Diamond75/Panda3DS
1
0
Fork 0
mirror of https://github.com/wheremyfoodat/Panda3DS.git synced 2025-04-06 06:05:40 +12:00
Code Issues Projects Releases Packages Wiki Activity Actions

metal: implement texture decoder

Browse source
This commit is contained in:
Samuliak 2025-03-11 08:11:39 +01:00
parent c59ee99364
commit 3a654b3609
No known key found for this signature in database
11 changed files with 419 additions and 223 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
CMakeLists.txt
View file

@ -420,7 +420,7 @@ set(HEADER_FILES include/emulator.hpp include/helpers.hpp include/termcolor.hpp
if(IOS) if(IOS)
set(SOURCE_FILES ${SOURCE_FILES} src/miniaudio/miniaudio.m) set(SOURCE_FILES ${SOURCE_FILES} src/miniaudio/miniaudio.m)
target_compile_definitions(AlberCore PUBLIC "PANDA3DS_IOS=1") target_compile_definitions(AlberCore PUBLIC "PANDA3DS_IOS=1")
if (IOS_SIMULATOR_BUILD) if (IOS_SIMULATOR_BUILD)
target_compile_definitions(AlberCore PUBLIC "PANDA3DS_IOS_SIMULATOR=1") target_compile_definitions(AlberCore PUBLIC "PANDA3DS_IOS_SIMULATOR=1")
endif() endif()
@ -604,14 +604,16 @@ if(ENABLE_METAL AND APPLE)
include/renderer_mtl/mtl_common.hpp include/renderer_mtl/mtl_common.hpp
include/renderer_mtl/pica_to_mtl.hpp include/renderer_mtl/pica_to_mtl.hpp
include/renderer_mtl/objc_helper.hpp include/renderer_mtl/objc_helper.hpp
include/renderer_mtl/texture_decoder.hpp
) )
set(RENDERER_MTL_SOURCE_FILES src/core/renderer_mtl/metal_cpp_impl.cpp set(RENDERER_MTL_SOURCE_FILES src/core/renderer_mtl/metal_cpp_impl.cpp
src/core/renderer_mtl/renderer_mtl.cpp src/core/renderer_mtl/renderer_mtl.cpp
src/core/renderer_mtl/mtl_texture.cpp src/core/renderer_mtl/mtl_texture.cpp
src/core/renderer_mtl/mtl_etc1.cpp
src/core/renderer_mtl/mtl_lut_texture.cpp src/core/renderer_mtl/mtl_lut_texture.cpp
src/core/renderer_mtl/pica_to_mtl.cpp
src/core/renderer_mtl/objc_helper.mm src/core/renderer_mtl/objc_helper.mm
src/core/renderer_mtl/texture_decoder.cpp
src/host_shaders/metal_shaders.metal src/host_shaders/metal_shaders.metal
src/host_shaders/metal_blit.metal src/host_shaders/metal_blit.metal
#src/host_shaders/metal_copy_to_lut_texture.metal #src/host_shaders/metal_copy_to_lut_texture.metal
@ -808,7 +810,7 @@ if(NOT BUILD_HYDRA_CORE AND NOT BUILD_LIBRETRO_CORE)
elseif(BUILD_HYDRA_CORE) elseif(BUILD_HYDRA_CORE)
target_compile_definitions(AlberCore PRIVATE PANDA3DS_HYDRA_CORE=1) target_compile_definitions(AlberCore PRIVATE PANDA3DS_HYDRA_CORE=1)
include_directories(third_party/hydra_core/include) include_directories(third_party/hydra_core/include)
set(SHARED_SOURCE_FILES src/hydra_core.cpp) set(SHARED_SOURCE_FILES src/hydra_core.cpp)
if(IOS) if(IOS)
set(SHARED_SOURCE_FILES ${SHARED_SOURCE_FILES} src/ios_driver.mm) set(SHARED_SOURCE_FILES ${SHARED_SOURCE_FILES} src/ios_driver.mm)

28
include/renderer_mtl/mtl_texture.hpp
View file

@ -8,8 +8,9 @@
#include "boost/icl/interval.hpp" #include "boost/icl/interval.hpp"
#include "helpers.hpp" #include "helpers.hpp"
#include "math_util.hpp" #include "math_util.hpp"
#include "opengl.hpp"
#include "renderer_mtl/pica_to_mtl.hpp" #include "renderer_mtl/pica_to_mtl.hpp"
// TODO: remove dependency on OpenGL
#include "opengl.hpp"
template <typename T> template <typename T>
using Interval = boost::icl::right_open_interval<T>; using Interval = boost::icl::right_open_interval<T>;
@ -52,22 +53,19 @@ namespace Metal {
void free(); void free();
u64 sizeInBytes(); u64 sizeInBytes();
u8 decodeTexelU8(u32 u, u32 v, PICA::TextureFmt fmt, std::span<const u8> data); u8 decodeTexelBGR8ToRGBA8(u32 u, u32 v, std::span<const u8> data);
u16 decodeTexelU16(u32 u, u32 v, PICA::TextureFmt fmt, std::span<const u8> data); u8 decodeTexelA1BGR5ToRGBA8(u32 u, u32 v, std::span<const u8> data);
u32 decodeTexelU32(u32 u, u32 v, PICA::TextureFmt fmt, std::span<const u8> data); u8 decodeTexelB5G6R5ToRGBA8(u32 u, u32 v, std::span<const u8> data);
u8 decodeTexelABGR4ToRGBA8(u32 u, u32 v, std::span<const u8> data);
// Get the morton interleave offset of a texel based on its U and V values u8 decodeTexelAI8ToRGBA8(u32 u, u32 v, std::span<const u8> data);
static u32 mortonInterleave(u32 u, u32 v); u8 decodeTexelI8ToRGBA8(u32 u, u32 v, std::span<const u8> data);
// Get the byte offset of texel (u, v) in the texture u8 decodeTexelAI4ToRGBA4(u32 u, u32 v, std::span<const u8> data);
static u32 getSwizzledOffset(u32 u, u32 v, u32 width, u32 bytesPerPixel); u8 decodeTexelAI4ToRGBA8(u32 u, u32 v, std::span<const u8> data);
static u32 getSwizzledOffset_4bpp(u32 u, u32 v, u32 width); u8 decodeTexelI4ToRGBA4(u32 u, u32 v, std::span<const u8> data);
u8 decodeTexelI4ToRGBA8(u32 u, u32 v, std::span<const u8> data);
u8 decodeTexelA4ToA8(u32 u, u32 v, std::span<const u8> data);
// Returns the format of this texture as a string // Returns the format of this texture as a string
std::string_view formatToString() { return PICA::textureFormatToString(format); } std::string_view formatToString() { return PICA::textureFormatToString(format); }
// Returns the texel at coordinates (u, v) of an ETC1(A4) texture
// TODO: Make hasAlpha a template parameter
u32 getTexelETC(bool hasAlpha, u32 u, u32 v, u32 width, std::span<const u8> data);
u32 decodeETC(u32 alpha, u32 u, u32 v, u64 colourData);
}; };
} // namespace Metal } // namespace Metal

44
include/renderer_mtl/pica_to_mtl.hpp
View file

@ -3,49 +3,21 @@
#include <Metal/Metal.hpp> #include <Metal/Metal.hpp>
#include "PICA/regs.hpp" #include "PICA/regs.hpp"
// TODO: remove dependency on OpenGL
#include "opengl.hpp"
namespace PICA { namespace PICA {
struct PixelFormatInfo { struct PixelFormatInfo {
MTL::PixelFormat pixelFormat; MTL::PixelFormat pixelFormat;
size_t bytesPerTexel; size_t bytesPerTexel;
void (*decoder)(OpenGL::uvec2, u32, u32, std::span<const u8>, std::vector<u8>&);
bool needsSwizzle{false};
// TODO: swizzle
}; };
// iOS, at least on simulator, doesn't support a lot of more "exotic" texture formats, so we avoid them tehre extern PixelFormatInfo pixelFormatInfos[14];
#ifndef PANDA3DS_IOS
constexpr PixelFormatInfo pixelFormatInfos[14] = { void checkForPixelFormatSupport(MTL::Device* device);
{MTL::PixelFormatRGBA8Unorm, 4}, // RGBA8
{MTL::PixelFormatRGBA8Unorm, 4}, // RGB8
{MTL::PixelFormatBGR5A1Unorm, 2}, // RGBA5551
{MTL::PixelFormatB5G6R5Unorm, 2}, // RGB565
{MTL::PixelFormatABGR4Unorm, 2}, // RGBA4
{MTL::PixelFormatRGBA8Unorm, 4}, // IA8
{MTL::PixelFormatRG8Unorm, 2}, // RG8
{MTL::PixelFormatRGBA8Unorm, 4}, // I8
{MTL::PixelFormatA8Unorm, 1}, // A8
{MTL::PixelFormatABGR4Unorm, 2}, // IA4
{MTL::PixelFormatABGR4Unorm, 2}, // I4
{MTL::PixelFormatA8Unorm, 1}, // A4
{MTL::PixelFormatRGBA8Unorm, 4}, // ETC1
{MTL::PixelFormatRGBA8Unorm, 4}, // ETC1A4
};
#else
constexpr PixelFormatInfo pixelFormatInfos[14] = {
{MTL::PixelFormatRGBA8Unorm, 4}, // RGBA8
{MTL::PixelFormatRGBA8Unorm, 4}, // RGB8
{MTL::PixelFormatBGR5A1Unorm, 2}, // RGBA5551
{MTL::PixelFormatRGBA8Unorm, 4}, // RGB565
{MTL::PixelFormatRGBA8Unorm, 4}, // RGBA4
{MTL::PixelFormatRGBA8Unorm, 4}, // IA8
{MTL::PixelFormatRG8Unorm, 2}, // RG8
{MTL::PixelFormatRGBA8Unorm, 4}, // I8
{MTL::PixelFormatA8Unorm, 1}, // A8
{MTL::PixelFormatRGBA8Unorm, 4}, // IA4
{MTL::PixelFormatRGBA8Unorm, 4}, // I4
{MTL::PixelFormatA8Unorm, 1}, // A4
{MTL::PixelFormatRGBA8Unorm, 4}, // ETC1
{MTL::PixelFormatRGBA8Unorm, 4}, // ETC1A4
};
#endif
inline PixelFormatInfo getPixelFormatInfo(TextureFmt format) { return pixelFormatInfos[static_cast<int>(format)]; } inline PixelFormatInfo getPixelFormatInfo(TextureFmt format) { return pixelFormatInfos[static_cast<int>(format)]; }

23
include/renderer_mtl/texture_decoder.hpp Normal file
View file

@ -0,0 +1,23 @@
#pragma once
#include "helpers.hpp"
// TODO: remove dependency on OpenGL
#include "opengl.hpp"
void decodeTexelABGR8ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelBGR8ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelA1BGR5ToBGR5A1(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelA1BGR5ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelB5G6R5ToB5G6R5(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelB5G6R5ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelABGR4ToABGR4(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelABGR4ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelAI8ToRG8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelGR8ToRG8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelI8ToR8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelA8ToA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelAI4ToRG8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelI4ToR8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelA4ToA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelETC1ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);
void decodeTexelETC1A4ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData);

116
src/core/renderer_mtl/mtl_etc1.cpp
View file

@ -1,116 +0,0 @@
#include <algorithm>
#include "colour.hpp"
#include "renderer_mtl/mtl_texture.hpp"
#include "renderer_mtl/renderer_mtl.hpp"
using namespace Helpers;
namespace Metal {
static constexpr u32 signExtend3To32(u32 val) {
return (u32)(s32(val) << 29 >> 29);
}
u32 Texture::getTexelETC(bool hasAlpha, u32 u, u32 v, u32 width, std::span<const u8> data) {
// Pixel offset of the 8x8 tile based on u, v and the width of the texture
u32 offs = ((u & ~7) * 8) + ((v & ~7) * width);
if (!hasAlpha) {
offs >>= 1;
}
// In-tile offsets for u/v
u &= 7;
v &= 7;
// ETC1(A4) also subdivide the 8x8 tile to 4 4x4 tiles
// Each tile is 8 bytes for ETC1, but since ETC1A4 has 4 alpha bits per pixel, that becomes 16 bytes
const u32 subTileSize = hasAlpha ? 16 : 8;
const u32 subTileIndex = (u / 4) + 2 * (v / 4); // Which of the 4 subtiles is this texel in?
// In-subtile offsets for u/v
u &= 3;
v &= 3;
offs += subTileSize * subTileIndex;
u32 alpha;
const u64* ptr = reinterpret_cast<const u64*>(data.data() + offs); // Cast to u64*
if (hasAlpha) {
// First 64 bits of the 4x4 subtile are alpha data
const u64 alphaData = *ptr++;
alpha = Colour::convert4To8Bit((alphaData >> (4 * (u * 4 + v))) & 0xf);
} else {
alpha = 0xff; // ETC1 without alpha uses ff for every pixel
}
// Next 64 bits of the subtile are colour data
u64 colourData = *ptr;
return decodeETC(alpha, u, v, colourData);
}
u32 Texture::decodeETC(u32 alpha, u32 u, u32 v, u64 colourData) {
static constexpr u32 modifiers[8][2] = {
{2, 8}, {5, 17}, {9, 29}, {13, 42}, {18, 60}, {24, 80}, {33, 106}, {47, 183},
};
// Parse colour data for 4x4 block
const u32 subindices = getBits<0, 16, u32>(colourData);
const u32 negationFlags = getBits<16, 16, u32>(colourData);
const bool flip = getBit<32>(colourData);
const bool diffMode = getBit<33>(colourData);
// Note: index1 is indeed stored on the higher bits, with index2 in the lower bits
const u32 tableIndex1 = getBits<37, 3, u32>(colourData);
const u32 tableIndex2 = getBits<34, 3, u32>(colourData);
const u32 texelIndex = u * 4 + v; // Index of the texel in the block
if (flip) std::swap(u, v);
s32 r, g, b;
if (diffMode) {
r = getBits<59, 5, s32>(colourData);
g = getBits<51, 5, s32>(colourData);
b = getBits<43, 5, s32>(colourData);
if (u >= 2) {
r += signExtend3To32(getBits<56, 3, u32>(colourData));
g += signExtend3To32(getBits<48, 3, u32>(colourData));
b += signExtend3To32(getBits<40, 3, u32>(colourData));
}
// Expand from 5 to 8 bits per channel
r = Colour::convert5To8Bit(r);
g = Colour::convert5To8Bit(g);
b = Colour::convert5To8Bit(b);
} else {
if (u < 2) {
r = getBits<60, 4, s32>(colourData);
g = getBits<52, 4, s32>(colourData);
b = getBits<44, 4, s32>(colourData);
} else {
r = getBits<56, 4, s32>(colourData);
g = getBits<48, 4, s32>(colourData);
b = getBits<40, 4, s32>(colourData);
}
// Expand from 4 to 8 bits per channel
r = Colour::convert4To8Bit(r);
g = Colour::convert4To8Bit(g);
b = Colour::convert4To8Bit(b);
}
const u32 index = (u < 2) ? tableIndex1 : tableIndex2;
s32 modifier = modifiers[index][(subindices >> texelIndex) & 1];
if (((negationFlags >> texelIndex) & 1) != 0) {
modifier = -modifier;
}
r = std::clamp(r + modifier, 0, 255);
g = std::clamp(g + modifier, 0, 255);
b = std::clamp(b + modifier, 0, 255);
return (alpha << 24) | (u32(b) << 16) | (u32(g) << 8) | u32(r);
}
} // namespace Metal

58
src/core/renderer_mtl/mtl_texture.cpp
View file

@ -99,35 +99,7 @@ namespace Metal {
} }
} }
// u and v are the UVs of the relevant texel /*
// Texture data is stored interleaved in Morton order, ie in a Z - order curve as shown here
// https://en.wikipedia.org/wiki/Z-order_curve
// Textures are split into 8x8 tiles.This function returns the in - tile offset depending on the u & v of the texel
// The in - tile offset is the sum of 2 offsets, one depending on the value of u % 8 and the other on the value of y % 8
// As documented in this picture https ://en.wikipedia.org/wiki/File:Moser%E2%80%93de_Bruijn_addition.svg
u32 Texture::mortonInterleave(u32 u, u32 v) {
static constexpr u32 xOffsets[] = {0, 1, 4, 5, 16, 17, 20, 21};
static constexpr u32 yOffsets[] = {0, 2, 8, 10, 32, 34, 40, 42};
return xOffsets[u & 7] + yOffsets[v & 7];
}
// Get the byte offset of texel (u, v) in the texture
u32 Texture::getSwizzledOffset(u32 u, u32 v, u32 width, u32 bytesPerPixel) {
u32 offset = ((u & ~7) * 8) + ((v & ~7) * width); // Offset of the 8x8 tile the texel belongs to
offset += mortonInterleave(u, v); // Add the in-tile offset of the texel
return offset * bytesPerPixel;
}
// Same as the above code except we need to divide by 2 because 4 bits is smaller than a byte
u32 Texture::getSwizzledOffset_4bpp(u32 u, u32 v, u32 width) {
u32 offset = ((u & ~7) * 8) + ((v & ~7) * width); // Offset of the 8x8 tile the texel belongs to
offset += mortonInterleave(u, v); // Add the in-tile offset of the texel
return offset / 2;
}
u8 Texture::decodeTexelU8(u32 u, u32 v, PICA::TextureFmt fmt, std::span<const u8> data) { u8 Texture::decodeTexelU8(u32 u, u32 v, PICA::TextureFmt fmt, std::span<const u8> data) {
switch (fmt) { switch (fmt) {
case PICA::TextureFmt::A4: { case PICA::TextureFmt::A4: {
@ -331,32 +303,18 @@ namespace Metal {
default: Helpers::panic("[Texture::DecodeTexel] Unimplemented format = %d", static_cast<int>(fmt)); default: Helpers::panic("[Texture::DecodeTexel] Unimplemented format = %d", static_cast<int>(fmt));
} }
} }
*/
void Texture::decodeTexture(std::span<const u8> data) { void Texture::decodeTexture(std::span<const u8> data) {
std::vector<u8> decoded; std::vector<u8> decoded;
decoded.reserve(u64(size.u()) * u64(size.v()) * formatInfo.bytesPerTexel); decoded.reserve(u64(size.u()) * u64(size.v()) * formatInfo.bytesPerTexel);
// Decode texels line by line // Decode texels line by line
for (u32 v = 0; v < size.v(); v++) { for (u32 v = 0; v < size.v(); v++) {
for (u32 u = 0; u < size.u(); u++) { for (u32 u = 0; u < size.u(); u++) {
if (formatInfo.bytesPerTexel == 1) { formatInfo.decoder(size, u, v, data, decoded);
u8 texel = decodeTexelU8(u, v, format, data); }
decoded.push_back(texel); }
} else if (formatInfo.bytesPerTexel == 2) {
u16 texel = decodeTexelU16(u, v, format, data);
decoded.push_back((texel & 0x00ff) >> 0);
decoded.push_back((texel & 0xff00) >> 8);
} else if (formatInfo.bytesPerTexel == 4) {
u32 texel = decodeTexelU32(u, v, format, data);
decoded.push_back((texel & 0x000000ff) >> 0);
decoded.push_back((texel & 0x0000ff00) >> 8);
decoded.push_back((texel & 0x00ff0000) >> 16);
decoded.push_back((texel & 0xff000000) >> 24);
} else {
Helpers::panic("[Texture::decodeTexture] Unimplemented bytesPerTexel (%u)", formatInfo.bytesPerTexel);
}
}
}
texture->replaceRegion(MTL::Region(0, 0, size.u(), size.v()), 0, 0, decoded.data(), formatInfo.bytesPerTexel * size.u(), 0); texture->replaceRegion(MTL::Region(0, 0, size.u(), size.v()), 0, 0, decoded.data(), formatInfo.bytesPerTexel * size.u(), 0);
} }

33
src/core/renderer_mtl/pica_to_mtl.cpp Normal file
View file

@ -0,0 +1,33 @@
#include "renderer_mtl/pica_to_mtl.hpp"
#include "renderer_mtl/texture_decoder.hpp"
using namespace Helpers;
namespace PICA {
PixelFormatInfo pixelFormatInfos[14] = {
{MTL::PixelFormatRGBA8Unorm, 4, decodeTexelABGR8ToRGBA8}, // RGBA8
{MTL::PixelFormatRGBA8Unorm, 4, decodeTexelBGR8ToRGBA8}, // RGB8
{MTL::PixelFormatBGR5A1Unorm, 2, decodeTexelA1BGR5ToBGR5A1}, // RGBA5551
{MTL::PixelFormatB5G6R5Unorm, 2, decodeTexelB5G6R5ToB5G6R5}, // RGB565
{MTL::PixelFormatABGR4Unorm, 2, decodeTexelABGR4ToABGR4}, // RGBA4
{MTL::PixelFormatRG8Unorm, 2, decodeTexelAI8ToRG8}, // IA8
{MTL::PixelFormatRG8Unorm, 2, decodeTexelGR8ToRG8}, // RG8
{MTL::PixelFormatR8Unorm, 1, decodeTexelI8ToR8}, // I8
{MTL::PixelFormatA8Unorm, 1, decodeTexelA8ToA8}, // A8
{MTL::PixelFormatRG8Unorm, 2, decodeTexelAI4ToRG8}, // IA4
{MTL::PixelFormatR8Unorm, 1, decodeTexelI4ToR8}, // I4
{MTL::PixelFormatA8Unorm, 1, decodeTexelA4ToA8}, // A4
{MTL::PixelFormatRGBA8Unorm, 4, decodeTexelETC1ToRGBA8}, // ETC1
{MTL::PixelFormatRGBA8Unorm, 4, decodeTexelETC1A4ToRGBA8}, // ETC1A4
};
void checkForPixelFormatSupport(MTL::Device* device) {
if (!device->supportsFamily(MTL::GPUFamilyApple1)) {
// TODO
throw;
}
}
}

2
src/core/renderer_mtl/renderer_mtl.cpp
View file

@ -141,9 +141,9 @@ void RendererMTL::initGraphicsContext(SDL_Window* window) {
device = MTL::CreateSystemDefaultDevice(); device = MTL::CreateSystemDefaultDevice();
metalLayer->setDevice(device); metalLayer->setDevice(device);
#endif #endif
checkForPixelFormatSupport(device);
commandQueue = device->newCommandQueue(); commandQueue = device->newCommandQueue();
printf("C++ device pointer: %p\n", device);
// Textures // Textures
MTL::TextureDescriptor* textureDescriptor = MTL::TextureDescriptor::alloc()->init(); MTL::TextureDescriptor* textureDescriptor = MTL::TextureDescriptor::alloc()->init();

326
src/core/renderer_mtl/texture_decoder.cpp Normal file
View file

@ -0,0 +1,326 @@
#include "renderer_mtl/texture_decoder.hpp"
#include <array>
#include <string>
#include "math_util.hpp"
#include "colour.hpp"
using namespace Helpers;
// u and v are the UVs of the relevant texel
// Texture data is stored interleaved in Morton order, ie in a Z - order curve as shown here
// https://en.wikipedia.org/wiki/Z-order_curve
// Textures are split into 8x8 tiles.This function returns the in - tile offset depending on the u & v of the texel
// The in - tile offset is the sum of 2 offsets, one depending on the value of u % 8 and the other on the value of y % 8
// As documented in this picture https ://en.wikipedia.org/wiki/File:Moser%E2%80%93de_Bruijn_addition.svg
u32 mortonInterleave(u32 u, u32 v) {
static constexpr u32 xOffsets[] = {0, 1, 4, 5, 16, 17, 20, 21};
static constexpr u32 yOffsets[] = {0, 2, 8, 10, 32, 34, 40, 42};
return xOffsets[u & 7] + yOffsets[v & 7];
}
// Get the byte offset of texel (u, v) in the texture
u32 getSwizzledOffset(u32 u, u32 v, u32 width, u32 bytesPerPixel) {
u32 offset = ((u & ~7) * 8) + ((v & ~7) * width); // Offset of the 8x8 tile the texel belongs to
offset += mortonInterleave(u, v); // Add the in-tile offset of the texel
return offset * bytesPerPixel;
}
// Same as the above code except we need to divide by 2 because 4 bits is smaller than a byte
u32 getSwizzledOffset_4bpp(u32 u, u32 v, u32 width) {
u32 offset = ((u & ~7) * 8) + ((v & ~7) * width); // Offset of the 8x8 tile the texel belongs to
offset += mortonInterleave(u, v); // Add the in-tile offset of the texel
return offset / 2;
}
void decodeTexelABGR8ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
const u32 offset = getSwizzledOffset(u, v, size.u(), 4);
const u8 alpha = inData[offset];
const u8 b = inData[offset + 1];
const u8 g = inData[offset + 2];
const u8 r = inData[offset + 3];
outData.push_back(r);
outData.push_back(g);
outData.push_back(b);
outData.push_back(alpha);
}
void decodeTexelBGR8ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
const u32 offset = getSwizzledOffset(u, v, size.u(), 3);
const u8 b = inData[offset];
const u8 g = inData[offset + 1];
const u8 r = inData[offset + 2];
outData.push_back(r);
outData.push_back(g);
outData.push_back(b);
outData.push_back(0xff);
}
void decodeTexelA1BGR5ToBGR5A1(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
const u32 offset = getSwizzledOffset(u, v, size.u(), 2);
const u16 texel = u16(inData[offset]) | (u16(inData[offset + 1]) << 8);
u8 alpha = getBit<0>(texel);
u8 b = getBits<1, 5, u8>(texel);
u8 g = getBits<6, 5, u8>(texel);
u8 r = getBits<11, 5, u8>(texel);
u16 outTexel = (alpha << 15) | (r << 10) | (g << 5) | b;
outData.push_back(outTexel & 0xff);
outData.push_back((outTexel >> 8) & 0xff);
}
void decodeTexelA1BGR5ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
const u32 offset = getSwizzledOffset(u, v, size.u(), 2);
const u16 texel = u16(inData[offset]) | (u16(inData[offset + 1]) << 8);
u8 alpha = getBit<0>(texel) ? 0xff : 0;
u8 b = Colour::convert5To8Bit(getBits<1, 5, u8>(texel));
u8 g = Colour::convert5To8Bit(getBits<6, 5, u8>(texel));
u8 r = Colour::convert5To8Bit(getBits<11, 5, u8>(texel));
outData.push_back(r);
outData.push_back(g);
outData.push_back(b);
outData.push_back(alpha);
}
void decodeTexelB5G6R5ToB5G6R5(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
const u32 offset = getSwizzledOffset(u, v, size.u(), 2);
const u16 texel = u16(inData[offset]) | (u16(inData[offset + 1]) << 8);
outData.push_back(texel & 0xff);
outData.push_back((texel >> 8) & 0xff);
}
void decodeTexelB5G6R5ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
const u32 offset = getSwizzledOffset(u, v, size.u(), 2);
const u16 texel = u16(inData[offset]) | (u16(inData[offset + 1]) << 8);
const u8 b = Colour::convert5To8Bit(getBits<0, 5, u8>(texel));
const u8 g = Colour::convert6To8Bit(getBits<5, 6, u8>(texel));
const u8 r = Colour::convert5To8Bit(getBits<11, 5, u8>(texel));
outData.push_back(r);
outData.push_back(g);
outData.push_back(b);
outData.push_back(0xff);
}
void decodeTexelABGR4ToABGR4(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
u32 offset = getSwizzledOffset(u, v, size.u(), 2);
u16 texel = u16(inData[offset]) | (u16(inData[offset + 1]) << 8);
u8 alpha = getBits<0, 4, u8>(texel);
u8 b = getBits<4, 4, u8>(texel);
u8 g = getBits<8, 4, u8>(texel);
u8 r = getBits<12, 4, u8>(texel);
outData.push_back((b << 4) | alpha);
outData.push_back((r << 4) | g);
}
void decodeTexelABGR4ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
u32 offset = getSwizzledOffset(u, v, size.u(), 2);
u16 texel = u16(inData[offset]) | (u16(inData[offset + 1]) << 8);
u8 alpha = Colour::convert4To8Bit(getBits<0, 4, u8>(texel));
u8 b = Colour::convert4To8Bit(getBits<4, 4, u8>(texel));
u8 g = Colour::convert4To8Bit(getBits<8, 4, u8>(texel));
u8 r = Colour::convert4To8Bit(getBits<12, 4, u8>(texel));
outData.push_back(r);
outData.push_back(g);
outData.push_back(b);
outData.push_back(alpha);
}
void decodeTexelAI8ToRG8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
u32 offset = getSwizzledOffset(u, v, size.u(), 2);
// Same as I8 except each pixel gets its own alpha value too
const u8 alpha = inData[offset];
const u8 intensity = inData[offset + 1];
outData.push_back(intensity);
outData.push_back(alpha);
}
void decodeTexelGR8ToRG8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
u32 offset = getSwizzledOffset(u, v, size.u(), 2);
constexpr u8 b = 0;
const u8 g = inData[offset];
const u8 r = inData[offset + 1];
outData.push_back(r);
outData.push_back(g);
}
void decodeTexelI8ToR8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
u32 offset = getSwizzledOffset(u, v, size.u(), 1);
const u8 intensity = inData[offset];
outData.push_back(intensity);
}
void decodeTexelA8ToA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
u32 offset = getSwizzledOffset(u, v, size.u(), 1);
const u8 alpha = inData[offset];
outData.push_back(alpha);
}
void decodeTexelAI4ToRG8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
const u32 offset = getSwizzledOffset(u, v, size.u(), 1);
const u8 texel = inData[offset];
const u8 alpha = Colour::convert4To8Bit(texel & 0xf);
const u8 intensity = Colour::convert4To8Bit(texel >> 4);
outData.push_back(intensity);
outData.push_back(alpha);
}
void decodeTexelI4ToR8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
u32 offset = getSwizzledOffset_4bpp(u, v, size.u());
// For odd U coordinates, grab the top 4 bits, and the low 4 bits for even coordinates
u8 intensity = inData[offset] >> ((u % 2) ? 4 : 0);
intensity = Colour::convert4To8Bit(getBits<0, 4>(intensity));
outData.push_back(intensity);
}
void decodeTexelA4ToA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
const u32 offset = getSwizzledOffset_4bpp(u, v, size.u());
// For odd U coordinates, grab the top 4 bits, and the low 4 bits for even coordinates
u8 alpha = inData[offset] >> ((u % 2) ? 4 : 0);
alpha = Colour::convert4To8Bit(getBits<0, 4>(alpha));
outData.push_back(alpha);
}
static constexpr u32 signExtend3To32(u32 val) {
return (u32)(s32(val) << 29 >> 29);
}
void decodeETC(u32 u, u32 v, u64 colourData, u32 alpha, std::vector<u8>& outData) {
static constexpr u32 modifiers[8][2] = {
{2, 8}, {5, 17}, {9, 29}, {13, 42}, {18, 60}, {24, 80}, {33, 106}, {47, 183},
};
// Parse colour data for 4x4 block
const u32 subindices = getBits<0, 16, u32>(colourData);
const u32 negationFlags = getBits<16, 16, u32>(colourData);
const bool flip = getBit<32>(colourData);
const bool diffMode = getBit<33>(colourData);
// Note: index1 is indeed stored on the higher bits, with index2 in the lower bits
const u32 tableIndex1 = getBits<37, 3, u32>(colourData);
const u32 tableIndex2 = getBits<34, 3, u32>(colourData);
const u32 texelIndex = u * 4 + v; // Index of the texel in the block
if (flip) std::swap(u, v);
s32 r, g, b;
if (diffMode) {
r = getBits<59, 5, s32>(colourData);
g = getBits<51, 5, s32>(colourData);
b = getBits<43, 5, s32>(colourData);
if (u >= 2) {
r += signExtend3To32(getBits<56, 3, u32>(colourData));
g += signExtend3To32(getBits<48, 3, u32>(colourData));
b += signExtend3To32(getBits<40, 3, u32>(colourData));
}
// Expand from 5 to 8 bits per channel
r = Colour::convert5To8Bit(r);
g = Colour::convert5To8Bit(g);
b = Colour::convert5To8Bit(b);
} else {
if (u < 2) {
r = getBits<60, 4, s32>(colourData);
g = getBits<52, 4, s32>(colourData);
b = getBits<44, 4, s32>(colourData);
} else {
r = getBits<56, 4, s32>(colourData);
g = getBits<48, 4, s32>(colourData);
b = getBits<40, 4, s32>(colourData);
}
// Expand from 4 to 8 bits per channel
r = Colour::convert4To8Bit(r);
g = Colour::convert4To8Bit(g);
b = Colour::convert4To8Bit(b);
}
const u32 index = (u < 2) ? tableIndex1 : tableIndex2;
s32 modifier = modifiers[index][(subindices >> texelIndex) & 1];
if (((negationFlags >> texelIndex) & 1) != 0) {
modifier = -modifier;
}
r = std::clamp(r + modifier, 0, 255);
g = std::clamp(g + modifier, 0, 255);
b = std::clamp(b + modifier, 0, 255);
outData.push_back(r);
outData.push_back(g);
outData.push_back(b);
outData.push_back(alpha);
}
template <bool hasAlpha>
void getTexelETC(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
// Pixel offset of the 8x8 tile based on u, v and the width of the texture
u32 offs = ((u & ~7) * 8) + ((v & ~7) * size.u());
if (!hasAlpha) {
offs >>= 1;
}
// In-tile offsets for u/v
u &= 7;
v &= 7;
// ETC1(A4) also subdivide the 8x8 tile to 4 4x4 tiles
// Each tile is 8 bytes for ETC1, but since ETC1A4 has 4 alpha bits per pixel, that becomes 16 bytes
const u32 subTileSize = hasAlpha ? 16 : 8;
const u32 subTileIndex = (u / 4) + 2 * (v / 4); // Which of the 4 subtiles is this texel in?
// In-subtile offsets for u/v
u &= 3;
v &= 3;
offs += subTileSize * subTileIndex;
u32 alpha;
const u64* ptr = reinterpret_cast<const u64*>(inData.data() + offs); // Cast to u64*
if (hasAlpha) {
// First 64 bits of the 4x4 subtile are alpha data
const u64 alphaData = *ptr++;
alpha = Colour::convert4To8Bit((alphaData >> (4 * (u * 4 + v))) & 0xf);
} else {
alpha = 0xff; // ETC1 without alpha uses ff for every pixel
}
// Next 64 bits of the subtile are colour data
u64 colourData = *ptr;
decodeETC(u, v, colourData, alpha, outData);
}
void decodeTexelETC1ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
getTexelETC<false>(size, u, v, inData, outData);
}
void decodeTexelETC1A4ToRGBA8(OpenGL::uvec2 size, u32 u, u32 v, std::span<const u8> inData, std::vector<u8>& outData) {
getTexelETC<true>(size, u, v, inData, outData);
}

2
third_party/LuaJIT vendored

@ -1 +1 @@
Subproject commit 8bf7686d820f868eae1a522c481fee09c18c90b9 Subproject commit 41edf0959b9504d36dd85f5f16893c004ea7d7ba

2
third_party/oaknut vendored

@ -1 +1 @@
Subproject commit 790374d7e66257b1f8ed89d798e5dcfb5363af05 Subproject commit 94c726ce0338b054eb8cb5ea91de8fe6c19f4392