FFmpeg coverage


Directory: ../../../ffmpeg/
File: src/libavcodec/vp9lpf.c
Date: 2024-03-29 01:21:52
Exec Total Coverage
Lines: 112 112 100.0%
Functions: 3 3 100.0%
Branches: 60 60 100.0%

Line Branch Exec Source
1 /*
2 * VP9 compatible video decoder
3 *
4 * Copyright (C) 2013 Ronald S. Bultje <rsbultje gmail com>
5 * Copyright (C) 2013 Clément Bœsch <u pkh me>
6 *
7 * This file is part of FFmpeg.
8 *
9 * FFmpeg is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
13 *
14 * FFmpeg is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
18 *
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with FFmpeg; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 */
23
24 #include "vp9dec.h"
25
26 149298 static av_always_inline void filter_plane_cols(VP9Context *s, int col, int ss_h, int ss_v,
27 uint8_t *lvl, uint8_t (*mask)[4],
28 uint8_t *dst, ptrdiff_t ls)
29 {
30 149298 int y, x, bytesperpixel = s->bytesperpixel;
31
32 // filter edges between columns (e.g. block1 | block2)
33
2/2
✓ Branch 0 taken 401440 times.
✓ Branch 1 taken 149298 times.
550738 for (y = 0; y < 8; y += 2 << ss_v, dst += 16 * ls, lvl += 16 << ss_v) {
34 401440 uint8_t *ptr = dst, *l = lvl, *hmask1 = mask[y], *hmask2 = mask[y + 1 + ss_v];
35 401440 unsigned hm1 = hmask1[0] | hmask1[1] | hmask1[2], hm13 = hmask1[3];
36 401440 unsigned hm2 = hmask2[1] | hmask2[2], hm23 = hmask2[3];
37 401440 unsigned hm = hm1 | hm2 | hm13 | hm23;
38
39
2/2
✓ Branch 0 taken 1790348 times.
✓ Branch 1 taken 401440 times.
2191788 for (x = 1; hm & ~(x - 1); x <<= 1, ptr += 8 * bytesperpixel >> ss_h) {
40
4/4
✓ Branch 0 taken 287136 times.
✓ Branch 1 taken 1503212 times.
✓ Branch 2 taken 233721 times.
✓ Branch 3 taken 53415 times.
1790348 if (col || x > 1) {
41
2/2
✓ Branch 0 taken 1177739 times.
✓ Branch 1 taken 559194 times.
1736933 if (hm1 & x) {
42 1177739 int L = *l, H = L >> 4;
43 1177739 int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
44
45
2/2
✓ Branch 0 taken 345292 times.
✓ Branch 1 taken 832447 times.
1177739 if (hmask1[0] & x) {
46
2/2
✓ Branch 0 taken 343504 times.
✓ Branch 1 taken 1788 times.
345292 if (hmask2[0] & x) {
47 av_assert2(l[8 << ss_v] == L);
48 343504 s->dsp.loop_filter_16[0](ptr, ls, E, I, H);
49 } else {
50 1788 s->dsp.loop_filter_8[2][0](ptr, ls, E, I, H);
51 }
52
2/2
✓ Branch 0 taken 766320 times.
✓ Branch 1 taken 66127 times.
832447 } else if (hm2 & x) {
53 766320 L = l[8 << ss_v];
54 766320 H |= (L >> 4) << 8;
55 766320 E |= s->filter_lut.mblim_lut[L] << 8;
56 766320 I |= s->filter_lut.lim_lut[L] << 8;
57 766320 s->dsp.loop_filter_mix2[!!(hmask1[1] & x)]
58 766320 [!!(hmask2[1] & x)]
59 766320 [0](ptr, ls, E, I, H);
60 } else {
61 66127 s->dsp.loop_filter_8[!!(hmask1[1] & x)]
62 66127 [0](ptr, ls, E, I, H);
63 }
64
2/2
✓ Branch 0 taken 33532 times.
✓ Branch 1 taken 525662 times.
559194 } else if (hm2 & x) {
65 33532 int L = l[8 << ss_v], H = L >> 4;
66 33532 int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
67
68 33532 s->dsp.loop_filter_8[!!(hmask2[1] & x)]
69 33532 [0](ptr + 8 * ls, ls, E, I, H);
70 }
71 }
72
2/2
✓ Branch 0 taken 867642 times.
✓ Branch 1 taken 922706 times.
1790348 if (ss_h) {
73
2/2
✓ Branch 0 taken 362726 times.
✓ Branch 1 taken 504916 times.
867642 if (x & 0xAA)
74 362726 l += 2;
75 } else {
76
2/2
✓ Branch 0 taken 115223 times.
✓ Branch 1 taken 807483 times.
922706 if (hm13 & x) {
77 115223 int L = *l, H = L >> 4;
78 115223 int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
79
80
2/2
✓ Branch 0 taken 66821 times.
✓ Branch 1 taken 48402 times.
115223 if (hm23 & x) {
81 66821 L = l[8 << ss_v];
82 66821 H |= (L >> 4) << 8;
83 66821 E |= s->filter_lut.mblim_lut[L] << 8;
84 66821 I |= s->filter_lut.lim_lut[L] << 8;
85 66821 s->dsp.loop_filter_mix2[0][0][0](ptr + 4 * bytesperpixel, ls, E, I, H);
86 } else {
87 48402 s->dsp.loop_filter_8[0][0](ptr + 4 * bytesperpixel, ls, E, I, H);
88 }
89
2/2
✓ Branch 0 taken 39533 times.
✓ Branch 1 taken 767950 times.
807483 } else if (hm23 & x) {
90 39533 int L = l[8 << ss_v], H = L >> 4;
91 39533 int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
92
93 39533 s->dsp.loop_filter_8[0][0](ptr + 8 * ls + 4 * bytesperpixel, ls, E, I, H);
94 }
95 922706 l++;
96 }
97 }
98 }
99 149298 }
100
101 149298 static av_always_inline void filter_plane_rows(VP9Context *s, int row, int ss_h, int ss_v,
102 uint8_t *lvl, uint8_t (*mask)[4],
103 uint8_t *dst, ptrdiff_t ls)
104 {
105 149298 int y, x, bytesperpixel = s->bytesperpixel;
106
107 // block1
108 // filter edges between rows (e.g. ------)
109 // block2
110
2/2
✓ Branch 0 taken 1194384 times.
✓ Branch 1 taken 149298 times.
1343682 for (y = 0; y < 8; y++, dst += 8 * ls >> ss_v) {
111 1194384 uint8_t *ptr = dst, *l = lvl, *vmask = mask[y];
112 1194384 unsigned vm = vmask[0] | vmask[1] | vmask[2], vm3 = vmask[3];
113
114
2/2
✓ Branch 0 taken 1385779 times.
✓ Branch 1 taken 1194384 times.
2580163 for (x = 1; vm & ~(x - 1); x <<= (2 << ss_h), ptr += 16 * bytesperpixel, l += 2 << ss_h) {
115
4/4
✓ Branch 0 taken 293184 times.
✓ Branch 1 taken 1092595 times.
✓ Branch 2 taken 224709 times.
✓ Branch 3 taken 68475 times.
1385779 if (row || y) {
116
2/2
✓ Branch 0 taken 1169997 times.
✓ Branch 1 taken 147307 times.
1317304 if (vm & x) {
117 1169997 int L = *l, H = L >> 4;
118 1169997 int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
119
120
2/2
✓ Branch 0 taken 347705 times.
✓ Branch 1 taken 822292 times.
1169997 if (vmask[0] & x) {
121
2/2
✓ Branch 0 taken 347386 times.
✓ Branch 1 taken 319 times.
347705 if (vmask[0] & (x << (1 + ss_h))) {
122 av_assert2(l[1 + ss_h] == L);
123 347386 s->dsp.loop_filter_16[1](ptr, ls, E, I, H);
124 } else {
125 319 s->dsp.loop_filter_8[2][1](ptr, ls, E, I, H);
126 }
127
2/2
✓ Branch 0 taken 754663 times.
✓ Branch 1 taken 67629 times.
822292 } else if (vm & (x << (1 + ss_h))) {
128 754663 L = l[1 + ss_h];
129 754663 H |= (L >> 4) << 8;
130 754663 E |= s->filter_lut.mblim_lut[L] << 8;
131 754663 I |= s->filter_lut.lim_lut[L] << 8;
132 754663 s->dsp.loop_filter_mix2[!!(vmask[1] & x)]
133 754663 [!!(vmask[1] & (x << (1 + ss_h)))]
134 754663 [1](ptr, ls, E, I, H);
135 } else {
136 67629 s->dsp.loop_filter_8[!!(vmask[1] & x)]
137 67629 [1](ptr, ls, E, I, H);
138 }
139
2/2
✓ Branch 0 taken 35358 times.
✓ Branch 1 taken 111949 times.
147307 } else if (vm & (x << (1 + ss_h))) {
140 35358 int L = l[1 + ss_h], H = L >> 4;
141 35358 int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
142
143 35358 s->dsp.loop_filter_8[!!(vmask[1] & (x << (1 + ss_h)))]
144 35358 [1](ptr + 8 * bytesperpixel, ls, E, I, H);
145 }
146 }
147
2/2
✓ Branch 0 taken 754557 times.
✓ Branch 1 taken 631222 times.
1385779 if (!ss_v) {
148
2/2
✓ Branch 0 taken 110039 times.
✓ Branch 1 taken 644518 times.
754557 if (vm3 & x) {
149 110039 int L = *l, H = L >> 4;
150 110039 int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
151
152
2/2
✓ Branch 0 taken 65543 times.
✓ Branch 1 taken 44496 times.
110039 if (vm3 & (x << (1 + ss_h))) {
153 65543 L = l[1 + ss_h];
154 65543 H |= (L >> 4) << 8;
155 65543 E |= s->filter_lut.mblim_lut[L] << 8;
156 65543 I |= s->filter_lut.lim_lut[L] << 8;
157 65543 s->dsp.loop_filter_mix2[0][0][1](ptr + ls * 4, ls, E, I, H);
158 } else {
159 44496 s->dsp.loop_filter_8[0][1](ptr + ls * 4, ls, E, I, H);
160 }
161
2/2
✓ Branch 0 taken 42225 times.
✓ Branch 1 taken 602293 times.
644518 } else if (vm3 & (x << (1 + ss_h))) {
162 42225 int L = l[1 + ss_h], H = L >> 4;
163 42225 int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
164
165 42225 s->dsp.loop_filter_8[0][1](ptr + ls * 4 + 8 * bytesperpixel, ls, E, I, H);
166 }
167 }
168 }
169
2/2
✓ Branch 0 taken 783008 times.
✓ Branch 1 taken 411376 times.
1194384 if (ss_v) {
170
2/2
✓ Branch 0 taken 391504 times.
✓ Branch 1 taken 391504 times.
783008 if (y & 1)
171 391504 lvl += 16;
172 } else {
173 411376 lvl += 8;
174 }
175 }
176 149298 }
177
178 49766 void ff_vp9_loopfilter_sb(AVCodecContext *avctx, VP9Filter *lflvl,
179 int row, int col, ptrdiff_t yoff, ptrdiff_t uvoff)
180 {
181 49766 VP9Context *s = avctx->priv_data;
182 49766 AVFrame *f = s->s.frames[CUR_FRAME].tf.f;
183 49766 uint8_t *dst = f->data[0] + yoff;
184 49766 ptrdiff_t ls_y = f->linesize[0], ls_uv = f->linesize[1];
185 49766 uint8_t (*uv_masks)[8][4] = lflvl->mask[s->ss_h | s->ss_v];
186 int p;
187
188 /* FIXME: In how far can we interleave the v/h loopfilter calls? E.g.
189 * if you think of them as acting on a 8x8 block max, we can interleave
190 * each v/h within the single x loop, but that only works if we work on
191 * 8 pixel blocks, and we won't always do that (we want at least 16px
192 * to use SSE2 optimizations, perhaps 32 for AVX2) */
193
194 49766 filter_plane_cols(s, col, 0, 0, lflvl->level, lflvl->mask[0][0], dst, ls_y);
195 49766 filter_plane_rows(s, row, 0, 0, lflvl->level, lflvl->mask[0][1], dst, ls_y);
196
197
2/2
✓ Branch 0 taken 99532 times.
✓ Branch 1 taken 49766 times.
149298 for (p = 0; p < 2; p++) {
198 99532 dst = f->data[1 + p] + uvoff;
199 99532 filter_plane_cols(s, col, s->ss_h, s->ss_v, lflvl->level, uv_masks[0], dst, ls_uv);
200 99532 filter_plane_rows(s, row, s->ss_h, s->ss_v, lflvl->level, uv_masks[1], dst, ls_uv);
201 }
202 49766 }
203