early-access version 3089

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 3088.
+This is the source code for early-access 3089.
 
 ## Legal Notice
 

src/shader_recompiler/CMakeLists.txt

@@ -224,6 +224,7 @@ add_library(shader_recompiler STATIC
     ir_opt/lower_fp16_to_fp32.cpp
     ir_opt/lower_int64_to_int32.cpp
     ir_opt/passes.h
+    ir_opt/position_pass.cpp
     ir_opt/rescaling_pass.cpp
     ir_opt/ssa_rewrite_pass.cpp
     ir_opt/texture_pass.cpp

src/shader_recompiler/backend/glasm/emit_glasm.cpp

@@ -450,6 +450,9 @@ std::string EmitGLASM(const Profile& profile, const RuntimeInfo& runtime_info, I
     if (program.info.uses_rescaling_uniform) {
         header += "PARAM scaling[1]={program.local[0..0]};";
     }
+    if (program.info.uses_render_area) {
+        header += "PARAM render_area[1]={program.local[1..1]};";
+    }
     header += "TEMP ";
     for (size_t index = 0; index < ctx.reg_alloc.NumUsedRegisters(); ++index) {
         header += fmt::format("R{},", index);

src/shader_recompiler/backend/glasm/emit_glasm_bitwise_conversion.cpp

@@ -43,6 +43,10 @@ void EmitBitCastU64F64(EmitContext&, IR::Inst& inst, const IR::Value& value) {
     Alias(inst, value);
 }
 
+void EmitBitCastS32F32(EmitContext&, IR::Inst& inst, const IR::Value& value) {
+    Alias(inst, value);
+}
+
 void EmitBitCastF16U16(EmitContext&, IR::Inst& inst, const IR::Value& value) {
     Alias(inst, value);
 }

src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp

@@ -386,7 +386,7 @@ void EmitInvocationInfo(EmitContext& ctx, IR::Inst& inst) {
         ctx.Add("SHL.U {}.x, primitive.vertexcount, 16;", inst);
         break;
     default:
-        LOG_WARNING(Shader, "(STUBBED) EmitInvocationInfo");
+        LOG_WARNING(Shader, "(STUBBED) called");
         ctx.Add("MOV.S {}.x,0x00ff0000;", inst);
     }
 }
@@ -408,6 +408,10 @@ void EmitResolutionDownFactor(EmitContext& ctx, IR::Inst& inst) {
     ctx.Add("MOV.F {}.x,scaling[0].z;", inst);
 }
 
+void EmitRenderArea(EmitContext& ctx, IR::Inst& inst) {
+    ctx.Add("MOV.F {},render_area[0];", inst);
+}
+
 void EmitLoadLocal(EmitContext& ctx, IR::Inst& inst, ScalarU32 word_offset) {
     ctx.Add("MOV.U {},lmem[{}].x;", inst, word_offset);
 }

src/shader_recompiler/backend/glasm/emit_glasm_instructions.h

@@ -74,6 +74,7 @@ void EmitSampleId(EmitContext& ctx, IR::Inst& inst);
 void EmitIsHelperInvocation(EmitContext& ctx, IR::Inst& inst);
 void EmitYDirection(EmitContext& ctx, IR::Inst& inst);
 void EmitResolutionDownFactor(EmitContext& ctx, IR::Inst& inst);
+void EmitRenderArea(EmitContext& ctx, IR::Inst& inst);
 void EmitLoadLocal(EmitContext& ctx, IR::Inst& inst, ScalarU32 word_offset);
 void EmitWriteLocal(EmitContext& ctx, ScalarU32 word_offset, ScalarU32 value);
 void EmitUndefU1(EmitContext& ctx, IR::Inst& inst);
@@ -196,6 +197,7 @@ void EmitSelectF64(EmitContext& ctx, ScalarS32 cond, Register true_value, Regist
 void EmitBitCastU16F16(EmitContext& ctx, IR::Inst& inst, const IR::Value& value);
 void EmitBitCastU32F32(EmitContext& ctx, IR::Inst& inst, const IR::Value& value);
 void EmitBitCastU64F64(EmitContext& ctx, IR::Inst& inst, const IR::Value& value);
+void EmitBitCastS32F32(EmitContext& ctx, IR::Inst& inst, const IR::Value& value);
 void EmitBitCastF16U16(EmitContext& ctx, IR::Inst& inst, const IR::Value& value);
 void EmitBitCastF32U32(EmitContext& ctx, IR::Inst& inst, const IR::Value& value);
 void EmitBitCastF64U64(EmitContext& ctx, IR::Inst& inst, const IR::Value& value);

src/shader_recompiler/backend/glasm/glasm_emit_context.cpp

@@ -95,10 +95,9 @@ EmitContext::EmitContext(IR::Program& program, Bindings& bindings, const Profile
     if (info.uses_invocation_id) {
         Add("ATTRIB primitive_invocation=primitive.invocation;");
     }
-    if (info.uses_invocation_info) {
+    if (info.uses_invocation_info &&
-        if (stage == Stage::TessellationControl || stage == Stage::TessellationEval) {
+        (stage == Stage::TessellationControl || stage == Stage::TessellationEval)) {
-            Add("ATTRIB primitive_vertexcount = primitive.vertexcount;");
+        Add("ATTRIB primitive_vertexcount = primitive.vertexcount;");
-        }
     }
     if (info.stores_tess_level_outer) {
         Add("OUTPUT result_patch_tessouter[]={{result.patch.tessouter[0..3]}};");

src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp

@@ -48,6 +48,10 @@ void EmitBitCastU64F64(EmitContext& ctx, IR::Inst& inst, std::string_view value)
     ctx.AddU64("{}=doubleBitsToUint64({});", inst, value);
 }
 
+void EmitBitCastS32F32(EmitContext& ctx, IR::Inst& inst, std::string_view value) {
+    ctx.AddF32("{}=ftoi({});", inst, value);
+}
+
 void EmitBitCastF16U16([[maybe_unused]] EmitContext& ctx, [[maybe_unused]] IR::Inst& inst) {
     NotImplemented();
 }

src/shader_recompiler/backend/glsl/emit_glsl_context_get_set.cpp

@@ -406,7 +406,7 @@ void EmitInvocationInfo(EmitContext& ctx, IR::Inst& inst) {
         ctx.AddU32("{}=uint(gl_PatchVerticesIn)<<16;", inst);
         break;
     default:
-        LOG_WARNING(Shader, "(STUBBED) EmitInvocationInfo");
+        LOG_WARNING(Shader, "(STUBBED) called");
         ctx.AddU32("{}=uint(0x00ff0000);", inst);
     }
 }
@@ -428,6 +428,10 @@ void EmitResolutionDownFactor(EmitContext& ctx, IR::Inst& inst) {
     ctx.AddF32("{}=scaling.z;", inst);
 }
 
+void EmitRenderArea(EmitContext& ctx, IR::Inst& inst) {
+    ctx.AddF32x4("{}=render_area;", inst);
+}
+
 void EmitLoadLocal(EmitContext& ctx, IR::Inst& inst, std::string_view word_offset) {
     ctx.AddU32("{}=lmem[{}];", inst, word_offset);
 }

src/shader_recompiler/backend/glsl/emit_glsl_instructions.h

@@ -88,6 +88,7 @@ void EmitSampleId(EmitContext& ctx, IR::Inst& inst);
 void EmitIsHelperInvocation(EmitContext& ctx, IR::Inst& inst);
 void EmitYDirection(EmitContext& ctx, IR::Inst& inst);
 void EmitResolutionDownFactor(EmitContext& ctx, IR::Inst& inst);
+void EmitRenderArea(EmitContext& ctx, IR::Inst& inst);
 void EmitLoadLocal(EmitContext& ctx, IR::Inst& inst, std::string_view word_offset);
 void EmitWriteLocal(EmitContext& ctx, std::string_view word_offset, std::string_view value);
 void EmitUndefU1(EmitContext& ctx, IR::Inst& inst);
@@ -230,6 +231,7 @@ void EmitSelectF64(EmitContext& ctx, IR::Inst& inst, std::string_view cond,
 void EmitBitCastU16F16(EmitContext& ctx, IR::Inst& inst);
 void EmitBitCastU32F32(EmitContext& ctx, IR::Inst& inst, std::string_view value);
 void EmitBitCastU64F64(EmitContext& ctx, IR::Inst& inst, std::string_view value);
+void EmitBitCastS32F32(EmitContext& ctx, IR::Inst& inst, std::string_view value);
 void EmitBitCastF16U16(EmitContext& ctx, IR::Inst& inst);
 void EmitBitCastF32U32(EmitContext& ctx, IR::Inst& inst, std::string_view value);
 void EmitBitCastF64U64(EmitContext& ctx, IR::Inst& inst, std::string_view value);

src/shader_recompiler/backend/glsl/glsl_emit_context.cpp

@@ -358,6 +358,9 @@ EmitContext::EmitContext(IR::Program& program, Bindings& bindings, const Profile
     if (info.uses_rescaling_uniform) {
         header += "layout(location=0) uniform vec4 scaling;";
     }
+    if (info.uses_render_area) {
+        header += "layout(location=1) uniform vec4 render_area;";
+    }
     DefineConstantBuffers(bindings);
     DefineConstantBufferIndirect();
     DefineStorageBuffers(bindings);

src/shader_recompiler/backend/spirv/emit_spirv.h

@@ -23,8 +23,12 @@ struct RescalingLayout {
     alignas(16) std::array<u32, NUM_IMAGE_SCALING_WORDS> rescaling_images;
     u32 down_factor;
 };
+struct RenderAreaLayout {
+    std::array<f32, 4> render_area;
+};
 constexpr u32 RESCALING_LAYOUT_WORDS_OFFSET = offsetof(RescalingLayout, rescaling_textures);
 constexpr u32 RESCALING_LAYOUT_DOWN_FACTOR_OFFSET = offsetof(RescalingLayout, down_factor);
+constexpr u32 RENDERAREA_LAYOUT_OFFSET = offsetof(RenderAreaLayout, render_area);
 
 [[nodiscard]] std::vector<u32> EmitSPIRV(const Profile& profile, const RuntimeInfo& runtime_info,
                                          IR::Program& program, Bindings& bindings);

src/shader_recompiler/backend/spirv/emit_spirv_bitwise_conversion.cpp

@@ -18,6 +18,10 @@ void EmitBitCastU64F64(EmitContext&) {
     throw NotImplementedException("SPIR-V Instruction");
 }
 
+void EmitBitCastS32F32(EmitContext&) {
+    throw NotImplementedException("SPIR-V Instruction");
+}
+
 void EmitBitCastF16U16(EmitContext&) {
     throw NotImplementedException("SPIR-V Instruction");
 }

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -353,7 +353,6 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
     case IR::Attribute::TessellationEvaluationPointV:
         return ctx.OpLoad(ctx.F32[1],
                           ctx.OpAccessChain(ctx.input_f32, ctx.tess_coord, ctx.Const(1U)));
-
     default:
         throw NotImplementedException("Read attribute {}", attr);
     }
@@ -519,9 +518,8 @@ Id EmitInvocationInfo(EmitContext& ctx) {
     case Stage::TessellationEval:
         return ctx.OpShiftLeftLogical(ctx.U32[1], ctx.OpLoad(ctx.U32[1], ctx.patch_vertices_in),
                                       ctx.Const(16u));
-        break;
     default:
-        LOG_WARNING(Shader, "(STUBBED) EmitInvocationInfo");
+        LOG_WARNING(Shader, "(STUBBED) called");
         return ctx.Const(0x00ff0000u);
     }
 }
@@ -550,6 +548,17 @@ Id EmitResolutionDownFactor(EmitContext& ctx) {
     }
 }
 
+Id EmitRenderArea(EmitContext& ctx) {
+    if (ctx.profile.unified_descriptor_binding) {
+        const Id pointer_type{ctx.TypePointer(spv::StorageClass::PushConstant, ctx.F32[4])};
+        const Id index{ctx.Const(ctx.render_are_member_index)};
+        const Id pointer{ctx.OpAccessChain(pointer_type, ctx.render_area_push_constant, index)};
+        return ctx.OpLoad(ctx.F32[4], pointer);
+    } else {
+        throw NotImplementedException("SPIR-V Instruction");
+    }
+}
+
 Id EmitLoadLocal(EmitContext& ctx, Id word_offset) {
     const Id pointer{ctx.OpAccessChain(ctx.private_u32, ctx.local_memory, word_offset)};
     return ctx.OpLoad(ctx.U32[1], pointer);

src/shader_recompiler/backend/spirv/emit_spirv_instructions.h

@@ -77,6 +77,7 @@ Id EmitSampleId(EmitContext& ctx);
 Id EmitIsHelperInvocation(EmitContext& ctx);
 Id EmitYDirection(EmitContext& ctx);
 Id EmitResolutionDownFactor(EmitContext& ctx);
+Id EmitRenderArea(EmitContext& ctx);
 Id EmitLoadLocal(EmitContext& ctx, Id word_offset);
 void EmitWriteLocal(EmitContext& ctx, Id word_offset, Id value);
 Id EmitUndefU1(EmitContext& ctx);
@@ -178,7 +179,8 @@ Id EmitSelectF64(EmitContext& ctx, Id cond, Id true_value, Id false_value);
 void EmitBitCastU16F16(EmitContext& ctx);
 Id EmitBitCastU32F32(EmitContext& ctx, Id value);
 void EmitBitCastU64F64(EmitContext& ctx);
-void EmitBitCastF16U16(EmitContext& ctx);
+void EmitBitCastS32F32(EmitContext& ctx);
+void EmitBitCastF16U16(EmitContext&);
 Id EmitBitCastF32U32(EmitContext& ctx, Id value);
 void EmitBitCastF64U64(EmitContext& ctx);
 Id EmitPackUint2x32(EmitContext& ctx, Id value);

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -473,6 +473,7 @@ EmitContext::EmitContext(const Profile& profile_, const RuntimeInfo& runtime_inf
     DefineAttributeMemAccess(program.info);
     DefineGlobalMemoryFunctions(program.info);
     DefineRescalingInput(program.info);
+    DefineRenderArea(program.info);
 }
 
 EmitContext::~EmitContext() = default;
@@ -982,6 +983,36 @@ void EmitContext::DefineRescalingInputUniformConstant() {
     }
 }
 
+void EmitContext::DefineRenderArea(const Info& info) {
+    if (!info.uses_render_area) {
+        return;
+    }
+
+    if (profile.unified_descriptor_binding) {
+        boost::container::static_vector<Id, 1> members{};
+        u32 member_index{0};
+
+        members.push_back(F32[4]);
+        render_are_member_index = member_index++;
+
+        const Id push_constant_struct{TypeStruct(std::span(members.data(), members.size()))};
+        Decorate(push_constant_struct, spv::Decoration::Block);
+        Name(push_constant_struct, "RenderAreaInfo");
+
+        MemberDecorate(push_constant_struct, render_are_member_index, spv::Decoration::Offset, 0);
+        MemberName(push_constant_struct, render_are_member_index, "render_area");
+
+        const Id pointer_type{TypePointer(spv::StorageClass::PushConstant, push_constant_struct)};
+        render_area_push_constant =
+            AddGlobalVariable(pointer_type, spv::StorageClass::PushConstant);
+        Name(render_area_push_constant, "render_area_push_constants");
+
+        if (profile.supported_spirv >= 0x00010400) {
+            interfaces.push_back(render_area_push_constant);
+        }
+    }
+}
+
 void EmitContext::DefineConstantBuffers(const Info& info, u32& binding) {
     if (info.constant_buffer_descriptors.empty()) {
         return;

src/shader_recompiler/backend/spirv/spirv_emit_context.h

@@ -244,6 +244,9 @@ public:
     u32 texture_rescaling_index{};
     u32 image_rescaling_index{};
 
+    Id render_area_push_constant{};
+    u32 render_are_member_index{};
+
     Id local_memory{};
 
     Id shared_memory_u8{};
@@ -319,6 +322,7 @@ private:
     void DefineRescalingInput(const Info& info);
     void DefineRescalingInputPushConstant();
     void DefineRescalingInputUniformConstant();
+    void DefineRenderArea(const Info& info);
 
     void DefineInputs(const IR::Program& program);
     void DefineOutputs(const IR::Program& program);

src/shader_recompiler/environment.h

@@ -22,6 +22,10 @@ public:
 
     [[nodiscard]] virtual TextureType ReadTextureType(u32 raw_handle) = 0;
 
+    [[nodiscard]] virtual TexturePixelFormat ReadTexturePixelFormat(u32 raw_handle) = 0;
+
+    [[nodiscard]] virtual u32 ReadViewportTransformState() = 0;
+
     [[nodiscard]] virtual u32 TextureBoundBuffer() const = 0;
 
     [[nodiscard]] virtual u32 LocalMemorySize() const = 0;

src/shader_recompiler/frontend/ir/ir_emitter.cpp

@@ -382,6 +382,14 @@ F32 IREmitter::ResolutionDownFactor() {
     return Inst<F32>(Opcode::ResolutionDownFactor);
 }
 
+F32 IREmitter::RenderAreaWidth() {
+    return F32(CompositeExtract(Inst<Value>(Opcode::RenderArea), 0));
+}
+
+F32 IREmitter::RenderAreaHeight() {
+    return F32(CompositeExtract(Inst<Value>(Opcode::RenderArea), 1));
+}
+
 U32 IREmitter::LaneId() {
     return Inst<U32>(Opcode::LaneId);
 }
@@ -687,6 +695,11 @@ IR::U32 IREmitter::BitCast<IR::U32, IR::F32>(const IR::F32& value) {
     return Inst<IR::U32>(Opcode::BitCastU32F32, value);
 }
 
+template <>
+IR::S32 IREmitter::BitCast<IR::S32, IR::F32>(const IR::F32& value) {
+    return Inst<IR::S32>(Opcode::BitCastS32F32, value);
+}
+
 template <>
 IR::F32 IREmitter::BitCast<IR::F32, IR::U32>(const IR::U32& value) {
     return Inst<IR::F32>(Opcode::BitCastF32U32, value);

src/shader_recompiler/frontend/ir/ir_emitter.h

@@ -104,6 +104,9 @@ public:
 
     [[nodiscard]] F32 ResolutionDownFactor();
 
+    [[nodiscard]] F32 RenderAreaWidth();
+    [[nodiscard]] F32 RenderAreaHeight();
+
     [[nodiscard]] U32 LaneId();
 
     [[nodiscard]] U32 LoadGlobalU8(const U64& address);

src/shader_recompiler/frontend/ir/opcodes.h

@@ -37,6 +37,7 @@ constexpr Type U8{Type::U8};
 constexpr Type U16{Type::U16};
 constexpr Type U32{Type::U32};
 constexpr Type U64{Type::U64};
+constexpr Type S32{Type::S32};
 constexpr Type F16{Type::F16};
 constexpr Type F32{Type::F32};
 constexpr Type F64{Type::F64};

src/shader_recompiler/frontend/ir/opcodes.inc

@@ -64,6 +64,7 @@ OPCODE(SampleId,                                            U32,
 OPCODE(IsHelperInvocation,                                  U1,                                                                                             )
 OPCODE(YDirection,                                          F32,                                                                                            )
 OPCODE(ResolutionDownFactor,                                F32,                                                                                            )
+OPCODE(RenderArea,                                          F32x4,                                                                                          )
 
 // Undefined
 OPCODE(UndefU1,                                             U1,                                                                                             )
@@ -174,6 +175,7 @@ OPCODE(SelectF64,                                           F64,            U1,
 OPCODE(BitCastU16F16,                                       U16,            F16,                                                                            )
 OPCODE(BitCastU32F32,                                       U32,            F32,                                                                            )
 OPCODE(BitCastU64F64,                                       U64,            F64,                                                                            )
+OPCODE(BitCastS32F32,                                       S32,            F32,                                                                            )
 OPCODE(BitCastF16U16,                                       F16,            U16,                                                                            )
 OPCODE(BitCastF32U32,                                       F32,            U32,                                                                            )
 OPCODE(BitCastF64U64,                                       F64,            U64,                                                                            )

src/shader_recompiler/frontend/ir/type.h

@@ -24,21 +24,22 @@ enum class Type {
     U16 = 1 << 7,
     U32 = 1 << 8,
     U64 = 1 << 9,
-    F16 = 1 << 10,
+    S32 = 1 << 10,
-    F32 = 1 << 11,
+    F16 = 1 << 11,
-    F64 = 1 << 12,
+    F32 = 1 << 12,
-    U32x2 = 1 << 13,
+    F64 = 1 << 13,
-    U32x3 = 1 << 14,
+    U32x2 = 1 << 14,
-    U32x4 = 1 << 15,
+    U32x3 = 1 << 15,
-    F16x2 = 1 << 16,
+    U32x4 = 1 << 16,
-    F16x3 = 1 << 17,
+    F16x2 = 1 << 17,
-    F16x4 = 1 << 18,
+    F16x3 = 1 << 18,
-    F32x2 = 1 << 19,
+    F16x4 = 1 << 19,
-    F32x3 = 1 << 20,
+    F32x2 = 1 << 20,
-    F32x4 = 1 << 21,
+    F32x3 = 1 << 21,
-    F64x2 = 1 << 22,
+    F32x4 = 1 << 22,
-    F64x3 = 1 << 23,
+    F64x2 = 1 << 23,
-    F64x4 = 1 << 24,
+    F64x3 = 1 << 24,
+    F64x4 = 1 << 25,
 };
 DECLARE_ENUM_FLAG_OPERATORS(Type)
 

src/shader_recompiler/frontend/ir/value.cpp

@@ -23,6 +23,8 @@ Value::Value(u16 value) noexcept : type{Type::U16}, imm_u16{value} {}
 
 Value::Value(u32 value) noexcept : type{Type::U32}, imm_u32{value} {}
 
+Value::Value(s32 value) noexcept : type{Type::S32}, imm_s32{value} {}
+
 Value::Value(f32 value) noexcept : type{Type::F32}, imm_f32{value} {}
 
 Value::Value(u64 value) noexcept : type{Type::U64}, imm_u64{value} {}
@@ -69,6 +71,7 @@ bool Value::operator==(const Value& other) const {
         return imm_u16 == other.imm_u16;
     case Type::U32:
     case Type::F32:
+    case Type::S32:
         return imm_u32 == other.imm_u32;
     case Type::U64:
     case Type::F64:

src/shader_recompiler/frontend/ir/value.h

@@ -44,6 +44,7 @@ public:
     explicit Value(u8 value) noexcept;
     explicit Value(u16 value) noexcept;
     explicit Value(u32 value) noexcept;
+    explicit Value(s32 value) noexcept;
     explicit Value(f32 value) noexcept;
     explicit Value(u64 value) noexcept;
     explicit Value(f64 value) noexcept;
@@ -66,6 +67,7 @@ public:
     [[nodiscard]] u8 U8() const;
     [[nodiscard]] u16 U16() const;
     [[nodiscard]] u32 U32() const;
+    [[nodiscard]] s32 S32() const;
     [[nodiscard]] f32 F32() const;
     [[nodiscard]] u64 U64() const;
     [[nodiscard]] f64 F64() const;
@@ -85,6 +87,7 @@ private:
         u8 imm_u8;
         u16 imm_u16;
         u32 imm_u32;
+        s32 imm_s32;
         f32 imm_f32;
         u64 imm_u64;
         f64 imm_f64;
@@ -266,6 +269,7 @@ using U8 = TypedValue<Type::U8>;
 using U16 = TypedValue<Type::U16>;
 using U32 = TypedValue<Type::U32>;
 using U64 = TypedValue<Type::U64>;
+using S32 = TypedValue<Type::S32>;
 using F16 = TypedValue<Type::F16>;
 using F32 = TypedValue<Type::F32>;
 using F64 = TypedValue<Type::F64>;
@@ -377,6 +381,14 @@ inline u32 Value::U32() const {
     return imm_u32;
 }
 
+inline s32 Value::S32() const {
+    if (IsIdentity()) {
+        return inst->Arg(0).S32();
+    }
+    DEBUG_ASSERT(type == Type::S32);
+    return imm_s32;
+}
+
 inline f32 Value::F32() const {
     if (IsIdentity()) {
         return inst->Arg(0).F32();

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -220,8 +220,10 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
 
     Optimization::ConstantPropagationPass(program);
 
+    Optimization::PositionPass(env, program);
+
     Optimization::GlobalMemoryToStorageBufferPass(program);
-    Optimization::TexturePass(env, program);
+    Optimization::TexturePass(env, program, host_info);
 
     if (Settings::values.resolution_info.active) {
         Optimization::RescalingPass(program);

src/shader_recompiler/host_translate_info.h

@@ -13,6 +13,7 @@ struct HostTranslateInfo {
     bool support_float16{};      ///< True when the device supports 16-bit floats
     bool support_int64{};        ///< True when the device supports 64-bit integers
     bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
+    bool support_snorm_render_buffer{}; ///< True when the device supports SNORM render buffers
 };
 
 } // namespace Shader

src/shader_recompiler/ir_opt/passes.h

@@ -6,6 +6,10 @@
 #include "shader_recompiler/environment.h"
 #include "shader_recompiler/frontend/ir/program.h"
 
+namespace Shader {
+struct HostTranslateInfo;
+}
+
 namespace Shader::Optimization {
 
 void CollectShaderInfoPass(Environment& env, IR::Program& program);
@@ -17,7 +21,8 @@ void LowerFp16ToFp32(IR::Program& program);
 void LowerInt64ToInt32(IR::Program& program);
 void RescalingPass(IR::Program& program);
 void SsaRewritePass(IR::Program& program);
-void TexturePass(Environment& env, IR::Program& program);
+void PositionPass(Environment& env, IR::Program& program);
+void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo& host_info);
 void VerificationPass(const IR::Program& program);
 
 // Dual Vertex

src/shader_recompiler/ir_opt/position_pass.cpp

@@ -0,0 +1,77 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <boost/container/small_vector.hpp>
+
+#include "shader_recompiler/frontend/ir/basic_block.h"
+#include "shader_recompiler/frontend/ir/ir_emitter.h"
+#include "shader_recompiler/frontend/ir/value.h"
+#include "shader_recompiler/ir_opt/passes.h"
+
+namespace Shader::Optimization {
+
+namespace {
+struct PositionInst {
+    IR::Inst* inst;
+    IR::Block* block;
+    IR::Attribute attr;
+};
+using PositionInstVector = boost::container::small_vector<PositionInst, 24>;
+} // Anonymous namespace
+
+void PositionPass(Environment& env, IR::Program& program) {
+    if (env.ShaderStage() != Stage::VertexB || env.ReadViewportTransformState()) {
+        return;
+    }
+
+    Info& info{program.info};
+    info.uses_render_area = true;
+
+    PositionInstVector to_replace;
+    for (IR::Block* const block : program.post_order_blocks) {
+        for (IR::Inst& inst : block->Instructions()) {
+            switch (inst.GetOpcode()) {
+            case IR::Opcode::SetAttribute: {
+                const IR::Attribute attr{inst.Arg(0).Attribute()};
+                switch (attr) {
+                case IR::Attribute::PositionX:
+                case IR::Attribute::PositionY: {
+                    to_replace.push_back(PositionInst{.inst = &inst, .block = block, .attr = attr});
+                    break;
+                }
+                default:
+                    break;
+                }
+                break;
+            }
+            default:
+                break;
+            }
+        }
+    }
+
+    for (PositionInst& position_inst : to_replace) {
+        IR::IREmitter ir{*position_inst.block,
+                         IR::Block::InstructionList::s_iterator_to(*position_inst.inst)};
+        const IR::F32 value(position_inst.inst->Arg(1));
+        const IR::F32F64 scale(ir.Imm32(2.f));
+        const IR::F32 negative_one{ir.Imm32(-1.f)};
+        switch (position_inst.attr) {
+        case IR::Attribute::PositionX: {
+            position_inst.inst->SetArg(
+                1,
+                ir.FPFma(value, ir.FPMul(ir.FPRecip(ir.RenderAreaWidth()), scale), negative_one));
+            break;
+        }
+        case IR::Attribute::PositionY: {
+            position_inst.inst->SetArg(
+                1,
+                ir.FPFma(value, ir.FPMul(ir.FPRecip(ir.RenderAreaHeight()), scale), negative_one));
+            break;
+        }
+        default:
+            break;
+        }
+    }
+}
+} // namespace Shader::Optimization

src/shader_recompiler/ir_opt/texture_pass.cpp

@@ -11,6 +11,7 @@
 #include "shader_recompiler/frontend/ir/basic_block.h"
 #include "shader_recompiler/frontend/ir/breadth_first_search.h"
 #include "shader_recompiler/frontend/ir/ir_emitter.h"
+#include "shader_recompiler/host_translate_info.h"
 #include "shader_recompiler/ir_opt/passes.h"
 #include "shader_recompiler/shader_info.h"
 
@@ -363,6 +364,14 @@ TextureType ReadTextureType(Environment& env, const ConstBufferAddr& cbuf) {
     return env.ReadTextureType(lhs_raw | rhs_raw);
 }
 
+TexturePixelFormat ReadTexturePixelFormat(Environment& env, const ConstBufferAddr& cbuf) {
+    const u32 secondary_index{cbuf.has_secondary ? cbuf.secondary_index : cbuf.index};
+    const u32 secondary_offset{cbuf.has_secondary ? cbuf.secondary_offset : cbuf.offset};
+    const u32 lhs_raw{env.ReadCbufValue(cbuf.index, cbuf.offset)};
+    const u32 rhs_raw{env.ReadCbufValue(secondary_index, secondary_offset)};
+    return env.ReadTexturePixelFormat(lhs_raw | rhs_raw);
+}
+
 class Descriptors {
 public:
     explicit Descriptors(TextureBufferDescriptors& texture_buffer_descriptors_,
@@ -451,9 +460,41 @@ void PatchImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
                ir.FPMul(IR::F32(ir.CompositeExtract(coord, 1)),
                         ir.FPRecip(ir.ConvertUToF(32, 32, ir.CompositeExtract(texture_size, 1))))));
 }
+
+void PatchTexelFetch(IR::Block& block, IR::Inst& inst, TexturePixelFormat pixel_format) {
+    const auto it{IR::Block::InstructionList::s_iterator_to(inst)};
+    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
+    auto get_max_value = [pixel_format]() -> float {
+        switch (pixel_format) {
+        case TexturePixelFormat::A8B8G8R8_SNORM:
+        case TexturePixelFormat::R8G8_SNORM:
+        case TexturePixelFormat::R8_SNORM:
+            return 1.f / std::numeric_limits<char>::max();
+        case TexturePixelFormat::R16G16B16A16_SNORM:
+        case TexturePixelFormat::R16G16_SNORM:
+        case TexturePixelFormat::R16_SNORM:
+            return 1.f / std::numeric_limits<short>::max();
+        default:
+            throw InvalidArgument("Invalid texture pixel format");
+        }
+    };
+
+    const IR::Value new_inst{&*block.PrependNewInst(it, inst)};
+    const IR::F32 x(ir.CompositeExtract(new_inst, 0));
+    const IR::F32 y(ir.CompositeExtract(new_inst, 1));
+    const IR::F32 z(ir.CompositeExtract(new_inst, 2));
+    const IR::F32 w(ir.CompositeExtract(new_inst, 3));
+    const IR::F16F32F64 max_value(ir.Imm32(get_max_value()));
+    const IR::Value converted =
+        ir.CompositeConstruct(ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::S32>(x)), max_value),
+                              ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::S32>(y)), max_value),
+                              ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::S32>(z)), max_value),
+                              ir.FPMul(ir.ConvertSToF(32, 32, ir.BitCast<IR::S32>(w)), max_value));
+    inst.ReplaceUsesWith(converted);
+}
 } // Anonymous namespace
 
-void TexturePass(Environment& env, IR::Program& program) {
+void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo& host_info) {
     TextureInstVector to_replace;
     for (IR::Block* const block : program.post_order_blocks) {
         for (IR::Inst& inst : block->Instructions()) {
@@ -597,6 +638,14 @@ void TexturePass(Environment& env, IR::Program& program) {
         } else {
             inst->SetArg(0, IR::Value{});
         }
+
+        if (!host_info.support_snorm_render_buffer && inst->GetOpcode() == IR::Opcode::ImageFetch &&
+            flags.type == TextureType::Buffer) {
+            const auto pixel_format = ReadTexturePixelFormat(env, cbuf);
+            if (pixel_format != TexturePixelFormat::OTHER) {
+                PatchTexelFetch(*texture_inst.block, *texture_inst.inst, pixel_format);
+            }
+        }
     }
 }
 

src/shader_recompiler/shader_info.h

@@ -29,6 +29,16 @@ enum class TextureType : u32 {
 };
 constexpr u32 NUM_TEXTURE_TYPES = 9;
 
+enum class TexturePixelFormat : u32 {
+    A8B8G8R8_SNORM,
+    R8_SNORM,
+    R8G8_SNORM,
+    R16G16B16A16_SNORM,
+    R16G16_SNORM,
+    R16_SNORM,
+    OTHER
+};
+
 enum class ImageFormat : u32 {
     Typeless,
     R8_UINT,
@@ -183,6 +193,7 @@ struct Info {
     bool uses_shadow_lod{};
     bool uses_rescaling_uniform{};
     bool uses_cbuf_indirect{};
+    bool uses_render_area{};
 
     IR::Type used_constant_buffer_types{};
     IR::Type used_storage_buffer_types{};

src/video_core/engines/maxwell_3d.h

@@ -2970,7 +2970,7 @@ public:
                 CullFace gl_cull_face;                                                 ///< 0x1920
                 Viewport::PixelCenter viewport_pixel_center;                           ///< 0x1924
                 INSERT_PADDING_BYTES_NOINIT(0x4);
-                u32 viewport_scale_offset_enbled;                                      ///< 0x192C
+                u32 viewport_scale_offset_enabled;                                      ///< 0x192C
                 INSERT_PADDING_BYTES_NOINIT(0xC);
                 ViewportClipControl viewport_clip_control;                             ///< 0x193C
                 UserClip::Op user_clip_op;                                             ///< 0x1940
@@ -3482,7 +3482,7 @@ ASSERT_REG_POSITION(gl_cull_test_enabled, 0x1918);
 ASSERT_REG_POSITION(gl_front_face, 0x191C);
 ASSERT_REG_POSITION(gl_cull_face, 0x1920);
 ASSERT_REG_POSITION(viewport_pixel_center, 0x1924);
-ASSERT_REG_POSITION(viewport_scale_offset_enbled, 0x192C);
+ASSERT_REG_POSITION(viewport_scale_offset_enabled, 0x192C);
 ASSERT_REG_POSITION(viewport_clip_control, 0x193C);
 ASSERT_REG_POSITION(user_clip_op, 0x1940);
 ASSERT_REG_POSITION(render_enable_override, 0x1944);

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -29,17 +29,17 @@ constexpr std::array PROGRAM_LUT{
 [[nodiscard]] GLenum GetTextureBufferFormat(GLenum gl_format) {
     switch (gl_format) {
     case GL_RGBA8_SNORM:
-        return GL_RGBA8;
+        return GL_RGBA8I;
     case GL_R8_SNORM:
-        return GL_R8;
+        return GL_R8I;
     case GL_RGBA16_SNORM:
-        return GL_RGBA16;
+        return GL_RGBA16I;
     case GL_R16_SNORM:
-        return GL_R16;
+        return GL_R16I;
     case GL_RG16_SNORM:
-        return GL_RG16;
+        return GL_RG16I;
     case GL_RG8_SNORM:
-        return GL_RG8;
+        return GL_RG8I;
     default:
         return gl_format;
     }
@@ -96,9 +96,6 @@ GLuint Buffer::View(u32 offset, u32 size, PixelFormat format) {
     texture.Create(GL_TEXTURE_BUFFER);
     const GLenum gl_format{MaxwellToGL::GetFormatTuple(format).internal_format};
     const GLenum texture_format{GetTextureBufferFormat(gl_format)};
-    if (texture_format != gl_format) {
-        LOG_WARNING(Render_OpenGL, "Emulating SNORM texture buffer with UNORM.");
-    }
     glTextureBufferRange(texture.handle, texture_format, buffer.handle, offset, size);
     views.push_back({
         .offset = offset,

src/video_core/renderer_opengl/gl_graphics_pipeline.cpp

@@ -503,6 +503,17 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
                                    float_image_scaling_mask, down_factor, 0.0f);
             }
         }
+        if (info.uses_render_area) {
+            const auto render_area_width(static_cast<GLfloat>(regs.surface_clip.width));
+            const auto render_area_height(static_cast<GLfloat>(regs.surface_clip.height));
+            if (use_assembly) {
+                glProgramLocalParameter4fARB(AssemblyStage(stage), 1, render_area_width,
+                                             render_area_height, 0.0f, 0.0f);
+            } else {
+                glProgramUniform4f(source_programs[stage].handle, 1, render_area_width,
+                                   render_area_height, 0.0f, 0.0f);
+            }
+        }
     }};
     if constexpr (Spec::enabled_stages[0]) {
         prepare_stage(0);

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -618,6 +618,16 @@ void RasterizerOpenGL::SyncViewport() {
             }
             flags[Dirty::Viewport0 + index] = false;
 
+            if (!regs.viewport_scale_offset_enabled) {
+                const auto x = static_cast<GLfloat>(regs.surface_clip.x);
+                const auto y = static_cast<GLfloat>(regs.surface_clip.y);
+                const auto width = static_cast<GLfloat>(regs.surface_clip.width);
+                const auto height = static_cast<GLfloat>(regs.surface_clip.height);
+                glViewportIndexedf(static_cast<GLuint>(index), x, y, width != 0.0f ? width : 1.0f,
+                                   height != 0.0f ? height : 1.0f);
+                continue;
+            }
+
             const auto& src = regs.viewport_transform[index];
             GLfloat x = conv(src.translate_x - src.scale_x);
             GLfloat y = conv(src.translate_y - src.scale_y);

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -49,7 +49,7 @@ using VideoCommon::LoadPipelines;
 using VideoCommon::SerializePipeline;
 using Context = ShaderContext::Context;
 
-constexpr u32 CACHE_VERSION = 6;
+constexpr u32 CACHE_VERSION = 7;
 
 template <typename Container>
 auto MakeSpan(Container& container) {
@@ -76,7 +76,7 @@ Shader::RuntimeInfo MakeRuntimeInfo(const GraphicsPipelineKey& key,
         }
         break;
     case Shader::Stage::TessellationEval:
-        // invert the face
+        // Flip the face, as opengl's drawing is also flipped
         info.tess_clockwise = key.tessellation_clockwise == 0;
         info.tess_primitive = [&key] {
             switch (key.tessellation_primitive) {
@@ -219,6 +219,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .support_float16 = false,
           .support_int64 = device.HasShaderInt64(),
           .needs_demote_reorder = device.IsAmd(),
+          .support_snorm_render_buffer = false,
       } {
     if (use_asynchronous_shaders) {
         workers = CreateWorkers();

src/video_core/renderer_opengl/gl_state_tracker.cpp

@@ -70,8 +70,8 @@ void SetupDirtyViewports(Tables& tables) {
     FillBlock(tables[1], OFF(viewport_transform), NUM(viewport_transform), Viewports);
     FillBlock(tables[1], OFF(viewports), NUM(viewports), Viewports);
 
-    tables[0][OFF(viewport_scale_offset_enbled)] = ViewportTransform;
+    tables[0][OFF(viewport_scale_offset_enabled)] = ViewportTransform;
-    tables[1][OFF(viewport_scale_offset_enbled)] = Viewports;
+    tables[1][OFF(viewport_scale_offset_enabled)] = Viewports;
 }
 
 void SetupDirtyScissors(Tables& tables) {

src/video_core/renderer_vulkan/pipeline_helper.h

@@ -68,13 +68,15 @@ public:
     }
 
     vk::PipelineLayout CreatePipelineLayout(VkDescriptorSetLayout descriptor_set_layout) const {
+        using Shader::Backend::SPIRV::RenderAreaLayout;
         using Shader::Backend::SPIRV::RescalingLayout;
         const u32 size_offset = is_compute ? sizeof(RescalingLayout::down_factor) : 0u;
         const VkPushConstantRange range{
             .stageFlags = static_cast<VkShaderStageFlags>(
                 is_compute ? VK_SHADER_STAGE_COMPUTE_BIT : VK_SHADER_STAGE_ALL_GRAPHICS),
             .offset = 0,
-            .size = static_cast<u32>(sizeof(RescalingLayout)) - size_offset,
+            .size = static_cast<u32>(sizeof(RescalingLayout)) - size_offset +
+                    static_cast<u32>(sizeof(RenderAreaLayout)),
         };
         return device->GetLogical().CreatePipelineLayout({
             .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
@@ -167,6 +169,12 @@ private:
     u32 image_bit{1u};
 };
 
+class RenderAreaPushConstant {
+public:
+    bool uses_render_area{};
+    std::array<f32, 4> words{};
+};
+
 inline void PushImageDescriptors(TextureCache& texture_cache,
                                  UpdateDescriptorQueue& update_descriptor_queue,
                                  const Shader::Info& info, RescalingPushConstant& rescaling,

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -31,6 +31,7 @@ namespace {
 using boost::container::small_vector;
 using boost::container::static_vector;
 using Shader::ImageBufferDescriptor;
+using Shader::Backend::SPIRV::RENDERAREA_LAYOUT_OFFSET;
 using Shader::Backend::SPIRV::RESCALING_LAYOUT_DOWN_FACTOR_OFFSET;
 using Shader::Backend::SPIRV::RESCALING_LAYOUT_WORDS_OFFSET;
 using Tegra::Texture::TexturePair;
@@ -433,12 +434,19 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
     update_descriptor_queue.Acquire();
 
     RescalingPushConstant rescaling;
+    RenderAreaPushConstant render_area;
     const VkSampler* samplers_it{samplers.data()};
     const VideoCommon::ImageViewInOut* views_it{views.data()};
     const auto prepare_stage{[&](size_t stage) LAMBDA_FORCEINLINE {
         buffer_cache.BindHostStageBuffers(stage);
         PushImageDescriptors(texture_cache, update_descriptor_queue, stage_infos[stage], rescaling,
                              samplers_it, views_it);
+        const auto& info{stage_infos[0]};
+        if (info.uses_render_area) {
+            render_area.uses_render_area = true;
+            render_area.words = {static_cast<float>(regs.surface_clip.width),
+                                 static_cast<float>(regs.surface_clip.height)};
+        }
     }};
     if constexpr (Spec::enabled_stages[0]) {
         prepare_stage(0);
@@ -455,10 +463,11 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
     if constexpr (Spec::enabled_stages[4]) {
         prepare_stage(4);
     }
-    ConfigureDraw(rescaling);
+    ConfigureDraw(rescaling, render_area);
 }
 
-void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling) {
+void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
+                                     const RenderAreaPushConstant& render_area) {
     texture_cache.UpdateRenderTargets(false);
     scheduler.RequestRenderpass(texture_cache.GetFramebuffer());
 
@@ -474,7 +483,9 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling) {
     const bool bind_pipeline{scheduler.UpdateGraphicsPipeline(this)};
     const void* const descriptor_data{update_descriptor_queue.UpdateData()};
     scheduler.Record([this, descriptor_data, bind_pipeline, rescaling_data = rescaling.Data(),
-                      is_rescaling, update_rescaling](vk::CommandBuffer cmdbuf) {
+                      is_rescaling, update_rescaling,
+                      uses_render_area = render_area.uses_render_area,
+                      render_area_data = render_area.words](vk::CommandBuffer cmdbuf) {
         if (bind_pipeline) {
             cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline);
         }
@@ -488,6 +499,11 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling) {
                                  RESCALING_LAYOUT_DOWN_FACTOR_OFFSET, sizeof(scale_down_factor),
                                  &scale_down_factor);
         }
+        if (uses_render_area) {
+            cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
+                                 RENDERAREA_LAYOUT_OFFSET, sizeof(render_area_data),
+                                 &render_area_data);
+        }
         if (!descriptor_set_layout) {
             return;
         }

src/video_core/renderer_vulkan/vk_graphics_pipeline.h

@@ -62,6 +62,7 @@ class Device;
 class PipelineStatistics;
 class RenderPassCache;
 class RescalingPushConstant;
+class RenderAreaPushConstant;
 class Scheduler;
 class UpdateDescriptorQueue;
 
@@ -119,7 +120,8 @@ private:
     template <typename Spec>
     void ConfigureImpl(bool is_indexed);
 
-    void ConfigureDraw(const RescalingPushConstant& rescaling);
+    void ConfigureDraw(const RescalingPushConstant& rescaling,
+                       const RenderAreaPushConstant& render_are);
 
     void MakePipeline(VkRenderPass render_pass);
 

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -53,7 +53,7 @@ using VideoCommon::FileEnvironment;
 using VideoCommon::GenericEnvironment;
 using VideoCommon::GraphicsEnvironment;
 
-constexpr u32 CACHE_VERSION = 6;
+constexpr u32 CACHE_VERSION = 7;
 
 template <typename Container>
 auto MakeSpan(Container& container) {
@@ -326,6 +326,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .support_int64 = device.IsShaderInt64Supported(),
         .needs_demote_reorder = driver_id == VK_DRIVER_ID_AMD_PROPRIETARY_KHR ||
                                 driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE_KHR,
+        .support_snorm_render_buffer = true,
     };
 }
 

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -683,6 +683,22 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& reg
     if (!state_tracker.TouchViewports()) {
         return;
     }
+    if (!regs.viewport_scale_offset_enabled) {
+        const auto x = static_cast<float>(regs.surface_clip.x);
+        const auto y = static_cast<float>(regs.surface_clip.y);
+        const auto width = static_cast<float>(regs.surface_clip.width);
+        const auto height = static_cast<float>(regs.surface_clip.height);
+        VkViewport viewport{
+            .x = x,
+            .y = y,
+            .width = width != 0.0f ? width : 1.0f,
+            .height = height != 0.0f ? height : 1.0f,
+            .minDepth = 0.0f,
+            .maxDepth = 1.0f,
+        };
+        scheduler.Record([viewport](vk::CommandBuffer cmdbuf) { cmdbuf.SetViewport(0, viewport); });
+        return;
+    }
     const bool is_rescaling{texture_cache.IsRescaling()};
     const float scale = is_rescaling ? Settings::values.resolution_info.up_factor : 1.0f;
     const std::array viewports{

src/video_core/renderer_vulkan/vk_state_tracker.cpp

@@ -51,7 +51,7 @@ Flags MakeInvalidationFlags() {
 void SetupDirtyViewports(Tables& tables) {
     FillBlock(tables[0], OFF(viewport_transform), NUM(viewport_transform), Viewports);
     FillBlock(tables[0], OFF(viewports), NUM(viewports), Viewports);
-    tables[0][OFF(viewport_scale_offset_enbled)] = Viewports;
+    tables[0][OFF(viewport_scale_offset_enabled)] = Viewports;
     tables[1][OFF(window_origin)] = Viewports;
 }
 

src/video_core/shader_environment.cpp

@@ -19,6 +19,7 @@
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/memory_manager.h"
 #include "video_core/shader_environment.h"
+#include "video_core/texture_cache/format_lookup_table.h"
 #include "video_core/textures/texture.h"
 
 namespace VideoCommon {
@@ -33,7 +34,7 @@ static u64 MakeCbufKey(u32 index, u32 offset) {
     return (static_cast<u64>(index) << 32) | offset;
 }
 
-static Shader::TextureType ConvertType(const Tegra::Texture::TICEntry& entry) {
+static Shader::TextureType ConvertTextureType(const Tegra::Texture::TICEntry& entry) {
     switch (entry.texture_type) {
     case Tegra::Texture::TextureType::Texture1D:
         return Shader::TextureType::Color1D;
@@ -59,6 +60,26 @@ static Shader::TextureType ConvertType(const Tegra::Texture::TICEntry& entry) {
     }
 }
 
+static Shader::TexturePixelFormat ConvertTexturePixelFormat(const Tegra::Texture::TICEntry& entry) {
+    switch (PixelFormatFromTextureInfo(entry.format, entry.r_type, entry.g_type, entry.b_type,
+                                       entry.a_type, entry.srgb_conversion)) {
+    case VideoCore::Surface::PixelFormat::A8B8G8R8_SNORM:
+        return Shader::TexturePixelFormat::A8B8G8R8_SNORM;
+    case VideoCore::Surface::PixelFormat::R8_SNORM:
+        return Shader::TexturePixelFormat::R8_SNORM;
+    case VideoCore::Surface::PixelFormat::R8G8_SNORM:
+        return Shader::TexturePixelFormat::R8G8_SNORM;
+    case VideoCore::Surface::PixelFormat::R16G16B16A16_SNORM:
+        return Shader::TexturePixelFormat::R16G16B16A16_SNORM;
+    case VideoCore::Surface::PixelFormat::R16G16_SNORM:
+        return Shader::TexturePixelFormat::R16G16_SNORM;
+    case VideoCore::Surface::PixelFormat::R16_SNORM:
+        return Shader::TexturePixelFormat::R16_SNORM;
+    default:
+        return Shader::TexturePixelFormat::OTHER;
+    }
+}
+
 static std::string_view StageToPrefix(Shader::Stage stage) {
     switch (stage) {
     case Shader::Stage::VertexB:
@@ -178,22 +199,31 @@ void GenericEnvironment::Dump(u64 hash) {
 void GenericEnvironment::Serialize(std::ofstream& file) const {
     const u64 code_size{static_cast<u64>(CachedSize())};
     const u64 num_texture_types{static_cast<u64>(texture_types.size())};
+    const u64 num_texture_pixel_formats{static_cast<u64>(texture_pixel_formats.size())};
     const u64 num_cbuf_values{static_cast<u64>(cbuf_values.size())};
 
     file.write(reinterpret_cast<const char*>(&code_size), sizeof(code_size))
         .write(reinterpret_cast<const char*>(&num_texture_types), sizeof(num_texture_types))
+        .write(reinterpret_cast<const char*>(&num_texture_pixel_formats),
+               sizeof(num_texture_pixel_formats))
         .write(reinterpret_cast<const char*>(&num_cbuf_values), sizeof(num_cbuf_values))
         .write(reinterpret_cast<const char*>(&local_memory_size), sizeof(local_memory_size))
         .write(reinterpret_cast<const char*>(&texture_bound), sizeof(texture_bound))
         .write(reinterpret_cast<const char*>(&start_address), sizeof(start_address))
         .write(reinterpret_cast<const char*>(&cached_lowest), sizeof(cached_lowest))
         .write(reinterpret_cast<const char*>(&cached_highest), sizeof(cached_highest))
+        .write(reinterpret_cast<const char*>(&viewport_transform_state),
+               sizeof(viewport_transform_state))
         .write(reinterpret_cast<const char*>(&stage), sizeof(stage))
         .write(reinterpret_cast<const char*>(code.data()), code_size);
     for (const auto& [key, type] : texture_types) {
         file.write(reinterpret_cast<const char*>(&key), sizeof(key))
             .write(reinterpret_cast<const char*>(&type), sizeof(type));
     }
+    for (const auto& [key, format] : texture_pixel_formats) {
+        file.write(reinterpret_cast<const char*>(&key), sizeof(key))
+            .write(reinterpret_cast<const char*>(&format), sizeof(format));
+    }
     for (const auto& [key, type] : cbuf_values) {
         file.write(reinterpret_cast<const char*>(&key), sizeof(key))
             .write(reinterpret_cast<const char*>(&type), sizeof(type));
@@ -237,15 +267,13 @@ std::optional<u64> GenericEnvironment::TryFindSize() {
     return std::nullopt;
 }
 
-Shader::TextureType GenericEnvironment::ReadTextureTypeImpl(GPUVAddr tic_addr, u32 tic_limit,
+Tegra::Texture::TICEntry GenericEnvironment::ReadTextureInfo(GPUVAddr tic_addr, u32 tic_limit,
-                                                            bool via_header_index, u32 raw) {
+                                                             bool via_header_index, u32 raw) {
     const auto handle{Tegra::Texture::TexturePair(raw, via_header_index)};
     const GPUVAddr descriptor_addr{tic_addr + handle.first * sizeof(Tegra::Texture::TICEntry)};
     Tegra::Texture::TICEntry entry;
     gpu_memory->ReadBlock(descriptor_addr, &entry, sizeof(entry));
-    const Shader::TextureType result{ConvertType(entry)};
+    return entry;
-    texture_types.emplace(raw, result);
-    return result;
 }
 
 GraphicsEnvironment::GraphicsEnvironment(Tegra::Engines::Maxwell3D& maxwell3d_,
@@ -305,8 +333,27 @@ u32 GraphicsEnvironment::ReadCbufValue(u32 cbuf_index, u32 cbuf_offset) {
 Shader::TextureType GraphicsEnvironment::ReadTextureType(u32 handle) {
     const auto& regs{maxwell3d->regs};
     const bool via_header_index{regs.sampler_binding == Maxwell::SamplerBinding::ViaHeaderBinding};
-    return ReadTextureTypeImpl(regs.tex_header.Address(), regs.tex_header.limit, via_header_index,
+    auto entry =
-                               handle);
+        ReadTextureInfo(regs.tex_header.Address(), regs.tex_header.limit, via_header_index, handle);
+    const Shader::TextureType result{ConvertTextureType(entry)};
+    texture_types.emplace(handle, result);
+    return result;
+}
+
+Shader::TexturePixelFormat GraphicsEnvironment::ReadTexturePixelFormat(u32 handle) {
+    const auto& regs{maxwell3d->regs};
+    const bool via_header_index{regs.sampler_binding == Maxwell::SamplerBinding::ViaHeaderBinding};
+    auto entry =
+        ReadTextureInfo(regs.tex_header.Address(), regs.tex_header.limit, via_header_index, handle);
+    const Shader::TexturePixelFormat result(ConvertTexturePixelFormat(entry));
+    texture_pixel_formats.emplace(handle, result);
+    return result;
+}
+
+u32 GraphicsEnvironment::ReadViewportTransformState() {
+    const auto& regs{maxwell3d->regs};
+    viewport_transform_state = regs.viewport_scale_offset_enabled;
+    return viewport_transform_state;
 }
 
 ComputeEnvironment::ComputeEnvironment(Tegra::Engines::KeplerCompute& kepler_compute_,
@@ -337,21 +384,41 @@ u32 ComputeEnvironment::ReadCbufValue(u32 cbuf_index, u32 cbuf_offset) {
 Shader::TextureType ComputeEnvironment::ReadTextureType(u32 handle) {
     const auto& regs{kepler_compute->regs};
     const auto& qmd{kepler_compute->launch_description};
-    return ReadTextureTypeImpl(regs.tic.Address(), regs.tic.limit, qmd.linked_tsc != 0, handle);
+    auto entry = ReadTextureInfo(regs.tic.Address(), regs.tic.limit, qmd.linked_tsc != 0, handle);
+    const Shader::TextureType result{ConvertTextureType(entry)};
+    texture_types.emplace(handle, result);
+    return result;
+}
+
+Shader::TexturePixelFormat ComputeEnvironment::ReadTexturePixelFormat(u32 handle) {
+    const auto& regs{kepler_compute->regs};
+    const auto& qmd{kepler_compute->launch_description};
+    auto entry = ReadTextureInfo(regs.tic.Address(), regs.tic.limit, qmd.linked_tsc != 0, handle);
+    const Shader::TexturePixelFormat result(ConvertTexturePixelFormat(entry));
+    texture_pixel_formats.emplace(handle, result);
+    return result;
+}
+
+u32 ComputeEnvironment::ReadViewportTransformState() {
+    return viewport_transform_state;
 }
 
 void FileEnvironment::Deserialize(std::ifstream& file) {
     u64 code_size{};
     u64 num_texture_types{};
+    u64 num_texture_pixel_formats{};
     u64 num_cbuf_values{};
     file.read(reinterpret_cast<char*>(&code_size), sizeof(code_size))
         .read(reinterpret_cast<char*>(&num_texture_types), sizeof(num_texture_types))
+        .read(reinterpret_cast<char*>(&num_texture_pixel_formats),
+              sizeof(num_texture_pixel_formats))
         .read(reinterpret_cast<char*>(&num_cbuf_values), sizeof(num_cbuf_values))
         .read(reinterpret_cast<char*>(&local_memory_size), sizeof(local_memory_size))
         .read(reinterpret_cast<char*>(&texture_bound), sizeof(texture_bound))
         .read(reinterpret_cast<char*>(&start_address), sizeof(start_address))
         .read(reinterpret_cast<char*>(&read_lowest), sizeof(read_lowest))
         .read(reinterpret_cast<char*>(&read_highest), sizeof(read_highest))
+        .read(reinterpret_cast<char*>(&viewport_transform_state), sizeof(viewport_transform_state))
         .read(reinterpret_cast<char*>(&stage), sizeof(stage));
     code = std::make_unique<u64[]>(Common::DivCeil(code_size, sizeof(u64)));
     file.read(reinterpret_cast<char*>(code.get()), code_size);
@@ -362,6 +429,13 @@ void FileEnvironment::Deserialize(std::ifstream& file) {
             .read(reinterpret_cast<char*>(&type), sizeof(type));
         texture_types.emplace(key, type);
     }
+    for (size_t i = 0; i < num_texture_pixel_formats; ++i) {
+        u32 key;
+        Shader::TexturePixelFormat format;
+        file.read(reinterpret_cast<char*>(&key), sizeof(key))
+            .read(reinterpret_cast<char*>(&format), sizeof(format));
+        texture_pixel_formats.emplace(key, format);
+    }
     for (size_t i = 0; i < num_cbuf_values; ++i) {
         u64 key;
         u32 value;
@@ -409,6 +483,18 @@ Shader::TextureType FileEnvironment::ReadTextureType(u32 handle) {
     return it->second;
 }
 
+Shader::TexturePixelFormat FileEnvironment::ReadTexturePixelFormat(u32 handle) {
+    const auto it{texture_pixel_formats.find(handle)};
+    if (it == texture_pixel_formats.end()) {
+        throw Shader::LogicError("Uncached read texture pixel format");
+    }
+    return it->second;
+}
+
+u32 FileEnvironment::ReadViewportTransformState() {
+    return viewport_transform_state;
+}
+
 u32 FileEnvironment::LocalMemorySize() const {
     return local_memory_size;
 }

src/video_core/shader_environment.h

@@ -63,14 +63,15 @@ public:
 protected:
     std::optional<u64> TryFindSize();
 
-    Shader::TextureType ReadTextureTypeImpl(GPUVAddr tic_addr, u32 tic_limit, bool via_header_index,
+    Tegra::Texture::TICEntry ReadTextureInfo(GPUVAddr tic_addr, u32 tic_limit,
-                                            u32 raw);
+                                             bool via_header_index, u32 raw);
 
     Tegra::MemoryManager* gpu_memory{};
     GPUVAddr program_base{};
 
     std::vector<u64> code;
     std::unordered_map<u32, Shader::TextureType> texture_types;
+    std::unordered_map<u32, Shader::TexturePixelFormat> texture_pixel_formats;
     std::unordered_map<u64, u32> cbuf_values;
 
     u32 local_memory_size{};
@@ -85,6 +86,8 @@ protected:
     u32 cached_highest = 0;
     u32 initial_offset = 0;
 
+    u32 viewport_transform_state = 1;
+
     bool has_unbound_instructions = false;
 };
 
@@ -102,6 +105,10 @@ public:
 
     Shader::TextureType ReadTextureType(u32 handle) override;
 
+    Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
+
+    u32 ReadViewportTransformState() override;
+
 private:
     Tegra::Engines::Maxwell3D* maxwell3d{};
     size_t stage_index{};
@@ -120,6 +127,10 @@ public:
 
     Shader::TextureType ReadTextureType(u32 handle) override;
 
+    Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
+
+    u32 ReadViewportTransformState() override;
+
 private:
     Tegra::Engines::KeplerCompute* kepler_compute{};
 };
@@ -143,6 +154,10 @@ public:
 
     [[nodiscard]] Shader::TextureType ReadTextureType(u32 handle) override;
 
+    [[nodiscard]] Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
+
+    [[nodiscard]] u32 ReadViewportTransformState() override;
+
     [[nodiscard]] u32 LocalMemorySize() const override;
 
     [[nodiscard]] u32 SharedMemorySize() const override;
@@ -156,6 +171,7 @@ public:
 private:
     std::unique_ptr<u64[]> code;
     std::unordered_map<u32, Shader::TextureType> texture_types;
+    std::unordered_map<u32, Shader::TexturePixelFormat> texture_pixel_formats;
     std::unordered_map<u64, u32> cbuf_values;
     std::array<u32, 3> workgroup_size{};
     u32 local_memory_size{};
@@ -164,6 +180,7 @@ private:
     u32 read_lowest{};
     u32 read_highest{};
     u32 initial_offset{};
+    u32 viewport_transform_state = 1;
 };
 
 void SerializePipeline(std::span<const char> key, std::span<const GenericEnvironment* const> envs,

src/video_core/texture_cache/util.cpp

@@ -516,7 +516,6 @@ void SwizzleBlockLinearImage(Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr
     const u32 num_blocks_per_layer = NumBlocks(level_size, tile_size);
     const u32 host_bytes_per_layer = num_blocks_per_layer * bytes_per_block;
 
-    UNIMPLEMENTED_IF(info.tile_width_spacing > 0);
     UNIMPLEMENTED_IF(copy.image_offset.x != 0);
     UNIMPLEMENTED_IF(copy.image_offset.y != 0);
     UNIMPLEMENTED_IF(copy.image_offset.z != 0);