Line | Branch | Exec | Source |
---|---|---|---|
1 | /* | ||
2 | * rectangle filling function | ||
3 | * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at> | ||
4 | * | ||
5 | * This file is part of FFmpeg. | ||
6 | * | ||
7 | * FFmpeg is free software; you can redistribute it and/or | ||
8 | * modify it under the terms of the GNU Lesser General Public | ||
9 | * License as published by the Free Software Foundation; either | ||
10 | * version 2.1 of the License, or (at your option) any later version. | ||
11 | * | ||
12 | * FFmpeg is distributed in the hope that it will be useful, | ||
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
15 | * Lesser General Public License for more details. | ||
16 | * | ||
17 | * You should have received a copy of the GNU Lesser General Public | ||
18 | * License along with FFmpeg; if not, write to the Free Software | ||
19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
20 | */ | ||
21 | |||
22 | /** | ||
23 | * @file | ||
24 | * useful rectangle filling function | ||
25 | * @author Michael Niedermayer <michaelni@gmx.at> | ||
26 | */ | ||
27 | |||
28 | #ifndef AVCODEC_RECTANGLE_H | ||
29 | #define AVCODEC_RECTANGLE_H | ||
30 | |||
31 | #include "config.h" | ||
32 | #include "libavutil/common.h" | ||
33 | #include "libavutil/avassert.h" | ||
34 | |||
35 | /** | ||
36 | * fill a rectangle. | ||
37 | * @param h height of the rectangle, should be a constant | ||
38 | * @param w width of the rectangle, should be a constant | ||
39 | * @param size the size of val (1, 2 or 4), should be a constant | ||
40 | */ | ||
41 | 86971382 | static av_always_inline void fill_rectangle(void *vp, int w, int h, int stride, uint32_t val, int size){ | |
42 | 86971382 | uint8_t *p= (uint8_t*)vp; | |
43 | av_assert2(size==1 || size==2 || size==4); | ||
44 | av_assert2(w<=4); | ||
45 | |||
46 | 86971382 | w *= size; | |
47 | 86971382 | stride *= size; | |
48 | |||
49 | av_assert2((((long)vp)&(FFMIN(w, 8<<(HAVE_NEON|ARCH_PPC|HAVE_MMX))-1)) == 0); | ||
50 | av_assert2((stride&(w-1))==0); | ||
51 |
2/2✓ Branch 0 taken 24531362 times.
✓ Branch 1 taken 62440020 times.
|
86971382 | if(w==2){ |
52 |
1/2✓ Branch 0 taken 24531362 times.
✗ Branch 1 not taken.
|
24531362 | const uint16_t v= size==4 ? val : val*0x0101; |
53 | 24531362 | *(uint16_t*)(p + 0*stride)= v; | |
54 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 24531362 times.
|
24531362 | if(h==1) return; |
55 | 24531362 | *(uint16_t*)(p + 1*stride)= v; | |
56 |
2/2✓ Branch 0 taken 22839562 times.
✓ Branch 1 taken 1691800 times.
|
24531362 | if(h==2) return; |
57 | 1691800 | *(uint16_t*)(p + 2*stride)= v; | |
58 | 1691800 | *(uint16_t*)(p + 3*stride)= v; | |
59 |
2/2✓ Branch 0 taken 37520145 times.
✓ Branch 1 taken 24919875 times.
|
62440020 | }else if(w==4){ |
60 |
4/4✓ Branch 0 taken 36607655 times.
✓ Branch 1 taken 912490 times.
✓ Branch 2 taken 1817318 times.
✓ Branch 3 taken 34790337 times.
|
37520145 | const uint32_t v= size==4 ? val : size==2 ? val*0x00010001 : val*0x01010101; |
61 | 37520145 | *(uint32_t*)(p + 0*stride)= v; | |
62 |
2/2✓ Branch 0 taken 48 times.
✓ Branch 1 taken 37520097 times.
|
37520145 | if(h==1) return; |
63 | 37520097 | *(uint32_t*)(p + 1*stride)= v; | |
64 |
2/2✓ Branch 0 taken 3590870 times.
✓ Branch 1 taken 33929227 times.
|
37520097 | if(h==2) return; |
65 | 33929227 | *(uint32_t*)(p + 2*stride)= v; | |
66 | 33929227 | *(uint32_t*)(p + 3*stride)= v; | |
67 |
2/2✓ Branch 0 taken 9971816 times.
✓ Branch 1 taken 14948059 times.
|
24919875 | }else if(w==8){ |
68 | // gcc cannot optimize 64-bit math on x86_32 | ||
69 | #if HAVE_FAST_64BIT | ||
70 |
2/2✓ Branch 0 taken 4847331 times.
✓ Branch 1 taken 5124485 times.
|
9971816 | const uint64_t v= size==2 ? val*0x0001000100010001ULL : val*0x0100000001ULL; |
71 | 9971816 | *(uint64_t*)(p + 0*stride)= v; | |
72 |
2/2✓ Branch 0 taken 64 times.
✓ Branch 1 taken 9971752 times.
|
9971816 | if(h==1) return; |
73 | 9971752 | *(uint64_t*)(p + 1*stride)= v; | |
74 |
2/2✓ Branch 0 taken 4627673 times.
✓ Branch 1 taken 5344079 times.
|
9971752 | if(h==2) return; |
75 | 5344079 | *(uint64_t*)(p + 2*stride)= v; | |
76 | 5344079 | *(uint64_t*)(p + 3*stride)= v; | |
77 |
1/2✓ Branch 0 taken 14948059 times.
✗ Branch 1 not taken.
|
14948059 | }else if(w==16){ |
78 | 14948059 | const uint64_t v= val*0x0100000001ULL; | |
79 | 14948059 | *(uint64_t*)(p + 0+0*stride)= v; | |
80 | 14948059 | *(uint64_t*)(p + 8+0*stride)= v; | |
81 | 14948059 | *(uint64_t*)(p + 0+1*stride)= v; | |
82 | 14948059 | *(uint64_t*)(p + 8+1*stride)= v; | |
83 |
2/2✓ Branch 0 taken 1922602 times.
✓ Branch 1 taken 13025457 times.
|
14948059 | if(h==2) return; |
84 | 13025457 | *(uint64_t*)(p + 0+2*stride)= v; | |
85 | 13025457 | *(uint64_t*)(p + 8+2*stride)= v; | |
86 | 13025457 | *(uint64_t*)(p + 0+3*stride)= v; | |
87 | 13025457 | *(uint64_t*)(p + 8+3*stride)= v; | |
88 | #else | ||
89 | const uint32_t v= size==2 ? val*0x00010001 : val; | ||
90 | *(uint32_t*)(p + 0+0*stride)= v; | ||
91 | *(uint32_t*)(p + 4+0*stride)= v; | ||
92 | if(h==1) return; | ||
93 | *(uint32_t*)(p + 0+1*stride)= v; | ||
94 | *(uint32_t*)(p + 4+1*stride)= v; | ||
95 | if(h==2) return; | ||
96 | *(uint32_t*)(p + 0+2*stride)= v; | ||
97 | *(uint32_t*)(p + 4+2*stride)= v; | ||
98 | *(uint32_t*)(p + 0+3*stride)= v; | ||
99 | *(uint32_t*)(p + 4+3*stride)= v; | ||
100 | }else if(w==16){ | ||
101 | *(uint32_t*)(p + 0+0*stride)= val; | ||
102 | *(uint32_t*)(p + 4+0*stride)= val; | ||
103 | *(uint32_t*)(p + 8+0*stride)= val; | ||
104 | *(uint32_t*)(p +12+0*stride)= val; | ||
105 | *(uint32_t*)(p + 0+1*stride)= val; | ||
106 | *(uint32_t*)(p + 4+1*stride)= val; | ||
107 | *(uint32_t*)(p + 8+1*stride)= val; | ||
108 | *(uint32_t*)(p +12+1*stride)= val; | ||
109 | if(h==2) return; | ||
110 | *(uint32_t*)(p + 0+2*stride)= val; | ||
111 | *(uint32_t*)(p + 4+2*stride)= val; | ||
112 | *(uint32_t*)(p + 8+2*stride)= val; | ||
113 | *(uint32_t*)(p +12+2*stride)= val; | ||
114 | *(uint32_t*)(p + 0+3*stride)= val; | ||
115 | *(uint32_t*)(p + 4+3*stride)= val; | ||
116 | *(uint32_t*)(p + 8+3*stride)= val; | ||
117 | *(uint32_t*)(p +12+3*stride)= val; | ||
118 | #endif | ||
119 | }else | ||
120 | av_assert2(0); | ||
121 | av_assert2(h==4); | ||
122 | } | ||
123 | |||
124 | #endif /* AVCODEC_RECTANGLE_H */ | ||
125 |