8 files changed, 143 insertions, 26 deletions

diff --git a/externals/nihstro b/externals/nihstro
-Subproject 4a78588b308564f7ebae193e0ae00d9a0d5741d
+Subproject 81f1804a43f625e3a1a20752c0db70a41341038
diff --git a/src/core/core.cpp b/src/core/core.cpp
index bb2ed7a92..b5c258230 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -61,10 +61,6 @@ int Init() {
     g_sys_core = new ARM_DynCom(USER32MODE);
     g_app_core = new ARM_DynCom(USER32MODE);
 
-    // TODO: Whenever TLS is implemented, this should contain
-    // the address of the 0x200-byte TLS
-    g_app_core->SetCP15Register(CP15_THREAD_URO, Memory::TLS_AREA_VADDR);
-
     LOG_DEBUG(Core, "Initialized OK");
     return 0;
 }
diff --git a/src/core/hle/kernel/session.h b/src/core/hle/kernel/session.h
index 0fd18148a..8c3886ffd 100644
--- a/src/core/hle/kernel/session.h
+++ b/src/core/hle/kernel/session.h
@@ -5,6 +5,7 @@
 #pragma once
 
 #include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/thread.h"
 #include "core/mem_map.h"
 
 namespace Kernel {
@@ -12,12 +13,15 @@ namespace Kernel {
 static const int kCommandHeaderOffset = 0x80; ///< Offset into command buffer of header
 
 /**
- * Returns a pointer to the command buffer in kernel memory
+ * Returns a pointer to the command buffer in the current thread's TLS
+ * TODO(Subv): This is not entirely correct, the command buffer should be copied from
+ * the thread's TLS to an intermediate buffer in kernel memory, and then copied again to
+ * the service handler process' memory.
  * @param offset Optional offset into command buffer
  * @return Pointer to command buffer
  */
-inline static u32* GetCommandBuffer(const int offset=0) {
-    return (u32*)Memory::GetPointer(Memory::TLS_AREA_VADDR + kCommandHeaderOffset + offset);
+inline static u32* GetCommandBuffer(const int offset = 0) {
+    return (u32*)Memory::GetPointer(GetCurrentThread()->GetTLSAddress() + kCommandHeaderOffset + offset);
 }
 
 /**
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index 0a3fd7cb1..5de8f9a73 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -197,6 +197,7 @@ static void SwitchContext(Thread* new_thread) {
         new_thread->current_priority = new_thread->nominal_priority;
 
         Core::g_app_core->LoadContext(new_thread->context);
+        Core::g_app_core->SetCP15Register(CP15_THREAD_URO, new_thread->GetTLSAddress());
     } else {
         current_thread = nullptr;
     }
@@ -402,6 +403,12 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point,
     thread->name = std::move(name);
     thread->callback_handle = wakeup_callback_handle_table.Create(thread).MoveFrom();
 
+    VAddr tls_address = Memory::TLS_AREA_VADDR + (thread->thread_id - 1) * 0x200;
+
+    ASSERT_MSG(tls_address < Memory::TLS_AREA_VADDR_END, "Too many threads");
+
+    thread->tls_address = tls_address;
+
     // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used
     // to initialize the context
     Core::g_app_core->ResetContext(thread->context, stack_top, entry_point, arg);
@@ -495,6 +502,10 @@ void Thread::SetWaitSynchronizationOutput(s32 output) {
     context.cpu_registers[1] = output;
 }
 
+VAddr Thread::GetTLSAddress() const {
+    return tls_address;
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
 void ThreadingInit() {
diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h
index 9958b16e6..6891c8c2f 100644
--- a/src/core/hle/kernel/thread.h
+++ b/src/core/hle/kernel/thread.h
@@ -135,6 +135,12 @@ public:
      */
     void Stop();
 
+    /*
+     * Returns the Thread Local Storage address of the current thread
+     * @returns VAddr of the thread's TLS
+     */
+    VAddr GetTLSAddress() const;
+
     Core::ThreadContext context;
 
     u32 thread_id;
@@ -150,6 +156,8 @@ public:
 
     s32 processor_id;
 
+    VAddr tls_address; ///< Address of the Thread Local Storage of the thread
+
     /// Mutexes currently held by this thread, which will be released when it exits.
     boost::container::flat_set<SharedPtr<Mutex>> held_mutexes;
 
diff --git a/src/video_core/pica.h b/src/video_core/pica.h
index e4a91058c..5e169ff69 100644
--- a/src/video_core/pica.h
+++ b/src/video_core/pica.h
@@ -226,7 +226,8 @@ struct Regs {
             Texture1               = 0x4,
             Texture2               = 0x5,
             Texture3               = 0x6,
-            // 0x7-0xc = primary color??
+
+            PreviousBuffer         = 0xd,
             Constant               = 0xe,
             Previous               = 0xf,
         };
@@ -299,7 +300,18 @@ struct Regs {
             BitField<24, 8, u32> const_a;
         };
 
-        INSERT_PADDING_WORDS(0x1);
+        union {
+            BitField< 0, 2, u32> color_scale;
+            BitField<16, 2, u32> alpha_scale;
+        };
+
+        inline unsigned GetColorMultiplier() const {
+            return (color_scale < 3) ? (1 << color_scale) : 1;
+        }
+
+        inline unsigned GetAlphaMultiplier() const {
+            return (alpha_scale < 3) ? (1 << alpha_scale) : 1;
+        }
     };
 
     TevStageConfig tev_stage0;
@@ -309,11 +321,36 @@ struct Regs {
     TevStageConfig tev_stage2;
     INSERT_PADDING_WORDS(0x3);
     TevStageConfig tev_stage3;
-    INSERT_PADDING_WORDS(0x13);
+    INSERT_PADDING_WORDS(0x3);
+
+    union {
+        // Tev stages 0-3 write their output to the combiner buffer if the corresponding bit in
+        // these masks are set
+        BitField< 8, 4, u32> update_mask_rgb;
+        BitField<12, 4, u32> update_mask_a;
+
+        bool TevStageUpdatesCombinerBufferColor(unsigned stage_index) const {
+            return (stage_index < 4) && (update_mask_rgb & (1 << stage_index));
+        }
+
+        bool TevStageUpdatesCombinerBufferAlpha(unsigned stage_index) const {
+            return (stage_index < 4) && (update_mask_a & (1 << stage_index));
+        }
+    } tev_combiner_buffer_input;
+    
+    INSERT_PADDING_WORDS(0xf);
     TevStageConfig tev_stage4;
     INSERT_PADDING_WORDS(0x3);
     TevStageConfig tev_stage5;
-    INSERT_PADDING_WORDS(0x3);
+
+    union {
+        BitField< 0, 8, u32> r;
+        BitField< 8, 8, u32> g;
+        BitField<16, 8, u32> b;
+        BitField<24, 8, u32> a;
+    } tev_combiner_buffer_color;
+
+    INSERT_PADDING_WORDS(0x2);
 
     const std::array<Regs::TevStageConfig,6> GetTevStages() const {
         return { tev_stage0, tev_stage1,
@@ -426,9 +463,7 @@ struct Regs {
         D24S8  = 3
     };
 
-    /*
-     * Returns the number of bytes in the specified depth format
-     */
+    // Returns the number of bytes in the specified depth format
     static u32 BytesPerDepthPixel(DepthFormat format) {
         switch (format) {
         case DepthFormat::D16:
@@ -443,6 +478,20 @@ struct Regs {
         }
     }
 
+    // Returns the number of bits per depth component of the specified depth format
+    static u32 DepthBitsPerPixel(DepthFormat format) {
+        switch (format) {
+        case DepthFormat::D16:
+            return 16;
+        case DepthFormat::D24:
+        case DepthFormat::D24S8:
+            return 24;
+        default:
+            LOG_CRITICAL(HW_GPU, "Unknown depth format %u", format);
+            UNIMPLEMENTED();
+        }
+    }
+
     struct {
         // Components are laid out in reverse byte order, most significant bits first.
         enum ColorFormat : u32 {
@@ -784,8 +833,10 @@ struct Regs {
         ADD_FIELD(tev_stage1);
         ADD_FIELD(tev_stage2);
         ADD_FIELD(tev_stage3);
+        ADD_FIELD(tev_combiner_buffer_input);
         ADD_FIELD(tev_stage4);
         ADD_FIELD(tev_stage5);
+        ADD_FIELD(tev_combiner_buffer_color);
         ADD_FIELD(output_merger);
         ADD_FIELD(framebuffer);
         ADD_FIELD(vertex_attributes);
@@ -859,8 +910,10 @@ ASSERT_REG_POSITION(tev_stage0, 0xc0);
 ASSERT_REG_POSITION(tev_stage1, 0xc8);
 ASSERT_REG_POSITION(tev_stage2, 0xd0);
 ASSERT_REG_POSITION(tev_stage3, 0xd8);
+ASSERT_REG_POSITION(tev_combiner_buffer_input, 0xe0);
 ASSERT_REG_POSITION(tev_stage4, 0xf0);
 ASSERT_REG_POSITION(tev_stage5, 0xf8);
+ASSERT_REG_POSITION(tev_combiner_buffer_color, 0xfd);
 ASSERT_REG_POSITION(output_merger, 0x100);
 ASSERT_REG_POSITION(framebuffer, 0x110);
 ASSERT_REG_POSITION(vertex_attributes, 0x200);
diff --git a/src/video_core/rasterizer.cpp b/src/video_core/rasterizer.cpp
index 3b3fef484..46a326bb4 100644
--- a/src/video_core/rasterizer.cpp
+++ b/src/video_core/rasterizer.cpp
@@ -90,7 +90,7 @@ static const Math::Vec4<u8> GetPixel(int x, int y) {
         UNIMPLEMENTED();
     }
 
-    return {};
+    return {0, 0, 0, 0};
 }
 
 static u32 GetDepth(int x, int y) {
@@ -376,7 +376,13 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0,
             // with some basic arithmetic. Alpha combiners can be configured separately but work
             // analogously.
             Math::Vec4<u8> combiner_output;
-            for (const auto& tev_stage : tev_stages) {
+            Math::Vec4<u8> combiner_buffer = {
+                registers.tev_combiner_buffer_color.r, registers.tev_combiner_buffer_color.g,
+                registers.tev_combiner_buffer_color.b, registers.tev_combiner_buffer_color.a
+            };
+
+            for (unsigned tev_stage_index = 0; tev_stage_index < tev_stages.size(); ++tev_stage_index) {
+                const auto& tev_stage = tev_stages[tev_stage_index];
                 using Source = Regs::TevStageConfig::Source;
                 using ColorModifier = Regs::TevStageConfig::ColorModifier;
                 using AlphaModifier = Regs::TevStageConfig::AlphaModifier;
@@ -398,6 +404,9 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0,
                     case Source::Texture2:
                         return texture_color[2];
 
+                    case Source::PreviousBuffer:
+                        return combiner_buffer;
+
                     case Source::Constant:
                         return {tev_stage.const_r, tev_stage.const_g, tev_stage.const_b, tev_stage.const_a};
 
@@ -407,7 +416,7 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0,
                     default:
                         LOG_ERROR(HW_GPU, "Unknown color combiner source %d\n", (int)source);
                         UNIMPLEMENTED();
-                        return {};
+                        return {0, 0, 0, 0};
                     }
                 };
 
@@ -490,6 +499,16 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0,
                         return result.Cast<u8>();
                     }
 
+                    case Operation::AddSigned:
+                    {
+                        // TODO(bunnei): Verify that the color conversion from (float) 0.5f to (byte) 128 is correct
+                        auto result = input[0].Cast<int>() + input[1].Cast<int>() - Math::MakeVec<int>(128, 128, 128);
+                        result.r() = MathUtil::Clamp<int>(result.r(), 0, 255);
+                        result.g() = MathUtil::Clamp<int>(result.g(), 0, 255);
+                        result.b() = MathUtil::Clamp<int>(result.b(), 0, 255);
+                        return result.Cast<u8>();
+                    }
+
                     case Operation::Lerp:
                         return ((input[0] * input[2] + input[1] * (Math::MakeVec<u8>(255, 255, 255) - input[2]).Cast<u8>()) / 255).Cast<u8>();
 
@@ -524,7 +543,7 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0,
                     default:
                         LOG_ERROR(HW_GPU, "Unknown color combiner operation %d\n", (int)op);
                         UNIMPLEMENTED();
-                        return {};
+                        return {0, 0, 0};
                     }
                 };
 
@@ -578,7 +597,20 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0,
                 };
                 auto alpha_output = AlphaCombine(tev_stage.alpha_op, alpha_result);
 
-                combiner_output = Math::MakeVec(color_output, alpha_output);
+                combiner_output[0] = std::min((unsigned)255, color_output.r() * tev_stage.GetColorMultiplier());
+                combiner_output[1] = std::min((unsigned)255, color_output.g() * tev_stage.GetColorMultiplier());
+                combiner_output[2] = std::min((unsigned)255, color_output.b() * tev_stage.GetColorMultiplier());
+                combiner_output[3] = std::min((unsigned)255, alpha_output * tev_stage.GetAlphaMultiplier());
+
+                if (registers.tev_combiner_buffer_input.TevStageUpdatesCombinerBufferColor(tev_stage_index)) {
+                    combiner_buffer.r() = combiner_output.r();
+                    combiner_buffer.g() = combiner_output.g();
+                    combiner_buffer.b() = combiner_output.b();
+                }
+
+                if (registers.tev_combiner_buffer_input.TevStageUpdatesCombinerBufferAlpha(tev_stage_index)) {
+                    combiner_buffer.a() = combiner_output.a();
+                }
             }
 
             if (registers.output_merger.alpha_test.enable) {
@@ -624,9 +656,10 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0,
 
             // TODO: Does depth indeed only get written even if depth testing is enabled?
             if (registers.output_merger.depth_test_enable) {
-                u16 z = (u16)((v0.screenpos[2].ToFloat32() * w0 +
-                            v1.screenpos[2].ToFloat32() * w1 +
-                            v2.screenpos[2].ToFloat32() * w2) * 65535.f / wsum);
+                unsigned num_bits = Pica::Regs::DepthBitsPerPixel(registers.framebuffer.depth_format);
+                u32 z = (u32)((v0.screenpos[2].ToFloat32() * w0 +
+                               v1.screenpos[2].ToFloat32() * w1 +
+                               v2.screenpos[2].ToFloat32() * w2) * ((1 << num_bits) - 1) / wsum);
                 u32 ref_z = GetDepth(x >> 4, y >> 4);
 
                 bool pass = false;
diff --git a/src/video_core/vertex_shader.cpp b/src/video_core/vertex_shader.cpp
index 51f4e58bf..885b7de59 100644
--- a/src/video_core/vertex_shader.cpp
+++ b/src/video_core/vertex_shader.cpp
@@ -235,6 +235,15 @@ static void ProcessShaderCode(VertexShaderState& state) {
                 break;
             }
 
+            case OpCode::Id::FLR:
+                for (int i = 0; i < 4; ++i) {
+                    if (!swizzle.DestComponentEnabled(i))
+                        continue;
+
+                    dest[i] = float24::FromFloat32(std::floor(src1[i].ToFloat32()));
+                }
+                break;
+
             case OpCode::Id::MAX:
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
@@ -366,12 +375,15 @@ static void ProcessShaderCode(VertexShaderState& state) {
 
         case OpCode::Type::MultiplyAdd:
         {
-            if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MAD) {
+            if ((instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MAD) || 
+                (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI)) {
                 const SwizzlePattern& swizzle = *(SwizzlePattern*)&swizzle_data[instr.mad.operand_desc_id];
 
-                const float24* src1_ = LookupSourceRegister(instr.mad.src1);
-                const float24* src2_ = LookupSourceRegister(instr.mad.src2);
-                const float24* src3_ = LookupSourceRegister(instr.mad.src3);
+                bool is_inverted = (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI);
+
+                const float24* src1_ = LookupSourceRegister(instr.mad.GetSrc1(is_inverted));
+                const float24* src2_ = LookupSourceRegister(instr.mad.GetSrc2(is_inverted));
+                const float24* src3_ = LookupSourceRegister(instr.mad.GetSrc3(is_inverted));
 
                 const bool negate_src1 = ((bool)swizzle.negate_src1 != false);
                 const bool negate_src2 = ((bool)swizzle.negate_src2 != false);