1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
|
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "vis_proto.h"
/***************************************************************/
typedef int t_s32;
typedef unsigned int t_u32;
#if defined(__sparcv9)
typedef long t_s64;
typedef unsigned long t_u64;
#else
typedef long long t_s64;
typedef unsigned long long t_u64;
#endif
typedef double t_d64;
/***************************************************************/
typedef union {
t_d64 d64;
struct {
t_s32 i0;
t_s32 i1;
} i32s;
} d64_2_i32;
/***************************************************************/
#define BUFF_SIZE 256
#define A_BITS 19
#define A_MASK ((1 << A_BITS) - 1)
/***************************************************************/
static t_u64 mask_cnst[] = {
0x8000000080000000ull
};
/***************************************************************/
#define DEF_VARS(N) \
t_d64 *py = (t_d64 *)y; \
t_d64 mask = *((t_d64 *)mask_cnst); \
t_d64 ca = (1u << 31) - 1; \
t_d64 da = (t_d64)a; \
t_s64 buff[N], s; \
d64_2_i32 dy
/***************************************************************/
#define MUL_U32_S64_2(i) \
dy.d64 = vis_fxnor(mask, py[i]); \
buff[2 * (i)] = (ca - (t_d64)dy.i32s.i0) * da; \
buff[2 * (i) + 1] = (ca - (t_d64)dy.i32s.i1) * da
#define MUL_U32_S64_2_D(i) \
dy.d64 = vis_fxnor(mask, py[i]); \
d0 = ca - (t_d64)dy.i32s.i0; \
d1 = ca - (t_d64)dy.i32s.i1; \
buff[4 * (i)] = (t_s64)(d0 * da); \
buff[4 * (i) + 1] = (t_s64)(d0 * db); \
buff[4 * (i) + 2] = (t_s64)(d1 * da); \
buff[4 * (i) + 3] = (t_s64)(d1 * db)
/***************************************************************/
#define ADD_S64_U32(i) \
s = buff[i] + x[i] + c; \
z[i] = s; \
c = (s >> 32)
#define ADD_S64_U32_D(i) \
s = buff[2 * (i)] + (((t_s64)(buff[2 * (i) + 1])) << A_BITS) + x[i] + uc; \
z[i] = s; \
uc = ((t_u64)s >> 32)
/***************************************************************/
#define MUL_U32_S64_8(i) \
MUL_U32_S64_2(i); \
MUL_U32_S64_2(i + 1); \
MUL_U32_S64_2(i + 2); \
MUL_U32_S64_2(i + 3)
#define MUL_U32_S64_D_8(i) \
MUL_U32_S64_2_D(i); \
MUL_U32_S64_2_D(i + 1); \
MUL_U32_S64_2_D(i + 2); \
MUL_U32_S64_2_D(i + 3)
/***************************************************************/
#define ADD_S64_U32_8(i) \
ADD_S64_U32(i); \
ADD_S64_U32(i + 1); \
ADD_S64_U32(i + 2); \
ADD_S64_U32(i + 3); \
ADD_S64_U32(i + 4); \
ADD_S64_U32(i + 5); \
ADD_S64_U32(i + 6); \
ADD_S64_U32(i + 7)
#define ADD_S64_U32_D_8(i) \
ADD_S64_U32_D(i); \
ADD_S64_U32_D(i + 1); \
ADD_S64_U32_D(i + 2); \
ADD_S64_U32_D(i + 3); \
ADD_S64_U32_D(i + 4); \
ADD_S64_U32_D(i + 5); \
ADD_S64_U32_D(i + 6); \
ADD_S64_U32_D(i + 7)
/***************************************************************/
t_u32
mul_add(t_u32 *z, t_u32 *x, t_u32 *y, int n, t_u32 a)
{
if (a < (1 << A_BITS)) {
if (n == 8) {
DEF_VARS(8);
t_s32 c = 0;
MUL_U32_S64_8(0);
ADD_S64_U32_8(0);
return c;
} else if (n == 16) {
DEF_VARS(16);
t_s32 c = 0;
MUL_U32_S64_8(0);
MUL_U32_S64_8(4);
ADD_S64_U32_8(0);
ADD_S64_U32_8(8);
return c;
} else {
DEF_VARS(BUFF_SIZE);
t_s32 i, c = 0;
#pragma pipeloop(0)
for (i = 0; i < (n + 1) / 2; i++) {
MUL_U32_S64_2(i);
}
#pragma pipeloop(0)
for (i = 0; i < n; i++) {
ADD_S64_U32(i);
}
return c;
}
} else {
if (n == 8) {
DEF_VARS(2 * 8);
t_d64 d0, d1, db;
t_u32 uc = 0;
da = (t_d64)(a & A_MASK);
db = (t_d64)(a >> A_BITS);
MUL_U32_S64_D_8(0);
ADD_S64_U32_D_8(0);
return uc;
} else if (n == 16) {
DEF_VARS(2 * 16);
t_d64 d0, d1, db;
t_u32 uc = 0;
da = (t_d64)(a & A_MASK);
db = (t_d64)(a >> A_BITS);
MUL_U32_S64_D_8(0);
MUL_U32_S64_D_8(4);
ADD_S64_U32_D_8(0);
ADD_S64_U32_D_8(8);
return uc;
} else {
DEF_VARS(2 * BUFF_SIZE);
t_d64 d0, d1, db;
t_u32 i, uc = 0;
da = (t_d64)(a & A_MASK);
db = (t_d64)(a >> A_BITS);
#pragma pipeloop(0)
for (i = 0; i < (n + 1) / 2; i++) {
MUL_U32_S64_2_D(i);
}
#pragma pipeloop(0)
for (i = 0; i < n; i++) {
ADD_S64_U32_D(i);
}
return uc;
}
}
}
/***************************************************************/
t_u32
mul_add_inp(t_u32 *x, t_u32 *y, int n, t_u32 a)
{
return mul_add(x, x, y, n, a);
}
/***************************************************************/
|
