-
Notifications
You must be signed in to change notification settings - Fork 35
/
opencl_lm_kernel_params.h
170 lines (151 loc) · 5.03 KB
/
opencl_lm_kernel_params.h
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
#include "opencl_lm_hst_dev_shared.h"
#include "opencl_device_info.h"
typedef unsigned WORD vtype;
/*
* Some devices/drivers has problems with the optimized 'goto' program flow.
* Some AMD driver versions can't build the "fast goto" version but those who
* can runs faster. Hawaii on 14.9 fails, Tahiti on 14.9 does not (!?).
*
* Nvidia can build either kernel but GTX980 is significantly faster with the
* "safe goto" version (7% faster for one salt, 16% for many salts).
*
* OSX' Intel HD4000 driver [1.2(Sep25 2014 22:26:04)] fails building the
* "fast goto" version.
*/
#if nvidia_sm_5x(DEVICE_INFO) || gpu_intel(DEVICE_INFO) || \
(gpu_amd(DEVICE_INFO) && DEV_VER_MAJOR >= 1573 && !defined(__Tahiti__)) || \
(gpu_amd(DEVICE_INFO) && DEV_VER_MAJOR >= 1702)
//#warning Using 'safe goto' kernel
#define SAFE_GOTO
#else
//#warning Using 'fast goto' kernel
#endif
#if no_byte_addressable(DEVICE_INFO)
#define RV7xx
#endif
#if gpu_nvidia(DEVICE_INFO)
#define _NV
#endif
#define vxorf(a, b) \
((a) ^ (b))
#define vnot(dst, a) \
(dst) = ~(a)
#define vand(dst, a, b) \
(dst) = (a) & (b)
#define vor(dst, a, b) \
(dst) = (a) | (b)
#define vandn(dst, a, b) \
(dst) = (a) & ~(b)
#define vxor(dst, a, b) \
(dst) = vxorf((a), (b))
#define vshl(dst, src, shift) \
(dst) = (src) << (shift)
#define vshr(dst, src, shift) \
(dst) = (src) >> (shift)
#define vshl1(dst, src) \
vshl((dst), (src), 1)
#if HAVE_LUT3
#define vsel(dst, a, b, c) (dst) = lut3(a, b, c, 0xd8)
#elif defined(_NV) || __CPU__
#define vsel(dst, a, b, c) \
(dst) = (((a) & ~(c)) ^ ((b) & (c)))
#else
#define vsel(dst, a, b, c) \
(dst) = bitselect((a),(b),(c))
#endif
#if defined(_NV) || __CPU__
#include "opencl_sboxes.h"
#else
#include "opencl_sboxes-s.h"
#endif
#define vst_private(dst, ofs, src) \
*((__private vtype *)((__private lm_vector *)&(dst) + (ofs))) = (src)
#define lm_clear_block_8(j) \
vst_private(B[j] , 0, zero); \
vst_private(B[j] , 1, zero); \
vst_private(B[j] , 2, zero); \
vst_private(B[j] , 3, zero); \
vst_private(B[j] , 4, zero); \
vst_private(B[j] , 5, zero); \
vst_private(B[j] , 6, zero); \
vst_private(B[j] , 7, zero);
#define lm_clear_block \
lm_clear_block_8(0); \
lm_clear_block_8(8); \
lm_clear_block_8(16); \
lm_clear_block_8(24); \
lm_clear_block_8(32); \
lm_clear_block_8(40); \
lm_clear_block_8(48); \
lm_clear_block_8(56);
#if BITMAP_SIZE_BITS_LESS_ONE < 0xffffffff
#define BITMAP_SIZE_BITS (BITMAP_SIZE_BITS_LESS_ONE + 1)
#else
/*undefined, cause error.*/
#endif
#define GET_HASH_0(hash, x, k, bits) \
for (bit = bits; bit < k; bit++) \
hash |= ((((uint)B[bit]) >> x) & 1) << bit;
#define GET_HASH_1(hash, x, k, bits) \
for (bit = bits; bit < k; bit++) \
hash |= ((((uint)B[32 + bit]) >> x) & 1) << bit;
inline void cmp_final(__private unsigned lm_vector *B,
__private unsigned int *binary,
__global unsigned int *offset_table,
__global unsigned int *hash_table,
volatile __global uint *output,
volatile __global uint *bitmap_dupe,
unsigned int depth,
unsigned int section,
unsigned int iter)
{
unsigned long hash;
unsigned int hash_table_index, t, bit;
#if SELECT_CMP_STEPS > 1
GET_HASH_0(binary[0], depth, 32, REQ_BITMAP_BITS);
GET_HASH_1(binary[1], depth, 32, REQ_BITMAP_BITS);
#else
binary[0] = 0;
GET_HASH_0(binary[0], depth, 32, 0);
GET_HASH_1(binary[1], depth, 32, REQ_BITMAP_BITS);
#endif
hash = ((unsigned long)binary[1] << 32) | (unsigned long)binary[0];
hash += (unsigned long)offset_table[hash % OFFSET_TABLE_SIZE];
hash_table_index = hash % HASH_TABLE_SIZE;
if (hash_table[hash_table_index + HASH_TABLE_SIZE] == binary[1])
if (hash_table[hash_table_index] == binary[0])
if (!(atomic_or(&bitmap_dupe[hash_table_index/32], (1U << (hash_table_index % 32))) & (1U << (hash_table_index % 32)))) {
t = atomic_inc(&output[0]);
output[1 + 3 * t] = (section * 32) + depth;
output[2 + 3 * t] = iter;
output[3 + 3 * t] = hash_table_index;
}
}
inline void cmp( __private unsigned lm_vector *B,
__global unsigned int *offset_table,
__global unsigned int *hash_table,
__global unsigned int *bitmaps,
volatile __global uint *output,
volatile __global uint *bitmap_dupe,
int section, unsigned int iter) {
unsigned int value[2] , i, bit, bitmap_index;
for (i = 0; i < 32; i++){
#if SELECT_CMP_STEPS > 1
value[0] = 0;
value[1] = 0;
GET_HASH_0(value[0], i, REQ_BITMAP_BITS, 0);
GET_HASH_1(value[1], i, REQ_BITMAP_BITS, 0);
bitmap_index = value[1] & (BITMAP_SIZE_BITS - 1);
bit = (bitmaps[bitmap_index >> 5] >> (bitmap_index & 31)) & 1U;
bitmap_index = value[0] & (BITMAP_SIZE_BITS - 1);
bit &= (bitmaps[(BITMAP_SIZE_BITS >> 5) + (bitmap_index >> 5)] >> (bitmap_index & 31)) & 1U;
#else
value[1] = 0;
GET_HASH_1(value[1], i, REQ_BITMAP_BITS, 0);
bitmap_index = value[1] & BITMAP_SIZE_BITS_LESS_ONE;
bit = (bitmaps[bitmap_index >> 5] >> (bitmap_index & 31)) & 1U;
#endif
if (bit)
cmp_final(B, value, offset_table, hash_table, output, bitmap_dupe, i, section, iter);
}
}