]> git.sesse.net Git - ffmpeg/blob - tests/checkasm/pixblockdsp.c
lavu: move LOCAL_ALIGNED from internal.h to mem_internal.h
[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 #include "libavutil/mem_internal.h"
28
29 #define BUF_UNITS 8
30 #define BUF_SIZE (BUF_UNITS * 128 + 8 * BUF_UNITS)
31
32 #define randomize_buffers()                 \
33     do {                                    \
34         int i;                              \
35         for (i = 0; i < BUF_SIZE; i += 4) { \
36             uint32_t r = rnd();             \
37             AV_WN32A(src10 + i, r);         \
38             AV_WN32A(src11 + i, r);         \
39             r = rnd();                      \
40             AV_WN32A(src20 + i, r);         \
41             AV_WN32A(src21 + i, r);         \
42             r = rnd();                      \
43             AV_WN32A(dst0_ + i, r);         \
44             AV_WN32A(dst1_ + i, r);         \
45         }                                   \
46     } while (0)
47
48 #define check_get_pixels(type, aligned)                                                    \
49     do {                                                                                   \
50         int i;                                                                             \
51         declare_func_emms(AV_CPU_FLAG_MMX, void, int16_t *block, const uint8_t *pixels, ptrdiff_t line_size);    \
52                                                                                            \
53         for (i = 0; i < BUF_UNITS; i++) {                                              \
54             int src_offset = i * 64 * sizeof(type) + (aligned ? 8 : 1) * i;                \
55             int dst_offset = i * 64; /* dst must be aligned */                             \
56             randomize_buffers();                                                           \
57             call_ref(dst0 + dst_offset, src10 + src_offset, 8);                            \
58             call_new(dst1 + dst_offset, src11 + src_offset, 8);                            \
59             if (memcmp(src10, src11, BUF_SIZE)|| memcmp(dst0, dst1, BUF_SIZE)) \
60                 fail();                                                                    \
61             bench_new(dst1 + dst_offset, src11 + src_offset, 8);                           \
62         }                                                                                  \
63     } while (0)
64
65 #define check_diff_pixels(type, aligned)                                                   \
66     do {                                                                                   \
67         int i;                                                                             \
68         declare_func_emms(AV_CPU_FLAG_MMX, void, int16_t *av_restrict block, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride); \
69                                                                                            \
70         for (i = 0; i < BUF_UNITS; i++) {                                              \
71             int src_offset = i * 64 * sizeof(type) + (aligned ? 8 : 1) * i;                \
72             int dst_offset = i * 64; /* dst must be aligned */                             \
73             randomize_buffers();                                                           \
74             call_ref(dst0 + dst_offset, src10 + src_offset, src20 + src_offset, 8);        \
75             call_new(dst1 + dst_offset, src11 + src_offset, src21 + src_offset, 8);        \
76             if (memcmp(src10, src11, BUF_SIZE) || memcmp(src20, src21, BUF_SIZE) || memcmp(dst0, dst1, BUF_SIZE)) \
77                 fail();                                                                    \
78             bench_new(dst1 + dst_offset, src11 + src_offset, src21 + src_offset, 8);       \
79         }                                                                                  \
80     } while (0)
81
82 void checkasm_check_pixblockdsp(void)
83 {
84     LOCAL_ALIGNED_16(uint8_t, src10, [BUF_SIZE]);
85     LOCAL_ALIGNED_16(uint8_t, src11, [BUF_SIZE]);
86     LOCAL_ALIGNED_16(uint8_t, src20, [BUF_SIZE]);
87     LOCAL_ALIGNED_16(uint8_t, src21, [BUF_SIZE]);
88     LOCAL_ALIGNED_16(uint8_t, dst0_, [BUF_SIZE]);
89     LOCAL_ALIGNED_16(uint8_t, dst1_, [BUF_SIZE]);
90     uint16_t *dst0 = (uint16_t *)dst0_;
91     uint16_t *dst1 = (uint16_t *)dst1_;
92     PixblockDSPContext h;
93     AVCodecContext avctx = {
94         .bits_per_raw_sample = 8,
95     };
96
97     ff_pixblockdsp_init(&h, &avctx);
98
99     if (check_func(h.get_pixels, "get_pixels"))
100         check_get_pixels(uint8_t, 1);
101     if (check_func(h.get_pixels_unaligned, "get_pixels_unaligned"))
102         check_get_pixels(uint8_t, 0);
103
104     report("get_pixels");
105
106     if (check_func(h.diff_pixels, "diff_pixels"))
107         check_diff_pixels(uint8_t, 1);
108     if (check_func(h.diff_pixels_unaligned, "diff_pixels_unaligned"))
109         check_diff_pixels(uint8_t, 0);
110
111     report("diff_pixels");
112 }