aboutsummaryrefslogtreecommitdiffstats
path: root/Transceiver52M/x86/convert.c
blob: 862a2e738ad1427e2f8fd872f9771a2a0f3b0c5a (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
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
/*
 * 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.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include <malloc.h>
#include <string.h>
#include "convert.h"

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#ifdef HAVE_SSE3
#include <xmmintrin.h>
#include <emmintrin.h>

#ifdef HAVE_SSE4_1
#include <smmintrin.h>

/* 16*N 16-bit signed integer converted to single precision floats */
static void _sse_convert_si16_ps_16n(float *restrict out,
				     const short *restrict in,
				     int len)
{
	__m128i m0, m1, m2, m3, m4, m5;
	__m128 m6, m7, m8, m9;

	for (int i = 0; i < len / 16; i++) {
		/* Load (unaligned) packed floats */
		m0 = _mm_loadu_si128((__m128i *) &in[16 * i + 0]);
		m1 = _mm_loadu_si128((__m128i *) &in[16 * i + 8]);

		/* Unpack */
		m2 = _mm_cvtepi16_epi32(m0);
		m4 = _mm_cvtepi16_epi32(m1);
		m0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1, 0, 3, 2));
		m1 = _mm_shuffle_epi32(m1, _MM_SHUFFLE(1, 0, 3, 2));
		m3 = _mm_cvtepi16_epi32(m0);
		m5 = _mm_cvtepi16_epi32(m1);

		/* Convert */
		m6 = _mm_cvtepi32_ps(m2);
		m7 = _mm_cvtepi32_ps(m3);
		m8 = _mm_cvtepi32_ps(m4);
		m9 = _mm_cvtepi32_ps(m5);

		/* Store */
		_mm_storeu_ps(&out[16 * i + 0], m6);
		_mm_storeu_ps(&out[16 * i + 4], m7);
		_mm_storeu_ps(&out[16 * i + 8], m8);
		_mm_storeu_ps(&out[16 * i + 12], m9);
	}
}

/* 16*N 16-bit signed integer conversion with remainder */
static void _sse_convert_si16_ps(float *restrict out,
				 const short *restrict in,
				 int len)
{
	int start = len / 16 * 16;

	_sse_convert_si16_ps_16n(out, in, len);

	for (int i = 0; i < len % 16; i++)
		out[start + i] = in[start + i];
}
#endif /* HAVE_SSE4_1 */

/* 8*N single precision floats scaled and converted to 16-bit signed integer */
static 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 */
static 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 */
static 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);
	}
}
#else /* HAVE_SSE3 */
static void convert_scale_ps_si16(short *out, const float *in,
				  float scale, int len)
{
	for (int i = 0; i < len; i++)
		out[i] = in[i] * scale;
}
#endif

#ifndef HAVE_SSE4_1
static void convert_si16_ps(float *out, const short *in, int len)
{
	for (int i = 0; i < len; i++)
		out[i] = in[i];
}
#endif

void convert_float_short(short *out, const float *in, float scale, int len)
{
	void (*conv_func)(short *, const float *, float, int);

#ifdef HAVE_SSE3
	if (!(len % 16))
		conv_func = _sse_convert_scale_ps_si16_16n;
	else if (!(len % 8))
		conv_func = _sse_convert_scale_ps_si16_8n;
	else
		conv_func = _sse_convert_scale_ps_si16;
#else
	conv_func = convert_scale_ps_si16;
#endif

	conv_func(out, in, scale, len);
}

void convert_short_float(float *out, const short *in, int len)
{
	void (*conv_func) (float *, const short *, int);

#ifdef HAVE_SSE4_1
	if (!(len % 16))
		conv_func = _sse_convert_si16_ps_16n;
	else
		conv_func = _sse_convert_si16_ps;
#else
	conv_func = convert_si16_ps;
#endif

	conv_func(out, in, len);
}