aa52709b087194d69231dc46f4e15e25dfa7f4b7
[deb_x265.git] / source / common / vec / dct-sse41.cpp
1 /*****************************************************************************
2 * Copyright (C) 2013 x265 project
3 *
4 * Authors: Steve Borho <steve@borho.org>
5 * Mandar Gurav <mandar@multicorewareinc.com>
6 * Deepthi Devaki Akkoorath <deepthidevaki@multicorewareinc.com>
7 * Mahesh Pittala <mahesh@multicorewareinc.com>
8 * Rajesh Paulraj <rajesh@multicorewareinc.com>
9 * Min Chen <min.chen@multicorewareinc.com>
10 * Praveen Kumar Tiwari <praveen@multicorewareinc.com>
11 * Nabajit Deka <nabajit@multicorewareinc.com>
12 *
13 * This program is free software; you can redistribute it and/or modify
14 * it under the terms of the GNU General Public License as published by
15 * the Free Software Foundation; either version 2 of the License, or
16 * (at your option) any later version.
17 *
18 * This program is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21 * GNU General Public License for more details.
22 *
23 * You should have received a copy of the GNU General Public License
24 * along with this program; if not, write to the Free Software
25 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
26 *
27 * This program is also available under a commercial proprietary license.
28 * For more information, contact us at license @ x265.com.
29 *****************************************************************************/
30
31 #include "common.h"
32 #include "primitives.h"
33 #include <xmmintrin.h> // SSE
34 #include <smmintrin.h> // SSE4.1
35
36 using namespace x265;
37
38 namespace {
39 void dequant_scaling(const int16_t* quantCoef, const int32_t *deQuantCoef, int32_t* coef, int num, int per, int shift)
40 {
41 X265_CHECK(num <= 32 * 32, "dequant num too large\n");
42
43 int valueToAdd;
44
45 shift += 4;
46
47 if (shift > per)
48 {
49 valueToAdd = 1 << (shift - per - 1);
50 __m128i IAdd = _mm_set1_epi32(valueToAdd);
51
52 for (int n = 0; n < num; n = n + 8)
53 {
54 __m128i quantCoef1, quantCoef2, deQuantCoef1, deQuantCoef2, quantCoef12, sign;
55
56 quantCoef12 = _mm_loadu_si128((__m128i*)(quantCoef + n));
57
58 deQuantCoef1 = _mm_loadu_si128((__m128i*)(deQuantCoef + n));
59 deQuantCoef2 = _mm_loadu_si128((__m128i*)(deQuantCoef + n + 4));
60
61 sign = _mm_srai_epi16(quantCoef12, 15);
62 quantCoef1 = _mm_unpacklo_epi16(quantCoef12, sign);
63 quantCoef2 = _mm_unpackhi_epi16(quantCoef12, sign);
64
65 quantCoef1 = _mm_sra_epi32(_mm_add_epi32(_mm_mullo_epi32(quantCoef1, deQuantCoef1), IAdd), _mm_cvtsi32_si128(shift - per));
66 quantCoef2 = _mm_sra_epi32(_mm_add_epi32(_mm_mullo_epi32(quantCoef2, deQuantCoef2), IAdd), _mm_cvtsi32_si128(shift - per));
67
68 quantCoef12 = _mm_packs_epi32(quantCoef1, quantCoef2);
69 sign = _mm_srai_epi16(quantCoef12, 15);
70 quantCoef1 = _mm_unpacklo_epi16(quantCoef12, sign);
71 _mm_storeu_si128((__m128i*)(coef + n), quantCoef1);
72 quantCoef2 = _mm_unpackhi_epi16(quantCoef12, sign);
73 _mm_storeu_si128((__m128i*)(coef + n + 4), quantCoef2);
74 }
75 }
76 else
77 {
78 for (int n = 0; n < num; n = n + 8)
79 {
80 __m128i quantCoef1, quantCoef2, deQuantCoef1, deQuantCoef2, quantCoef12, sign;
81
82 quantCoef12 = _mm_loadu_si128((__m128i*)(quantCoef + n));
83
84 deQuantCoef1 = _mm_loadu_si128((__m128i*)(deQuantCoef + n));
85 deQuantCoef2 = _mm_loadu_si128((__m128i*)(deQuantCoef + n + 4));
86
87 sign = _mm_srai_epi16(quantCoef12, 15);
88 quantCoef1 = _mm_unpacklo_epi16(quantCoef12, sign);
89 quantCoef2 = _mm_unpackhi_epi16(quantCoef12, sign);
90
91 quantCoef1 = _mm_mullo_epi32(quantCoef1, deQuantCoef1);
92 quantCoef2 = _mm_mullo_epi32(quantCoef2, deQuantCoef2);
93
94 quantCoef12 = _mm_packs_epi32(quantCoef1, quantCoef2);
95 sign = _mm_srai_epi16(quantCoef12, 15);
96 quantCoef1 = _mm_unpacklo_epi16(quantCoef12, sign);
97 quantCoef2 = _mm_unpackhi_epi16(quantCoef12, sign);
98
99 quantCoef1 = _mm_sll_epi32(quantCoef1, _mm_cvtsi32_si128(per - shift));
100 quantCoef2 = _mm_sll_epi32(quantCoef2, _mm_cvtsi32_si128(per - shift));
101
102 quantCoef12 = _mm_packs_epi32(quantCoef1, quantCoef2);
103 sign = _mm_srai_epi16(quantCoef12, 15);
104 quantCoef1 = _mm_unpacklo_epi16(quantCoef12, sign);
105 _mm_storeu_si128((__m128i*)(coef + n), quantCoef1);
106 quantCoef2 = _mm_unpackhi_epi16(quantCoef12, sign);
107 _mm_storeu_si128((__m128i*)(coef + n + 4), quantCoef2);
108 }
109 }
110 }
111 }
112
113 namespace x265 {
114 void Setup_Vec_DCTPrimitives_sse41(EncoderPrimitives &p)
115 {
116 p.dequant_scaling = dequant_scaling;
117 }
118 }