/src/libavif/ext/libyuv/source/convert.cc
Line | Count | Source |
1 | | /* |
2 | | * Copyright 2011 The LibYuv Project Authors. All rights reserved. |
3 | | * |
4 | | * Use of this source code is governed by a BSD-style license |
5 | | * that can be found in the LICENSE file in the root of the source |
6 | | * tree. An additional intellectual property rights grant can be found |
7 | | * in the file PATENTS. All contributing project authors may |
8 | | * be found in the AUTHORS file in the root of the source tree. |
9 | | */ |
10 | | |
11 | | #include "libyuv/convert.h" |
12 | | |
13 | | #include "libyuv/basic_types.h" |
14 | | #include "libyuv/cpu_id.h" |
15 | | #include "libyuv/planar_functions.h" |
16 | | #include "libyuv/rotate.h" |
17 | | #include "libyuv/row.h" |
18 | | #include "libyuv/scale.h" // For ScalePlane() |
19 | | #include "libyuv/scale_row.h" // For FixedDiv |
20 | | #include "libyuv/scale_uv.h" // For UVScale() |
21 | | |
22 | | #ifdef __cplusplus |
23 | | namespace libyuv { |
24 | | extern "C" { |
25 | | #endif |
26 | | |
27 | | // Subsample amount uses a shift. |
28 | | // v is value |
29 | | // a is amount to add to round up |
30 | | // s is shift to subsample down |
31 | 0 | #define SUBSAMPLE(v, a, s) (v < 0) ? (-((-v + a) >> s)) : ((v + a) >> s) |
32 | 0 | static __inline int Abs(int v) { |
33 | 0 | return v >= 0 ? v : -v; |
34 | 0 | } |
35 | | |
36 | | // Any I4xx To I420 format |
37 | | static int I4xxToI420(const uint8_t* src_y, |
38 | | int src_stride_y, |
39 | | const uint8_t* src_u, |
40 | | int src_stride_u, |
41 | | const uint8_t* src_v, |
42 | | int src_stride_v, |
43 | | uint8_t* dst_y, |
44 | | int dst_stride_y, |
45 | | uint8_t* dst_u, |
46 | | int dst_stride_u, |
47 | | uint8_t* dst_v, |
48 | | int dst_stride_v, |
49 | | int src_y_width, |
50 | | int src_y_height, |
51 | | int src_uv_width, |
52 | 0 | int src_uv_height) { |
53 | 0 | const int dst_y_width = src_y_width; |
54 | 0 | const int dst_y_height = Abs(src_y_height); |
55 | 0 | const int dst_uv_width = SUBSAMPLE(dst_y_width, 1, 1); |
56 | 0 | const int dst_uv_height = SUBSAMPLE(dst_y_height, 1, 1); |
57 | 0 | int r; |
58 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || |
59 | 0 | src_y_width <= 0 || src_y_height == 0 || src_uv_width <= 0 || |
60 | 0 | src_uv_height == 0) { |
61 | 0 | return -1; |
62 | 0 | } |
63 | 0 | if (dst_y) { |
64 | 0 | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, src_y_width, |
65 | 0 | src_y_height); |
66 | 0 | } |
67 | 0 | r = ScalePlane(src_u, src_stride_u, src_uv_width, src_uv_height, dst_u, |
68 | 0 | dst_stride_u, dst_uv_width, dst_uv_height, kFilterBilinear); |
69 | 0 | if (r != 0) { |
70 | 0 | return r; |
71 | 0 | } |
72 | 0 | r = ScalePlane(src_v, src_stride_v, src_uv_width, src_uv_height, dst_v, |
73 | 0 | dst_stride_v, dst_uv_width, dst_uv_height, kFilterBilinear); |
74 | 0 | return r; |
75 | 0 | } |
76 | | |
77 | | // Copy I420 with optional vertical flipping using negative height. |
78 | | LIBYUV_API |
79 | | int I420Copy(const uint8_t* src_y, |
80 | | int src_stride_y, |
81 | | const uint8_t* src_u, |
82 | | int src_stride_u, |
83 | | const uint8_t* src_v, |
84 | | int src_stride_v, |
85 | | uint8_t* dst_y, |
86 | | int dst_stride_y, |
87 | | uint8_t* dst_u, |
88 | | int dst_stride_u, |
89 | | uint8_t* dst_v, |
90 | | int dst_stride_v, |
91 | | int width, |
92 | 0 | int height) { |
93 | 0 | int halfwidth = (width + 1) >> 1; |
94 | 0 | int halfheight = (height + 1) >> 1; |
95 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 || |
96 | 0 | height == 0) { |
97 | 0 | return -1; |
98 | 0 | } |
99 | | // Negative height means invert the image. |
100 | 0 | if (height < 0) { |
101 | 0 | height = -height; |
102 | 0 | halfheight = (height + 1) >> 1; |
103 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
104 | 0 | src_u = src_u + (halfheight - 1) * src_stride_u; |
105 | 0 | src_v = src_v + (halfheight - 1) * src_stride_v; |
106 | 0 | src_stride_y = -src_stride_y; |
107 | 0 | src_stride_u = -src_stride_u; |
108 | 0 | src_stride_v = -src_stride_v; |
109 | 0 | } |
110 | |
|
111 | 0 | if (dst_y) { |
112 | 0 | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
113 | 0 | } |
114 | | // Copy UV planes. |
115 | 0 | CopyPlane(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, halfheight); |
116 | 0 | CopyPlane(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, halfheight); |
117 | 0 | return 0; |
118 | 0 | } |
119 | | |
120 | | // Copy I010 with optional flipping. |
121 | | LIBYUV_API |
122 | | int I010Copy(const uint16_t* src_y, |
123 | | int src_stride_y, |
124 | | const uint16_t* src_u, |
125 | | int src_stride_u, |
126 | | const uint16_t* src_v, |
127 | | int src_stride_v, |
128 | | uint16_t* dst_y, |
129 | | int dst_stride_y, |
130 | | uint16_t* dst_u, |
131 | | int dst_stride_u, |
132 | | uint16_t* dst_v, |
133 | | int dst_stride_v, |
134 | | int width, |
135 | 0 | int height) { |
136 | 0 | int halfwidth = (width + 1) >> 1; |
137 | 0 | int halfheight = (height + 1) >> 1; |
138 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 || |
139 | 0 | height == 0) { |
140 | 0 | return -1; |
141 | 0 | } |
142 | | // Negative height means invert the image. |
143 | 0 | if (height < 0) { |
144 | 0 | height = -height; |
145 | 0 | halfheight = (height + 1) >> 1; |
146 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
147 | 0 | src_u = src_u + (halfheight - 1) * src_stride_u; |
148 | 0 | src_v = src_v + (halfheight - 1) * src_stride_v; |
149 | 0 | src_stride_y = -src_stride_y; |
150 | 0 | src_stride_u = -src_stride_u; |
151 | 0 | src_stride_v = -src_stride_v; |
152 | 0 | } |
153 | |
|
154 | 0 | if (dst_y) { |
155 | 0 | CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
156 | 0 | } |
157 | | // Copy UV planes. |
158 | 0 | CopyPlane_16(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, halfheight); |
159 | 0 | CopyPlane_16(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, halfheight); |
160 | 0 | return 0; |
161 | 0 | } |
162 | | |
163 | | static int Planar16bitTo8bit(const uint16_t* src_y, |
164 | | int src_stride_y, |
165 | | const uint16_t* src_u, |
166 | | int src_stride_u, |
167 | | const uint16_t* src_v, |
168 | | int src_stride_v, |
169 | | uint8_t* dst_y, |
170 | | int dst_stride_y, |
171 | | uint8_t* dst_u, |
172 | | int dst_stride_u, |
173 | | uint8_t* dst_v, |
174 | | int dst_stride_v, |
175 | | int width, |
176 | | int height, |
177 | | int subsample_x, |
178 | | int subsample_y, |
179 | 0 | int depth) { |
180 | 0 | int uv_width = SUBSAMPLE(width, subsample_x, subsample_x); |
181 | 0 | int uv_height = SUBSAMPLE(height, subsample_y, subsample_y); |
182 | 0 | int scale = 1 << (24 - depth); |
183 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 || |
184 | 0 | height == 0) { |
185 | 0 | return -1; |
186 | 0 | } |
187 | | // Negative height means invert the image. |
188 | 0 | if (height < 0) { |
189 | 0 | height = -height; |
190 | 0 | uv_height = -uv_height; |
191 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
192 | 0 | src_u = src_u + (uv_height - 1) * src_stride_u; |
193 | 0 | src_v = src_v + (uv_height - 1) * src_stride_v; |
194 | 0 | src_stride_y = -src_stride_y; |
195 | 0 | src_stride_u = -src_stride_u; |
196 | 0 | src_stride_v = -src_stride_v; |
197 | 0 | } |
198 | | |
199 | | // Convert Y plane. |
200 | 0 | if (dst_y) { |
201 | 0 | Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width, |
202 | 0 | height); |
203 | 0 | } |
204 | | // Convert UV planes. |
205 | 0 | Convert16To8Plane(src_u, src_stride_u, dst_u, dst_stride_u, scale, uv_width, |
206 | 0 | uv_height); |
207 | 0 | Convert16To8Plane(src_v, src_stride_v, dst_v, dst_stride_v, scale, uv_width, |
208 | 0 | uv_height); |
209 | 0 | return 0; |
210 | 0 | } |
211 | | |
212 | | static int I41xToI420(const uint16_t* src_y, |
213 | | int src_stride_y, |
214 | | const uint16_t* src_u, |
215 | | int src_stride_u, |
216 | | const uint16_t* src_v, |
217 | | int src_stride_v, |
218 | | uint8_t* dst_y, |
219 | | int dst_stride_y, |
220 | | uint8_t* dst_u, |
221 | | int dst_stride_u, |
222 | | uint8_t* dst_v, |
223 | | int dst_stride_v, |
224 | | int width, |
225 | | int height, |
226 | 0 | int depth) { |
227 | 0 | const int scale = 1 << (24 - depth); |
228 | |
|
229 | 0 | if (width <= 0 || height == 0) { |
230 | 0 | return -1; |
231 | 0 | } |
232 | | // Negative height means invert the image. |
233 | 0 | if (height < 0) { |
234 | 0 | height = -height; |
235 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
236 | 0 | src_u = src_u + (height - 1) * src_stride_u; |
237 | 0 | src_v = src_v + (height - 1) * src_stride_v; |
238 | 0 | src_stride_y = -src_stride_y; |
239 | 0 | src_stride_u = -src_stride_u; |
240 | 0 | src_stride_v = -src_stride_v; |
241 | 0 | } |
242 | |
|
243 | 0 | { |
244 | 0 | const int uv_width = SUBSAMPLE(width, 1, 1); |
245 | 0 | const int uv_height = SUBSAMPLE(height, 1, 1); |
246 | |
|
247 | 0 | Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width, |
248 | 0 | height); |
249 | 0 | ScalePlaneDown2_16To8(width, height, uv_width, uv_height, src_stride_u, |
250 | 0 | dst_stride_u, src_u, dst_u, scale, kFilterBilinear); |
251 | 0 | ScalePlaneDown2_16To8(width, height, uv_width, uv_height, src_stride_v, |
252 | 0 | dst_stride_v, src_v, dst_v, scale, kFilterBilinear); |
253 | 0 | } |
254 | 0 | return 0; |
255 | 0 | } |
256 | | |
257 | | static int I21xToI420(const uint16_t* src_y, |
258 | | int src_stride_y, |
259 | | const uint16_t* src_u, |
260 | | int src_stride_u, |
261 | | const uint16_t* src_v, |
262 | | int src_stride_v, |
263 | | uint8_t* dst_y, |
264 | | int dst_stride_y, |
265 | | uint8_t* dst_u, |
266 | | int dst_stride_u, |
267 | | uint8_t* dst_v, |
268 | | int dst_stride_v, |
269 | | int width, |
270 | | int height, |
271 | 0 | int depth) { |
272 | 0 | const int scale = 1 << (24 - depth); |
273 | |
|
274 | 0 | if (width <= 0 || height == 0) { |
275 | 0 | return -1; |
276 | 0 | } |
277 | | // Negative height means invert the image. |
278 | 0 | if (height < 0) { |
279 | 0 | height = -height; |
280 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
281 | 0 | src_u = src_u + (height - 1) * src_stride_u; |
282 | 0 | src_v = src_v + (height - 1) * src_stride_v; |
283 | 0 | src_stride_y = -src_stride_y; |
284 | 0 | src_stride_u = -src_stride_u; |
285 | 0 | src_stride_v = -src_stride_v; |
286 | 0 | } |
287 | |
|
288 | 0 | { |
289 | 0 | const int uv_width = SUBSAMPLE(width, 1, 1); |
290 | 0 | const int uv_height = SUBSAMPLE(height, 1, 1); |
291 | 0 | const int dy = FixedDiv(height, uv_height); |
292 | |
|
293 | 0 | Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width, |
294 | 0 | height); |
295 | 0 | ScalePlaneVertical_16To8(height, uv_width, uv_height, src_stride_u, |
296 | 0 | dst_stride_u, src_u, dst_u, 0, 32768, dy, |
297 | 0 | /*bpp=*/1, scale, kFilterBilinear); |
298 | 0 | ScalePlaneVertical_16To8(height, uv_width, uv_height, src_stride_v, |
299 | 0 | dst_stride_v, src_v, dst_v, 0, 32768, dy, |
300 | 0 | /*bpp=*/1, scale, kFilterBilinear); |
301 | 0 | } |
302 | 0 | return 0; |
303 | 0 | } |
304 | | |
305 | | // Convert 10 bit YUV to 8 bit. |
306 | | LIBYUV_API |
307 | | int I010ToI420(const uint16_t* src_y, |
308 | | int src_stride_y, |
309 | | const uint16_t* src_u, |
310 | | int src_stride_u, |
311 | | const uint16_t* src_v, |
312 | | int src_stride_v, |
313 | | uint8_t* dst_y, |
314 | | int dst_stride_y, |
315 | | uint8_t* dst_u, |
316 | | int dst_stride_u, |
317 | | uint8_t* dst_v, |
318 | | int dst_stride_v, |
319 | | int width, |
320 | 0 | int height) { |
321 | 0 | return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v, |
322 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, |
323 | 0 | dst_stride_u, dst_v, dst_stride_v, width, height, 1, |
324 | 0 | 1, 10); |
325 | 0 | } |
326 | | |
327 | | LIBYUV_API |
328 | | int I210ToI420(const uint16_t* src_y, |
329 | | int src_stride_y, |
330 | | const uint16_t* src_u, |
331 | | int src_stride_u, |
332 | | const uint16_t* src_v, |
333 | | int src_stride_v, |
334 | | uint8_t* dst_y, |
335 | | int dst_stride_y, |
336 | | uint8_t* dst_u, |
337 | | int dst_stride_u, |
338 | | uint8_t* dst_v, |
339 | | int dst_stride_v, |
340 | | int width, |
341 | 0 | int height) { |
342 | 0 | return I21xToI420(src_y, src_stride_y, src_u, src_stride_u, src_v, |
343 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u, |
344 | 0 | dst_v, dst_stride_v, width, height, 10); |
345 | 0 | } |
346 | | |
347 | | LIBYUV_API |
348 | | int I210ToI422(const uint16_t* src_y, |
349 | | int src_stride_y, |
350 | | const uint16_t* src_u, |
351 | | int src_stride_u, |
352 | | const uint16_t* src_v, |
353 | | int src_stride_v, |
354 | | uint8_t* dst_y, |
355 | | int dst_stride_y, |
356 | | uint8_t* dst_u, |
357 | | int dst_stride_u, |
358 | | uint8_t* dst_v, |
359 | | int dst_stride_v, |
360 | | int width, |
361 | 0 | int height) { |
362 | 0 | return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v, |
363 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, |
364 | 0 | dst_stride_u, dst_v, dst_stride_v, width, height, 1, |
365 | 0 | 0, 10); |
366 | 0 | } |
367 | | |
368 | | LIBYUV_API |
369 | | int I410ToI420(const uint16_t* src_y, |
370 | | int src_stride_y, |
371 | | const uint16_t* src_u, |
372 | | int src_stride_u, |
373 | | const uint16_t* src_v, |
374 | | int src_stride_v, |
375 | | uint8_t* dst_y, |
376 | | int dst_stride_y, |
377 | | uint8_t* dst_u, |
378 | | int dst_stride_u, |
379 | | uint8_t* dst_v, |
380 | | int dst_stride_v, |
381 | | int width, |
382 | 0 | int height) { |
383 | 0 | return I41xToI420(src_y, src_stride_y, src_u, src_stride_u, src_v, |
384 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u, |
385 | 0 | dst_v, dst_stride_v, width, height, 10); |
386 | 0 | } |
387 | | |
388 | | LIBYUV_API |
389 | | int I410ToI444(const uint16_t* src_y, |
390 | | int src_stride_y, |
391 | | const uint16_t* src_u, |
392 | | int src_stride_u, |
393 | | const uint16_t* src_v, |
394 | | int src_stride_v, |
395 | | uint8_t* dst_y, |
396 | | int dst_stride_y, |
397 | | uint8_t* dst_u, |
398 | | int dst_stride_u, |
399 | | uint8_t* dst_v, |
400 | | int dst_stride_v, |
401 | | int width, |
402 | 0 | int height) { |
403 | 0 | return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v, |
404 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, |
405 | 0 | dst_stride_u, dst_v, dst_stride_v, width, height, 0, |
406 | 0 | 0, 10); |
407 | 0 | } |
408 | | |
409 | | LIBYUV_API |
410 | | int I012ToI420(const uint16_t* src_y, |
411 | | int src_stride_y, |
412 | | const uint16_t* src_u, |
413 | | int src_stride_u, |
414 | | const uint16_t* src_v, |
415 | | int src_stride_v, |
416 | | uint8_t* dst_y, |
417 | | int dst_stride_y, |
418 | | uint8_t* dst_u, |
419 | | int dst_stride_u, |
420 | | uint8_t* dst_v, |
421 | | int dst_stride_v, |
422 | | int width, |
423 | 0 | int height) { |
424 | 0 | return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v, |
425 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, |
426 | 0 | dst_stride_u, dst_v, dst_stride_v, width, height, 1, |
427 | 0 | 1, 12); |
428 | 0 | } |
429 | | |
430 | | LIBYUV_API |
431 | | int I212ToI422(const uint16_t* src_y, |
432 | | int src_stride_y, |
433 | | const uint16_t* src_u, |
434 | | int src_stride_u, |
435 | | const uint16_t* src_v, |
436 | | int src_stride_v, |
437 | | uint8_t* dst_y, |
438 | | int dst_stride_y, |
439 | | uint8_t* dst_u, |
440 | | int dst_stride_u, |
441 | | uint8_t* dst_v, |
442 | | int dst_stride_v, |
443 | | int width, |
444 | 0 | int height) { |
445 | 0 | return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v, |
446 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, |
447 | 0 | dst_stride_u, dst_v, dst_stride_v, width, height, 1, |
448 | 0 | 0, 12); |
449 | 0 | } |
450 | | |
451 | | LIBYUV_API |
452 | | int I212ToI420(const uint16_t* src_y, |
453 | | int src_stride_y, |
454 | | const uint16_t* src_u, |
455 | | int src_stride_u, |
456 | | const uint16_t* src_v, |
457 | | int src_stride_v, |
458 | | uint8_t* dst_y, |
459 | | int dst_stride_y, |
460 | | uint8_t* dst_u, |
461 | | int dst_stride_u, |
462 | | uint8_t* dst_v, |
463 | | int dst_stride_v, |
464 | | int width, |
465 | 0 | int height) { |
466 | 0 | return I21xToI420(src_y, src_stride_y, src_u, src_stride_u, src_v, |
467 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u, |
468 | 0 | dst_v, dst_stride_v, width, height, 12); |
469 | 0 | } |
470 | | |
471 | | LIBYUV_API |
472 | | int I412ToI444(const uint16_t* src_y, |
473 | | int src_stride_y, |
474 | | const uint16_t* src_u, |
475 | | int src_stride_u, |
476 | | const uint16_t* src_v, |
477 | | int src_stride_v, |
478 | | uint8_t* dst_y, |
479 | | int dst_stride_y, |
480 | | uint8_t* dst_u, |
481 | | int dst_stride_u, |
482 | | uint8_t* dst_v, |
483 | | int dst_stride_v, |
484 | | int width, |
485 | 0 | int height) { |
486 | 0 | return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v, |
487 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, |
488 | 0 | dst_stride_u, dst_v, dst_stride_v, width, height, 0, |
489 | 0 | 0, 12); |
490 | 0 | } |
491 | | |
492 | | LIBYUV_API |
493 | | int I412ToI420(const uint16_t* src_y, |
494 | | int src_stride_y, |
495 | | const uint16_t* src_u, |
496 | | int src_stride_u, |
497 | | const uint16_t* src_v, |
498 | | int src_stride_v, |
499 | | uint8_t* dst_y, |
500 | | int dst_stride_y, |
501 | | uint8_t* dst_u, |
502 | | int dst_stride_u, |
503 | | uint8_t* dst_v, |
504 | | int dst_stride_v, |
505 | | int width, |
506 | 0 | int height) { |
507 | 0 | return I41xToI420(src_y, src_stride_y, src_u, src_stride_u, src_v, |
508 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u, |
509 | 0 | dst_v, dst_stride_v, width, height, 12); |
510 | 0 | } |
511 | | |
512 | | // Any Ix10 To I010 format |
513 | | static int Ix10ToI010(const uint16_t* src_y, |
514 | | int src_stride_y, |
515 | | const uint16_t* src_u, |
516 | | int src_stride_u, |
517 | | const uint16_t* src_v, |
518 | | int src_stride_v, |
519 | | uint16_t* dst_y, |
520 | | int dst_stride_y, |
521 | | uint16_t* dst_u, |
522 | | int dst_stride_u, |
523 | | uint16_t* dst_v, |
524 | | int dst_stride_v, |
525 | | int width, |
526 | | int height, |
527 | | int subsample_x, |
528 | 0 | int subsample_y) { |
529 | 0 | const int dst_y_width = width; |
530 | 0 | const int dst_y_height = Abs(height); |
531 | 0 | const int src_uv_width = SUBSAMPLE(width, subsample_x, subsample_x); |
532 | 0 | const int src_uv_height = SUBSAMPLE(height, subsample_y, subsample_y); |
533 | 0 | const int dst_uv_width = SUBSAMPLE(dst_y_width, 1, 1); |
534 | 0 | const int dst_uv_height = SUBSAMPLE(dst_y_height, 1, 1); |
535 | 0 | int r; |
536 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 || |
537 | 0 | height == 0) { |
538 | 0 | return -1; |
539 | 0 | } |
540 | 0 | if (dst_y) { |
541 | 0 | CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
542 | 0 | } |
543 | 0 | r = ScalePlane_12(src_u, src_stride_u, src_uv_width, src_uv_height, dst_u, |
544 | 0 | dst_stride_u, dst_uv_width, dst_uv_height, kFilterBilinear); |
545 | 0 | if (r != 0) { |
546 | 0 | return r; |
547 | 0 | } |
548 | 0 | r = ScalePlane_12(src_v, src_stride_v, src_uv_width, src_uv_height, dst_v, |
549 | 0 | dst_stride_v, dst_uv_width, dst_uv_height, kFilterBilinear); |
550 | 0 | return r; |
551 | 0 | } |
552 | | |
553 | | LIBYUV_API |
554 | | int I410ToI010(const uint16_t* src_y, |
555 | | int src_stride_y, |
556 | | const uint16_t* src_u, |
557 | | int src_stride_u, |
558 | | const uint16_t* src_v, |
559 | | int src_stride_v, |
560 | | uint16_t* dst_y, |
561 | | int dst_stride_y, |
562 | | uint16_t* dst_u, |
563 | | int dst_stride_u, |
564 | | uint16_t* dst_v, |
565 | | int dst_stride_v, |
566 | | int width, |
567 | 0 | int height) { |
568 | 0 | return Ix10ToI010(src_y, src_stride_y, src_u, src_stride_u, src_v, |
569 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u, |
570 | 0 | dst_v, dst_stride_v, width, height, 0, 0); |
571 | 0 | } |
572 | | |
573 | | LIBYUV_API |
574 | | int I210ToI010(const uint16_t* src_y, |
575 | | int src_stride_y, |
576 | | const uint16_t* src_u, |
577 | | int src_stride_u, |
578 | | const uint16_t* src_v, |
579 | | int src_stride_v, |
580 | | uint16_t* dst_y, |
581 | | int dst_stride_y, |
582 | | uint16_t* dst_u, |
583 | | int dst_stride_u, |
584 | | uint16_t* dst_v, |
585 | | int dst_stride_v, |
586 | | int width, |
587 | 0 | int height) { |
588 | 0 | return Ix10ToI010(src_y, src_stride_y, src_u, src_stride_u, src_v, |
589 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u, |
590 | 0 | dst_v, dst_stride_v, width, height, 1, 0); |
591 | 0 | } |
592 | | |
593 | | // Any I[420]1[02] to P[420]1[02] format |
594 | | static int IxxxToPxxx(const uint16_t* src_y, |
595 | | int src_stride_y, |
596 | | const uint16_t* src_u, |
597 | | int src_stride_u, |
598 | | const uint16_t* src_v, |
599 | | int src_stride_v, |
600 | | uint16_t* dst_y, |
601 | | int dst_stride_y, |
602 | | uint16_t* dst_uv, |
603 | | int dst_stride_uv, |
604 | | int width, |
605 | | int height, |
606 | | int subsample_x, |
607 | | int subsample_y, |
608 | 0 | int depth) { |
609 | 0 | const int uv_width = SUBSAMPLE(width, subsample_x, subsample_x); |
610 | 0 | const int uv_height = SUBSAMPLE(height, subsample_y, subsample_y); |
611 | 0 | if (width <= 0 || height == 0) { |
612 | 0 | return -1; |
613 | 0 | } |
614 | | |
615 | 0 | ConvertToMSBPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height, |
616 | 0 | depth); |
617 | 0 | MergeUVPlane_16(src_u, src_stride_u, src_v, src_stride_v, dst_uv, |
618 | 0 | dst_stride_uv, uv_width, uv_height, depth); |
619 | 0 | return 0; |
620 | 0 | } |
621 | | |
622 | | LIBYUV_API |
623 | | int I010ToP010(const uint16_t* src_y, |
624 | | int src_stride_y, |
625 | | const uint16_t* src_u, |
626 | | int src_stride_u, |
627 | | const uint16_t* src_v, |
628 | | int src_stride_v, |
629 | | uint16_t* dst_y, |
630 | | int dst_stride_y, |
631 | | uint16_t* dst_uv, |
632 | | int dst_stride_uv, |
633 | | int width, |
634 | 0 | int height) { |
635 | 0 | return IxxxToPxxx(src_y, src_stride_y, src_u, src_stride_u, src_v, |
636 | 0 | src_stride_v, dst_y, dst_stride_y, dst_uv, dst_stride_uv, |
637 | 0 | width, height, 1, 1, 10); |
638 | 0 | } |
639 | | |
640 | | LIBYUV_API |
641 | | int I010ToNV12(const uint16_t* src_y, |
642 | | int src_stride_y, |
643 | | const uint16_t* src_u, |
644 | | int src_stride_u, |
645 | | const uint16_t* src_v, |
646 | | int src_stride_v, |
647 | | uint8_t* dst_y, |
648 | | int dst_stride_y, |
649 | | uint8_t* dst_uv, |
650 | | int dst_stride_uv, |
651 | | int width, |
652 | 0 | int height) { |
653 | 0 | int y; |
654 | 0 | int halfwidth = (width + 1) >> 1; |
655 | 0 | int halfheight = (height + 1) >> 1; |
656 | 0 | const int scale = 16385; // 16384 for 10 bits |
657 | 0 | void (*Convert16To8Row)(const uint16_t* src_y, uint8_t* dst_y, int scale, |
658 | 0 | int width) = Convert16To8Row_C; |
659 | 0 | void (*MergeUVRow)(const uint8_t* src_u, const uint8_t* src_v, |
660 | 0 | uint8_t* dst_uv, int width) = MergeUVRow_C; |
661 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_uv || width <= 0 || |
662 | 0 | height == 0) { |
663 | 0 | return -1; |
664 | 0 | } |
665 | | // Negative height means invert the image. |
666 | 0 | if (height < 0) { |
667 | 0 | height = -height; |
668 | 0 | halfheight = (height + 1) >> 1; |
669 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
670 | 0 | src_u = src_u + (halfheight - 1) * src_stride_u; |
671 | 0 | src_v = src_v + (halfheight - 1) * src_stride_v; |
672 | 0 | src_stride_y = -src_stride_y; |
673 | 0 | src_stride_u = -src_stride_u; |
674 | 0 | src_stride_v = -src_stride_v; |
675 | 0 | } |
676 | | #if defined(HAS_CONVERT16TO8ROW_NEON) |
677 | | if (TestCpuFlag(kCpuHasNEON)) { |
678 | | Convert16To8Row = Convert16To8Row_Any_NEON; |
679 | | if (IS_ALIGNED(width, 16)) { |
680 | | Convert16To8Row = Convert16To8Row_NEON; |
681 | | } |
682 | | } |
683 | | #endif |
684 | | #if defined(HAS_CONVERT16TO8ROW_SME) |
685 | | if (TestCpuFlag(kCpuHasSME)) { |
686 | | Convert16To8Row = Convert16To8Row_SME; |
687 | | } |
688 | | #endif |
689 | 0 | #if defined(HAS_CONVERT16TO8ROW_SSSE3) |
690 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
691 | 0 | Convert16To8Row = Convert16To8Row_Any_SSSE3; |
692 | 0 | if (IS_ALIGNED(width, 16)) { |
693 | 0 | Convert16To8Row = Convert16To8Row_SSSE3; |
694 | 0 | } |
695 | 0 | } |
696 | 0 | #endif |
697 | 0 | #if defined(HAS_CONVERT16TO8ROW_AVX2) |
698 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
699 | 0 | Convert16To8Row = Convert16To8Row_Any_AVX2; |
700 | 0 | if (IS_ALIGNED(width, 32)) { |
701 | 0 | Convert16To8Row = Convert16To8Row_AVX2; |
702 | 0 | } |
703 | 0 | } |
704 | 0 | #endif |
705 | 0 | #if defined(HAS_CONVERT16TO8ROW_AVX512BW) |
706 | 0 | if (TestCpuFlag(kCpuHasAVX512BW)) { |
707 | 0 | Convert16To8Row = Convert16To8Row_Any_AVX512BW; |
708 | 0 | if (IS_ALIGNED(width, 64)) { |
709 | 0 | Convert16To8Row = Convert16To8Row_AVX512BW; |
710 | 0 | } |
711 | 0 | } |
712 | 0 | #endif |
713 | |
|
714 | 0 | #if defined(HAS_MERGEUVROW_SSE2) |
715 | 0 | if (TestCpuFlag(kCpuHasSSE2)) { |
716 | 0 | MergeUVRow = MergeUVRow_Any_SSE2; |
717 | 0 | if (IS_ALIGNED(halfwidth, 16)) { |
718 | 0 | MergeUVRow = MergeUVRow_SSE2; |
719 | 0 | } |
720 | 0 | } |
721 | 0 | #endif |
722 | 0 | #if defined(HAS_MERGEUVROW_AVX2) |
723 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
724 | 0 | MergeUVRow = MergeUVRow_Any_AVX2; |
725 | 0 | if (IS_ALIGNED(halfwidth, 16)) { |
726 | 0 | MergeUVRow = MergeUVRow_AVX2; |
727 | 0 | } |
728 | 0 | } |
729 | 0 | #endif |
730 | 0 | #if defined(HAS_MERGEUVROW_AVX512BW) |
731 | 0 | if (TestCpuFlag(kCpuHasAVX512BW)) { |
732 | 0 | MergeUVRow = MergeUVRow_Any_AVX512BW; |
733 | 0 | if (IS_ALIGNED(halfwidth, 32)) { |
734 | 0 | MergeUVRow = MergeUVRow_AVX512BW; |
735 | 0 | } |
736 | 0 | } |
737 | 0 | #endif |
738 | | #if defined(HAS_MERGEUVROW_NEON) |
739 | | if (TestCpuFlag(kCpuHasNEON)) { |
740 | | MergeUVRow = MergeUVRow_Any_NEON; |
741 | | if (IS_ALIGNED(halfwidth, 16)) { |
742 | | MergeUVRow = MergeUVRow_NEON; |
743 | | } |
744 | | } |
745 | | #endif |
746 | | #if defined(HAS_MERGEUVROW_SME) |
747 | | if (TestCpuFlag(kCpuHasSME)) { |
748 | | MergeUVRow = MergeUVRow_SME; |
749 | | } |
750 | | #endif |
751 | | #if defined(HAS_MERGEUVROW_LSX) |
752 | | if (TestCpuFlag(kCpuHasLSX)) { |
753 | | MergeUVRow = MergeUVRow_Any_LSX; |
754 | | if (IS_ALIGNED(halfwidth, 16)) { |
755 | | MergeUVRow = MergeUVRow_LSX; |
756 | | } |
757 | | } |
758 | | #endif |
759 | | #if defined(HAS_MERGEUVROW_RVV) |
760 | | if (TestCpuFlag(kCpuHasRVV)) { |
761 | | MergeUVRow = MergeUVRow_RVV; |
762 | | } |
763 | | #endif |
764 | | |
765 | | // Convert Y plane. |
766 | 0 | if (dst_y) { |
767 | 0 | Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width, |
768 | 0 | height); |
769 | 0 | } |
770 | |
|
771 | 0 | { |
772 | | // Allocate a row of uv. |
773 | 0 | align_buffer_64(row_u, ((halfwidth + 31) & ~31) * 2); |
774 | 0 | uint8_t* row_v = row_u + ((halfwidth + 31) & ~31); |
775 | 0 | if (!row_u) |
776 | 0 | return 1; |
777 | | |
778 | 0 | for (y = 0; y < halfheight; ++y) { |
779 | 0 | Convert16To8Row(src_u, row_u, scale, halfwidth); |
780 | 0 | Convert16To8Row(src_v, row_v, scale, halfwidth); |
781 | 0 | MergeUVRow(row_u, row_v, dst_uv, halfwidth); |
782 | 0 | src_u += src_stride_u; |
783 | 0 | src_v += src_stride_v; |
784 | 0 | dst_uv += dst_stride_uv; |
785 | 0 | } |
786 | 0 | free_aligned_buffer_64(row_u); |
787 | 0 | } |
788 | 0 | return 0; |
789 | 0 | } |
790 | | |
791 | | LIBYUV_API |
792 | | int I210ToP210(const uint16_t* src_y, |
793 | | int src_stride_y, |
794 | | const uint16_t* src_u, |
795 | | int src_stride_u, |
796 | | const uint16_t* src_v, |
797 | | int src_stride_v, |
798 | | uint16_t* dst_y, |
799 | | int dst_stride_y, |
800 | | uint16_t* dst_uv, |
801 | | int dst_stride_uv, |
802 | | int width, |
803 | 0 | int height) { |
804 | 0 | return IxxxToPxxx(src_y, src_stride_y, src_u, src_stride_u, src_v, |
805 | 0 | src_stride_v, dst_y, dst_stride_y, dst_uv, dst_stride_uv, |
806 | 0 | width, height, 1, 0, 10); |
807 | 0 | } |
808 | | |
809 | | LIBYUV_API |
810 | | int I012ToP012(const uint16_t* src_y, |
811 | | int src_stride_y, |
812 | | const uint16_t* src_u, |
813 | | int src_stride_u, |
814 | | const uint16_t* src_v, |
815 | | int src_stride_v, |
816 | | uint16_t* dst_y, |
817 | | int dst_stride_y, |
818 | | uint16_t* dst_uv, |
819 | | int dst_stride_uv, |
820 | | int width, |
821 | 0 | int height) { |
822 | 0 | return IxxxToPxxx(src_y, src_stride_y, src_u, src_stride_u, src_v, |
823 | 0 | src_stride_v, dst_y, dst_stride_y, dst_uv, dst_stride_uv, |
824 | 0 | width, height, 1, 1, 12); |
825 | 0 | } |
826 | | |
827 | | LIBYUV_API |
828 | | int I212ToP212(const uint16_t* src_y, |
829 | | int src_stride_y, |
830 | | const uint16_t* src_u, |
831 | | int src_stride_u, |
832 | | const uint16_t* src_v, |
833 | | int src_stride_v, |
834 | | uint16_t* dst_y, |
835 | | int dst_stride_y, |
836 | | uint16_t* dst_uv, |
837 | | int dst_stride_uv, |
838 | | int width, |
839 | 0 | int height) { |
840 | 0 | return IxxxToPxxx(src_y, src_stride_y, src_u, src_stride_u, src_v, |
841 | 0 | src_stride_v, dst_y, dst_stride_y, dst_uv, dst_stride_uv, |
842 | 0 | width, height, 1, 0, 12); |
843 | 0 | } |
844 | | |
845 | | // 422 chroma is 1/2 width, 1x height |
846 | | // 420 chroma is 1/2 width, 1/2 height |
847 | | LIBYUV_API |
848 | | int I422ToI420(const uint8_t* src_y, |
849 | | int src_stride_y, |
850 | | const uint8_t* src_u, |
851 | | int src_stride_u, |
852 | | const uint8_t* src_v, |
853 | | int src_stride_v, |
854 | | uint8_t* dst_y, |
855 | | int dst_stride_y, |
856 | | uint8_t* dst_u, |
857 | | int dst_stride_u, |
858 | | uint8_t* dst_v, |
859 | | int dst_stride_v, |
860 | | int width, |
861 | 0 | int height) { |
862 | 0 | const int src_uv_width = SUBSAMPLE(width, 1, 1); |
863 | 0 | return I4xxToI420(src_y, src_stride_y, src_u, src_stride_u, src_v, |
864 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u, |
865 | 0 | dst_v, dst_stride_v, width, height, src_uv_width, height); |
866 | 0 | } |
867 | | |
868 | | LIBYUV_API |
869 | | int I422ToI210(const uint8_t* src_y, |
870 | | int src_stride_y, |
871 | | const uint8_t* src_u, |
872 | | int src_stride_u, |
873 | | const uint8_t* src_v, |
874 | | int src_stride_v, |
875 | | uint16_t* dst_y, |
876 | | int dst_stride_y, |
877 | | uint16_t* dst_u, |
878 | | int dst_stride_u, |
879 | | uint16_t* dst_v, |
880 | | int dst_stride_v, |
881 | | int width, |
882 | 0 | int height) { |
883 | 0 | int halfwidth = (width + 1) >> 1; |
884 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 || |
885 | 0 | height == 0) { |
886 | 0 | return -1; |
887 | 0 | } |
888 | | // Negative height means invert the image. |
889 | 0 | if (height < 0) { |
890 | 0 | height = -height; |
891 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
892 | 0 | src_u = src_u + (height - 1) * src_stride_u; |
893 | 0 | src_v = src_v + (height - 1) * src_stride_v; |
894 | 0 | src_stride_y = -src_stride_y; |
895 | 0 | src_stride_u = -src_stride_u; |
896 | 0 | src_stride_v = -src_stride_v; |
897 | 0 | } |
898 | | |
899 | | // Convert Y plane. |
900 | 0 | Convert8To16Plane(src_y, src_stride_y, dst_y, dst_stride_y, 1024, width, |
901 | 0 | height); |
902 | | // Convert UV planes. |
903 | 0 | Convert8To16Plane(src_u, src_stride_u, dst_u, dst_stride_u, 1024, halfwidth, |
904 | 0 | height); |
905 | 0 | Convert8To16Plane(src_v, src_stride_v, dst_v, dst_stride_v, 1024, halfwidth, |
906 | 0 | height); |
907 | 0 | return 0; |
908 | 0 | } |
909 | | |
910 | | // TODO(fbarchard): Implement row conversion. |
911 | | LIBYUV_API |
912 | | int I422ToNV21(const uint8_t* src_y, |
913 | | int src_stride_y, |
914 | | const uint8_t* src_u, |
915 | | int src_stride_u, |
916 | | const uint8_t* src_v, |
917 | | int src_stride_v, |
918 | | uint8_t* dst_y, |
919 | | int dst_stride_y, |
920 | | uint8_t* dst_vu, |
921 | | int dst_stride_vu, |
922 | | int width, |
923 | 0 | int height) { |
924 | 0 | int halfwidth = (width + 1) >> 1; |
925 | 0 | int halfheight = (height + 1) >> 1; |
926 | | // Negative height means invert the image. |
927 | 0 | if (height < 0) { |
928 | 0 | height = -height; |
929 | 0 | halfheight = (height + 1) >> 1; |
930 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
931 | 0 | src_u = src_u + (height - 1) * src_stride_u; |
932 | 0 | src_v = src_v + (height - 1) * src_stride_v; |
933 | 0 | src_stride_y = -src_stride_y; |
934 | 0 | src_stride_u = -src_stride_u; |
935 | 0 | src_stride_v = -src_stride_v; |
936 | 0 | } |
937 | | |
938 | | // Allocate u and v buffers |
939 | 0 | align_buffer_64(plane_u, halfwidth * halfheight * 2); |
940 | 0 | uint8_t* plane_v = plane_u + halfwidth * halfheight; |
941 | 0 | if (!plane_u) |
942 | 0 | return 1; |
943 | | |
944 | 0 | I422ToI420(src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v, |
945 | 0 | dst_y, dst_stride_y, plane_u, halfwidth, plane_v, halfwidth, width, |
946 | 0 | height); |
947 | 0 | MergeUVPlane(plane_v, halfwidth, plane_u, halfwidth, dst_vu, dst_stride_vu, |
948 | 0 | halfwidth, halfheight); |
949 | 0 | free_aligned_buffer_64(plane_u); |
950 | 0 | return 0; |
951 | 0 | } |
952 | | |
953 | | LIBYUV_API |
954 | | int MM21ToNV12(const uint8_t* src_y, |
955 | | int src_stride_y, |
956 | | const uint8_t* src_uv, |
957 | | int src_stride_uv, |
958 | | uint8_t* dst_y, |
959 | | int dst_stride_y, |
960 | | uint8_t* dst_uv, |
961 | | int dst_stride_uv, |
962 | | int width, |
963 | 0 | int height) { |
964 | 0 | if (!src_uv || !dst_uv || width <= 0) { |
965 | 0 | return -1; |
966 | 0 | } |
967 | | |
968 | 0 | int sign = height < 0 ? -1 : 1; |
969 | |
|
970 | 0 | if (dst_y) { |
971 | 0 | DetilePlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height, 32); |
972 | 0 | } |
973 | 0 | DetilePlane(src_uv, src_stride_uv, dst_uv, dst_stride_uv, (width + 1) & ~1, |
974 | 0 | (height + sign) / 2, 16); |
975 | |
|
976 | 0 | return 0; |
977 | 0 | } |
978 | | |
979 | | LIBYUV_API |
980 | | int MM21ToI420(const uint8_t* src_y, |
981 | | int src_stride_y, |
982 | | const uint8_t* src_uv, |
983 | | int src_stride_uv, |
984 | | uint8_t* dst_y, |
985 | | int dst_stride_y, |
986 | | uint8_t* dst_u, |
987 | | int dst_stride_u, |
988 | | uint8_t* dst_v, |
989 | | int dst_stride_v, |
990 | | int width, |
991 | 0 | int height) { |
992 | 0 | int sign = height < 0 ? -1 : 1; |
993 | |
|
994 | 0 | if (!src_uv || !dst_u || !dst_v || width <= 0) { |
995 | 0 | return -1; |
996 | 0 | } |
997 | | |
998 | 0 | if (dst_y) { |
999 | 0 | DetilePlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height, 32); |
1000 | 0 | } |
1001 | 0 | DetileSplitUVPlane(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v, |
1002 | 0 | dst_stride_v, (width + 1) & ~1, (height + sign) / 2, 16); |
1003 | |
|
1004 | 0 | return 0; |
1005 | 0 | } |
1006 | | |
1007 | | LIBYUV_API |
1008 | | int MM21ToYUY2(const uint8_t* src_y, |
1009 | | int src_stride_y, |
1010 | | const uint8_t* src_uv, |
1011 | | int src_stride_uv, |
1012 | | uint8_t* dst_yuy2, |
1013 | | int dst_stride_yuy2, |
1014 | | int width, |
1015 | 0 | int height) { |
1016 | 0 | if (!src_y || !src_uv || !dst_yuy2 || width <= 0) { |
1017 | 0 | return -1; |
1018 | 0 | } |
1019 | | |
1020 | 0 | DetileToYUY2(src_y, src_stride_y, src_uv, src_stride_uv, dst_yuy2, |
1021 | 0 | dst_stride_yuy2, width, height, 32); |
1022 | |
|
1023 | 0 | return 0; |
1024 | 0 | } |
1025 | | |
1026 | | // Convert MT2T into P010. See tinyurl.com/mtk-10bit-video-format for format |
1027 | | // documentation. |
1028 | | // TODO(greenjustin): Add an MT2T to I420 conversion. |
1029 | | LIBYUV_API |
1030 | | int MT2TToP010(const uint8_t* src_y, |
1031 | | int src_stride_y, |
1032 | | const uint8_t* src_uv, |
1033 | | int src_stride_uv, |
1034 | | uint16_t* dst_y, |
1035 | | int dst_stride_y, |
1036 | | uint16_t* dst_uv, |
1037 | | int dst_stride_uv, |
1038 | | int width, |
1039 | 0 | int height) { |
1040 | 0 | if (width <= 0 || !height || !src_uv || !dst_uv) { |
1041 | 0 | return -1; |
1042 | 0 | } |
1043 | | |
1044 | 0 | { |
1045 | 0 | int uv_width = (width + 1) & ~1; |
1046 | 0 | int uv_height = (height + 1) / 2; |
1047 | 0 | int y = 0; |
1048 | 0 | const int tile_width = 16; |
1049 | 0 | const int y_tile_height = 32; |
1050 | 0 | const int uv_tile_height = 16; |
1051 | 0 | int padded_width = (width + tile_width - 1) & ~(tile_width - 1); |
1052 | 0 | int y_tile_row_size = padded_width * y_tile_height * 10 / 8; |
1053 | 0 | int uv_tile_row_size = padded_width * uv_tile_height * 10 / 8; |
1054 | 0 | size_t row_buf_size = padded_width * y_tile_height * sizeof(uint16_t); |
1055 | 0 | void (*UnpackMT2T)(const uint8_t* src, uint16_t* dst, size_t size) = |
1056 | 0 | UnpackMT2T_C; |
1057 | 0 | align_buffer_64(row_buf, row_buf_size); |
1058 | 0 | if (!row_buf) |
1059 | 0 | return 1; |
1060 | | |
1061 | | #if defined(HAS_UNPACKMT2T_NEON) |
1062 | | if (TestCpuFlag(kCpuHasNEON)) { |
1063 | | UnpackMT2T = UnpackMT2T_NEON; |
1064 | | } |
1065 | | #endif |
1066 | | // Negative height means invert the image. |
1067 | 0 | if (height < 0) { |
1068 | 0 | height = -height; |
1069 | 0 | uv_height = (height + 1) / 2; |
1070 | 0 | if (dst_y) { |
1071 | 0 | dst_y = dst_y + (height - 1) * dst_stride_y; |
1072 | 0 | dst_stride_y = -dst_stride_y; |
1073 | 0 | } |
1074 | 0 | dst_uv = dst_uv + (uv_height - 1) * dst_stride_uv; |
1075 | 0 | dst_stride_uv = -dst_stride_uv; |
1076 | 0 | } |
1077 | | |
1078 | | // Unpack and detile Y in rows of tiles |
1079 | 0 | if (src_y && dst_y) { |
1080 | 0 | for (y = 0; y < (height & ~(y_tile_height - 1)); y += y_tile_height) { |
1081 | 0 | UnpackMT2T(src_y, (uint16_t*)row_buf, y_tile_row_size); |
1082 | 0 | DetilePlane_16((uint16_t*)row_buf, padded_width, dst_y, dst_stride_y, |
1083 | 0 | width, y_tile_height, y_tile_height); |
1084 | 0 | src_y += src_stride_y * y_tile_height; |
1085 | 0 | dst_y += dst_stride_y * y_tile_height; |
1086 | 0 | } |
1087 | 0 | if (height & (y_tile_height - 1)) { |
1088 | 0 | UnpackMT2T(src_y, (uint16_t*)row_buf, y_tile_row_size); |
1089 | 0 | DetilePlane_16((uint16_t*)row_buf, padded_width, dst_y, dst_stride_y, |
1090 | 0 | width, height & (y_tile_height - 1), y_tile_height); |
1091 | 0 | } |
1092 | 0 | } |
1093 | | |
1094 | | // Unpack and detile UV plane |
1095 | 0 | for (y = 0; y < (uv_height & ~(uv_tile_height - 1)); y += uv_tile_height) { |
1096 | 0 | UnpackMT2T(src_uv, (uint16_t*)row_buf, uv_tile_row_size); |
1097 | 0 | DetilePlane_16((uint16_t*)row_buf, padded_width, dst_uv, dst_stride_uv, |
1098 | 0 | uv_width, uv_tile_height, uv_tile_height); |
1099 | 0 | src_uv += src_stride_uv * uv_tile_height; |
1100 | 0 | dst_uv += dst_stride_uv * uv_tile_height; |
1101 | 0 | } |
1102 | 0 | if (uv_height & (uv_tile_height - 1)) { |
1103 | 0 | UnpackMT2T(src_uv, (uint16_t*)row_buf, uv_tile_row_size); |
1104 | 0 | DetilePlane_16((uint16_t*)row_buf, padded_width, dst_uv, dst_stride_uv, |
1105 | 0 | uv_width, uv_height & (uv_tile_height - 1), |
1106 | 0 | uv_tile_height); |
1107 | 0 | } |
1108 | 0 | free_aligned_buffer_64(row_buf); |
1109 | 0 | } |
1110 | 0 | return 0; |
1111 | 0 | } |
1112 | | |
1113 | | #ifdef I422TONV21_ROW_VERSION |
1114 | | // Unittest fails for this version. |
1115 | | // 422 chroma is 1/2 width, 1x height |
1116 | | // 420 chroma is 1/2 width, 1/2 height |
1117 | | // Swap src_u and src_v to implement I422ToNV12 |
1118 | | LIBYUV_API |
1119 | | int I422ToNV21(const uint8_t* src_y, |
1120 | | int src_stride_y, |
1121 | | const uint8_t* src_u, |
1122 | | int src_stride_u, |
1123 | | const uint8_t* src_v, |
1124 | | int src_stride_v, |
1125 | | uint8_t* dst_y, |
1126 | | int dst_stride_y, |
1127 | | uint8_t* dst_vu, |
1128 | | int dst_stride_vu, |
1129 | | int width, |
1130 | | int height) { |
1131 | | int y; |
1132 | | void (*MergeUVRow)(const uint8_t* src_u, const uint8_t* src_v, |
1133 | | uint8_t* dst_uv, int width) = MergeUVRow_C; |
1134 | | void (*InterpolateRow)(uint8_t* dst_ptr, const uint8_t* src_ptr, |
1135 | | ptrdiff_t src_stride, int dst_width, |
1136 | | int source_y_fraction) = InterpolateRow_C; |
1137 | | int halfwidth = (width + 1) >> 1; |
1138 | | int halfheight = (height + 1) >> 1; |
1139 | | if ((!src_y && dst_y) || !src_u || !src_v || !dst_vu || width <= 0 || |
1140 | | height == 0) { |
1141 | | return -1; |
1142 | | } |
1143 | | // Negative height means invert the image. |
1144 | | if (height < 0) { |
1145 | | height = -height; |
1146 | | halfheight = (height + 1) >> 1; |
1147 | | src_y = src_y + (height - 1) * src_stride_y; |
1148 | | src_u = src_u + (halfheight - 1) * src_stride_u; |
1149 | | src_v = src_v + (halfheight - 1) * src_stride_v; |
1150 | | src_stride_y = -src_stride_y; |
1151 | | src_stride_u = -src_stride_u; |
1152 | | src_stride_v = -src_stride_v; |
1153 | | } |
1154 | | #if defined(HAS_MERGEUVROW_SSE2) |
1155 | | if (TestCpuFlag(kCpuHasSSE2)) { |
1156 | | MergeUVRow = MergeUVRow_Any_SSE2; |
1157 | | if (IS_ALIGNED(halfwidth, 16)) { |
1158 | | MergeUVRow = MergeUVRow_SSE2; |
1159 | | } |
1160 | | } |
1161 | | #endif |
1162 | | #if defined(HAS_MERGEUVROW_AVX2) |
1163 | | if (TestCpuFlag(kCpuHasAVX2)) { |
1164 | | MergeUVRow = MergeUVRow_Any_AVX2; |
1165 | | if (IS_ALIGNED(halfwidth, 16)) { |
1166 | | MergeUVRow = MergeUVRow_AVX2; |
1167 | | } |
1168 | | } |
1169 | | #endif |
1170 | | #if defined(HAS_MERGEUVROW_AVX512BW) |
1171 | | if (TestCpuFlag(kCpuHasAVX512BW)) { |
1172 | | MergeUVRow = MergeUVRow_Any_AVX512BW; |
1173 | | if (IS_ALIGNED(halfwidth, 32)) { |
1174 | | MergeUVRow = MergeUVRow_AVX512BW; |
1175 | | } |
1176 | | } |
1177 | | #endif |
1178 | | #if defined(HAS_MERGEUVROW_NEON) |
1179 | | if (TestCpuFlag(kCpuHasNEON)) { |
1180 | | MergeUVRow = MergeUVRow_Any_NEON; |
1181 | | if (IS_ALIGNED(halfwidth, 16)) { |
1182 | | MergeUVRow = MergeUVRow_NEON; |
1183 | | } |
1184 | | } |
1185 | | #endif |
1186 | | #if defined(HAS_MERGEUVROW_SME) |
1187 | | if (TestCpuFlag(kCpuHasSME)) { |
1188 | | MergeUVRow = MergeUVRow_SME; |
1189 | | } |
1190 | | #endif |
1191 | | #if defined(HAS_MERGEUVROW_LSX) |
1192 | | if (TestCpuFlag(kCpuHasLSX)) { |
1193 | | MergeUVRow = MergeUVRow_Any_LSX; |
1194 | | if (IS_ALIGNED(halfwidth, 16)) { |
1195 | | MergeUVRow = MergeUVRow_LSX; |
1196 | | } |
1197 | | } |
1198 | | #endif |
1199 | | #if defined(HAS_MERGEUVROW_RVV) |
1200 | | if (TestCpuFlag(kCpuHasRVV)) { |
1201 | | MergeUVRow = MergeUVRow_RVV; |
1202 | | } |
1203 | | #endif |
1204 | | #if defined(HAS_INTERPOLATEROW_SSSE3) |
1205 | | if (TestCpuFlag(kCpuHasSSSE3)) { |
1206 | | InterpolateRow = InterpolateRow_Any_SSSE3; |
1207 | | if (IS_ALIGNED(width, 16)) { |
1208 | | InterpolateRow = InterpolateRow_SSSE3; |
1209 | | } |
1210 | | } |
1211 | | #endif |
1212 | | #if defined(HAS_INTERPOLATEROW_AVX2) |
1213 | | if (TestCpuFlag(kCpuHasAVX2)) { |
1214 | | InterpolateRow = InterpolateRow_Any_AVX2; |
1215 | | if (IS_ALIGNED(width, 32)) { |
1216 | | InterpolateRow = InterpolateRow_AVX2; |
1217 | | } |
1218 | | } |
1219 | | #endif |
1220 | | #if defined(HAS_INTERPOLATEROW_NEON) |
1221 | | if (TestCpuFlag(kCpuHasNEON)) { |
1222 | | InterpolateRow = InterpolateRow_Any_NEON; |
1223 | | if (IS_ALIGNED(width, 16)) { |
1224 | | InterpolateRow = InterpolateRow_NEON; |
1225 | | } |
1226 | | } |
1227 | | #endif |
1228 | | #if defined(HAS_INTERPOLATEROW_SME) |
1229 | | if (TestCpuFlag(kCpuHasSME)) { |
1230 | | InterpolateRow = InterpolateRow_SME; |
1231 | | } |
1232 | | #endif |
1233 | | #if defined(HAS_INTERPOLATEROW_LSX) |
1234 | | if (TestCpuFlag(kCpuHasLSX)) { |
1235 | | InterpolateRow = InterpolateRow_Any_LSX; |
1236 | | if (IS_ALIGNED(width, 32)) { |
1237 | | InterpolateRow = InterpolateRow_LSX; |
1238 | | } |
1239 | | } |
1240 | | #endif |
1241 | | #if defined(HAS_INTERPOLATEROW_RVV) |
1242 | | if (TestCpuFlag(kCpuHasRVV)) { |
1243 | | InterpolateRow = InterpolateRow_RVV; |
1244 | | } |
1245 | | #endif |
1246 | | |
1247 | | if (dst_y) { |
1248 | | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, halfwidth, height); |
1249 | | } |
1250 | | { |
1251 | | // Allocate 2 rows of vu. |
1252 | | int awidth = halfwidth * 2; |
1253 | | align_buffer_64(row_vu_0, awidth * 2); |
1254 | | uint8_t* row_vu_1 = row_vu_0 + awidth; |
1255 | | if (!row_vu_0) |
1256 | | return 1; |
1257 | | |
1258 | | for (y = 0; y < height - 1; y += 2) { |
1259 | | MergeUVRow(src_v, src_u, row_vu_0, halfwidth); |
1260 | | MergeUVRow(src_v + src_stride_v, src_u + src_stride_u, row_vu_1, |
1261 | | halfwidth); |
1262 | | InterpolateRow(dst_vu, row_vu_0, awidth, awidth, 128); |
1263 | | src_u += src_stride_u * 2; |
1264 | | src_v += src_stride_v * 2; |
1265 | | dst_vu += dst_stride_vu; |
1266 | | } |
1267 | | if (height & 1) { |
1268 | | MergeUVRow(src_v, src_u, dst_vu, halfwidth); |
1269 | | } |
1270 | | free_aligned_buffer_64(row_vu_0); |
1271 | | } |
1272 | | return 0; |
1273 | | } |
1274 | | #endif // I422TONV21_ROW_VERSION |
1275 | | |
1276 | | // 444 chroma is 1x width, 1x height |
1277 | | // 420 chroma is 1/2 width, 1/2 height |
1278 | | LIBYUV_API |
1279 | | int I444ToI420(const uint8_t* src_y, |
1280 | | int src_stride_y, |
1281 | | const uint8_t* src_u, |
1282 | | int src_stride_u, |
1283 | | const uint8_t* src_v, |
1284 | | int src_stride_v, |
1285 | | uint8_t* dst_y, |
1286 | | int dst_stride_y, |
1287 | | uint8_t* dst_u, |
1288 | | int dst_stride_u, |
1289 | | uint8_t* dst_v, |
1290 | | int dst_stride_v, |
1291 | | int width, |
1292 | 0 | int height) { |
1293 | 0 | return I4xxToI420(src_y, src_stride_y, src_u, src_stride_u, src_v, |
1294 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u, |
1295 | 0 | dst_v, dst_stride_v, width, height, width, height); |
1296 | 0 | } |
1297 | | |
1298 | | LIBYUV_API |
1299 | | int I444ToNV12(const uint8_t* src_y, |
1300 | | int src_stride_y, |
1301 | | const uint8_t* src_u, |
1302 | | int src_stride_u, |
1303 | | const uint8_t* src_v, |
1304 | | int src_stride_v, |
1305 | | uint8_t* dst_y, |
1306 | | int dst_stride_y, |
1307 | | uint8_t* dst_uv, |
1308 | | int dst_stride_uv, |
1309 | | int width, |
1310 | 0 | int height) { |
1311 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_uv || width <= 0 || |
1312 | 0 | height == 0) { |
1313 | 0 | return -1; |
1314 | 0 | } |
1315 | | // Negative height means invert the image. |
1316 | 0 | if (height < 0) { |
1317 | 0 | height = -height; |
1318 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
1319 | 0 | src_u = src_u + (height - 1) * src_stride_u; |
1320 | 0 | src_v = src_v + (height - 1) * src_stride_v; |
1321 | 0 | src_stride_y = -src_stride_y; |
1322 | 0 | src_stride_u = -src_stride_u; |
1323 | 0 | src_stride_v = -src_stride_v; |
1324 | 0 | } |
1325 | 0 | if (dst_y) { |
1326 | 0 | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
1327 | 0 | } |
1328 | 0 | HalfMergeUVPlane(src_u, src_stride_u, src_v, src_stride_v, dst_uv, |
1329 | 0 | dst_stride_uv, width, height); |
1330 | 0 | return 0; |
1331 | 0 | } |
1332 | | |
1333 | | LIBYUV_API |
1334 | | int I444ToNV21(const uint8_t* src_y, |
1335 | | int src_stride_y, |
1336 | | const uint8_t* src_u, |
1337 | | int src_stride_u, |
1338 | | const uint8_t* src_v, |
1339 | | int src_stride_v, |
1340 | | uint8_t* dst_y, |
1341 | | int dst_stride_y, |
1342 | | uint8_t* dst_vu, |
1343 | | int dst_stride_vu, |
1344 | | int width, |
1345 | 0 | int height) { |
1346 | 0 | return I444ToNV12(src_y, src_stride_y, src_v, src_stride_v, src_u, |
1347 | 0 | src_stride_u, dst_y, dst_stride_y, dst_vu, dst_stride_vu, |
1348 | 0 | width, height); |
1349 | 0 | } |
1350 | | |
1351 | | // I400 is greyscale typically used in MJPG |
1352 | | LIBYUV_API |
1353 | | int I400ToI420(const uint8_t* src_y, |
1354 | | int src_stride_y, |
1355 | | uint8_t* dst_y, |
1356 | | int dst_stride_y, |
1357 | | uint8_t* dst_u, |
1358 | | int dst_stride_u, |
1359 | | uint8_t* dst_v, |
1360 | | int dst_stride_v, |
1361 | | int width, |
1362 | 0 | int height) { |
1363 | 0 | int halfwidth = (width + 1) >> 1; |
1364 | 0 | int halfheight = (height + 1) >> 1; |
1365 | 0 | if ((!src_y && dst_y) || !dst_u || !dst_v || width <= 0 || height == 0) { |
1366 | 0 | return -1; |
1367 | 0 | } |
1368 | | // Negative height means invert the image. |
1369 | 0 | if (height < 0) { |
1370 | 0 | height = -height; |
1371 | 0 | halfheight = (height + 1) >> 1; |
1372 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
1373 | 0 | src_stride_y = -src_stride_y; |
1374 | 0 | } |
1375 | 0 | if (dst_y) { |
1376 | 0 | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
1377 | 0 | } |
1378 | 0 | SetPlane(dst_u, dst_stride_u, halfwidth, halfheight, 128); |
1379 | 0 | SetPlane(dst_v, dst_stride_v, halfwidth, halfheight, 128); |
1380 | 0 | return 0; |
1381 | 0 | } |
1382 | | |
1383 | | // I400 is greyscale typically used in MJPG |
1384 | | LIBYUV_API |
1385 | | int I400ToNV21(const uint8_t* src_y, |
1386 | | int src_stride_y, |
1387 | | uint8_t* dst_y, |
1388 | | int dst_stride_y, |
1389 | | uint8_t* dst_vu, |
1390 | | int dst_stride_vu, |
1391 | | int width, |
1392 | 0 | int height) { |
1393 | 0 | int halfwidth = (width + 1) >> 1; |
1394 | 0 | int halfheight = (height + 1) >> 1; |
1395 | 0 | if ((!src_y && dst_y) || !dst_vu || width <= 0 || height == 0) { |
1396 | 0 | return -1; |
1397 | 0 | } |
1398 | | // Negative height means invert the image. |
1399 | 0 | if (height < 0) { |
1400 | 0 | height = -height; |
1401 | 0 | halfheight = (height + 1) >> 1; |
1402 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
1403 | 0 | src_stride_y = -src_stride_y; |
1404 | 0 | } |
1405 | 0 | if (dst_y) { |
1406 | 0 | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
1407 | 0 | } |
1408 | 0 | SetPlane(dst_vu, dst_stride_vu, halfwidth * 2, halfheight, 128); |
1409 | 0 | return 0; |
1410 | 0 | } |
1411 | | |
1412 | | // Convert NV12 to I420. |
1413 | | // TODO(fbarchard): Consider inverting destination. Faster on ARM with prfm. |
1414 | | LIBYUV_API |
1415 | | int NV12ToI420(const uint8_t* src_y, |
1416 | | int src_stride_y, |
1417 | | const uint8_t* src_uv, |
1418 | | int src_stride_uv, |
1419 | | uint8_t* dst_y, |
1420 | | int dst_stride_y, |
1421 | | uint8_t* dst_u, |
1422 | | int dst_stride_u, |
1423 | | uint8_t* dst_v, |
1424 | | int dst_stride_v, |
1425 | | int width, |
1426 | 0 | int height) { |
1427 | 0 | int halfwidth = (width + 1) >> 1; |
1428 | 0 | int halfheight = (height + 1) >> 1; |
1429 | 0 | if ((!src_y && dst_y) || !src_uv || !dst_u || !dst_v || width <= 0 || |
1430 | 0 | height == 0) { |
1431 | 0 | return -1; |
1432 | 0 | } |
1433 | | // Negative height means invert the image. |
1434 | 0 | if (height < 0) { |
1435 | 0 | height = -height; |
1436 | 0 | halfheight = (height + 1) >> 1; |
1437 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
1438 | 0 | src_uv = src_uv + (halfheight - 1) * src_stride_uv; |
1439 | 0 | src_stride_y = -src_stride_y; |
1440 | 0 | src_stride_uv = -src_stride_uv; |
1441 | 0 | } |
1442 | | // Coalesce rows. |
1443 | 0 | if (src_stride_y == width && dst_stride_y == width) { |
1444 | 0 | width *= height; |
1445 | 0 | height = 1; |
1446 | 0 | src_stride_y = dst_stride_y = 0; |
1447 | 0 | } |
1448 | | // Coalesce rows. |
1449 | 0 | if (src_stride_uv == halfwidth * 2 && dst_stride_u == halfwidth && |
1450 | 0 | dst_stride_v == halfwidth) { |
1451 | 0 | halfwidth *= halfheight; |
1452 | 0 | halfheight = 1; |
1453 | 0 | src_stride_uv = dst_stride_u = dst_stride_v = 0; |
1454 | 0 | } |
1455 | |
|
1456 | 0 | if (dst_y) { |
1457 | 0 | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
1458 | 0 | } |
1459 | | |
1460 | | // Split UV plane - NV12 / NV21 |
1461 | 0 | SplitUVPlane(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v, dst_stride_v, |
1462 | 0 | halfwidth, halfheight); |
1463 | |
|
1464 | 0 | return 0; |
1465 | 0 | } |
1466 | | |
1467 | | // Convert NV21 to I420. Same as NV12 but u and v pointers swapped. |
1468 | | LIBYUV_API |
1469 | | int NV21ToI420(const uint8_t* src_y, |
1470 | | int src_stride_y, |
1471 | | const uint8_t* src_vu, |
1472 | | int src_stride_vu, |
1473 | | uint8_t* dst_y, |
1474 | | int dst_stride_y, |
1475 | | uint8_t* dst_u, |
1476 | | int dst_stride_u, |
1477 | | uint8_t* dst_v, |
1478 | | int dst_stride_v, |
1479 | | int width, |
1480 | 0 | int height) { |
1481 | 0 | return NV12ToI420(src_y, src_stride_y, src_vu, src_stride_vu, dst_y, |
1482 | 0 | dst_stride_y, dst_v, dst_stride_v, dst_u, dst_stride_u, |
1483 | 0 | width, height); |
1484 | 0 | } |
1485 | | |
1486 | | LIBYUV_API |
1487 | | int NV12ToNV24(const uint8_t* src_y, |
1488 | | int src_stride_y, |
1489 | | const uint8_t* src_uv, |
1490 | | int src_stride_uv, |
1491 | | uint8_t* dst_y, |
1492 | | int dst_stride_y, |
1493 | | uint8_t* dst_uv, |
1494 | | int dst_stride_uv, |
1495 | | int width, |
1496 | 0 | int height) { |
1497 | 0 | int r; |
1498 | 0 | if ((!src_y && dst_y) || !src_uv || !dst_uv || width <= 0 || height == 0) { |
1499 | 0 | return -1; |
1500 | 0 | } |
1501 | | |
1502 | 0 | if (dst_y) { |
1503 | 0 | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
1504 | 0 | } |
1505 | 0 | r = UVScale(src_uv, src_stride_uv, SUBSAMPLE(width, 1, 1), |
1506 | 0 | SUBSAMPLE(height, 1, 1), dst_uv, dst_stride_uv, Abs(width), |
1507 | 0 | Abs(height), kFilterBilinear); |
1508 | 0 | return r; |
1509 | 0 | } |
1510 | | |
1511 | | LIBYUV_API |
1512 | | int NV16ToNV24(const uint8_t* src_y, |
1513 | | int src_stride_y, |
1514 | | const uint8_t* src_uv, |
1515 | | int src_stride_uv, |
1516 | | uint8_t* dst_y, |
1517 | | int dst_stride_y, |
1518 | | uint8_t* dst_uv, |
1519 | | int dst_stride_uv, |
1520 | | int width, |
1521 | 0 | int height) { |
1522 | 0 | int r; |
1523 | 0 | if ((!src_y && dst_y) || !src_uv || !dst_uv || width <= 0 || height == 0) { |
1524 | 0 | return -1; |
1525 | 0 | } |
1526 | | |
1527 | 0 | if (dst_y) { |
1528 | 0 | CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
1529 | 0 | } |
1530 | 0 | r = UVScale(src_uv, src_stride_uv, SUBSAMPLE(width, 1, 1), height, dst_uv, |
1531 | 0 | dst_stride_uv, Abs(width), Abs(height), kFilterBilinear); |
1532 | 0 | return r; |
1533 | 0 | } |
1534 | | |
1535 | | // Any P[420]1[02] to I[420]1[02] format |
1536 | | static int PxxxToIxxx(const uint16_t* src_y, |
1537 | | int src_stride_y, |
1538 | | const uint16_t* src_uv, |
1539 | | int src_stride_uv, |
1540 | | uint16_t* dst_y, |
1541 | | int dst_stride_y, |
1542 | | uint16_t* dst_u, |
1543 | | int dst_stride_u, |
1544 | | uint16_t* dst_v, |
1545 | | int dst_stride_v, |
1546 | | int width, |
1547 | | int height, |
1548 | | int subsample_x, |
1549 | | int subsample_y, |
1550 | 0 | int depth) { |
1551 | 0 | const int uv_width = SUBSAMPLE(width, subsample_x, subsample_x); |
1552 | 0 | const int uv_height = SUBSAMPLE(height, subsample_y, subsample_y); |
1553 | 0 | if (!src_y || !dst_y || !src_uv || !dst_u || !dst_v || width <= 0 || |
1554 | 0 | height == 0) { |
1555 | 0 | return -1; |
1556 | 0 | } |
1557 | 0 | ConvertToLSBPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height, |
1558 | 0 | depth); |
1559 | 0 | SplitUVPlane_16(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v, |
1560 | 0 | dst_stride_v, uv_width, uv_height, depth); |
1561 | 0 | return 0; |
1562 | 0 | } |
1563 | | |
1564 | | LIBYUV_API |
1565 | | int P010ToI010(const uint16_t* src_y, |
1566 | | int src_stride_y, |
1567 | | const uint16_t* src_uv, |
1568 | | int src_stride_uv, |
1569 | | uint16_t* dst_y, |
1570 | | int dst_stride_y, |
1571 | | uint16_t* dst_u, |
1572 | | int dst_stride_u, |
1573 | | uint16_t* dst_v, |
1574 | | int dst_stride_v, |
1575 | | int width, |
1576 | 0 | int height) { |
1577 | 0 | return PxxxToIxxx(src_y, src_stride_y, src_uv, src_stride_uv, dst_y, |
1578 | 0 | dst_stride_y, dst_u, dst_stride_u, dst_v, dst_stride_v, |
1579 | 0 | width, height, 1, 1, 10); |
1580 | 0 | } |
1581 | | |
1582 | | LIBYUV_API |
1583 | | int P012ToI012(const uint16_t* src_y, |
1584 | | int src_stride_y, |
1585 | | const uint16_t* src_uv, |
1586 | | int src_stride_uv, |
1587 | | uint16_t* dst_y, |
1588 | | int dst_stride_y, |
1589 | | uint16_t* dst_u, |
1590 | | int dst_stride_u, |
1591 | | uint16_t* dst_v, |
1592 | | int dst_stride_v, |
1593 | | int width, |
1594 | 0 | int height) { |
1595 | 0 | return PxxxToIxxx(src_y, src_stride_y, src_uv, src_stride_uv, dst_y, |
1596 | 0 | dst_stride_y, dst_u, dst_stride_u, dst_v, dst_stride_v, |
1597 | 0 | width, height, 1, 1, 12); |
1598 | 0 | } |
1599 | | |
1600 | | LIBYUV_API |
1601 | | int P010ToP410(const uint16_t* src_y, |
1602 | | int src_stride_y, |
1603 | | const uint16_t* src_uv, |
1604 | | int src_stride_uv, |
1605 | | uint16_t* dst_y, |
1606 | | int dst_stride_y, |
1607 | | uint16_t* dst_uv, |
1608 | | int dst_stride_uv, |
1609 | | int width, |
1610 | 0 | int height) { |
1611 | 0 | int r; |
1612 | 0 | if ((!src_y && dst_y) || !src_uv || !dst_uv || width <= 0 || height == 0) { |
1613 | 0 | return -1; |
1614 | 0 | } |
1615 | | |
1616 | 0 | if (dst_y) { |
1617 | 0 | CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
1618 | 0 | } |
1619 | 0 | r = UVScale_16(src_uv, src_stride_uv, SUBSAMPLE(width, 1, 1), |
1620 | 0 | SUBSAMPLE(height, 1, 1), dst_uv, dst_stride_uv, Abs(width), |
1621 | 0 | Abs(height), kFilterBilinear); |
1622 | 0 | return r; |
1623 | 0 | } |
1624 | | |
1625 | | LIBYUV_API |
1626 | | int P210ToP410(const uint16_t* src_y, |
1627 | | int src_stride_y, |
1628 | | const uint16_t* src_uv, |
1629 | | int src_stride_uv, |
1630 | | uint16_t* dst_y, |
1631 | | int dst_stride_y, |
1632 | | uint16_t* dst_uv, |
1633 | | int dst_stride_uv, |
1634 | | int width, |
1635 | 0 | int height) { |
1636 | 0 | int r; |
1637 | 0 | if ((!src_y && dst_y) || !src_uv || !dst_uv || width <= 0 || height == 0) { |
1638 | 0 | return -1; |
1639 | 0 | } |
1640 | | |
1641 | 0 | if (dst_y) { |
1642 | 0 | CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height); |
1643 | 0 | } |
1644 | 0 | r = UVScale_16(src_uv, src_stride_uv, SUBSAMPLE(width, 1, 1), height, dst_uv, |
1645 | 0 | dst_stride_uv, Abs(width), Abs(height), kFilterBilinear); |
1646 | 0 | return r; |
1647 | 0 | } |
1648 | | |
1649 | | // Convert YUY2 to I420. |
1650 | | LIBYUV_API |
1651 | | int YUY2ToI420(const uint8_t* src_yuy2, |
1652 | | int src_stride_yuy2, |
1653 | | uint8_t* dst_y, |
1654 | | int dst_stride_y, |
1655 | | uint8_t* dst_u, |
1656 | | int dst_stride_u, |
1657 | | uint8_t* dst_v, |
1658 | | int dst_stride_v, |
1659 | | int width, |
1660 | 0 | int height) { |
1661 | 0 | int y; |
1662 | 0 | void (*YUY2ToUVRow)(const uint8_t* src_yuy2, int src_stride_yuy2, |
1663 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
1664 | 0 | YUY2ToUVRow_C; |
1665 | 0 | void (*YUY2ToYRow)(const uint8_t* src_yuy2, uint8_t* dst_y, int width) = |
1666 | 0 | YUY2ToYRow_C; |
1667 | | // Negative height means invert the image. |
1668 | 0 | if (height < 0) { |
1669 | 0 | height = -height; |
1670 | 0 | src_yuy2 = src_yuy2 + (height - 1) * src_stride_yuy2; |
1671 | 0 | src_stride_yuy2 = -src_stride_yuy2; |
1672 | 0 | } |
1673 | 0 | #if defined(HAS_YUY2TOYROW_SSE2) |
1674 | 0 | if (TestCpuFlag(kCpuHasSSE2)) { |
1675 | 0 | YUY2ToUVRow = YUY2ToUVRow_Any_SSE2; |
1676 | 0 | YUY2ToYRow = YUY2ToYRow_Any_SSE2; |
1677 | 0 | if (IS_ALIGNED(width, 16)) { |
1678 | 0 | YUY2ToUVRow = YUY2ToUVRow_SSE2; |
1679 | 0 | YUY2ToYRow = YUY2ToYRow_SSE2; |
1680 | 0 | } |
1681 | 0 | } |
1682 | 0 | #endif |
1683 | 0 | #if defined(HAS_YUY2TOYROW_AVX2) |
1684 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
1685 | 0 | YUY2ToUVRow = YUY2ToUVRow_Any_AVX2; |
1686 | 0 | YUY2ToYRow = YUY2ToYRow_Any_AVX2; |
1687 | 0 | if (IS_ALIGNED(width, 32)) { |
1688 | 0 | YUY2ToUVRow = YUY2ToUVRow_AVX2; |
1689 | 0 | YUY2ToYRow = YUY2ToYRow_AVX2; |
1690 | 0 | } |
1691 | 0 | } |
1692 | 0 | #endif |
1693 | | #if defined(HAS_YUY2TOYROW_NEON) |
1694 | | if (TestCpuFlag(kCpuHasNEON)) { |
1695 | | YUY2ToYRow = YUY2ToYRow_Any_NEON; |
1696 | | YUY2ToUVRow = YUY2ToUVRow_Any_NEON; |
1697 | | if (IS_ALIGNED(width, 16)) { |
1698 | | YUY2ToYRow = YUY2ToYRow_NEON; |
1699 | | YUY2ToUVRow = YUY2ToUVRow_NEON; |
1700 | | } |
1701 | | } |
1702 | | #endif |
1703 | | #if defined(HAS_YUY2TOYROW_LSX) && defined(HAS_YUY2TOUVROW_LSX) |
1704 | | if (TestCpuFlag(kCpuHasLSX)) { |
1705 | | YUY2ToYRow = YUY2ToYRow_Any_LSX; |
1706 | | YUY2ToUVRow = YUY2ToUVRow_Any_LSX; |
1707 | | if (IS_ALIGNED(width, 16)) { |
1708 | | YUY2ToYRow = YUY2ToYRow_LSX; |
1709 | | YUY2ToUVRow = YUY2ToUVRow_LSX; |
1710 | | } |
1711 | | } |
1712 | | #endif |
1713 | | #if defined(HAS_YUY2TOYROW_LASX) && defined(HAS_YUY2TOUVROW_LASX) |
1714 | | if (TestCpuFlag(kCpuHasLASX)) { |
1715 | | YUY2ToYRow = YUY2ToYRow_Any_LASX; |
1716 | | YUY2ToUVRow = YUY2ToUVRow_Any_LASX; |
1717 | | if (IS_ALIGNED(width, 32)) { |
1718 | | YUY2ToYRow = YUY2ToYRow_LASX; |
1719 | | YUY2ToUVRow = YUY2ToUVRow_LASX; |
1720 | | } |
1721 | | } |
1722 | | #endif |
1723 | |
|
1724 | 0 | for (y = 0; y < height - 1; y += 2) { |
1725 | 0 | YUY2ToUVRow(src_yuy2, src_stride_yuy2, dst_u, dst_v, width); |
1726 | 0 | YUY2ToYRow(src_yuy2, dst_y, width); |
1727 | 0 | YUY2ToYRow(src_yuy2 + src_stride_yuy2, dst_y + dst_stride_y, width); |
1728 | 0 | src_yuy2 += src_stride_yuy2 * 2; |
1729 | 0 | dst_y += dst_stride_y * 2; |
1730 | 0 | dst_u += dst_stride_u; |
1731 | 0 | dst_v += dst_stride_v; |
1732 | 0 | } |
1733 | 0 | if (height & 1) { |
1734 | 0 | YUY2ToUVRow(src_yuy2, 0, dst_u, dst_v, width); |
1735 | 0 | YUY2ToYRow(src_yuy2, dst_y, width); |
1736 | 0 | } |
1737 | 0 | return 0; |
1738 | 0 | } |
1739 | | |
1740 | | // Convert UYVY to I420. |
1741 | | LIBYUV_API |
1742 | | int UYVYToI420(const uint8_t* src_uyvy, |
1743 | | int src_stride_uyvy, |
1744 | | uint8_t* dst_y, |
1745 | | int dst_stride_y, |
1746 | | uint8_t* dst_u, |
1747 | | int dst_stride_u, |
1748 | | uint8_t* dst_v, |
1749 | | int dst_stride_v, |
1750 | | int width, |
1751 | 0 | int height) { |
1752 | 0 | int y; |
1753 | 0 | void (*UYVYToUVRow)(const uint8_t* src_uyvy, int src_stride_uyvy, |
1754 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
1755 | 0 | UYVYToUVRow_C; |
1756 | 0 | void (*UYVYToYRow)(const uint8_t* src_uyvy, uint8_t* dst_y, int width) = |
1757 | 0 | UYVYToYRow_C; |
1758 | | // Negative height means invert the image. |
1759 | 0 | if (height < 0) { |
1760 | 0 | height = -height; |
1761 | 0 | src_uyvy = src_uyvy + (height - 1) * src_stride_uyvy; |
1762 | 0 | src_stride_uyvy = -src_stride_uyvy; |
1763 | 0 | } |
1764 | 0 | #if defined(HAS_UYVYTOYROW_SSE2) |
1765 | 0 | if (TestCpuFlag(kCpuHasSSE2)) { |
1766 | 0 | UYVYToUVRow = UYVYToUVRow_Any_SSE2; |
1767 | 0 | UYVYToYRow = UYVYToYRow_Any_SSE2; |
1768 | 0 | if (IS_ALIGNED(width, 16)) { |
1769 | 0 | UYVYToUVRow = UYVYToUVRow_SSE2; |
1770 | 0 | UYVYToYRow = UYVYToYRow_SSE2; |
1771 | 0 | } |
1772 | 0 | } |
1773 | 0 | #endif |
1774 | 0 | #if defined(HAS_UYVYTOYROW_AVX2) |
1775 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
1776 | 0 | UYVYToUVRow = UYVYToUVRow_Any_AVX2; |
1777 | 0 | UYVYToYRow = UYVYToYRow_Any_AVX2; |
1778 | 0 | if (IS_ALIGNED(width, 32)) { |
1779 | 0 | UYVYToUVRow = UYVYToUVRow_AVX2; |
1780 | 0 | UYVYToYRow = UYVYToYRow_AVX2; |
1781 | 0 | } |
1782 | 0 | } |
1783 | 0 | #endif |
1784 | | #if defined(HAS_UYVYTOYROW_NEON) |
1785 | | if (TestCpuFlag(kCpuHasNEON)) { |
1786 | | UYVYToYRow = UYVYToYRow_Any_NEON; |
1787 | | UYVYToUVRow = UYVYToUVRow_Any_NEON; |
1788 | | if (IS_ALIGNED(width, 16)) { |
1789 | | UYVYToYRow = UYVYToYRow_NEON; |
1790 | | UYVYToUVRow = UYVYToUVRow_NEON; |
1791 | | } |
1792 | | } |
1793 | | #endif |
1794 | | #if defined(HAS_UYVYTOYROW_LSX) |
1795 | | if (TestCpuFlag(kCpuHasLSX)) { |
1796 | | UYVYToYRow = UYVYToYRow_Any_LSX; |
1797 | | UYVYToUVRow = UYVYToUVRow_Any_LSX; |
1798 | | if (IS_ALIGNED(width, 16)) { |
1799 | | UYVYToYRow = UYVYToYRow_LSX; |
1800 | | UYVYToUVRow = UYVYToUVRow_LSX; |
1801 | | } |
1802 | | } |
1803 | | #endif |
1804 | | #if defined(HAS_UYVYTOYROW_LSX) |
1805 | | if (TestCpuFlag(kCpuHasLSX)) { |
1806 | | UYVYToYRow = UYVYToYRow_Any_LSX; |
1807 | | UYVYToUVRow = UYVYToUVRow_Any_LSX; |
1808 | | if (IS_ALIGNED(width, 16)) { |
1809 | | UYVYToYRow = UYVYToYRow_LSX; |
1810 | | UYVYToUVRow = UYVYToUVRow_LSX; |
1811 | | } |
1812 | | } |
1813 | | #endif |
1814 | | #if defined(HAS_UYVYTOYROW_LASX) |
1815 | | if (TestCpuFlag(kCpuHasLASX)) { |
1816 | | UYVYToYRow = UYVYToYRow_Any_LASX; |
1817 | | UYVYToUVRow = UYVYToUVRow_Any_LASX; |
1818 | | if (IS_ALIGNED(width, 32)) { |
1819 | | UYVYToYRow = UYVYToYRow_LASX; |
1820 | | UYVYToUVRow = UYVYToUVRow_LASX; |
1821 | | } |
1822 | | } |
1823 | | #endif |
1824 | |
|
1825 | 0 | for (y = 0; y < height - 1; y += 2) { |
1826 | 0 | UYVYToUVRow(src_uyvy, src_stride_uyvy, dst_u, dst_v, width); |
1827 | 0 | UYVYToYRow(src_uyvy, dst_y, width); |
1828 | 0 | UYVYToYRow(src_uyvy + src_stride_uyvy, dst_y + dst_stride_y, width); |
1829 | 0 | src_uyvy += src_stride_uyvy * 2; |
1830 | 0 | dst_y += dst_stride_y * 2; |
1831 | 0 | dst_u += dst_stride_u; |
1832 | 0 | dst_v += dst_stride_v; |
1833 | 0 | } |
1834 | 0 | if (height & 1) { |
1835 | 0 | UYVYToUVRow(src_uyvy, 0, dst_u, dst_v, width); |
1836 | 0 | UYVYToYRow(src_uyvy, dst_y, width); |
1837 | 0 | } |
1838 | 0 | return 0; |
1839 | 0 | } |
1840 | | |
1841 | | // Convert AYUV to NV12. |
1842 | | LIBYUV_API |
1843 | | int AYUVToNV12(const uint8_t* src_ayuv, |
1844 | | int src_stride_ayuv, |
1845 | | uint8_t* dst_y, |
1846 | | int dst_stride_y, |
1847 | | uint8_t* dst_uv, |
1848 | | int dst_stride_uv, |
1849 | | int width, |
1850 | 0 | int height) { |
1851 | 0 | int y; |
1852 | 0 | void (*AYUVToUVRow)(const uint8_t* src_ayuv, int src_stride_ayuv, |
1853 | 0 | uint8_t* dst_uv, int width) = AYUVToUVRow_C; |
1854 | 0 | void (*AYUVToYRow)(const uint8_t* src_ayuv, uint8_t* dst_y, int width) = |
1855 | 0 | AYUVToYRow_C; |
1856 | | // Negative height means invert the image. |
1857 | 0 | if (height < 0) { |
1858 | 0 | height = -height; |
1859 | 0 | src_ayuv = src_ayuv + (height - 1) * src_stride_ayuv; |
1860 | 0 | src_stride_ayuv = -src_stride_ayuv; |
1861 | 0 | } |
1862 | | // place holders for future intel code |
1863 | | #if defined(HAS_AYUVTOYROW_SSE2) |
1864 | | if (TestCpuFlag(kCpuHasSSE2)) { |
1865 | | AYUVToUVRow = AYUVToUVRow_Any_SSE2; |
1866 | | AYUVToYRow = AYUVToYRow_Any_SSE2; |
1867 | | if (IS_ALIGNED(width, 16)) { |
1868 | | AYUVToUVRow = AYUVToUVRow_SSE2; |
1869 | | AYUVToYRow = AYUVToYRow_SSE2; |
1870 | | } |
1871 | | } |
1872 | | #endif |
1873 | | #if defined(HAS_AYUVTOYROW_AVX2) |
1874 | | if (TestCpuFlag(kCpuHasAVX2)) { |
1875 | | AYUVToUVRow = AYUVToUVRow_Any_AVX2; |
1876 | | AYUVToYRow = AYUVToYRow_Any_AVX2; |
1877 | | if (IS_ALIGNED(width, 32)) { |
1878 | | AYUVToUVRow = AYUVToUVRow_AVX2; |
1879 | | AYUVToYRow = AYUVToYRow_AVX2; |
1880 | | } |
1881 | | } |
1882 | | #endif |
1883 | |
|
1884 | | #if defined(HAS_AYUVTOYROW_NEON) |
1885 | | if (TestCpuFlag(kCpuHasNEON)) { |
1886 | | AYUVToYRow = AYUVToYRow_Any_NEON; |
1887 | | AYUVToUVRow = AYUVToUVRow_Any_NEON; |
1888 | | if (IS_ALIGNED(width, 16)) { |
1889 | | AYUVToYRow = AYUVToYRow_NEON; |
1890 | | AYUVToUVRow = AYUVToUVRow_NEON; |
1891 | | } |
1892 | | } |
1893 | | #endif |
1894 | | #if defined(HAS_AYUVTOUVROW_SVE2) |
1895 | | if (TestCpuFlag(kCpuHasSVE2)) { |
1896 | | AYUVToUVRow = AYUVToUVRow_Any_SVE2; |
1897 | | if (IS_ALIGNED(width, 2)) { |
1898 | | AYUVToUVRow = AYUVToUVRow_SVE2; |
1899 | | } |
1900 | | } |
1901 | | #endif |
1902 | |
|
1903 | 0 | for (y = 0; y < height - 1; y += 2) { |
1904 | 0 | AYUVToUVRow(src_ayuv, src_stride_ayuv, dst_uv, width); |
1905 | 0 | AYUVToYRow(src_ayuv, dst_y, width); |
1906 | 0 | AYUVToYRow(src_ayuv + src_stride_ayuv, dst_y + dst_stride_y, width); |
1907 | 0 | src_ayuv += src_stride_ayuv * 2; |
1908 | 0 | dst_y += dst_stride_y * 2; |
1909 | 0 | dst_uv += dst_stride_uv; |
1910 | 0 | } |
1911 | 0 | if (height & 1) { |
1912 | 0 | AYUVToUVRow(src_ayuv, 0, dst_uv, width); |
1913 | 0 | AYUVToYRow(src_ayuv, dst_y, width); |
1914 | 0 | } |
1915 | 0 | return 0; |
1916 | 0 | } |
1917 | | |
1918 | | // Convert AYUV to NV21. |
1919 | | LIBYUV_API |
1920 | | int AYUVToNV21(const uint8_t* src_ayuv, |
1921 | | int src_stride_ayuv, |
1922 | | uint8_t* dst_y, |
1923 | | int dst_stride_y, |
1924 | | uint8_t* dst_vu, |
1925 | | int dst_stride_vu, |
1926 | | int width, |
1927 | 0 | int height) { |
1928 | 0 | int y; |
1929 | 0 | void (*AYUVToVURow)(const uint8_t* src_ayuv, int src_stride_ayuv, |
1930 | 0 | uint8_t* dst_vu, int width) = AYUVToVURow_C; |
1931 | 0 | void (*AYUVToYRow)(const uint8_t* src_ayuv, uint8_t* dst_y, int width) = |
1932 | 0 | AYUVToYRow_C; |
1933 | | // Negative height means invert the image. |
1934 | 0 | if (height < 0) { |
1935 | 0 | height = -height; |
1936 | 0 | src_ayuv = src_ayuv + (height - 1) * src_stride_ayuv; |
1937 | 0 | src_stride_ayuv = -src_stride_ayuv; |
1938 | 0 | } |
1939 | | // place holders for future intel code |
1940 | | #if defined(HAS_AYUVTOYROW_SSE2) |
1941 | | if (TestCpuFlag(kCpuHasSSE2)) { |
1942 | | AYUVToVURow = AYUVToVURow_Any_SSE2; |
1943 | | AYUVToYRow = AYUVToYRow_Any_SSE2; |
1944 | | if (IS_ALIGNED(width, 16)) { |
1945 | | AYUVToVURow = AYUVToVURow_SSE2; |
1946 | | AYUVToYRow = AYUVToYRow_SSE2; |
1947 | | } |
1948 | | } |
1949 | | #endif |
1950 | | #if defined(HAS_AYUVTOYROW_AVX2) |
1951 | | if (TestCpuFlag(kCpuHasAVX2)) { |
1952 | | AYUVToVURow = AYUVToVURow_Any_AVX2; |
1953 | | AYUVToYRow = AYUVToYRow_Any_AVX2; |
1954 | | if (IS_ALIGNED(width, 32)) { |
1955 | | AYUVToVURow = AYUVToVURow_AVX2; |
1956 | | AYUVToYRow = AYUVToYRow_AVX2; |
1957 | | } |
1958 | | } |
1959 | | #endif |
1960 | |
|
1961 | | #if defined(HAS_AYUVTOYROW_NEON) |
1962 | | if (TestCpuFlag(kCpuHasNEON)) { |
1963 | | AYUVToYRow = AYUVToYRow_Any_NEON; |
1964 | | AYUVToVURow = AYUVToVURow_Any_NEON; |
1965 | | if (IS_ALIGNED(width, 16)) { |
1966 | | AYUVToYRow = AYUVToYRow_NEON; |
1967 | | AYUVToVURow = AYUVToVURow_NEON; |
1968 | | } |
1969 | | } |
1970 | | #endif |
1971 | | #if defined(HAS_AYUVTOVUROW_SVE2) |
1972 | | if (TestCpuFlag(kCpuHasSVE2)) { |
1973 | | AYUVToVURow = AYUVToVURow_Any_SVE2; |
1974 | | if (IS_ALIGNED(width, 2)) { |
1975 | | AYUVToVURow = AYUVToVURow_SVE2; |
1976 | | } |
1977 | | } |
1978 | | #endif |
1979 | |
|
1980 | 0 | for (y = 0; y < height - 1; y += 2) { |
1981 | 0 | AYUVToVURow(src_ayuv, src_stride_ayuv, dst_vu, width); |
1982 | 0 | AYUVToYRow(src_ayuv, dst_y, width); |
1983 | 0 | AYUVToYRow(src_ayuv + src_stride_ayuv, dst_y + dst_stride_y, width); |
1984 | 0 | src_ayuv += src_stride_ayuv * 2; |
1985 | 0 | dst_y += dst_stride_y * 2; |
1986 | 0 | dst_vu += dst_stride_vu; |
1987 | 0 | } |
1988 | 0 | if (height & 1) { |
1989 | 0 | AYUVToVURow(src_ayuv, 0, dst_vu, width); |
1990 | 0 | AYUVToYRow(src_ayuv, dst_y, width); |
1991 | 0 | } |
1992 | 0 | return 0; |
1993 | 0 | } |
1994 | | |
1995 | | // Convert ARGB to I420. |
1996 | | LIBYUV_API |
1997 | | int ARGBToI420(const uint8_t* src_argb, |
1998 | | int src_stride_argb, |
1999 | | uint8_t* dst_y, |
2000 | | int dst_stride_y, |
2001 | | uint8_t* dst_u, |
2002 | | int dst_stride_u, |
2003 | | uint8_t* dst_v, |
2004 | | int dst_stride_v, |
2005 | | int width, |
2006 | 0 | int height) { |
2007 | 0 | int y; |
2008 | 0 | void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb, |
2009 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2010 | 0 | ARGBToUVRow_C; |
2011 | 0 | void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
2012 | 0 | ARGBToYRow_C; |
2013 | 0 | if (!src_argb || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
2014 | 0 | return -1; |
2015 | 0 | } |
2016 | | // Negative height means invert the image. |
2017 | 0 | if (height < 0) { |
2018 | 0 | height = -height; |
2019 | 0 | src_argb = src_argb + (height - 1) * src_stride_argb; |
2020 | 0 | src_stride_argb = -src_stride_argb; |
2021 | 0 | } |
2022 | | #if defined(HAS_ARGBTOYROW_NEON) |
2023 | | if (TestCpuFlag(kCpuHasNEON)) { |
2024 | | ARGBToYRow = ARGBToYRow_Any_NEON; |
2025 | | if (IS_ALIGNED(width, 16)) { |
2026 | | ARGBToYRow = ARGBToYRow_NEON; |
2027 | | } |
2028 | | } |
2029 | | #endif |
2030 | | #if defined(HAS_ARGBTOYROW_NEON_DOTPROD) |
2031 | | if (TestCpuFlag(kCpuHasNeonDotProd)) { |
2032 | | ARGBToYRow = ARGBToYRow_Any_NEON_DotProd; |
2033 | | if (IS_ALIGNED(width, 16)) { |
2034 | | ARGBToYRow = ARGBToYRow_NEON_DotProd; |
2035 | | } |
2036 | | } |
2037 | | #endif |
2038 | | #if defined(HAS_ARGBTOUVROW_NEON) |
2039 | | if (TestCpuFlag(kCpuHasNEON)) { |
2040 | | ARGBToUVRow = ARGBToUVRow_Any_NEON; |
2041 | | if (IS_ALIGNED(width, 16)) { |
2042 | | ARGBToUVRow = ARGBToUVRow_NEON; |
2043 | | } |
2044 | | } |
2045 | | #endif |
2046 | | #if defined(HAS_ARGBTOUVROW_NEON_I8MM) |
2047 | | if (TestCpuFlag(kCpuHasNeonI8MM)) { |
2048 | | ARGBToUVRow = ARGBToUVRow_Any_NEON_I8MM; |
2049 | | if (IS_ALIGNED(width, 16)) { |
2050 | | ARGBToUVRow = ARGBToUVRow_NEON_I8MM; |
2051 | | } |
2052 | | } |
2053 | | #endif |
2054 | | #if defined(HAS_ARGBTOUVROW_SVE2) |
2055 | | if (TestCpuFlag(kCpuHasSVE2)) { |
2056 | | ARGBToUVRow = ARGBToUVRow_Any_SVE2; |
2057 | | if (IS_ALIGNED(width, 2)) { |
2058 | | ARGBToUVRow = ARGBToUVRow_SVE2; |
2059 | | } |
2060 | | } |
2061 | | #endif |
2062 | | #if defined(HAS_ARGBTOUVROW_SME) |
2063 | | if (TestCpuFlag(kCpuHasSME)) { |
2064 | | ARGBToUVRow = ARGBToUVRow_Any_SME; |
2065 | | if (IS_ALIGNED(width, 2)) { |
2066 | | ARGBToUVRow = ARGBToUVRow_SME; |
2067 | | } |
2068 | | } |
2069 | | #endif |
2070 | 0 | #if defined(HAS_ARGBTOYROW_SSSE3) |
2071 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2072 | 0 | ARGBToYRow = ARGBToYRow_Any_SSSE3; |
2073 | 0 | if (IS_ALIGNED(width, 16)) { |
2074 | 0 | ARGBToYRow = ARGBToYRow_SSSE3; |
2075 | 0 | } |
2076 | 0 | } |
2077 | 0 | #endif |
2078 | 0 | #if defined(HAS_ARGBTOUVROW_SSSE3) |
2079 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2080 | 0 | ARGBToUVRow = ARGBToUVRow_Any_SSSE3; |
2081 | 0 | if (IS_ALIGNED(width, 16)) { |
2082 | 0 | ARGBToUVRow = ARGBToUVRow_SSSE3; |
2083 | 0 | } |
2084 | 0 | } |
2085 | 0 | #endif |
2086 | 0 | #if defined(HAS_ARGBTOYROW_AVX2) |
2087 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2088 | 0 | ARGBToYRow = ARGBToYRow_Any_AVX2; |
2089 | 0 | if (IS_ALIGNED(width, 32)) { |
2090 | 0 | ARGBToYRow = ARGBToYRow_AVX2; |
2091 | 0 | } |
2092 | 0 | } |
2093 | 0 | #endif |
2094 | 0 | #if defined(HAS_ARGBTOUVROW_AVX2) |
2095 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2096 | 0 | ARGBToUVRow = ARGBToUVRow_Any_AVX2; |
2097 | 0 | if (IS_ALIGNED(width, 32)) { |
2098 | 0 | ARGBToUVRow = ARGBToUVRow_AVX2; |
2099 | 0 | } |
2100 | 0 | } |
2101 | 0 | #endif |
2102 | | #if defined(HAS_ARGBTOYROW_LSX) |
2103 | | if (TestCpuFlag(kCpuHasLSX)) { |
2104 | | ARGBToYRow = ARGBToYRow_Any_LSX; |
2105 | | if (IS_ALIGNED(width, 16)) { |
2106 | | ARGBToYRow = ARGBToYRow_LSX; |
2107 | | } |
2108 | | } |
2109 | | #endif |
2110 | | #if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX) |
2111 | | if (TestCpuFlag(kCpuHasLSX)) { |
2112 | | ARGBToYRow = ARGBToYRow_Any_LSX; |
2113 | | ARGBToUVRow = ARGBToUVRow_Any_LSX; |
2114 | | if (IS_ALIGNED(width, 16)) { |
2115 | | ARGBToYRow = ARGBToYRow_LSX; |
2116 | | ARGBToUVRow = ARGBToUVRow_LSX; |
2117 | | } |
2118 | | } |
2119 | | #endif |
2120 | | #if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX) |
2121 | | if (TestCpuFlag(kCpuHasLASX)) { |
2122 | | ARGBToYRow = ARGBToYRow_Any_LASX; |
2123 | | ARGBToUVRow = ARGBToUVRow_Any_LASX; |
2124 | | if (IS_ALIGNED(width, 32)) { |
2125 | | ARGBToYRow = ARGBToYRow_LASX; |
2126 | | ARGBToUVRow = ARGBToUVRow_LASX; |
2127 | | } |
2128 | | } |
2129 | | #endif |
2130 | |
|
2131 | 0 | for (y = 0; y < height - 1; y += 2) { |
2132 | 0 | ARGBToUVRow(src_argb, src_stride_argb, dst_u, dst_v, width); |
2133 | 0 | ARGBToYRow(src_argb, dst_y, width); |
2134 | 0 | ARGBToYRow(src_argb + src_stride_argb, dst_y + dst_stride_y, width); |
2135 | 0 | src_argb += src_stride_argb * 2; |
2136 | 0 | dst_y += dst_stride_y * 2; |
2137 | 0 | dst_u += dst_stride_u; |
2138 | 0 | dst_v += dst_stride_v; |
2139 | 0 | } |
2140 | 0 | if (height & 1) { |
2141 | 0 | ARGBToUVRow(src_argb, 0, dst_u, dst_v, width); |
2142 | 0 | ARGBToYRow(src_argb, dst_y, width); |
2143 | 0 | } |
2144 | 0 | return 0; |
2145 | 0 | } |
2146 | | |
2147 | | #ifdef USE_EXTRACTALPHA |
2148 | | // Convert ARGB to I420 with Alpha |
2149 | | // The following version calls ARGBExtractAlpha on the full image. |
2150 | | LIBYUV_API |
2151 | | int ARGBToI420Alpha(const uint8_t* src_argb, |
2152 | | int src_stride_argb, |
2153 | | uint8_t* dst_y, |
2154 | | int dst_stride_y, |
2155 | | uint8_t* dst_u, |
2156 | | int dst_stride_u, |
2157 | | uint8_t* dst_v, |
2158 | | int dst_stride_v, |
2159 | | uint8_t* dst_a, |
2160 | | int dst_stride_a, |
2161 | | int width, |
2162 | | int height) { |
2163 | | int r = ARGBToI420(src_argb, src_stride_argb, dst_y, dst_stride_y, dst_u, |
2164 | | dst_stride_u, dst_v, dst_stride_v, width, height); |
2165 | | if (r == 0) { |
2166 | | r = ARGBExtractAlpha(src_argb, src_stride_argb, dst_a, dst_stride_a, width, |
2167 | | height); |
2168 | | } |
2169 | | return r; |
2170 | | } |
2171 | | #else // USE_EXTRACTALPHA |
2172 | | // Convert ARGB to I420 with Alpha |
2173 | | LIBYUV_API |
2174 | | int ARGBToI420Alpha(const uint8_t* src_argb, |
2175 | | int src_stride_argb, |
2176 | | uint8_t* dst_y, |
2177 | | int dst_stride_y, |
2178 | | uint8_t* dst_u, |
2179 | | int dst_stride_u, |
2180 | | uint8_t* dst_v, |
2181 | | int dst_stride_v, |
2182 | | uint8_t* dst_a, |
2183 | | int dst_stride_a, |
2184 | | int width, |
2185 | 0 | int height) { |
2186 | 0 | int y; |
2187 | 0 | void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb, |
2188 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2189 | 0 | ARGBToUVRow_C; |
2190 | 0 | void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
2191 | 0 | ARGBToYRow_C; |
2192 | 0 | void (*ARGBExtractAlphaRow)(const uint8_t* src_argb, uint8_t* dst_a, |
2193 | 0 | int width) = ARGBExtractAlphaRow_C; |
2194 | 0 | if (!src_argb || !dst_y || !dst_u || !dst_v || !dst_a || width <= 0 || |
2195 | 0 | height == 0) { |
2196 | 0 | return -1; |
2197 | 0 | } |
2198 | | // Negative height means invert the image. |
2199 | 0 | if (height < 0) { |
2200 | 0 | height = -height; |
2201 | 0 | src_argb = src_argb + (height - 1) * src_stride_argb; |
2202 | 0 | src_stride_argb = -src_stride_argb; |
2203 | 0 | } |
2204 | | #if defined(HAS_ARGBTOYROW_NEON) |
2205 | | if (TestCpuFlag(kCpuHasNEON)) { |
2206 | | ARGBToYRow = ARGBToYRow_Any_NEON; |
2207 | | if (IS_ALIGNED(width, 16)) { |
2208 | | ARGBToYRow = ARGBToYRow_NEON; |
2209 | | } |
2210 | | } |
2211 | | #endif |
2212 | | #if defined(HAS_ARGBTOYROW_NEON_DOTPROD) |
2213 | | if (TestCpuFlag(kCpuHasNeonDotProd)) { |
2214 | | ARGBToYRow = ARGBToYRow_Any_NEON_DotProd; |
2215 | | if (IS_ALIGNED(width, 16)) { |
2216 | | ARGBToYRow = ARGBToYRow_NEON_DotProd; |
2217 | | } |
2218 | | } |
2219 | | #endif |
2220 | | #if defined(HAS_ARGBTOUVROW_NEON) |
2221 | | if (TestCpuFlag(kCpuHasNEON)) { |
2222 | | ARGBToUVRow = ARGBToUVRow_Any_NEON; |
2223 | | if (IS_ALIGNED(width, 16)) { |
2224 | | ARGBToUVRow = ARGBToUVRow_NEON; |
2225 | | } |
2226 | | } |
2227 | | #endif |
2228 | | #if defined(HAS_ARGBTOUVROW_NEON_I8MM) |
2229 | | if (TestCpuFlag(kCpuHasNeonI8MM)) { |
2230 | | ARGBToUVRow = ARGBToUVRow_Any_NEON_I8MM; |
2231 | | if (IS_ALIGNED(width, 16)) { |
2232 | | ARGBToUVRow = ARGBToUVRow_NEON_I8MM; |
2233 | | } |
2234 | | } |
2235 | | #endif |
2236 | | #if defined(HAS_ARGBTOUVROW_SVE2) |
2237 | | if (TestCpuFlag(kCpuHasSVE2)) { |
2238 | | ARGBToUVRow = ARGBToUVRow_Any_SVE2; |
2239 | | if (IS_ALIGNED(width, 2)) { |
2240 | | ARGBToUVRow = ARGBToUVRow_SVE2; |
2241 | | } |
2242 | | } |
2243 | | #endif |
2244 | | #if defined(HAS_ARGBTOUVROW_SME) |
2245 | | if (TestCpuFlag(kCpuHasSME)) { |
2246 | | ARGBToUVRow = ARGBToUVRow_Any_SME; |
2247 | | if (IS_ALIGNED(width, 2)) { |
2248 | | ARGBToUVRow = ARGBToUVRow_SME; |
2249 | | } |
2250 | | } |
2251 | | #endif |
2252 | 0 | #if defined(HAS_ARGBTOYROW_SSSE3) |
2253 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2254 | 0 | ARGBToYRow = ARGBToYRow_Any_SSSE3; |
2255 | 0 | if (IS_ALIGNED(width, 16)) { |
2256 | 0 | ARGBToYRow = ARGBToYRow_SSSE3; |
2257 | 0 | } |
2258 | 0 | } |
2259 | 0 | #endif |
2260 | 0 | #if defined(HAS_ARGBTOUVROW_SSSE3) |
2261 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2262 | 0 | ARGBToUVRow = ARGBToUVRow_Any_SSSE3; |
2263 | 0 | if (IS_ALIGNED(width, 16)) { |
2264 | 0 | ARGBToUVRow = ARGBToUVRow_SSSE3; |
2265 | 0 | } |
2266 | 0 | } |
2267 | 0 | #endif |
2268 | 0 | #if defined(HAS_ARGBTOYROW_AVX2) |
2269 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2270 | 0 | ARGBToYRow = ARGBToYRow_Any_AVX2; |
2271 | 0 | if (IS_ALIGNED(width, 32)) { |
2272 | 0 | ARGBToYRow = ARGBToYRow_AVX2; |
2273 | 0 | } |
2274 | 0 | } |
2275 | 0 | #endif |
2276 | 0 | #if defined(HAS_ARGBTOUVROW_AVX2) |
2277 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2278 | 0 | ARGBToUVRow = ARGBToUVRow_Any_AVX2; |
2279 | 0 | if (IS_ALIGNED(width, 32)) { |
2280 | 0 | ARGBToUVRow = ARGBToUVRow_AVX2; |
2281 | 0 | } |
2282 | 0 | } |
2283 | 0 | #endif |
2284 | | #if defined(HAS_ARGBTOYROW_LSX) |
2285 | | if (TestCpuFlag(kCpuHasLSX)) { |
2286 | | ARGBToYRow = ARGBToYRow_Any_LSX; |
2287 | | if (IS_ALIGNED(width, 16)) { |
2288 | | ARGBToYRow = ARGBToYRow_LSX; |
2289 | | } |
2290 | | } |
2291 | | #endif |
2292 | | #if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX) |
2293 | | if (TestCpuFlag(kCpuHasLASX)) { |
2294 | | ARGBToYRow = ARGBToYRow_Any_LASX; |
2295 | | ARGBToUVRow = ARGBToUVRow_Any_LASX; |
2296 | | if (IS_ALIGNED(width, 32)) { |
2297 | | ARGBToYRow = ARGBToYRow_LASX; |
2298 | | ARGBToUVRow = ARGBToUVRow_LASX; |
2299 | | } |
2300 | | } |
2301 | | #endif |
2302 | 0 | #if defined(HAS_ARGBEXTRACTALPHAROW_SSE2) |
2303 | 0 | if (TestCpuFlag(kCpuHasSSE2)) { |
2304 | 0 | ARGBExtractAlphaRow = IS_ALIGNED(width, 8) ? ARGBExtractAlphaRow_SSE2 |
2305 | 0 | : ARGBExtractAlphaRow_Any_SSE2; |
2306 | 0 | } |
2307 | 0 | #endif |
2308 | 0 | #if defined(HAS_ARGBEXTRACTALPHAROW_AVX2) |
2309 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2310 | 0 | ARGBExtractAlphaRow = IS_ALIGNED(width, 32) ? ARGBExtractAlphaRow_AVX2 |
2311 | 0 | : ARGBExtractAlphaRow_Any_AVX2; |
2312 | 0 | } |
2313 | 0 | #endif |
2314 | | #if defined(HAS_ARGBEXTRACTALPHAROW_NEON) |
2315 | | if (TestCpuFlag(kCpuHasNEON)) { |
2316 | | ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_NEON |
2317 | | : ARGBExtractAlphaRow_Any_NEON; |
2318 | | } |
2319 | | #endif |
2320 | | #if defined(HAS_ARGBEXTRACTALPHAROW_LSX) |
2321 | | if (TestCpuFlag(kCpuHasLSX)) { |
2322 | | ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_LSX |
2323 | | : ARGBExtractAlphaRow_Any_LSX; |
2324 | | } |
2325 | | #endif |
2326 | | #if defined(HAS_ARGBEXTRACTALPHAROW_RVV) |
2327 | | if (TestCpuFlag(kCpuHasRVV)) { |
2328 | | ARGBExtractAlphaRow = ARGBExtractAlphaRow_RVV; |
2329 | | } |
2330 | | #endif |
2331 | |
|
2332 | 0 | for (y = 0; y < height - 1; y += 2) { |
2333 | 0 | ARGBToUVRow(src_argb, src_stride_argb, dst_u, dst_v, width); |
2334 | 0 | ARGBToYRow(src_argb, dst_y, width); |
2335 | 0 | ARGBToYRow(src_argb + src_stride_argb, dst_y + dst_stride_y, width); |
2336 | 0 | ARGBExtractAlphaRow(src_argb, dst_a, width); |
2337 | 0 | ARGBExtractAlphaRow(src_argb + src_stride_argb, dst_a + dst_stride_a, |
2338 | 0 | width); |
2339 | 0 | src_argb += src_stride_argb * 2; |
2340 | 0 | dst_y += dst_stride_y * 2; |
2341 | 0 | dst_u += dst_stride_u; |
2342 | 0 | dst_v += dst_stride_v; |
2343 | 0 | dst_a += dst_stride_a * 2; |
2344 | 0 | } |
2345 | 0 | if (height & 1) { |
2346 | 0 | ARGBToUVRow(src_argb, 0, dst_u, dst_v, width); |
2347 | 0 | ARGBToYRow(src_argb, dst_y, width); |
2348 | 0 | ARGBExtractAlphaRow(src_argb, dst_a, width); |
2349 | 0 | } |
2350 | 0 | return 0; |
2351 | 0 | } |
2352 | | #endif // USE_EXTRACTALPHA |
2353 | | |
2354 | | // Convert BGRA to I420. |
2355 | | LIBYUV_API |
2356 | | int BGRAToI420(const uint8_t* src_bgra, |
2357 | | int src_stride_bgra, |
2358 | | uint8_t* dst_y, |
2359 | | int dst_stride_y, |
2360 | | uint8_t* dst_u, |
2361 | | int dst_stride_u, |
2362 | | uint8_t* dst_v, |
2363 | | int dst_stride_v, |
2364 | | int width, |
2365 | 0 | int height) { |
2366 | 0 | int y; |
2367 | 0 | void (*BGRAToUVRow)(const uint8_t* src_bgra0, int src_stride_bgra, |
2368 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2369 | 0 | BGRAToUVRow_C; |
2370 | 0 | void (*BGRAToYRow)(const uint8_t* src_bgra, uint8_t* dst_y, int width) = |
2371 | 0 | BGRAToYRow_C; |
2372 | 0 | if (!src_bgra || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
2373 | 0 | return -1; |
2374 | 0 | } |
2375 | | // Negative height means invert the image. |
2376 | 0 | if (height < 0) { |
2377 | 0 | height = -height; |
2378 | 0 | src_bgra = src_bgra + (height - 1) * src_stride_bgra; |
2379 | 0 | src_stride_bgra = -src_stride_bgra; |
2380 | 0 | } |
2381 | | #if defined(HAS_BGRATOYROW_NEON) |
2382 | | if (TestCpuFlag(kCpuHasNEON)) { |
2383 | | BGRAToYRow = BGRAToYRow_Any_NEON; |
2384 | | if (IS_ALIGNED(width, 16)) { |
2385 | | BGRAToYRow = BGRAToYRow_NEON; |
2386 | | } |
2387 | | } |
2388 | | #endif |
2389 | | #if defined(HAS_BGRATOYROW_NEON_DOTPROD) |
2390 | | if (TestCpuFlag(kCpuHasNeonDotProd)) { |
2391 | | BGRAToYRow = BGRAToYRow_Any_NEON_DotProd; |
2392 | | if (IS_ALIGNED(width, 16)) { |
2393 | | BGRAToYRow = BGRAToYRow_NEON_DotProd; |
2394 | | } |
2395 | | } |
2396 | | #endif |
2397 | | #if defined(HAS_BGRATOUVROW_NEON) |
2398 | | if (TestCpuFlag(kCpuHasNEON)) { |
2399 | | BGRAToUVRow = BGRAToUVRow_Any_NEON; |
2400 | | if (IS_ALIGNED(width, 16)) { |
2401 | | BGRAToUVRow = BGRAToUVRow_NEON; |
2402 | | } |
2403 | | } |
2404 | | #endif |
2405 | | #if defined(HAS_BGRATOUVROW_NEON_I8MM) |
2406 | | if (TestCpuFlag(kCpuHasNeonI8MM)) { |
2407 | | BGRAToUVRow = BGRAToUVRow_Any_NEON_I8MM; |
2408 | | if (IS_ALIGNED(width, 16)) { |
2409 | | BGRAToUVRow = BGRAToUVRow_NEON_I8MM; |
2410 | | } |
2411 | | } |
2412 | | #endif |
2413 | | #if defined(HAS_BGRATOUVROW_SVE2) |
2414 | | if (TestCpuFlag(kCpuHasSVE2)) { |
2415 | | BGRAToUVRow = BGRAToUVRow_Any_SVE2; |
2416 | | if (IS_ALIGNED(width, 2)) { |
2417 | | BGRAToUVRow = BGRAToUVRow_SVE2; |
2418 | | } |
2419 | | } |
2420 | | #endif |
2421 | | #if defined(HAS_BGRATOUVROW_SME) |
2422 | | if (TestCpuFlag(kCpuHasSME)) { |
2423 | | BGRAToUVRow = BGRAToUVRow_Any_SME; |
2424 | | if (IS_ALIGNED(width, 2)) { |
2425 | | BGRAToUVRow = BGRAToUVRow_SME; |
2426 | | } |
2427 | | } |
2428 | | #endif |
2429 | 0 | #if defined(HAS_BGRATOYROW_SSSE3) |
2430 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2431 | 0 | BGRAToYRow = BGRAToYRow_Any_SSSE3; |
2432 | 0 | if (IS_ALIGNED(width, 16)) { |
2433 | 0 | BGRAToYRow = BGRAToYRow_SSSE3; |
2434 | 0 | } |
2435 | 0 | } |
2436 | 0 | #endif |
2437 | 0 | #if defined(HAS_BGRATOUVROW_SSSE3) |
2438 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2439 | 0 | BGRAToUVRow = BGRAToUVRow_Any_SSSE3; |
2440 | 0 | if (IS_ALIGNED(width, 16)) { |
2441 | 0 | BGRAToUVRow = BGRAToUVRow_SSSE3; |
2442 | 0 | } |
2443 | 0 | } |
2444 | 0 | #endif |
2445 | | #if defined(HAS_BGRATOYROW_AVX2) |
2446 | | if (TestCpuFlag(kCpuHasAVX2)) { |
2447 | | BGRAToYRow = BGRAToYRow_Any_AVX2; |
2448 | | if (IS_ALIGNED(width, 32)) { |
2449 | | BGRAToYRow = BGRAToYRow_AVX2; |
2450 | | } |
2451 | | } |
2452 | | #endif |
2453 | | #if defined(HAS_BGRATOUVROW_AVX2) |
2454 | | if (TestCpuFlag(kCpuHasAVX2)) { |
2455 | | BGRAToUVRow = BGRAToUVRow_Any_AVX2; |
2456 | | if (IS_ALIGNED(width, 32)) { |
2457 | | BGRAToUVRow = BGRAToUVRow_AVX2; |
2458 | | } |
2459 | | } |
2460 | | #endif |
2461 | | #if defined(HAS_BGRATOYROW_LSX) && defined(HAS_BGRATOUVROW_LSX) |
2462 | | if (TestCpuFlag(kCpuHasLSX)) { |
2463 | | BGRAToYRow = BGRAToYRow_Any_LSX; |
2464 | | BGRAToUVRow = BGRAToUVRow_Any_LSX; |
2465 | | if (IS_ALIGNED(width, 16)) { |
2466 | | BGRAToYRow = BGRAToYRow_LSX; |
2467 | | BGRAToUVRow = BGRAToUVRow_LSX; |
2468 | | } |
2469 | | } |
2470 | | #endif |
2471 | | #if defined(HAS_BGRATOYROW_LASX) |
2472 | | if (TestCpuFlag(kCpuHasLASX)) { |
2473 | | BGRAToYRow = BGRAToYRow_Any_LASX; |
2474 | | if (IS_ALIGNED(width, 32)) { |
2475 | | BGRAToYRow = BGRAToYRow_LASX; |
2476 | | } |
2477 | | } |
2478 | | #endif |
2479 | | #if defined(HAS_BGRATOYROW_RVV) |
2480 | | if (TestCpuFlag(kCpuHasRVV)) { |
2481 | | BGRAToYRow = BGRAToYRow_RVV; |
2482 | | } |
2483 | | #endif |
2484 | |
|
2485 | 0 | for (y = 0; y < height - 1; y += 2) { |
2486 | 0 | BGRAToUVRow(src_bgra, src_stride_bgra, dst_u, dst_v, width); |
2487 | 0 | BGRAToYRow(src_bgra, dst_y, width); |
2488 | 0 | BGRAToYRow(src_bgra + src_stride_bgra, dst_y + dst_stride_y, width); |
2489 | 0 | src_bgra += src_stride_bgra * 2; |
2490 | 0 | dst_y += dst_stride_y * 2; |
2491 | 0 | dst_u += dst_stride_u; |
2492 | 0 | dst_v += dst_stride_v; |
2493 | 0 | } |
2494 | 0 | if (height & 1) { |
2495 | 0 | BGRAToUVRow(src_bgra, 0, dst_u, dst_v, width); |
2496 | 0 | BGRAToYRow(src_bgra, dst_y, width); |
2497 | 0 | } |
2498 | 0 | return 0; |
2499 | 0 | } |
2500 | | |
2501 | | // Convert ABGR to I420. |
2502 | | LIBYUV_API |
2503 | | int ABGRToI420(const uint8_t* src_abgr, |
2504 | | int src_stride_abgr, |
2505 | | uint8_t* dst_y, |
2506 | | int dst_stride_y, |
2507 | | uint8_t* dst_u, |
2508 | | int dst_stride_u, |
2509 | | uint8_t* dst_v, |
2510 | | int dst_stride_v, |
2511 | | int width, |
2512 | 0 | int height) { |
2513 | 0 | int y; |
2514 | 0 | void (*ABGRToUVRow)(const uint8_t* src_abgr0, int src_stride_abgr, |
2515 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2516 | 0 | ABGRToUVRow_C; |
2517 | 0 | void (*ABGRToYRow)(const uint8_t* src_abgr, uint8_t* dst_y, int width) = |
2518 | 0 | ABGRToYRow_C; |
2519 | 0 | if (!src_abgr || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
2520 | 0 | return -1; |
2521 | 0 | } |
2522 | | // Negative height means invert the image. |
2523 | 0 | if (height < 0) { |
2524 | 0 | height = -height; |
2525 | 0 | src_abgr = src_abgr + (height - 1) * src_stride_abgr; |
2526 | 0 | src_stride_abgr = -src_stride_abgr; |
2527 | 0 | } |
2528 | 0 | #if defined(HAS_ABGRTOYROW_SSSE3) |
2529 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2530 | 0 | ABGRToYRow = ABGRToYRow_Any_SSSE3; |
2531 | 0 | if (IS_ALIGNED(width, 16)) { |
2532 | 0 | ABGRToYRow = ABGRToYRow_SSSE3; |
2533 | 0 | } |
2534 | 0 | } |
2535 | 0 | #endif |
2536 | 0 | #if defined(HAS_ABGRTOUVROW_SSSE3) |
2537 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2538 | 0 | ABGRToUVRow = ABGRToUVRow_Any_SSSE3; |
2539 | 0 | if (IS_ALIGNED(width, 16)) { |
2540 | 0 | ABGRToUVRow = ABGRToUVRow_SSSE3; |
2541 | 0 | } |
2542 | 0 | } |
2543 | 0 | #endif |
2544 | 0 | #if defined(HAS_ABGRTOYROW_AVX2) |
2545 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2546 | 0 | ABGRToYRow = ABGRToYRow_Any_AVX2; |
2547 | 0 | if (IS_ALIGNED(width, 32)) { |
2548 | 0 | ABGRToYRow = ABGRToYRow_AVX2; |
2549 | 0 | } |
2550 | 0 | } |
2551 | 0 | #endif |
2552 | 0 | #if defined(HAS_ABGRTOUVROW_AVX2) |
2553 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2554 | 0 | ABGRToUVRow = ABGRToUVRow_Any_AVX2; |
2555 | 0 | if (IS_ALIGNED(width, 32)) { |
2556 | 0 | ABGRToUVRow = ABGRToUVRow_AVX2; |
2557 | 0 | } |
2558 | 0 | } |
2559 | 0 | #endif |
2560 | | #if defined(HAS_ABGRTOYROW_NEON) |
2561 | | if (TestCpuFlag(kCpuHasNEON)) { |
2562 | | ABGRToYRow = ABGRToYRow_Any_NEON; |
2563 | | if (IS_ALIGNED(width, 16)) { |
2564 | | ABGRToYRow = ABGRToYRow_NEON; |
2565 | | } |
2566 | | } |
2567 | | #endif |
2568 | | #if defined(HAS_ABGRTOYROW_NEON_DOTPROD) |
2569 | | if (TestCpuFlag(kCpuHasNeonDotProd)) { |
2570 | | ABGRToYRow = ABGRToYRow_Any_NEON_DotProd; |
2571 | | if (IS_ALIGNED(width, 16)) { |
2572 | | ABGRToYRow = ABGRToYRow_NEON_DotProd; |
2573 | | } |
2574 | | } |
2575 | | #endif |
2576 | | #if defined(HAS_ABGRTOUVROW_NEON) |
2577 | | if (TestCpuFlag(kCpuHasNEON)) { |
2578 | | ABGRToUVRow = ABGRToUVRow_Any_NEON; |
2579 | | if (IS_ALIGNED(width, 16)) { |
2580 | | ABGRToUVRow = ABGRToUVRow_NEON; |
2581 | | } |
2582 | | } |
2583 | | #endif |
2584 | | #if defined(HAS_ABGRTOUVROW_NEON_I8MM) |
2585 | | if (TestCpuFlag(kCpuHasNeonI8MM)) { |
2586 | | ABGRToUVRow = ABGRToUVRow_Any_NEON_I8MM; |
2587 | | if (IS_ALIGNED(width, 16)) { |
2588 | | ABGRToUVRow = ABGRToUVRow_NEON_I8MM; |
2589 | | } |
2590 | | } |
2591 | | #endif |
2592 | | #if defined(HAS_ABGRTOUVROW_SVE2) |
2593 | | if (TestCpuFlag(kCpuHasSVE2)) { |
2594 | | ABGRToUVRow = ABGRToUVRow_Any_SVE2; |
2595 | | if (IS_ALIGNED(width, 2)) { |
2596 | | ABGRToUVRow = ABGRToUVRow_SVE2; |
2597 | | } |
2598 | | } |
2599 | | #endif |
2600 | | #if defined(HAS_ABGRTOUVROW_SME) |
2601 | | if (TestCpuFlag(kCpuHasSME)) { |
2602 | | ABGRToUVRow = ABGRToUVRow_Any_SME; |
2603 | | if (IS_ALIGNED(width, 2)) { |
2604 | | ABGRToUVRow = ABGRToUVRow_SME; |
2605 | | } |
2606 | | } |
2607 | | #endif |
2608 | | #if defined(HAS_ABGRTOYROW_LSX) && defined(HAS_ABGRTOUVROW_LSX) |
2609 | | if (TestCpuFlag(kCpuHasLSX)) { |
2610 | | ABGRToYRow = ABGRToYRow_Any_LSX; |
2611 | | ABGRToUVRow = ABGRToUVRow_Any_LSX; |
2612 | | if (IS_ALIGNED(width, 16)) { |
2613 | | ABGRToYRow = ABGRToYRow_LSX; |
2614 | | ABGRToUVRow = ABGRToUVRow_LSX; |
2615 | | } |
2616 | | } |
2617 | | #endif |
2618 | | #if defined(HAS_ABGRTOYROW_LASX) |
2619 | | if (TestCpuFlag(kCpuHasLASX)) { |
2620 | | ABGRToYRow = ABGRToYRow_Any_LASX; |
2621 | | if (IS_ALIGNED(width, 32)) { |
2622 | | ABGRToYRow = ABGRToYRow_LASX; |
2623 | | } |
2624 | | } |
2625 | | #endif |
2626 | | #if defined(HAS_ABGRTOYROW_RVV) |
2627 | | if (TestCpuFlag(kCpuHasRVV)) { |
2628 | | ABGRToYRow = ABGRToYRow_RVV; |
2629 | | } |
2630 | | #endif |
2631 | |
|
2632 | 0 | for (y = 0; y < height - 1; y += 2) { |
2633 | 0 | ABGRToUVRow(src_abgr, src_stride_abgr, dst_u, dst_v, width); |
2634 | 0 | ABGRToYRow(src_abgr, dst_y, width); |
2635 | 0 | ABGRToYRow(src_abgr + src_stride_abgr, dst_y + dst_stride_y, width); |
2636 | 0 | src_abgr += src_stride_abgr * 2; |
2637 | 0 | dst_y += dst_stride_y * 2; |
2638 | 0 | dst_u += dst_stride_u; |
2639 | 0 | dst_v += dst_stride_v; |
2640 | 0 | } |
2641 | 0 | if (height & 1) { |
2642 | 0 | ABGRToUVRow(src_abgr, 0, dst_u, dst_v, width); |
2643 | 0 | ABGRToYRow(src_abgr, dst_y, width); |
2644 | 0 | } |
2645 | 0 | return 0; |
2646 | 0 | } |
2647 | | |
2648 | | // Convert RGBA to I420. |
2649 | | LIBYUV_API |
2650 | | int RGBAToI420(const uint8_t* src_rgba, |
2651 | | int src_stride_rgba, |
2652 | | uint8_t* dst_y, |
2653 | | int dst_stride_y, |
2654 | | uint8_t* dst_u, |
2655 | | int dst_stride_u, |
2656 | | uint8_t* dst_v, |
2657 | | int dst_stride_v, |
2658 | | int width, |
2659 | 0 | int height) { |
2660 | 0 | int y; |
2661 | 0 | void (*RGBAToUVRow)(const uint8_t* src_rgba0, int src_stride_rgba, |
2662 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2663 | 0 | RGBAToUVRow_C; |
2664 | 0 | void (*RGBAToYRow)(const uint8_t* src_rgba, uint8_t* dst_y, int width) = |
2665 | 0 | RGBAToYRow_C; |
2666 | 0 | if (!src_rgba || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
2667 | 0 | return -1; |
2668 | 0 | } |
2669 | | // Negative height means invert the image. |
2670 | 0 | if (height < 0) { |
2671 | 0 | height = -height; |
2672 | 0 | src_rgba = src_rgba + (height - 1) * src_stride_rgba; |
2673 | 0 | src_stride_rgba = -src_stride_rgba; |
2674 | 0 | } |
2675 | 0 | #if defined(HAS_RGBATOYROW_SSSE3) |
2676 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2677 | 0 | RGBAToYRow = RGBAToYRow_Any_SSSE3; |
2678 | 0 | if (IS_ALIGNED(width, 16)) { |
2679 | 0 | RGBAToYRow = RGBAToYRow_SSSE3; |
2680 | 0 | } |
2681 | 0 | } |
2682 | 0 | #endif |
2683 | 0 | #if defined(HAS_RGBATOUVROW_SSSE3) |
2684 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2685 | 0 | RGBAToUVRow = RGBAToUVRow_Any_SSSE3; |
2686 | 0 | if (IS_ALIGNED(width, 16)) { |
2687 | 0 | RGBAToUVRow = RGBAToUVRow_SSSE3; |
2688 | 0 | } |
2689 | 0 | } |
2690 | 0 | #endif |
2691 | | #if defined(HAS_RGBATOYROW_NEON) |
2692 | | if (TestCpuFlag(kCpuHasNEON)) { |
2693 | | RGBAToYRow = RGBAToYRow_Any_NEON; |
2694 | | if (IS_ALIGNED(width, 16)) { |
2695 | | RGBAToYRow = RGBAToYRow_NEON; |
2696 | | } |
2697 | | } |
2698 | | #endif |
2699 | | #if defined(HAS_RGBATOYROW_NEON_DOTPROD) |
2700 | | if (TestCpuFlag(kCpuHasNeonDotProd)) { |
2701 | | RGBAToYRow = RGBAToYRow_Any_NEON_DotProd; |
2702 | | if (IS_ALIGNED(width, 16)) { |
2703 | | RGBAToYRow = RGBAToYRow_NEON_DotProd; |
2704 | | } |
2705 | | } |
2706 | | #endif |
2707 | | #if defined(HAS_RGBATOUVROW_NEON) |
2708 | | if (TestCpuFlag(kCpuHasNEON)) { |
2709 | | RGBAToUVRow = RGBAToUVRow_Any_NEON; |
2710 | | if (IS_ALIGNED(width, 16)) { |
2711 | | RGBAToUVRow = RGBAToUVRow_NEON; |
2712 | | } |
2713 | | } |
2714 | | #endif |
2715 | | #if defined(HAS_RGBATOUVROW_NEON_I8MM) |
2716 | | if (TestCpuFlag(kCpuHasNeonI8MM)) { |
2717 | | RGBAToUVRow = RGBAToUVRow_Any_NEON_I8MM; |
2718 | | if (IS_ALIGNED(width, 16)) { |
2719 | | RGBAToUVRow = RGBAToUVRow_NEON_I8MM; |
2720 | | } |
2721 | | } |
2722 | | #endif |
2723 | | #if defined(HAS_RGBATOUVROW_SVE2) |
2724 | | if (TestCpuFlag(kCpuHasSVE2)) { |
2725 | | RGBAToUVRow = RGBAToUVRow_Any_SVE2; |
2726 | | if (IS_ALIGNED(width, 2)) { |
2727 | | RGBAToUVRow = RGBAToUVRow_SVE2; |
2728 | | } |
2729 | | } |
2730 | | #endif |
2731 | | #if defined(HAS_RGBATOUVROW_SME) |
2732 | | if (TestCpuFlag(kCpuHasSME)) { |
2733 | | RGBAToUVRow = RGBAToUVRow_Any_SME; |
2734 | | if (IS_ALIGNED(width, 2)) { |
2735 | | RGBAToUVRow = RGBAToUVRow_SME; |
2736 | | } |
2737 | | } |
2738 | | #endif |
2739 | | #if defined(HAS_RGBATOYROW_LSX) && defined(HAS_RGBATOUVROW_LSX) |
2740 | | if (TestCpuFlag(kCpuHasLSX)) { |
2741 | | RGBAToYRow = RGBAToYRow_Any_LSX; |
2742 | | RGBAToUVRow = RGBAToUVRow_Any_LSX; |
2743 | | if (IS_ALIGNED(width, 16)) { |
2744 | | RGBAToYRow = RGBAToYRow_LSX; |
2745 | | RGBAToUVRow = RGBAToUVRow_LSX; |
2746 | | } |
2747 | | } |
2748 | | #endif |
2749 | | #if defined(HAS_RGBATOYROW_LASX) |
2750 | | if (TestCpuFlag(kCpuHasNEON)) { |
2751 | | RGBAToYRow = RGBAToYRow_Any_LASX; |
2752 | | if (IS_ALIGNED(width, 32)) { |
2753 | | RGBAToYRow = RGBAToYRow_LASX; |
2754 | | } |
2755 | | } |
2756 | | #endif |
2757 | | #if defined(HAS_RGBATOYROW_RVV) |
2758 | | if (TestCpuFlag(kCpuHasRVV)) { |
2759 | | RGBAToYRow = RGBAToYRow_RVV; |
2760 | | } |
2761 | | #endif |
2762 | |
|
2763 | 0 | for (y = 0; y < height - 1; y += 2) { |
2764 | 0 | RGBAToUVRow(src_rgba, src_stride_rgba, dst_u, dst_v, width); |
2765 | 0 | RGBAToYRow(src_rgba, dst_y, width); |
2766 | 0 | RGBAToYRow(src_rgba + src_stride_rgba, dst_y + dst_stride_y, width); |
2767 | 0 | src_rgba += src_stride_rgba * 2; |
2768 | 0 | dst_y += dst_stride_y * 2; |
2769 | 0 | dst_u += dst_stride_u; |
2770 | 0 | dst_v += dst_stride_v; |
2771 | 0 | } |
2772 | 0 | if (height & 1) { |
2773 | 0 | RGBAToUVRow(src_rgba, 0, dst_u, dst_v, width); |
2774 | 0 | RGBAToYRow(src_rgba, dst_y, width); |
2775 | 0 | } |
2776 | 0 | return 0; |
2777 | 0 | } |
2778 | | |
2779 | | // Enabled if 1 pass is available |
2780 | | #if (defined(HAS_RGB24TOYROW_NEON) || defined(HAS_RGB24TOYROW_LSX) || \ |
2781 | | defined(HAS_RGB24TOYROW_RVV)) |
2782 | | #define HAS_RGB24TOYROW |
2783 | | #endif |
2784 | | |
2785 | | // Convert RGB24 to I420. |
2786 | | LIBYUV_API |
2787 | | int RGB24ToI420(const uint8_t* src_rgb24, |
2788 | | int src_stride_rgb24, |
2789 | | uint8_t* dst_y, |
2790 | | int dst_stride_y, |
2791 | | uint8_t* dst_u, |
2792 | | int dst_stride_u, |
2793 | | uint8_t* dst_v, |
2794 | | int dst_stride_v, |
2795 | | int width, |
2796 | 0 | int height) { |
2797 | 0 | int y; |
2798 | | #if defined(HAS_RGB24TOYROW) |
2799 | | void (*RGB24ToUVRow)(const uint8_t* src_rgb24, int src_stride_rgb24, |
2800 | | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2801 | | RGB24ToUVRow_C; |
2802 | | void (*RGB24ToYRow)(const uint8_t* src_rgb24, uint8_t* dst_y, int width) = |
2803 | | RGB24ToYRow_C; |
2804 | | #else |
2805 | 0 | void (*RGB24ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) = |
2806 | 0 | RGB24ToARGBRow_C; |
2807 | 0 | void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb, |
2808 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2809 | 0 | ARGBToUVRow_C; |
2810 | 0 | void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
2811 | 0 | ARGBToYRow_C; |
2812 | 0 | #endif |
2813 | 0 | if (!src_rgb24 || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
2814 | 0 | return -1; |
2815 | 0 | } |
2816 | | // Negative height means invert the image. |
2817 | 0 | if (height < 0) { |
2818 | 0 | height = -height; |
2819 | 0 | src_rgb24 = src_rgb24 + (height - 1) * src_stride_rgb24; |
2820 | 0 | src_stride_rgb24 = -src_stride_rgb24; |
2821 | 0 | } |
2822 | |
|
2823 | | #if defined(HAS_RGB24TOYROW) |
2824 | | |
2825 | | // Neon version does direct RGB24 to YUV. |
2826 | | #if defined(HAS_RGB24TOYROW_NEON) && defined(HAS_RGB24TOUVROW_NEON) |
2827 | | if (TestCpuFlag(kCpuHasNEON)) { |
2828 | | RGB24ToUVRow = RGB24ToUVRow_Any_NEON; |
2829 | | RGB24ToYRow = RGB24ToYRow_Any_NEON; |
2830 | | if (IS_ALIGNED(width, 16)) { |
2831 | | RGB24ToYRow = RGB24ToYRow_NEON; |
2832 | | RGB24ToUVRow = RGB24ToUVRow_NEON; |
2833 | | } |
2834 | | } |
2835 | | #endif |
2836 | | #if defined(HAS_RGB24TOYROW_LSX) && defined(HAS_RGB24TOUVROW_LSX) |
2837 | | if (TestCpuFlag(kCpuHasLSX)) { |
2838 | | RGB24ToUVRow = RGB24ToUVRow_Any_LSX; |
2839 | | RGB24ToYRow = RGB24ToYRow_Any_LSX; |
2840 | | if (IS_ALIGNED(width, 16)) { |
2841 | | RGB24ToYRow = RGB24ToYRow_LSX; |
2842 | | RGB24ToUVRow = RGB24ToUVRow_LSX; |
2843 | | } |
2844 | | } |
2845 | | #endif |
2846 | | #if defined(HAS_RGB24TOYROW_LASX) && defined(HAS_RGB24TOUVROW_LASX) |
2847 | | if (TestCpuFlag(kCpuHasLASX)) { |
2848 | | RGB24ToUVRow = RGB24ToUVRow_Any_LASX; |
2849 | | RGB24ToYRow = RGB24ToYRow_Any_LASX; |
2850 | | if (IS_ALIGNED(width, 32)) { |
2851 | | RGB24ToYRow = RGB24ToYRow_LASX; |
2852 | | RGB24ToUVRow = RGB24ToUVRow_LASX; |
2853 | | } |
2854 | | } |
2855 | | #endif |
2856 | | #if defined(HAS_RGB24TOYROW_RVV) |
2857 | | if (TestCpuFlag(kCpuHasRVV)) { |
2858 | | RGB24ToYRow = RGB24ToYRow_RVV; |
2859 | | } |
2860 | | #endif |
2861 | | |
2862 | | // Other platforms do intermediate conversion from RGB24 to ARGB. |
2863 | | #else // HAS_RGB24TOYROW |
2864 | |
|
2865 | 0 | #if defined(HAS_RGB24TOARGBROW_SSSE3) |
2866 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2867 | 0 | RGB24ToARGBRow = RGB24ToARGBRow_Any_SSSE3; |
2868 | 0 | if (IS_ALIGNED(width, 16)) { |
2869 | 0 | RGB24ToARGBRow = RGB24ToARGBRow_SSSE3; |
2870 | 0 | } |
2871 | 0 | } |
2872 | 0 | #endif |
2873 | 0 | #if defined(HAS_ARGBTOYROW_SSSE3) |
2874 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2875 | 0 | ARGBToYRow = ARGBToYRow_Any_SSSE3; |
2876 | 0 | if (IS_ALIGNED(width, 16)) { |
2877 | 0 | ARGBToYRow = ARGBToYRow_SSSE3; |
2878 | 0 | } |
2879 | 0 | } |
2880 | 0 | #endif |
2881 | 0 | #if defined(HAS_ARGBTOYROW_AVX2) |
2882 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2883 | 0 | ARGBToYRow = ARGBToYRow_Any_AVX2; |
2884 | 0 | if (IS_ALIGNED(width, 32)) { |
2885 | 0 | ARGBToYRow = ARGBToYRow_AVX2; |
2886 | 0 | } |
2887 | 0 | } |
2888 | 0 | #endif |
2889 | 0 | #if defined(HAS_ARGBTOUVROW_SSSE3) |
2890 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
2891 | 0 | ARGBToUVRow = ARGBToUVRow_Any_SSSE3; |
2892 | 0 | if (IS_ALIGNED(width, 16)) { |
2893 | 0 | ARGBToUVRow = ARGBToUVRow_SSSE3; |
2894 | 0 | } |
2895 | 0 | } |
2896 | 0 | #endif |
2897 | 0 | #if defined(HAS_ARGBTOUVROW_AVX2) |
2898 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
2899 | 0 | ARGBToUVRow = ARGBToUVRow_Any_AVX2; |
2900 | 0 | if (IS_ALIGNED(width, 32)) { |
2901 | 0 | ARGBToUVRow = ARGBToUVRow_AVX2; |
2902 | 0 | } |
2903 | 0 | } |
2904 | 0 | #endif |
2905 | 0 | #endif // HAS_RGB24TOYROW |
2906 | |
|
2907 | 0 | { |
2908 | 0 | #if !defined(HAS_RGB24TOYROW) |
2909 | | // Allocate 2 rows of ARGB. |
2910 | 0 | const int row_size = (width * 4 + 31) & ~31; |
2911 | 0 | align_buffer_64(row, row_size * 2); |
2912 | 0 | if (!row) |
2913 | 0 | return 1; |
2914 | 0 | #endif |
2915 | | |
2916 | 0 | for (y = 0; y < height - 1; y += 2) { |
2917 | | #if defined(HAS_RGB24TOYROW) |
2918 | | RGB24ToUVRow(src_rgb24, src_stride_rgb24, dst_u, dst_v, width); |
2919 | | RGB24ToYRow(src_rgb24, dst_y, width); |
2920 | | RGB24ToYRow(src_rgb24 + src_stride_rgb24, dst_y + dst_stride_y, width); |
2921 | | #else |
2922 | 0 | RGB24ToARGBRow(src_rgb24, row, width); |
2923 | 0 | RGB24ToARGBRow(src_rgb24 + src_stride_rgb24, row + row_size, width); |
2924 | 0 | ARGBToUVRow(row, row_size, dst_u, dst_v, width); |
2925 | 0 | ARGBToYRow(row, dst_y, width); |
2926 | 0 | ARGBToYRow(row + row_size, dst_y + dst_stride_y, width); |
2927 | 0 | #endif |
2928 | 0 | src_rgb24 += src_stride_rgb24 * 2; |
2929 | 0 | dst_y += dst_stride_y * 2; |
2930 | 0 | dst_u += dst_stride_u; |
2931 | 0 | dst_v += dst_stride_v; |
2932 | 0 | } |
2933 | 0 | if (height & 1) { |
2934 | | #if defined(HAS_RGB24TOYROW) |
2935 | | RGB24ToUVRow(src_rgb24, 0, dst_u, dst_v, width); |
2936 | | RGB24ToYRow(src_rgb24, dst_y, width); |
2937 | | #else |
2938 | 0 | RGB24ToARGBRow(src_rgb24, row, width); |
2939 | 0 | ARGBToUVRow(row, 0, dst_u, dst_v, width); |
2940 | 0 | ARGBToYRow(row, dst_y, width); |
2941 | 0 | #endif |
2942 | 0 | } |
2943 | 0 | #if !defined(HAS_RGB24TOYROW) |
2944 | 0 | free_aligned_buffer_64(row); |
2945 | 0 | #endif |
2946 | 0 | } |
2947 | 0 | return 0; |
2948 | 0 | } |
2949 | | #undef HAS_RGB24TOYROW |
2950 | | |
2951 | | // Enabled if 1 pass is available |
2952 | | #if defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_RVV) |
2953 | | #define HAS_RGB24TOYJROW |
2954 | | #endif |
2955 | | |
2956 | | // Convert RGB24 to J420. |
2957 | | LIBYUV_API |
2958 | | int RGB24ToJ420(const uint8_t* src_rgb24, |
2959 | | int src_stride_rgb24, |
2960 | | uint8_t* dst_y, |
2961 | | int dst_stride_y, |
2962 | | uint8_t* dst_u, |
2963 | | int dst_stride_u, |
2964 | | uint8_t* dst_v, |
2965 | | int dst_stride_v, |
2966 | | int width, |
2967 | 0 | int height) { |
2968 | 0 | int y; |
2969 | | #if defined(HAS_RGB24TOYJROW) |
2970 | | void (*RGB24ToUVJRow)(const uint8_t* src_rgb24, int src_stride_rgb24, |
2971 | | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2972 | | RGB24ToUVJRow_C; |
2973 | | void (*RGB24ToYJRow)(const uint8_t* src_rgb24, uint8_t* dst_y, int width) = |
2974 | | RGB24ToYJRow_C; |
2975 | | #else |
2976 | 0 | void (*RGB24ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) = |
2977 | 0 | RGB24ToARGBRow_C; |
2978 | 0 | void (*ARGBToUVJRow)(const uint8_t* src_argb0, int src_stride_argb, |
2979 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
2980 | 0 | ARGBToUVJRow_C; |
2981 | 0 | void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
2982 | 0 | ARGBToYJRow_C; |
2983 | 0 | #endif |
2984 | 0 | if (!src_rgb24 || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
2985 | 0 | return -1; |
2986 | 0 | } |
2987 | | // Negative height means invert the image. |
2988 | 0 | if (height < 0) { |
2989 | 0 | height = -height; |
2990 | 0 | src_rgb24 = src_rgb24 + (height - 1) * src_stride_rgb24; |
2991 | 0 | src_stride_rgb24 = -src_stride_rgb24; |
2992 | 0 | } |
2993 | |
|
2994 | | #if defined(HAS_RGB24TOYJROW) |
2995 | | |
2996 | | // Neon version does direct RGB24 to YUV. |
2997 | | #if defined(HAS_RGB24TOYJROW_NEON) && defined(HAS_RGB24TOUVJROW_NEON) |
2998 | | if (TestCpuFlag(kCpuHasNEON)) { |
2999 | | RGB24ToUVJRow = RGB24ToUVJRow_Any_NEON; |
3000 | | RGB24ToYJRow = RGB24ToYJRow_Any_NEON; |
3001 | | if (IS_ALIGNED(width, 16)) { |
3002 | | RGB24ToYJRow = RGB24ToYJRow_NEON; |
3003 | | RGB24ToUVJRow = RGB24ToUVJRow_NEON; |
3004 | | } |
3005 | | } |
3006 | | #endif |
3007 | | #if defined(HAS_RGB24TOYJROW_LSX) |
3008 | | if (TestCpuFlag(kCpuHasLSX)) { |
3009 | | RGB24ToYJRow = RGB24ToYJRow_Any_LSX; |
3010 | | if (IS_ALIGNED(width, 16)) { |
3011 | | RGB24ToYJRow = RGB24ToYJRow_LSX; |
3012 | | } |
3013 | | } |
3014 | | #endif |
3015 | | #if defined(HAS_RGB24TOYJROW_LASX) |
3016 | | if (TestCpuFlag(kCpuHasLASX)) { |
3017 | | RGB24ToYJRow = RGB24ToYJRow_Any_LASX; |
3018 | | if (IS_ALIGNED(width, 32)) { |
3019 | | RGB24ToYJRow = RGB24ToYJRow_LASX; |
3020 | | } |
3021 | | } |
3022 | | #endif |
3023 | | #if defined(HAS_RGB24TOYJROW_RVV) |
3024 | | if (TestCpuFlag(kCpuHasRVV)) { |
3025 | | RGB24ToYJRow = RGB24ToYJRow_RVV; |
3026 | | } |
3027 | | #endif |
3028 | | |
3029 | | // Other platforms do intermediate conversion from RGB24 to ARGB. |
3030 | | #else // HAS_RGB24TOYJROW |
3031 | |
|
3032 | 0 | #if defined(HAS_RGB24TOARGBROW_SSSE3) |
3033 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3034 | 0 | RGB24ToARGBRow = RGB24ToARGBRow_Any_SSSE3; |
3035 | 0 | if (IS_ALIGNED(width, 16)) { |
3036 | 0 | RGB24ToARGBRow = RGB24ToARGBRow_SSSE3; |
3037 | 0 | } |
3038 | 0 | } |
3039 | 0 | #endif |
3040 | 0 | #if defined(HAS_ARGBTOYJROW_SSSE3) |
3041 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3042 | 0 | ARGBToYJRow = ARGBToYJRow_Any_SSSE3; |
3043 | 0 | if (IS_ALIGNED(width, 16)) { |
3044 | 0 | ARGBToYJRow = ARGBToYJRow_SSSE3; |
3045 | 0 | } |
3046 | 0 | } |
3047 | 0 | #endif |
3048 | 0 | #if defined(HAS_ARGBTOYJROW_AVX2) |
3049 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3050 | 0 | ARGBToYJRow = ARGBToYJRow_Any_AVX2; |
3051 | 0 | if (IS_ALIGNED(width, 32)) { |
3052 | 0 | ARGBToYJRow = ARGBToYJRow_AVX2; |
3053 | 0 | } |
3054 | 0 | } |
3055 | 0 | #endif |
3056 | 0 | #if defined(HAS_ARGBTOUVJROW_SSSE3) |
3057 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3058 | 0 | ARGBToUVJRow = ARGBToUVJRow_Any_SSSE3; |
3059 | 0 | if (IS_ALIGNED(width, 16)) { |
3060 | 0 | ARGBToUVJRow = ARGBToUVJRow_SSSE3; |
3061 | 0 | } |
3062 | 0 | } |
3063 | 0 | #endif |
3064 | 0 | #if defined(HAS_ARGBTOUVJROW_AVX2) |
3065 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3066 | 0 | ARGBToUVJRow = ARGBToUVJRow_Any_AVX2; |
3067 | 0 | if (IS_ALIGNED(width, 32)) { |
3068 | 0 | ARGBToUVJRow = ARGBToUVJRow_AVX2; |
3069 | 0 | } |
3070 | 0 | } |
3071 | 0 | #endif |
3072 | 0 | #endif // HAS_RGB24TOYJROW |
3073 | |
|
3074 | 0 | { |
3075 | 0 | #if !defined(HAS_RGB24TOYJROW) |
3076 | | // Allocate 2 rows of ARGB. |
3077 | 0 | const int row_size = (width * 4 + 31) & ~31; |
3078 | 0 | align_buffer_64(row, row_size * 2); |
3079 | 0 | if (!row) |
3080 | 0 | return 1; |
3081 | 0 | #endif |
3082 | | |
3083 | 0 | for (y = 0; y < height - 1; y += 2) { |
3084 | | #if defined(HAS_RGB24TOYJROW) |
3085 | | RGB24ToUVJRow(src_rgb24, src_stride_rgb24, dst_u, dst_v, width); |
3086 | | RGB24ToYJRow(src_rgb24, dst_y, width); |
3087 | | RGB24ToYJRow(src_rgb24 + src_stride_rgb24, dst_y + dst_stride_y, width); |
3088 | | #else |
3089 | 0 | RGB24ToARGBRow(src_rgb24, row, width); |
3090 | 0 | RGB24ToARGBRow(src_rgb24 + src_stride_rgb24, row + row_size, width); |
3091 | 0 | ARGBToUVJRow(row, row_size, dst_u, dst_v, width); |
3092 | 0 | ARGBToYJRow(row, dst_y, width); |
3093 | 0 | ARGBToYJRow(row + row_size, dst_y + dst_stride_y, width); |
3094 | 0 | #endif |
3095 | 0 | src_rgb24 += src_stride_rgb24 * 2; |
3096 | 0 | dst_y += dst_stride_y * 2; |
3097 | 0 | dst_u += dst_stride_u; |
3098 | 0 | dst_v += dst_stride_v; |
3099 | 0 | } |
3100 | 0 | if (height & 1) { |
3101 | | #if defined(HAS_RGB24TOYJROW) |
3102 | | RGB24ToUVJRow(src_rgb24, 0, dst_u, dst_v, width); |
3103 | | RGB24ToYJRow(src_rgb24, dst_y, width); |
3104 | | #else |
3105 | 0 | RGB24ToARGBRow(src_rgb24, row, width); |
3106 | 0 | ARGBToUVJRow(row, 0, dst_u, dst_v, width); |
3107 | 0 | ARGBToYJRow(row, dst_y, width); |
3108 | 0 | #endif |
3109 | 0 | } |
3110 | 0 | #if !defined(HAS_RGB24TOYJROW) |
3111 | 0 | free_aligned_buffer_64(row); |
3112 | 0 | #endif |
3113 | 0 | } |
3114 | 0 | return 0; |
3115 | 0 | } |
3116 | | #undef HAS_RGB24TOYJROW |
3117 | | |
3118 | | // Enabled if 1 pass is available |
3119 | | #if (defined(HAS_RAWTOYROW_NEON) || defined(HAS_RAWTOYROW_LSX) || \ |
3120 | | defined(HAS_RAWTOYROW_RVV)) |
3121 | | #define HAS_RAWTOYROW |
3122 | | #endif |
3123 | | |
3124 | | // Convert RAW to I420. |
3125 | | LIBYUV_API |
3126 | | int RAWToI420(const uint8_t* src_raw, |
3127 | | int src_stride_raw, |
3128 | | uint8_t* dst_y, |
3129 | | int dst_stride_y, |
3130 | | uint8_t* dst_u, |
3131 | | int dst_stride_u, |
3132 | | uint8_t* dst_v, |
3133 | | int dst_stride_v, |
3134 | | int width, |
3135 | 0 | int height) { |
3136 | 0 | int y; |
3137 | | #if defined(HAS_RAWTOYROW) |
3138 | | void (*RAWToUVRow)(const uint8_t* src_raw, int src_stride_raw, uint8_t* dst_u, |
3139 | | uint8_t* dst_v, int width) = RAWToUVRow_C; |
3140 | | void (*RAWToYRow)(const uint8_t* src_raw, uint8_t* dst_y, int width) = |
3141 | | RAWToYRow_C; |
3142 | | #else |
3143 | 0 | void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) = |
3144 | 0 | RAWToARGBRow_C; |
3145 | 0 | void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb, |
3146 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
3147 | 0 | ARGBToUVRow_C; |
3148 | 0 | void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
3149 | 0 | ARGBToYRow_C; |
3150 | 0 | #endif |
3151 | 0 | if (!src_raw || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
3152 | 0 | return -1; |
3153 | 0 | } |
3154 | | // Negative height means invert the image. |
3155 | 0 | if (height < 0) { |
3156 | 0 | height = -height; |
3157 | 0 | src_raw = src_raw + (height - 1) * src_stride_raw; |
3158 | 0 | src_stride_raw = -src_stride_raw; |
3159 | 0 | } |
3160 | |
|
3161 | | #if defined(HAS_RAWTOYROW) |
3162 | | |
3163 | | // Neon version does direct RAW to YUV. |
3164 | | #if defined(HAS_RAWTOYROW_NEON) && defined(HAS_RAWTOUVROW_NEON) |
3165 | | if (TestCpuFlag(kCpuHasNEON)) { |
3166 | | RAWToUVRow = RAWToUVRow_Any_NEON; |
3167 | | RAWToYRow = RAWToYRow_Any_NEON; |
3168 | | if (IS_ALIGNED(width, 16)) { |
3169 | | RAWToYRow = RAWToYRow_NEON; |
3170 | | RAWToUVRow = RAWToUVRow_NEON; |
3171 | | } |
3172 | | } |
3173 | | #endif |
3174 | | #if defined(HAS_RAWTOYROW_LSX) && defined(HAS_RAWTOUVROW_LSX) |
3175 | | if (TestCpuFlag(kCpuHasLSX)) { |
3176 | | RAWToUVRow = RAWToUVRow_Any_LSX; |
3177 | | RAWToYRow = RAWToYRow_Any_LSX; |
3178 | | if (IS_ALIGNED(width, 16)) { |
3179 | | RAWToYRow = RAWToYRow_LSX; |
3180 | | RAWToUVRow = RAWToUVRow_LSX; |
3181 | | } |
3182 | | } |
3183 | | #endif |
3184 | | #if defined(HAS_RAWTOYROW_LASX) && defined(HAS_RAWTOUVROW_LASX) |
3185 | | if (TestCpuFlag(kCpuHasLASX)) { |
3186 | | RAWToUVRow = RAWToUVRow_Any_LASX; |
3187 | | RAWToYRow = RAWToYRow_Any_LASX; |
3188 | | if (IS_ALIGNED(width, 32)) { |
3189 | | RAWToYRow = RAWToYRow_LASX; |
3190 | | RAWToUVRow = RAWToUVRow_LASX; |
3191 | | } |
3192 | | } |
3193 | | #endif |
3194 | | #if defined(HAS_RAWTOYROW_RVV) |
3195 | | if (TestCpuFlag(kCpuHasRVV)) { |
3196 | | RAWToYRow = RAWToYRow_RVV; |
3197 | | } |
3198 | | #endif |
3199 | | |
3200 | | // Other platforms do intermediate conversion from RAW to ARGB. |
3201 | | #else // HAS_RAWTOYROW |
3202 | |
|
3203 | 0 | #if defined(HAS_RAWTOARGBROW_SSSE3) |
3204 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3205 | 0 | RAWToARGBRow = RAWToARGBRow_Any_SSSE3; |
3206 | 0 | if (IS_ALIGNED(width, 16)) { |
3207 | 0 | RAWToARGBRow = RAWToARGBRow_SSSE3; |
3208 | 0 | } |
3209 | 0 | } |
3210 | 0 | #endif |
3211 | 0 | #if defined(HAS_RAWTOARGBROW_AVX2) |
3212 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3213 | 0 | RAWToARGBRow = RAWToARGBRow_Any_AVX2; |
3214 | 0 | if (IS_ALIGNED(width, 32)) { |
3215 | 0 | RAWToARGBRow = RAWToARGBRow_AVX2; |
3216 | 0 | } |
3217 | 0 | } |
3218 | 0 | #endif |
3219 | 0 | #if defined(HAS_ARGBTOYROW_SSSE3) |
3220 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3221 | 0 | ARGBToYRow = ARGBToYRow_Any_SSSE3; |
3222 | 0 | if (IS_ALIGNED(width, 16)) { |
3223 | 0 | ARGBToYRow = ARGBToYRow_SSSE3; |
3224 | 0 | } |
3225 | 0 | } |
3226 | 0 | #endif |
3227 | 0 | #if defined(HAS_ARGBTOYROW_AVX2) |
3228 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3229 | 0 | ARGBToYRow = ARGBToYRow_Any_AVX2; |
3230 | 0 | if (IS_ALIGNED(width, 32)) { |
3231 | 0 | ARGBToYRow = ARGBToYRow_AVX2; |
3232 | 0 | } |
3233 | 0 | } |
3234 | 0 | #endif |
3235 | 0 | #if defined(HAS_ARGBTOUVROW_SSSE3) |
3236 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3237 | 0 | ARGBToUVRow = ARGBToUVRow_Any_SSSE3; |
3238 | 0 | if (IS_ALIGNED(width, 16)) { |
3239 | 0 | ARGBToUVRow = ARGBToUVRow_SSSE3; |
3240 | 0 | } |
3241 | 0 | } |
3242 | 0 | #endif |
3243 | 0 | #if defined(HAS_ARGBTOUVROW_AVX2) |
3244 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3245 | 0 | ARGBToUVRow = ARGBToUVRow_Any_AVX2; |
3246 | 0 | if (IS_ALIGNED(width, 32)) { |
3247 | 0 | ARGBToUVRow = ARGBToUVRow_AVX2; |
3248 | 0 | } |
3249 | 0 | } |
3250 | 0 | #endif |
3251 | 0 | #endif // HAS_RAWTOYROW |
3252 | |
|
3253 | 0 | { |
3254 | 0 | #if !defined(HAS_RAWTOYROW) |
3255 | | // Allocate 2 rows of ARGB. |
3256 | 0 | const int row_size = (width * 4 + 31) & ~31; |
3257 | 0 | align_buffer_64(row, row_size * 2); |
3258 | 0 | if (!row) |
3259 | 0 | return 1; |
3260 | 0 | #endif |
3261 | | |
3262 | 0 | for (y = 0; y < height - 1; y += 2) { |
3263 | | #if defined(HAS_RAWTOYROW) |
3264 | | RAWToUVRow(src_raw, src_stride_raw, dst_u, dst_v, width); |
3265 | | RAWToYRow(src_raw, dst_y, width); |
3266 | | RAWToYRow(src_raw + src_stride_raw, dst_y + dst_stride_y, width); |
3267 | | #else |
3268 | 0 | RAWToARGBRow(src_raw, row, width); |
3269 | 0 | RAWToARGBRow(src_raw + src_stride_raw, row + row_size, width); |
3270 | 0 | ARGBToUVRow(row, row_size, dst_u, dst_v, width); |
3271 | 0 | ARGBToYRow(row, dst_y, width); |
3272 | 0 | ARGBToYRow(row + row_size, dst_y + dst_stride_y, width); |
3273 | 0 | #endif |
3274 | 0 | src_raw += src_stride_raw * 2; |
3275 | 0 | dst_y += dst_stride_y * 2; |
3276 | 0 | dst_u += dst_stride_u; |
3277 | 0 | dst_v += dst_stride_v; |
3278 | 0 | } |
3279 | 0 | if (height & 1) { |
3280 | | #if defined(HAS_RAWTOYROW) |
3281 | | RAWToUVRow(src_raw, 0, dst_u, dst_v, width); |
3282 | | RAWToYRow(src_raw, dst_y, width); |
3283 | | #else |
3284 | 0 | RAWToARGBRow(src_raw, row, width); |
3285 | 0 | ARGBToUVRow(row, 0, dst_u, dst_v, width); |
3286 | 0 | ARGBToYRow(row, dst_y, width); |
3287 | 0 | #endif |
3288 | 0 | } |
3289 | 0 | #if !defined(HAS_RAWTOYROW) |
3290 | 0 | free_aligned_buffer_64(row); |
3291 | 0 | #endif |
3292 | 0 | } |
3293 | 0 | return 0; |
3294 | 0 | } |
3295 | | #undef HAS_RAWTOYROW |
3296 | | |
3297 | | // Enabled if 1 pass is available |
3298 | | #if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_RVV) |
3299 | | #define HAS_RAWTOYJROW |
3300 | | #endif |
3301 | | |
3302 | | // Convert RAW to J420. |
3303 | | LIBYUV_API |
3304 | | int RAWToJ420(const uint8_t* src_raw, |
3305 | | int src_stride_raw, |
3306 | | uint8_t* dst_y, |
3307 | | int dst_stride_y, |
3308 | | uint8_t* dst_u, |
3309 | | int dst_stride_u, |
3310 | | uint8_t* dst_v, |
3311 | | int dst_stride_v, |
3312 | | int width, |
3313 | 15 | int height) { |
3314 | 15 | int y; |
3315 | | #if defined(HAS_RAWTOYJROW) |
3316 | | void (*RAWToUVJRow)(const uint8_t* src_raw, int src_stride_raw, |
3317 | | uint8_t* dst_u, uint8_t* dst_v, int width) = |
3318 | | RAWToUVJRow_C; |
3319 | | void (*RAWToYJRow)(const uint8_t* src_raw, uint8_t* dst_y, int width) = |
3320 | | RAWToYJRow_C; |
3321 | | #else |
3322 | 15 | void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) = |
3323 | 15 | RAWToARGBRow_C; |
3324 | 15 | void (*ARGBToUVJRow)(const uint8_t* src_argb0, int src_stride_argb, |
3325 | 15 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
3326 | 15 | ARGBToUVJRow_C; |
3327 | 15 | void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
3328 | 15 | ARGBToYJRow_C; |
3329 | 15 | #endif |
3330 | 15 | if (!src_raw || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
3331 | 0 | return -1; |
3332 | 0 | } |
3333 | | // Negative height means invert the image. |
3334 | 15 | if (height < 0) { |
3335 | 0 | height = -height; |
3336 | 0 | src_raw = src_raw + (height - 1) * src_stride_raw; |
3337 | 0 | src_stride_raw = -src_stride_raw; |
3338 | 0 | } |
3339 | | |
3340 | | #if defined(HAS_RAWTOYJROW) |
3341 | | |
3342 | | // Neon version does direct RAW to YUV. |
3343 | | #if defined(HAS_RAWTOYJROW_NEON) && defined(HAS_RAWTOUVJROW_NEON) |
3344 | | if (TestCpuFlag(kCpuHasNEON)) { |
3345 | | RAWToUVJRow = RAWToUVJRow_Any_NEON; |
3346 | | RAWToYJRow = RAWToYJRow_Any_NEON; |
3347 | | if (IS_ALIGNED(width, 16)) { |
3348 | | RAWToYJRow = RAWToYJRow_NEON; |
3349 | | RAWToUVJRow = RAWToUVJRow_NEON; |
3350 | | } |
3351 | | } |
3352 | | #endif |
3353 | | #if defined(HAS_RAWTOYJROW_LSX) && defined(HAS_RAWTOUVJROW_LSX) |
3354 | | if (TestCpuFlag(kCpuHasLSX)) { |
3355 | | RAWToUVJRow = RAWToUVJRow_Any_LSX; |
3356 | | RAWToYJRow = RAWToYJRow_Any_LSX; |
3357 | | if (IS_ALIGNED(width, 16)) { |
3358 | | RAWToYJRow = RAWToYJRow_LSX; |
3359 | | RAWToUVJRow = RAWToUVJRow_LSX; |
3360 | | } |
3361 | | } |
3362 | | #endif |
3363 | | #if defined(HAS_RAWTOYJROW_LASX) && defined(HAS_RAWTOUVJROW_LASX) |
3364 | | if (TestCpuFlag(kCpuHasLASX)) { |
3365 | | RAWToUVJRow = RAWToUVJRow_Any_LASX; |
3366 | | RAWToYJRow = RAWToYJRow_Any_LASX; |
3367 | | if (IS_ALIGNED(width, 32)) { |
3368 | | RAWToYJRow = RAWToYJRow_LASX; |
3369 | | RAWToUVJRow = RAWToUVJRow_LASX; |
3370 | | } |
3371 | | } |
3372 | | #endif |
3373 | | #if defined(HAS_RAWTOYJROW_RVV) |
3374 | | if (TestCpuFlag(kCpuHasRVV)) { |
3375 | | RAWToYJRow = RAWToYJRow_RVV; |
3376 | | } |
3377 | | #endif |
3378 | | |
3379 | | // Other platforms do intermediate conversion from RAW to ARGB. |
3380 | | #else // HAS_RAWTOYJROW |
3381 | | |
3382 | 15 | #if defined(HAS_RAWTOARGBROW_SSSE3) |
3383 | 15 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3384 | 15 | RAWToARGBRow = RAWToARGBRow_Any_SSSE3; |
3385 | 15 | if (IS_ALIGNED(width, 16)) { |
3386 | 7 | RAWToARGBRow = RAWToARGBRow_SSSE3; |
3387 | 7 | } |
3388 | 15 | } |
3389 | 15 | #endif |
3390 | 15 | #if defined(HAS_RAWTOARGBROW_AVX2) |
3391 | 15 | if (TestCpuFlag(kCpuHasAVX2)) { |
3392 | 15 | RAWToARGBRow = RAWToARGBRow_Any_AVX2; |
3393 | 15 | if (IS_ALIGNED(width, 32)) { |
3394 | 2 | RAWToARGBRow = RAWToARGBRow_AVX2; |
3395 | 2 | } |
3396 | 15 | } |
3397 | 15 | #endif |
3398 | 15 | #if defined(HAS_ARGBTOYJROW_SSSE3) |
3399 | 15 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3400 | 15 | ARGBToYJRow = ARGBToYJRow_Any_SSSE3; |
3401 | 15 | if (IS_ALIGNED(width, 16)) { |
3402 | 7 | ARGBToYJRow = ARGBToYJRow_SSSE3; |
3403 | 7 | } |
3404 | 15 | } |
3405 | 15 | #endif |
3406 | 15 | #if defined(HAS_ARGBTOYJROW_AVX2) |
3407 | 15 | if (TestCpuFlag(kCpuHasAVX2)) { |
3408 | 15 | ARGBToYJRow = ARGBToYJRow_Any_AVX2; |
3409 | 15 | if (IS_ALIGNED(width, 32)) { |
3410 | 2 | ARGBToYJRow = ARGBToYJRow_AVX2; |
3411 | 2 | } |
3412 | 15 | } |
3413 | 15 | #endif |
3414 | 15 | #if defined(HAS_ARGBTOUVJROW_SSSE3) |
3415 | 15 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3416 | 15 | ARGBToUVJRow = ARGBToUVJRow_Any_SSSE3; |
3417 | 15 | if (IS_ALIGNED(width, 16)) { |
3418 | 7 | ARGBToUVJRow = ARGBToUVJRow_SSSE3; |
3419 | 7 | } |
3420 | 15 | } |
3421 | 15 | #endif |
3422 | 15 | #if defined(HAS_ARGBTOUVJROW_AVX2) |
3423 | 15 | if (TestCpuFlag(kCpuHasAVX2)) { |
3424 | 15 | ARGBToUVJRow = ARGBToUVJRow_Any_AVX2; |
3425 | 15 | if (IS_ALIGNED(width, 32)) { |
3426 | 2 | ARGBToUVJRow = ARGBToUVJRow_AVX2; |
3427 | 2 | } |
3428 | 15 | } |
3429 | 15 | #endif |
3430 | 15 | #endif // HAS_RAWTOYJROW |
3431 | | |
3432 | 15 | { |
3433 | 15 | #if !defined(HAS_RAWTOYJROW) |
3434 | | // Allocate 2 rows of ARGB. |
3435 | 15 | const int row_size = (width * 4 + 31) & ~31; |
3436 | 15 | align_buffer_64(row, row_size * 2); |
3437 | 15 | if (!row) |
3438 | 0 | return 1; |
3439 | 15 | #endif |
3440 | | |
3441 | 37.7k | for (y = 0; y < height - 1; y += 2) { |
3442 | | #if defined(HAS_RAWTOYJROW) |
3443 | | RAWToUVJRow(src_raw, src_stride_raw, dst_u, dst_v, width); |
3444 | | RAWToYJRow(src_raw, dst_y, width); |
3445 | | RAWToYJRow(src_raw + src_stride_raw, dst_y + dst_stride_y, width); |
3446 | | #else |
3447 | 37.7k | RAWToARGBRow(src_raw, row, width); |
3448 | 37.7k | RAWToARGBRow(src_raw + src_stride_raw, row + row_size, width); |
3449 | 37.7k | ARGBToUVJRow(row, row_size, dst_u, dst_v, width); |
3450 | 37.7k | ARGBToYJRow(row, dst_y, width); |
3451 | 37.7k | ARGBToYJRow(row + row_size, dst_y + dst_stride_y, width); |
3452 | 37.7k | #endif |
3453 | 37.7k | src_raw += src_stride_raw * 2; |
3454 | 37.7k | dst_y += dst_stride_y * 2; |
3455 | 37.7k | dst_u += dst_stride_u; |
3456 | 37.7k | dst_v += dst_stride_v; |
3457 | 37.7k | } |
3458 | 15 | if (height & 1) { |
3459 | | #if defined(HAS_RAWTOYJROW) |
3460 | | RAWToUVJRow(src_raw, 0, dst_u, dst_v, width); |
3461 | | RAWToYJRow(src_raw, dst_y, width); |
3462 | | #else |
3463 | 1 | RAWToARGBRow(src_raw, row, width); |
3464 | 1 | ARGBToUVJRow(row, 0, dst_u, dst_v, width); |
3465 | 1 | ARGBToYJRow(row, dst_y, width); |
3466 | 1 | #endif |
3467 | 1 | } |
3468 | 15 | #if !defined(HAS_RAWTOYJROW) |
3469 | 15 | free_aligned_buffer_64(row); |
3470 | 15 | #endif |
3471 | 15 | } |
3472 | 0 | return 0; |
3473 | 15 | } |
3474 | | #undef HAS_RAWTOYJROW |
3475 | | |
3476 | | // RAW big endian (rgb in memory) to I444 |
3477 | | // 2 step conversion of RAWToARGB then ARGBToY and ARGBToUV444 |
3478 | | LIBYUV_API |
3479 | | int RAWToI444(const uint8_t* src_raw, |
3480 | | int src_stride_raw, |
3481 | | uint8_t* dst_y, |
3482 | | int dst_stride_y, |
3483 | | uint8_t* dst_u, |
3484 | | int dst_stride_u, |
3485 | | uint8_t* dst_v, |
3486 | | int dst_stride_v, |
3487 | | int width, |
3488 | 0 | int height) { |
3489 | 0 | int y; |
3490 | 0 | void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) = |
3491 | 0 | RAWToARGBRow_C; |
3492 | 0 | void (*ARGBToYRow)(const uint8_t* src_raw, uint8_t* dst_y, int width) = |
3493 | 0 | ARGBToYRow_C; |
3494 | 0 | void (*ARGBToUV444Row)(const uint8_t* src_raw, uint8_t* dst_u, uint8_t* dst_v, |
3495 | 0 | int width) = ARGBToUV444Row_C; |
3496 | 0 | if (!src_raw || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
3497 | 0 | return -1; |
3498 | 0 | } |
3499 | 0 | if (height < 0) { |
3500 | 0 | height = -height; |
3501 | 0 | src_raw = src_raw + (height - 1) * src_stride_raw; |
3502 | 0 | src_stride_raw = -src_stride_raw; |
3503 | 0 | } |
3504 | | // TODO: add row coalesce when main loop handles large width in blocks |
3505 | | // TODO: implement UV444 or trim the ifdef below |
3506 | 0 | #if defined(HAS_ARGBTOUV444ROW_SSSE3) |
3507 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3508 | 0 | ARGBToUV444Row = ARGBToUV444Row_Any_SSSE3; |
3509 | 0 | if (IS_ALIGNED(width, 16)) { |
3510 | 0 | ARGBToUV444Row = ARGBToUV444Row_SSSE3; |
3511 | 0 | } |
3512 | 0 | } |
3513 | 0 | #endif |
3514 | 0 | #if defined(HAS_ARGBTOUV444ROW_AVX2) |
3515 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3516 | 0 | ARGBToUV444Row = ARGBToUV444Row_Any_AVX2; |
3517 | 0 | if (IS_ALIGNED(width, 32)) { |
3518 | 0 | ARGBToUV444Row = ARGBToUV444Row_AVX2; |
3519 | 0 | } |
3520 | 0 | } |
3521 | 0 | #endif |
3522 | | #if defined(HAS_ARGBTOUV444ROW_NEON) |
3523 | | if (TestCpuFlag(kCpuHasNEON)) { |
3524 | | ARGBToUV444Row = ARGBToUV444Row_Any_NEON; |
3525 | | if (IS_ALIGNED(width, 8)) { |
3526 | | ARGBToUV444Row = ARGBToUV444Row_NEON; |
3527 | | } |
3528 | | } |
3529 | | #endif |
3530 | | #if defined(HAS_ARGBTOUV444ROW_NEON_I8MM) |
3531 | | if (TestCpuFlag(kCpuHasNeonI8MM)) { |
3532 | | ARGBToUV444Row = ARGBToUV444Row_Any_NEON_I8MM; |
3533 | | if (IS_ALIGNED(width, 8)) { |
3534 | | ARGBToUV444Row = ARGBToUV444Row_NEON_I8MM; |
3535 | | } |
3536 | | } |
3537 | | #endif |
3538 | | #if defined(HAS_ARGBTOUV444ROW_LSX) |
3539 | | if (TestCpuFlag(kCpuHasLSX)) { |
3540 | | ARGBToUV444Row = ARGBToUV444Row_Any_LSX; |
3541 | | if (IS_ALIGNED(width, 16)) { |
3542 | | ARGBToUV444Row = ARGBToUV444Row_LSX; |
3543 | | } |
3544 | | } |
3545 | | #endif |
3546 | | #if defined(HAS_ARGBTOUV444ROW_LASX) |
3547 | | if (TestCpuFlag(kCpuHasLASX)) { |
3548 | | ARGBToUV444Row = ARGBToUV444Row_Any_LASX; |
3549 | | if (IS_ALIGNED(width, 32)) { |
3550 | | ARGBToUV444Row = ARGBToUV444Row_LASX; |
3551 | | } |
3552 | | } |
3553 | | #endif |
3554 | 0 | #if defined(HAS_ARGBTOYROW_SSSE3) |
3555 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3556 | 0 | ARGBToYRow = ARGBToYRow_Any_SSSE3; |
3557 | 0 | if (IS_ALIGNED(width, 16)) { |
3558 | 0 | ARGBToYRow = ARGBToYRow_SSSE3; |
3559 | 0 | } |
3560 | 0 | } |
3561 | 0 | #endif |
3562 | 0 | #if defined(HAS_ARGBTOYROW_AVX2) |
3563 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3564 | 0 | ARGBToYRow = ARGBToYRow_Any_AVX2; |
3565 | 0 | if (IS_ALIGNED(width, 32)) { |
3566 | 0 | ARGBToYRow = ARGBToYRow_AVX2; |
3567 | 0 | } |
3568 | 0 | } |
3569 | 0 | #endif |
3570 | | #if defined(HAS_ARGBTOYROW_NEON) |
3571 | | if (TestCpuFlag(kCpuHasNEON)) { |
3572 | | ARGBToYRow = ARGBToYRow_Any_NEON; |
3573 | | if (IS_ALIGNED(width, 16)) { |
3574 | | ARGBToYRow = ARGBToYRow_NEON; |
3575 | | } |
3576 | | } |
3577 | | #endif |
3578 | | #if defined(HAS_ARGBTOYROW_NEON_DOTPROD) |
3579 | | if (TestCpuFlag(kCpuHasNeonDotProd)) { |
3580 | | ARGBToYRow = ARGBToYRow_Any_NEON_DotProd; |
3581 | | if (IS_ALIGNED(width, 16)) { |
3582 | | ARGBToYRow = ARGBToYRow_NEON_DotProd; |
3583 | | } |
3584 | | } |
3585 | | #endif |
3586 | | #if defined(HAS_ARGBTOYROW_LSX) |
3587 | | if (TestCpuFlag(kCpuHasLSX)) { |
3588 | | ARGBToYRow = ARGBToYRow_Any_LSX; |
3589 | | if (IS_ALIGNED(width, 16)) { |
3590 | | ARGBToYRow = ARGBToYRow_LSX; |
3591 | | } |
3592 | | } |
3593 | | #endif |
3594 | | #if defined(HAS_ARGBTOYROW_LASX) |
3595 | | if (TestCpuFlag(kCpuHasLASX)) { |
3596 | | ARGBToYRow = ARGBToYRow_Any_LASX; |
3597 | | if (IS_ALIGNED(width, 32)) { |
3598 | | ARGBToYRow = ARGBToYRow_LASX; |
3599 | | } |
3600 | | } |
3601 | | #endif |
3602 | | #if defined(HAS_ARGBTOYROW_RVV) |
3603 | | if (TestCpuFlag(kCpuHasRVV)) { |
3604 | | ARGBToYRow = ARGBToYRow_RVV; |
3605 | | } |
3606 | | #endif |
3607 | |
|
3608 | 0 | #if defined(HAS_RAWTOARGBROW_SSSE3) |
3609 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3610 | 0 | RAWToARGBRow = RAWToARGBRow_Any_SSSE3; |
3611 | 0 | if (IS_ALIGNED(width, 16)) { |
3612 | 0 | RAWToARGBRow = RAWToARGBRow_SSSE3; |
3613 | 0 | } |
3614 | 0 | } |
3615 | 0 | #endif |
3616 | 0 | #if defined(HAS_RAWTOARGBROW_AVX2) |
3617 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3618 | 0 | RAWToARGBRow = RAWToARGBRow_Any_AVX2; |
3619 | 0 | if (IS_ALIGNED(width, 32)) { |
3620 | 0 | RAWToARGBRow = RAWToARGBRow_AVX2; |
3621 | 0 | } |
3622 | 0 | } |
3623 | 0 | #endif |
3624 | | #if defined(HAS_RAWTOARGBROW_NEON) |
3625 | | if (TestCpuFlag(kCpuHasNEON)) { |
3626 | | RAWToARGBRow = RAWToARGBRow_Any_NEON; |
3627 | | if (IS_ALIGNED(width, 8)) { |
3628 | | RAWToARGBRow = RAWToARGBRow_NEON; |
3629 | | } |
3630 | | } |
3631 | | #endif |
3632 | | #if defined(HAS_RAWTOARGBROW_SVE2) |
3633 | | if (TestCpuFlag(kCpuHasSVE2)) { |
3634 | | RAWToARGBRow = RAWToARGBRow_SVE2; |
3635 | | } |
3636 | | #endif |
3637 | | #if defined(HAS_RAWTOARGBROW_LSX) |
3638 | | if (TestCpuFlag(kCpuHasLSX)) { |
3639 | | RAWToARGBRow = RAWToARGBRow_Any_LSX; |
3640 | | if (IS_ALIGNED(width, 16)) { |
3641 | | RAWToARGBRow = RAWToARGBRow_LSX; |
3642 | | } |
3643 | | } |
3644 | | #endif |
3645 | | #if defined(HAS_RAWTOARGBROW_LASX) |
3646 | | if (TestCpuFlag(kCpuHasLASX)) { |
3647 | | RAWToARGBRow = RAWToARGBRow_Any_LASX; |
3648 | | if (IS_ALIGNED(width, 32)) { |
3649 | | RAWToARGBRow = RAWToARGBRow_LASX; |
3650 | | } |
3651 | | } |
3652 | | #endif |
3653 | | #if defined(HAS_RAWTOARGBROW_RVV) |
3654 | | if (TestCpuFlag(kCpuHasRVV)) { |
3655 | | RAWToARGBRow = RAWToARGBRow_RVV; |
3656 | | } |
3657 | | #endif |
3658 | |
|
3659 | 0 | { |
3660 | | // Allocate a row of ARGB. |
3661 | 0 | const int row_size = width * 4; |
3662 | 0 | align_buffer_64(row, row_size); |
3663 | 0 | if (!row) |
3664 | 0 | return 1; |
3665 | | |
3666 | 0 | for (y = 0; y < height; ++y) { |
3667 | 0 | RAWToARGBRow(src_raw, row, width); |
3668 | 0 | ARGBToUV444Row(row, dst_u, dst_v, width); |
3669 | 0 | ARGBToYRow(row, dst_y, width); |
3670 | 0 | src_raw += src_stride_raw; |
3671 | 0 | dst_y += dst_stride_y; |
3672 | 0 | dst_u += dst_stride_u; |
3673 | 0 | dst_v += dst_stride_v; |
3674 | 0 | } |
3675 | 0 | free_aligned_buffer_64(row); |
3676 | 0 | } |
3677 | 0 | return 0; |
3678 | 0 | } |
3679 | | |
3680 | | // RAW big endian (rgb in memory) to J444 |
3681 | | // 2 step conversion of RAWToARGB then ARGBToYJ and ARGBToUVJ444 |
3682 | | LIBYUV_API |
3683 | | int RAWToJ444(const uint8_t* src_raw, |
3684 | | int src_stride_raw, |
3685 | | uint8_t* dst_y, |
3686 | | int dst_stride_y, |
3687 | | uint8_t* dst_u, |
3688 | | int dst_stride_u, |
3689 | | uint8_t* dst_v, |
3690 | | int dst_stride_v, |
3691 | | int width, |
3692 | 22 | int height) { |
3693 | 22 | int y; |
3694 | 22 | void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) = |
3695 | 22 | RAWToARGBRow_C; |
3696 | 22 | void (*ARGBToYJRow)(const uint8_t* src_raw, uint8_t* dst_y, int width) = |
3697 | 22 | ARGBToYJRow_C; |
3698 | 22 | void (*ARGBToUVJ444Row)(const uint8_t* src_raw, uint8_t* dst_u, |
3699 | 22 | uint8_t* dst_v, int width) = ARGBToUVJ444Row_C; |
3700 | 22 | if (!src_raw || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
3701 | 0 | return -1; |
3702 | 0 | } |
3703 | 22 | if (height < 0) { |
3704 | 0 | height = -height; |
3705 | 0 | src_raw = src_raw + (height - 1) * src_stride_raw; |
3706 | 0 | src_stride_raw = -src_stride_raw; |
3707 | 0 | } |
3708 | | // TODO: add row coalesce when main loop handles large width in blocks |
3709 | 22 | #if defined(HAS_ARGBTOUVJ444ROW_SSSE3) |
3710 | 22 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3711 | 22 | ARGBToUVJ444Row = ARGBToUVJ444Row_Any_SSSE3; |
3712 | 22 | if (IS_ALIGNED(width, 16)) { |
3713 | 7 | ARGBToUVJ444Row = ARGBToUVJ444Row_SSSE3; |
3714 | 7 | } |
3715 | 22 | } |
3716 | 22 | #endif |
3717 | 22 | #if defined(HAS_ARGBTOUVJ444ROW_AVX2) |
3718 | 22 | if (TestCpuFlag(kCpuHasAVX2)) { |
3719 | 22 | ARGBToUVJ444Row = ARGBToUVJ444Row_Any_AVX2; |
3720 | 22 | if (IS_ALIGNED(width, 32)) { |
3721 | 1 | ARGBToUVJ444Row = ARGBToUVJ444Row_AVX2; |
3722 | 1 | } |
3723 | 22 | } |
3724 | 22 | #endif |
3725 | | #if defined(HAS_ARGBTOUVJ444ROW_NEON) |
3726 | | if (TestCpuFlag(kCpuHasNEON)) { |
3727 | | ARGBToUVJ444Row = ARGBToUVJ444Row_Any_NEON; |
3728 | | if (IS_ALIGNED(width, 8)) { |
3729 | | ARGBToUVJ444Row = ARGBToUVJ444Row_NEON; |
3730 | | } |
3731 | | } |
3732 | | #endif |
3733 | | #if defined(HAS_ARGBTOUVJ444ROW_NEON_I8MM) |
3734 | | if (TestCpuFlag(kCpuHasNeonI8MM)) { |
3735 | | ARGBToUVJ444Row = ARGBToUVJ444Row_Any_NEON_I8MM; |
3736 | | if (IS_ALIGNED(width, 8)) { |
3737 | | ARGBToUVJ444Row = ARGBToUVJ444Row_NEON_I8MM; |
3738 | | } |
3739 | | } |
3740 | | #endif |
3741 | | #if defined(HAS_ARGBTOUVJ444ROW_LSX) |
3742 | | if (TestCpuFlag(kCpuHasLSX)) { |
3743 | | ARGBToUVJ444Row = ARGBToUVJ444Row_Any_LSX; |
3744 | | if (IS_ALIGNED(width, 16)) { |
3745 | | ARGBToUVJ444Row = ARGBToUVJ444Row_LSX; |
3746 | | } |
3747 | | } |
3748 | | #endif |
3749 | | #if defined(HAS_ARGBTOUVJ444ROW_LASX) |
3750 | | if (TestCpuFlag(kCpuHasLASX)) { |
3751 | | ARGBToUVJ444Row = ARGBToUVJ444Row_Any_LASX; |
3752 | | if (IS_ALIGNED(width, 32)) { |
3753 | | ARGBToUVJ444Row = ARGBToUVJ444Row_LASX; |
3754 | | } |
3755 | | } |
3756 | | #endif |
3757 | 22 | #if defined(HAS_ARGBTOYJROW_SSSE3) |
3758 | 22 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3759 | 22 | ARGBToYJRow = ARGBToYJRow_Any_SSSE3; |
3760 | 22 | if (IS_ALIGNED(width, 16)) { |
3761 | 7 | ARGBToYJRow = ARGBToYJRow_SSSE3; |
3762 | 7 | } |
3763 | 22 | } |
3764 | 22 | #endif |
3765 | 22 | #if defined(HAS_ARGBTOYJROW_AVX2) |
3766 | 22 | if (TestCpuFlag(kCpuHasAVX2)) { |
3767 | 22 | ARGBToYJRow = ARGBToYJRow_Any_AVX2; |
3768 | 22 | if (IS_ALIGNED(width, 32)) { |
3769 | 1 | ARGBToYJRow = ARGBToYJRow_AVX2; |
3770 | 1 | } |
3771 | 22 | } |
3772 | 22 | #endif |
3773 | | #if defined(HAS_ARGBTOYJROW_NEON) |
3774 | | if (TestCpuFlag(kCpuHasNEON)) { |
3775 | | ARGBToYJRow = ARGBToYJRow_Any_NEON; |
3776 | | if (IS_ALIGNED(width, 16)) { |
3777 | | ARGBToYJRow = ARGBToYJRow_NEON; |
3778 | | } |
3779 | | } |
3780 | | #endif |
3781 | | #if defined(HAS_ARGBTOYJROW_NEON_DOTPROD) |
3782 | | if (TestCpuFlag(kCpuHasNeonDotProd)) { |
3783 | | ARGBToYJRow = ARGBToYJRow_Any_NEON_DotProd; |
3784 | | if (IS_ALIGNED(width, 16)) { |
3785 | | ARGBToYJRow = ARGBToYJRow_NEON_DotProd; |
3786 | | } |
3787 | | } |
3788 | | #endif |
3789 | | #if defined(HAS_ARGBTOYJROW_LSX) |
3790 | | if (TestCpuFlag(kCpuHasLSX)) { |
3791 | | ARGBToYJRow = ARGBToYJRow_Any_LSX; |
3792 | | if (IS_ALIGNED(width, 16)) { |
3793 | | ARGBToYJRow = ARGBToYJRow_LSX; |
3794 | | } |
3795 | | } |
3796 | | #endif |
3797 | | #if defined(HAS_ARGBTOYJROW_LASX) |
3798 | | if (TestCpuFlag(kCpuHasLASX)) { |
3799 | | ARGBToYJRow = ARGBToYJRow_Any_LASX; |
3800 | | if (IS_ALIGNED(width, 32)) { |
3801 | | ARGBToYJRow = ARGBToYJRow_LASX; |
3802 | | } |
3803 | | } |
3804 | | #endif |
3805 | | #if defined(HAS_ARGBTOYJROW_RVV) |
3806 | | if (TestCpuFlag(kCpuHasRVV)) { |
3807 | | ARGBToYJRow = ARGBToYJRow_RVV; |
3808 | | } |
3809 | | #endif |
3810 | | |
3811 | 22 | #if defined(HAS_RAWTOARGBROW_SSSE3) |
3812 | 22 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3813 | 22 | RAWToARGBRow = RAWToARGBRow_Any_SSSE3; |
3814 | 22 | if (IS_ALIGNED(width, 16)) { |
3815 | 7 | RAWToARGBRow = RAWToARGBRow_SSSE3; |
3816 | 7 | } |
3817 | 22 | } |
3818 | 22 | #endif |
3819 | 22 | #if defined(HAS_RAWTOARGBROW_AVX2) |
3820 | 22 | if (TestCpuFlag(kCpuHasAVX2)) { |
3821 | 22 | RAWToARGBRow = RAWToARGBRow_Any_AVX2; |
3822 | 22 | if (IS_ALIGNED(width, 32)) { |
3823 | 1 | RAWToARGBRow = RAWToARGBRow_AVX2; |
3824 | 1 | } |
3825 | 22 | } |
3826 | 22 | #endif |
3827 | | #if defined(HAS_RAWTOARGBROW_NEON) |
3828 | | if (TestCpuFlag(kCpuHasNEON)) { |
3829 | | RAWToARGBRow = RAWToARGBRow_Any_NEON; |
3830 | | if (IS_ALIGNED(width, 8)) { |
3831 | | RAWToARGBRow = RAWToARGBRow_NEON; |
3832 | | } |
3833 | | } |
3834 | | #endif |
3835 | | #if defined(HAS_RAWTOARGBROW_SVE2) |
3836 | | if (TestCpuFlag(kCpuHasSVE2)) { |
3837 | | RAWToARGBRow = RAWToARGBRow_SVE2; |
3838 | | } |
3839 | | #endif |
3840 | | #if defined(HAS_RAWTOARGBROW_LSX) |
3841 | | if (TestCpuFlag(kCpuHasLSX)) { |
3842 | | RAWToARGBRow = RAWToARGBRow_Any_LSX; |
3843 | | if (IS_ALIGNED(width, 16)) { |
3844 | | RAWToARGBRow = RAWToARGBRow_LSX; |
3845 | | } |
3846 | | } |
3847 | | #endif |
3848 | | #if defined(HAS_RAWTOARGBROW_LASX) |
3849 | | if (TestCpuFlag(kCpuHasLASX)) { |
3850 | | RAWToARGBRow = RAWToARGBRow_Any_LASX; |
3851 | | if (IS_ALIGNED(width, 32)) { |
3852 | | RAWToARGBRow = RAWToARGBRow_LASX; |
3853 | | } |
3854 | | } |
3855 | | #endif |
3856 | | #if defined(HAS_RAWTOARGBROW_RVV) |
3857 | | if (TestCpuFlag(kCpuHasRVV)) { |
3858 | | RAWToARGBRow = RAWToARGBRow_RVV; |
3859 | | } |
3860 | | #endif |
3861 | | |
3862 | 22 | { |
3863 | | // Allocate a row of ARGB. |
3864 | 22 | const int row_size = width * 4; |
3865 | 22 | align_buffer_64(row, row_size); |
3866 | 22 | if (!row) |
3867 | 0 | return 1; |
3868 | | |
3869 | 129k | for (y = 0; y < height; ++y) { |
3870 | 129k | RAWToARGBRow(src_raw, row, width); |
3871 | 129k | ARGBToUVJ444Row(row, dst_u, dst_v, width); |
3872 | 129k | ARGBToYJRow(row, dst_y, width); |
3873 | 129k | src_raw += src_stride_raw; |
3874 | 129k | dst_y += dst_stride_y; |
3875 | 129k | dst_u += dst_stride_u; |
3876 | 129k | dst_v += dst_stride_v; |
3877 | 129k | } |
3878 | 22 | free_aligned_buffer_64(row); |
3879 | 22 | } |
3880 | 0 | return 0; |
3881 | 22 | } |
3882 | | |
3883 | | // Convert RGB565 to I420. |
3884 | | LIBYUV_API |
3885 | | int RGB565ToI420(const uint8_t* src_rgb565, |
3886 | | int src_stride_rgb565, |
3887 | | uint8_t* dst_y, |
3888 | | int dst_stride_y, |
3889 | | uint8_t* dst_u, |
3890 | | int dst_stride_u, |
3891 | | uint8_t* dst_v, |
3892 | | int dst_stride_v, |
3893 | | int width, |
3894 | 0 | int height) { |
3895 | 0 | int y; |
3896 | | #if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \ |
3897 | | defined(HAS_RGB565TOYROW_LASX)) |
3898 | | void (*RGB565ToUVRow)(const uint8_t* src_rgb565, int src_stride_rgb565, |
3899 | | uint8_t* dst_u, uint8_t* dst_v, int width) = |
3900 | | RGB565ToUVRow_C; |
3901 | | void (*RGB565ToYRow)(const uint8_t* src_rgb565, uint8_t* dst_y, int width) = |
3902 | | RGB565ToYRow_C; |
3903 | | #else |
3904 | 0 | void (*RGB565ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, |
3905 | 0 | int width) = RGB565ToARGBRow_C; |
3906 | 0 | void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb, |
3907 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
3908 | 0 | ARGBToUVRow_C; |
3909 | 0 | void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
3910 | 0 | ARGBToYRow_C; |
3911 | 0 | #endif |
3912 | 0 | if (!src_rgb565 || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) { |
3913 | 0 | return -1; |
3914 | 0 | } |
3915 | | // Negative height means invert the image. |
3916 | 0 | if (height < 0) { |
3917 | 0 | height = -height; |
3918 | 0 | src_rgb565 = src_rgb565 + (height - 1) * src_stride_rgb565; |
3919 | 0 | src_stride_rgb565 = -src_stride_rgb565; |
3920 | 0 | } |
3921 | | |
3922 | | // Neon version does direct RGB565 to YUV. |
3923 | | #if defined(HAS_RGB565TOYROW_NEON) |
3924 | | if (TestCpuFlag(kCpuHasNEON)) { |
3925 | | RGB565ToYRow = RGB565ToYRow_Any_NEON; |
3926 | | if (IS_ALIGNED(width, 16)) { |
3927 | | RGB565ToYRow = RGB565ToYRow_NEON; |
3928 | | } |
3929 | | } |
3930 | | #endif |
3931 | | // Neon version does direct RGB565 to YUV. |
3932 | | #if defined(HAS_RGB565TOUVROW_NEON) |
3933 | | if (TestCpuFlag(kCpuHasNEON)) { |
3934 | | RGB565ToUVRow = RGB565ToUVRow_Any_NEON; |
3935 | | if (IS_ALIGNED(width, 16)) { |
3936 | | RGB565ToUVRow = RGB565ToUVRow_NEON; |
3937 | | } |
3938 | | } |
3939 | | #endif |
3940 | | // LSX version does direct RGB565 to YUV. |
3941 | | #if defined(HAS_RGB565TOYROW_LSX) && defined(HAS_RGB565TOUVROW_LSX) |
3942 | | if (TestCpuFlag(kCpuHasLSX)) { |
3943 | | RGB565ToUVRow = RGB565ToUVRow_Any_LSX; |
3944 | | RGB565ToYRow = RGB565ToYRow_Any_LSX; |
3945 | | if (IS_ALIGNED(width, 16)) { |
3946 | | RGB565ToYRow = RGB565ToYRow_LSX; |
3947 | | RGB565ToUVRow = RGB565ToUVRow_LSX; |
3948 | | } |
3949 | | } |
3950 | | #endif |
3951 | | #if defined(HAS_RGB565TOYROW_LASX) && defined(HAS_RGB565TOUVROW_LASX) |
3952 | | if (TestCpuFlag(kCpuHasLASX)) { |
3953 | | RGB565ToUVRow = RGB565ToUVRow_Any_LASX; |
3954 | | RGB565ToYRow = RGB565ToYRow_Any_LASX; |
3955 | | if (IS_ALIGNED(width, 32)) { |
3956 | | RGB565ToYRow = RGB565ToYRow_LASX; |
3957 | | RGB565ToUVRow = RGB565ToUVRow_LASX; |
3958 | | } |
3959 | | } |
3960 | | #endif |
3961 | | // Other platforms do intermediate conversion from RGB565 to ARGB. |
3962 | 0 | #if defined(HAS_RGB565TOARGBROW_SSE2) |
3963 | 0 | if (TestCpuFlag(kCpuHasSSE2)) { |
3964 | 0 | RGB565ToARGBRow = RGB565ToARGBRow_Any_SSE2; |
3965 | 0 | if (IS_ALIGNED(width, 8)) { |
3966 | 0 | RGB565ToARGBRow = RGB565ToARGBRow_SSE2; |
3967 | 0 | } |
3968 | 0 | } |
3969 | 0 | #endif |
3970 | | #if defined(HAS_RGB565TOARGBROW_AVX2) |
3971 | | if (TestCpuFlag(kCpuHasAVX2)) { |
3972 | | RGB565ToARGBRow = RGB565ToARGBRow_Any_AVX2; |
3973 | | if (IS_ALIGNED(width, 16)) { |
3974 | | RGB565ToARGBRow = RGB565ToARGBRow_AVX2; |
3975 | | } |
3976 | | } |
3977 | | #endif |
3978 | 0 | #if defined(HAS_ARGBTOYROW_SSSE3) |
3979 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3980 | 0 | ARGBToYRow = ARGBToYRow_Any_SSSE3; |
3981 | 0 | if (IS_ALIGNED(width, 16)) { |
3982 | 0 | ARGBToYRow = ARGBToYRow_SSSE3; |
3983 | 0 | } |
3984 | 0 | } |
3985 | 0 | #endif |
3986 | 0 | #if defined(HAS_ARGBTOYROW_AVX2) |
3987 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
3988 | 0 | ARGBToYRow = ARGBToYRow_Any_AVX2; |
3989 | 0 | if (IS_ALIGNED(width, 32)) { |
3990 | 0 | ARGBToYRow = ARGBToYRow_AVX2; |
3991 | 0 | } |
3992 | 0 | } |
3993 | 0 | #endif |
3994 | 0 | #if defined(HAS_ARGBTOUVROW_SSSE3) |
3995 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
3996 | 0 | ARGBToUVRow = ARGBToUVRow_Any_SSSE3; |
3997 | 0 | if (IS_ALIGNED(width, 16)) { |
3998 | 0 | ARGBToUVRow = ARGBToUVRow_SSSE3; |
3999 | 0 | } |
4000 | 0 | } |
4001 | 0 | #endif |
4002 | 0 | #if defined(HAS_ARGBTOUVROW_AVX2) |
4003 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
4004 | 0 | ARGBToUVRow = ARGBToUVRow_Any_AVX2; |
4005 | 0 | if (IS_ALIGNED(width, 32)) { |
4006 | 0 | ARGBToUVRow = ARGBToUVRow_AVX2; |
4007 | 0 | } |
4008 | 0 | } |
4009 | 0 | #endif |
4010 | 0 | { |
4011 | 0 | #if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \ |
4012 | 0 | defined(HAS_RGB565TOYROW_LASX)) |
4013 | | // Allocate 2 rows of ARGB. |
4014 | 0 | const int row_size = (width * 4 + 31) & ~31; |
4015 | 0 | align_buffer_64(row, row_size * 2); |
4016 | 0 | if (!row) |
4017 | 0 | return 1; |
4018 | 0 | #endif |
4019 | 0 | for (y = 0; y < height - 1; y += 2) { |
4020 | | #if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \ |
4021 | | defined(HAS_RGB565TOYROW_LASX)) |
4022 | | RGB565ToUVRow(src_rgb565, src_stride_rgb565, dst_u, dst_v, width); |
4023 | | RGB565ToYRow(src_rgb565, dst_y, width); |
4024 | | RGB565ToYRow(src_rgb565 + src_stride_rgb565, dst_y + dst_stride_y, width); |
4025 | | #else |
4026 | 0 | RGB565ToARGBRow(src_rgb565, row, width); |
4027 | 0 | RGB565ToARGBRow(src_rgb565 + src_stride_rgb565, row + row_size, width); |
4028 | 0 | ARGBToUVRow(row, row_size, dst_u, dst_v, width); |
4029 | 0 | ARGBToYRow(row, dst_y, width); |
4030 | 0 | ARGBToYRow(row + row_size, dst_y + dst_stride_y, width); |
4031 | 0 | #endif |
4032 | 0 | src_rgb565 += src_stride_rgb565 * 2; |
4033 | 0 | dst_y += dst_stride_y * 2; |
4034 | 0 | dst_u += dst_stride_u; |
4035 | 0 | dst_v += dst_stride_v; |
4036 | 0 | } |
4037 | 0 | if (height & 1) { |
4038 | | #if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \ |
4039 | | defined(HAS_RGB565TOYROW_LASX)) |
4040 | | RGB565ToUVRow(src_rgb565, 0, dst_u, dst_v, width); |
4041 | | RGB565ToYRow(src_rgb565, dst_y, width); |
4042 | | #else |
4043 | 0 | RGB565ToARGBRow(src_rgb565, row, width); |
4044 | 0 | ARGBToUVRow(row, 0, dst_u, dst_v, width); |
4045 | 0 | ARGBToYRow(row, dst_y, width); |
4046 | 0 | #endif |
4047 | 0 | } |
4048 | 0 | #if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \ |
4049 | 0 | defined(HAS_RGB565TOYROW_LASX)) |
4050 | 0 | free_aligned_buffer_64(row); |
4051 | 0 | #endif |
4052 | 0 | } |
4053 | 0 | return 0; |
4054 | 0 | } |
4055 | | |
4056 | | // Convert ARGB1555 to I420. |
4057 | | LIBYUV_API |
4058 | | int ARGB1555ToI420(const uint8_t* src_argb1555, |
4059 | | int src_stride_argb1555, |
4060 | | uint8_t* dst_y, |
4061 | | int dst_stride_y, |
4062 | | uint8_t* dst_u, |
4063 | | int dst_stride_u, |
4064 | | uint8_t* dst_v, |
4065 | | int dst_stride_v, |
4066 | | int width, |
4067 | 0 | int height) { |
4068 | 0 | int y; |
4069 | | #if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \ |
4070 | | defined(HAS_ARGB1555TOYROW_LASX)) |
4071 | | void (*ARGB1555ToUVRow)(const uint8_t* src_argb1555, int src_stride_argb1555, |
4072 | | uint8_t* dst_u, uint8_t* dst_v, int width) = |
4073 | | ARGB1555ToUVRow_C; |
4074 | | void (*ARGB1555ToYRow)(const uint8_t* src_argb1555, uint8_t* dst_y, |
4075 | | int width) = ARGB1555ToYRow_C; |
4076 | | #else |
4077 | 0 | void (*ARGB1555ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, |
4078 | 0 | int width) = ARGB1555ToARGBRow_C; |
4079 | 0 | void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb, |
4080 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
4081 | 0 | ARGBToUVRow_C; |
4082 | 0 | void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
4083 | 0 | ARGBToYRow_C; |
4084 | 0 | #endif |
4085 | 0 | if (!src_argb1555 || !dst_y || !dst_u || !dst_v || width <= 0 || |
4086 | 0 | height == 0) { |
4087 | 0 | return -1; |
4088 | 0 | } |
4089 | | // Negative height means invert the image. |
4090 | 0 | if (height < 0) { |
4091 | 0 | height = -height; |
4092 | 0 | src_argb1555 = src_argb1555 + (height - 1) * src_stride_argb1555; |
4093 | 0 | src_stride_argb1555 = -src_stride_argb1555; |
4094 | 0 | } |
4095 | | |
4096 | | // Neon version does direct ARGB1555 to YUV. |
4097 | | #if defined(HAS_ARGB1555TOYROW_NEON) |
4098 | | if (TestCpuFlag(kCpuHasNEON)) { |
4099 | | ARGB1555ToYRow = ARGB1555ToYRow_Any_NEON; |
4100 | | if (IS_ALIGNED(width, 8)) { |
4101 | | ARGB1555ToYRow = ARGB1555ToYRow_NEON; |
4102 | | } |
4103 | | } |
4104 | | #endif |
4105 | | #if defined(HAS_ARGB1555TOUVROW_NEON) |
4106 | | if (TestCpuFlag(kCpuHasNEON)) { |
4107 | | ARGB1555ToUVRow = ARGB1555ToUVRow_Any_NEON; |
4108 | | if (IS_ALIGNED(width, 16)) { |
4109 | | ARGB1555ToUVRow = ARGB1555ToUVRow_NEON; |
4110 | | } |
4111 | | } |
4112 | | #endif |
4113 | | |
4114 | | // LSX version does direct ARGB1555 to YUV. |
4115 | | #if defined(HAS_ARGB1555TOYROW_LSX) && defined(HAS_ARGB1555TOUVROW_LSX) |
4116 | | if (TestCpuFlag(kCpuHasLSX)) { |
4117 | | ARGB1555ToUVRow = ARGB1555ToUVRow_Any_LSX; |
4118 | | ARGB1555ToYRow = ARGB1555ToYRow_Any_LSX; |
4119 | | if (IS_ALIGNED(width, 16)) { |
4120 | | ARGB1555ToYRow = ARGB1555ToYRow_LSX; |
4121 | | ARGB1555ToUVRow = ARGB1555ToUVRow_LSX; |
4122 | | } |
4123 | | } |
4124 | | #endif |
4125 | | #if defined(HAS_ARGB1555TOYROW_LASX) && defined(HAS_ARGB1555TOUVROW_LASX) |
4126 | | if (TestCpuFlag(kCpuHasLASX)) { |
4127 | | ARGB1555ToUVRow = ARGB1555ToUVRow_Any_LASX; |
4128 | | ARGB1555ToYRow = ARGB1555ToYRow_Any_LASX; |
4129 | | if (IS_ALIGNED(width, 32)) { |
4130 | | ARGB1555ToYRow = ARGB1555ToYRow_LASX; |
4131 | | ARGB1555ToUVRow = ARGB1555ToUVRow_LASX; |
4132 | | } |
4133 | | } |
4134 | | #endif |
4135 | | |
4136 | | // Other platforms do intermediate conversion from ARGB1555 to ARGB. |
4137 | 0 | #if defined(HAS_ARGB1555TOARGBROW_SSE2) |
4138 | 0 | if (TestCpuFlag(kCpuHasSSE2)) { |
4139 | 0 | ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_SSE2; |
4140 | 0 | if (IS_ALIGNED(width, 8)) { |
4141 | 0 | ARGB1555ToARGBRow = ARGB1555ToARGBRow_SSE2; |
4142 | 0 | } |
4143 | 0 | } |
4144 | 0 | #endif |
4145 | | #if defined(HAS_ARGB1555TOARGBROW_AVX2) |
4146 | | if (TestCpuFlag(kCpuHasAVX2)) { |
4147 | | ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_AVX2; |
4148 | | if (IS_ALIGNED(width, 16)) { |
4149 | | ARGB1555ToARGBRow = ARGB1555ToARGBRow_AVX2; |
4150 | | } |
4151 | | } |
4152 | | #endif |
4153 | 0 | #if defined(HAS_ARGBTOYROW_SSSE3) |
4154 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
4155 | 0 | ARGBToYRow = ARGBToYRow_Any_SSSE3; |
4156 | 0 | if (IS_ALIGNED(width, 16)) { |
4157 | 0 | ARGBToYRow = ARGBToYRow_SSSE3; |
4158 | 0 | } |
4159 | 0 | } |
4160 | 0 | #endif |
4161 | 0 | #if defined(HAS_ARGBTOUVROW_SSSE3) |
4162 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
4163 | 0 | ARGBToUVRow = ARGBToUVRow_Any_SSSE3; |
4164 | 0 | if (IS_ALIGNED(width, 16)) { |
4165 | 0 | ARGBToUVRow = ARGBToUVRow_SSSE3; |
4166 | 0 | } |
4167 | 0 | } |
4168 | 0 | #endif |
4169 | 0 | #if defined(HAS_ARGBTOYROW_AVX2) |
4170 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
4171 | 0 | ARGBToYRow = ARGBToYRow_Any_AVX2; |
4172 | 0 | if (IS_ALIGNED(width, 32)) { |
4173 | 0 | ARGBToYRow = ARGBToYRow_AVX2; |
4174 | 0 | } |
4175 | 0 | } |
4176 | 0 | #endif |
4177 | 0 | #if defined(HAS_ARGBTOUVROW_AVX2) |
4178 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
4179 | 0 | ARGBToUVRow = ARGBToUVRow_Any_AVX2; |
4180 | 0 | if (IS_ALIGNED(width, 32)) { |
4181 | 0 | ARGBToUVRow = ARGBToUVRow_AVX2; |
4182 | 0 | } |
4183 | 0 | } |
4184 | 0 | #endif |
4185 | 0 | { |
4186 | 0 | #if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \ |
4187 | 0 | defined(HAS_ARGB1555TOYROW_LASX)) |
4188 | | // Allocate 2 rows of ARGB. |
4189 | 0 | const int row_size = (width * 4 + 31) & ~31; |
4190 | 0 | align_buffer_64(row, row_size * 2); |
4191 | 0 | if (!row) |
4192 | 0 | return 1; |
4193 | 0 | #endif |
4194 | | |
4195 | 0 | for (y = 0; y < height - 1; y += 2) { |
4196 | | #if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \ |
4197 | | defined(HAS_ARGB1555TOYROW_LASX)) |
4198 | | ARGB1555ToUVRow(src_argb1555, src_stride_argb1555, dst_u, dst_v, width); |
4199 | | ARGB1555ToYRow(src_argb1555, dst_y, width); |
4200 | | ARGB1555ToYRow(src_argb1555 + src_stride_argb1555, dst_y + dst_stride_y, |
4201 | | width); |
4202 | | #else |
4203 | 0 | ARGB1555ToARGBRow(src_argb1555, row, width); |
4204 | 0 | ARGB1555ToARGBRow(src_argb1555 + src_stride_argb1555, row + row_size, |
4205 | 0 | width); |
4206 | 0 | ARGBToUVRow(row, row_size, dst_u, dst_v, width); |
4207 | 0 | ARGBToYRow(row, dst_y, width); |
4208 | 0 | ARGBToYRow(row + row_size, dst_y + dst_stride_y, width); |
4209 | 0 | #endif |
4210 | 0 | src_argb1555 += src_stride_argb1555 * 2; |
4211 | 0 | dst_y += dst_stride_y * 2; |
4212 | 0 | dst_u += dst_stride_u; |
4213 | 0 | dst_v += dst_stride_v; |
4214 | 0 | } |
4215 | 0 | if (height & 1) { |
4216 | | #if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \ |
4217 | | defined(HAS_ARGB1555TOYROW_LASX)) |
4218 | | ARGB1555ToUVRow(src_argb1555, 0, dst_u, dst_v, width); |
4219 | | ARGB1555ToYRow(src_argb1555, dst_y, width); |
4220 | | #else |
4221 | 0 | ARGB1555ToARGBRow(src_argb1555, row, width); |
4222 | 0 | ARGBToUVRow(row, 0, dst_u, dst_v, width); |
4223 | 0 | ARGBToYRow(row, dst_y, width); |
4224 | 0 | #endif |
4225 | 0 | } |
4226 | 0 | #if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \ |
4227 | 0 | defined(HAS_ARGB1555TOYROW_LASX)) |
4228 | 0 | free_aligned_buffer_64(row); |
4229 | 0 | #endif |
4230 | 0 | } |
4231 | 0 | return 0; |
4232 | 0 | } |
4233 | | |
4234 | | // Convert ARGB4444 to I420. |
4235 | | LIBYUV_API |
4236 | | int ARGB4444ToI420(const uint8_t* src_argb4444, |
4237 | | int src_stride_argb4444, |
4238 | | uint8_t* dst_y, |
4239 | | int dst_stride_y, |
4240 | | uint8_t* dst_u, |
4241 | | int dst_stride_u, |
4242 | | uint8_t* dst_v, |
4243 | | int dst_stride_v, |
4244 | | int width, |
4245 | 0 | int height) { |
4246 | 0 | int y; |
4247 | | #if defined(HAS_ARGB4444TOYROW_NEON) |
4248 | | void (*ARGB4444ToUVRow)(const uint8_t* src_argb4444, int src_stride_argb4444, |
4249 | | uint8_t* dst_u, uint8_t* dst_v, int width) = |
4250 | | ARGB4444ToUVRow_C; |
4251 | | void (*ARGB4444ToYRow)(const uint8_t* src_argb4444, uint8_t* dst_y, |
4252 | | int width) = ARGB4444ToYRow_C; |
4253 | | #else |
4254 | 0 | void (*ARGB4444ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, |
4255 | 0 | int width) = ARGB4444ToARGBRow_C; |
4256 | 0 | void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb, |
4257 | 0 | uint8_t* dst_u, uint8_t* dst_v, int width) = |
4258 | 0 | ARGBToUVRow_C; |
4259 | 0 | void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) = |
4260 | 0 | ARGBToYRow_C; |
4261 | 0 | #endif |
4262 | 0 | if (!src_argb4444 || !dst_y || !dst_u || !dst_v || width <= 0 || |
4263 | 0 | height == 0) { |
4264 | 0 | return -1; |
4265 | 0 | } |
4266 | | // Negative height means invert the image. |
4267 | 0 | if (height < 0) { |
4268 | 0 | height = -height; |
4269 | 0 | src_argb4444 = src_argb4444 + (height - 1) * src_stride_argb4444; |
4270 | 0 | src_stride_argb4444 = -src_stride_argb4444; |
4271 | 0 | } |
4272 | | |
4273 | | // Neon version does direct ARGB4444 to YUV. |
4274 | | #if defined(HAS_ARGB4444TOYROW_NEON) |
4275 | | if (TestCpuFlag(kCpuHasNEON)) { |
4276 | | ARGB4444ToYRow = ARGB4444ToYRow_Any_NEON; |
4277 | | if (IS_ALIGNED(width, 8)) { |
4278 | | ARGB4444ToYRow = ARGB4444ToYRow_NEON; |
4279 | | } |
4280 | | } |
4281 | | #endif |
4282 | | #if defined(HAS_ARGB4444TOUVROW_NEON) |
4283 | | if (TestCpuFlag(kCpuHasNEON)) { |
4284 | | ARGB4444ToUVRow = ARGB4444ToUVRow_Any_NEON; |
4285 | | if (IS_ALIGNED(width, 16)) { |
4286 | | ARGB4444ToUVRow = ARGB4444ToUVRow_NEON; |
4287 | | } |
4288 | | } |
4289 | | #endif |
4290 | 0 | #if defined(HAS_ARGB4444TOARGBROW_SSE2) |
4291 | 0 | if (TestCpuFlag(kCpuHasSSE2)) { |
4292 | 0 | ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_SSE2; |
4293 | 0 | if (IS_ALIGNED(width, 8)) { |
4294 | 0 | ARGB4444ToARGBRow = ARGB4444ToARGBRow_SSE2; |
4295 | 0 | } |
4296 | 0 | } |
4297 | 0 | #endif |
4298 | | #if defined(HAS_ARGB4444TOARGBROW_AVX2) |
4299 | | if (TestCpuFlag(kCpuHasAVX2)) { |
4300 | | ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_AVX2; |
4301 | | if (IS_ALIGNED(width, 16)) { |
4302 | | ARGB4444ToARGBRow = ARGB4444ToARGBRow_AVX2; |
4303 | | } |
4304 | | } |
4305 | | #endif |
4306 | | #if defined(HAS_ARGB4444TOARGBROW_LSX) |
4307 | | if (TestCpuFlag(kCpuHasLSX)) { |
4308 | | ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LSX; |
4309 | | if (IS_ALIGNED(width, 16)) { |
4310 | | ARGB4444ToARGBRow = ARGB4444ToARGBRow_LSX; |
4311 | | } |
4312 | | } |
4313 | | #endif |
4314 | | #if defined(HAS_ARGB4444TOARGBROW_LASX) |
4315 | | if (TestCpuFlag(kCpuHasLASX)) { |
4316 | | ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LASX; |
4317 | | if (IS_ALIGNED(width, 32)) { |
4318 | | ARGB4444ToARGBRow = ARGB4444ToARGBRow_LASX; |
4319 | | } |
4320 | | } |
4321 | | #endif |
4322 | 0 | #if defined(HAS_ARGBTOYROW_SSSE3) |
4323 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
4324 | 0 | ARGBToYRow = ARGBToYRow_Any_SSSE3; |
4325 | 0 | if (IS_ALIGNED(width, 16)) { |
4326 | 0 | ARGBToYRow = ARGBToYRow_SSSE3; |
4327 | 0 | } |
4328 | 0 | } |
4329 | 0 | #endif |
4330 | 0 | #if defined(HAS_ARGBTOUVROW_SSSE3) |
4331 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
4332 | 0 | ARGBToUVRow = ARGBToUVRow_Any_SSSE3; |
4333 | 0 | if (IS_ALIGNED(width, 16)) { |
4334 | 0 | ARGBToUVRow = ARGBToUVRow_SSSE3; |
4335 | 0 | } |
4336 | 0 | } |
4337 | 0 | #endif |
4338 | 0 | #if defined(HAS_ARGBTOYROW_AVX2) |
4339 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
4340 | 0 | ARGBToYRow = ARGBToYRow_Any_AVX2; |
4341 | 0 | if (IS_ALIGNED(width, 32)) { |
4342 | 0 | ARGBToYRow = ARGBToYRow_AVX2; |
4343 | 0 | } |
4344 | 0 | } |
4345 | 0 | #endif |
4346 | 0 | #if defined(HAS_ARGBTOUVROW_AVX2) |
4347 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
4348 | 0 | ARGBToUVRow = ARGBToUVRow_Any_AVX2; |
4349 | 0 | if (IS_ALIGNED(width, 32)) { |
4350 | 0 | ARGBToUVRow = ARGBToUVRow_AVX2; |
4351 | 0 | } |
4352 | 0 | } |
4353 | 0 | #endif |
4354 | | #if defined(HAS_ARGBTOYROW_LSX) |
4355 | | if (TestCpuFlag(kCpuHasLSX)) { |
4356 | | ARGBToYRow = ARGBToYRow_Any_LSX; |
4357 | | if (IS_ALIGNED(width, 16)) { |
4358 | | ARGBToYRow = ARGBToYRow_LSX; |
4359 | | } |
4360 | | } |
4361 | | #endif |
4362 | | #if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX) |
4363 | | if (TestCpuFlag(kCpuHasLSX)) { |
4364 | | ARGBToYRow = ARGBToYRow_Any_LSX; |
4365 | | ARGBToUVRow = ARGBToUVRow_Any_LSX; |
4366 | | if (IS_ALIGNED(width, 16)) { |
4367 | | ARGBToYRow = ARGBToYRow_LSX; |
4368 | | ARGBToUVRow = ARGBToUVRow_LSX; |
4369 | | } |
4370 | | } |
4371 | | #endif |
4372 | | #if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX) |
4373 | | if (TestCpuFlag(kCpuHasLASX)) { |
4374 | | ARGBToYRow = ARGBToYRow_Any_LASX; |
4375 | | ARGBToUVRow = ARGBToUVRow_Any_LASX; |
4376 | | if (IS_ALIGNED(width, 32)) { |
4377 | | ARGBToYRow = ARGBToYRow_LASX; |
4378 | | ARGBToUVRow = ARGBToUVRow_LASX; |
4379 | | } |
4380 | | } |
4381 | | #endif |
4382 | |
|
4383 | 0 | { |
4384 | 0 | #if !(defined(HAS_ARGB4444TOYROW_NEON)) |
4385 | | // Allocate 2 rows of ARGB. |
4386 | 0 | const int row_size = (width * 4 + 31) & ~31; |
4387 | 0 | align_buffer_64(row, row_size * 2); |
4388 | 0 | if (!row) |
4389 | 0 | return 1; |
4390 | 0 | #endif |
4391 | | |
4392 | 0 | for (y = 0; y < height - 1; y += 2) { |
4393 | | #if defined(HAS_ARGB4444TOYROW_NEON) |
4394 | | ARGB4444ToUVRow(src_argb4444, src_stride_argb4444, dst_u, dst_v, width); |
4395 | | ARGB4444ToYRow(src_argb4444, dst_y, width); |
4396 | | ARGB4444ToYRow(src_argb4444 + src_stride_argb4444, dst_y + dst_stride_y, |
4397 | | width); |
4398 | | #else |
4399 | 0 | ARGB4444ToARGBRow(src_argb4444, row, width); |
4400 | 0 | ARGB4444ToARGBRow(src_argb4444 + src_stride_argb4444, row + row_size, |
4401 | 0 | width); |
4402 | 0 | ARGBToUVRow(row, row_size, dst_u, dst_v, width); |
4403 | 0 | ARGBToYRow(row, dst_y, width); |
4404 | 0 | ARGBToYRow(row + row_size, dst_y + dst_stride_y, width); |
4405 | 0 | #endif |
4406 | 0 | src_argb4444 += src_stride_argb4444 * 2; |
4407 | 0 | dst_y += dst_stride_y * 2; |
4408 | 0 | dst_u += dst_stride_u; |
4409 | 0 | dst_v += dst_stride_v; |
4410 | 0 | } |
4411 | 0 | if (height & 1) { |
4412 | | #if defined(HAS_ARGB4444TOYROW_NEON) |
4413 | | ARGB4444ToUVRow(src_argb4444, 0, dst_u, dst_v, width); |
4414 | | ARGB4444ToYRow(src_argb4444, dst_y, width); |
4415 | | #else |
4416 | 0 | ARGB4444ToARGBRow(src_argb4444, row, width); |
4417 | 0 | ARGBToUVRow(row, 0, dst_u, dst_v, width); |
4418 | 0 | ARGBToYRow(row, dst_y, width); |
4419 | 0 | #endif |
4420 | 0 | } |
4421 | 0 | #if !(defined(HAS_ARGB4444TOYROW_NEON)) |
4422 | 0 | free_aligned_buffer_64(row); |
4423 | 0 | #endif |
4424 | 0 | } |
4425 | 0 | return 0; |
4426 | 0 | } |
4427 | | |
4428 | | // Convert RGB24 to J400. |
4429 | | LIBYUV_API |
4430 | | int RGB24ToJ400(const uint8_t* src_rgb24, |
4431 | | int src_stride_rgb24, |
4432 | | uint8_t* dst_yj, |
4433 | | int dst_stride_yj, |
4434 | | int width, |
4435 | 0 | int height) { |
4436 | 0 | int y; |
4437 | 0 | void (*RGB24ToYJRow)(const uint8_t* src_rgb24, uint8_t* dst_yj, int width) = |
4438 | 0 | RGB24ToYJRow_C; |
4439 | 0 | if (!src_rgb24 || !dst_yj || width <= 0 || height == 0) { |
4440 | 0 | return -1; |
4441 | 0 | } |
4442 | 0 | if (height < 0) { |
4443 | 0 | height = -height; |
4444 | 0 | src_rgb24 = src_rgb24 + (height - 1) * src_stride_rgb24; |
4445 | 0 | src_stride_rgb24 = -src_stride_rgb24; |
4446 | 0 | } |
4447 | | // Coalesce rows. |
4448 | 0 | if (src_stride_rgb24 == width * 3 && dst_stride_yj == width) { |
4449 | 0 | width *= height; |
4450 | 0 | height = 1; |
4451 | 0 | src_stride_rgb24 = dst_stride_yj = 0; |
4452 | 0 | } |
4453 | 0 | #if defined(HAS_RGB24TOYJROW_SSSE3) |
4454 | 0 | if (TestCpuFlag(kCpuHasSSSE3)) { |
4455 | 0 | RGB24ToYJRow = RGB24ToYJRow_Any_SSSE3; |
4456 | 0 | if (IS_ALIGNED(width, 16)) { |
4457 | 0 | RGB24ToYJRow = RGB24ToYJRow_SSSE3; |
4458 | 0 | } |
4459 | 0 | } |
4460 | 0 | #endif |
4461 | 0 | #if defined(HAS_RGB24TOYJROW_AVX2) |
4462 | 0 | if (TestCpuFlag(kCpuHasAVX2)) { |
4463 | 0 | RGB24ToYJRow = RGB24ToYJRow_Any_AVX2; |
4464 | 0 | if (IS_ALIGNED(width, 32)) { |
4465 | 0 | RGB24ToYJRow = RGB24ToYJRow_AVX2; |
4466 | 0 | } |
4467 | 0 | } |
4468 | 0 | #endif |
4469 | | #if defined(HAS_RGB24TOYJROW_NEON) |
4470 | | if (TestCpuFlag(kCpuHasNEON)) { |
4471 | | RGB24ToYJRow = RGB24ToYJRow_Any_NEON; |
4472 | | if (IS_ALIGNED(width, 16)) { |
4473 | | RGB24ToYJRow = RGB24ToYJRow_NEON; |
4474 | | } |
4475 | | } |
4476 | | #endif |
4477 | | #if defined(HAS_RGB24TOYJROW_LSX) |
4478 | | if (TestCpuFlag(kCpuHasLSX)) { |
4479 | | RGB24ToYJRow = RGB24ToYJRow_Any_LSX; |
4480 | | if (IS_ALIGNED(width, 16)) { |
4481 | | RGB24ToYJRow = RGB24ToYJRow_LSX; |
4482 | | } |
4483 | | } |
4484 | | #endif |
4485 | | #if defined(HAS_RGB24TOYJROW_LASX) |
4486 | | if (TestCpuFlag(kCpuHasLASX)) { |
4487 | | RGB24ToYJRow = RGB24ToYJRow_Any_LASX; |
4488 | | if (IS_ALIGNED(width, 32)) { |
4489 | | RGB24ToYJRow = RGB24ToYJRow_LASX; |
4490 | | } |
4491 | | } |
4492 | | #endif |
4493 | | #if defined(HAS_RGB24TOYJROW_RVV) |
4494 | | if (TestCpuFlag(kCpuHasRVV)) { |
4495 | | RGB24ToYJRow = RGB24ToYJRow_RVV; |
4496 | | } |
4497 | | #endif |
4498 | |
|
4499 | 0 | for (y = 0; y < height; ++y) { |
4500 | 0 | RGB24ToYJRow(src_rgb24, dst_yj, width); |
4501 | 0 | src_rgb24 += src_stride_rgb24; |
4502 | 0 | dst_yj += dst_stride_yj; |
4503 | 0 | } |
4504 | 0 | return 0; |
4505 | 0 | } |
4506 | | |
4507 | | // Convert RAW to J400. |
4508 | | LIBYUV_API |
4509 | | int RAWToJ400(const uint8_t* src_raw, |
4510 | | int src_stride_raw, |
4511 | | uint8_t* dst_yj, |
4512 | | int dst_stride_yj, |
4513 | | int width, |
4514 | 6 | int height) { |
4515 | 6 | int y; |
4516 | 6 | void (*RAWToYJRow)(const uint8_t* src_raw, uint8_t* dst_yj, int width) = |
4517 | 6 | RAWToYJRow_C; |
4518 | 6 | if (!src_raw || !dst_yj || width <= 0 || height == 0) { |
4519 | 0 | return -1; |
4520 | 0 | } |
4521 | | |
4522 | 6 | if (height < 0) { |
4523 | 0 | height = -height; |
4524 | 0 | src_raw = src_raw + (height - 1) * src_stride_raw; |
4525 | 0 | src_stride_raw = -src_stride_raw; |
4526 | 0 | } |
4527 | | // Coalesce rows. |
4528 | 6 | if (src_stride_raw == width * 3 && dst_stride_yj == width) { |
4529 | 6 | width *= height; |
4530 | 6 | height = 1; |
4531 | 6 | src_stride_raw = dst_stride_yj = 0; |
4532 | 6 | } |
4533 | | |
4534 | 6 | #if defined(HAS_RAWTOYJROW_SSSE3) |
4535 | 6 | if (TestCpuFlag(kCpuHasSSSE3)) { |
4536 | 6 | RAWToYJRow = RAWToYJRow_Any_SSSE3; |
4537 | 6 | if (IS_ALIGNED(width, 16)) { |
4538 | 2 | RAWToYJRow = RAWToYJRow_SSSE3; |
4539 | 2 | } |
4540 | 6 | } |
4541 | 6 | #endif |
4542 | 6 | #if defined(HAS_RAWTOYJROW_AVX2) |
4543 | 6 | if (TestCpuFlag(kCpuHasAVX2)) { |
4544 | 6 | RAWToYJRow = RAWToYJRow_Any_AVX2; |
4545 | 6 | if (IS_ALIGNED(width, 32)) { |
4546 | 1 | RAWToYJRow = RAWToYJRow_AVX2; |
4547 | 1 | } |
4548 | 6 | } |
4549 | 6 | #endif |
4550 | | #if defined(HAS_RAWTOYJROW_NEON) |
4551 | | if (TestCpuFlag(kCpuHasNEON)) { |
4552 | | RAWToYJRow = RAWToYJRow_Any_NEON; |
4553 | | if (IS_ALIGNED(width, 16)) { |
4554 | | RAWToYJRow = RAWToYJRow_NEON; |
4555 | | } |
4556 | | } |
4557 | | #endif |
4558 | | #if defined(HAS_RAWTOYJROW_LSX) |
4559 | | if (TestCpuFlag(kCpuHasLSX)) { |
4560 | | RAWToYJRow = RAWToYJRow_Any_LSX; |
4561 | | if (IS_ALIGNED(width, 16)) { |
4562 | | RAWToYJRow = RAWToYJRow_LSX; |
4563 | | } |
4564 | | } |
4565 | | #endif |
4566 | | #if defined(HAS_RAWTOYJROW_LASX) |
4567 | | if (TestCpuFlag(kCpuHasLASX)) { |
4568 | | RAWToYJRow = RAWToYJRow_Any_LASX; |
4569 | | if (IS_ALIGNED(width, 32)) { |
4570 | | RAWToYJRow = RAWToYJRow_LASX; |
4571 | | } |
4572 | | } |
4573 | | #endif |
4574 | | #if defined(HAS_RAWTOYJROW_RVV) |
4575 | | if (TestCpuFlag(kCpuHasRVV)) { |
4576 | | RAWToYJRow = RAWToYJRow_RVV; |
4577 | | } |
4578 | | #endif |
4579 | | |
4580 | 12 | for (y = 0; y < height; ++y) { |
4581 | 6 | RAWToYJRow(src_raw, dst_yj, width); |
4582 | 6 | src_raw += src_stride_raw; |
4583 | 6 | dst_yj += dst_stride_yj; |
4584 | 6 | } |
4585 | 6 | return 0; |
4586 | 6 | } |
4587 | | |
4588 | | // Convert Android420 to I420. |
4589 | | LIBYUV_API |
4590 | | int Android420ToI420(const uint8_t* src_y, |
4591 | | int src_stride_y, |
4592 | | const uint8_t* src_u, |
4593 | | int src_stride_u, |
4594 | | const uint8_t* src_v, |
4595 | | int src_stride_v, |
4596 | | int src_pixel_stride_uv, |
4597 | | uint8_t* dst_y, |
4598 | | int dst_stride_y, |
4599 | | uint8_t* dst_u, |
4600 | | int dst_stride_u, |
4601 | | uint8_t* dst_v, |
4602 | | int dst_stride_v, |
4603 | | int width, |
4604 | 0 | int height) { |
4605 | 0 | return Android420ToI420Rotate(src_y, src_stride_y, src_u, src_stride_u, src_v, |
4606 | 0 | src_stride_v, src_pixel_stride_uv, dst_y, |
4607 | 0 | dst_stride_y, dst_u, dst_stride_u, dst_v, |
4608 | 0 | dst_stride_v, width, height, kRotate0); |
4609 | 0 | } |
4610 | | |
4611 | | // depth is source bits measured from lsb; For msb use 16 |
4612 | | static int Biplanar16bitTo8bit(const uint16_t* src_y, |
4613 | | int src_stride_y, |
4614 | | const uint16_t* src_uv, |
4615 | | int src_stride_uv, |
4616 | | uint8_t* dst_y, |
4617 | | int dst_stride_y, |
4618 | | uint8_t* dst_uv, |
4619 | | int dst_stride_uv, |
4620 | | int width, |
4621 | | int height, |
4622 | | int subsample_x, |
4623 | | int subsample_y, |
4624 | 0 | int depth) { |
4625 | 0 | int uv_width = SUBSAMPLE(width, subsample_x, subsample_x); |
4626 | 0 | int uv_height = SUBSAMPLE(height, subsample_y, subsample_y); |
4627 | 0 | int scale = 1 << (24 - depth); |
4628 | 0 | if ((!src_y && dst_y) || !src_uv || !dst_uv || width <= 0 || height == 0) { |
4629 | 0 | return -1; |
4630 | 0 | } |
4631 | | // Negative height means invert the image. |
4632 | 0 | if (height < 0) { |
4633 | 0 | height = -height; |
4634 | 0 | uv_height = -uv_height; |
4635 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
4636 | 0 | src_uv = src_uv + (uv_height - 1) * src_stride_uv; |
4637 | 0 | src_stride_y = -src_stride_y; |
4638 | 0 | src_stride_uv = -src_stride_uv; |
4639 | 0 | } |
4640 | | |
4641 | | // Convert Y plane. |
4642 | 0 | if (dst_y) { |
4643 | 0 | Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width, |
4644 | 0 | height); |
4645 | 0 | } |
4646 | | // Convert UV planes. |
4647 | 0 | Convert16To8Plane(src_uv, src_stride_uv, dst_uv, dst_stride_uv, scale, |
4648 | 0 | uv_width * 2, uv_height); |
4649 | 0 | return 0; |
4650 | 0 | } |
4651 | | |
4652 | | // Convert 10 bit P010 to 8 bit NV12. |
4653 | | // Depth set to 16 because P010 uses 10 msb and this function keeps the upper 8 |
4654 | | // bits of the specified number of bits. |
4655 | | LIBYUV_API |
4656 | | int P010ToNV12(const uint16_t* src_y, |
4657 | | int src_stride_y, |
4658 | | const uint16_t* src_uv, |
4659 | | int src_stride_uv, |
4660 | | uint8_t* dst_y, |
4661 | | int dst_stride_y, |
4662 | | uint8_t* dst_uv, |
4663 | | int dst_stride_uv, |
4664 | | int width, |
4665 | 0 | int height) { |
4666 | 0 | return Biplanar16bitTo8bit(src_y, src_stride_y, src_uv, src_stride_uv, dst_y, |
4667 | 0 | dst_stride_y, dst_uv, dst_stride_uv, width, height, |
4668 | 0 | 1, 1, 16); |
4669 | 0 | } |
4670 | | |
4671 | | static int Planar8bitTo8bit(const uint8_t* src_y, |
4672 | | int src_stride_y, |
4673 | | const uint8_t* src_u, |
4674 | | int src_stride_u, |
4675 | | const uint8_t* src_v, |
4676 | | int src_stride_v, |
4677 | | uint8_t* dst_y, |
4678 | | int dst_stride_y, |
4679 | | uint8_t* dst_u, |
4680 | | int dst_stride_u, |
4681 | | uint8_t* dst_v, |
4682 | | int dst_stride_v, |
4683 | | int width, |
4684 | | int height, |
4685 | | int subsample_x, |
4686 | | int subsample_y, |
4687 | | int scale_y, |
4688 | | int bias_y, |
4689 | | int scale_uv, |
4690 | 0 | int bias_uv) { |
4691 | 0 | int uv_width = SUBSAMPLE(width, subsample_x, subsample_x); |
4692 | 0 | int uv_height = SUBSAMPLE(height, subsample_y, subsample_y); |
4693 | 0 | if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 || |
4694 | 0 | height == 0) { |
4695 | 0 | return -1; |
4696 | 0 | } |
4697 | | // Negative height means invert the image. |
4698 | 0 | if (height < 0) { |
4699 | 0 | height = -height; |
4700 | 0 | uv_height = -uv_height; |
4701 | 0 | src_y = src_y + (height - 1) * src_stride_y; |
4702 | 0 | src_u = src_u + (uv_height - 1) * src_stride_u; |
4703 | 0 | src_v = src_v + (uv_height - 1) * src_stride_v; |
4704 | 0 | src_stride_y = -src_stride_y; |
4705 | 0 | src_stride_u = -src_stride_u; |
4706 | 0 | src_stride_v = -src_stride_v; |
4707 | 0 | } |
4708 | | |
4709 | | // Convert Y plane. |
4710 | 0 | if (dst_y) { |
4711 | 0 | Convert8To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale_y, bias_y, |
4712 | 0 | width, height); |
4713 | 0 | } |
4714 | | // Convert UV planes. |
4715 | 0 | Convert8To8Plane(src_u, src_stride_u, dst_u, dst_stride_u, scale_uv, bias_uv, |
4716 | 0 | uv_width, uv_height); |
4717 | 0 | Convert8To8Plane(src_v, src_stride_v, dst_v, dst_stride_v, scale_uv, bias_uv, |
4718 | 0 | uv_width, uv_height); |
4719 | 0 | return 0; |
4720 | 0 | } |
4721 | | |
4722 | | LIBYUV_API |
4723 | | int J420ToI420(const uint8_t* src_y, |
4724 | | int src_stride_y, |
4725 | | const uint8_t* src_u, |
4726 | | int src_stride_u, |
4727 | | const uint8_t* src_v, |
4728 | | int src_stride_v, |
4729 | | uint8_t* dst_y, |
4730 | | int dst_stride_y, |
4731 | | uint8_t* dst_u, |
4732 | | int dst_stride_u, |
4733 | | uint8_t* dst_v, |
4734 | | int dst_stride_v, |
4735 | | int width, |
4736 | 0 | int height) { |
4737 | 0 | return Planar8bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v, |
4738 | 0 | src_stride_v, dst_y, dst_stride_y, dst_u, |
4739 | 0 | dst_stride_u, dst_v, dst_stride_v, width, height, 1, |
4740 | 0 | 1, 220, 16, 225, 16); |
4741 | 0 | } |
4742 | | |
4743 | | #ifdef __cplusplus |
4744 | | } // extern "C" |
4745 | | } // namespace libyuv |
4746 | | #endif |