stream_encoder_intrin_sse2.c
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
/* libFLAC - Free Lossless Audio Codec library
* Copyright (C) 2000-2009 Josh Coalson
* Copyright (C) 2011-2016 Xiph.Org Foundation
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* - Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
*
* - Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* - Neither the name of the Xiph.org Foundation nor the names of its
* contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#ifdef HAVE_CONFIG_H
# include <config.h>
#endif
#include "private/cpu.h"
#ifndef FLAC__NO_ASM
#if (defined FLAC__CPU_IA32 || defined FLAC__CPU_X86_64) && FLAC__HAS_X86INTRIN
#include "private/stream_encoder.h"
#include "private/bitmath.h"
#ifdef FLAC__SSE2_SUPPORTED
#include <stdlib.h> /* for abs() */
#include <emmintrin.h> /* SSE2 */
#include "FLAC/assert.h"
#include "share/compat.h"
FLAC__SSE_TARGET("sse2")
static inline __m128i local_abs_epi32(__m128i val)
{
__m128i mask = _mm_srai_epi32(val, 31);
val = _mm_xor_si128(val, mask);
val = _mm_sub_epi32(val, mask);
return val;
}
FLAC__SSE_TARGET("sse2")
void FLAC__precompute_partition_info_sums_intrin_sse2(const FLAC__int32 residual[], FLAC__uint64 abs_residual_partition_sums[],
62
uint32_t residual_samples, uint32_t predictor_order, uint32_t min_partition_order, uint32_t max_partition_order, uint32_t bps)
64
65
const uint32_t default_partition_samples = (residual_samples + predictor_order) >> max_partition_order;
uint32_t partitions = 1u << max_partition_order;
66
67
68
69
70
FLAC__ASSERT(default_partition_samples > predictor_order);
/* first do max_partition_order */
{
71
72
const uint32_t threshold = 32 - FLAC__bitmath_ilog2(default_partition_samples);
uint32_t partition, residual_sample, end = (uint32_t)(-(int32_t)predictor_order);
73
74
75
76
if(bps + FLAC__MAX_EXTRA_RESIDUAL_BPS < threshold) {
for(partition = residual_sample = 0; partition < partitions; partition++) {
__m128i mm_sum = _mm_setzero_si128();
77
uint32_t e1, e3;
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
end += default_partition_samples;
e1 = (residual_sample + 3) & ~3; e3 = end & ~3;
if(e1 > end)
e1 = end; /* try flac -l 1 -b 16 and you'll be here */
/* assumption: residual[] is properly aligned so (residual + e1) is properly aligned too and _mm_loadu_si128() is fast */
for( ; residual_sample < e1; residual_sample++) {
__m128i mm_res = local_abs_epi32(_mm_cvtsi32_si128(residual[residual_sample]));
mm_sum = _mm_add_epi32(mm_sum, mm_res);
}
for( ; residual_sample < e3; residual_sample+=4) {
__m128i mm_res = local_abs_epi32(_mm_loadu_si128((const __m128i*)(residual+residual_sample)));
mm_sum = _mm_add_epi32(mm_sum, mm_res);
}
for( ; residual_sample < end; residual_sample++) {
__m128i mm_res = local_abs_epi32(_mm_cvtsi32_si128(residual[residual_sample]));
mm_sum = _mm_add_epi32(mm_sum, mm_res);
}
100
101
mm_sum = _mm_add_epi32(mm_sum, _mm_shuffle_epi32(mm_sum, _MM_SHUFFLE(1,0,3,2)));
mm_sum = _mm_add_epi32(mm_sum, _mm_shufflelo_epi16(mm_sum, _MM_SHUFFLE(1,0,3,2)));
102
abs_residual_partition_sums[partition] = (FLAC__uint32)_mm_cvtsi128_si32(mm_sum);
103
104
/* workaround for MSVC bugs (at least versions 2015 and 2017 are affected) */
#if (defined _MSC_VER) && (defined FLAC__CPU_X86_64)
105
106
107
108
109
110
111
abs_residual_partition_sums[partition] &= 0xFFFFFFFF;
#endif
}
}
else { /* have to pessimistically use 64 bits for accumulator */
for(partition = residual_sample = 0; partition < partitions; partition++) {
__m128i mm_sum = _mm_setzero_si128();
112
uint32_t e1, e3;
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
end += default_partition_samples;
e1 = (residual_sample + 1) & ~1; e3 = end & ~1;
FLAC__ASSERT(e1 <= end);
for( ; residual_sample < e1; residual_sample++) {
__m128i mm_res = local_abs_epi32(_mm_cvtsi32_si128(residual[residual_sample])); /* 0 0 0 |r0| == 00 |r0_64| */
mm_sum = _mm_add_epi64(mm_sum, mm_res);
}
for( ; residual_sample < e3; residual_sample+=2) {
__m128i mm_res = local_abs_epi32(_mm_loadl_epi64((const __m128i*)(residual+residual_sample))); /* 0 0 |r1| |r0| */
mm_res = _mm_shuffle_epi32(mm_res, _MM_SHUFFLE(3,1,2,0)); /* 0 |r1| 0 |r0| == |r1_64| |r0_64| */
mm_sum = _mm_add_epi64(mm_sum, mm_res);
}
for( ; residual_sample < end; residual_sample++) {
__m128i mm_res = local_abs_epi32(_mm_cvtsi32_si128(residual[residual_sample]));
mm_sum = _mm_add_epi64(mm_sum, mm_res);
}
mm_sum = _mm_add_epi64(mm_sum, _mm_srli_si128(mm_sum, 8));
_mm_storel_epi64((__m128i*)(abs_residual_partition_sums+partition), mm_sum);
}
}
}
/* now merge partitions for lower orders */
{
142
uint32_t from_partition = 0, to_partition = partitions;
143
144
int partition_order;
for(partition_order = (int)max_partition_order - 1; partition_order >= (int)min_partition_order; partition_order--) {
146
147
148
149
150
151
152
153
154
155
156
157
158
159
partitions >>= 1;
for(i = 0; i < partitions; i++) {
abs_residual_partition_sums[to_partition++] =
abs_residual_partition_sums[from_partition ] +
abs_residual_partition_sums[from_partition+1];
from_partition += 2;
}
}
}
}
#endif /* FLAC__SSE2_SUPPORTED */
#endif /* (FLAC__CPU_IA32 || FLAC__CPU_X86_64) && FLAC__HAS_X86INTRIN */
#endif /* FLAC__NO_ASM */