xref: /aosp_15_r20/external/libaom/av1/common/x86/reconinter_ssse3.c (revision 77c1e3ccc04c968bd2bc212e87364f250e820521)
1 /*
2  * Copyright (c) 2018, Alliance for Open Media. All rights reserved.
3  *
4  * This source code is subject to the terms of the BSD 2 Clause License and
5  * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6  * was not distributed with this source code in the LICENSE file, you can
7  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8  * Media Patent License 1.0 was not distributed with this source code in the
9  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10  */
11 
12 #include "config/av1_rtcd.h"
13 
14 #if CONFIG_AV1_HIGHBITDEPTH
15 
16 #include <tmmintrin.h>
17 
18 #include "aom/aom_integer.h"
19 #include "aom_dsp/blend.h"
20 #include "aom_dsp/x86/synonyms.h"
21 #include "av1/common/blockd.h"
22 
av1_build_compound_diffwtd_mask_highbd_ssse3(uint8_t * mask,DIFFWTD_MASK_TYPE mask_type,const uint8_t * src0,int src0_stride,const uint8_t * src1,int src1_stride,int h,int w,int bd)23 void av1_build_compound_diffwtd_mask_highbd_ssse3(
24     uint8_t *mask, DIFFWTD_MASK_TYPE mask_type, const uint8_t *src0,
25     int src0_stride, const uint8_t *src1, int src1_stride, int h, int w,
26     int bd) {
27   if (w < 8) {
28     av1_build_compound_diffwtd_mask_highbd_c(mask, mask_type, src0, src0_stride,
29                                              src1, src1_stride, h, w, bd);
30   } else {
31     assert(bd >= 8);
32     assert((w % 8) == 0);
33     assert(mask_type == DIFFWTD_38 || mask_type == DIFFWTD_38_INV);
34     const __m128i x0 = _mm_setzero_si128();
35     const __m128i xAOM_BLEND_A64_MAX_ALPHA =
36         _mm_set1_epi16(AOM_BLEND_A64_MAX_ALPHA);
37     const int mask_base = 38;
38     const __m128i xmask_base = _mm_set1_epi16(mask_base);
39     const uint16_t *ssrc0 = CONVERT_TO_SHORTPTR(src0);
40     const uint16_t *ssrc1 = CONVERT_TO_SHORTPTR(src1);
41     if (bd == 8) {
42       if (mask_type == DIFFWTD_38_INV) {
43         for (int i = 0; i < h; ++i) {
44           for (int j = 0; j < w; j += 8) {
45             __m128i s0 = _mm_loadu_si128((const __m128i *)&ssrc0[j]);
46             __m128i s1 = _mm_loadu_si128((const __m128i *)&ssrc1[j]);
47             __m128i diff = _mm_srai_epi16(_mm_abs_epi16(_mm_sub_epi16(s0, s1)),
48                                           DIFF_FACTOR_LOG2);
49             __m128i m = _mm_min_epi16(
50                 _mm_max_epi16(x0, _mm_add_epi16(diff, xmask_base)),
51                 xAOM_BLEND_A64_MAX_ALPHA);
52             m = _mm_sub_epi16(xAOM_BLEND_A64_MAX_ALPHA, m);
53             m = _mm_packus_epi16(m, m);
54             _mm_storel_epi64((__m128i *)&mask[j], m);
55           }
56           ssrc0 += src0_stride;
57           ssrc1 += src1_stride;
58           mask += w;
59         }
60       } else {
61         for (int i = 0; i < h; ++i) {
62           for (int j = 0; j < w; j += 8) {
63             __m128i s0 = _mm_loadu_si128((const __m128i *)&ssrc0[j]);
64             __m128i s1 = _mm_loadu_si128((const __m128i *)&ssrc1[j]);
65             __m128i diff = _mm_srai_epi16(_mm_abs_epi16(_mm_sub_epi16(s0, s1)),
66                                           DIFF_FACTOR_LOG2);
67             __m128i m = _mm_min_epi16(
68                 _mm_max_epi16(x0, _mm_add_epi16(diff, xmask_base)),
69                 xAOM_BLEND_A64_MAX_ALPHA);
70             m = _mm_packus_epi16(m, m);
71             _mm_storel_epi64((__m128i *)&mask[j], m);
72           }
73           ssrc0 += src0_stride;
74           ssrc1 += src1_stride;
75           mask += w;
76         }
77       }
78     } else {
79       const __m128i xshift = _mm_set1_epi64x(bd - 8 + DIFF_FACTOR_LOG2);
80       if (mask_type == DIFFWTD_38_INV) {
81         for (int i = 0; i < h; ++i) {
82           for (int j = 0; j < w; j += 8) {
83             __m128i s0 = _mm_loadu_si128((const __m128i *)&ssrc0[j]);
84             __m128i s1 = _mm_loadu_si128((const __m128i *)&ssrc1[j]);
85             __m128i diff =
86                 _mm_sra_epi16(_mm_abs_epi16(_mm_sub_epi16(s0, s1)), xshift);
87             __m128i m = _mm_min_epi16(
88                 _mm_max_epi16(x0, _mm_add_epi16(diff, xmask_base)),
89                 xAOM_BLEND_A64_MAX_ALPHA);
90             m = _mm_sub_epi16(xAOM_BLEND_A64_MAX_ALPHA, m);
91             m = _mm_packus_epi16(m, m);
92             _mm_storel_epi64((__m128i *)&mask[j], m);
93           }
94           ssrc0 += src0_stride;
95           ssrc1 += src1_stride;
96           mask += w;
97         }
98       } else {
99         for (int i = 0; i < h; ++i) {
100           for (int j = 0; j < w; j += 8) {
101             __m128i s0 = _mm_loadu_si128((const __m128i *)&ssrc0[j]);
102             __m128i s1 = _mm_loadu_si128((const __m128i *)&ssrc1[j]);
103             __m128i diff =
104                 _mm_sra_epi16(_mm_abs_epi16(_mm_sub_epi16(s0, s1)), xshift);
105             __m128i m = _mm_min_epi16(
106                 _mm_max_epi16(x0, _mm_add_epi16(diff, xmask_base)),
107                 xAOM_BLEND_A64_MAX_ALPHA);
108             m = _mm_packus_epi16(m, m);
109             _mm_storel_epi64((__m128i *)&mask[j], m);
110           }
111           ssrc0 += src0_stride;
112           ssrc1 += src1_stride;
113           mask += w;
114         }
115       }
116     }
117   }
118 }
119 
120 #endif  // CONFIG_AV1_HIGHBITDEPTH
121