2 * Copyright (c) 2015 Henrik Gramner
4 * This file is part of FFmpeg.
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.
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.
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.
23 #include "libavcodec/avcodec.h"
24 #include "libavcodec/h264pred.h"
25 #include "libavutil/common.h"
26 #include "libavutil/internal.h"
27 #include "libavutil/intreadwrite.h"
29 static const int codec_ids[4] = { AV_CODEC_ID_H264, AV_CODEC_ID_VP8, AV_CODEC_ID_RV40, AV_CODEC_ID_SVQ3 };
31 static const char * const pred4x4_modes[4][15] = {
33 [VERT_PRED ] = "vertical",
34 [HOR_PRED ] = "horizontal",
36 [DIAG_DOWN_LEFT_PRED ] = "down_left",
37 [DIAG_DOWN_RIGHT_PRED] = "down_right",
38 [VERT_RIGHT_PRED ] = "vertical_right",
39 [HOR_DOWN_PRED ] = "horizontal_right",
40 [VERT_LEFT_PRED ] = "vertical_left",
41 [HOR_UP_PRED ] = "horizontal_up",
42 [LEFT_DC_PRED ] = "left_dc",
43 [TOP_DC_PRED ] = "top_dc",
44 [DC_128_PRED ] = "dc_128",
47 [VERT_PRED ] = "vertical_vp8",
48 [HOR_PRED ] = "horizontal_vp8",
49 [VERT_LEFT_PRED] = "vertical_left_vp8",
50 [TM_VP8_PRED ] = "tm_vp8",
51 [DC_127_PRED ] = "dc_127_vp8",
52 [DC_129_PRED ] = "dc_129_vp8",
55 [DIAG_DOWN_LEFT_PRED ] = "down_left_rv40",
56 [VERT_LEFT_PRED ] = "vertical_left_rv40",
57 [HOR_UP_PRED ] = "horizontal_up_rv40",
58 [DIAG_DOWN_LEFT_PRED_RV40_NODOWN] = "down_left_nodown_rv40",
59 [HOR_UP_PRED_RV40_NODOWN ] = "horizontal_up_nodown_rv40",
60 [VERT_LEFT_PRED_RV40_NODOWN ] = "vertical_left_nodown_rv40",
63 [DIAG_DOWN_LEFT_PRED] = "down_left_svq3",
67 static const char * const pred8x8_modes[4][11] = {
70 [HOR_PRED8x8 ] = "horizontal",
71 [VERT_PRED8x8 ] = "vertical",
72 [PLANE_PRED8x8 ] = "plane",
73 [LEFT_DC_PRED8x8 ] = "left_dc",
74 [TOP_DC_PRED8x8 ] = "top_dc",
75 [DC_128_PRED8x8 ] = "dc_128",
76 [ALZHEIMER_DC_L0T_PRED8x8] = "mad_cow_dc_l0t",
77 [ALZHEIMER_DC_0LT_PRED8x8] = "mad_cow_dc_0lt",
78 [ALZHEIMER_DC_L00_PRED8x8] = "mad_cow_dc_l00",
79 [ALZHEIMER_DC_0L0_PRED8x8] = "mad_cow_dc_0l0",
82 [PLANE_PRED8x8 ] = "tm_vp8",
83 [DC_127_PRED8x8] = "dc_127_vp8",
84 [DC_129_PRED8x8] = "dc_129_vp8",
87 [DC_PRED8x8 ] = "dc_rv40",
88 [LEFT_DC_PRED8x8] = "left_dc_rv40",
89 [TOP_DC_PRED8x8 ] = "top_dc_rv40",
91 /* nothing for SVQ3 */
94 static const char * const pred16x16_modes[4][9] = {
97 [HOR_PRED8x8 ] = "horizontal",
98 [VERT_PRED8x8 ] = "vertical",
99 [PLANE_PRED8x8 ] = "plane",
100 [LEFT_DC_PRED8x8] = "left_dc",
101 [TOP_DC_PRED8x8 ] = "top_dc",
102 [DC_128_PRED8x8 ] = "dc_128",
105 [PLANE_PRED8x8 ] = "tm_vp8",
106 [DC_127_PRED8x8] = "dc_127_vp8",
107 [DC_129_PRED8x8] = "dc_129_vp8",
110 [PLANE_PRED8x8] = "plane_rv40",
113 [PLANE_PRED8x8] = "plane_svq3",
117 static const uint32_t pixel_mask[3] = { 0xffffffff, 0x01ff01ff, 0x03ff03ff };
119 #define SIZEOF_PIXEL ((bit_depth + 7) / 8)
120 #define BUF_SIZE (3 * 16 * 17)
122 #define check_pred_func(func, name, mode_name) \
123 (mode_name && ((codec_ids[codec] == AV_CODEC_ID_H264) ? \
124 check_func(func, "pred%s_%s_%d", name, mode_name, bit_depth) : \
125 check_func(func, "pred%s_%s", name, mode_name)))
127 #define randomize_buffers() \
129 uint32_t mask = pixel_mask[bit_depth - 8]; \
131 for (i = 0; i < BUF_SIZE; i += 4) { \
132 uint32_t r = rnd() & mask; \
133 AV_WN32A(buf0 + i, r); \
134 AV_WN32A(buf1 + i, r); \
138 #define src0 (buf0 + 4 * 16) /* Offset to allow room for top and left */
139 #define src1 (buf1 + 4 * 16)
141 static void check_pred4x4(H264PredContext *h, uint8_t *buf0, uint8_t *buf1,
142 int codec, int chroma_format, int bit_depth)
144 if (chroma_format == 1) {
145 uint8_t *topright = buf0 + 2*16;
147 declare_func_emms(AV_CPU_FLAG_MMX | AV_CPU_FLAG_MMXEXT, void, uint8_t *src, const uint8_t *topright, ptrdiff_t stride);
149 for (pred_mode = 0; pred_mode < 15; pred_mode++) {
150 if (check_pred_func(h->pred4x4[pred_mode], "4x4", pred4x4_modes[codec][pred_mode])) {
152 call_ref(src0, topright, 12*SIZEOF_PIXEL);
153 call_new(src1, topright, 12*SIZEOF_PIXEL);
154 if (memcmp(buf0, buf1, BUF_SIZE))
156 bench_new(src1, topright, 12*SIZEOF_PIXEL);
162 static void check_pred8x8(H264PredContext *h, uint8_t *buf0, uint8_t *buf1,
163 int codec, int chroma_format, int bit_depth)
166 declare_func_emms(AV_CPU_FLAG_MMX | AV_CPU_FLAG_MMXEXT, void, uint8_t *src, ptrdiff_t stride);
168 for (pred_mode = 0; pred_mode < 11; pred_mode++) {
169 if (check_pred_func(h->pred8x8[pred_mode], (chroma_format == 2) ? "8x16" : "8x8",
170 pred8x8_modes[codec][pred_mode])) {
172 call_ref(src0, 24*SIZEOF_PIXEL);
173 call_new(src1, 24*SIZEOF_PIXEL);
174 if (memcmp(buf0, buf1, BUF_SIZE))
176 bench_new(src1, 24*SIZEOF_PIXEL);
181 static void check_pred16x16(H264PredContext *h, uint8_t *buf0, uint8_t *buf1,
182 int codec, int chroma_format, int bit_depth)
184 if (chroma_format == 1) {
186 declare_func_emms(AV_CPU_FLAG_MMX | AV_CPU_FLAG_MMXEXT, void, uint8_t *src, ptrdiff_t stride);
188 for (pred_mode = 0; pred_mode < 9; pred_mode++) {
189 if (check_pred_func(h->pred16x16[pred_mode], "16x16", pred16x16_modes[codec][pred_mode])) {
193 if (memcmp(buf0, buf1, BUF_SIZE))
201 static void check_pred8x8l(H264PredContext *h, uint8_t *buf0, uint8_t *buf1,
202 int codec, int chroma_format, int bit_depth)
204 if (chroma_format == 1 && codec_ids[codec] == AV_CODEC_ID_H264) {
206 declare_func_emms(AV_CPU_FLAG_MMXEXT, void, uint8_t *src, int topleft, int topright, ptrdiff_t stride);
208 for (pred_mode = 0; pred_mode < 12; pred_mode++) {
209 if (check_pred_func(h->pred8x8l[pred_mode], "8x8l", pred4x4_modes[codec][pred_mode])) {
211 for (neighbors = 0; neighbors <= 0xc000; neighbors += 0x4000) {
212 int has_topleft = neighbors & 0x8000;
213 int has_topright = neighbors & 0x4000;
215 if ((pred_mode == DIAG_DOWN_RIGHT_PRED || pred_mode == VERT_RIGHT_PRED) && !has_topleft)
216 continue; /* Those aren't allowed according to the spec */
219 call_ref(src0, has_topleft, has_topright, 24*SIZEOF_PIXEL);
220 call_new(src1, has_topleft, has_topright, 24*SIZEOF_PIXEL);
221 if (memcmp(buf0, buf1, BUF_SIZE))
223 bench_new(src1, has_topleft, has_topright, 24*SIZEOF_PIXEL);
230 /* TODO: Add tests for H.264 lossless H/V prediction */
232 void checkasm_check_h264pred(void)
234 static const struct {
235 void (*func)(H264PredContext*, uint8_t*, uint8_t*, int, int, int);
238 { check_pred4x4, "pred4x4" },
239 { check_pred8x8, "pred8x8" },
240 { check_pred16x16, "pred16x16" },
241 { check_pred8x8l, "pred8x8l" },
244 LOCAL_ALIGNED_16(uint8_t, buf0, [BUF_SIZE]);
245 LOCAL_ALIGNED_16(uint8_t, buf1, [BUF_SIZE]);
247 int test, codec, chroma_format, bit_depth;
249 for (test = 0; test < FF_ARRAY_ELEMS(tests); test++) {
250 for (codec = 0; codec < 4; codec++) {
251 int codec_id = codec_ids[codec];
252 for (bit_depth = 8; bit_depth <= (codec_id == AV_CODEC_ID_H264 ? 10 : 8); bit_depth++)
253 for (chroma_format = 1; chroma_format <= (codec_id == AV_CODEC_ID_H264 ? 2 : 1); chroma_format++) {
254 ff_h264_pred_init(&h, codec_id, bit_depth, chroma_format);
255 tests[test].func(&h, buf0, buf1, codec, chroma_format, bit_depth);
258 report("%s", tests[test].name);