OLD | NEW |
1 /* | 1 /* |
2 * Copyright 2014 Google Inc. | 2 * Copyright 2014 Google Inc. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license that can be | 4 * Use of this source code is governed by a BSD-style license that can be |
5 * found in the LICENSE file. | 5 * found in the LICENSE file. |
6 */ | 6 */ |
7 | 7 |
8 #include "SkTextureCompressor.h" | 8 #include "SkTextureCompressor.h" |
9 #include "SkTextureCompressor_Blitter.h" | 9 #include "SkTextureCompressor_Blitter.h" |
10 #include "SkTextureCompressor_Utils.h" | 10 #include "SkTextureCompressor_Utils.h" |
(...skipping 257 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
268 return compress_heterogeneous_r11eac_block(block); | 268 return compress_heterogeneous_r11eac_block(block); |
269 } | 269 } |
270 | 270 |
271 // This function is used by R11 EAC to compress 4x4 blocks | 271 // This function is used by R11 EAC to compress 4x4 blocks |
272 // of 8-bit alpha into 64-bit values that comprise the compressed data. | 272 // of 8-bit alpha into 64-bit values that comprise the compressed data. |
273 // We need to make sure that the dimensions of the src pixels are divisible | 273 // We need to make sure that the dimensions of the src pixels are divisible |
274 // by 4, and copy 4x4 blocks one at a time for compression. | 274 // by 4, and copy 4x4 blocks one at a time for compression. |
275 typedef uint64_t (*A84x4To64BitProc)(const uint8_t block[]); | 275 typedef uint64_t (*A84x4To64BitProc)(const uint8_t block[]); |
276 | 276 |
277 static bool compress_4x4_a8_to_64bit(uint8_t* dst, const uint8_t* src, | 277 static bool compress_4x4_a8_to_64bit(uint8_t* dst, const uint8_t* src, |
278 int width, int height, int rowBytes, | 278 int width, int height, size_t rowBytes, |
279 A84x4To64BitProc proc) { | 279 A84x4To64BitProc proc) { |
280 // Make sure that our data is well-formed enough to be considered for compre
ssion | 280 // Make sure that our data is well-formed enough to be considered for compre
ssion |
281 if (0 == width || 0 == height || (width % 4) != 0 || (height % 4) != 0) { | 281 if (0 == width || 0 == height || (width % 4) != 0 || (height % 4) != 0) { |
282 return false; | 282 return false; |
283 } | 283 } |
284 | 284 |
285 int blocksX = width >> 2; | 285 int blocksX = width >> 2; |
286 int blocksY = height >> 2; | 286 int blocksY = height >> 2; |
287 | 287 |
288 uint8_t block[16]; | 288 uint8_t block[16]; |
(...skipping 125 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
414 x = (x & (0xFFFULL << 36)) | ((x & 0xFFFFFFULL) << 12) | ((x >> 24) & 0xFFFU
LL); | 414 x = (x & (0xFFFULL << 36)) | ((x & 0xFFFFFFULL) << 12) | ((x >> 24) & 0xFFFU
LL); |
415 #endif | 415 #endif |
416 | 416 |
417 // x: 00 00 00 00 00 00 00 00 a e i m b f j n c g k o d h l p | 417 // x: 00 00 00 00 00 00 00 00 a e i m b f j n c g k o d h l p |
418 return x; | 418 return x; |
419 } | 419 } |
420 | 420 |
421 // This function follows the same basic procedure as compress_heterogeneous_r11e
ac_block | 421 // This function follows the same basic procedure as compress_heterogeneous_r11e
ac_block |
422 // above when COMPRESS_R11_EAC_FAST is defined, but it avoids a few loads/stores
and | 422 // above when COMPRESS_R11_EAC_FAST is defined, but it avoids a few loads/stores
and |
423 // tries to optimize where it can using SIMD. | 423 // tries to optimize where it can using SIMD. |
424 static uint64_t compress_r11eac_block_fast(const uint8_t* src, int rowBytes) { | 424 static uint64_t compress_r11eac_block_fast(const uint8_t* src, size_t rowBytes)
{ |
425 // Store each row of alpha values in an integer | 425 // Store each row of alpha values in an integer |
426 const uint32_t alphaRow1 = *(reinterpret_cast<const uint32_t*>(src)); | 426 const uint32_t alphaRow1 = *(reinterpret_cast<const uint32_t*>(src)); |
427 const uint32_t alphaRow2 = *(reinterpret_cast<const uint32_t*>(src + rowByte
s)); | 427 const uint32_t alphaRow2 = *(reinterpret_cast<const uint32_t*>(src + rowByte
s)); |
428 const uint32_t alphaRow3 = *(reinterpret_cast<const uint32_t*>(src + 2*rowBy
tes)); | 428 const uint32_t alphaRow3 = *(reinterpret_cast<const uint32_t*>(src + 2*rowBy
tes)); |
429 const uint32_t alphaRow4 = *(reinterpret_cast<const uint32_t*>(src + 3*rowBy
tes)); | 429 const uint32_t alphaRow4 = *(reinterpret_cast<const uint32_t*>(src + 3*rowBy
tes)); |
430 | 430 |
431 // Check for solid blocks. The explanations for these values | 431 // Check for solid blocks. The explanations for these values |
432 // can be found in the comments of compress_r11eac_block above | 432 // can be found in the comments of compress_r11eac_block above |
433 if (alphaRow1 == alphaRow2 && alphaRow1 == alphaRow3 && alphaRow1 == alphaRo
w4) { | 433 if (alphaRow1 == alphaRow2 && alphaRow1 == alphaRow3 && alphaRow1 == alphaRo
w4) { |
434 if (0 == alphaRow1) { | 434 if (0 == alphaRow1) { |
(...skipping 20 matching lines...) Expand all Loading... |
455 // taken care of in interleave6. | 455 // taken care of in interleave6. |
456 const uint32_t r1r2 = (indexRow1 << 3) | indexRow2; | 456 const uint32_t r1r2 = (indexRow1 << 3) | indexRow2; |
457 const uint32_t r3r4 = (indexRow3 << 3) | indexRow4; | 457 const uint32_t r3r4 = (indexRow3 << 3) | indexRow4; |
458 const uint64_t indices = interleave6(r1r2, r3r4); | 458 const uint64_t indices = interleave6(r1r2, r3r4); |
459 | 459 |
460 // Return the packed incdices in the least significant bits with the magic h
eader | 460 // Return the packed incdices in the least significant bits with the magic h
eader |
461 return SkEndian_SwapBE64(0x8490000000000000ULL | indices); | 461 return SkEndian_SwapBE64(0x8490000000000000ULL | indices); |
462 } | 462 } |
463 | 463 |
464 static bool compress_a8_to_r11eac_fast(uint8_t* dst, const uint8_t* src, | 464 static bool compress_a8_to_r11eac_fast(uint8_t* dst, const uint8_t* src, |
465 int width, int height, int rowBytes) { | 465 int width, int height, size_t rowBytes) { |
466 // Make sure that our data is well-formed enough to be considered for compre
ssion | 466 // Make sure that our data is well-formed enough to be considered for compre
ssion |
467 if (0 == width || 0 == height || (width % 4) != 0 || (height % 4) != 0) { | 467 if (0 == width || 0 == height || (width % 4) != 0 || (height % 4) != 0) { |
468 return false; | 468 return false; |
469 } | 469 } |
470 | 470 |
471 const int blocksX = width >> 2; | 471 const int blocksX = width >> 2; |
472 const int blocksY = height >> 2; | 472 const int blocksY = height >> 2; |
473 | 473 |
474 uint64_t* encPtr = reinterpret_cast<uint64_t*>(dst); | 474 uint64_t* encPtr = reinterpret_cast<uint64_t*>(dst); |
475 for (int y = 0; y < blocksY; ++y) { | 475 for (int y = 0; y < blocksY; ++y) { |
(...skipping 136 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
612 // the R11EAC indices directly correspond to pixel values. | 612 // the R11EAC indices directly correspond to pixel values. |
613 SkFAIL("Implement me!"); | 613 SkFAIL("Implement me!"); |
614 } | 614 } |
615 #endif | 615 #endif |
616 }; | 616 }; |
617 | 617 |
618 //////////////////////////////////////////////////////////////////////////////// | 618 //////////////////////////////////////////////////////////////////////////////// |
619 | 619 |
620 namespace SkTextureCompressor { | 620 namespace SkTextureCompressor { |
621 | 621 |
622 bool CompressA8ToR11EAC(uint8_t* dst, const uint8_t* src, int width, int height,
int rowBytes) { | 622 bool CompressA8ToR11EAC(uint8_t* dst, const uint8_t* src, int width, int height,
size_t rowBytes) { |
623 | 623 |
624 #if (COMPRESS_R11_EAC_SLOW) || (COMPRESS_R11_EAC_FAST) | 624 #if (COMPRESS_R11_EAC_SLOW) || (COMPRESS_R11_EAC_FAST) |
625 | 625 |
626 return compress_4x4_a8_to_64bit(dst, src, width, height, rowBytes, compress_
r11eac_block); | 626 return compress_4x4_a8_to_64bit(dst, src, width, height, rowBytes, compress_
r11eac_block); |
627 | 627 |
628 #elif COMPRESS_R11_EAC_FASTEST | 628 #elif COMPRESS_R11_EAC_FASTEST |
629 | 629 |
630 return compress_a8_to_r11eac_fast(dst, src, width, height, rowBytes); | 630 return compress_a8_to_r11eac_fast(dst, src, width, height, rowBytes); |
631 | 631 |
632 #else | 632 #else |
(...skipping 28 matching lines...) Expand all Loading... |
661 for (int j = 0; j < height; j += 4) { | 661 for (int j = 0; j < height; j += 4) { |
662 for (int i = 0; i < width; i += 4) { | 662 for (int i = 0; i < width; i += 4) { |
663 decompress_r11_eac_block(dst + i, dstRowBytes, src); | 663 decompress_r11_eac_block(dst + i, dstRowBytes, src); |
664 src += 8; | 664 src += 8; |
665 } | 665 } |
666 dst += 4 * dstRowBytes; | 666 dst += 4 * dstRowBytes; |
667 } | 667 } |
668 } | 668 } |
669 | 669 |
670 } // namespace SkTextureCompressor | 670 } // namespace SkTextureCompressor |
OLD | NEW |