summaryrefslogtreecommitdiffstats
path: root/src/video_core/renderer_opengl
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core/renderer_opengl')
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp9
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.h18
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp55
3 files changed, 70 insertions, 12 deletions
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 3f385484f..86682d7cb 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -167,6 +167,7 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
{GL_RG8, GL_RG, GL_BYTE, ComponentType::SNorm, false}, // RG8S
{GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // RG32UI
{GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // R32UI
+ {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X8
// Depth formats
{GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT, ComponentType::Float, false}, // Z32F
@@ -213,6 +214,7 @@ static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType
static bool IsPixelFormatASTC(PixelFormat format) {
switch (format) {
case PixelFormat::ASTC_2D_4X4:
+ case PixelFormat::ASTC_2D_8X8:
return true;
default:
return false;
@@ -223,6 +225,8 @@ static std::pair<u32, u32> GetASTCBlockSize(PixelFormat format) {
switch (format) {
case PixelFormat::ASTC_2D_4X4:
return {4, 4};
+ case PixelFormat::ASTC_2D_8X8:
+ return {8, 8};
default:
LOG_CRITICAL(HW_GPU, "Unhandled format: {}", static_cast<u32>(format));
UNREACHABLE();
@@ -327,6 +331,7 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, std::size_t, VAddr),
MortonCopy<true, PixelFormat::RG8S>,
MortonCopy<true, PixelFormat::RG32UI>,
MortonCopy<true, PixelFormat::R32UI>,
+ MortonCopy<true, PixelFormat::ASTC_2D_8X8>,
MortonCopy<true, PixelFormat::Z32F>,
MortonCopy<true, PixelFormat::Z16>,
MortonCopy<true, PixelFormat::Z24S8>,
@@ -386,6 +391,7 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, std::size_t, VAddr),
MortonCopy<false, PixelFormat::RG8S>,
MortonCopy<false, PixelFormat::RG32UI>,
MortonCopy<false, PixelFormat::R32UI>,
+ nullptr,
MortonCopy<false, PixelFormat::Z32F>,
MortonCopy<false, PixelFormat::Z16>,
MortonCopy<false, PixelFormat::Z24S8>,
@@ -544,7 +550,8 @@ static void ConvertG8R8ToR8G8(std::vector<u8>& data, u32 width, u32 height) {
static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
u32 width, u32 height) {
switch (pixel_format) {
- case PixelFormat::ASTC_2D_4X4: {
+ case PixelFormat::ASTC_2D_4X4:
+ case PixelFormat::ASTC_2D_8X8: {
// Convert ASTC pixel formats to RGBA8, as most desktop GPUs do not support ASTC.
u32 block_width{};
u32 block_height{};
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index aafac9a20..d7a4bc37f 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -70,19 +70,20 @@ struct SurfaceParams {
RG8S = 42,
RG32UI = 43,
R32UI = 44,
+ ASTC_2D_8X8 = 45,
MaxColorFormat,
// Depth formats
- Z32F = 45,
- Z16 = 46,
+ Z32F = 46,
+ Z16 = 47,
MaxDepthFormat,
// DepthStencil formats
- Z24S8 = 47,
- S8Z24 = 48,
- Z32FS8 = 49,
+ Z24S8 = 48,
+ S8Z24 = 49,
+ Z32FS8 = 50,
MaxDepthStencilFormat,
@@ -192,6 +193,7 @@ struct SurfaceParams {
1, // RG8S
1, // RG32UI
1, // R32UI
+ 4, // ASTC_2D_8X8
1, // Z32F
1, // Z16
1, // Z24S8
@@ -253,6 +255,7 @@ struct SurfaceParams {
16, // RG8S
64, // RG32UI
32, // R32UI
+ 16, // ASTC_2D_8X8
32, // Z32F
16, // Z16
32, // Z24S8
@@ -316,6 +319,8 @@ struct SurfaceParams {
return PixelFormat::R11FG11FB10F;
case Tegra::RenderTargetFormat::B5G6R5_UNORM:
return PixelFormat::B5G6R5U;
+ case Tegra::RenderTargetFormat::BGR5A1_UNORM:
+ return PixelFormat::A1B5G5R5U;
case Tegra::RenderTargetFormat::RGBA32_UINT:
return PixelFormat::RGBA32UI;
case Tegra::RenderTargetFormat::R8_UNORM:
@@ -522,6 +527,8 @@ struct SurfaceParams {
return PixelFormat::BC6H_SF16;
case Tegra::Texture::TextureFormat::ASTC_2D_4X4:
return PixelFormat::ASTC_2D_4X4;
+ case Tegra::Texture::TextureFormat::ASTC_2D_8X8:
+ return PixelFormat::ASTC_2D_8X8;
case Tegra::Texture::TextureFormat::R16_G16:
switch (component_type) {
case Tegra::Texture::ComponentType::FLOAT:
@@ -576,6 +583,7 @@ struct SurfaceParams {
case Tegra::RenderTargetFormat::RG16_UNORM:
case Tegra::RenderTargetFormat::R16_UNORM:
case Tegra::RenderTargetFormat::B5G6R5_UNORM:
+ case Tegra::RenderTargetFormat::BGR5A1_UNORM:
case Tegra::RenderTargetFormat::RG8_UNORM:
case Tegra::RenderTargetFormat::RGBA16_UNORM:
return ComponentType::UNorm;
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index d58a65935..15f628acd 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -1772,13 +1772,34 @@ private:
case OpCode::Type::Memory: {
switch (opcode->GetId()) {
case OpCode::Id::LD_A: {
- ASSERT_MSG(instr.attribute.fmt20.size == 0, "untested");
// Note: Shouldn't this be interp mode flat? As in no interpolation made.
+ ASSERT_MSG(instr.gpr8.Value() == Register::ZeroIndex,
+ "Indirect attribute loads are not supported");
+ ASSERT_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) == 0,
+ "Unaligned attribute loads are not supported");
Tegra::Shader::IpaMode input_mode{Tegra::Shader::IpaInterpMode::Perspective,
Tegra::Shader::IpaSampleMode::Default};
- regs.SetRegisterToInputAttibute(instr.gpr0, instr.attribute.fmt20.element,
- instr.attribute.fmt20.index, input_mode);
+
+ u32 next_element = instr.attribute.fmt20.element;
+ u32 next_index = static_cast<u32>(instr.attribute.fmt20.index.Value());
+
+ const auto LoadNextElement = [&](u32 reg_offset) {
+ regs.SetRegisterToInputAttibute(instr.gpr0.Value() + reg_offset, next_element,
+ static_cast<Attribute::Index>(next_index),
+ input_mode);
+
+ // Load the next attribute element into the following register. If the element
+ // to load goes beyond the vec4 size, load the first element of the next
+ // attribute.
+ next_element = (next_element + 1) % 4;
+ next_index = next_index + (next_element == 0 ? 1 : 0);
+ };
+
+ const u32 num_words = static_cast<u32>(instr.attribute.fmt20.size.Value()) + 1;
+ for (u32 reg_offset = 0; reg_offset < num_words; ++reg_offset) {
+ LoadNextElement(reg_offset);
+ }
break;
}
case OpCode::Id::LD_C: {
@@ -1820,9 +1841,31 @@ private:
break;
}
case OpCode::Id::ST_A: {
- ASSERT_MSG(instr.attribute.fmt20.size == 0, "untested");
- regs.SetOutputAttributeToRegister(instr.attribute.fmt20.index,
- instr.attribute.fmt20.element, instr.gpr0);
+ ASSERT_MSG(instr.gpr8.Value() == Register::ZeroIndex,
+ "Indirect attribute loads are not supported");
+ ASSERT_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) == 0,
+ "Unaligned attribute loads are not supported");
+
+ u32 next_element = instr.attribute.fmt20.element;
+ u32 next_index = static_cast<u32>(instr.attribute.fmt20.index.Value());
+
+ const auto StoreNextElement = [&](u32 reg_offset) {
+ regs.SetOutputAttributeToRegister(static_cast<Attribute::Index>(next_index),
+ next_element,
+ instr.gpr0.Value() + reg_offset);
+
+ // Load the next attribute element into the following register. If the element
+ // to load goes beyond the vec4 size, load the first element of the next
+ // attribute.
+ next_element = (next_element + 1) % 4;
+ next_index = next_index + (next_element == 0 ? 1 : 0);
+ };
+
+ const u32 num_words = static_cast<u32>(instr.attribute.fmt20.size.Value()) + 1;
+ for (u32 reg_offset = 0; reg_offset < num_words; ++reg_offset) {
+ StoreNextElement(reg_offset);
+ }
+
break;
}
case OpCode::Id::TEX: {
generated by cgit v1.2.3 (git 2.25.1) at 2024-08-03 00:10:56 +0200