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
|
//
// Copyright 2011-2012 Ettus Research LLC
// Copyright 2018 Ettus Research, a National Instruments Company
//
// SPDX-License-Identifier: GPL-3.0-or-later
//
#include "convert_common.hpp"
#include <uhd/utils/byteswap.hpp>
#include <emmintrin.h>
using namespace uhd::convert;
DECLARE_CONVERTER(sc16_item32_le, 1, fc64, 1, PRIORITY_SIMD)
{
const item32_t* input = reinterpret_cast<const item32_t*>(inputs[0]);
fc64_t* output = reinterpret_cast<fc64_t*>(outputs[0]);
const __m128d scalar = _mm_set1_pd(scale_factor / (1 << 16));
const __m128i zeroi = _mm_setzero_si128();
#define convert_item32_1_to_fc64_1_nswap_guts(_al_) \
for (; i + 3 < nsamps; i += 4) { \
/* load from input */ \
__m128i tmpi = _mm_loadu_si128(reinterpret_cast<const __m128i*>(input + i)); \
\
/* unpack + swap 16-bit pairs */ \
tmpi = _mm_shufflelo_epi16(tmpi, _MM_SHUFFLE(2, 3, 0, 1)); \
tmpi = _mm_shufflehi_epi16(tmpi, _MM_SHUFFLE(2, 3, 0, 1)); \
__m128i tmpilo = _mm_unpacklo_epi16(zeroi, tmpi); /* value in upper 16 bits */ \
__m128i tmpihi = _mm_unpackhi_epi16(zeroi, tmpi); \
\
/* convert and scale */ \
__m128d tmp0 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
tmpilo = _mm_unpackhi_epi64(tmpilo, zeroi); \
__m128d tmp1 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
__m128d tmp2 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
tmpihi = _mm_unpackhi_epi64(tmpihi, zeroi); \
__m128d tmp3 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
\
/* store to output */ \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 0), tmp0); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 1), tmp1); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 2), tmp2); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 3), tmp3); \
}
size_t i = 0;
// dispatch according to alignment
if ((size_t(output) & 0xf) == 0) {
convert_item32_1_to_fc64_1_nswap_guts(_)
} else {
convert_item32_1_to_fc64_1_nswap_guts(u_)
}
// convert remainder
item32_sc16_to_xx<uhd::htowx>(input + i, output + i, nsamps - i, scale_factor);
}
DECLARE_CONVERTER(sc16_item32_be, 1, fc64, 1, PRIORITY_SIMD)
{
const item32_t* input = reinterpret_cast<const item32_t*>(inputs[0]);
fc64_t* output = reinterpret_cast<fc64_t*>(outputs[0]);
const __m128d scalar = _mm_set1_pd(scale_factor / (1 << 16));
const __m128i zeroi = _mm_setzero_si128();
#define convert_item32_1_to_fc64_1_bswap_guts(_al_) \
for (; i + 3 < nsamps; i += 4) { \
/* load from input */ \
__m128i tmpi = _mm_loadu_si128(reinterpret_cast<const __m128i*>(input + i)); \
\
/* byteswap + unpack -> byteswap 16 bit words */ \
tmpi = _mm_or_si128(_mm_srli_epi16(tmpi, 8), _mm_slli_epi16(tmpi, 8)); \
__m128i tmpilo = _mm_unpacklo_epi16(zeroi, tmpi); /* value in upper 16 bits */ \
__m128i tmpihi = _mm_unpackhi_epi16(zeroi, tmpi); \
\
/* convert and scale */ \
__m128d tmp0 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
tmpilo = _mm_unpackhi_epi64(tmpilo, zeroi); \
__m128d tmp1 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
__m128d tmp2 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
tmpihi = _mm_unpackhi_epi64(tmpihi, zeroi); \
__m128d tmp3 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
\
/* store to output */ \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 0), tmp0); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 1), tmp1); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 2), tmp2); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 3), tmp3); \
}
size_t i = 0;
// dispatch according to alignment
if ((size_t(output) & 0xf) == 0) {
convert_item32_1_to_fc64_1_bswap_guts(_)
} else {
convert_item32_1_to_fc64_1_bswap_guts(u_)
}
// convert remainder
item32_sc16_to_xx<uhd::htonx>(input + i, output + i, nsamps - i, scale_factor);
}
DECLARE_CONVERTER(sc16_chdr, 1, fc64, 1, PRIORITY_SIMD)
{
const sc16_t* input = reinterpret_cast<const sc16_t*>(inputs[0]);
fc64_t* output = reinterpret_cast<fc64_t*>(outputs[0]);
const __m128d scalar = _mm_set1_pd(scale_factor / (1 << 16));
const __m128i zeroi = _mm_setzero_si128();
#define convert_chdr_1_to_fc64_1_guts(_al_) \
for (; i + 3 < nsamps; i += 4) { \
/* load from input */ \
__m128i tmpi = _mm_loadu_si128(reinterpret_cast<const __m128i*>(input + i)); \
\
/* unpack 16-bit pairs */ \
__m128i tmpilo = _mm_unpacklo_epi16(zeroi, tmpi); /* value in upper 16 bits */ \
__m128i tmpihi = _mm_unpackhi_epi16(zeroi, tmpi); \
\
/* convert and scale */ \
__m128d tmp0 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
tmpilo = _mm_unpackhi_epi64(tmpilo, zeroi); \
__m128d tmp1 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
__m128d tmp2 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
tmpihi = _mm_unpackhi_epi64(tmpihi, zeroi); \
__m128d tmp3 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
\
/* store to output */ \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 0), tmp0); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 1), tmp1); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 2), tmp2); \
_mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 3), tmp3); \
}
size_t i = 0;
// dispatch according to alignment
if ((size_t(output) & 0xf) == 0) {
convert_chdr_1_to_fc64_1_guts(_)
} else {
convert_chdr_1_to_fc64_1_guts(u_)
}
// convert remainder
chdr_sc16_to_xx(input + i, output + i, nsamps - i, scale_factor);
}
|
