Commit | Line | Data |
---|---|---|
72b9787e JB |
1 | /***************************************************************************** |
2 | * Copyright (C) 2013 x265 project | |
3 | * | |
4 | * Authors: Steve Borho <steve@borho.org> | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License as published by | |
8 | * the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * along with this program; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA. | |
19 | * | |
20 | * This program is also available under a commercial proprietary license. | |
21 | * For more information, contact us at license @ x265.com. | |
22 | *****************************************************************************/ | |
23 | ||
24 | #include "pixelharness.h" | |
25 | #include "primitives.h" | |
26 | ||
27 | using namespace x265; | |
28 | ||
29 | PixelHarness::PixelHarness() | |
30 | { | |
31 | /* [0] --- Random values | |
32 | * [1] --- Minimum | |
33 | * [2] --- Maximum */ | |
34 | for (int i = 0; i < BUFFSIZE; i++) | |
35 | { | |
36 | pixel_test_buff[0][i] = rand() % PIXEL_MAX; | |
37 | short_test_buff[0][i] = (rand() % (2 * SMAX + 1)) - SMAX - 1; // max(SHORT_MIN, min(rand(), SMAX)); | |
38 | short_test_buff1[0][i] = rand() & PIXEL_MAX; // For block copy only | |
39 | short_test_buff2[0][i] = rand() % 16383; // for addAvg | |
40 | int_test_buff[0][i] = rand() % SHORT_MAX; | |
41 | ushort_test_buff[0][i] = rand() % ((1 << 16) - 1); | |
42 | uchar_test_buff[0][i] = rand() % ((1 << 8) - 1); | |
43 | ||
44 | pixel_test_buff[1][i] = PIXEL_MIN; | |
45 | short_test_buff[1][i] = SMIN; | |
46 | short_test_buff1[1][i] = PIXEL_MIN; | |
47 | short_test_buff2[1][i] = -16384; | |
48 | int_test_buff[1][i] = SHORT_MIN; | |
49 | ushort_test_buff[1][i] = PIXEL_MIN; | |
50 | uchar_test_buff[1][i] = PIXEL_MIN; | |
51 | ||
52 | pixel_test_buff[2][i] = PIXEL_MAX; | |
53 | short_test_buff[2][i] = SMAX; | |
54 | short_test_buff1[2][i] = PIXEL_MAX; | |
55 | short_test_buff2[2][i] = 16383; | |
56 | int_test_buff[2][i] = SHORT_MAX; | |
57 | ushort_test_buff[2][i] = ((1 << 16) - 1); | |
58 | uchar_test_buff[2][i] = 255; | |
59 | ||
60 | pbuf1[i] = rand() & PIXEL_MAX; | |
61 | pbuf2[i] = rand() & PIXEL_MAX; | |
62 | pbuf3[i] = rand() & PIXEL_MAX; | |
63 | pbuf4[i] = rand() & PIXEL_MAX; | |
64 | ||
65 | sbuf1[i] = (rand() % (2 * SMAX + 1)) - SMAX - 1; //max(SHORT_MIN, min(rand(), SMAX)); | |
66 | sbuf2[i] = (rand() % (2 * SMAX + 1)) - SMAX - 1; //max(SHORT_MIN, min(rand(), SMAX)); | |
67 | ibuf1[i] = (rand() % (2 * SMAX + 1)) - SMAX - 1; | |
68 | psbuf1[i] = (rand() % 65) - 32; // range is between -32 to 32 | |
69 | sbuf3[i] = rand() % PIXEL_MAX; // for blockcopy only | |
70 | } | |
71 | } | |
72 | ||
73 | bool PixelHarness::check_pixelcmp(pixelcmp_t ref, pixelcmp_t opt) | |
74 | { | |
75 | int j = 0; | |
76 | intptr_t stride = STRIDE; | |
77 | ||
78 | for (int i = 0; i < ITERS; i++) | |
79 | { | |
80 | int index1 = rand() % TEST_CASES; | |
81 | int index2 = rand() % TEST_CASES; | |
82 | int vres = (int)checked(opt, pixel_test_buff[index1], stride, pixel_test_buff[index2] + j, stride); | |
83 | int cres = ref(pixel_test_buff[index1], stride, pixel_test_buff[index2] + j, stride); | |
84 | if (vres != cres) | |
85 | return false; | |
86 | ||
87 | reportfail(); | |
88 | j += INCR; | |
89 | } | |
90 | ||
91 | return true; | |
92 | } | |
93 | ||
94 | bool PixelHarness::check_pixelcmp_sp(pixelcmp_sp_t ref, pixelcmp_sp_t opt) | |
95 | { | |
96 | int j = 0; | |
97 | intptr_t stride = STRIDE; | |
98 | ||
99 | for (int i = 0; i < ITERS; i++) | |
100 | { | |
101 | int index1 = rand() % TEST_CASES; | |
102 | int index2 = rand() % TEST_CASES; | |
103 | int vres = (int)checked(opt, short_test_buff[index1], stride, pixel_test_buff[index2] + j, stride); | |
104 | int cres = ref(short_test_buff[index1], stride, pixel_test_buff[index2] + j, stride); | |
105 | if (vres != cres) | |
106 | return false; | |
107 | ||
108 | reportfail(); | |
109 | j += INCR; | |
110 | } | |
111 | ||
112 | return true; | |
113 | } | |
114 | ||
115 | bool PixelHarness::check_pixelcmp_ss(pixelcmp_ss_t ref, pixelcmp_ss_t opt) | |
116 | { | |
117 | int j = 0; | |
118 | intptr_t stride = STRIDE; | |
119 | ||
120 | for (int i = 0; i < ITERS; i++) | |
121 | { | |
122 | int index1 = rand() % TEST_CASES; | |
123 | int index2 = rand() % TEST_CASES; | |
124 | int vres = (int)checked(opt, short_test_buff[index1], stride, short_test_buff[index2] + j, stride); | |
125 | int cres = ref(short_test_buff[index1], stride, short_test_buff[index2] + j, stride); | |
126 | if (vres != cres) | |
127 | return false; | |
128 | ||
129 | reportfail(); | |
130 | j += INCR; | |
131 | } | |
132 | ||
133 | return true; | |
134 | } | |
135 | ||
136 | bool PixelHarness::check_pixelcmp_x3(pixelcmp_x3_t ref, pixelcmp_x3_t opt) | |
137 | { | |
138 | ALIGN_VAR_16(int, cres[16]); | |
139 | ALIGN_VAR_16(int, vres[16]); | |
140 | int j = 0; | |
141 | intptr_t stride = FENC_STRIDE - 5; | |
142 | for (int i = 0; i < ITERS; i++) | |
143 | { | |
144 | int index1 = rand() % TEST_CASES; | |
145 | int index2 = rand() % TEST_CASES; | |
146 | checked(opt, pixel_test_buff[index1], | |
147 | pixel_test_buff[index2] + j, | |
148 | pixel_test_buff[index2] + j + 1, | |
149 | pixel_test_buff[index2] + j + 2, stride, &vres[0]); | |
150 | ref(pixel_test_buff[index1], | |
151 | pixel_test_buff[index2] + j, | |
152 | pixel_test_buff[index2] + j + 1, | |
153 | pixel_test_buff[index2] + j + 2, stride, &cres[0]); | |
154 | if ((vres[0] != cres[0]) || ((vres[1] != cres[1])) || ((vres[2] != cres[2]))) | |
155 | return false; | |
156 | ||
157 | reportfail(); | |
158 | j += INCR; | |
159 | } | |
160 | ||
161 | return true; | |
162 | } | |
163 | ||
164 | bool PixelHarness::check_pixelcmp_x4(pixelcmp_x4_t ref, pixelcmp_x4_t opt) | |
165 | { | |
166 | ALIGN_VAR_16(int, cres[16]); | |
167 | ALIGN_VAR_16(int, vres[16]); | |
168 | int j = 0; | |
169 | intptr_t stride = FENC_STRIDE - 5; | |
170 | for (int i = 0; i < ITERS; i++) | |
171 | { | |
172 | int index1 = rand() % TEST_CASES; | |
173 | int index2 = rand() % TEST_CASES; | |
174 | checked(opt, pixel_test_buff[index1], | |
175 | pixel_test_buff[index2] + j, | |
176 | pixel_test_buff[index2] + j + 1, | |
177 | pixel_test_buff[index2] + j + 2, | |
178 | pixel_test_buff[index2] + j + 3, stride, &vres[0]); | |
179 | ref(pixel_test_buff[index1], | |
180 | pixel_test_buff[index2] + j, | |
181 | pixel_test_buff[index2] + j + 1, | |
182 | pixel_test_buff[index2] + j + 2, | |
183 | pixel_test_buff[index2] + j + 3, stride, &cres[0]); | |
184 | ||
185 | if ((vres[0] != cres[0]) || ((vres[1] != cres[1])) || ((vres[2] != cres[2])) || ((vres[3] != cres[3]))) | |
186 | return false; | |
187 | ||
188 | reportfail(); | |
189 | j += INCR; | |
190 | } | |
191 | ||
192 | return true; | |
193 | } | |
194 | ||
195 | bool PixelHarness::check_calresidual(calcresidual_t ref, calcresidual_t opt) | |
196 | { | |
197 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
198 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
199 | memset(ref_dest, 0, 64 * 64 * sizeof(int16_t)); | |
200 | memset(opt_dest, 0, 64 * 64 * sizeof(int16_t)); | |
201 | ||
202 | int j = 0; | |
203 | intptr_t stride = STRIDE; | |
204 | for (int i = 0; i < ITERS; i++) | |
205 | { | |
206 | int index = i % TEST_CASES; | |
207 | checked(opt, pbuf1 + j, pixel_test_buff[index] + j, opt_dest, stride); | |
208 | ref(pbuf1 + j, pixel_test_buff[index] + j, ref_dest, stride); | |
209 | ||
210 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
211 | return false; | |
212 | ||
213 | reportfail(); | |
214 | j += INCR; | |
215 | } | |
216 | ||
217 | return true; | |
218 | } | |
219 | ||
220 | bool PixelHarness::check_ssd_s(pixel_ssd_s_t ref, pixel_ssd_s_t opt) | |
221 | { | |
222 | int j = 0; | |
223 | for (int i = 0; i < ITERS; i++) | |
224 | { | |
225 | // NOTE: stride must be multiple of 16, because minimum block is 4x4 | |
226 | int stride = (STRIDE + (rand() % STRIDE)) & ~15; | |
227 | int cres = ref(sbuf1 + j, stride); | |
228 | int vres = (int)checked(opt, sbuf1 + j, (intptr_t)stride); | |
229 | ||
230 | if (cres != vres) | |
231 | { | |
232 | return false; | |
233 | } | |
234 | ||
235 | reportfail(); | |
236 | j += INCR; | |
237 | } | |
238 | ||
239 | return true; | |
240 | } | |
241 | ||
242 | bool PixelHarness::check_weightp(weightp_sp_t ref, weightp_sp_t opt) | |
243 | { | |
244 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
245 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
246 | ||
247 | memset(ref_dest, 0, 64 * 64 * sizeof(pixel)); | |
248 | memset(opt_dest, 0, 64 * 64 * sizeof(pixel)); | |
249 | int j = 0; | |
250 | int width = 2 * (rand() % 32 + 1); | |
251 | int height = 8; | |
252 | int w0 = rand() % 128; | |
253 | int shift = rand() % 15; | |
254 | int round = shift ? (1 << (shift - 1)) : 0; | |
255 | int offset = (rand() % 256) - 128; | |
256 | intptr_t stride = 64; | |
257 | for (int i = 0; i < ITERS; i++) | |
258 | { | |
259 | int index = i % TEST_CASES; | |
260 | checked(opt, short_test_buff[index] + j, opt_dest, stride, stride, width, height, w0, round, shift, offset); | |
261 | ref(short_test_buff[index] + j, ref_dest, stride, stride, width, height, w0, round, shift, offset); | |
262 | ||
263 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
264 | return false; | |
265 | ||
266 | reportfail(); | |
267 | j += INCR; | |
268 | } | |
269 | ||
270 | return true; | |
271 | } | |
272 | ||
273 | bool PixelHarness::check_weightp(weightp_pp_t ref, weightp_pp_t opt) | |
274 | { | |
275 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
276 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
277 | ||
278 | memset(ref_dest, 0, 64 * 64 * sizeof(pixel)); | |
279 | memset(opt_dest, 0, 64 * 64 * sizeof(pixel)); | |
280 | int j = 0; | |
281 | int width = 16 * (rand() % 4 + 1); | |
282 | int height = 8; | |
283 | int w0 = rand() % 128; | |
284 | int shift = rand() % 15; | |
285 | int round = shift ? (1 << (shift - 1)) : 0; | |
286 | int offset = (rand() % 256) - 128; | |
287 | intptr_t stride = 64; | |
288 | for (int i = 0; i < ITERS; i++) | |
289 | { | |
290 | int index = i % TEST_CASES; | |
291 | checked(opt, pixel_test_buff[index] + j, opt_dest, stride, width, height, w0, round, shift, offset); | |
292 | ref(pixel_test_buff[index] + j, ref_dest, stride, width, height, w0, round, shift, offset); | |
293 | ||
294 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
295 | return false; | |
296 | ||
297 | reportfail(); | |
298 | j += INCR; | |
299 | } | |
300 | ||
301 | return true; | |
302 | } | |
303 | ||
304 | bool PixelHarness::check_downscale_t(downscale_t ref, downscale_t opt) | |
305 | { | |
306 | ALIGN_VAR_16(pixel, ref_destf[32 * 32]); | |
307 | ALIGN_VAR_16(pixel, opt_destf[32 * 32]); | |
308 | ||
309 | ALIGN_VAR_16(pixel, ref_desth[32 * 32]); | |
310 | ALIGN_VAR_16(pixel, opt_desth[32 * 32]); | |
311 | ||
312 | ALIGN_VAR_16(pixel, ref_destv[32 * 32]); | |
313 | ALIGN_VAR_16(pixel, opt_destv[32 * 32]); | |
314 | ||
315 | ALIGN_VAR_16(pixel, ref_destc[32 * 32]); | |
316 | ALIGN_VAR_16(pixel, opt_destc[32 * 32]); | |
317 | ||
318 | intptr_t src_stride = 64; | |
319 | intptr_t dst_stride = 32; | |
320 | int bx = 32; | |
321 | int by = 32; | |
322 | int j = 0; | |
323 | for (int i = 0; i < ITERS; i++) | |
324 | { | |
325 | int index = i % TEST_CASES; | |
326 | ref(pixel_test_buff[index] + j, ref_destf, ref_desth, ref_destv, | |
327 | ref_destc, src_stride, dst_stride, bx, by); | |
328 | checked(opt, pixel_test_buff[index] + j, opt_destf, opt_desth, opt_destv, | |
329 | opt_destc, src_stride, dst_stride, bx, by); | |
330 | ||
331 | if (memcmp(ref_destf, opt_destf, 32 * 32 * sizeof(pixel))) | |
332 | return false; | |
333 | if (memcmp(ref_desth, opt_desth, 32 * 32 * sizeof(pixel))) | |
334 | return false; | |
335 | if (memcmp(ref_destv, opt_destv, 32 * 32 * sizeof(pixel))) | |
336 | return false; | |
337 | if (memcmp(ref_destc, opt_destc, 32 * 32 * sizeof(pixel))) | |
338 | return false; | |
339 | ||
340 | reportfail(); | |
341 | j += INCR; | |
342 | } | |
343 | ||
344 | return true; | |
345 | } | |
346 | ||
b53f7c52 | 347 | bool PixelHarness::check_cpy2Dto1D_shl_t(cpy2Dto1D_shl_t ref, cpy2Dto1D_shl_t opt) |
72b9787e JB |
348 | { |
349 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
350 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
351 | ||
352 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
353 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
354 | ||
355 | int j = 0; | |
356 | intptr_t stride = STRIDE; | |
357 | for (int i = 0; i < ITERS; i++) | |
358 | { | |
359 | int shift = (rand() % 7 + 1); | |
360 | ||
361 | int index = i % TEST_CASES; | |
b53f7c52 JB |
362 | checked(opt, opt_dest, short_test_buff[index] + j, stride, shift); |
363 | ref(ref_dest, short_test_buff[index] + j, stride, shift); | |
72b9787e JB |
364 | |
365 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
366 | return false; | |
367 | ||
368 | reportfail(); | |
369 | j += INCR; | |
370 | } | |
371 | ||
372 | return true; | |
373 | } | |
374 | ||
b53f7c52 | 375 | bool PixelHarness::check_cpy2Dto1D_shr_t(cpy2Dto1D_shr_t ref, cpy2Dto1D_shr_t opt) |
72b9787e JB |
376 | { |
377 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
378 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
379 | ||
380 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
381 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
382 | ||
383 | int j = 0; | |
384 | intptr_t stride = STRIDE; | |
385 | for (int i = 0; i < ITERS; i++) | |
386 | { | |
387 | int shift = (rand() % 7 + 1); | |
388 | ||
389 | int index = i % TEST_CASES; | |
b53f7c52 JB |
390 | checked(opt, opt_dest, short_test_buff[index] + j, stride, shift); |
391 | ref(ref_dest, short_test_buff[index] + j, stride, shift); | |
72b9787e JB |
392 | |
393 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
394 | return false; | |
395 | ||
396 | reportfail(); | |
397 | j += INCR; | |
398 | } | |
399 | ||
400 | return true; | |
401 | } | |
402 | ||
403 | bool PixelHarness::check_copy_cnt_t(copy_cnt_t ref, copy_cnt_t opt) | |
404 | { | |
405 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
406 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
407 | ||
408 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
409 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
410 | ||
411 | int j = 0; | |
412 | intptr_t stride = STRIDE; | |
413 | for (int i = 0; i < ITERS; i++) | |
414 | { | |
415 | int index = i % TEST_CASES; | |
416 | int opt_cnt = (int)checked(opt, opt_dest, short_test_buff1[index] + j, stride); | |
417 | int ref_cnt = ref(ref_dest, short_test_buff1[index] + j, stride); | |
418 | ||
419 | if ((ref_cnt != opt_cnt) || memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
420 | return false; | |
421 | ||
422 | reportfail(); | |
423 | j += INCR; | |
424 | } | |
425 | ||
426 | return true; | |
427 | } | |
428 | ||
b53f7c52 | 429 | bool PixelHarness::check_cpy1Dto2D_shl_t(cpy1Dto2D_shl_t ref, cpy1Dto2D_shl_t opt) |
72b9787e JB |
430 | { |
431 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
432 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
433 | ||
434 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
435 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
436 | ||
437 | int j = 0; | |
438 | intptr_t stride = STRIDE; | |
439 | for (int i = 0; i < ITERS; i++) | |
440 | { | |
441 | int shift = (rand() % 7 + 1); | |
442 | ||
443 | int index = i % TEST_CASES; | |
b53f7c52 JB |
444 | checked(opt, opt_dest, short_test_buff[index] + j, stride, shift); |
445 | ref(ref_dest, short_test_buff[index] + j, stride, shift); | |
72b9787e JB |
446 | |
447 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
448 | return false; | |
449 | ||
450 | reportfail(); | |
451 | j += INCR; | |
452 | } | |
453 | ||
454 | return true; | |
455 | } | |
456 | ||
b53f7c52 | 457 | bool PixelHarness::check_cpy1Dto2D_shr_t(cpy1Dto2D_shr_t ref, cpy1Dto2D_shr_t opt) |
72b9787e JB |
458 | { |
459 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
460 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
461 | ||
462 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
463 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
464 | ||
465 | int j = 0; | |
466 | intptr_t stride = STRIDE; | |
467 | for (int i = 0; i < ITERS; i++) | |
468 | { | |
469 | int shift = (rand() % 7 + 1); | |
470 | ||
471 | int index = i % TEST_CASES; | |
472 | checked(opt, opt_dest, short_test_buff[index] + j, stride, shift); | |
473 | ref(ref_dest, short_test_buff[index] + j, stride, shift); | |
474 | ||
475 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
476 | return false; | |
477 | ||
478 | reportfail(); | |
479 | j += INCR; | |
480 | } | |
481 | ||
482 | return true; | |
483 | } | |
484 | ||
485 | bool PixelHarness::check_pixelavg_pp(pixelavg_pp_t ref, pixelavg_pp_t opt) | |
486 | { | |
487 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
488 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
489 | ||
490 | int j = 0; | |
491 | ||
492 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
493 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
494 | ||
495 | intptr_t stride = STRIDE; | |
496 | for (int i = 0; i < ITERS; i++) | |
497 | { | |
498 | int index1 = rand() % TEST_CASES; | |
499 | int index2 = rand() % TEST_CASES; | |
500 | checked(ref, ref_dest, stride, pixel_test_buff[index1] + j, | |
501 | stride, pixel_test_buff[index2] + j, stride, 32); | |
502 | opt(opt_dest, stride, pixel_test_buff[index1] + j, | |
503 | stride, pixel_test_buff[index2] + j, stride, 32); | |
504 | ||
505 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
506 | return false; | |
507 | ||
508 | reportfail(); | |
509 | j += INCR; | |
510 | } | |
511 | ||
512 | return true; | |
513 | } | |
514 | ||
515 | bool PixelHarness::check_copy_pp(copy_pp_t ref, copy_pp_t opt) | |
516 | { | |
517 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
518 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
519 | ||
520 | // we don't know the partition size so we are checking the entire output buffer so | |
521 | // we must initialize the buffers | |
522 | memset(ref_dest, 0, sizeof(ref_dest)); | |
523 | memset(opt_dest, 0, sizeof(opt_dest)); | |
524 | ||
525 | int j = 0; | |
526 | intptr_t stride = STRIDE; | |
527 | for (int i = 0; i < ITERS; i++) | |
528 | { | |
529 | int index = i % TEST_CASES; | |
530 | checked(opt, opt_dest, stride, pixel_test_buff[index] + j, stride); | |
531 | ref(ref_dest, stride, pixel_test_buff[index] + j, stride); | |
532 | ||
533 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
534 | return false; | |
535 | ||
536 | reportfail(); | |
537 | j += INCR; | |
538 | } | |
539 | ||
540 | return true; | |
541 | } | |
542 | ||
543 | bool PixelHarness::check_copy_sp(copy_sp_t ref, copy_sp_t opt) | |
544 | { | |
545 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
546 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
547 | ||
548 | // we don't know the partition size so we are checking the entire output buffer so | |
549 | // we must initialize the buffers | |
550 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
551 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
552 | ||
553 | int j = 0; | |
554 | intptr_t stride1 = 64, stride2 = STRIDE; | |
555 | for (int i = 0; i < ITERS; i++) | |
556 | { | |
557 | int index = i % TEST_CASES; | |
558 | checked(opt, opt_dest, stride1, short_test_buff1[index] + j, stride2); | |
559 | ref(ref_dest, stride1, short_test_buff1[index] + j, stride2); | |
560 | ||
561 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
562 | return false; | |
563 | ||
564 | reportfail(); | |
565 | j += INCR; | |
566 | } | |
567 | ||
568 | return true; | |
569 | } | |
570 | ||
571 | bool PixelHarness::check_copy_ps(copy_ps_t ref, copy_ps_t opt) | |
572 | { | |
573 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
574 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
575 | ||
576 | // we don't know the partition size so we are checking the entire output buffer so | |
577 | // we must initialize the buffers | |
578 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
579 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
580 | ||
581 | int j = 0; | |
582 | intptr_t stride = STRIDE; | |
583 | for (int i = 0; i < ITERS; i++) | |
584 | { | |
585 | int index = i % TEST_CASES; | |
586 | checked(opt, opt_dest, stride, pixel_test_buff[index] + j, stride); | |
587 | ref(ref_dest, stride, pixel_test_buff[index] + j, stride); | |
588 | ||
589 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
590 | return false; | |
591 | ||
592 | reportfail(); | |
593 | j += INCR; | |
594 | } | |
595 | ||
596 | return true; | |
597 | } | |
598 | ||
599 | bool PixelHarness::check_copy_ss(copy_ss_t ref, copy_ss_t opt) | |
600 | { | |
601 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
602 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
603 | ||
604 | // we don't know the partition size so we are checking the entire output buffer so | |
605 | // we must initialize the buffers | |
606 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
607 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
608 | ||
609 | int j = 0; | |
610 | intptr_t stride = STRIDE; | |
611 | for (int i = 0; i < ITERS; i++) | |
612 | { | |
613 | int index = i % TEST_CASES; | |
614 | checked(opt, opt_dest, stride, short_test_buff1[index] + j, stride); | |
615 | ref(ref_dest, stride, short_test_buff1[index] + j, stride); | |
616 | ||
617 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
618 | return false; | |
619 | ||
620 | reportfail(); | |
621 | j += INCR; | |
622 | } | |
623 | ||
624 | return true; | |
625 | } | |
626 | ||
627 | bool PixelHarness::check_blockfill_s(blockfill_s_t ref, blockfill_s_t opt) | |
628 | { | |
629 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
630 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
631 | ||
632 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
633 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
634 | ||
635 | intptr_t stride = 64; | |
636 | for (int i = 0; i < ITERS; i++) | |
637 | { | |
638 | int16_t value = (rand() % SHORT_MAX) + 1; | |
639 | ||
640 | checked(opt, opt_dest, stride, value); | |
641 | ref(ref_dest, stride, value); | |
642 | ||
643 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
644 | return false; | |
645 | ||
646 | reportfail(); | |
647 | } | |
648 | ||
649 | return true; | |
650 | } | |
651 | ||
652 | bool PixelHarness::check_pixel_sub_ps(pixel_sub_ps_t ref, pixel_sub_ps_t opt) | |
653 | { | |
654 | ALIGN_VAR_16(int16_t, ref_dest[64 * 64]); | |
655 | ALIGN_VAR_16(int16_t, opt_dest[64 * 64]); | |
656 | ||
657 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
658 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
659 | ||
660 | int j = 0; | |
661 | intptr_t stride2 = 64, stride = STRIDE; | |
662 | for (int i = 0; i < 1; i++) | |
663 | { | |
664 | int index1 = rand() % TEST_CASES; | |
665 | int index2 = rand() % TEST_CASES; | |
666 | checked(opt, opt_dest, stride2, pixel_test_buff[index1] + j, | |
667 | pixel_test_buff[index2] + j, stride, stride); | |
668 | ref(ref_dest, stride2, pixel_test_buff[index1] + j, | |
669 | pixel_test_buff[index2] + j, stride, stride); | |
670 | ||
671 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int16_t))) | |
672 | return false; | |
673 | ||
674 | reportfail(); | |
675 | j += INCR; | |
676 | } | |
677 | ||
678 | return true; | |
679 | } | |
680 | ||
681 | bool PixelHarness::check_scale_pp(scale_t ref, scale_t opt) | |
682 | { | |
683 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
684 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
685 | ||
686 | memset(ref_dest, 0, sizeof(ref_dest)); | |
687 | memset(opt_dest, 0, sizeof(opt_dest)); | |
688 | ||
689 | int j = 0; | |
690 | intptr_t stride = STRIDE; | |
691 | for (int i = 0; i < ITERS; i++) | |
692 | { | |
693 | int index = i % TEST_CASES; | |
694 | checked(opt, opt_dest, pixel_test_buff[index] + j, stride); | |
695 | ref(ref_dest, pixel_test_buff[index] + j, stride); | |
696 | ||
697 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
698 | return false; | |
699 | ||
700 | reportfail(); | |
701 | j += INCR; | |
702 | } | |
703 | ||
704 | return true; | |
705 | } | |
706 | ||
707 | bool PixelHarness::check_transpose(transpose_t ref, transpose_t opt) | |
708 | { | |
709 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
710 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
711 | ||
712 | memset(ref_dest, 0, sizeof(ref_dest)); | |
713 | memset(opt_dest, 0, sizeof(opt_dest)); | |
714 | ||
715 | int j = 0; | |
716 | intptr_t stride = STRIDE; | |
717 | for (int i = 0; i < ITERS; i++) | |
718 | { | |
719 | int index = i % TEST_CASES; | |
720 | checked(opt, opt_dest, pixel_test_buff[index] + j, stride); | |
721 | ref(ref_dest, pixel_test_buff[index] + j, stride); | |
722 | ||
723 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
724 | return false; | |
725 | ||
726 | reportfail(); | |
727 | j += INCR; | |
728 | } | |
729 | ||
730 | return true; | |
731 | } | |
732 | ||
733 | bool PixelHarness::check_pixel_add_ps(pixel_add_ps_t ref, pixel_add_ps_t opt) | |
734 | { | |
735 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
736 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
737 | ||
738 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
739 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
740 | ||
741 | int j = 0; | |
742 | intptr_t stride2 = 64, stride = STRIDE; | |
743 | for (int i = 0; i < ITERS; i++) | |
744 | { | |
745 | int index1 = rand() % TEST_CASES; | |
746 | int index2 = rand() % TEST_CASES; | |
747 | checked(opt, opt_dest, stride2, pixel_test_buff[index1] + j, short_test_buff[index2] + j, stride, stride); | |
748 | ref(ref_dest, stride2, pixel_test_buff[index1] + j, short_test_buff[index2] + j, stride, stride); | |
749 | ||
750 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
751 | return false; | |
752 | ||
753 | reportfail(); | |
754 | j += INCR; | |
755 | } | |
756 | ||
757 | return true; | |
758 | } | |
759 | ||
760 | bool PixelHarness::check_pixel_var(var_t ref, var_t opt) | |
761 | { | |
762 | int j = 0; | |
763 | ||
764 | intptr_t stride = STRIDE; | |
765 | ||
766 | for (int i = 0; i < ITERS; i++) | |
767 | { | |
768 | int index = i % TEST_CASES; | |
769 | uint64_t vres = checked(opt, pixel_test_buff[index], stride); | |
770 | uint64_t cres = ref(pixel_test_buff[index], stride); | |
771 | if (vres != cres) | |
772 | return false; | |
773 | ||
774 | reportfail(); | |
775 | j += INCR; | |
776 | } | |
777 | ||
778 | return true; | |
779 | } | |
780 | ||
781 | bool PixelHarness::check_ssim_4x4x2_core(ssim_4x4x2_core_t ref, ssim_4x4x2_core_t opt) | |
782 | { | |
783 | ALIGN_VAR_32(int, sum0[2][4]); | |
784 | ALIGN_VAR_32(int, sum1[2][4]); | |
785 | ||
786 | for (int i = 0; i < ITERS; i++) | |
787 | { | |
788 | intptr_t stride = rand() % 64; | |
789 | int index1 = rand() % TEST_CASES; | |
790 | int index2 = rand() % TEST_CASES; | |
791 | ref(pixel_test_buff[index1] + i, stride, pixel_test_buff[index2] + i, stride, sum0); | |
792 | checked(opt, pixel_test_buff[index1] + i, stride, pixel_test_buff[index2] + i, stride, sum1); | |
793 | ||
794 | if (memcmp(sum0, sum1, sizeof(sum0))) | |
795 | return false; | |
796 | ||
797 | reportfail(); | |
798 | } | |
799 | ||
800 | return true; | |
801 | } | |
802 | ||
803 | /* TODO: This function causes crashes when checked. Is this a real bug? */ | |
804 | bool PixelHarness::check_ssim_end(ssim_end4_t ref, ssim_end4_t opt) | |
805 | { | |
806 | ALIGN_VAR_32(int, sum0[5][4]); | |
807 | ALIGN_VAR_32(int, sum1[5][4]); | |
808 | ||
809 | for (int i = 0; i < ITERS; i++) | |
810 | { | |
811 | for (int j = 0; j < 5; j++) | |
812 | { | |
813 | for (int k = 0; k < 4; k++) | |
814 | { | |
815 | sum0[j][k] = rand() % (1 << 12); | |
816 | sum1[j][k] = rand() % (1 << 12); | |
817 | } | |
818 | } | |
819 | ||
820 | int width = (rand() % 4) + 1; // range[1-4] | |
821 | float cres = ref(sum0, sum1, width); | |
822 | float vres = checked_float(opt, sum0, sum1, width); | |
823 | if (fabs(vres - cres) > 0.00001) | |
824 | return false; | |
825 | ||
826 | reportfail(); | |
827 | } | |
828 | ||
829 | return true; | |
830 | } | |
831 | ||
832 | bool PixelHarness::check_addAvg(addAvg_t ref, addAvg_t opt) | |
833 | { | |
834 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
835 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
836 | ||
837 | int j = 0; | |
838 | ||
839 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
840 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
841 | intptr_t stride = STRIDE; | |
842 | ||
843 | for (int i = 0; i < ITERS; i++) | |
844 | { | |
845 | int index1 = rand() % TEST_CASES; | |
846 | int index2 = rand() % TEST_CASES; | |
847 | ref(short_test_buff2[index1] + j, short_test_buff2[index2] + j, ref_dest, stride, stride, stride); | |
848 | checked(opt, short_test_buff2[index1] + j, short_test_buff2[index2] + j, opt_dest, stride, stride, stride); | |
849 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
850 | return false; | |
851 | ||
852 | reportfail(); | |
853 | j += INCR; | |
854 | } | |
855 | ||
856 | return true; | |
857 | } | |
858 | ||
859 | bool PixelHarness::check_saoCuOrgE0_t(saoCuOrgE0_t ref, saoCuOrgE0_t opt) | |
860 | { | |
861 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
862 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
863 | ||
864 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
865 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
866 | ||
867 | int j = 0; | |
868 | ||
869 | for (int i = 0; i < ITERS; i++) | |
870 | { | |
871 | int width = 16 * (rand() % 4 + 1); | |
872 | int8_t sign = rand() % 3; | |
873 | if (sign == 2) | |
874 | { | |
875 | sign = -1; | |
876 | } | |
877 | ||
878 | ref(ref_dest, psbuf1 + j, width, sign); | |
879 | checked(opt, opt_dest, psbuf1 + j, width, sign); | |
880 | ||
881 | if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel))) | |
882 | return false; | |
883 | ||
884 | reportfail(); | |
885 | j += INCR; | |
886 | } | |
887 | ||
888 | return true; | |
889 | } | |
890 | ||
891 | bool PixelHarness::check_planecopy_sp(planecopy_sp_t ref, planecopy_sp_t opt) | |
892 | { | |
893 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
894 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
895 | ||
896 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
897 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
898 | ||
899 | int width = 16 + rand() % 48; | |
900 | int height = 16 + rand() % 48; | |
901 | intptr_t srcStride = 64; | |
902 | intptr_t dstStride = width; | |
903 | int j = 0; | |
904 | ||
905 | for (int i = 0; i < ITERS; i++) | |
906 | { | |
907 | int index = i % TEST_CASES; | |
908 | checked(opt, ushort_test_buff[index] + j, srcStride, opt_dest, dstStride, width, height, (int)8, (uint16_t)255); | |
909 | ref(ushort_test_buff[index] + j, srcStride, ref_dest, dstStride, width, height, (int)8, (uint16_t)255); | |
910 | ||
911 | if (memcmp(ref_dest, opt_dest, width * height * sizeof(pixel))) | |
912 | return false; | |
913 | ||
914 | reportfail(); | |
915 | j += INCR; | |
916 | } | |
917 | ||
918 | return true; | |
919 | } | |
920 | ||
921 | bool PixelHarness::check_planecopy_cp(planecopy_cp_t ref, planecopy_cp_t opt) | |
922 | { | |
923 | ALIGN_VAR_16(pixel, ref_dest[64 * 64]); | |
924 | ALIGN_VAR_16(pixel, opt_dest[64 * 64]); | |
925 | ||
926 | memset(ref_dest, 0xCD, sizeof(ref_dest)); | |
927 | memset(opt_dest, 0xCD, sizeof(opt_dest)); | |
928 | ||
929 | int width = 16 + rand() % 48; | |
930 | int height = 16 + rand() % 48; | |
931 | intptr_t srcStride = 64; | |
932 | intptr_t dstStride = width; | |
933 | int j = 0; | |
934 | ||
935 | for (int i = 0; i < ITERS; i++) | |
936 | { | |
937 | int index = i % TEST_CASES; | |
938 | checked(opt, uchar_test_buff[index] + j, srcStride, opt_dest, dstStride, width, height, (int)2); | |
939 | ref(uchar_test_buff[index] + j, srcStride, ref_dest, dstStride, width, height, (int)2); | |
940 | ||
941 | if (memcmp(ref_dest, opt_dest, width * height * sizeof(pixel))) | |
942 | return false; | |
943 | ||
944 | reportfail(); | |
945 | j += INCR; | |
946 | } | |
947 | ||
948 | return true; | |
949 | } | |
950 | ||
951 | bool PixelHarness::testPartition(int part, const EncoderPrimitives& ref, const EncoderPrimitives& opt) | |
952 | { | |
953 | if (opt.satd[part]) | |
954 | { | |
955 | if (!check_pixelcmp(ref.satd[part], opt.satd[part])) | |
956 | { | |
957 | printf("satd[%s]: failed!\n", lumaPartStr[part]); | |
958 | return false; | |
959 | } | |
960 | } | |
961 | ||
962 | if (opt.sa8d_inter[part]) | |
963 | { | |
964 | if (!check_pixelcmp(ref.sa8d_inter[part], opt.sa8d_inter[part])) | |
965 | { | |
966 | printf("sa8d_inter[%s]: failed!\n", lumaPartStr[part]); | |
967 | return false; | |
968 | } | |
969 | } | |
970 | ||
971 | if (opt.sad[part]) | |
972 | { | |
973 | if (!check_pixelcmp(ref.sad[part], opt.sad[part])) | |
974 | { | |
975 | printf("sad[%s]: failed!\n", lumaPartStr[part]); | |
976 | return false; | |
977 | } | |
978 | } | |
979 | ||
980 | if (opt.sse_pp[part]) | |
981 | { | |
982 | if (!check_pixelcmp(ref.sse_pp[part], opt.sse_pp[part])) | |
983 | { | |
984 | printf("sse_pp[%s]: failed!\n", lumaPartStr[part]); | |
985 | return false; | |
986 | } | |
987 | } | |
988 | ||
989 | if (opt.sse_sp[part]) | |
990 | { | |
991 | if (!check_pixelcmp_sp(ref.sse_sp[part], opt.sse_sp[part])) | |
992 | { | |
993 | printf("sse_sp[%s]: failed!\n", lumaPartStr[part]); | |
994 | return false; | |
995 | } | |
996 | } | |
997 | ||
998 | if (opt.sse_ss[part]) | |
999 | { | |
1000 | if (!check_pixelcmp_ss(ref.sse_ss[part], opt.sse_ss[part])) | |
1001 | { | |
1002 | printf("sse_ss[%s]: failed!\n", lumaPartStr[part]); | |
1003 | return false; | |
1004 | } | |
1005 | } | |
1006 | ||
1007 | if (opt.sad_x3[part]) | |
1008 | { | |
1009 | if (!check_pixelcmp_x3(ref.sad_x3[part], opt.sad_x3[part])) | |
1010 | { | |
1011 | printf("sad_x3[%s]: failed!\n", lumaPartStr[part]); | |
1012 | return false; | |
1013 | } | |
1014 | } | |
1015 | ||
1016 | if (opt.sad_x4[part]) | |
1017 | { | |
1018 | if (!check_pixelcmp_x4(ref.sad_x4[part], opt.sad_x4[part])) | |
1019 | { | |
1020 | printf("sad_x4[%s]: failed!\n", lumaPartStr[part]); | |
1021 | return false; | |
1022 | } | |
1023 | } | |
1024 | ||
1025 | if (opt.pixelavg_pp[part]) | |
1026 | { | |
1027 | if (!check_pixelavg_pp(ref.pixelavg_pp[part], opt.pixelavg_pp[part])) | |
1028 | { | |
1029 | printf("pixelavg_pp[%s]: failed!\n", lumaPartStr[part]); | |
1030 | return false; | |
1031 | } | |
1032 | } | |
1033 | ||
1034 | if (opt.luma_copy_pp[part]) | |
1035 | { | |
1036 | if (!check_copy_pp(ref.luma_copy_pp[part], opt.luma_copy_pp[part])) | |
1037 | { | |
1038 | printf("luma_copy_pp[%s] failed\n", lumaPartStr[part]); | |
1039 | return false; | |
1040 | } | |
1041 | } | |
1042 | ||
1043 | if (opt.luma_copy_sp[part]) | |
1044 | { | |
1045 | if (!check_copy_sp(ref.luma_copy_sp[part], opt.luma_copy_sp[part])) | |
1046 | { | |
1047 | printf("luma_copy_sp[%s] failed\n", lumaPartStr[part]); | |
1048 | return false; | |
1049 | } | |
1050 | } | |
1051 | ||
1052 | if (opt.luma_copy_ps[part]) | |
1053 | { | |
1054 | if (!check_copy_ps(ref.luma_copy_ps[part], opt.luma_copy_ps[part])) | |
1055 | { | |
1056 | printf("luma_copy_ps[%s] failed\n", lumaPartStr[part]); | |
1057 | return false; | |
1058 | } | |
1059 | } | |
1060 | ||
1061 | if (opt.luma_copy_ss[part]) | |
1062 | { | |
1063 | if (!check_copy_ss(ref.luma_copy_ss[part], opt.luma_copy_ss[part])) | |
1064 | { | |
1065 | printf("luma_copy_ss[%s] failed\n", lumaPartStr[part]); | |
1066 | return false; | |
1067 | } | |
1068 | } | |
1069 | ||
1070 | if (opt.luma_addAvg[part]) | |
1071 | { | |
1072 | if (!check_addAvg(ref.luma_addAvg[part], opt.luma_addAvg[part])) | |
1073 | { | |
1074 | printf("luma_addAvg[%s] failed\n", lumaPartStr[part]); | |
1075 | return false; | |
1076 | } | |
1077 | } | |
1078 | ||
1079 | if (part < NUM_SQUARE_BLOCKS) | |
1080 | { | |
1081 | if (opt.luma_sub_ps[part]) | |
1082 | { | |
1083 | if (!check_pixel_sub_ps(ref.luma_sub_ps[part], opt.luma_sub_ps[part])) | |
1084 | { | |
1085 | printf("luma_sub_ps[%s] failed\n", lumaPartStr[part]); | |
1086 | return false; | |
1087 | } | |
1088 | } | |
1089 | ||
1090 | if (opt.luma_add_ps[part]) | |
1091 | { | |
1092 | if (!check_pixel_add_ps(ref.luma_add_ps[part], opt.luma_add_ps[part])) | |
1093 | { | |
1094 | printf("luma_add_ps[%s] failed\n", lumaPartStr[part]); | |
1095 | return false; | |
1096 | } | |
1097 | } | |
1098 | } | |
1099 | ||
1100 | for (int i = 0; i < X265_CSP_COUNT; i++) | |
1101 | { | |
1102 | if (opt.chroma[i].copy_pp[part]) | |
1103 | { | |
1104 | if (!check_copy_pp(ref.chroma[i].copy_pp[part], opt.chroma[i].copy_pp[part])) | |
1105 | { | |
1106 | printf("chroma_copy_pp[%s][%s] failed\n", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1107 | return false; | |
1108 | } | |
1109 | } | |
1110 | if (opt.chroma[i].copy_sp[part]) | |
1111 | { | |
1112 | if (!check_copy_sp(ref.chroma[i].copy_sp[part], opt.chroma[i].copy_sp[part])) | |
1113 | { | |
1114 | printf("chroma_copy_sp[%s][%s] failed\n", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1115 | return false; | |
1116 | } | |
1117 | } | |
1118 | if (opt.chroma[i].copy_ps[part]) | |
1119 | { | |
1120 | if (!check_copy_ps(ref.chroma[i].copy_ps[part], opt.chroma[i].copy_ps[part])) | |
1121 | { | |
1122 | printf("chroma_copy_ps[%s][%s] failed\n", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1123 | return false; | |
1124 | } | |
1125 | } | |
1126 | if (opt.chroma[i].copy_ss[part]) | |
1127 | { | |
1128 | if (!check_copy_ss(ref.chroma[i].copy_ss[part], opt.chroma[i].copy_ss[part])) | |
1129 | { | |
1130 | printf("chroma_copy_ss[%s][%s] failed\n", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1131 | return false; | |
1132 | } | |
1133 | } | |
1134 | if (opt.chroma[i].addAvg[part]) | |
1135 | { | |
1136 | if (!check_addAvg(ref.chroma[i].addAvg[part], opt.chroma[i].addAvg[part])) | |
1137 | { | |
1138 | printf("chroma_addAvg[%s][%s] failed\n", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1139 | return false; | |
1140 | } | |
1141 | } | |
1142 | if (part < NUM_SQUARE_BLOCKS) | |
1143 | { | |
1144 | if (opt.chroma[i].sub_ps[part]) | |
1145 | { | |
1146 | if (!check_pixel_sub_ps(ref.chroma[i].sub_ps[part], opt.chroma[i].sub_ps[part])) | |
1147 | { | |
1148 | printf("chroma_sub_ps[%s][%s] failed\n", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1149 | return false; | |
1150 | } | |
1151 | } | |
1152 | if (opt.chroma[i].add_ps[part]) | |
1153 | { | |
1154 | if (!check_pixel_add_ps(ref.chroma[i].add_ps[part], opt.chroma[i].add_ps[part])) | |
1155 | { | |
1156 | printf("chroma_add_ps[%s][%s] failed\n", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1157 | return false; | |
1158 | } | |
1159 | } | |
1160 | } | |
1161 | } | |
1162 | ||
1163 | return true; | |
1164 | } | |
1165 | ||
1166 | bool PixelHarness::testCorrectness(const EncoderPrimitives& ref, const EncoderPrimitives& opt) | |
1167 | { | |
1168 | for (int size = 4; size <= 64; size *= 2) | |
1169 | { | |
1170 | int part = partitionFromSizes(size, size); // 2Nx2N | |
1171 | if (!testPartition(part, ref, opt)) return false; | |
1172 | ||
1173 | if (size > 4) | |
1174 | { | |
1175 | part = partitionFromSizes(size, size >> 1); // 2NxN | |
1176 | if (!testPartition(part, ref, opt)) return false; | |
1177 | part = partitionFromSizes(size >> 1, size); // Nx2N | |
1178 | if (!testPartition(part, ref, opt)) return false; | |
1179 | } | |
1180 | if (size > 8) | |
1181 | { | |
1182 | // 4 AMP modes | |
1183 | part = partitionFromSizes(size, size >> 2); | |
1184 | if (!testPartition(part, ref, opt)) return false; | |
1185 | part = partitionFromSizes(size, 3 * (size >> 2)); | |
1186 | if (!testPartition(part, ref, opt)) return false; | |
1187 | ||
1188 | part = partitionFromSizes(size >> 2, size); | |
1189 | if (!testPartition(part, ref, opt)) return false; | |
1190 | part = partitionFromSizes(3 * (size >> 2), size); | |
1191 | if (!testPartition(part, ref, opt)) return false; | |
1192 | } | |
1193 | } | |
1194 | ||
1195 | for (int i = 0; i < NUM_SQUARE_BLOCKS; i++) | |
1196 | { | |
1197 | if (opt.calcresidual[i]) | |
1198 | { | |
1199 | if (!check_calresidual(ref.calcresidual[i], opt.calcresidual[i])) | |
1200 | { | |
1201 | printf("calcresidual width: %d failed!\n", 4 << i); | |
1202 | return false; | |
1203 | } | |
1204 | } | |
1205 | if (opt.sa8d[i]) | |
1206 | { | |
1207 | if (!check_pixelcmp(ref.sa8d[i], opt.sa8d[i])) | |
1208 | { | |
1209 | printf("sa8d[%dx%d]: failed!\n", 4 << i, 4 << i); | |
1210 | return false; | |
1211 | } | |
1212 | } | |
1213 | ||
1214 | if ((i <= BLOCK_32x32) && opt.ssd_s[i]) | |
1215 | { | |
1216 | if (!check_ssd_s(ref.ssd_s[i], opt.ssd_s[i])) | |
1217 | { | |
1218 | printf("ssd_s[%dx%d]: failed!\n", 4 << i, 4 << i); | |
1219 | return false; | |
1220 | } | |
1221 | } | |
1222 | ||
1223 | if (opt.blockfill_s[i]) | |
1224 | { | |
1225 | if (!check_blockfill_s(ref.blockfill_s[i], opt.blockfill_s[i])) | |
1226 | { | |
1227 | printf("blockfill_s[%dx%d]: failed!\n", 4 << i, 4 << i); | |
1228 | return false; | |
1229 | } | |
1230 | } | |
1231 | if (opt.transpose[i]) | |
1232 | { | |
1233 | if (!check_transpose(ref.transpose[i], opt.transpose[i])) | |
1234 | { | |
1235 | printf("transpose[%dx%d] failed\n", 4 << i, 4 << i); | |
1236 | return false; | |
1237 | } | |
1238 | } | |
1239 | ||
1240 | if (opt.var[i]) | |
1241 | { | |
1242 | if (!check_pixel_var(ref.var[i], opt.var[i])) | |
1243 | { | |
1244 | printf("var[%dx%d] failed\n", 4 << i, 4 << i); | |
1245 | return false; | |
1246 | } | |
1247 | } | |
1248 | ||
1249 | if ((i < BLOCK_64x64) && opt.copy_cnt[i]) | |
1250 | { | |
1251 | if (!check_copy_cnt_t(ref.copy_cnt[i], opt.copy_cnt[i])) | |
1252 | { | |
1253 | printf("copy_cnt[%dx%d] failed!\n", 4 << i, 4 << i); | |
1254 | return false; | |
1255 | } | |
1256 | } | |
1257 | ||
b53f7c52 | 1258 | if ((i < BLOCK_64x64) && opt.cpy2Dto1D_shl[i]) |
72b9787e | 1259 | { |
b53f7c52 | 1260 | if (!check_cpy2Dto1D_shl_t(ref.cpy2Dto1D_shl[i], opt.cpy2Dto1D_shl[i])) |
72b9787e | 1261 | { |
b53f7c52 | 1262 | printf("cpy2Dto1D_shl failed!\n"); |
72b9787e JB |
1263 | return false; |
1264 | } | |
1265 | } | |
1266 | ||
b53f7c52 | 1267 | if ((i < BLOCK_64x64) && opt.cpy2Dto1D_shr[i]) |
72b9787e | 1268 | { |
b53f7c52 | 1269 | if (!check_cpy2Dto1D_shr_t(ref.cpy2Dto1D_shr[i], opt.cpy2Dto1D_shr[i])) |
72b9787e | 1270 | { |
b53f7c52 | 1271 | printf("cpy2Dto1D_shr failed!\n"); |
72b9787e JB |
1272 | return false; |
1273 | } | |
1274 | } | |
1275 | ||
b53f7c52 | 1276 | if ((i < BLOCK_64x64) && opt.cpy1Dto2D_shl[i]) |
72b9787e | 1277 | { |
b53f7c52 | 1278 | if (!check_cpy1Dto2D_shl_t(ref.cpy1Dto2D_shl[i], opt.cpy1Dto2D_shl[i])) |
72b9787e | 1279 | { |
b53f7c52 | 1280 | printf("cpy1Dto2D_shl[%dx%d] failed!\n", 4 << i, 4 << i); |
72b9787e JB |
1281 | return false; |
1282 | } | |
1283 | } | |
1284 | ||
b53f7c52 | 1285 | if ((i < BLOCK_64x64) && opt.cpy1Dto2D_shr[i]) |
72b9787e | 1286 | { |
b53f7c52 JB |
1287 | if (!check_cpy1Dto2D_shr_t(ref.cpy1Dto2D_shr[i], opt.cpy1Dto2D_shr[i])) |
1288 | { | |
1289 | printf("cpy1Dto2D_shr[%dx%d] failed!\n", 4 << i, 4 << i); | |
1290 | return false; | |
1291 | } | |
72b9787e JB |
1292 | } |
1293 | } | |
1294 | ||
1295 | if (opt.weight_pp) | |
1296 | { | |
1297 | if (!check_weightp(ref.weight_pp, opt.weight_pp)) | |
1298 | { | |
1299 | printf("Weighted Prediction (pixel) failed!\n"); | |
1300 | return false; | |
1301 | } | |
1302 | } | |
1303 | ||
1304 | if (opt.weight_sp) | |
1305 | { | |
1306 | if (!check_weightp(ref.weight_sp, opt.weight_sp)) | |
1307 | { | |
1308 | printf("Weighted Prediction (short) failed!\n"); | |
1309 | return false; | |
1310 | } | |
1311 | } | |
1312 | ||
b53f7c52 | 1313 | if (opt.frameInitLowres) |
72b9787e | 1314 | { |
b53f7c52 | 1315 | if (!check_downscale_t(ref.frameInitLowres, opt.frameInitLowres)) |
72b9787e JB |
1316 | { |
1317 | printf("downscale failed!\n"); | |
1318 | return false; | |
1319 | } | |
1320 | } | |
1321 | ||
1322 | if (opt.scale1D_128to64) | |
1323 | { | |
1324 | if (!check_scale_pp(ref.scale1D_128to64, opt.scale1D_128to64)) | |
1325 | { | |
1326 | printf("scale1D_128to64 failed!\n"); | |
1327 | return false; | |
1328 | } | |
1329 | } | |
1330 | ||
1331 | if (opt.scale2D_64to32) | |
1332 | { | |
1333 | if (!check_scale_pp(ref.scale2D_64to32, opt.scale2D_64to32)) | |
1334 | { | |
1335 | printf("scale2D_64to32 failed!\n"); | |
1336 | return false; | |
1337 | } | |
1338 | } | |
1339 | ||
1340 | if (opt.ssim_4x4x2_core) | |
1341 | { | |
1342 | if (!check_ssim_4x4x2_core(ref.ssim_4x4x2_core, opt.ssim_4x4x2_core)) | |
1343 | { | |
1344 | printf("ssim_end_4 failed!\n"); | |
1345 | return false; | |
1346 | } | |
1347 | } | |
1348 | ||
1349 | if (opt.ssim_end_4) | |
1350 | { | |
1351 | if (!check_ssim_end(ref.ssim_end_4, opt.ssim_end_4)) | |
1352 | { | |
1353 | printf("ssim_end_4 failed!\n"); | |
1354 | return false; | |
1355 | } | |
1356 | } | |
1357 | ||
1358 | if (opt.saoCuOrgE0) | |
1359 | { | |
1360 | if (!check_saoCuOrgE0_t(ref.saoCuOrgE0, opt.saoCuOrgE0)) | |
1361 | { | |
1362 | printf("SAO_EO_0 failed\n"); | |
1363 | return false; | |
1364 | } | |
1365 | } | |
1366 | ||
1367 | if (opt.planecopy_sp) | |
1368 | { | |
1369 | if (!check_planecopy_sp(ref.planecopy_sp, opt.planecopy_sp)) | |
1370 | { | |
1371 | printf("planecopy_sp failed\n"); | |
1372 | return false; | |
1373 | } | |
1374 | } | |
1375 | ||
1376 | if (opt.planecopy_cp) | |
1377 | { | |
1378 | if (!check_planecopy_cp(ref.planecopy_cp, opt.planecopy_cp)) | |
1379 | { | |
1380 | printf("planecopy_cp failed\n"); | |
1381 | return false; | |
1382 | } | |
1383 | } | |
1384 | ||
72b9787e JB |
1385 | return true; |
1386 | } | |
1387 | ||
1388 | void PixelHarness::measurePartition(int part, const EncoderPrimitives& ref, const EncoderPrimitives& opt) | |
1389 | { | |
1390 | ALIGN_VAR_16(int, cres[16]); | |
1391 | pixel *fref = pbuf2 + 2 * INCR; | |
1392 | char header[128]; | |
1393 | #define HEADER(str, ...) sprintf(header, str, __VA_ARGS__); printf("%22s", header); | |
1394 | ||
1395 | if (opt.satd[part]) | |
1396 | { | |
1397 | HEADER("satd[%s]", lumaPartStr[part]); | |
1398 | REPORT_SPEEDUP(opt.satd[part], ref.satd[part], pbuf1, STRIDE, fref, STRIDE); | |
1399 | } | |
1400 | ||
1401 | if (opt.pixelavg_pp[part]) | |
1402 | { | |
1403 | HEADER("avg_pp[%s]", lumaPartStr[part]); | |
1404 | REPORT_SPEEDUP(opt.pixelavg_pp[part], ref.pixelavg_pp[part], pbuf1, STRIDE, pbuf2, STRIDE, pbuf3, STRIDE, 32); | |
1405 | } | |
1406 | ||
1407 | if (opt.sa8d_inter[part]) | |
1408 | { | |
1409 | HEADER("sa8d[%s]", lumaPartStr[part]); | |
1410 | REPORT_SPEEDUP(opt.sa8d_inter[part], ref.sa8d_inter[part], pbuf1, STRIDE, fref, STRIDE); | |
1411 | } | |
1412 | ||
1413 | if (opt.sad[part]) | |
1414 | { | |
1415 | HEADER("sad[%s]", lumaPartStr[part]); | |
1416 | REPORT_SPEEDUP(opt.sad[part], ref.sad[part], pbuf1, STRIDE, fref, STRIDE); | |
1417 | } | |
1418 | ||
1419 | if (opt.sad_x3[part]) | |
1420 | { | |
1421 | HEADER("sad_x3[%s]", lumaPartStr[part]); | |
1422 | REPORT_SPEEDUP(opt.sad_x3[part], ref.sad_x3[part], pbuf1, fref, fref + 1, fref - 1, FENC_STRIDE + 5, &cres[0]); | |
1423 | } | |
1424 | ||
1425 | if (opt.sad_x4[part]) | |
1426 | { | |
1427 | HEADER("sad_x4[%s]", lumaPartStr[part]); | |
1428 | REPORT_SPEEDUP(opt.sad_x4[part], ref.sad_x4[part], pbuf1, fref, fref + 1, fref - 1, fref - INCR, FENC_STRIDE + 5, &cres[0]); | |
1429 | } | |
1430 | ||
1431 | if (opt.sse_pp[part]) | |
1432 | { | |
1433 | HEADER("sse_pp[%s]", lumaPartStr[part]); | |
1434 | REPORT_SPEEDUP(opt.sse_pp[part], ref.sse_pp[part], pbuf1, STRIDE, fref, STRIDE); | |
1435 | } | |
1436 | ||
1437 | if (opt.sse_sp[part]) | |
1438 | { | |
1439 | HEADER("sse_sp[%s]", lumaPartStr[part]); | |
1440 | REPORT_SPEEDUP(opt.sse_sp[part], ref.sse_sp[part], (int16_t*)pbuf1, STRIDE, fref, STRIDE); | |
1441 | } | |
1442 | ||
1443 | if (opt.sse_ss[part]) | |
1444 | { | |
1445 | HEADER("sse_ss[%s]", lumaPartStr[part]); | |
1446 | REPORT_SPEEDUP(opt.sse_ss[part], ref.sse_ss[part], (int16_t*)pbuf1, STRIDE, (int16_t*)fref, STRIDE); | |
1447 | } | |
1448 | ||
1449 | if (opt.luma_copy_pp[part]) | |
1450 | { | |
1451 | HEADER("luma_copy_pp[%s]", lumaPartStr[part]); | |
1452 | REPORT_SPEEDUP(opt.luma_copy_pp[part], ref.luma_copy_pp[part], pbuf1, 64, pbuf2, 128); | |
1453 | } | |
1454 | ||
1455 | if (opt.luma_copy_sp[part]) | |
1456 | { | |
1457 | HEADER("luma_copy_sp[%s]", lumaPartStr[part]); | |
1458 | REPORT_SPEEDUP(opt.luma_copy_sp[part], ref.luma_copy_sp[part], pbuf1, 64, sbuf3, 128); | |
1459 | } | |
1460 | ||
1461 | if (opt.luma_copy_ps[part]) | |
1462 | { | |
1463 | HEADER("luma_copy_ps[%s]", lumaPartStr[part]); | |
1464 | REPORT_SPEEDUP(opt.luma_copy_ps[part], ref.luma_copy_ps[part], sbuf1, 64, pbuf1, 128); | |
1465 | } | |
1466 | if (opt.luma_copy_ss[part]) | |
1467 | { | |
1468 | HEADER("luma_copy_ss[%s]", lumaPartStr[part]); | |
1469 | REPORT_SPEEDUP(opt.luma_copy_ss[part], ref.luma_copy_ss[part], sbuf1, 64, sbuf2, 128); | |
1470 | } | |
1471 | if (opt.luma_addAvg[part]) | |
1472 | { | |
1473 | HEADER("luma_addAvg[%s]", lumaPartStr[part]); | |
1474 | REPORT_SPEEDUP(opt.luma_addAvg[part], ref.luma_addAvg[part], sbuf1, sbuf2, pbuf1, STRIDE, STRIDE, STRIDE); | |
1475 | } | |
1476 | if (part < NUM_SQUARE_BLOCKS) | |
1477 | { | |
1478 | if (opt.luma_sub_ps[part]) | |
1479 | { | |
1480 | HEADER("luma_sub_ps[%s]", lumaPartStr[part]); | |
1481 | REPORT_SPEEDUP(opt.luma_sub_ps[part], ref.luma_sub_ps[part], (int16_t*)pbuf1, FENC_STRIDE, pbuf2, pbuf1, STRIDE, STRIDE); | |
1482 | } | |
1483 | if (opt.luma_add_ps[part]) | |
1484 | { | |
1485 | HEADER("luma_add_ps[%s]", lumaPartStr[part]); | |
1486 | REPORT_SPEEDUP(opt.luma_add_ps[part], ref.luma_add_ps[part], pbuf1, FENC_STRIDE, pbuf2, sbuf1, STRIDE, STRIDE); | |
1487 | } | |
1488 | } | |
1489 | ||
1490 | for (int i = 0; i < X265_CSP_COUNT; i++) | |
1491 | { | |
1492 | if (opt.chroma[i].copy_pp[part]) | |
1493 | { | |
1494 | HEADER("[%s] copy_pp[%s]", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1495 | REPORT_SPEEDUP(opt.chroma[i].copy_pp[part], ref.chroma[i].copy_pp[part], pbuf1, 64, pbuf2, 128); | |
1496 | } | |
1497 | if (opt.chroma[i].copy_sp[part]) | |
1498 | { | |
1499 | HEADER("[%s] copy_sp[%s]", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1500 | REPORT_SPEEDUP(opt.chroma[i].copy_sp[part], ref.chroma[i].copy_sp[part], pbuf1, 64, sbuf3, 128); | |
1501 | } | |
1502 | if (opt.chroma[i].copy_ps[part]) | |
1503 | { | |
1504 | HEADER("[%s] copy_ps[%s]", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1505 | REPORT_SPEEDUP(opt.chroma[i].copy_ps[part], ref.chroma[i].copy_ps[part], sbuf1, 64, pbuf1, 128); | |
1506 | } | |
1507 | if (opt.chroma[i].copy_ss[part]) | |
1508 | { | |
1509 | HEADER("[%s] copy_ss[%s]", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1510 | REPORT_SPEEDUP(opt.chroma[i].copy_ss[part], ref.chroma[i].copy_ss[part], sbuf1, 64, sbuf2, 128); | |
1511 | } | |
1512 | if (opt.chroma[i].addAvg[part]) | |
1513 | { | |
1514 | HEADER("[%s] addAvg[%s]", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1515 | REPORT_SPEEDUP(opt.chroma[i].addAvg[part], ref.chroma[i].addAvg[part], sbuf1, sbuf2, pbuf1, STRIDE, STRIDE, STRIDE); | |
1516 | } | |
1517 | if (part < NUM_SQUARE_BLOCKS) | |
1518 | { | |
1519 | if (opt.chroma[i].sub_ps[part]) | |
1520 | { | |
1521 | HEADER("[%s] sub_ps[%s]", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1522 | REPORT_SPEEDUP(opt.chroma[i].sub_ps[part], ref.chroma[i].sub_ps[part], (int16_t*)pbuf1, FENC_STRIDE, pbuf2, pbuf1, STRIDE, STRIDE); | |
1523 | } | |
1524 | if (opt.chroma[i].add_ps[part]) | |
1525 | { | |
1526 | HEADER("[%s] add_ps[%s]", x265_source_csp_names[i], chromaPartStr[i][part]); | |
1527 | REPORT_SPEEDUP(opt.chroma[i].add_ps[part], ref.chroma[i].add_ps[part], pbuf1, FENC_STRIDE, pbuf2, sbuf1, STRIDE, STRIDE); | |
1528 | } | |
1529 | } | |
1530 | } | |
1531 | ||
1532 | #undef HEADER | |
1533 | } | |
1534 | ||
1535 | void PixelHarness::measureSpeed(const EncoderPrimitives& ref, const EncoderPrimitives& opt) | |
1536 | { | |
1537 | char header[128]; | |
1538 | ||
1539 | #define HEADER(str, ...) sprintf(header, str, __VA_ARGS__); printf("%22s", header); | |
1540 | #define HEADER0(str) printf("%22s", str); | |
1541 | ||
1542 | for (int size = 4; size <= 64; size *= 2) | |
1543 | { | |
1544 | int part = partitionFromSizes(size, size); // 2Nx2N | |
1545 | measurePartition(part, ref, opt); | |
1546 | ||
1547 | if (size > 4) | |
1548 | { | |
1549 | part = partitionFromSizes(size, size >> 1); // 2NxN | |
1550 | measurePartition(part, ref, opt); | |
1551 | part = partitionFromSizes(size >> 1, size); // Nx2N | |
1552 | measurePartition(part, ref, opt); | |
1553 | } | |
1554 | if (size > 8) | |
1555 | { | |
1556 | // 4 AMP modes | |
1557 | part = partitionFromSizes(size, size >> 2); | |
1558 | measurePartition(part, ref, opt); | |
1559 | part = partitionFromSizes(size, 3 * (size >> 2)); | |
1560 | measurePartition(part, ref, opt); | |
1561 | ||
1562 | part = partitionFromSizes(size >> 2, size); | |
1563 | measurePartition(part, ref, opt); | |
1564 | part = partitionFromSizes(3 * (size >> 2), size); | |
1565 | measurePartition(part, ref, opt); | |
1566 | } | |
1567 | } | |
1568 | ||
1569 | for (int i = 0; i < NUM_SQUARE_BLOCKS; i++) | |
1570 | { | |
1571 | if ((i <= BLOCK_32x32) && opt.ssd_s[i]) | |
1572 | { | |
1573 | HEADER("ssd_s[%dx%d]", 4 << i, 4 << i); | |
1574 | REPORT_SPEEDUP(opt.ssd_s[i], ref.ssd_s[i], sbuf1, STRIDE); | |
1575 | } | |
1576 | if (opt.sa8d[i]) | |
1577 | { | |
1578 | HEADER("sa8d[%dx%d]", 4 << i, 4 << i); | |
1579 | REPORT_SPEEDUP(opt.sa8d[i], ref.sa8d[i], pbuf1, STRIDE, pbuf2, STRIDE); | |
1580 | } | |
1581 | if (opt.calcresidual[i]) | |
1582 | { | |
1583 | HEADER("residual[%dx%d]", 4 << i, 4 << i); | |
1584 | REPORT_SPEEDUP(opt.calcresidual[i], ref.calcresidual[i], pbuf1, pbuf2, sbuf1, 64); | |
1585 | } | |
1586 | ||
1587 | if (opt.blockfill_s[i]) | |
1588 | { | |
1589 | HEADER("blkfill[%dx%d]", 4 << i, 4 << i); | |
1590 | REPORT_SPEEDUP(opt.blockfill_s[i], ref.blockfill_s[i], sbuf1, 64, SHORT_MAX); | |
1591 | } | |
1592 | ||
1593 | if (opt.transpose[i]) | |
1594 | { | |
1595 | HEADER("transpose[%dx%d]", 4 << i, 4 << i); | |
1596 | REPORT_SPEEDUP(opt.transpose[i], ref.transpose[i], pbuf1, pbuf2, STRIDE); | |
1597 | } | |
1598 | ||
1599 | if (opt.var[i]) | |
1600 | { | |
1601 | HEADER("var[%dx%d]", 4 << i, 4 << i); | |
1602 | REPORT_SPEEDUP(opt.var[i], ref.var[i], pbuf1, STRIDE); | |
1603 | } | |
1604 | ||
b53f7c52 | 1605 | if ((i < BLOCK_64x64) && opt.cpy2Dto1D_shl[i]) |
72b9787e | 1606 | { |
b53f7c52 JB |
1607 | HEADER("cpy2Dto1D_shl[%dx%d]", 4 << i, 4 << i); |
1608 | REPORT_SPEEDUP(opt.cpy2Dto1D_shl[i], ref.cpy2Dto1D_shl[i], sbuf1, sbuf2, STRIDE, MAX_TR_DYNAMIC_RANGE - X265_DEPTH - (i + 2)); | |
72b9787e JB |
1609 | } |
1610 | ||
b53f7c52 | 1611 | if ((i < BLOCK_64x64) && opt.cpy2Dto1D_shr[i]) |
72b9787e | 1612 | { |
b53f7c52 JB |
1613 | HEADER("cpy2Dto1D_shr[%dx%d]", 4 << i, 4 << i); |
1614 | REPORT_SPEEDUP(opt.cpy2Dto1D_shr[i], ref.cpy2Dto1D_shr[i], sbuf1, sbuf2, STRIDE, 3); | |
72b9787e JB |
1615 | } |
1616 | ||
b53f7c52 | 1617 | if ((i < BLOCK_64x64) && opt.cpy1Dto2D_shl[i]) |
72b9787e | 1618 | { |
b53f7c52 JB |
1619 | HEADER("cpy1Dto2D_shl[%dx%d]", 4 << i, 4 << i); |
1620 | REPORT_SPEEDUP(opt.cpy1Dto2D_shl[i], ref.cpy1Dto2D_shl[i], sbuf1, sbuf2, STRIDE, 64); | |
72b9787e JB |
1621 | } |
1622 | ||
b53f7c52 | 1623 | if ((i < BLOCK_64x64) && opt.cpy1Dto2D_shr[i]) |
72b9787e | 1624 | { |
b53f7c52 JB |
1625 | HEADER("cpy1Dto2D_shr[%dx%d]", 4 << i, 4 << i); |
1626 | REPORT_SPEEDUP(opt.cpy1Dto2D_shr[i], ref.cpy1Dto2D_shr[i], sbuf1, sbuf2, STRIDE, 64); | |
72b9787e JB |
1627 | } |
1628 | ||
b53f7c52 JB |
1629 | if ((i < BLOCK_64x64) && opt.copy_cnt[i]) |
1630 | { | |
1631 | HEADER("copy_cnt[%dx%d]", 4 << i, 4 << i); | |
1632 | REPORT_SPEEDUP(opt.copy_cnt[i], ref.copy_cnt[i], sbuf1, sbuf2, STRIDE); | |
1633 | } | |
72b9787e JB |
1634 | } |
1635 | ||
1636 | if (opt.weight_pp) | |
1637 | { | |
1638 | HEADER0("weight_pp"); | |
1639 | REPORT_SPEEDUP(opt.weight_pp, ref.weight_pp, pbuf1, pbuf2, 64, 32, 32, 128, 1 << 9, 10, 100); | |
1640 | } | |
1641 | ||
1642 | if (opt.weight_sp) | |
1643 | { | |
1644 | HEADER0("weight_sp"); | |
1645 | REPORT_SPEEDUP(opt.weight_sp, ref.weight_sp, (int16_t*)sbuf1, pbuf1, 64, 64, 32, 32, 128, 1 << 9, 10, 100); | |
1646 | } | |
1647 | ||
b53f7c52 | 1648 | if (opt.frameInitLowres) |
72b9787e JB |
1649 | { |
1650 | HEADER0("downscale"); | |
b53f7c52 | 1651 | REPORT_SPEEDUP(opt.frameInitLowres, ref.frameInitLowres, pbuf2, pbuf1, pbuf2, pbuf3, pbuf4, 64, 64, 64, 64); |
72b9787e JB |
1652 | } |
1653 | ||
1654 | if (opt.scale1D_128to64) | |
1655 | { | |
1656 | HEADER0("scale1D_128to64"); | |
1657 | REPORT_SPEEDUP(opt.scale1D_128to64, ref.scale1D_128to64, pbuf2, pbuf1, 64); | |
1658 | } | |
1659 | ||
1660 | if (opt.scale2D_64to32) | |
1661 | { | |
1662 | HEADER0("scale2D_64to32"); | |
1663 | REPORT_SPEEDUP(opt.scale2D_64to32, ref.scale2D_64to32, pbuf2, pbuf1, 64); | |
1664 | } | |
1665 | ||
1666 | if (opt.ssim_4x4x2_core) | |
1667 | { | |
1668 | HEADER0("ssim_4x4x2_core"); | |
1669 | REPORT_SPEEDUP(opt.ssim_4x4x2_core, ref.ssim_4x4x2_core, pbuf1, 64, pbuf2, 64, (int(*)[4])sbuf1); | |
1670 | } | |
1671 | ||
1672 | if (opt.ssim_end_4) | |
1673 | { | |
1674 | HEADER0("ssim_end_4"); | |
1675 | REPORT_SPEEDUP(opt.ssim_end_4, ref.ssim_end_4, (int(*)[4])pbuf2, (int(*)[4])pbuf1, 4); | |
1676 | } | |
1677 | ||
1678 | if (opt.saoCuOrgE0) | |
1679 | { | |
1680 | HEADER0("SAO_EO_0"); | |
1681 | REPORT_SPEEDUP(opt.saoCuOrgE0, ref.saoCuOrgE0, pbuf1, psbuf1, 64, 1); | |
1682 | } | |
1683 | ||
1684 | if (opt.planecopy_sp) | |
1685 | { | |
1686 | HEADER0("planecopy_sp"); | |
1687 | REPORT_SPEEDUP(opt.planecopy_sp, ref.planecopy_sp, ushort_test_buff[0], 64, pbuf1, 64, 64, 64, 8, 255); | |
1688 | } | |
1689 | ||
1690 | if (opt.planecopy_cp) | |
1691 | { | |
1692 | HEADER0("planecopy_cp"); | |
1693 | REPORT_SPEEDUP(opt.planecopy_cp, ref.planecopy_cp, uchar_test_buff[0], 64, pbuf1, 64, 64, 64, 2); | |
1694 | } | |
72b9787e | 1695 | } |