diff options
Diffstat (limited to '')
| -rw-r--r-- | src/video_core/textures/astc.cpp | 138 |
1 files changed, 60 insertions, 78 deletions
diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp index 3c4ad1c9d..b1feacae9 100644 --- a/src/video_core/textures/astc.cpp +++ b/src/video_core/textures/astc.cpp | |||
| @@ -25,16 +25,15 @@ | |||
| 25 | 25 | ||
| 26 | class BitStream { | 26 | class BitStream { |
| 27 | public: | 27 | public: |
| 28 | BitStream(unsigned char* ptr, int nBits = 0, int start_offset = 0) | 28 | explicit BitStream(unsigned char* ptr, int nBits = 0, int start_offset = 0) |
| 29 | : m_BitsWritten(0), m_BitsRead(0), m_NumBits(nBits), m_CurByte(ptr), | 29 | : m_NumBits(nBits), m_CurByte(ptr), m_NextBit(start_offset % 8) {} |
| 30 | m_NextBit(start_offset % 8), done(false) {} | 30 | |
| 31 | ~BitStream() = default; | ||
| 31 | 32 | ||
| 32 | int GetBitsWritten() const { | 33 | int GetBitsWritten() const { |
| 33 | return m_BitsWritten; | 34 | return m_BitsWritten; |
| 34 | } | 35 | } |
| 35 | 36 | ||
| 36 | ~BitStream() {} | ||
| 37 | |||
| 38 | void WriteBitsR(unsigned int val, unsigned int nBits) { | 37 | void WriteBitsR(unsigned int val, unsigned int nBits) { |
| 39 | for (unsigned int i = 0; i < nBits; i++) { | 38 | for (unsigned int i = 0; i < nBits; i++) { |
| 40 | WriteBit((val >> (nBits - i - 1)) & 1); | 39 | WriteBit((val >> (nBits - i - 1)) & 1); |
| @@ -95,33 +94,28 @@ private: | |||
| 95 | done = done || ++m_BitsWritten >= m_NumBits; | 94 | done = done || ++m_BitsWritten >= m_NumBits; |
| 96 | } | 95 | } |
| 97 | 96 | ||
| 98 | int m_BitsWritten; | 97 | int m_BitsWritten = 0; |
| 99 | const int m_NumBits; | 98 | const int m_NumBits; |
| 100 | unsigned char* m_CurByte; | 99 | unsigned char* m_CurByte; |
| 101 | int m_NextBit; | 100 | int m_NextBit = 0; |
| 102 | int m_BitsRead; | 101 | int m_BitsRead = 0; |
| 103 | 102 | ||
| 104 | bool done; | 103 | bool done = false; |
| 105 | }; | 104 | }; |
| 106 | 105 | ||
| 107 | template <typename IntType> | 106 | template <typename IntType> |
| 108 | class Bits { | 107 | class Bits { |
| 109 | private: | ||
| 110 | const IntType& m_Bits; | ||
| 111 | |||
| 112 | // Don't copy | ||
| 113 | Bits() {} | ||
| 114 | Bits(const Bits&) {} | ||
| 115 | Bits& operator=(const Bits&) {} | ||
| 116 | |||
| 117 | public: | 108 | public: |
| 118 | explicit Bits(IntType& v) : m_Bits(v) {} | 109 | explicit Bits(const IntType& v) : m_Bits(v) {} |
| 119 | 110 | ||
| 120 | uint8_t operator[](uint32_t bitPos) { | 111 | Bits(const Bits&) = delete; |
| 112 | Bits& operator=(const Bits&) = delete; | ||
| 113 | |||
| 114 | uint8_t operator[](uint32_t bitPos) const { | ||
| 121 | return static_cast<uint8_t>((m_Bits >> bitPos) & 1); | 115 | return static_cast<uint8_t>((m_Bits >> bitPos) & 1); |
| 122 | } | 116 | } |
| 123 | 117 | ||
| 124 | IntType operator()(uint32_t start, uint32_t end) { | 118 | IntType operator()(uint32_t start, uint32_t end) const { |
| 125 | if (start == end) { | 119 | if (start == end) { |
| 126 | return (*this)[start]; | 120 | return (*this)[start]; |
| 127 | } else if (start > end) { | 121 | } else if (start > end) { |
| @@ -133,6 +127,9 @@ public: | |||
| 133 | uint64_t mask = (1 << (end - start + 1)) - 1; | 127 | uint64_t mask = (1 << (end - start + 1)) - 1; |
| 134 | return (m_Bits >> start) & mask; | 128 | return (m_Bits >> start) & mask; |
| 135 | } | 129 | } |
| 130 | |||
| 131 | private: | ||
| 132 | const IntType& m_Bits; | ||
| 136 | }; | 133 | }; |
| 137 | 134 | ||
| 138 | enum EIntegerEncoding { eIntegerEncoding_JustBits, eIntegerEncoding_Quint, eIntegerEncoding_Trit }; | 135 | enum EIntegerEncoding { eIntegerEncoding_JustBits, eIntegerEncoding_Quint, eIntegerEncoding_Trit }; |
| @@ -186,12 +183,12 @@ public: | |||
| 186 | m_QuintValue = val; | 183 | m_QuintValue = val; |
| 187 | } | 184 | } |
| 188 | 185 | ||
| 189 | bool MatchesEncoding(const IntegerEncodedValue& other) { | 186 | bool MatchesEncoding(const IntegerEncodedValue& other) const { |
| 190 | return m_Encoding == other.m_Encoding && m_NumBits == other.m_NumBits; | 187 | return m_Encoding == other.m_Encoding && m_NumBits == other.m_NumBits; |
| 191 | } | 188 | } |
| 192 | 189 | ||
| 193 | // Returns the number of bits required to encode nVals values. | 190 | // Returns the number of bits required to encode nVals values. |
| 194 | uint32_t GetBitLength(uint32_t nVals) { | 191 | uint32_t GetBitLength(uint32_t nVals) const { |
| 195 | uint32_t totalBits = m_NumBits * nVals; | 192 | uint32_t totalBits = m_NumBits * nVals; |
| 196 | if (m_Encoding == eIntegerEncoding_Trit) { | 193 | if (m_Encoding == eIntegerEncoding_Trit) { |
| 197 | totalBits += (nVals * 8 + 4) / 5; | 194 | totalBits += (nVals * 8 + 4) / 5; |
| @@ -382,19 +379,15 @@ private: | |||
| 382 | namespace ASTCC { | 379 | namespace ASTCC { |
| 383 | 380 | ||
| 384 | struct TexelWeightParams { | 381 | struct TexelWeightParams { |
| 385 | uint32_t m_Width; | 382 | uint32_t m_Width = 0; |
| 386 | uint32_t m_Height; | 383 | uint32_t m_Height = 0; |
| 387 | bool m_bDualPlane; | 384 | bool m_bDualPlane = false; |
| 388 | uint32_t m_MaxWeight; | 385 | uint32_t m_MaxWeight = 0; |
| 389 | bool m_bError; | 386 | bool m_bError = false; |
| 390 | bool m_bVoidExtentLDR; | 387 | bool m_bVoidExtentLDR = false; |
| 391 | bool m_bVoidExtentHDR; | 388 | bool m_bVoidExtentHDR = false; |
| 392 | 389 | ||
| 393 | TexelWeightParams() { | 390 | uint32_t GetPackedBitSize() const { |
| 394 | memset(this, 0, sizeof(*this)); | ||
| 395 | } | ||
| 396 | |||
| 397 | uint32_t GetPackedBitSize() { | ||
| 398 | // How many indices do we have? | 391 | // How many indices do we have? |
| 399 | uint32_t nIdxs = m_Height * m_Width; | 392 | uint32_t nIdxs = m_Height * m_Width; |
| 400 | if (m_bDualPlane) { | 393 | if (m_bDualPlane) { |
| @@ -413,7 +406,7 @@ struct TexelWeightParams { | |||
| 413 | } | 406 | } |
| 414 | }; | 407 | }; |
| 415 | 408 | ||
| 416 | TexelWeightParams DecodeBlockInfo(BitStream& strm) { | 409 | static TexelWeightParams DecodeBlockInfo(BitStream& strm) { |
| 417 | TexelWeightParams params; | 410 | TexelWeightParams params; |
| 418 | 411 | ||
| 419 | // Read the entire block mode all at once | 412 | // Read the entire block mode all at once |
| @@ -612,8 +605,8 @@ TexelWeightParams DecodeBlockInfo(BitStream& strm) { | |||
| 612 | return params; | 605 | return params; |
| 613 | } | 606 | } |
| 614 | 607 | ||
| 615 | void FillVoidExtentLDR(BitStream& strm, uint32_t* const outBuf, uint32_t blockWidth, | 608 | static void FillVoidExtentLDR(BitStream& strm, uint32_t* const outBuf, uint32_t blockWidth, |
| 616 | uint32_t blockHeight) { | 609 | uint32_t blockHeight) { |
| 617 | // Don't actually care about the void extent, just read the bits... | 610 | // Don't actually care about the void extent, just read the bits... |
| 618 | for (int i = 0; i < 4; ++i) { | 611 | for (int i = 0; i < 4; ++i) { |
| 619 | strm.ReadBits(13); | 612 | strm.ReadBits(13); |
| @@ -628,23 +621,25 @@ void FillVoidExtentLDR(BitStream& strm, uint32_t* const outBuf, uint32_t blockWi | |||
| 628 | uint32_t rgba = (r >> 8) | (g & 0xFF00) | (static_cast<uint32_t>(b) & 0xFF00) << 8 | | 621 | uint32_t rgba = (r >> 8) | (g & 0xFF00) | (static_cast<uint32_t>(b) & 0xFF00) << 8 | |
| 629 | (static_cast<uint32_t>(a) & 0xFF00) << 16; | 622 | (static_cast<uint32_t>(a) & 0xFF00) << 16; |
| 630 | 623 | ||
| 631 | for (uint32_t j = 0; j < blockHeight; j++) | 624 | for (uint32_t j = 0; j < blockHeight; j++) { |
| 632 | for (uint32_t i = 0; i < blockWidth; i++) { | 625 | for (uint32_t i = 0; i < blockWidth; i++) { |
| 633 | outBuf[j * blockWidth + i] = rgba; | 626 | outBuf[j * blockWidth + i] = rgba; |
| 634 | } | 627 | } |
| 628 | } | ||
| 635 | } | 629 | } |
| 636 | 630 | ||
| 637 | void FillError(uint32_t* outBuf, uint32_t blockWidth, uint32_t blockHeight) { | 631 | static void FillError(uint32_t* outBuf, uint32_t blockWidth, uint32_t blockHeight) { |
| 638 | for (uint32_t j = 0; j < blockHeight; j++) | 632 | for (uint32_t j = 0; j < blockHeight; j++) { |
| 639 | for (uint32_t i = 0; i < blockWidth; i++) { | 633 | for (uint32_t i = 0; i < blockWidth; i++) { |
| 640 | outBuf[j * blockWidth + i] = 0xFFFF00FF; | 634 | outBuf[j * blockWidth + i] = 0xFFFF00FF; |
| 641 | } | 635 | } |
| 636 | } | ||
| 642 | } | 637 | } |
| 643 | 638 | ||
| 644 | // Replicates low numBits such that [(toBit - 1):(toBit - 1 - fromBit)] | 639 | // Replicates low numBits such that [(toBit - 1):(toBit - 1 - fromBit)] |
| 645 | // is the same as [(numBits - 1):0] and repeats all the way down. | 640 | // is the same as [(numBits - 1):0] and repeats all the way down. |
| 646 | template <typename IntType> | 641 | template <typename IntType> |
| 647 | IntType Replicate(const IntType& val, uint32_t numBits, uint32_t toBit) { | 642 | static IntType Replicate(const IntType& val, uint32_t numBits, uint32_t toBit) { |
| 648 | if (numBits == 0) | 643 | if (numBits == 0) |
| 649 | return 0; | 644 | return 0; |
| 650 | if (toBit == 0) | 645 | if (toBit == 0) |
| @@ -668,27 +663,15 @@ IntType Replicate(const IntType& val, uint32_t numBits, uint32_t toBit) { | |||
| 668 | 663 | ||
| 669 | class Pixel { | 664 | class Pixel { |
| 670 | protected: | 665 | protected: |
| 671 | typedef int16_t ChannelType; | 666 | using ChannelType = int16_t; |
| 672 | uint8_t m_BitDepth[4]; | 667 | uint8_t m_BitDepth[4] = {8, 8, 8, 8}; |
| 673 | int16_t color[4]; | 668 | int16_t color[4] = {}; |
| 674 | 669 | ||
| 675 | public: | 670 | public: |
| 676 | Pixel() { | 671 | Pixel() = default; |
| 677 | for (int i = 0; i < 4; i++) { | 672 | Pixel(ChannelType a, ChannelType r, ChannelType g, ChannelType b, unsigned bitDepth = 8) |
| 678 | m_BitDepth[i] = 8; | 673 | : m_BitDepth{uint8_t(bitDepth), uint8_t(bitDepth), uint8_t(bitDepth), uint8_t(bitDepth)}, |
| 679 | color[i] = 0; | 674 | color{a, r, g, b} {} |
| 680 | } | ||
| 681 | } | ||
| 682 | |||
| 683 | Pixel(ChannelType a, ChannelType r, ChannelType g, ChannelType b, unsigned bitDepth = 8) { | ||
| 684 | for (int i = 0; i < 4; i++) | ||
| 685 | m_BitDepth[i] = bitDepth; | ||
| 686 | |||
| 687 | color[0] = a; | ||
| 688 | color[1] = r; | ||
| 689 | color[2] = g; | ||
| 690 | color[3] = b; | ||
| 691 | } | ||
| 692 | 675 | ||
| 693 | // Changes the depth of each pixel. This scales the values to | 676 | // Changes the depth of each pixel. This scales the values to |
| 694 | // the appropriate bit depth by either truncating the least | 677 | // the appropriate bit depth by either truncating the least |
| @@ -807,8 +790,8 @@ public: | |||
| 807 | } | 790 | } |
| 808 | }; | 791 | }; |
| 809 | 792 | ||
| 810 | void DecodeColorValues(uint32_t* out, uint8_t* data, uint32_t* modes, const uint32_t nPartitions, | 793 | static void DecodeColorValues(uint32_t* out, uint8_t* data, const uint32_t* modes, |
| 811 | const uint32_t nBitsForColorData) { | 794 | const uint32_t nPartitions, const uint32_t nBitsForColorData) { |
| 812 | // First figure out how many color values we have | 795 | // First figure out how many color values we have |
| 813 | uint32_t nValues = 0; | 796 | uint32_t nValues = 0; |
| 814 | for (uint32_t i = 0; i < nPartitions; i++) { | 797 | for (uint32_t i = 0; i < nPartitions; i++) { |
| @@ -844,8 +827,7 @@ void DecodeColorValues(uint32_t* out, uint8_t* data, uint32_t* modes, const uint | |||
| 844 | // Once we have the decoded values, we need to dequantize them to the 0-255 range | 827 | // Once we have the decoded values, we need to dequantize them to the 0-255 range |
| 845 | // This procedure is outlined in ASTC spec C.2.13 | 828 | // This procedure is outlined in ASTC spec C.2.13 |
| 846 | uint32_t outIdx = 0; | 829 | uint32_t outIdx = 0; |
| 847 | std::vector<IntegerEncodedValue>::const_iterator itr; | 830 | for (auto itr = decodedColorValues.begin(); itr != decodedColorValues.end(); ++itr) { |
| 848 | for (itr = decodedColorValues.begin(); itr != decodedColorValues.end(); itr++) { | ||
| 849 | // Have we already decoded all that we need? | 831 | // Have we already decoded all that we need? |
| 850 | if (outIdx >= nValues) { | 832 | if (outIdx >= nValues) { |
| 851 | break; | 833 | break; |
| @@ -978,7 +960,7 @@ void DecodeColorValues(uint32_t* out, uint8_t* data, uint32_t* modes, const uint | |||
| 978 | } | 960 | } |
| 979 | } | 961 | } |
| 980 | 962 | ||
| 981 | uint32_t UnquantizeTexelWeight(const IntegerEncodedValue& val) { | 963 | static uint32_t UnquantizeTexelWeight(const IntegerEncodedValue& val) { |
| 982 | uint32_t bitval = val.GetBitValue(); | 964 | uint32_t bitval = val.GetBitValue(); |
| 983 | uint32_t bitlen = val.BaseBitLength(); | 965 | uint32_t bitlen = val.BaseBitLength(); |
| 984 | 966 | ||
| @@ -1067,17 +1049,18 @@ uint32_t UnquantizeTexelWeight(const IntegerEncodedValue& val) { | |||
| 1067 | return result; | 1049 | return result; |
| 1068 | } | 1050 | } |
| 1069 | 1051 | ||
| 1070 | void UnquantizeTexelWeights(uint32_t out[2][144], std::vector<IntegerEncodedValue>& weights, | 1052 | static void UnquantizeTexelWeights(uint32_t out[2][144], |
| 1071 | const TexelWeightParams& params, const uint32_t blockWidth, | 1053 | const std::vector<IntegerEncodedValue>& weights, |
| 1072 | const uint32_t blockHeight) { | 1054 | const TexelWeightParams& params, const uint32_t blockWidth, |
| 1055 | const uint32_t blockHeight) { | ||
| 1073 | uint32_t weightIdx = 0; | 1056 | uint32_t weightIdx = 0; |
| 1074 | uint32_t unquantized[2][144]; | 1057 | uint32_t unquantized[2][144]; |
| 1075 | std::vector<IntegerEncodedValue>::const_iterator itr; | 1058 | |
| 1076 | for (itr = weights.begin(); itr != weights.end(); itr++) { | 1059 | for (auto itr = weights.begin(); itr != weights.end(); ++itr) { |
| 1077 | unquantized[0][weightIdx] = UnquantizeTexelWeight(*itr); | 1060 | unquantized[0][weightIdx] = UnquantizeTexelWeight(*itr); |
| 1078 | 1061 | ||
| 1079 | if (params.m_bDualPlane) { | 1062 | if (params.m_bDualPlane) { |
| 1080 | itr++; | 1063 | ++itr; |
| 1081 | unquantized[1][weightIdx] = UnquantizeTexelWeight(*itr); | 1064 | unquantized[1][weightIdx] = UnquantizeTexelWeight(*itr); |
| 1082 | if (itr == weights.end()) { | 1065 | if (itr == weights.end()) { |
| 1083 | break; | 1066 | break; |
| @@ -1261,8 +1244,8 @@ static inline uint32_t Select2DPartition(int32_t seed, int32_t x, int32_t y, int | |||
| 1261 | } | 1244 | } |
| 1262 | 1245 | ||
| 1263 | // Section C.2.14 | 1246 | // Section C.2.14 |
| 1264 | void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValues, | 1247 | static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValues, |
| 1265 | uint32_t colorEndpointMode) { | 1248 | uint32_t colorEndpointMode) { |
| 1266 | #define READ_UINT_VALUES(N) \ | 1249 | #define READ_UINT_VALUES(N) \ |
| 1267 | uint32_t v[N]; \ | 1250 | uint32_t v[N]; \ |
| 1268 | for (uint32_t i = 0; i < N; i++) { \ | 1251 | for (uint32_t i = 0; i < N; i++) { \ |
| @@ -1382,8 +1365,8 @@ void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValues, | |||
| 1382 | #undef READ_INT_VALUES | 1365 | #undef READ_INT_VALUES |
| 1383 | } | 1366 | } |
| 1384 | 1367 | ||
| 1385 | void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth, const uint32_t blockHeight, | 1368 | static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth, |
| 1386 | uint32_t* outBuf) { | 1369 | const uint32_t blockHeight, uint32_t* outBuf) { |
| 1387 | BitStream strm(inBuf); | 1370 | BitStream strm(inBuf); |
| 1388 | TexelWeightParams weightParams = DecodeBlockInfo(strm); | 1371 | TexelWeightParams weightParams = DecodeBlockInfo(strm); |
| 1389 | 1372 | ||
| @@ -1617,8 +1600,7 @@ namespace Tegra::Texture::ASTC { | |||
| 1617 | std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint32_t height, | 1600 | std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint32_t height, |
| 1618 | uint32_t block_width, uint32_t block_height) { | 1601 | uint32_t block_width, uint32_t block_height) { |
| 1619 | uint32_t blockIdx = 0; | 1602 | uint32_t blockIdx = 0; |
| 1620 | std::vector<uint8_t> outData; | 1603 | std::vector<uint8_t> outData(height * width * 4); |
| 1621 | outData.resize(height * width * 4); | ||
| 1622 | for (uint32_t j = 0; j < height; j += block_height) { | 1604 | for (uint32_t j = 0; j < height; j += block_height) { |
| 1623 | for (uint32_t i = 0; i < width; i += block_width) { | 1605 | for (uint32_t i = 0; i < width; i += block_width) { |
| 1624 | 1606 | ||