zoomvtools 2.0.0

Video motion vector analysis utilities in pure Rust
Documentation 
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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336

#[cfg(all(target_arch = "x86_64", feature = "simd"))]
mod avx2;
#[cfg(all(target_arch = "x86_64", feature = "simd"))]
mod avx512;
mod rust;

#[cfg(test)]
mod tests;

use std::{mem::size_of, num::NonZeroUsize};

use crate::util::Pixel;

/// Function pointer for SAD (Sum of Absolute Differences) with block size baked in.
pub type SadFn = unsafe fn(
    src: *const u8,
    src_pitch: NonZeroUsize,
    ref_: *const u8,
    ref_pitch: NonZeroUsize,
) -> u64;

/// Selects the appropriate SAD function for the given block size.
///
/// # Panics
/// Panics if the block size is not supported.
#[must_use]
#[inline]
pub fn select_sad<T: Pixel>(width: NonZeroUsize, height: NonZeroUsize) -> SadFn {
    #[cfg(all(target_arch = "x86_64", feature = "simd"))]
    if cpudetect::x86_64::is_znver5_compatible() {
        match (size_of::<T>(), width.get(), height.get()) {
            // PERF: Equivalent to AVX2 on test machine
            (1, 2, 2) => return avx512::get_sad_u8::<2, 2>,
            (1, 2, 4) => return avx512::get_sad_u8::<2, 4>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 4, 2) => return avx512::get_sad_u8::<4, 2>,
            (1, 4, 4) => return avx512::get_sad_u8::<4, 4>,
            (1, 4, 8) => return avx512::get_sad_u8::<4, 8>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 8, 1) => return avx512::get_sad_u8::<8, 1>,
            (1, 8, 2) => return avx512::get_sad_u8::<8, 2>,
            (1, 8, 4) => return avx512::get_sad_u8::<8, 4>,
            (1, 8, 8) => return avx512::get_sad_u8::<8, 8>,
            (1, 8, 16) => return avx512::get_sad_u8::<8, 16>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 16, 1) => return avx512::get_sad_u8::<16, 1>,
            (1, 16, 2) => return avx512::get_sad_u8::<16, 2>,
            (1, 16, 4) => return avx512::get_sad_u8::<16, 4>,
            (1, 16, 8) => return avx512::get_sad_u8::<16, 8>,
            (1, 16, 16) => return avx512::get_sad_u8::<16, 16>,
            (1, 16, 32) => return avx512::get_sad_u8::<16, 32>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 32, 8) => return avx512::get_sad_u8::<32, 8>,
            (1, 32, 16) => return avx512::get_sad_u8::<32, 16>,
            (1, 32, 32) => return avx512::get_sad_u8::<32, 32>,
            (1, 32, 64) => return avx512::get_sad_u8::<32, 64>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 64, 16) => return avx512::get_sad_u8::<64, 16>,
            (1, 64, 32) => return avx512::get_sad_u8::<64, 32>,
            (1, 64, 64) => return avx512::get_sad_u8::<64, 64>,
            (1, 64, 128) => return avx512::get_sad_u8::<64, 128>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 128, 32) => return avx512::get_sad_u8::<128, 32>,
            (1, 128, 64) => return avx512::get_sad_u8::<128, 64>,
            (1, 128, 128) => return avx512::get_sad_u8::<128, 128>,
            // PERF: Equivalent to AVX2 on test machine
            (2, 2, 2) => return avx512::get_sad_u16::<2, 2>,
            (2, 2, 4) => return avx512::get_sad_u16::<2, 4>,
            // PERF: Equivalent to AVX2 on test machine
            (2, 4, 2) => return avx512::get_sad_u16::<4, 2>,
            (2, 4, 4) => return avx512::get_sad_u16::<4, 4>,
            (2, 4, 8) => return avx512::get_sad_u16::<4, 8>,
            // PERF: Equivalent to AVX2 on test machine
            (2, 8, 1) => return avx512::get_sad_u16::<8, 1>,
            (2, 8, 2) => return avx512::get_sad_u16::<8, 2>,
            (2, 8, 4) => return avx512::get_sad_u16::<8, 4>,
            (2, 8, 8) => return avx512::get_sad_u16::<8, 8>,
            (2, 8, 16) => return avx512::get_sad_u16::<8, 16>,
            // PERF: 10% faster than AVX2 on test machine
            (2, 16, 1) => return avx512::get_sad_u16::<16, 1>,
            (2, 16, 2) => return avx512::get_sad_u16::<16, 2>,
            (2, 16, 4) => return avx512::get_sad_u16::<16, 4>,
            (2, 16, 8) => return avx512::get_sad_u16::<16, 8>,
            (2, 16, 16) => return avx512::get_sad_u16::<16, 16>,
            (2, 16, 32) => return avx512::get_sad_u16::<16, 32>,
            // PERF: 20% faster than AVX2 on test machine
            (2, 32, 8) => return avx512::get_sad_u16::<32, 8>,
            (2, 32, 16) => return avx512::get_sad_u16::<32, 16>,
            (2, 32, 32) => return avx512::get_sad_u16::<32, 32>,
            (2, 32, 64) => return avx512::get_sad_u16::<32, 64>,
            // PERF: 40% faster than AVX2 on test machine
            (2, 64, 16) => return avx512::get_sad_u16::<64, 16>,
            (2, 64, 32) => return avx512::get_sad_u16::<64, 32>,
            (2, 64, 64) => return avx512::get_sad_u16::<64, 64>,
            (2, 64, 128) => return avx512::get_sad_u16::<64, 128>,
            // PERF: 20% faster than AVX2 on test machine
            (2, 128, 32) => return avx512::get_sad_u16::<128, 32>,
            (2, 128, 64) => return avx512::get_sad_u16::<128, 64>,
            (2, 128, 128) => return avx512::get_sad_u16::<128, 128>,
            _ => {}
        };
    }

    #[cfg(all(target_arch = "x86_64", feature = "simd"))]
    if cpudetect::x86_64::is_x86_64_v4_compatible() {
        match (size_of::<T>(), width.get(), height.get()) {
            // TODO: 5% slower than scalar
            #[cfg(feature = "experimental")]
            (1, 2, 2) => return avx512::get_sad_u8::<2, 2>,
            #[cfg(feature = "experimental")]
            (1, 2, 4) => return avx512::get_sad_u8::<2, 4>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 4, 2) => return avx512::get_sad_u8::<4, 2>,
            (1, 4, 4) => return avx512::get_sad_u8::<4, 4>,
            (1, 4, 8) => return avx512::get_sad_u8::<4, 8>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 8, 1) => return avx512::get_sad_u8::<8, 1>,
            (1, 8, 2) => return avx512::get_sad_u8::<8, 2>,
            (1, 8, 4) => return avx512::get_sad_u8::<8, 4>,
            (1, 8, 8) => return avx512::get_sad_u8::<8, 8>,
            (1, 8, 16) => return avx512::get_sad_u8::<8, 16>,
            // PERF: Equivalent to AVX2 on test machine
            (1, 16, 1) => return avx512::get_sad_u8::<16, 1>,
            (1, 16, 2) => return avx512::get_sad_u8::<16, 2>,
            (1, 16, 4) => return avx512::get_sad_u8::<16, 4>,
            (1, 16, 8) => return avx512::get_sad_u8::<16, 8>,
            (1, 16, 16) => return avx512::get_sad_u8::<16, 16>,
            (1, 16, 32) => return avx512::get_sad_u8::<16, 32>,
            // TODO: 2% slower than AVX2 on test machine
            #[cfg(feature = "experimental")]
            (1, 32, 8) => return avx512::get_sad_u8::<32, 8>,
            #[cfg(feature = "experimental")]
            (1, 32, 16) => return avx512::get_sad_u8::<32, 16>,
            #[cfg(feature = "experimental")]
            (1, 32, 32) => return avx512::get_sad_u8::<32, 32>,
            #[cfg(feature = "experimental")]
            (1, 32, 64) => return avx512::get_sad_u8::<32, 64>,
            // TODO: 10% slower than AVX2 on test machine
            #[cfg(feature = "experimental")]
            (1, 64, 16) => return avx512::get_sad_u8::<64, 16>,
            #[cfg(feature = "experimental")]
            (1, 64, 32) => return avx512::get_sad_u8::<64, 32>,
            #[cfg(feature = "experimental")]
            (1, 64, 64) => return avx512::get_sad_u8::<64, 64>,
            #[cfg(feature = "experimental")]
            (1, 64, 128) => return avx512::get_sad_u8::<64, 128>,
            // TODO: 2% slower than AVX2 on test machine
            #[cfg(feature = "experimental")]
            (1, 128, 32) => return avx512::get_sad_u8::<128, 32>,
            #[cfg(feature = "experimental")]
            (1, 128, 64) => return avx512::get_sad_u8::<128, 64>,
            #[cfg(feature = "experimental")]
            (1, 128, 128) => return avx512::get_sad_u8::<128, 128>,
            // TODO: 2% slower than scalar
            #[cfg(feature = "experimental")]
            (2, 2, 2) => return avx512::get_sad_u16::<2, 2>,
            #[cfg(feature = "experimental")]
            (2, 2, 4) => return avx512::get_sad_u16::<2, 4>,
            // PERF: Equivalent to AVX2 on test machine
            (2, 4, 2) => return avx512::get_sad_u16::<4, 2>,
            (2, 4, 4) => return avx512::get_sad_u16::<4, 4>,
            (2, 4, 8) => return avx512::get_sad_u16::<4, 8>,
            // PERF: Equivalent to AVX2 on test machine
            (2, 8, 1) => return avx512::get_sad_u16::<8, 1>,
            (2, 8, 2) => return avx512::get_sad_u16::<8, 2>,
            (2, 8, 4) => return avx512::get_sad_u16::<8, 4>,
            (2, 8, 8) => return avx512::get_sad_u16::<8, 8>,
            (2, 8, 16) => return avx512::get_sad_u16::<8, 16>,
            // TODO: 5% slower than AVX2 on test machine
            #[cfg(feature = "experimental")]
            (2, 16, 1) => return avx512::get_sad_u16::<16, 1>,
            #[cfg(feature = "experimental")]
            (2, 16, 2) => return avx512::get_sad_u16::<16, 2>,
            #[cfg(feature = "experimental")]
            (2, 16, 4) => return avx512::get_sad_u16::<16, 4>,
            #[cfg(feature = "experimental")]
            (2, 16, 8) => return avx512::get_sad_u16::<16, 8>,
            #[cfg(feature = "experimental")]
            (2, 16, 16) => return avx512::get_sad_u16::<16, 16>,
            #[cfg(feature = "experimental")]
            (2, 16, 32) => return avx512::get_sad_u16::<16, 32>,
            // PERF: 3-5% faster than AVX2 on test machine
            (2, 32, 8) => return avx512::get_sad_u16::<32, 8>,
            (2, 32, 16) => return avx512::get_sad_u16::<32, 16>,
            (2, 32, 32) => return avx512::get_sad_u16::<32, 32>,
            (2, 32, 64) => return avx512::get_sad_u16::<32, 64>,
            // PERF: Equivalent to AVX2 on test machine
            (2, 64, 16) => return avx512::get_sad_u16::<64, 16>,
            (2, 64, 32) => return avx512::get_sad_u16::<64, 32>,
            (2, 64, 64) => return avx512::get_sad_u16::<64, 64>,
            (2, 64, 128) => return avx512::get_sad_u16::<64, 128>,
            // PERF: 5% faster than AVX2 on test machine
            (2, 128, 32) => return avx512::get_sad_u16::<128, 32>,
            (2, 128, 64) => return avx512::get_sad_u16::<128, 64>,
            (2, 128, 128) => return avx512::get_sad_u16::<128, 128>,
            _ => {}
        };
    }

    #[cfg(all(target_arch = "x86_64", feature = "simd"))]
    if cpudetect::x86_64::is_x86_64_v3_compatible() {
        match (size_of::<T>(), width.get(), height.get()) {
            // TODO: 5% slower than scalar
            #[cfg(feature = "experimental")]
            (1, 2, 2) => return avx2::get_sad_u8::<2, 2>,
            #[cfg(feature = "experimental")]
            (1, 2, 4) => return avx2::get_sad_u8::<2, 4>,
            // PERF: 2% faster than scalar on test machine
            (1, 4, 2) => return avx2::get_sad_u8::<4, 2>,
            (1, 4, 4) => return avx2::get_sad_u8::<4, 4>,
            (1, 4, 8) => return avx2::get_sad_u8::<4, 8>,
            // PERF: 75% faster than scalar on test machine
            (1, 8, 1) => return avx2::get_sad_u8::<8, 1>,
            (1, 8, 2) => return avx2::get_sad_u8::<8, 2>,
            (1, 8, 4) => return avx2::get_sad_u8::<8, 4>,
            (1, 8, 8) => return avx2::get_sad_u8::<8, 8>,
            (1, 8, 16) => return avx2::get_sad_u8::<8, 16>,
            // PERF: 85% faster than scalar on test machine
            (1, 16, 1) => return avx2::get_sad_u8::<16, 1>,
            (1, 16, 2) => return avx2::get_sad_u8::<16, 2>,
            (1, 16, 4) => return avx2::get_sad_u8::<16, 4>,
            (1, 16, 8) => return avx2::get_sad_u8::<16, 8>,
            (1, 16, 16) => return avx2::get_sad_u8::<16, 16>,
            (1, 16, 32) => return avx2::get_sad_u8::<16, 32>,
            // PERF: 85% faster than scalar on test machine
            (1, 32, 8) => return avx2::get_sad_u8::<32, 8>,
            (1, 32, 16) => return avx2::get_sad_u8::<32, 16>,
            (1, 32, 32) => return avx2::get_sad_u8::<32, 32>,
            (1, 32, 64) => return avx2::get_sad_u8::<32, 64>,
            // PERF: 50-90% faster than scalar on test machine
            (1, 64, 16) => return avx2::get_sad_u8::<64, 16>,
            (1, 64, 32) => return avx2::get_sad_u8::<64, 32>,
            (1, 64, 64) => return avx2::get_sad_u8::<64, 64>,
            (1, 64, 128) => return avx2::get_sad_u8::<64, 128>,
            // PERF: 50-90% faster than scalar on test machine
            (1, 128, 32) => return avx2::get_sad_u8::<128, 32>,
            (1, 128, 64) => return avx2::get_sad_u8::<128, 64>,
            (1, 128, 128) => return avx2::get_sad_u8::<128, 128>,
            // TODO: 2% slower than scalar
            #[cfg(feature = "experimental")]
            (2, 2, 2) => return avx2::get_sad_u16::<2, 2>,
            #[cfg(feature = "experimental")]
            (2, 2, 4) => return avx2::get_sad_u16::<2, 4>,
            // PERF: 10% faster than scalar on test machine
            (2, 4, 2) => return avx2::get_sad_u16::<4, 2>,
            (2, 4, 4) => return avx2::get_sad_u16::<4, 4>,
            (2, 4, 8) => return avx2::get_sad_u16::<4, 8>,
            // PERF: 75% faster than scalar on test machine
            (2, 8, 1) => return avx2::get_sad_u16::<8, 1>,
            (2, 8, 2) => return avx2::get_sad_u16::<8, 2>,
            (2, 8, 4) => return avx2::get_sad_u16::<8, 4>,
            (2, 8, 8) => return avx2::get_sad_u16::<8, 8>,
            (2, 8, 16) => return avx2::get_sad_u16::<8, 16>,
            // PERF: 75% faster than scalar on test machine
            (2, 16, 1) => return avx2::get_sad_u16::<16, 1>,
            (2, 16, 2) => return avx2::get_sad_u16::<16, 2>,
            (2, 16, 4) => return avx2::get_sad_u16::<16, 4>,
            (2, 16, 8) => return avx2::get_sad_u16::<16, 8>,
            (2, 16, 16) => return avx2::get_sad_u16::<16, 16>,
            (2, 16, 32) => return avx2::get_sad_u16::<16, 32>,
            // PERF: 70% faster than scalar on test machine
            (2, 32, 8) => return avx2::get_sad_u16::<32, 8>,
            (2, 32, 16) => return avx2::get_sad_u16::<32, 16>,
            (2, 32, 32) => return avx2::get_sad_u16::<32, 32>,
            (2, 32, 64) => return avx2::get_sad_u16::<32, 64>,
            // PERF: 40-70% faster than scalar on test machine
            (2, 64, 16) => return avx2::get_sad_u16::<64, 16>,
            (2, 64, 32) => return avx2::get_sad_u16::<64, 32>,
            (2, 64, 64) => return avx2::get_sad_u16::<64, 64>,
            (2, 64, 128) => return avx2::get_sad_u16::<64, 128>,
            // PERF: 40-70% faster than scalar on test machine
            (2, 128, 32) => return avx2::get_sad_u16::<128, 32>,
            (2, 128, 64) => return avx2::get_sad_u16::<128, 64>,
            (2, 128, 128) => return avx2::get_sad_u16::<128, 128>,
            _ => {}
        };
    }

    match (size_of::<T>(), width.get(), height.get()) {
        (1, 2, 2) => rust::get_sad_u8::<2, 2>,
        (1, 2, 4) => rust::get_sad_u8::<2, 4>,
        (1, 4, 2) => rust::get_sad_u8::<4, 2>,
        (1, 4, 4) => rust::get_sad_u8::<4, 4>,
        (1, 4, 8) => rust::get_sad_u8::<4, 8>,
        (1, 8, 1) => rust::get_sad_u8::<8, 1>,
        (1, 8, 2) => rust::get_sad_u8::<8, 2>,
        (1, 8, 4) => rust::get_sad_u8::<8, 4>,
        (1, 8, 8) => rust::get_sad_u8::<8, 8>,
        (1, 8, 16) => rust::get_sad_u8::<8, 16>,
        (1, 16, 1) => rust::get_sad_u8::<16, 1>,
        (1, 16, 2) => rust::get_sad_u8::<16, 2>,
        (1, 16, 4) => rust::get_sad_u8::<16, 4>,
        (1, 16, 8) => rust::get_sad_u8::<16, 8>,
        (1, 16, 16) => rust::get_sad_u8::<16, 16>,
        (1, 16, 32) => rust::get_sad_u8::<16, 32>,
        (1, 32, 8) => rust::get_sad_u8::<32, 8>,
        (1, 32, 16) => rust::get_sad_u8::<32, 16>,
        (1, 32, 32) => rust::get_sad_u8::<32, 32>,
        (1, 32, 64) => rust::get_sad_u8::<32, 64>,
        (1, 64, 16) => rust::get_sad_u8::<64, 16>,
        (1, 64, 32) => rust::get_sad_u8::<64, 32>,
        (1, 64, 64) => rust::get_sad_u8::<64, 64>,
        (1, 64, 128) => rust::get_sad_u8::<64, 128>,
        (1, 128, 32) => rust::get_sad_u8::<128, 32>,
        (1, 128, 64) => rust::get_sad_u8::<128, 64>,
        (1, 128, 128) => rust::get_sad_u8::<128, 128>,
        (2, 2, 2) => rust::get_sad_u16::<2, 2>,
        (2, 2, 4) => rust::get_sad_u16::<2, 4>,
        (2, 4, 2) => rust::get_sad_u16::<4, 2>,
        (2, 4, 4) => rust::get_sad_u16::<4, 4>,
        (2, 4, 8) => rust::get_sad_u16::<4, 8>,
        (2, 8, 1) => rust::get_sad_u16::<8, 1>,
        (2, 8, 2) => rust::get_sad_u16::<8, 2>,
        (2, 8, 4) => rust::get_sad_u16::<8, 4>,
        (2, 8, 8) => rust::get_sad_u16::<8, 8>,
        (2, 8, 16) => rust::get_sad_u16::<8, 16>,
        (2, 16, 1) => rust::get_sad_u16::<16, 1>,
        (2, 16, 2) => rust::get_sad_u16::<16, 2>,
        (2, 16, 4) => rust::get_sad_u16::<16, 4>,
        (2, 16, 8) => rust::get_sad_u16::<16, 8>,
        (2, 16, 16) => rust::get_sad_u16::<16, 16>,
        (2, 16, 32) => rust::get_sad_u16::<16, 32>,
        (2, 32, 8) => rust::get_sad_u16::<32, 8>,
        (2, 32, 16) => rust::get_sad_u16::<32, 16>,
        (2, 32, 32) => rust::get_sad_u16::<32, 32>,
        (2, 32, 64) => rust::get_sad_u16::<32, 64>,
        (2, 64, 16) => rust::get_sad_u16::<64, 16>,
        (2, 64, 32) => rust::get_sad_u16::<64, 32>,
        (2, 64, 64) => rust::get_sad_u16::<64, 64>,
        (2, 64, 128) => rust::get_sad_u16::<64, 128>,
        (2, 128, 32) => rust::get_sad_u16::<128, 32>,
        (2, 128, 64) => rust::get_sad_u16::<128, 64>,
        (2, 128, 128) => rust::get_sad_u16::<128, 128>,
        _ => panic!("unsupported block size for SAD: {}x{}", width, height),
    }
}