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
|
/*
* Intel SSE + AVX Viterbi decoder
*
* Copyright (C) 2013, 2014 Thomas Tsou <tom@tsou.cc>
*
* All Rights Reserved
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License along
* with this program; if not, write to the Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
*/
#include <stdint.h>
#include "config.h"
#include <emmintrin.h>
#include <tmmintrin.h>
#include <xmmintrin.h>
#include <immintrin.h>
#if defined(HAVE_SSE4_1)
#include <smmintrin.h>
#endif
#define SSE_ALIGN 16
/* Broadcast 16-bit integer
* Repeat the low 16-bit integer to all elements of the 128-bit SSE
* register. Only AVX2 has a dedicated broadcast instruction; use repeat
* unpacks for SSE only architectures. This is a destructive operation and
* the source register is overwritten.
*
* Input:
* M0 - Low 16-bit element is read
*
* Output:
* M0 - Contains broadcasted values
*/
#define SSE_BROADCAST(M0) \
{ \
M0 = _mm_broadcastw_epi16(M0); \
}
/**
* Include common SSE implementation
*/
#include <viterbi_sse_common.h>
/* Aligned Memory Allocator
* SSE requires 16-byte memory alignment. We store relevant trellis values
* (accumulated sums, outputs, and path decisions) as 16 bit signed integers
* so the allocated memory is casted as such.
*/
__attribute__ ((visibility("hidden")))
int16_t *osmo_conv_sse_avx_vdec_malloc(size_t n)
{
return (int16_t *) _mm_malloc(sizeof(int16_t) * n, SSE_ALIGN);
}
__attribute__ ((visibility("hidden")))
void osmo_conv_sse_avx_vdec_free(int16_t *ptr)
{
_mm_free(ptr);
}
__attribute__ ((visibility("hidden")))
void osmo_conv_sse_avx_metrics_k5_n2(const int8_t *val,
const int16_t *out, int16_t *sums, int16_t *paths, int norm)
{
const int16_t _val[4] = { val[0], val[1], val[0], val[1] };
_sse_metrics_k5_n2(_val, out, sums, paths, norm);
}
__attribute__ ((visibility("hidden")))
void osmo_conv_sse_avx_metrics_k5_n3(const int8_t *val,
const int16_t *out, int16_t *sums, int16_t *paths, int norm)
{
const int16_t _val[4] = { val[0], val[1], val[2], 0 };
_sse_metrics_k5_n4(_val, out, sums, paths, norm);
}
__attribute__ ((visibility("hidden")))
void osmo_conv_sse_avx_metrics_k5_n4(const int8_t *val,
const int16_t *out, int16_t *sums, int16_t *paths, int norm)
{
const int16_t _val[4] = { val[0], val[1], val[2], val[3] };
_sse_metrics_k5_n4(_val, out, sums, paths, norm);
}
__attribute__ ((visibility("hidden")))
void osmo_conv_sse_avx_metrics_k7_n2(const int8_t *val,
const int16_t *out, int16_t *sums, int16_t *paths, int norm)
{
const int16_t _val[4] = { val[0], val[1], val[0], val[1] };
_sse_metrics_k7_n2(_val, out, sums, paths, norm);
}
__attribute__ ((visibility("hidden")))
void osmo_conv_sse_avx_metrics_k7_n3(const int8_t *val,
const int16_t *out, int16_t *sums, int16_t *paths, int norm)
{
const int16_t _val[4] = { val[0], val[1], val[2], 0 };
_sse_metrics_k7_n4(_val, out, sums, paths, norm);
}
__attribute__ ((visibility("hidden")))
void osmo_conv_sse_avx_metrics_k7_n4(const int8_t *val,
const int16_t *out, int16_t *sums, int16_t *paths, int norm)
{
const int16_t _val[4] = { val[0], val[1], val[2], val[3] };
_sse_metrics_k7_n4(_val, out, sums, paths, norm);
}
|
