summaryrefslogtreecommitdiffstats
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/engines/shader_bytecode.h20
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp5
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp10
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.h15
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp14
-rw-r--r--src/video_core/textures/decoders.cpp3
6 files changed, 56 insertions, 11 deletions
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 3b70efeec..2bc1782ad 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -194,6 +194,13 @@ enum class UniformType : u64 {
Double = 5,
};
+enum class IMinMaxExchange : u64 {
+ None = 0,
+ XLo = 1,
+ XMed = 2,
+ XHi = 3,
+};
+
union Instruction {
Instruction& operator=(const Instruction& instr) {
value = instr.value;
@@ -279,6 +286,13 @@ union Instruction {
} alu_integer;
union {
+ BitField<39, 3, u64> pred;
+ BitField<42, 1, u64> negate_pred;
+ BitField<43, 2, IMinMaxExchange> exchange;
+ BitField<48, 1, u64> is_signed;
+ } imnmx;
+
+ union {
BitField<54, 1, u64> saturate;
BitField<56, 1, u64> negate_a;
} iadd32i;
@@ -700,9 +714,9 @@ private:
INST("0100110001100---", Id::FMNMX_C, Type::Arithmetic, "FMNMX_C"),
INST("0101110001100---", Id::FMNMX_R, Type::Arithmetic, "FMNMX_R"),
INST("0011100-01100---", Id::FMNMX_IMM, Type::Arithmetic, "FMNMX_IMM"),
- INST("0100110000100---", Id::IMNMX_C, Type::Arithmetic, "FMNMX_IMM"),
- INST("0101110000100---", Id::IMNMX_R, Type::Arithmetic, "FMNMX_IMM"),
- INST("0011100-00100---", Id::IMNMX_IMM, Type::Arithmetic, "FMNMX_IMM"),
+ INST("0100110000100---", Id::IMNMX_C, Type::ArithmeticInteger, "IMNMX_C"),
+ INST("0101110000100---", Id::IMNMX_R, Type::ArithmeticInteger, "IMNMX_R"),
+ INST("0011100-00100---", Id::IMNMX_IMM, Type::ArithmeticInteger, "IMNMX_IMM"),
INST("0100110000000---", Id::BFE_C, Type::Bfe, "BFE_C"),
INST("0101110000000---", Id::BFE_R, Type::Bfe, "BFE_R"),
INST("0011100-00000---", Id::BFE_IMM, Type::Bfe, "BFE_IMM"),
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index bacb389e1..ea138d402 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -775,10 +775,13 @@ void RasterizerOpenGL::SyncCullMode() {
state.cull.front_face = MaxwellToGL::FrontFace(regs.cull.front_face);
state.cull.mode = MaxwellToGL::CullFace(regs.cull.cull_face);
+ const bool flip_triangles{regs.screen_y_control.triangle_rast_flip == 0 ||
+ regs.viewport_transform[0].scale_y < 0.0f};
+
// If the GPU is configured to flip the rasterized triangles, then we need to flip the
// notion of front and back. Note: We flip the triangles when the value of the register is 0
// because OpenGL already does it for us.
- if (regs.screen_y_control.triangle_rast_flip == 0) {
+ if (flip_triangles) {
if (state.cull.front_face == GL_CCW)
state.cull.front_face = GL_CW;
else if (state.cull.front_face == GL_CW)
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 57d7763ff..323ff7408 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -102,6 +102,8 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
{GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXT45
{GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm, true}, // DXN1
+ {GL_COMPRESSED_RGBA_BPTC_UNORM_ARB, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
+ true}, // BC7U
{GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4
// DepthStencil formats
@@ -191,8 +193,9 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<true, PixelFormat::R11FG11FB10F>, MortonCopy<true, PixelFormat::RGBA32UI>,
MortonCopy<true, PixelFormat::DXT1>, MortonCopy<true, PixelFormat::DXT23>,
MortonCopy<true, PixelFormat::DXT45>, MortonCopy<true, PixelFormat::DXN1>,
- MortonCopy<true, PixelFormat::ASTC_2D_4X4>, MortonCopy<true, PixelFormat::Z24S8>,
- MortonCopy<true, PixelFormat::S8Z24>, MortonCopy<true, PixelFormat::Z32F>,
+ MortonCopy<true, PixelFormat::BC7U>, MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
+ MortonCopy<true, PixelFormat::Z24S8>, MortonCopy<true, PixelFormat::S8Z24>,
+ MortonCopy<true, PixelFormat::Z32F>,
};
static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
@@ -206,7 +209,8 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<false, PixelFormat::RGBA16F>,
MortonCopy<false, PixelFormat::R11FG11FB10F>,
MortonCopy<false, PixelFormat::RGBA32UI>,
- // TODO(Subv): Swizzling the DXT1/DXT23/DXT45/DXN1 formats is not yet supported
+ // TODO(Subv): Swizzling the DXT1/DXT23/DXT45/DXN1/BC7U formats is not yet supported
+ nullptr,
nullptr,
nullptr,
nullptr,
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index b4d7f8ebe..1bedae992 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -35,14 +35,15 @@ struct SurfaceParams {
DXT23 = 9,
DXT45 = 10,
DXN1 = 11, // This is also known as BC4
- ASTC_2D_4X4 = 12,
+ BC7U = 12,
+ ASTC_2D_4X4 = 13,
MaxColorFormat,
// DepthStencil formats
- Z24S8 = 13,
- S8Z24 = 14,
- Z32F = 15,
+ Z24S8 = 14,
+ S8Z24 = 15,
+ Z32F = 16,
MaxDepthStencilFormat,
@@ -92,6 +93,7 @@ struct SurfaceParams {
4, // DXT23
4, // DXT45
4, // DXN1
+ 4, // BC7U
4, // ASTC_2D_4X4
1, // Z24S8
1, // S8Z24
@@ -119,6 +121,7 @@ struct SurfaceParams {
128, // DXT23
128, // DXT45
64, // DXN1
+ 128, // BC7U
32, // ASTC_2D_4X4
32, // Z24S8
32, // S8Z24
@@ -192,6 +195,8 @@ struct SurfaceParams {
return PixelFormat::DXT45;
case Tegra::Texture::TextureFormat::DXN1:
return PixelFormat::DXN1;
+ case Tegra::Texture::TextureFormat::BC7U:
+ return PixelFormat::BC7U;
case Tegra::Texture::TextureFormat::ASTC_2D_4X4:
return PixelFormat::ASTC_2D_4X4;
default:
@@ -227,6 +232,8 @@ struct SurfaceParams {
return Tegra::Texture::TextureFormat::DXT45;
case PixelFormat::DXN1:
return Tegra::Texture::TextureFormat::DXN1;
+ case PixelFormat::BC7U:
+ return Tegra::Texture::TextureFormat::BC7U;
case PixelFormat::ASTC_2D_4X4:
return Tegra::Texture::TextureFormat::ASTC_2D_4X4;
default:
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index e817aca5a..5914077e8 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -1127,6 +1127,20 @@ private:
WriteLogicOperation(instr.gpr0, instr.alu.lop.operation, op_a, op_b);
break;
}
+ case OpCode::Id::IMNMX_C:
+ case OpCode::Id::IMNMX_R:
+ case OpCode::Id::IMNMX_IMM: {
+ ASSERT_MSG(instr.imnmx.exchange == Tegra::Shader::IMinMaxExchange::None,
+ "Unimplemented");
+ std::string condition =
+ GetPredicateCondition(instr.imnmx.pred, instr.imnmx.negate_pred != 0);
+ std::string parameters = op_a + ',' + op_b;
+ regs.SetRegisterToInteger(instr.gpr0, instr.imnmx.is_signed, 0,
+ '(' + condition + ") ? min(" + parameters + ") : max(" +
+ parameters + ')',
+ 1, 1);
+ break;
+ }
default: {
LOG_CRITICAL(HW_GPU, "Unhandled ArithmeticInteger instruction: {}",
opcode->GetName());
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp
index d5ab4e4f9..b3937b2fe 100644
--- a/src/video_core/textures/decoders.cpp
+++ b/src/video_core/textures/decoders.cpp
@@ -52,6 +52,7 @@ u32 BytesPerPixel(TextureFormat format) {
return 8;
case TextureFormat::DXT23:
case TextureFormat::DXT45:
+ case TextureFormat::BC7U:
// In this case a 'pixel' actually refers to a 4x4 tile.
return 16;
case TextureFormat::ASTC_2D_4X4:
@@ -98,6 +99,7 @@ std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width,
case TextureFormat::DXT23:
case TextureFormat::DXT45:
case TextureFormat::DXN1:
+ case TextureFormat::BC7U:
// In the DXT and DXN formats, each 4x4 tile is swizzled instead of just individual pixel
// values.
CopySwizzledData(width / 4, height / 4, bytes_per_pixel, bytes_per_pixel, data,
@@ -155,6 +157,7 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat
case TextureFormat::DXT23:
case TextureFormat::DXT45:
case TextureFormat::DXN1:
+ case TextureFormat::BC7U:
case TextureFormat::ASTC_2D_4X4:
case TextureFormat::A8R8G8B8:
case TextureFormat::A2B10G10R10:
generated by cgit v1.2.3 (git 2.25.1) at 2024-09-09 06:36:28 +0200