1/*
2 *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3 *
4 *  Use of this source code is governed by a BSD-style license
5 *  that can be found in the LICENSE file in the root of the source
6 *  tree. An additional intellectual property rights grant can be found
7 *  in the file PATENTS.  All contributing project authors may
8 *  be found in the AUTHORS file in the root of the source tree.
9 */
10
11#include "./vpx_config.h"
12#include "vp9/encoder/vp9_variance.h"
13#include "vp9/common/vp9_pragmas.h"
14#include "vpx_ports/mem.h"
15
16extern unsigned int vp9_get8x8var_mmx
17(
18  const unsigned char *src_ptr,
19  int  source_stride,
20  const unsigned char *ref_ptr,
21  int  recon_stride,
22  unsigned int *SSE,
23  int *Sum
24);
25extern unsigned int vp9_get4x4var_mmx
26(
27  const unsigned char *src_ptr,
28  int  source_stride,
29  const unsigned char *ref_ptr,
30  int  recon_stride,
31  unsigned int *SSE,
32  int *Sum
33);
34
35unsigned int vp9_variance4x4_mmx(
36  const unsigned char *src_ptr,
37  int  source_stride,
38  const unsigned char *ref_ptr,
39  int  recon_stride,
40  unsigned int *sse) {
41  unsigned int var;
42  int avg;
43
44  vp9_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg);
45  *sse = var;
46  return (var - (((unsigned int)avg * avg) >> 4));
47}
48
49unsigned int vp9_variance8x8_mmx(
50  const unsigned char *src_ptr,
51  int  source_stride,
52  const unsigned char *ref_ptr,
53  int  recon_stride,
54  unsigned int *sse) {
55  unsigned int var;
56  int avg;
57
58  vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg);
59  *sse = var;
60
61  return (var - (((unsigned int)avg * avg) >> 6));
62}
63
64unsigned int vp9_mse16x16_mmx(
65  const unsigned char *src_ptr,
66  int  source_stride,
67  const unsigned char *ref_ptr,
68  int  recon_stride,
69  unsigned int *sse) {
70  unsigned int sse0, sse1, sse2, sse3, var;
71  int sum0, sum1, sum2, sum3;
72
73
74  vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
75                    &sum0);
76  vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
77                    &sse1, &sum1);
78  vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
79                    ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2);
80  vp9_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride,
81                    ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
82
83  var = sse0 + sse1 + sse2 + sse3;
84  *sse = var;
85  return var;
86}
87
88
89unsigned int vp9_variance16x16_mmx(
90  const unsigned char *src_ptr,
91  int  source_stride,
92  const unsigned char *ref_ptr,
93  int  recon_stride,
94  unsigned int *sse) {
95  unsigned int sse0, sse1, sse2, sse3, var;
96  int sum0, sum1, sum2, sum3, avg;
97
98  vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
99                    &sum0);
100  vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
101                    &sse1, &sum1);
102  vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
103                    ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2);
104  vp9_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride,
105                    ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
106
107  var = sse0 + sse1 + sse2 + sse3;
108  avg = sum0 + sum1 + sum2 + sum3;
109  *sse = var;
110  return (var - (((unsigned int)avg * avg) >> 8));
111}
112
113unsigned int vp9_variance16x8_mmx(
114  const unsigned char *src_ptr,
115  int  source_stride,
116  const unsigned char *ref_ptr,
117  int  recon_stride,
118  unsigned int *sse) {
119  unsigned int sse0, sse1, var;
120  int sum0, sum1, avg;
121
122  vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
123                    &sum0);
124  vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
125                    &sse1, &sum1);
126
127  var = sse0 + sse1;
128  avg = sum0 + sum1;
129  *sse = var;
130  return (var - (((unsigned int)avg * avg) >> 7));
131}
132
133
134unsigned int vp9_variance8x16_mmx(
135  const unsigned char *src_ptr,
136  int  source_stride,
137  const unsigned char *ref_ptr,
138  int  recon_stride,
139  unsigned int *sse) {
140  unsigned int sse0, sse1, var;
141  int sum0, sum1, avg;
142
143  vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
144                    &sum0);
145  vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
146                    ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1);
147
148  var = sse0 + sse1;
149  avg = sum0 + sum1;
150  *sse = var;
151
152  return (var - (((unsigned int)avg * avg) >> 7));
153}
154