1 | // SPDX-License-Identifier: Apache-2.0 |
2 | // ---------------------------------------------------------------------------- |
3 | // Copyright 2011-2023 Arm Limited |
4 | // |
5 | // Licensed under the Apache License, Version 2.0 (the "License"); you may not |
6 | // use this file except in compliance with the License. You may obtain a copy |
7 | // of the License at: |
8 | // |
9 | // http://www.apache.org/licenses/LICENSE-2.0 |
10 | // |
11 | // Unless required by applicable law or agreed to in writing, software |
12 | // distributed under the License is distributed on an "AS IS" BASIS, WITHOUT |
13 | // WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the |
14 | // License for the specific language governing permissions and limitations |
15 | // under the License. |
16 | // ---------------------------------------------------------------------------- |
17 | |
18 | /** |
19 | * @brief Functions for generating partition tables on demand. |
20 | */ |
21 | |
22 | #include "astcenc_internal.h" |
23 | |
24 | /** @brief The number of 64-bit words needed to represent a canonical partition bit pattern. */ |
25 | #define BIT_PATTERN_WORDS (((ASTCENC_BLOCK_MAX_TEXELS * 2) + 63) / 64) |
26 | |
27 | /** |
28 | * @brief Generate a canonical representation of a partition pattern. |
29 | * |
30 | * The returned value stores two bits per texel, for up to 6x6x6 texels, where the two bits store |
31 | * the remapped texel index. Remapping ensures that we only match on the partition pattern, |
32 | * independent of the partition order generated by the hash. |
33 | * |
34 | * @param texel_count The number of texels in the block. |
35 | * @param partition_of_texel The partition assignments, in hash order. |
36 | * @param[out] bit_pattern The output bit pattern representation. |
37 | */ |
38 | static void generate_canonical_partitioning( |
39 | unsigned int texel_count, |
40 | const uint8_t* partition_of_texel, |
41 | uint64_t bit_pattern[BIT_PATTERN_WORDS] |
42 | ) { |
43 | // Clear the pattern |
44 | for (unsigned int i = 0; i < BIT_PATTERN_WORDS; i++) |
45 | { |
46 | bit_pattern[i] = 0; |
47 | } |
48 | |
49 | // Store a mapping to reorder the raw partitions so that the partitions are ordered such |
50 | // that the lowest texel index in partition N is smaller than the lowest texel index in |
51 | // partition N + 1. |
52 | int mapped_index[BLOCK_MAX_PARTITIONS]; |
53 | int map_weight_count = 0; |
54 | |
55 | for (unsigned int i = 0; i < BLOCK_MAX_PARTITIONS; i++) |
56 | { |
57 | mapped_index[i] = -1; |
58 | } |
59 | |
60 | for (unsigned int i = 0; i < texel_count; i++) |
61 | { |
62 | int index = partition_of_texel[i]; |
63 | if (mapped_index[index] < 0) |
64 | { |
65 | mapped_index[index] = map_weight_count++; |
66 | } |
67 | |
68 | uint64_t xlat_index = mapped_index[index]; |
69 | bit_pattern[i >> 5] |= xlat_index << (2 * (i & 0x1F)); |
70 | } |
71 | } |
72 | |
73 | /** |
74 | * @brief Compare two canonical patterns to see if they are the same. |
75 | * |
76 | * @param part1 The first canonical bit pattern to check. |
77 | * @param part2 The second canonical bit pattern to check. |
78 | * |
79 | * @return @c true if the patterns are the same, @c false otherwise. |
80 | */ |
81 | static bool compare_canonical_partitionings( |
82 | const uint64_t part1[BIT_PATTERN_WORDS], |
83 | const uint64_t part2[BIT_PATTERN_WORDS] |
84 | ) { |
85 | return (part1[0] == part2[0]) |
86 | #if BIT_PATTERN_WORDS > 1 |
87 | && (part1[1] == part2[1]) |
88 | #endif |
89 | #if BIT_PATTERN_WORDS > 2 |
90 | && (part1[2] == part2[2]) |
91 | #endif |
92 | #if BIT_PATTERN_WORDS > 3 |
93 | && (part1[3] == part2[3]) |
94 | #endif |
95 | #if BIT_PATTERN_WORDS > 4 |
96 | && (part1[4] == part2[4]) |
97 | #endif |
98 | #if BIT_PATTERN_WORDS > 5 |
99 | && (part1[5] == part2[5]) |
100 | #endif |
101 | #if BIT_PATTERN_WORDS > 6 |
102 | && (part1[6] == part2[6]) |
103 | #endif |
104 | ; |
105 | } |
106 | |
107 | /** |
108 | * @brief Hash function used for procedural partition assignment. |
109 | * |
110 | * @param inp The hash seed. |
111 | * |
112 | * @return The hashed value. |
113 | */ |
114 | static uint32_t hash52( |
115 | uint32_t inp |
116 | ) { |
117 | inp ^= inp >> 15; |
118 | |
119 | // (2^4 + 1) * (2^7 + 1) * (2^17 - 1) |
120 | inp *= 0xEEDE0891; |
121 | inp ^= inp >> 5; |
122 | inp += inp << 16; |
123 | inp ^= inp >> 7; |
124 | inp ^= inp >> 3; |
125 | inp ^= inp << 6; |
126 | inp ^= inp >> 17; |
127 | return inp; |
128 | } |
129 | |
130 | /** |
131 | * @brief Select texel assignment for a single coordinate. |
132 | * |
133 | * @param seed The seed - the partition index from the block. |
134 | * @param x The texel X coordinate in the block. |
135 | * @param y The texel Y coordinate in the block. |
136 | * @param z The texel Z coordinate in the block. |
137 | * @param partition_count The total partition count of this encoding. |
138 | * @param small_block @c true if the block has fewer than 32 texels. |
139 | * |
140 | * @return The assigned partition index for this texel. |
141 | */ |
142 | static uint8_t select_partition( |
143 | int seed, |
144 | int x, |
145 | int y, |
146 | int z, |
147 | int partition_count, |
148 | bool small_block |
149 | ) { |
150 | // For small blocks bias the coordinates to get better distribution |
151 | if (small_block) |
152 | { |
153 | x <<= 1; |
154 | y <<= 1; |
155 | z <<= 1; |
156 | } |
157 | |
158 | seed += (partition_count - 1) * 1024; |
159 | |
160 | uint32_t rnum = hash52(seed); |
161 | |
162 | uint8_t seed1 = rnum & 0xF; |
163 | uint8_t seed2 = (rnum >> 4) & 0xF; |
164 | uint8_t seed3 = (rnum >> 8) & 0xF; |
165 | uint8_t seed4 = (rnum >> 12) & 0xF; |
166 | uint8_t seed5 = (rnum >> 16) & 0xF; |
167 | uint8_t seed6 = (rnum >> 20) & 0xF; |
168 | uint8_t seed7 = (rnum >> 24) & 0xF; |
169 | uint8_t seed8 = (rnum >> 28) & 0xF; |
170 | uint8_t seed9 = (rnum >> 18) & 0xF; |
171 | uint8_t seed10 = (rnum >> 22) & 0xF; |
172 | uint8_t seed11 = (rnum >> 26) & 0xF; |
173 | uint8_t seed12 = ((rnum >> 30) | (rnum << 2)) & 0xF; |
174 | |
175 | // Squaring all the seeds in order to bias their distribution towards lower values. |
176 | seed1 *= seed1; |
177 | seed2 *= seed2; |
178 | seed3 *= seed3; |
179 | seed4 *= seed4; |
180 | seed5 *= seed5; |
181 | seed6 *= seed6; |
182 | seed7 *= seed7; |
183 | seed8 *= seed8; |
184 | seed9 *= seed9; |
185 | seed10 *= seed10; |
186 | seed11 *= seed11; |
187 | seed12 *= seed12; |
188 | |
189 | int sh1, sh2; |
190 | if (seed & 1) |
191 | { |
192 | sh1 = (seed & 2 ? 4 : 5); |
193 | sh2 = (partition_count == 3 ? 6 : 5); |
194 | } |
195 | else |
196 | { |
197 | sh1 = (partition_count == 3 ? 6 : 5); |
198 | sh2 = (seed & 2 ? 4 : 5); |
199 | } |
200 | |
201 | int sh3 = (seed & 0x10) ? sh1 : sh2; |
202 | |
203 | seed1 >>= sh1; |
204 | seed2 >>= sh2; |
205 | seed3 >>= sh1; |
206 | seed4 >>= sh2; |
207 | seed5 >>= sh1; |
208 | seed6 >>= sh2; |
209 | seed7 >>= sh1; |
210 | seed8 >>= sh2; |
211 | |
212 | seed9 >>= sh3; |
213 | seed10 >>= sh3; |
214 | seed11 >>= sh3; |
215 | seed12 >>= sh3; |
216 | |
217 | int a = seed1 * x + seed2 * y + seed11 * z + (rnum >> 14); |
218 | int b = seed3 * x + seed4 * y + seed12 * z + (rnum >> 10); |
219 | int c = seed5 * x + seed6 * y + seed9 * z + (rnum >> 6); |
220 | int d = seed7 * x + seed8 * y + seed10 * z + (rnum >> 2); |
221 | |
222 | // Apply the saw |
223 | a &= 0x3F; |
224 | b &= 0x3F; |
225 | c &= 0x3F; |
226 | d &= 0x3F; |
227 | |
228 | // Remove some of the components if we are to output < 4 partitions. |
229 | if (partition_count <= 3) |
230 | { |
231 | d = 0; |
232 | } |
233 | |
234 | if (partition_count <= 2) |
235 | { |
236 | c = 0; |
237 | } |
238 | |
239 | if (partition_count <= 1) |
240 | { |
241 | b = 0; |
242 | } |
243 | |
244 | uint8_t partition; |
245 | if (a >= b && a >= c && a >= d) |
246 | { |
247 | partition = 0; |
248 | } |
249 | else if (b >= c && b >= d) |
250 | { |
251 | partition = 1; |
252 | } |
253 | else if (c >= d) |
254 | { |
255 | partition = 2; |
256 | } |
257 | else |
258 | { |
259 | partition = 3; |
260 | } |
261 | |
262 | return partition; |
263 | } |
264 | |
265 | /** |
266 | * @brief Generate a single partition info structure. |
267 | * |
268 | * @param[out] bsd The block size information. |
269 | * @param partition_count The partition count of this partitioning. |
270 | * @param partition_index The partition index / seed of this partitioning. |
271 | * @param partition_remap_index The remapped partition index of this partitioning. |
272 | * @param[out] pi The partition info structure to populate. |
273 | * |
274 | * @return True if this is a useful partition index, False if we can skip it. |
275 | */ |
276 | static bool generate_one_partition_info_entry( |
277 | block_size_descriptor& bsd, |
278 | unsigned int partition_count, |
279 | unsigned int partition_index, |
280 | unsigned int partition_remap_index, |
281 | partition_info& pi |
282 | ) { |
283 | int texels_per_block = bsd.texel_count; |
284 | bool small_block = texels_per_block < 32; |
285 | |
286 | uint8_t *partition_of_texel = pi.partition_of_texel; |
287 | |
288 | // Assign texels to partitions |
289 | int texel_idx = 0; |
290 | int counts[BLOCK_MAX_PARTITIONS] { 0 }; |
291 | for (unsigned int z = 0; z < bsd.zdim; z++) |
292 | { |
293 | for (unsigned int y = 0; y < bsd.ydim; y++) |
294 | { |
295 | for (unsigned int x = 0; x < bsd.xdim; x++) |
296 | { |
297 | uint8_t part = select_partition(partition_index, x, y, z, partition_count, small_block); |
298 | pi.texels_of_partition[part][counts[part]++] = static_cast<uint8_t>(texel_idx++); |
299 | *partition_of_texel++ = part; |
300 | } |
301 | } |
302 | } |
303 | |
304 | // Fill loop tail so we can overfetch later |
305 | for (unsigned int i = 0; i < partition_count; i++) |
306 | { |
307 | int ptex_count = counts[i]; |
308 | int ptex_count_simd = round_up_to_simd_multiple_vla(ptex_count); |
309 | for (int j = ptex_count; j < ptex_count_simd; j++) |
310 | { |
311 | pi.texels_of_partition[i][j] = pi.texels_of_partition[i][ptex_count - 1]; |
312 | } |
313 | } |
314 | |
315 | // Populate the actual procedural partition count |
316 | if (counts[0] == 0) |
317 | { |
318 | pi.partition_count = 0; |
319 | } |
320 | else if (counts[1] == 0) |
321 | { |
322 | pi.partition_count = 1; |
323 | } |
324 | else if (counts[2] == 0) |
325 | { |
326 | pi.partition_count = 2; |
327 | } |
328 | else if (counts[3] == 0) |
329 | { |
330 | pi.partition_count = 3; |
331 | } |
332 | else |
333 | { |
334 | pi.partition_count = 4; |
335 | } |
336 | |
337 | // Populate the partition index |
338 | pi.partition_index = static_cast<uint16_t>(partition_index); |
339 | |
340 | // Populate the coverage bitmaps for 2/3/4 partitions |
341 | uint64_t* bitmaps { nullptr }; |
342 | if (partition_count == 2) |
343 | { |
344 | bitmaps = bsd.coverage_bitmaps_2[partition_remap_index]; |
345 | } |
346 | else if (partition_count == 3) |
347 | { |
348 | bitmaps = bsd.coverage_bitmaps_3[partition_remap_index]; |
349 | } |
350 | else if (partition_count == 4) |
351 | { |
352 | bitmaps = bsd.coverage_bitmaps_4[partition_remap_index]; |
353 | } |
354 | |
355 | for (unsigned int i = 0; i < BLOCK_MAX_PARTITIONS; i++) |
356 | { |
357 | pi.partition_texel_count[i] = static_cast<uint8_t>(counts[i]); |
358 | } |
359 | |
360 | // Valid partitionings have texels in all of the requested partitions |
361 | bool valid = pi.partition_count == partition_count; |
362 | |
363 | if (bitmaps) |
364 | { |
365 | // Populate the partition coverage bitmap |
366 | for (unsigned int i = 0; i < partition_count; i++) |
367 | { |
368 | bitmaps[i] = 0ULL; |
369 | } |
370 | |
371 | unsigned int texels_to_process = astc::min(bsd.texel_count, BLOCK_MAX_KMEANS_TEXELS); |
372 | for (unsigned int i = 0; i < texels_to_process; i++) |
373 | { |
374 | unsigned int idx = bsd.kmeans_texels[i]; |
375 | bitmaps[pi.partition_of_texel[idx]] |= 1ULL << i; |
376 | } |
377 | } |
378 | |
379 | return valid; |
380 | } |
381 | |
382 | static void build_partition_table_for_one_partition_count( |
383 | block_size_descriptor& bsd, |
384 | bool can_omit_partitionings, |
385 | unsigned int partition_count_cutoff, |
386 | unsigned int partition_count, |
387 | partition_info* ptab, |
388 | uint64_t* canonical_patterns |
389 | ) { |
390 | unsigned int next_index = 0; |
391 | bsd.partitioning_count_selected[partition_count - 1] = 0; |
392 | bsd.partitioning_count_all[partition_count - 1] = 0; |
393 | |
394 | // Skip tables larger than config max partition count if we can omit modes |
395 | if (can_omit_partitionings && (partition_count > partition_count_cutoff)) |
396 | { |
397 | return; |
398 | } |
399 | |
400 | // Iterate through twice |
401 | // - Pass 0: Keep selected partitionings |
402 | // - Pass 1: Keep non-selected partitionings (skip if in omit mode) |
403 | unsigned int max_iter = can_omit_partitionings ? 1 : 2; |
404 | |
405 | // Tracker for things we built in the first iteration |
406 | uint8_t build[BLOCK_MAX_PARTITIONINGS] { 0 }; |
407 | for (unsigned int x = 0; x < max_iter; x++) |
408 | { |
409 | for (unsigned int i = 0; i < BLOCK_MAX_PARTITIONINGS; i++) |
410 | { |
411 | // Don't include things we built in the first pass |
412 | if ((x == 1) && build[i]) |
413 | { |
414 | continue; |
415 | } |
416 | |
417 | bool keep_useful = generate_one_partition_info_entry(bsd, partition_count, i, next_index, ptab[next_index]); |
418 | if ((x == 0) && !keep_useful) |
419 | { |
420 | continue; |
421 | } |
422 | |
423 | generate_canonical_partitioning(bsd.texel_count, ptab[next_index].partition_of_texel, canonical_patterns + next_index * BIT_PATTERN_WORDS); |
424 | bool keep_canonical = true; |
425 | for (unsigned int j = 0; j < next_index; j++) |
426 | { |
427 | bool match = compare_canonical_partitionings(canonical_patterns + next_index * BIT_PATTERN_WORDS, canonical_patterns + j * BIT_PATTERN_WORDS); |
428 | if (match) |
429 | { |
430 | keep_canonical = false; |
431 | break; |
432 | } |
433 | } |
434 | |
435 | if (keep_useful && keep_canonical) |
436 | { |
437 | if (x == 0) |
438 | { |
439 | bsd.partitioning_packed_index[partition_count - 2][i] = static_cast<uint16_t>(next_index); |
440 | bsd.partitioning_count_selected[partition_count - 1]++; |
441 | bsd.partitioning_count_all[partition_count - 1]++; |
442 | build[i] = 1; |
443 | next_index++; |
444 | } |
445 | } |
446 | else |
447 | { |
448 | if (x == 1) |
449 | { |
450 | bsd.partitioning_packed_index[partition_count - 2][i] = static_cast<uint16_t>(next_index); |
451 | bsd.partitioning_count_all[partition_count - 1]++; |
452 | next_index++; |
453 | } |
454 | } |
455 | } |
456 | } |
457 | } |
458 | |
459 | /* See header for documentation. */ |
460 | void init_partition_tables( |
461 | block_size_descriptor& bsd, |
462 | bool can_omit_partitionings, |
463 | unsigned int partition_count_cutoff |
464 | ) { |
465 | partition_info* par_tab2 = bsd.partitionings; |
466 | partition_info* par_tab3 = par_tab2 + BLOCK_MAX_PARTITIONINGS; |
467 | partition_info* par_tab4 = par_tab3 + BLOCK_MAX_PARTITIONINGS; |
468 | partition_info* par_tab1 = par_tab4 + BLOCK_MAX_PARTITIONINGS; |
469 | |
470 | generate_one_partition_info_entry(bsd, 1, 0, 0, *par_tab1); |
471 | bsd.partitioning_count_selected[0] = 1; |
472 | bsd.partitioning_count_all[0] = 1; |
473 | |
474 | uint64_t* canonical_patterns = new uint64_t[BLOCK_MAX_PARTITIONINGS * BIT_PATTERN_WORDS]; |
475 | |
476 | build_partition_table_for_one_partition_count(bsd, can_omit_partitionings, partition_count_cutoff, 2, par_tab2, canonical_patterns); |
477 | build_partition_table_for_one_partition_count(bsd, can_omit_partitionings, partition_count_cutoff, 3, par_tab3, canonical_patterns); |
478 | build_partition_table_for_one_partition_count(bsd, can_omit_partitionings, partition_count_cutoff, 4, par_tab4, canonical_patterns); |
479 | |
480 | delete[] canonical_patterns; |
481 | } |
482 | |