/work/dav1d/src/ipred_prepare_tmpl.c
Line | Count | Source |
1 | | /* |
2 | | * Copyright © 2018, VideoLAN and dav1d authors |
3 | | * Copyright © 2018, Two Orioles, LLC |
4 | | * All rights reserved. |
5 | | * |
6 | | * Redistribution and use in source and binary forms, with or without |
7 | | * modification, are permitted provided that the following conditions are met: |
8 | | * |
9 | | * 1. Redistributions of source code must retain the above copyright notice, this |
10 | | * list of conditions and the following disclaimer. |
11 | | * |
12 | | * 2. Redistributions in binary form must reproduce the above copyright notice, |
13 | | * this list of conditions and the following disclaimer in the documentation |
14 | | * and/or other materials provided with the distribution. |
15 | | * |
16 | | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND |
17 | | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED |
18 | | * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE |
19 | | * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR |
20 | | * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES |
21 | | * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
22 | | * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND |
23 | | * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
24 | | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS |
25 | | * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
26 | | */ |
27 | | |
28 | | #include "config.h" |
29 | | |
30 | | #include <stdint.h> |
31 | | #include <string.h> |
32 | | |
33 | | #include "common/intops.h" |
34 | | |
35 | | #include "src/ipred_prepare.h" |
36 | | |
37 | | static const uint8_t av1_mode_conv[N_INTRA_PRED_MODES] |
38 | | [2 /* have_left */][2 /* have_top */] = |
39 | | { |
40 | | [DC_PRED] = { { DC_128_PRED, TOP_DC_PRED }, |
41 | | { LEFT_DC_PRED, DC_PRED } }, |
42 | | [PAETH_PRED] = { { DC_128_PRED, VERT_PRED }, |
43 | | { HOR_PRED, PAETH_PRED } }, |
44 | | }; |
45 | | |
46 | | static const uint8_t av1_mode_to_angle_map[8] = { |
47 | | 90, 180, 45, 135, 113, 157, 203, 67 |
48 | | }; |
49 | | |
50 | | static const struct { |
51 | | uint8_t needs_left:1; |
52 | | uint8_t needs_top:1; |
53 | | uint8_t needs_topleft:1; |
54 | | uint8_t needs_topright:1; |
55 | | uint8_t needs_bottomleft:1; |
56 | | } av1_intra_prediction_edges[N_IMPL_INTRA_PRED_MODES] = { |
57 | | [DC_PRED] = { .needs_top = 1, .needs_left = 1 }, |
58 | | [VERT_PRED] = { .needs_top = 1 }, |
59 | | [HOR_PRED] = { .needs_left = 1 }, |
60 | | [LEFT_DC_PRED] = { .needs_left = 1 }, |
61 | | [TOP_DC_PRED] = { .needs_top = 1 }, |
62 | | [DC_128_PRED] = { 0 }, |
63 | | [Z1_PRED] = { .needs_top = 1, .needs_topright = 1, |
64 | | .needs_topleft = 1 }, |
65 | | [Z2_PRED] = { .needs_left = 1, .needs_top = 1, .needs_topleft = 1 }, |
66 | | [Z3_PRED] = { .needs_left = 1, .needs_bottomleft = 1, |
67 | | .needs_topleft = 1 }, |
68 | | [SMOOTH_PRED] = { .needs_left = 1, .needs_top = 1 }, |
69 | | [SMOOTH_V_PRED] = { .needs_left = 1, .needs_top = 1 }, |
70 | | [SMOOTH_H_PRED] = { .needs_left = 1, .needs_top = 1 }, |
71 | | [PAETH_PRED] = { .needs_left = 1, .needs_top = 1, .needs_topleft = 1 }, |
72 | | [FILTER_PRED] = { .needs_left = 1, .needs_top = 1, .needs_topleft = 1 }, |
73 | | }; |
74 | | |
75 | | enum IntraPredMode |
76 | | bytefn(dav1d_prepare_intra_edges)(const int x, const int have_left, |
77 | | const int y, const int have_top, |
78 | | const int w, const int h, |
79 | | const enum EdgeFlags edge_flags, |
80 | | const pixel *const dst, |
81 | | const ptrdiff_t stride, |
82 | | const pixel *prefilter_toplevel_sb_edge, |
83 | | enum IntraPredMode mode, int *const angle, |
84 | | const int tw, const int th, const int filter_edge, |
85 | | pixel *const topleft_out HIGHBD_DECL_SUFFIX) |
86 | 5.71k | { |
87 | 5.71k | const int bitdepth = bitdepth_from_max(bitdepth_max); |
88 | 5.71k | assert(y < h && x < w); |
89 | | |
90 | 5.71k | switch (mode) { |
91 | 342 | case VERT_PRED: |
92 | 928 | case HOR_PRED: |
93 | 1.13k | case DIAG_DOWN_LEFT_PRED: |
94 | 1.24k | case DIAG_DOWN_RIGHT_PRED: |
95 | 1.40k | case VERT_RIGHT_PRED: |
96 | 1.60k | case HOR_DOWN_PRED: |
97 | 1.92k | case HOR_UP_PRED: |
98 | 2.14k | case VERT_LEFT_PRED: { |
99 | 2.14k | *angle = av1_mode_to_angle_map[mode - VERT_PRED] + 3 * *angle; |
100 | | |
101 | 2.14k | if (*angle <= 90) |
102 | 698 | mode = *angle < 90 && have_top ? Z1_PRED : VERT_PRED; |
103 | 1.44k | else if (*angle < 180) |
104 | 702 | mode = Z2_PRED; |
105 | 740 | else |
106 | 740 | mode = *angle > 180 && have_left ? Z3_PRED : HOR_PRED; |
107 | 2.14k | break; |
108 | 1.92k | } |
109 | 1.49k | case DC_PRED: |
110 | 1.74k | case PAETH_PRED: |
111 | 1.74k | mode = av1_mode_conv[mode][have_left][have_top]; |
112 | 1.74k | break; |
113 | 1.83k | default: |
114 | 1.83k | break; |
115 | 5.71k | } |
116 | | |
117 | 5.71k | const pixel *dst_top; |
118 | 5.71k | if (have_top && |
119 | 5.27k | (av1_intra_prediction_edges[mode].needs_top || |
120 | 676 | av1_intra_prediction_edges[mode].needs_topleft || |
121 | 288 | (av1_intra_prediction_edges[mode].needs_left && !have_left))) |
122 | 4.99k | { |
123 | 4.99k | if (prefilter_toplevel_sb_edge) { |
124 | 552 | dst_top = &prefilter_toplevel_sb_edge[x * 4]; |
125 | 4.44k | } else { |
126 | 4.44k | dst_top = &dst[-PXSTRIDE(stride)]; |
127 | 4.44k | } |
128 | 4.99k | } |
129 | | |
130 | 5.71k | if (av1_intra_prediction_edges[mode].needs_left) { |
131 | 5.00k | const int sz = th << 2; |
132 | 5.00k | pixel *const left = &topleft_out[-sz]; |
133 | | |
134 | 5.00k | if (have_left) { |
135 | 4.99k | const int px_have = imin(sz, (h - y) << 2); |
136 | | |
137 | 52.1k | for (int i = 0; i < px_have; i++) |
138 | 47.1k | left[sz - 1 - i] = dst[PXSTRIDE(stride) * i - 1]; |
139 | 4.99k | if (px_have < sz) |
140 | 0 | pixel_set(left, left[sz - px_have], sz - px_have); |
141 | 4.99k | } else { |
142 | 10 | pixel_set(left, have_top ? *dst_top : ((1 << bitdepth) >> 1) + 1, sz); |
143 | 10 | } |
144 | | |
145 | 5.00k | if (av1_intra_prediction_edges[mode].needs_bottomleft) { |
146 | 434 | const int have_bottomleft = (!have_left || y + th >= h) ? 0 : |
147 | 434 | (edge_flags & EDGE_I444_LEFT_HAS_BOTTOM); |
148 | | |
149 | 434 | if (have_bottomleft) { |
150 | 164 | const int px_have = imin(sz, (h - y - th) << 2); |
151 | | |
152 | 1.33k | for (int i = 0; i < px_have; i++) |
153 | 1.16k | left[-(i + 1)] = dst[(sz + i) * PXSTRIDE(stride) - 1]; |
154 | 164 | if (px_have < sz) |
155 | 0 | pixel_set(left - sz, left[-px_have], sz - px_have); |
156 | 270 | } else { |
157 | 270 | pixel_set(left - sz, left[0], sz); |
158 | 270 | } |
159 | 434 | } |
160 | 5.00k | } |
161 | | |
162 | 5.71k | if (av1_intra_prediction_edges[mode].needs_top) { |
163 | 4.81k | const int sz = tw << 2; |
164 | 4.81k | pixel *const top = &topleft_out[1]; |
165 | | |
166 | 4.81k | if (have_top) { |
167 | 4.60k | const int px_have = imin(sz, (w - x) << 2); |
168 | 4.60k | pixel_copy(top, dst_top, px_have); |
169 | 4.60k | if (px_have < sz) |
170 | 0 | pixel_set(top + px_have, top[px_have - 1], sz - px_have); |
171 | 4.60k | } else { |
172 | 212 | pixel_set(top, have_left ? dst[-1] : ((1 << bitdepth) >> 1) - 1, sz); |
173 | 212 | } |
174 | | |
175 | 4.81k | if (av1_intra_prediction_edges[mode].needs_topright) { |
176 | 504 | const int have_topright = (!have_top || x + tw >= w) ? 0 : |
177 | 504 | (edge_flags & EDGE_I444_TOP_HAS_RIGHT); |
178 | | |
179 | 504 | if (have_topright) { |
180 | 326 | const int px_have = imin(sz, (w - x - tw) << 2); |
181 | | |
182 | 326 | pixel_copy(top + sz, &dst_top[sz], px_have); |
183 | 326 | if (px_have < sz) |
184 | 0 | pixel_set(top + sz + px_have, top[sz + px_have - 1], |
185 | 0 | sz - px_have); |
186 | 326 | } else { |
187 | 178 | pixel_set(top + sz, top[sz - 1], sz); |
188 | 178 | } |
189 | 504 | } |
190 | 4.81k | } |
191 | | |
192 | 5.71k | if (av1_intra_prediction_edges[mode].needs_topleft) { |
193 | 2.54k | if (have_left) |
194 | 2.54k | *topleft_out = have_top ? dst_top[-1] : dst[-1]; |
195 | 4 | else |
196 | 4 | *topleft_out = have_top ? *dst_top : (1 << bitdepth) >> 1; |
197 | | |
198 | 2.54k | if (mode == Z2_PRED && tw + th >= 6 && filter_edge) |
199 | 150 | *topleft_out = ((topleft_out[-1] + topleft_out[1]) * 5 + |
200 | 150 | topleft_out[0] * 6 + 8) >> 4; |
201 | 2.54k | } |
202 | | |
203 | 5.71k | return mode; |
204 | 5.71k | } dav1d_prepare_intra_edges_8bpc Line | Count | Source | 86 | 5.71k | { | 87 | 5.71k | const int bitdepth = bitdepth_from_max(bitdepth_max); | 88 | 5.71k | assert(y < h && x < w); | 89 | | | 90 | 5.71k | switch (mode) { | 91 | 342 | case VERT_PRED: | 92 | 928 | case HOR_PRED: | 93 | 1.13k | case DIAG_DOWN_LEFT_PRED: | 94 | 1.24k | case DIAG_DOWN_RIGHT_PRED: | 95 | 1.40k | case VERT_RIGHT_PRED: | 96 | 1.60k | case HOR_DOWN_PRED: | 97 | 1.92k | case HOR_UP_PRED: | 98 | 2.14k | case VERT_LEFT_PRED: { | 99 | 2.14k | *angle = av1_mode_to_angle_map[mode - VERT_PRED] + 3 * *angle; | 100 | | | 101 | 2.14k | if (*angle <= 90) | 102 | 698 | mode = *angle < 90 && have_top ? Z1_PRED : VERT_PRED; | 103 | 1.44k | else if (*angle < 180) | 104 | 702 | mode = Z2_PRED; | 105 | 740 | else | 106 | 740 | mode = *angle > 180 && have_left ? Z3_PRED : HOR_PRED; | 107 | 2.14k | break; | 108 | 1.92k | } | 109 | 1.49k | case DC_PRED: | 110 | 1.74k | case PAETH_PRED: | 111 | 1.74k | mode = av1_mode_conv[mode][have_left][have_top]; | 112 | 1.74k | break; | 113 | 1.83k | default: | 114 | 1.83k | break; | 115 | 5.71k | } | 116 | | | 117 | 5.71k | const pixel *dst_top; | 118 | 5.71k | if (have_top && | 119 | 5.27k | (av1_intra_prediction_edges[mode].needs_top || | 120 | 676 | av1_intra_prediction_edges[mode].needs_topleft || | 121 | 288 | (av1_intra_prediction_edges[mode].needs_left && !have_left))) | 122 | 4.99k | { | 123 | 4.99k | if (prefilter_toplevel_sb_edge) { | 124 | 552 | dst_top = &prefilter_toplevel_sb_edge[x * 4]; | 125 | 4.44k | } else { | 126 | 4.44k | dst_top = &dst[-PXSTRIDE(stride)]; | 127 | 4.44k | } | 128 | 4.99k | } | 129 | | | 130 | 5.71k | if (av1_intra_prediction_edges[mode].needs_left) { | 131 | 5.00k | const int sz = th << 2; | 132 | 5.00k | pixel *const left = &topleft_out[-sz]; | 133 | | | 134 | 5.00k | if (have_left) { | 135 | 4.99k | const int px_have = imin(sz, (h - y) << 2); | 136 | | | 137 | 52.1k | for (int i = 0; i < px_have; i++) | 138 | 47.1k | left[sz - 1 - i] = dst[PXSTRIDE(stride) * i - 1]; | 139 | 4.99k | if (px_have < sz) | 140 | 0 | pixel_set(left, left[sz - px_have], sz - px_have); | 141 | 4.99k | } else { | 142 | 10 | pixel_set(left, have_top ? *dst_top : ((1 << bitdepth) >> 1) + 1, sz); | 143 | 10 | } | 144 | | | 145 | 5.00k | if (av1_intra_prediction_edges[mode].needs_bottomleft) { | 146 | 434 | const int have_bottomleft = (!have_left || y + th >= h) ? 0 : | 147 | 434 | (edge_flags & EDGE_I444_LEFT_HAS_BOTTOM); | 148 | | | 149 | 434 | if (have_bottomleft) { | 150 | 164 | const int px_have = imin(sz, (h - y - th) << 2); | 151 | | | 152 | 1.33k | for (int i = 0; i < px_have; i++) | 153 | 1.16k | left[-(i + 1)] = dst[(sz + i) * PXSTRIDE(stride) - 1]; | 154 | 164 | if (px_have < sz) | 155 | 0 | pixel_set(left - sz, left[-px_have], sz - px_have); | 156 | 270 | } else { | 157 | 270 | pixel_set(left - sz, left[0], sz); | 158 | 270 | } | 159 | 434 | } | 160 | 5.00k | } | 161 | | | 162 | 5.71k | if (av1_intra_prediction_edges[mode].needs_top) { | 163 | 4.81k | const int sz = tw << 2; | 164 | 4.81k | pixel *const top = &topleft_out[1]; | 165 | | | 166 | 4.81k | if (have_top) { | 167 | 4.60k | const int px_have = imin(sz, (w - x) << 2); | 168 | 4.60k | pixel_copy(top, dst_top, px_have); | 169 | 4.60k | if (px_have < sz) | 170 | 0 | pixel_set(top + px_have, top[px_have - 1], sz - px_have); | 171 | 4.60k | } else { | 172 | 212 | pixel_set(top, have_left ? dst[-1] : ((1 << bitdepth) >> 1) - 1, sz); | 173 | 212 | } | 174 | | | 175 | 4.81k | if (av1_intra_prediction_edges[mode].needs_topright) { | 176 | 504 | const int have_topright = (!have_top || x + tw >= w) ? 0 : | 177 | 504 | (edge_flags & EDGE_I444_TOP_HAS_RIGHT); | 178 | | | 179 | 504 | if (have_topright) { | 180 | 326 | const int px_have = imin(sz, (w - x - tw) << 2); | 181 | | | 182 | 326 | pixel_copy(top + sz, &dst_top[sz], px_have); | 183 | 326 | if (px_have < sz) | 184 | 0 | pixel_set(top + sz + px_have, top[sz + px_have - 1], | 185 | 0 | sz - px_have); | 186 | 326 | } else { | 187 | 178 | pixel_set(top + sz, top[sz - 1], sz); | 188 | 178 | } | 189 | 504 | } | 190 | 4.81k | } | 191 | | | 192 | 5.71k | if (av1_intra_prediction_edges[mode].needs_topleft) { | 193 | 2.54k | if (have_left) | 194 | 2.54k | *topleft_out = have_top ? dst_top[-1] : dst[-1]; | 195 | 4 | else | 196 | 4 | *topleft_out = have_top ? *dst_top : (1 << bitdepth) >> 1; | 197 | | | 198 | 2.54k | if (mode == Z2_PRED && tw + th >= 6 && filter_edge) | 199 | 150 | *topleft_out = ((topleft_out[-1] + topleft_out[1]) * 5 + | 200 | 150 | topleft_out[0] * 6 + 8) >> 4; | 201 | 2.54k | } | 202 | | | 203 | 5.71k | return mode; | 204 | 5.71k | } |
Unexecuted instantiation: dav1d_prepare_intra_edges_16bpc |