astc.h: Move data to cpp implementation

Moves leftover values that are no longer used by the gpu decoder back to the cpp implementation.
This commit is contained in:
ameerj 2021-07-04 22:07:53 -04:00
parent 7cf0958b06
commit 15c0c213b1
2 changed files with 63 additions and 64 deletions

View file

@ -521,35 +521,41 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
return params;
}
static void FillVoidExtentLDR(InputBitStream& strm, std::span<u32> outBuf, u32 blockWidth,
u32 blockHeight) {
// Don't actually care about the void extent, just read the bits...
for (s32 i = 0; i < 4; ++i) {
strm.ReadBits<13>();
// Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)]
// is the same as [(num_bits - 1):0] and repeats all the way down.
template <typename IntType>
static constexpr IntType Replicate(IntType val, u32 num_bits, u32 to_bit) {
if (num_bits == 0 || to_bit == 0) {
return 0;
}
const IntType v = val & static_cast<IntType>((1 << num_bits) - 1);
IntType res = v;
u32 reslen = num_bits;
while (reslen < to_bit) {
u32 comp = 0;
if (num_bits > to_bit - reslen) {
u32 newshift = to_bit - reslen;
comp = num_bits - newshift;
num_bits = newshift;
}
res = static_cast<IntType>(res << num_bits);
res = static_cast<IntType>(res | (v >> comp));
reslen += num_bits;
}
return res;
}
// Decode the RGBA components and renormalize them to the range [0, 255]
u16 r = static_cast<u16>(strm.ReadBits<16>());
u16 g = static_cast<u16>(strm.ReadBits<16>());
u16 b = static_cast<u16>(strm.ReadBits<16>());
u16 a = static_cast<u16>(strm.ReadBits<16>());
u32 rgba = (r >> 8) | (g & 0xFF00) | (static_cast<u32>(b) & 0xFF00) << 8 |
(static_cast<u32>(a) & 0xFF00) << 16;
for (u32 j = 0; j < blockHeight; j++) {
for (u32 i = 0; i < blockWidth; i++) {
outBuf[j * blockWidth + i] = rgba;
}
}
static constexpr std::size_t NumReplicateEntries(u32 num_bits) {
return std::size_t(1) << num_bits;
}
static void FillError(std::span<u32> outBuf, u32 blockWidth, u32 blockHeight) {
for (u32 j = 0; j < blockHeight; j++) {
for (u32 i = 0; i < blockWidth; i++) {
outBuf[j * blockWidth + i] = 0xFFFF00FF;
}
template <typename IntType, u32 num_bits, u32 to_bit>
static constexpr auto MakeReplicateTable() {
std::array<IntType, NumReplicateEntries(num_bits)> table{};
for (IntType value = 0; value < static_cast<IntType>(std::size(table)); ++value) {
table[value] = Replicate(value, num_bits, to_bit);
}
return table;
}
static constexpr auto REPLICATE_BYTE_TO_16_TABLE = MakeReplicateTable<u32, 8, 16>();
@ -572,6 +578,9 @@ static constexpr auto REPLICATE_2_BIT_TO_8_TABLE = MakeReplicateTable<u32, 2, 8>
static constexpr auto REPLICATE_3_BIT_TO_8_TABLE = MakeReplicateTable<u32, 3, 8>();
static constexpr auto REPLICATE_4_BIT_TO_8_TABLE = MakeReplicateTable<u32, 4, 8>();
static constexpr auto REPLICATE_5_BIT_TO_8_TABLE = MakeReplicateTable<u32, 5, 8>();
static constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable<u32, 6, 8>();
static constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable<u32, 7, 8>();
static constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable<u32, 8, 8>();
/// Use a precompiled table with the most common usages, if it's not in the expected range, fallback
/// to the runtime implementation
static constexpr u32 FastReplicateTo8(u32 value, u32 num_bits) {
@ -1316,6 +1325,37 @@ static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const u32*& colorValues,
#undef READ_INT_VALUES
}
static void FillVoidExtentLDR(InputBitStream& strm, std::span<u32> outBuf, u32 blockWidth,
u32 blockHeight) {
// Don't actually care about the void extent, just read the bits...
for (s32 i = 0; i < 4; ++i) {
strm.ReadBits<13>();
}
// Decode the RGBA components and renormalize them to the range [0, 255]
u16 r = static_cast<u16>(strm.ReadBits<16>());
u16 g = static_cast<u16>(strm.ReadBits<16>());
u16 b = static_cast<u16>(strm.ReadBits<16>());
u16 a = static_cast<u16>(strm.ReadBits<16>());
u32 rgba = (r >> 8) | (g & 0xFF00) | (static_cast<u32>(b) & 0xFF00) << 8 |
(static_cast<u32>(a) & 0xFF00) << 16;
for (u32 j = 0; j < blockHeight; j++) {
for (u32 i = 0; i < blockWidth; i++) {
outBuf[j * blockWidth + i] = rgba;
}
}
}
static void FillError(std::span<u32> outBuf, u32 blockWidth, u32 blockHeight) {
for (u32 j = 0; j < blockHeight; j++) {
for (u32 i = 0; i < blockWidth; i++) {
outBuf[j * blockWidth + i] = 0xFFFF00FF;
}
}
}
static void DecompressBlock(std::span<const u8, 16> inBuf, const u32 blockWidth,
const u32 blockHeight, std::span<u32, 12 * 12> outBuf) {
InputBitStream strm(inBuf);

View file

@ -79,47 +79,6 @@ constexpr std::array<IntegerEncodedValue, 256> MakeEncodedValues() {
constexpr std::array<IntegerEncodedValue, 256> ASTC_ENCODINGS_VALUES = MakeEncodedValues();
// Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)]
// is the same as [(num_bits - 1):0] and repeats all the way down.
template <typename IntType>
constexpr IntType Replicate(IntType val, u32 num_bits, u32 to_bit) {
if (num_bits == 0 || to_bit == 0) {
return 0;
}
const IntType v = val & static_cast<IntType>((1 << num_bits) - 1);
IntType res = v;
u32 reslen = num_bits;
while (reslen < to_bit) {
u32 comp = 0;
if (num_bits > to_bit - reslen) {
u32 newshift = to_bit - reslen;
comp = num_bits - newshift;
num_bits = newshift;
}
res = static_cast<IntType>(res << num_bits);
res = static_cast<IntType>(res | (v >> comp));
reslen += num_bits;
}
return res;
}
constexpr std::size_t NumReplicateEntries(u32 num_bits) {
return std::size_t(1) << num_bits;
}
template <typename IntType, u32 num_bits, u32 to_bit>
constexpr auto MakeReplicateTable() {
std::array<IntType, NumReplicateEntries(num_bits)> table{};
for (IntType value = 0; value < static_cast<IntType>(std::size(table)); ++value) {
table[value] = Replicate(value, num_bits, to_bit);
}
return table;
}
constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable<u32, 6, 8>();
constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable<u32, 7, 8>();
constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable<u32, 8, 8>();
void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
uint32_t block_width, uint32_t block_height, std::span<uint8_t> output);