1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
|
/* See LICENSE file for copyright and license details. */
#ifndef TYPE
#include "common.h"
USAGE("[-en] leftmost-stream ... rightmost-stream")
static int equal = 0;
static size_t max_frame_size;
#define FILE "blind-multiply-matrices.c"
#include "define-functions.h"
int
main(int argc, char *argv[])
{
struct stream *streams;
size_t n_streams, i, frames = 0;
int natural = 0, j;
char **rev_argv;
size_t max_width = 0, max_height = 0;
size_t width = 0, height = 0, w, h;
void (*process)(struct stream *streams, size_t n_streams);
ARGBEGIN {
case 'e':
equal = 1;
break;
case 'n':
natural = 1;
break;
default:
usage();
} ARGEND;
if (argc < 2)
usage();
if (natural) {
rev_argv = alloca((size_t)argc * sizeof(*rev_argv));
for (j = 0; j < argc; j++)
rev_argv[j] = argv[argc - 1 - j];
argv = rev_argv;
}
n_streams = (size_t)argc;
streams = ecalloc(n_streams, sizeof(*streams));
for (i = 0; i < n_streams; i++) {
eopen_stream(streams + i, argv[i]);
if (streams[i].frames && streams[i].frames < frames)
frames = streams[i].frames;
if (streams->width > max_width)
max_width = streams->width;
if (streams->height > max_height)
max_height = streams->height;
}
for (i = 1; i < n_streams; i++)
if (strcmp(streams->pixfmt, streams[i].pixfmt))
eprintf("videos use incompatible pixel formats\n");
width = streams[n_streams - 1].width;
height = streams[n_streams - 1].height;
for (i = n_streams - 1; i--;) {
if (streams[i].width != height)
eprintf("videos do not have the compatible geometry\n");
height = streams[i].height;
}
if (streams->encoding == DOUBLE)
process = process_lf;
else if (streams->encoding == FLOAT)
process = process_f;
else
eprintf("pixel format %s is not supported, try xyza\n", streams->pixfmt);
w = streams->width, streams->width = max_width;
h = streams->height, streams->height = max_height;
echeck_dimensions(streams, WIDTH | HEIGHT, NULL);
streams->width = width;
streams->height = height;
streams->frames = frames;
fprint_stream_head(stdout, streams);
streams->width = w;
streams->height = h;
efflush(stdout, "<stdout>");
max_frame_size = max_width * max_height * streams->pixel_size;
process(streams, n_streams);
free(streams);
return 0;
}
#else
static void
PROCESS(struct stream *streams, size_t n_streams)
{
typedef TYPE pixel_t[4];
pixel_t *res, *left, *right, *tmp;
size_t i, j, w, h, h2, x, y, k, r;
res = emalloc(max_frame_size);
left = emalloc(max_frame_size);
right = emalloc(max_frame_size);
while (eread_frame(streams + (n_streams - 1), res)) {
w = streams[n_streams - 1].width;
h = streams[n_streams - 1].height;
for (i = n_streams - 1; i--;) {
tmp = res, res = right, right = tmp;
if (!eread_frame(streams + i, left))
goto done;
h2 = streams[i].height;
memset(res, 0, w * h2 * streams->pixel_size);
/* XXX Is there any significant performance to be gained by transposing `right`? */
if (equal) {
for (y = r = 0; y < h2; y++) {
for (x = 0; x < w; x++, r++) {
for (k = 0; k < h; k++)
res[r][0] += left[y * h + k][0] * right[k * w + x][0];
for (j = 1; j < streams->n_chan; j++)
res[r][j] = res[r][0];
}
}
} else {
for (y = r = 0; y < h2; y++)
for (x = 0; x < w; x++, r++)
for (k = 0; k < h; k++)
for (j = 0; j < streams->n_chan; j++)
res[r][j] += left[y * h + k][j] * right[k * w + x][j];
}
h = h2;
}
ewriteall(STDOUT_FILENO, res, streams->frame_size, "<stdout>");
}
done:
free(res);
free(left);
free(right);
}
#endif
|
