FD.io VPP  v18.01-8-g0eacf49
Vector Packet Processing
math64.h
Go to the documentation of this file.
1 /*
2  * math64.h provides the 64 bit unsigned integer add, multiply followed by modulo operation
3  * The linux/math64.h provides divide and multiply 64 bit integers but:
4  * 1. multiply: mul_u64_u64_shr - only returns 64 bits of the result and has to be called
5  * twice to get the complete 128 bits of the result.
6  * 2. Modulo operation of the result of addition and multiplication of u64 that may result
7  * in integers > 64 bits is not supported
8  * Hence this header to combine add/multiply followed by modulo of u64 integrers
9  * always resulting in u64.
10  *
11  * Copyright (c) 2016 Cisco and/or its affiliates.
12  * Licensed under the Apache License, Version 2.0 (the "License");
13  * you may not use this file except in compliance with the License.
14  * You may obtain a copy of the License at:
15  *
16  * http://www.apache.org/licenses/LICENSE-2.0
17  *
18  * Unless required by applicable law or agreed to in writing, software
19  * distributed under the License is distributed on an "AS IS" BASIS,
20  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
21  * See the License for the specific language governing permissions and
22  * limitations under the License.
23  */
24 #ifndef include_vnet_math64_h
25 #define include_vnet_math64_h
26 #include <stdint.h>
27 
28 /*
29  * multiplies and returns result in hi and lo
30  */
31 static inline void mul64by64(u64 a, u64 b, u64 * hi, u64 * lo)
32 {
33  u64 a_lo = (u64) (uint32_t) a;
34  u64 a_hi = a >> 32;
35  u64 b_lo = (u64) (u32) b;
36  u64 b_hi = b >> 32;
37 
38  u64 p0 = a_lo * b_lo;
39  u64 p1 = a_lo * b_hi;
40  u64 p2 = a_hi * b_lo;
41  u64 p3 = a_hi * b_hi;
42 
43  u32 cy = (u32) (((p0 >> 32) + (u32) p1 + (u32) p2) >> 32);
44 
45  *lo = p0 + (p1 << 32) + (p2 << 32);
46  *hi = p3 + (p1 >> 32) + (p2 >> 32) + cy;
47  return;
48 }
49 
50 #define TWO64 18446744073709551616.0
51 
52 static inline u64 mod128by64(u64 x, u64 y, u64 m, double di)
53 {
54  u64 q1, q2, q;
55  u64 p1, p0;
56  double dq;
57 
58  /* calculate quotient first pass 53 bits */
59  dq = (TWO64 * (double)x + (double)y) * di;
60 
61  if (dq >= TWO64)
62  q1 = 0xfffffffffffff800L;
63  else
64  q1 = dq;
65 
66  /* q1 * m to compare the product to the dividend. */
67  mul64by64(q1, m, &p1, &p0);
68 
69  /* Adjust quotient. is it > actual result: */
70  if (x < p1 || (x == p1 && y < p0))
71  {
72  /* q1 > quotient. calculate abs remainder */
73  x = p1 - (x + (p0 < y));
74  y = p0 - y;
75 
76  /* use the remainder as new dividend to adjust quotient */
77  q2 = (u64) ((TWO64 * (double)x + (double)y) * di);
78  mul64by64(q2, m, &p1, &p0);
79 
80  q = q1 - q2;
81  if (x < p1 || (x == p1 && y <= p0))
82  {
83  y = p0 - y;
84  }
85  else
86  {
87  y = p0 - y;
88  y += m;
89  q--;
90  }
91  }
92  else
93  {
94  x = x - (p1 + (y < p0));
95  y = y - p0;
96 
97  q2 = (u64) ((TWO64 * (double)x + (double)y) * di);
98  mul64by64(q2, m, &p1, &p0);
99 
100  q = q1 + q2;
101  if (x < p1 || (x == p1 && y < p0))
102  {
103  y = y - p0;
104  y += m;
105  q--;
106  }
107  else
108  {
109  y = y - p0;
110  if (y >= m)
111  {
112  y -= m;
113  q++;
114  }
115  }
116  }
117 
118  return y;
119 }
120 
121 /*
122  * returns a % p
123  */
124 static inline u64 mod64by64(u64 a, u64 p, u64 primeinv)
125 {
126  return (mod128by64(0, a, p, primeinv));
127 }
128 
129 static inline void add64(u64 a, u64 b, u64 * whi, u64 * wlo)
130 {
131  *wlo = a + b;
132  if (*wlo < a)
133  *whi = 1;
134 
135 }
136 
137 /*
138  * returns (a + b)%p
139  */
140 static inline u64 add64_mod(u64 a, u64 b, u64 p, double pi)
141 {
142  u64 shi = 0, slo = 0;
143 
144  add64(a, b, &shi, &slo);
145  return (mod128by64(shi, slo, p, pi));
146 }
147 
148 /*
149  * returns (ab) % p
150  */
151 static inline u64 mul64_mod(u64 a, u64 b, u64 p, double pi)
152 {
153  u64 phi = 0, plo = 0;
154 
155  mul64by64(a, b, &phi, &plo);
156  return (mod128by64(phi, plo, p, pi));
157 }
158 
159 #endif
vmrglw vmrglh hi
a
Definition: bitmap.h:516
#define TWO64
Definition: math64.h:50
static void add64(u64 a, u64 b, u64 *whi, u64 *wlo)
Definition: math64.h:129
static u64 mul64_mod(u64 a, u64 b, u64 p, double pi)
Definition: math64.h:151
static u64 add64_mod(u64 a, u64 b, u64 p, double pi)
Definition: math64.h:140
static u64 mod128by64(u64 x, u64 y, u64 m, double di)
Definition: math64.h:52
add_epi add_epi sub_epi sub_epi adds_epu subs_epu i16x8 y
Definition: vector_sse2.h:293
void di(unformat_input_t *i)
Definition: unformat.c:163
unsigned long u64
Definition: types.h:89
static void mul64by64(u64 a, u64 b, u64 *hi, u64 *lo)
Definition: math64.h:31
static u64 mod64by64(u64 a, u64 p, u64 primeinv)
Definition: math64.h:124
unsigned int u32
Definition: types.h:88