1 /*
2 * Copyright (c) 2018 Yingming Fan <yingmingfan@gmail.com>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License along
17 * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
18 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
19 */
20
21 #include <string.h>
22
23 #include "libavutil/intreadwrite.h"
24 #include "libavutil/mem_internal.h"
25
26 #include "libavcodec/avcodec.h"
27
28 #include "libavcodec/hevcdsp.h"
29
30 #include "checkasm.h"
31
32 static const uint32_t pixel_mask[3] = { 0xffffffff, 0x03ff03ff, 0x0fff0fff };
33 static const uint32_t sao_size[5] = {8, 16, 32, 48, 64};
34
35 #define SIZEOF_PIXEL ((bit_depth + 7) / 8)
36 #define PIXEL_STRIDE (2*MAX_PB_SIZE + AV_INPUT_BUFFER_PADDING_SIZE) //same with sao_edge src_stride
37 #define BUF_SIZE (PIXEL_STRIDE * (64+2) * 2) //+2 for top and bottom row, *2 for high bit depth
38 #define OFFSET_THRESH (1 << (bit_depth - 5))
39 #define OFFSET_LENGTH 5
40
41 #define randomize_buffers(buf0, buf1, size) \
42 do { \
43 uint32_t mask = pixel_mask[(bit_depth - 8) >> 1]; \
44 int k; \
45 for (k = 0; k < size; k += 4) { \
46 uint32_t r = rnd() & mask; \
47 AV_WN32A(buf0 + k, r); \
48 AV_WN32A(buf1 + k, r); \
49 } \
50 } while (0)
51
52 #define randomize_buffers2(buf, size) \
53 do { \
54 uint32_t max_offset = OFFSET_THRESH; \
55 int k; \
56 if (bit_depth == 8) { \
57 for (k = 0; k < size; k++) { \
58 uint8_t r = rnd() % max_offset; \
59 buf[k] = r; \
60 } \
61 } else { \
62 for (k = 0; k < size; k++) { \
63 uint16_t r = rnd() % max_offset; \
64 buf[k] = r; \
65 } \
66 } \
67 } while (0)
68
check_sao_band(HEVCDSPContext h,int bit_depth)69 static void check_sao_band(HEVCDSPContext h, int bit_depth)
70 {
71 int i;
72 LOCAL_ALIGNED_32(uint8_t, dst0, [BUF_SIZE]);
73 LOCAL_ALIGNED_32(uint8_t, dst1, [BUF_SIZE]);
74 LOCAL_ALIGNED_32(uint8_t, src0, [BUF_SIZE]);
75 LOCAL_ALIGNED_32(uint8_t, src1, [BUF_SIZE]);
76 int16_t offset_val[OFFSET_LENGTH];
77 int left_class = rnd()%32;
78
79 for (i = 0; i <= 4; i++) {
80 int block_size = sao_size[i];
81 ptrdiff_t stride = PIXEL_STRIDE*SIZEOF_PIXEL;
82 declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *dst, uint8_t *src, ptrdiff_t dst_stride, ptrdiff_t src_stride,
83 int16_t *sao_offset_val, int sao_left_class, int width, int height);
84
85 randomize_buffers(src0, src1, BUF_SIZE);
86 randomize_buffers2(offset_val, OFFSET_LENGTH);
87 memset(dst0, 0, BUF_SIZE);
88 memset(dst1, 0, BUF_SIZE);
89
90 if (check_func(h.sao_band_filter[i], "hevc_sao_band_%dx%d_%d", block_size, block_size, bit_depth)) {
91 call_ref(dst0, src0, stride, stride, offset_val, left_class, block_size, block_size);
92 call_new(dst1, src1, stride, stride, offset_val, left_class, block_size, block_size);
93 if (memcmp(dst0, dst1, BUF_SIZE))
94 fail();
95 bench_new(dst1, src1, stride, stride, offset_val, left_class, block_size, block_size);
96 }
97 }
98 }
99
check_sao_edge(HEVCDSPContext h,int bit_depth)100 static void check_sao_edge(HEVCDSPContext h, int bit_depth)
101 {
102 int i;
103 LOCAL_ALIGNED_32(uint8_t, dst0, [BUF_SIZE]);
104 LOCAL_ALIGNED_32(uint8_t, dst1, [BUF_SIZE]);
105 LOCAL_ALIGNED_32(uint8_t, src0, [BUF_SIZE]);
106 LOCAL_ALIGNED_32(uint8_t, src1, [BUF_SIZE]);
107 int16_t offset_val[OFFSET_LENGTH];
108 int eo = rnd()%4;
109
110 for (i = 0; i <= 4; i++) {
111 int block_size = sao_size[i];
112 ptrdiff_t stride = PIXEL_STRIDE*SIZEOF_PIXEL;
113 int offset = (AV_INPUT_BUFFER_PADDING_SIZE + PIXEL_STRIDE)*SIZEOF_PIXEL;
114 declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *dst, uint8_t *src, ptrdiff_t stride_dst,
115 int16_t *sao_offset_val, int eo, int width, int height);
116
117 randomize_buffers(src0, src1, BUF_SIZE);
118 randomize_buffers2(offset_val, OFFSET_LENGTH);
119 memset(dst0, 0, BUF_SIZE);
120 memset(dst1, 0, BUF_SIZE);
121
122 if (check_func(h.sao_edge_filter[i], "hevc_sao_edge_%dx%d_%d", block_size, block_size, bit_depth)) {
123 call_ref(dst0, src0 + offset, stride, offset_val, eo, block_size, block_size);
124 call_new(dst1, src1 + offset, stride, offset_val, eo, block_size, block_size);
125 if (memcmp(dst0, dst1, BUF_SIZE))
126 fail();
127 bench_new(dst1, src1 + offset, stride, offset_val, eo, block_size, block_size);
128 }
129 }
130 }
131
checkasm_check_hevc_sao(void)132 void checkasm_check_hevc_sao(void)
133 {
134 int bit_depth;
135
136 for (bit_depth = 8; bit_depth <= 12; bit_depth += 2) {
137 HEVCDSPContext h;
138
139 ff_hevc_dsp_init(&h, bit_depth);
140 check_sao_band(h, bit_depth);
141 }
142 report("sao_band");
143
144 for (bit_depth = 8; bit_depth <= 12; bit_depth += 2) {
145 HEVCDSPContext h;
146
147 ff_hevc_dsp_init(&h, bit_depth);
148 check_sao_edge(h, bit_depth);
149 }
150 report("sao_edge");
151 }
152