summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--src/video_core/engines/shader_bytecode.h4
-rw-r--r--src/video_core/gpu.h1
-rw-r--r--src/video_core/memory_manager.cpp4
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp6
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp52
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.h21
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp44
-rw-r--r--src/video_core/textures/decoders.cpp4
8 files changed, 85 insertions, 51 deletions
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index cb4db0679..86fd64979 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -142,6 +142,7 @@ enum class PredCondition : u64 {
GreaterThan = 4,
NotEqual = 5,
GreaterEqual = 6,
+ NotEqualWithNan = 13,
// TODO(Subv): Other condition types
};
@@ -165,7 +166,6 @@ enum class SubOp : u64 {
Lg2 = 0x3,
Rcp = 0x4,
Rsq = 0x5,
- Min = 0x8,
};
enum class F2iRoundingOp : u64 {
@@ -209,7 +209,7 @@ union Instruction {
} pred;
BitField<19, 1, u64> negate_pred;
BitField<20, 8, Register> gpr20;
- BitField<20, 7, SubOp> sub_op;
+ BitField<20, 4, SubOp> sub_op;
BitField<28, 8, Register> gpr28;
BitField<39, 8, Register> gpr39;
BitField<48, 16, u64> opcode;
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 7b4e9b842..d0a4ac267 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -16,6 +16,7 @@ namespace Tegra {
enum class RenderTargetFormat : u32 {
NONE = 0x0,
RGBA32_FLOAT = 0xC0,
+ RGBA32_UINT = 0xC2,
RGBA16_FLOAT = 0xCA,
RGB10_A2_UNORM = 0xD1,
RGBA8_UNORM = 0xD5,
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index 5cefce9fc..2f814a184 100644
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -100,9 +100,9 @@ boost::optional<GPUVAddr> MemoryManager::FindFreeBlock(u64 size, u64 align) {
boost::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) {
VAddr base_addr = PageSlot(gpu_addr);
- ASSERT(base_addr != static_cast<u64>(PageStatus::Unmapped));
- if (base_addr == static_cast<u64>(PageStatus::Allocated)) {
+ if (base_addr == static_cast<u64>(PageStatus::Allocated) ||
+ base_addr == static_cast<u64>(PageStatus::Unmapped)) {
return {};
}
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 62ee45a36..45560fbee 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -636,7 +636,11 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, GLuint program,
glProgramUniform1i(program, uniform, current_bindpoint);
const auto texture = maxwell3d.GetStageTexture(entry.GetStage(), entry.GetOffset());
- ASSERT(texture.enabled);
+
+ if (!texture.enabled) {
+ state.texture_units[current_bindpoint].texture_2d = 0;
+ continue;
+ }
texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
Surface surface = res_cache.GetTextureSurface(texture);
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 63f5999ea..9410ddb4e 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -27,6 +27,7 @@ struct FormatTuple {
GLint internal_format;
GLenum format;
GLenum type;
+ ComponentType component_type;
bool compressed;
};
@@ -65,29 +66,33 @@ struct FormatTuple {
}
static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_format_tuples = {{
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false}, // ABGR8
- {GL_RGB, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, false}, // B5G6R5
- {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, false}, // A2B10G10R10
- {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, false}, // A1B5G5R5
- {GL_R8, GL_RED, GL_UNSIGNED_BYTE, false}, // R8
- {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, false}, // RGBA16F
- {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, false}, // R11FG11FB10F
- {GL_COMPRESSED_RGB_S3TC_DXT1_EXT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT1
- {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT23
- {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT45
- {GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, true}, // DXN1
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_4X4
+ {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, ComponentType::UNorm, false}, // ABGR8
+ {GL_RGB, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, ComponentType::UNorm, false}, // B5G6R5
+ {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, ComponentType::UNorm,
+ false}, // A2B10G10R10
+ {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, ComponentType::UNorm, false}, // A1B5G5R5
+ {GL_R8, GL_RED, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // R8
+ {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, ComponentType::Float, false}, // RGBA16F
+ {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, ComponentType::Float,
+ false}, // R11FG11FB10F
+ {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // RGBA32UI
+ {GL_COMPRESSED_RGB_S3TC_DXT1_EXT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
+ true}, // DXT1
+ {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
+ true}, // DXT23
+ {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_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4
}};
static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) {
const SurfaceType type = SurfaceParams::GetFormatType(pixel_format);
if (type == SurfaceType::ColorTexture) {
ASSERT(static_cast<size_t>(pixel_format) < tex_format_tuples.size());
- // For now only UNORM components are supported, or either R11FG11FB10F or RGBA16F which
- // are type FLOAT
- ASSERT(component_type == ComponentType::UNorm || pixel_format == PixelFormat::RGBA16F ||
- pixel_format == PixelFormat::R11FG11FB10F);
- return tex_format_tuples[static_cast<unsigned int>(pixel_format)];
+ auto& format = tex_format_tuples[static_cast<unsigned int>(pixel_format)];
+ ASSERT(component_type == format.component_type);
+ return format;
} else if (type == SurfaceType::Depth || type == SurfaceType::DepthStencil) {
// TODO(Subv): Implement depth formats
ASSERT_MSG(false, "Unimplemented");
@@ -166,9 +171,10 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<true, PixelFormat::ABGR8>, MortonCopy<true, PixelFormat::B5G6R5>,
MortonCopy<true, PixelFormat::A2B10G10R10>, MortonCopy<true, PixelFormat::A1B5G5R5>,
MortonCopy<true, PixelFormat::R8>, MortonCopy<true, PixelFormat::RGBA16F>,
- MortonCopy<true, PixelFormat::R11FG11FB10F>, 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::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>,
};
static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
@@ -181,6 +187,7 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<false, PixelFormat::R8>,
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
nullptr,
nullptr,
@@ -454,6 +461,11 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params) {
return {};
}
+ const auto& gpu = Core::System::GetInstance().GPU();
+ // Don't try to create any entries in the cache if the address of the texture is invalid.
+ if (gpu.memory_manager->GpuToCpuAddress(params.addr) == boost::none)
+ return {};
+
// Check for an exact match in existing surfaces
const auto& surface_key{SurfaceKey::Create(params)};
const auto& search{surface_cache.find(surface_key)};
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index 85e7c8888..99be250b4 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -30,11 +30,12 @@ struct SurfaceParams {
R8 = 4,
RGBA16F = 5,
R11FG11FB10F = 6,
- DXT1 = 7,
- DXT23 = 8,
- DXT45 = 9,
- DXN1 = 10, // This is also known as BC4
- ASTC_2D_4X4 = 11,
+ RGBA32UI = 7,
+ DXT1 = 8,
+ DXT23 = 9,
+ DXT45 = 10,
+ DXN1 = 11, // This is also known as BC4
+ ASTC_2D_4X4 = 12,
Max,
Invalid = 255,
@@ -77,6 +78,7 @@ struct SurfaceParams {
1, // R8
1, // RGBA16F
1, // R11FG11FB10F
+ 1, // RGBA32UI
4, // DXT1
4, // DXT23
4, // DXT45
@@ -100,6 +102,7 @@ struct SurfaceParams {
8, // R8
64, // RGBA16F
32, // R11FG11FB10F
+ 128, // RGBA32UI
64, // DXT1
128, // DXT23
128, // DXT45
@@ -125,6 +128,8 @@ struct SurfaceParams {
return PixelFormat::RGBA16F;
case Tegra::RenderTargetFormat::R11G11B10_FLOAT:
return PixelFormat::R11FG11FB10F;
+ case Tegra::RenderTargetFormat::RGBA32_UINT:
+ return PixelFormat::RGBA32UI;
default:
NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
UNREACHABLE();
@@ -148,6 +153,8 @@ struct SurfaceParams {
return PixelFormat::RGBA16F;
case Tegra::Texture::TextureFormat::BF10GF11RF11:
return PixelFormat::R11FG11FB10F;
+ case Tegra::Texture::TextureFormat::R32_G32_B32_A32:
+ return PixelFormat::RGBA32UI;
case Tegra::Texture::TextureFormat::DXT1:
return PixelFormat::DXT1;
case Tegra::Texture::TextureFormat::DXT23:
@@ -181,6 +188,8 @@ struct SurfaceParams {
return Tegra::Texture::TextureFormat::R16_G16_B16_A16;
case PixelFormat::R11FG11FB10F:
return Tegra::Texture::TextureFormat::BF10GF11RF11;
+ case PixelFormat::RGBA32UI:
+ return Tegra::Texture::TextureFormat::R32_G32_B32_A32;
case PixelFormat::DXT1:
return Tegra::Texture::TextureFormat::DXT1;
case PixelFormat::DXT23:
@@ -217,6 +226,8 @@ struct SurfaceParams {
case Tegra::RenderTargetFormat::RGBA16_FLOAT:
case Tegra::RenderTargetFormat::R11G11B10_FLOAT:
return ComponentType::Float;
+ case Tegra::RenderTargetFormat::RGBA32_UINT:
+ return ComponentType::UInt;
default:
NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
UNREACHABLE();
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 46eaad021..bbccf0bfd 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -719,21 +719,31 @@ private:
/**
* Returns the comparison string to use to compare two values in the 'set' family of
* instructions.
- * @params condition The condition used in the 'set'-family instruction.
+ * @param condition The condition used in the 'set'-family instruction.
+ * @param op_a First operand to use for the comparison.
+ * @param op_b Second operand to use for the comparison.
* @returns String corresponding to the GLSL operator that matches the desired comparison.
*/
- std::string GetPredicateComparison(Tegra::Shader::PredCondition condition) const {
+ std::string GetPredicateComparison(Tegra::Shader::PredCondition condition,
+ const std::string& op_a, const std::string& op_b) const {
using Tegra::Shader::PredCondition;
static const std::unordered_map<PredCondition, const char*> PredicateComparisonStrings = {
- {PredCondition::LessThan, "<"}, {PredCondition::Equal, "=="},
- {PredCondition::LessEqual, "<="}, {PredCondition::GreaterThan, ">"},
- {PredCondition::NotEqual, "!="}, {PredCondition::GreaterEqual, ">="},
+ {PredCondition::LessThan, "<"}, {PredCondition::Equal, "=="},
+ {PredCondition::LessEqual, "<="}, {PredCondition::GreaterThan, ">"},
+ {PredCondition::NotEqual, "!="}, {PredCondition::GreaterEqual, ">="},
+ {PredCondition::NotEqualWithNan, "!="},
};
- auto comparison = PredicateComparisonStrings.find(condition);
+ const auto& comparison{PredicateComparisonStrings.find(condition)};
ASSERT_MSG(comparison != PredicateComparisonStrings.end(),
"Unknown predicate comparison operation");
- return comparison->second;
+
+ std::string predicate{'(' + op_a + ") " + comparison->second + " (" + op_b + ')'};
+ if (condition == PredCondition::NotEqualWithNan) {
+ predicate += " || isnan(" + op_a + ") || isnan(" + op_b + ')';
+ }
+
+ return predicate;
}
/**
@@ -907,10 +917,6 @@ private:
regs.SetRegisterToFloat(instr.gpr0, 0, "inversesqrt(" + op_a + ')', 1, 1,
instr.alu.saturate_d);
break;
- case SubOp::Min:
- regs.SetRegisterToFloat(instr.gpr0, 0, "min(" + op_a + "," + op_b + ')', 1, 1,
- instr.alu.saturate_d);
- break;
default:
NGLOG_CRITICAL(HW_GPU, "Unhandled MUFU sub op: {0:x}",
static_cast<unsigned>(instr.sub_op.Value()));
@@ -1415,10 +1421,9 @@ private:
std::string second_pred =
GetPredicateCondition(instr.fsetp.pred39, instr.fsetp.neg_pred != 0);
- std::string comparator = GetPredicateComparison(instr.fsetp.cond);
std::string combiner = GetPredicateCombiner(instr.fsetp.op);
- std::string predicate = '(' + op_a + ") " + comparator + " (" + op_b + ')';
+ std::string predicate = GetPredicateComparison(instr.fsetp.cond, op_a, op_b);
// Set the primary predicate to the result of Predicate OP SecondPredicate
SetPredicate(instr.fsetp.pred3,
'(' + predicate + ") " + combiner + " (" + second_pred + ')');
@@ -1453,10 +1458,9 @@ private:
std::string second_pred =
GetPredicateCondition(instr.isetp.pred39, instr.isetp.neg_pred != 0);
- std::string comparator = GetPredicateComparison(instr.isetp.cond);
std::string combiner = GetPredicateCombiner(instr.isetp.op);
- std::string predicate = '(' + op_a + ") " + comparator + " (" + op_b + ')';
+ std::string predicate = GetPredicateComparison(instr.isetp.cond, op_a, op_b);
// Set the primary predicate to the result of Predicate OP SecondPredicate
SetPredicate(instr.isetp.pred3,
'(' + predicate + ") " + combiner + " (" + second_pred + ')');
@@ -1503,11 +1507,10 @@ private:
std::string second_pred =
GetPredicateCondition(instr.fset.pred39, instr.fset.neg_pred != 0);
- std::string comparator = GetPredicateComparison(instr.fset.cond);
std::string combiner = GetPredicateCombiner(instr.fset.op);
- std::string predicate = "(((" + op_a + ") " + comparator + " (" + op_b + ")) " +
- combiner + " (" + second_pred + "))";
+ std::string predicate = "((" + GetPredicateComparison(instr.fset.cond, op_a, op_b) +
+ ") " + combiner + " (" + second_pred + "))";
if (instr.fset.bf) {
regs.SetRegisterToFloat(instr.gpr0, 0, predicate + " ? 1.0 : 0.0", 1, 1);
@@ -1538,11 +1541,10 @@ private:
std::string second_pred =
GetPredicateCondition(instr.iset.pred39, instr.iset.neg_pred != 0);
- std::string comparator = GetPredicateComparison(instr.iset.cond);
std::string combiner = GetPredicateCombiner(instr.iset.op);
- std::string predicate = "(((" + op_a + ") " + comparator + " (" + op_b + ")) " +
- combiner + " (" + second_pred + "))";
+ std::string predicate = "((" + GetPredicateComparison(instr.iset.cond, op_a, op_b) +
+ ") " + combiner + " (" + second_pred + "))";
if (instr.iset.bf) {
regs.SetRegisterToFloat(instr.gpr0, 0, predicate + " ? 1.0 : 0.0", 1, 1);
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp
index 0db4367f1..eaf15da32 100644
--- a/src/video_core/textures/decoders.cpp
+++ b/src/video_core/textures/decoders.cpp
@@ -65,6 +65,8 @@ u32 BytesPerPixel(TextureFormat format) {
return 1;
case TextureFormat::R16_G16_B16_A16:
return 8;
+ case TextureFormat::R32_G32_B32_A32:
+ return 16;
default:
UNIMPLEMENTED_MSG("Format not implemented");
break;
@@ -94,6 +96,7 @@ std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width,
case TextureFormat::B5G6R5:
case TextureFormat::R8:
case TextureFormat::R16_G16_B16_A16:
+ case TextureFormat::R32_G32_B32_A32:
case TextureFormat::BF10GF11RF11:
case TextureFormat::ASTC_2D_4X4:
CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data,
@@ -124,6 +127,7 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat
case TextureFormat::B5G6R5:
case TextureFormat::R8:
case TextureFormat::BF10GF11RF11:
+ case TextureFormat::R32_G32_B32_A32:
// TODO(Subv): For the time being just forward the same data without any decoding.
rgba_data = texture_data;
break;