aboutsummaryrefslogtreecommitdiffstats
path: root/host/lib/convert/convert_with_sse2.cpp
blob: 52beea24a6235a6a7b62fcc8316c98653d756e4c (plain)
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
//
// Copyright 2011-2011 Ettus Research LLC
//
// 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 3 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, see <http://www.gnu.org/licenses/>.
//

#include "convert_common.hpp"
#include <uhd/utils/byteswap.hpp>
#include <emmintrin.h>

using namespace uhd::convert;

DECLARE_CONVERTER(convert_fc32_1_to_item32_1_nswap, PRIORITY_CUSTOM){
    const fc32_t *input = reinterpret_cast<const fc32_t *>(inputs[0]);
    item32_t *output = reinterpret_cast<item32_t *>(outputs[0]);

    __m128 scalar = _mm_set_ps1(float(scale_factor));

    //convert blocks of samples with intrinsics
    size_t i = 0; for (; i < (nsamps & ~0x3); i+=4){
        //load from input
        __m128 tmplo = _mm_loadu_ps(reinterpret_cast<const float *>(input+i+0));
        __m128 tmphi = _mm_loadu_ps(reinterpret_cast<const float *>(input+i+2));

        //convert and scale
        __m128i tmpilo = _mm_cvtps_epi32(_mm_mul_ps(tmplo, scalar));
        __m128i tmpihi = _mm_cvtps_epi32(_mm_mul_ps(tmphi, scalar));

        //pack + swap 16-bit pairs
        __m128i tmpi = _mm_packs_epi32(tmpilo, tmpihi);
        tmpi = _mm_shufflelo_epi16(tmpi, _MM_SHUFFLE(2, 3, 0, 1));
        tmpi = _mm_shufflehi_epi16(tmpi, _MM_SHUFFLE(2, 3, 0, 1));

        //store to output
        _mm_storeu_si128(reinterpret_cast<__m128i *>(output+i), tmpi);
    }

    //convert remainder
    for (; i < nsamps; i++){
        output[i] = fc32_to_item32(input[i], float(scale_factor));
    }
}

DECLARE_CONVERTER(convert_fc32_1_to_item32_1_bswap, PRIORITY_CUSTOM){
    const fc32_t *input = reinterpret_cast<const fc32_t *>(inputs[0]);
    item32_t *output = reinterpret_cast<item32_t *>(outputs[0]);

    __m128 scalar = _mm_set_ps1(float(scale_factor));

    //convert blocks of samples with intrinsics
    size_t i = 0; for (; i < (nsamps & ~0x3); i+=4){
        //load from input
        __m128 tmplo = _mm_loadu_ps(reinterpret_cast<const float *>(input+i+0));
        __m128 tmphi = _mm_loadu_ps(reinterpret_cast<const float *>(input+i+2));

        //convert and scale
        __m128i tmpilo = _mm_cvtps_epi32(_mm_mul_ps(tmplo, scalar));
        __m128i tmpihi = _mm_cvtps_epi32(_mm_mul_ps(tmphi, scalar));

        //pack + byteswap -> byteswap 16 bit words
        __m128i tmpi = _mm_packs_epi32(tmpilo, tmpihi);
        tmpi = _mm_or_si128(_mm_srli_epi16(tmpi, 8), _mm_slli_epi16(tmpi, 8));

        //store to output
        _mm_storeu_si128(reinterpret_cast<__m128i *>(output+i), tmpi);
    }

    //convert remainder
    for (; i < nsamps; i++){
        output[i] = uhd::byteswap(fc32_to_item32(input[i], float(scale_factor)));
    }
}

DECLARE_CONVERTER(convert_item32_1_to_fc32_1_nswap, PRIORITY_CUSTOM){
    const item32_t *input = reinterpret_cast<const item32_t *>(inputs[0]);
    fc32_t *output = reinterpret_cast<fc32_t *>(outputs[0]);

    __m128 scalar = _mm_set_ps1(float(scale_factor)/(1 << 16));
    __m128i zeroi = _mm_setzero_si128();

    //convert blocks of samples with intrinsics
    size_t i = 0; for (; i < (nsamps & ~0x3); 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
        __m128 tmplo = _mm_mul_ps(_mm_cvtepi32_ps(tmpilo), scalar);
        __m128 tmphi = _mm_mul_ps(_mm_cvtepi32_ps(tmpihi), scalar);

        //store to output
        _mm_storeu_ps(reinterpret_cast<float *>(output+i+0), tmplo);
        _mm_storeu_ps(reinterpret_cast<float *>(output+i+2), tmphi);
    }

    //convert remainder
    for (; i < nsamps; i++){
        output[i] = item32_to_fc32(input[i], float(scale_factor));
    }
}

DECLARE_CONVERTER(convert_item32_1_to_fc32_1_bswap, PRIORITY_CUSTOM){
    const item32_t *input = reinterpret_cast<const item32_t *>(inputs[0]);
    fc32_t *output = reinterpret_cast<fc32_t *>(outputs[0]);

    __m128 scalar = _mm_set_ps1(float(scale_factor)/(1 << 16));
    __m128i zeroi = _mm_setzero_si128();

    //convert blocks of samples with intrinsics
    size_t i = 0; for (; i < (nsamps & ~0x3); 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
        __m128 tmplo = _mm_mul_ps(_mm_cvtepi32_ps(tmpilo), scalar);
        __m128 tmphi = _mm_mul_ps(_mm_cvtepi32_ps(tmpihi), scalar);

        //store to output
        _mm_storeu_ps(reinterpret_cast<float *>(output+i+0), tmplo);
        _mm_storeu_ps(reinterpret_cast<float *>(output+i+2), tmphi);
    }

    //convert remainder
    for (; i < nsamps; i++){
        output[i] = item32_to_fc32(uhd::byteswap(input[i]), float(scale_factor));
    }
}