Add non420 code in multi-threaded loopfilter
[aom.git] / vp9 / encoder / vp9_extend.c
blobc9b2131426f851d6b2a521aa55d98bc17cd7c1e2
1 /*
2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
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 */
11 #include "vpx_mem/vpx_mem.h"
13 #include "vp9/common/vp9_common.h"
14 #include "vp9/encoder/vp9_extend.h"
16 static void copy_and_extend_plane(const uint8_t *src, int src_pitch,
17 uint8_t *dst, int dst_pitch,
18 int w, int h,
19 int extend_top, int extend_left,
20 int extend_bottom, int extend_right) {
21 int i, linesize;
23 // copy the left and right most columns out
24 const uint8_t *src_ptr1 = src;
25 const uint8_t *src_ptr2 = src + w - 1;
26 uint8_t *dst_ptr1 = dst - extend_left;
27 uint8_t *dst_ptr2 = dst + w;
29 for (i = 0; i < h; i++) {
30 vpx_memset(dst_ptr1, src_ptr1[0], extend_left);
31 vpx_memcpy(dst_ptr1 + extend_left, src_ptr1, w);
32 vpx_memset(dst_ptr2, src_ptr2[0], extend_right);
33 src_ptr1 += src_pitch;
34 src_ptr2 += src_pitch;
35 dst_ptr1 += dst_pitch;
36 dst_ptr2 += dst_pitch;
39 // Now copy the top and bottom lines into each line of the respective
40 // borders
41 src_ptr1 = dst - extend_left;
42 src_ptr2 = dst + dst_pitch * (h - 1) - extend_left;
43 dst_ptr1 = dst + dst_pitch * (-extend_top) - extend_left;
44 dst_ptr2 = dst + dst_pitch * (h) - extend_left;
45 linesize = extend_left + extend_right + w;
47 for (i = 0; i < extend_top; i++) {
48 vpx_memcpy(dst_ptr1, src_ptr1, linesize);
49 dst_ptr1 += dst_pitch;
52 for (i = 0; i < extend_bottom; i++) {
53 vpx_memcpy(dst_ptr2, src_ptr2, linesize);
54 dst_ptr2 += dst_pitch;
58 #if CONFIG_VP9_HIGHBITDEPTH
59 static void highbd_copy_and_extend_plane(const uint8_t *src8, int src_pitch,
60 uint8_t *dst8, int dst_pitch,
61 int w, int h,
62 int extend_top, int extend_left,
63 int extend_bottom, int extend_right) {
64 int i, linesize;
65 uint16_t *src = CONVERT_TO_SHORTPTR(src8);
66 uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
68 // copy the left and right most columns out
69 const uint16_t *src_ptr1 = src;
70 const uint16_t *src_ptr2 = src + w - 1;
71 uint16_t *dst_ptr1 = dst - extend_left;
72 uint16_t *dst_ptr2 = dst + w;
74 for (i = 0; i < h; i++) {
75 vpx_memset16(dst_ptr1, src_ptr1[0], extend_left);
76 vpx_memcpy(dst_ptr1 + extend_left, src_ptr1, w * sizeof(uint16_t));
77 vpx_memset16(dst_ptr2, src_ptr2[0], extend_right);
78 src_ptr1 += src_pitch;
79 src_ptr2 += src_pitch;
80 dst_ptr1 += dst_pitch;
81 dst_ptr2 += dst_pitch;
84 // Now copy the top and bottom lines into each line of the respective
85 // borders
86 src_ptr1 = dst - extend_left;
87 src_ptr2 = dst + dst_pitch * (h - 1) - extend_left;
88 dst_ptr1 = dst + dst_pitch * (-extend_top) - extend_left;
89 dst_ptr2 = dst + dst_pitch * (h) - extend_left;
90 linesize = extend_left + extend_right + w;
92 for (i = 0; i < extend_top; i++) {
93 vpx_memcpy(dst_ptr1, src_ptr1, linesize * sizeof(uint16_t));
94 dst_ptr1 += dst_pitch;
97 for (i = 0; i < extend_bottom; i++) {
98 vpx_memcpy(dst_ptr2, src_ptr2, linesize * sizeof(uint16_t));
99 dst_ptr2 += dst_pitch;
102 #endif // CONFIG_VP9_HIGHBITDEPTH
104 void vp9_copy_and_extend_frame(const YV12_BUFFER_CONFIG *src,
105 YV12_BUFFER_CONFIG *dst) {
106 // Extend src frame in buffer
107 // Altref filtering assumes 16 pixel extension
108 const int et_y = 16;
109 const int el_y = 16;
110 // Motion estimation may use src block variance with the block size up
111 // to 64x64, so the right and bottom need to be extended to 64 multiple
112 // or up to 16, whichever is greater.
113 const int eb_y = MAX(src->y_width + 16, ALIGN_POWER_OF_TWO(src->y_width, 6))
114 - src->y_crop_width;
115 const int er_y = MAX(src->y_height + 16, ALIGN_POWER_OF_TWO(src->y_height, 6))
116 - src->y_crop_height;
117 const int uv_width_subsampling = (src->uv_width != src->y_width);
118 const int uv_height_subsampling = (src->uv_height != src->y_height);
119 const int et_uv = et_y >> uv_height_subsampling;
120 const int el_uv = el_y >> uv_width_subsampling;
121 const int eb_uv = eb_y >> uv_height_subsampling;
122 const int er_uv = er_y >> uv_width_subsampling;
124 #if CONFIG_VP9_HIGHBITDEPTH
125 if (src->flags & YV12_FLAG_HIGHBITDEPTH) {
126 highbd_copy_and_extend_plane(src->y_buffer, src->y_stride,
127 dst->y_buffer, dst->y_stride,
128 src->y_crop_width, src->y_crop_height,
129 et_y, el_y, eb_y, er_y);
131 highbd_copy_and_extend_plane(src->u_buffer, src->uv_stride,
132 dst->u_buffer, dst->uv_stride,
133 src->uv_crop_width, src->uv_crop_height,
134 et_uv, el_uv, eb_uv, er_uv);
136 highbd_copy_and_extend_plane(src->v_buffer, src->uv_stride,
137 dst->v_buffer, dst->uv_stride,
138 src->uv_crop_width, src->uv_crop_height,
139 et_uv, el_uv, eb_uv, er_uv);
140 return;
142 #endif // CONFIG_VP9_HIGHBITDEPTH
144 copy_and_extend_plane(src->y_buffer, src->y_stride,
145 dst->y_buffer, dst->y_stride,
146 src->y_crop_width, src->y_crop_height,
147 et_y, el_y, eb_y, er_y);
149 copy_and_extend_plane(src->u_buffer, src->uv_stride,
150 dst->u_buffer, dst->uv_stride,
151 src->uv_crop_width, src->uv_crop_height,
152 et_uv, el_uv, eb_uv, er_uv);
154 copy_and_extend_plane(src->v_buffer, src->uv_stride,
155 dst->v_buffer, dst->uv_stride,
156 src->uv_crop_width, src->uv_crop_height,
157 et_uv, el_uv, eb_uv, er_uv);
160 void vp9_copy_and_extend_frame_with_rect(const YV12_BUFFER_CONFIG *src,
161 YV12_BUFFER_CONFIG *dst,
162 int srcy, int srcx,
163 int srch, int srcw) {
164 // If the side is not touching the bounder then don't extend.
165 const int et_y = srcy ? 0 : dst->border;
166 const int el_y = srcx ? 0 : dst->border;
167 const int eb_y = srcy + srch != src->y_height ? 0 :
168 dst->border + dst->y_height - src->y_height;
169 const int er_y = srcx + srcw != src->y_width ? 0 :
170 dst->border + dst->y_width - src->y_width;
171 const int src_y_offset = srcy * src->y_stride + srcx;
172 const int dst_y_offset = srcy * dst->y_stride + srcx;
174 const int et_uv = ROUND_POWER_OF_TWO(et_y, 1);
175 const int el_uv = ROUND_POWER_OF_TWO(el_y, 1);
176 const int eb_uv = ROUND_POWER_OF_TWO(eb_y, 1);
177 const int er_uv = ROUND_POWER_OF_TWO(er_y, 1);
178 const int src_uv_offset = ((srcy * src->uv_stride) >> 1) + (srcx >> 1);
179 const int dst_uv_offset = ((srcy * dst->uv_stride) >> 1) + (srcx >> 1);
180 const int srch_uv = ROUND_POWER_OF_TWO(srch, 1);
181 const int srcw_uv = ROUND_POWER_OF_TWO(srcw, 1);
183 copy_and_extend_plane(src->y_buffer + src_y_offset, src->y_stride,
184 dst->y_buffer + dst_y_offset, dst->y_stride,
185 srcw, srch,
186 et_y, el_y, eb_y, er_y);
188 copy_and_extend_plane(src->u_buffer + src_uv_offset, src->uv_stride,
189 dst->u_buffer + dst_uv_offset, dst->uv_stride,
190 srcw_uv, srch_uv,
191 et_uv, el_uv, eb_uv, er_uv);
193 copy_and_extend_plane(src->v_buffer + src_uv_offset, src->uv_stride,
194 dst->v_buffer + dst_uv_offset, dst->uv_stride,
195 srcw_uv, srch_uv,
196 et_uv, el_uv, eb_uv, er_uv);