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
|
/*
* SSE type conversions
* Copyright (C) 2013 Thomas Tsou <tom@tsou.cc>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*/
#include <malloc.h>
#include <string.h>
#include "convert_sse_3.h"
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#ifdef HAVE_SSE3
#include <xmmintrin.h>
#include <emmintrin.h>
/* 8*N single precision floats scaled and converted to 16-bit signed integer */
__attribute__((xray_never_instrument))
void _sse_convert_scale_ps_si16_8n(short *restrict out,
const float *restrict in,
float scale, int len)
{
__m128 m0, m1, m2;
__m128i m4, m5;
for (int i = 0; i < len / 8; i++) {
/* Load (unaligned) packed floats */
m0 = _mm_loadu_ps(&in[8 * i + 0]);
m1 = _mm_loadu_ps(&in[8 * i + 4]);
m2 = _mm_load1_ps(&scale);
/* Scale */
m0 = _mm_mul_ps(m0, m2);
m1 = _mm_mul_ps(m1, m2);
/* Convert */
m4 = _mm_cvtps_epi32(m0);
m5 = _mm_cvtps_epi32(m1);
/* Pack and store */
m5 = _mm_packs_epi32(m4, m5);
_mm_storeu_si128((__m128i *) & out[8 * i], m5);
}
}
/* 8*N single precision floats scaled and converted with remainder */
__attribute__((xray_never_instrument))
void _sse_convert_scale_ps_si16(short *restrict out,
const float *restrict in, float scale, int len)
{
int start = len / 8 * 8;
_sse_convert_scale_ps_si16_8n(out, in, scale, len);
for (int i = 0; i < len % 8; i++)
out[start + i] = in[start + i] * scale;
}
/* 16*N single precision floats scaled and converted to 16-bit signed integer */
__attribute__((xray_never_instrument))
void _sse_convert_scale_ps_si16_16n(short *restrict out,
const float *restrict in,
float scale, int len)
{
__m128 m0, m1, m2, m3, m4;
__m128i m5, m6, m7, m8;
for (int i = 0; i < len / 16; i++) {
/* Load (unaligned) packed floats */
m0 = _mm_loadu_ps(&in[16 * i + 0]);
m1 = _mm_loadu_ps(&in[16 * i + 4]);
m2 = _mm_loadu_ps(&in[16 * i + 8]);
m3 = _mm_loadu_ps(&in[16 * i + 12]);
m4 = _mm_load1_ps(&scale);
/* Scale */
m0 = _mm_mul_ps(m0, m4);
m1 = _mm_mul_ps(m1, m4);
m2 = _mm_mul_ps(m2, m4);
m3 = _mm_mul_ps(m3, m4);
/* Convert */
m5 = _mm_cvtps_epi32(m0);
m6 = _mm_cvtps_epi32(m1);
m7 = _mm_cvtps_epi32(m2);
m8 = _mm_cvtps_epi32(m3);
/* Pack and store */
m5 = _mm_packs_epi32(m5, m6);
m7 = _mm_packs_epi32(m7, m8);
_mm_storeu_si128((__m128i *) & out[16 * i + 0], m5);
_mm_storeu_si128((__m128i *) & out[16 * i + 8], m7);
}
}
#endif
|
