mul_u32_u32(ah, mul) << (32 - shift);
180 #endif /* mul_u64_u32_shr */
182 #ifndef mul_u64_u64_shr
183 static inline u64 mul_u64_u64_shr(u64 a, u64 b, unsigned int shift)
194 } rl, rm, rn, rh, a0, b0;
200 rl.ll = mul_u32_u32(a0.l.low, b0.l.low);
201 rm.ll = mul_u32_u32(a0.l.low, b0.l.high);
202 rn.ll = mul_u32_u32(a0.l.high, b0.l.low);
203 rh.ll = mul_u32_u32(a0.l.high, b0.l.high);
206 * Each of these lines computes a 64-bit intermediate result into "c",
207 * starting at bits 32-95. The low 32-bits go into the result of the
208 * multiplication, the high 32-bits are carried into the next step.
210 rl.l.high = c = (u64)rl.l.high + rm.l.low + rn.l.low;
211 rh.l.low = c = (c >> 32) + rm.l.high + rn.l.high + rh.l.low;
212 rh.l.high = (c >> 32) + rh.l.high;
215 * The 128-bit result of the multiplication is in rl.ll and rh.ll,
216 * shift it right and throw away the high part of the result.
221 return (rl.ll >> shift) | (rh.ll << (64 - shift));
222 return rh.ll >> (shift & 63);
224 #endif /* mul_u64_u64_shr */
228 #ifndef mul_u64_u32_div
229 static inline u64 mul_u64_u32_div(u64 a, u32 mul, u32 divisor)
243 rl.ll = mul_u32_u32(u.l.low, mul);
244 rh.ll = mul_u32_u32(u.l.high, mul) + rl.l.high;
246 /* Bits 32-63 of the result will be in rh.l.low. */
247 rl.l.high = do_div(rh.ll, divisor);
249 /* Bits 0-31 of the result will be in rl.l.low. */
250 do_div(rl.ll, divisor);
252 rl.l.high = rh.l.low;
255 #endif /* mul_u64_u32_div */
257 #endif /* _LINUX_MATH64_H */