xref: /aosp_15_r20/external/libaom/aom_dsp/x86/avg_intrin_sse4.c (revision 77c1e3ccc04c968bd2bc212e87364f250e820521)
1 /*
2  * Copyright (c) 2022, 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 <smmintrin.h>
13 
14 #include "config/aom_dsp_rtcd.h"
15 
16 // ref: [0 - 510]
17 // src: [0 - 510]
18 // bwl: {2, 3, 4, 5}
aom_vector_var_sse4_1(const int16_t * ref,const int16_t * src,int bwl)19 int aom_vector_var_sse4_1(const int16_t *ref, const int16_t *src, int bwl) {
20   const int width = 4 << bwl;
21   assert(width % 16 == 0);
22 
23   const __m128i k_one_epi16 = _mm_set1_epi16((int16_t)1);
24   __m128i mean = _mm_setzero_si128();
25   __m128i sse = _mm_setzero_si128();
26 
27   for (int i = 0; i < width; i += 16) {
28     const __m128i src_line = _mm_loadu_si128((const __m128i *)src);
29     const __m128i ref_line = _mm_loadu_si128((const __m128i *)ref);
30     const __m128i src_line2 = _mm_loadu_si128((const __m128i *)(src + 8));
31     const __m128i ref_line2 = _mm_loadu_si128((const __m128i *)(ref + 8));
32     __m128i diff = _mm_sub_epi16(ref_line, src_line);
33     const __m128i diff2 = _mm_sub_epi16(ref_line2, src_line2);
34     __m128i diff_sqr = _mm_madd_epi16(diff, diff);
35     const __m128i diff_sqr2 = _mm_madd_epi16(diff2, diff2);
36 
37     diff = _mm_add_epi16(diff, diff2);
38     diff_sqr = _mm_add_epi32(diff_sqr, diff_sqr2);
39     sse = _mm_add_epi32(sse, diff_sqr);
40     mean = _mm_add_epi16(mean, diff);
41 
42     src += 16;
43     ref += 16;
44   }
45 
46   // m0 m1 m2 m3
47   mean = _mm_madd_epi16(mean, k_one_epi16);
48   // m0+m1 m2+m3 s0+s1 s2+s3
49   __m128i result = _mm_hadd_epi32(mean, sse);
50   // m0+m1+m2+m3 s0+s1+s2+s3 x x
51   result = _mm_add_epi32(result, _mm_bsrli_si128(result, 4));
52 
53   // (mean * mean): dynamic range 31 bits.
54   const int mean_int = _mm_extract_epi32(result, 0);
55   const int sse_int = _mm_extract_epi32(result, 2);
56   const unsigned int mean_abs = abs(mean_int);
57   const int var = sse_int - ((mean_abs * mean_abs) >> (bwl + 2));
58   return var;
59 }
60