]> git.sesse.net Git - ffmpeg/blob - tests/checkasm/pixblockdsp.c
Merge commit 'c4642788e83b0858bca449f9b6e71ddb015dfa5d'
[ffmpeg] / tests / checkasm / pixblockdsp.c
1 /*
2  * Copyright (c) 2015 Tiancheng "Timothy" Gu
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 #include "checkasm.h"
23 #include "libavcodec/pixblockdsp.h"
24 #include "libavutil/common.h"
25 #include "libavutil/internal.h"
26 #include "libavutil/intreadwrite.h"
27
28 #define BUF_UNITS 8
29 #define BUF_SIZE (BUF_UNITS * 128 + 8 * BUF_UNITS)
30
31 #define randomize_buffers()                 \
32     do {                                    \
33         int i;                              \
34         for (i = 0; i < BUF_SIZE; i += 4) { \
35             uint32_t r = rnd();             \
36             AV_WN32A(src10 + i, r);         \
37             AV_WN32A(src11 + i, r);         \
38             r = rnd();                      \
39             AV_WN32A(src20 + i, r);         \
40             AV_WN32A(src21 + i, r);         \
41             r = rnd();                      \
42             AV_WN32A(dst0_ + i, r);         \
43             AV_WN32A(dst1_ + i, r);         \
44         }                                   \
45     } while (0)
46
47 #define check_get_pixels(type)                                                             \
48     do {                                                                                   \
49         int i;                                                                             \
50         declare_func_emms(AV_CPU_FLAG_MMX, void, int16_t *block, const uint8_t *pixels, ptrdiff_t line_size);    \
51                                                                                            \
52         for (i = 0; i < BUF_UNITS; i++) {                                              \
53             int src_offset = i * 64 * sizeof(type) + 8 * i; /* Test various alignments */      \
54             int dst_offset = i * 64; /* dst must be aligned */                             \
55             randomize_buffers();                                                           \
56             call_ref(dst0 + dst_offset, src10 + src_offset, 8);                            \
57             call_new(dst1 + dst_offset, src11 + src_offset, 8);                            \
58             if (memcmp(src10, src11, BUF_SIZE)|| memcmp(dst0, dst1, BUF_SIZE)) \
59                 fail();                                                                    \
60             bench_new(dst1 + dst_offset, src11 + src_offset, 8);                           \
61         }                                                                                  \
62     } while (0)
63
64 #define check_diff_pixels(type)                                                            \
65     do {                                                                                   \
66         int i;                                                                             \
67         declare_func_emms(AV_CPU_FLAG_MMX, void, int16_t *av_restrict block, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride); \
68                                                                                            \
69         for (i = 0; i < BUF_UNITS; i++) {                                              \
70             int src_offset = i * 64 * sizeof(type) + 8 * i; /* Test various alignments */      \
71             int dst_offset = i * 64; /* dst must be aligned */                             \
72             randomize_buffers();                                                           \
73             call_ref(dst0 + dst_offset, src10 + src_offset, src20 + src_offset, 8);        \
74             call_new(dst1 + dst_offset, src11 + src_offset, src21 + src_offset, 8);        \
75             if (memcmp(src10, src11, BUF_SIZE) || memcmp(src20, src21, BUF_SIZE) || memcmp(dst0, dst1, BUF_SIZE)) \
76                 fail();                                                                    \
77             bench_new(dst1 + dst_offset, src11 + src_offset, src21 + src_offset, 8);       \
78         }                                                                                  \
79     } while (0)
80
81 void checkasm_check_pixblockdsp(void)
82 {
83     LOCAL_ALIGNED_16(uint8_t, src10, [BUF_SIZE]);
84     LOCAL_ALIGNED_16(uint8_t, src11, [BUF_SIZE]);
85     LOCAL_ALIGNED_16(uint8_t, src20, [BUF_SIZE]);
86     LOCAL_ALIGNED_16(uint8_t, src21, [BUF_SIZE]);
87     LOCAL_ALIGNED_16(uint8_t, dst0_, [BUF_SIZE]);
88     LOCAL_ALIGNED_16(uint8_t, dst1_, [BUF_SIZE]);
89     uint16_t *dst0 = (uint16_t *)dst0_;
90     uint16_t *dst1 = (uint16_t *)dst1_;
91     PixblockDSPContext h;
92     AVCodecContext avctx = {
93         .bits_per_raw_sample = 8,
94     };
95
96     ff_pixblockdsp_init(&h, &avctx);
97
98     if (check_func(h.get_pixels, "get_pixels"))
99         check_get_pixels(uint8_t);
100
101     report("get_pixels");
102
103     if (check_func(h.diff_pixels, "diff_pixels"))
104         check_diff_pixels(uint8_t);
105
106     report("diff_pixels");
107 }