ce188d4d |
1 | /* libFLAC - Free Lossless Audio Codec library |
2 | * Copyright (C) 2000-2009 Josh Coalson |
3 | * Copyright (C) 2011-2016 Xiph.Org Foundation |
4 | * |
5 | * Redistribution and use in source and binary forms, with or without |
6 | * modification, are permitted provided that the following conditions |
7 | * are met: |
8 | * |
9 | * - Redistributions of source code must retain the above copyright |
10 | * notice, this list of conditions and the following disclaimer. |
11 | * |
12 | * - Redistributions in binary form must reproduce the above copyright |
13 | * notice, this list of conditions and the following disclaimer in the |
14 | * documentation and/or other materials provided with the distribution. |
15 | * |
16 | * - Neither the name of the Xiph.org Foundation nor the names of its |
17 | * contributors may be used to endorse or promote products derived from |
18 | * this software without specific prior written permission. |
19 | * |
20 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
21 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
22 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
23 | * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR |
24 | * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
25 | * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
26 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
27 | * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF |
28 | * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING |
29 | * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS |
30 | * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
31 | */ |
32 | |
33 | #ifdef HAVE_CONFIG_H |
34 | # include <config.h> |
35 | #endif |
36 | |
37 | #include "private/cpu.h" |
38 | |
39 | #ifndef FLAC__NO_ASM |
40 | #if (defined FLAC__CPU_IA32 || defined FLAC__CPU_X86_64) && FLAC__HAS_X86INTRIN |
41 | #include "private/stream_encoder.h" |
42 | #include "private/bitmath.h" |
43 | #ifdef FLAC__SSE2_SUPPORTED |
44 | |
45 | #include <stdlib.h> /* for abs() */ |
46 | #include <emmintrin.h> /* SSE2 */ |
47 | #include "FLAC/assert.h" |
48 | #include "share/compat.h" |
49 | |
50 | FLAC__SSE_TARGET("sse2") |
51 | static inline __m128i local_abs_epi32(__m128i val) |
52 | { |
53 | __m128i mask = _mm_srai_epi32(val, 31); |
54 | val = _mm_xor_si128(val, mask); |
55 | val = _mm_sub_epi32(val, mask); |
56 | return val; |
57 | } |
58 | |
59 | |
60 | FLAC__SSE_TARGET("sse2") |
61 | void FLAC__precompute_partition_info_sums_intrin_sse2(const FLAC__int32 residual[], FLAC__uint64 abs_residual_partition_sums[], |
62 | unsigned residual_samples, unsigned predictor_order, unsigned min_partition_order, unsigned max_partition_order, unsigned bps) |
63 | { |
64 | const unsigned default_partition_samples = (residual_samples + predictor_order) >> max_partition_order; |
65 | unsigned partitions = 1u << max_partition_order; |
66 | |
67 | FLAC__ASSERT(default_partition_samples > predictor_order); |
68 | |
69 | /* first do max_partition_order */ |
70 | { |
71 | const unsigned threshold = 32 - FLAC__bitmath_ilog2(default_partition_samples); |
72 | unsigned partition, residual_sample, end = (unsigned)(-(int)predictor_order); |
73 | |
74 | if(bps + FLAC__MAX_EXTRA_RESIDUAL_BPS < threshold) { |
75 | for(partition = residual_sample = 0; partition < partitions; partition++) { |
76 | __m128i mm_sum = _mm_setzero_si128(); |
77 | unsigned e1, e3; |
78 | end += default_partition_samples; |
79 | |
80 | e1 = (residual_sample + 3) & ~3; e3 = end & ~3; |
81 | if(e1 > end) |
82 | e1 = end; /* try flac -l 1 -b 16 and you'll be here */ |
83 | |
84 | /* assumption: residual[] is properly aligned so (residual + e1) is properly aligned too and _mm_loadu_si128() is fast */ |
85 | for( ; residual_sample < e1; residual_sample++) { |
86 | __m128i mm_res = local_abs_epi32(_mm_cvtsi32_si128(residual[residual_sample])); |
87 | mm_sum = _mm_add_epi32(mm_sum, mm_res); |
88 | } |
89 | |
90 | for( ; residual_sample < e3; residual_sample+=4) { |
91 | __m128i mm_res = local_abs_epi32(_mm_loadu_si128((const __m128i*)(residual+residual_sample))); |
92 | mm_sum = _mm_add_epi32(mm_sum, mm_res); |
93 | } |
94 | |
95 | for( ; residual_sample < end; residual_sample++) { |
96 | __m128i mm_res = local_abs_epi32(_mm_cvtsi32_si128(residual[residual_sample])); |
97 | mm_sum = _mm_add_epi32(mm_sum, mm_res); |
98 | } |
99 | |
100 | mm_sum = _mm_add_epi32(mm_sum, _mm_srli_si128(mm_sum, 8)); |
101 | mm_sum = _mm_add_epi32(mm_sum, _mm_srli_si128(mm_sum, 4)); |
102 | abs_residual_partition_sums[partition] = (FLAC__uint32)_mm_cvtsi128_si32(mm_sum); |
103 | /* workaround for a bug in MSVC2015U2 - see https://connect.microsoft.com/VisualStudio/feedback/details/2659191/incorrect-code-generation-for-x86-64 */ |
104 | #if (defined _MSC_VER) && (_MSC_FULL_VER == 190023918) && (defined FLAC__CPU_X86_64) |
105 | abs_residual_partition_sums[partition] &= 0xFFFFFFFF; |
106 | #endif |
107 | } |
108 | } |
109 | else { /* have to pessimistically use 64 bits for accumulator */ |
110 | for(partition = residual_sample = 0; partition < partitions; partition++) { |
111 | __m128i mm_sum = _mm_setzero_si128(); |
112 | unsigned e1, e3; |
113 | end += default_partition_samples; |
114 | |
115 | e1 = (residual_sample + 1) & ~1; e3 = end & ~1; |
116 | FLAC__ASSERT(e1 <= end); |
117 | |
118 | for( ; residual_sample < e1; residual_sample++) { |
119 | __m128i mm_res = local_abs_epi32(_mm_cvtsi32_si128(residual[residual_sample])); /* 0 0 0 |r0| == 00 |r0_64| */ |
120 | mm_sum = _mm_add_epi64(mm_sum, mm_res); |
121 | } |
122 | |
123 | for( ; residual_sample < e3; residual_sample+=2) { |
124 | __m128i mm_res = local_abs_epi32(_mm_loadl_epi64((const __m128i*)(residual+residual_sample))); /* 0 0 |r1| |r0| */ |
125 | mm_res = _mm_shuffle_epi32(mm_res, _MM_SHUFFLE(3,1,2,0)); /* 0 |r1| 0 |r0| == |r1_64| |r0_64| */ |
126 | mm_sum = _mm_add_epi64(mm_sum, mm_res); |
127 | } |
128 | |
129 | for( ; residual_sample < end; residual_sample++) { |
130 | __m128i mm_res = local_abs_epi32(_mm_cvtsi32_si128(residual[residual_sample])); |
131 | mm_sum = _mm_add_epi64(mm_sum, mm_res); |
132 | } |
133 | |
134 | mm_sum = _mm_add_epi64(mm_sum, _mm_srli_si128(mm_sum, 8)); |
135 | _mm_storel_epi64((__m128i*)(abs_residual_partition_sums+partition), mm_sum); |
136 | } |
137 | } |
138 | } |
139 | |
140 | /* now merge partitions for lower orders */ |
141 | { |
142 | unsigned from_partition = 0, to_partition = partitions; |
143 | int partition_order; |
144 | for(partition_order = (int)max_partition_order - 1; partition_order >= (int)min_partition_order; partition_order--) { |
145 | unsigned i; |
146 | partitions >>= 1; |
147 | for(i = 0; i < partitions; i++) { |
148 | abs_residual_partition_sums[to_partition++] = |
149 | abs_residual_partition_sums[from_partition ] + |
150 | abs_residual_partition_sums[from_partition+1]; |
151 | from_partition += 2; |
152 | } |
153 | } |
154 | } |
155 | } |
156 | |
157 | #endif /* FLAC__SSE2_SUPPORTED */ |
158 | #endif /* (FLAC__CPU_IA32 || FLAC__CPU_X86_64) && FLAC__HAS_X86INTRIN */ |
159 | #endif /* FLAC__NO_ASM */ |