GPU: Added Z buffer registers to Maxwell3D's reg structure.

GPU: Added TIC and TSC registers to the Maxwell3D register structure.

src/core/arm/arm_interface.h

@@ -25,22 +25,18 @@ public:
         VAddr tls_address;
     };
 
-    /**
-     * Runs the CPU for the given number of instructions
-     * @param num_instructions Number of instructions to run
-     */
-    void Run(int num_instructions) {
-        ExecuteInstructions(num_instructions);
-        this->num_instructions += num_instructions;
-    }
+    /// Runs the CPU until an event happens
+    virtual void Run() = 0;
 
     /// Step CPU by one instruction
-    void Step() {
-        Run(1);
-    }
+    virtual void Step() = 0;
 
+    /// Maps a backing memory region for the CPU
     virtual void MapBackingMemory(VAddr address, size_t size, u8* memory,
-                                  Kernel::VMAPermission perms) {}
+                                  Kernel::VMAPermission perms) = 0;
+
+    /// Unmaps a region of memory that was previously mapped using MapBackingMemory
+    virtual void UnmapMemory(VAddr address, size_t size) = 0;
 
     /// Clear all instruction cache
     virtual void ClearInstructionCache() = 0;
@@ -122,19 +118,4 @@ public:
 
     /// Prepare core for thread reschedule (if needed to correctly handle state)
     virtual void PrepareReschedule() = 0;
-
-    /// Getter for num_instructions
-    u64 GetNumInstructions() const {
-        return num_instructions;
-    }
-
-protected:
-    /**
-     * Executes the given number of instructions
-     * @param num_instructions Number of instructions to executes
-     */
-    virtual void ExecuteInstructions(int num_instructions) = 0;
-
-private:
-    u64 num_instructions = 0; ///< Number of instructions executed
 };

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -122,11 +122,22 @@ std::unique_ptr<Dynarmic::A64::Jit> MakeJit(const std::unique_ptr<ARM_Dynarmic_C
     return std::make_unique<Dynarmic::A64::Jit>(config);
 }
 
+void ARM_Dynarmic::Run() {
+    ASSERT(Memory::GetCurrentPageTable() == current_page_table);
+
+    jit->Run();
+}
+
+void ARM_Dynarmic::Step() {
+    cb->InterpreterFallback(jit->GetPC(), 1);
+}
+
 ARM_Dynarmic::ARM_Dynarmic()
     : cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), jit(MakeJit(cb)) {
     ARM_Interface::ThreadContext ctx;
     inner_unicorn.SaveContext(ctx);
     LoadContext(ctx);
+    PageTableChanged();
 }
 
 ARM_Dynarmic::~ARM_Dynarmic() = default;
@@ -136,6 +147,10 @@ void ARM_Dynarmic::MapBackingMemory(u64 address, size_t size, u8* memory,
     inner_unicorn.MapBackingMemory(address, size, memory, perms);
 }
 
+void ARM_Dynarmic::UnmapMemory(u64 address, size_t size) {
+    inner_unicorn.UnmapMemory(address, size);
+}
+
 void ARM_Dynarmic::SetPC(u64 pc) {
     jit->SetPC(pc);
 }
@@ -185,13 +200,6 @@ void ARM_Dynarmic::SetTlsAddress(u64 address) {
     cb->tpidrro_el0 = address;
 }
 
-void ARM_Dynarmic::ExecuteInstructions(int num_instructions) {
-    cb->ticks_remaining = num_instructions;
-    jit->Run();
-    CoreTiming::AddTicks(num_instructions - cb->num_interpreted_instructions);
-    cb->num_interpreted_instructions = 0;
-}
-
 void ARM_Dynarmic::SaveContext(ARM_Interface::ThreadContext& ctx) {
     ctx.cpu_registers = jit->GetRegisters();
     ctx.sp = jit->GetSP();
@@ -224,4 +232,5 @@ void ARM_Dynarmic::ClearInstructionCache() {
 
 void ARM_Dynarmic::PageTableChanged() {
     jit = MakeJit(cb);
+    current_page_table = Memory::GetCurrentPageTable();
 }

src/core/arm/dynarmic/arm_dynarmic.h

@@ -19,7 +19,7 @@ public:
 
     void MapBackingMemory(VAddr address, size_t size, u8* memory,
                           Kernel::VMAPermission perms) override;
-
+    void UnmapMemory(u64 address, size_t size) override;
     void SetPC(u64 pc) override;
     u64 GetPC() const override;
     u64 GetReg(int index) const override;
@@ -29,6 +29,8 @@ public:
     u32 GetVFPReg(int index) const override;
     void SetVFPReg(int index, u32 value) override;
     u32 GetCPSR() const override;
+    void Run() override;
+    void Step() override;
     void SetCPSR(u32 cpsr) override;
     VAddr GetTlsAddress() const override;
     void SetTlsAddress(VAddr address) override;
@@ -37,7 +39,6 @@ public:
     void LoadContext(const ThreadContext& ctx) override;
 
     void PrepareReschedule() override;
-    void ExecuteInstructions(int num_instructions) override;
 
     void ClearInstructionCache() override;
     void PageTableChanged() override;
@@ -47,4 +48,6 @@ private:
     std::unique_ptr<ARM_Dynarmic_Callbacks> cb;
     std::unique_ptr<Dynarmic::A64::Jit> jit;
     ARM_Unicorn inner_unicorn;
+
+    Memory::PageTable* current_page_table = nullptr;
 };

src/core/arm/unicorn/arm_unicorn.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <unicorn/arm64.h>
 #include "common/assert.h"
 #include "common/microprofile.h"
@@ -77,6 +78,10 @@ void ARM_Unicorn::MapBackingMemory(VAddr address, size_t size, u8* memory,
     CHECKED(uc_mem_map_ptr(uc, address, size, static_cast<u32>(perms), memory));
 }
 
+void ARM_Unicorn::UnmapMemory(VAddr address, size_t size) {
+    CHECKED(uc_mem_unmap(uc, address, size));
+}
+
 void ARM_Unicorn::SetPC(u64 pc) {
     CHECKED(uc_reg_write(uc, UC_ARM64_REG_PC, &pc));
 }
@@ -149,6 +154,14 @@ void ARM_Unicorn::SetTlsAddress(VAddr base) {
     CHECKED(uc_reg_write(uc, UC_ARM64_REG_TPIDRRO_EL0, &base));
 }
 
+void ARM_Unicorn::Run() {
+    ExecuteInstructions(std::max(CoreTiming::GetDowncount(), 0));
+}
+
+void ARM_Unicorn::Step() {
+    ExecuteInstructions(1);
+}
+
 MICROPROFILE_DEFINE(ARM_Jit, "ARM JIT", "ARM JIT", MP_RGB(255, 64, 64));
 
 void ARM_Unicorn::ExecuteInstructions(int num_instructions) {

src/core/arm/unicorn/arm_unicorn.h

@@ -14,6 +14,7 @@ public:
     ~ARM_Unicorn();
     void MapBackingMemory(VAddr address, size_t size, u8* memory,
                           Kernel::VMAPermission perms) override;
+    void UnmapMemory(VAddr address, size_t size) override;
     void SetPC(u64 pc) override;
     u64 GetPC() const override;
     u64 GetReg(int index) const override;
@@ -29,7 +30,9 @@ public:
     void SaveContext(ThreadContext& ctx) override;
     void LoadContext(const ThreadContext& ctx) override;
     void PrepareReschedule() override;
-    void ExecuteInstructions(int num_instructions) override;
+    void ExecuteInstructions(int num_instructions);
+    void Run() override;
+    void Step() override;
     void ClearInstructionCache() override;
     void PageTableChanged() override{};
 

src/core/core.cpp

@@ -26,7 +26,7 @@ namespace Core {
 
 /*static*/ System System::s_instance;
 
-System::ResultStatus System::RunLoop(int tight_loop) {
+System::ResultStatus System::RunLoop(bool tight_loop) {
     status = ResultStatus::Success;
     if (!cpu_core) {
         return ResultStatus::ErrorNotInitialized;
@@ -40,7 +40,7 @@ System::ResultStatus System::RunLoop(int tight_loop) {
         if (GDBStub::GetCpuHaltFlag()) {
             if (GDBStub::GetCpuStepFlag()) {
                 GDBStub::SetCpuStepFlag(false);
-                tight_loop = 1;
+                tight_loop = false;
             } else {
                 return ResultStatus::Success;
             }
@@ -56,7 +56,11 @@ System::ResultStatus System::RunLoop(int tight_loop) {
         PrepareReschedule();
     } else {
         CoreTiming::Advance();
-        cpu_core->Run(tight_loop);
+        if (tight_loop) {
+            cpu_core->Run();
+        } else {
+            cpu_core->Step();
+        }
     }
 
     HW::Update();
@@ -66,7 +70,7 @@ System::ResultStatus System::RunLoop(int tight_loop) {
 }
 
 System::ResultStatus System::SingleStep() {
-    return RunLoop(1);
+    return RunLoop(false);
 }
 
 System::ResultStatus System::Load(EmuWindow* emu_window, const std::string& filepath) {

src/core/core.h

@@ -53,10 +53,10 @@ public:
      * is not required to do a full dispatch with each instruction. NOTE: the number of instructions
      * requested is not guaranteed to run, as this will be interrupted preemptively if a hardware
      * update is requested (e.g. on a thread switch).
-     * @param tight_loop Number of instructions to execute.
+     * @param tight_loop If false, the CPU single-steps.
      * @return Result status, indicating whether or not the operation succeeded.
      */
-    ResultStatus RunLoop(int tight_loop = 100000);
+    ResultStatus RunLoop(bool tight_loop = true);
 
     /**
      * Step the CPU one instruction

src/core/hle/kernel/hle_ipc.cpp

@@ -7,6 +7,7 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/event.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/kernel/kernel.h"
@@ -26,6 +27,32 @@ void SessionRequestHandler::ClientDisconnected(SharedPtr<ServerSession> server_s
     boost::range::remove_erase(connected_sessions, server_session);
 }
 
+SharedPtr<Event> HLERequestContext::SleepClientThread(SharedPtr<Thread> thread,
+                                                      const std::string& reason, u64 timeout,
+                                                      WakeupCallback&& callback) {
+
+    // Put the client thread to sleep until the wait event is signaled or the timeout expires.
+    thread->wakeup_callback =
+        [context = *this, callback](ThreadWakeupReason reason, SharedPtr<Thread> thread,
+                                    SharedPtr<WaitObject> object, size_t index) mutable -> bool {
+        ASSERT(thread->status == THREADSTATUS_WAIT_HLE_EVENT);
+        callback(thread, context, reason);
+        context.WriteToOutgoingCommandBuffer(*thread);
+        return true;
+    };
+
+    auto event = Kernel::Event::Create(Kernel::ResetType::OneShot, "HLE Pause Event: " + reason);
+    thread->status = THREADSTATUS_WAIT_HLE_EVENT;
+    thread->wait_objects = {event};
+    event->AddWaitingThread(thread);
+
+    if (timeout > 0) {
+        thread->WakeAfterDelay(timeout);
+    }
+
+    return event;
+}
+
 HLERequestContext::HLERequestContext(SharedPtr<Kernel::ServerSession> server_session)
     : server_session(std::move(server_session)) {
     cmd_buf[0] = 0;
@@ -35,7 +62,7 @@ HLERequestContext::~HLERequestContext() = default;
 
 void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
     IPC::RequestParser rp(src_cmdbuf);
-    command_header = std::make_unique<IPC::CommandHeader>(rp.PopRaw<IPC::CommandHeader>());
+    command_header = std::make_shared<IPC::CommandHeader>(rp.PopRaw<IPC::CommandHeader>());
 
     if (command_header->type == IPC::CommandType::Close) {
         // Close does not populate the rest of the IPC header
@@ -45,7 +72,7 @@ void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
     // If handle descriptor is present, add size of it
     if (command_header->enable_handle_descriptor) {
         handle_descriptor_header =
-            std::make_unique<IPC::HandleDescriptorHeader>(rp.PopRaw<IPC::HandleDescriptorHeader>());
+            std::make_shared<IPC::HandleDescriptorHeader>(rp.PopRaw<IPC::HandleDescriptorHeader>());
         if (handle_descriptor_header->send_current_pid) {
             rp.Skip(2, false);
         }
@@ -88,7 +115,7 @@ void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
         // All outgoing domain messages have the domain header, if only incoming has it
         if (incoming || domain_message_header) {
             domain_message_header =
-                std::make_unique<IPC::DomainMessageHeader>(rp.PopRaw<IPC::DomainMessageHeader>());
+                std::make_shared<IPC::DomainMessageHeader>(rp.PopRaw<IPC::DomainMessageHeader>());
         } else {
             if (Session()->IsDomain())
                 LOG_WARNING(IPC, "Domain request has no DomainMessageHeader!");
@@ -96,7 +123,7 @@ void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
     }
 
     data_payload_header =
-        std::make_unique<IPC::DataPayloadHeader>(rp.PopRaw<IPC::DataPayloadHeader>());
+        std::make_shared<IPC::DataPayloadHeader>(rp.PopRaw<IPC::DataPayloadHeader>());
 
     data_payload_offset = rp.GetCurrentOffset();
 
@@ -159,8 +186,11 @@ ResultCode HLERequestContext::PopulateFromIncomingCommandBuffer(u32_le* src_cmdb
     return RESULT_SUCCESS;
 }
 
-ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(u32_le* dst_cmdbuf, Process& dst_process,
-                                                           HandleTable& dst_table) {
+ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(Thread& thread) {
+    std::array<u32, IPC::COMMAND_BUFFER_LENGTH> dst_cmdbuf;
+    Memory::ReadBlock(*thread.owner_process, thread.GetTLSAddress(), dst_cmdbuf.data(),
+                      dst_cmdbuf.size() * sizeof(u32));
+
     // The header was already built in the internal command buffer. Attempt to parse it to verify
     // the integrity and then copy it over to the target command buffer.
     ParseCommandBuffer(cmd_buf.data(), false);
@@ -171,7 +201,7 @@ ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(u32_le* dst_cmdbuf, P
     if (domain_message_header)
         size -= sizeof(IPC::DomainMessageHeader) / sizeof(u32);
 
-    std::copy_n(cmd_buf.begin(), size, dst_cmdbuf);
+    std::copy_n(cmd_buf.begin(), size, dst_cmdbuf.data());
 
     if (command_header->enable_handle_descriptor) {
         ASSERT_MSG(!move_objects.empty() || !copy_objects.empty(),
@@ -213,6 +243,11 @@ ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(u32_le* dst_cmdbuf, P
             dst_cmdbuf[domain_offset++] = static_cast<u32_le>(request_handlers.size());
         }
     }
+
+    // Copy the translated command buffer back into the thread's command buffer area.
+    Memory::WriteBlock(*thread.owner_process, thread.GetTLSAddress(), dst_cmdbuf.data(),
+                       dst_cmdbuf.size() * sizeof(u32));
+
     return RESULT_SUCCESS;
 }
 

src/core/hle/kernel/hle_ipc.h

@@ -6,6 +6,7 @@
 
 #include <array>
 #include <memory>
+#include <string>
 #include <vector>
 #include <boost/container/small_vector.hpp>
 #include "common/common_types.h"
@@ -13,6 +14,7 @@
 #include "core/hle/ipc.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/thread.h"
 
 namespace Service {
 class ServiceFrameworkBase;
@@ -24,6 +26,7 @@ class Domain;
 class HandleTable;
 class HLERequestContext;
 class Process;
+class Event;
 
 /**
  * Interface implemented by HLE Session handlers.
@@ -102,14 +105,31 @@ public:
         return server_session;
     }
 
+    using WakeupCallback = std::function<void(SharedPtr<Thread> thread, HLERequestContext& context,
+                                              ThreadWakeupReason reason)>;
+
+    /**
+     * Puts the specified guest thread to sleep until the returned event is signaled or until the
+     * specified timeout expires.
+     * @param thread Thread to be put to sleep.
+     * @param reason Reason for pausing the thread, to be used for debugging purposes.
+     * @param timeout Timeout in nanoseconds after which the thread will be awoken and the callback
+     * invoked with a Timeout reason.
+     * @param callback Callback to be invoked when the thread is resumed. This callback must write
+     * the entire command response once again, regardless of the state of it before this function
+     * was called.
+     * @returns Event that when signaled will resume the thread and call the callback function.
+     */
+    SharedPtr<Event> SleepClientThread(SharedPtr<Thread> thread, const std::string& reason,
+                                       u64 timeout, WakeupCallback&& callback);
+
     void ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming);
 
     /// Populates this context with data from the requesting process/thread.
     ResultCode PopulateFromIncomingCommandBuffer(u32_le* src_cmdbuf, Process& src_process,
                                                  HandleTable& src_table);
     /// Writes data from this context back to the requesting process/thread.
-    ResultCode WriteToOutgoingCommandBuffer(u32_le* dst_cmdbuf, Process& dst_process,
-                                            HandleTable& dst_table);
+    ResultCode WriteToOutgoingCommandBuffer(Thread& thread);
 
     u32_le GetCommand() const {
         return command;
@@ -139,7 +159,7 @@ public:
         return buffer_c_desciptors;
     }
 
-    const std::unique_ptr<IPC::DomainMessageHeader>& GetDomainMessageHeader() const {
+    const std::shared_ptr<IPC::DomainMessageHeader>& GetDomainMessageHeader() const {
         return domain_message_header;
     }
 
@@ -212,10 +232,10 @@ private:
     boost::container::small_vector<SharedPtr<Object>, 8> copy_objects;
     boost::container::small_vector<std::shared_ptr<SessionRequestHandler>, 8> domain_objects;
 
-    std::unique_ptr<IPC::CommandHeader> command_header;
-    std::unique_ptr<IPC::HandleDescriptorHeader> handle_descriptor_header;
-    std::unique_ptr<IPC::DataPayloadHeader> data_payload_header;
-    std::unique_ptr<IPC::DomainMessageHeader> domain_message_header;
+    std::shared_ptr<IPC::CommandHeader> command_header;
+    std::shared_ptr<IPC::HandleDescriptorHeader> handle_descriptor_header;
+    std::shared_ptr<IPC::DataPayloadHeader> data_payload_header;
+    std::shared_ptr<IPC::DomainMessageHeader> domain_message_header;
     std::vector<IPC::BufferDescriptorX> buffer_x_desciptors;
     std::vector<IPC::BufferDescriptorABW> buffer_a_desciptors;
     std::vector<IPC::BufferDescriptorABW> buffer_b_desciptors;

src/core/hle/kernel/kernel.h

@@ -33,10 +33,6 @@ enum class HandleType : u32 {
     ServerSession,
 };
 
-enum {
-    DEFAULT_STACK_SIZE = 0x10000,
-};
-
 enum class ResetType {
     OneShot,
     Sticky,

src/core/hle/kernel/process.cpp

@@ -117,11 +117,12 @@ void Process::ParseKernelCaps(const u32* kernel_caps, size_t len) {
 }
 
 void Process::Run(VAddr entry_point, s32 main_thread_priority, u32 stack_size) {
-    // Allocate and map stack
+    // Allocate and map the main thread stack
+    // TODO(bunnei): This is heap area that should be allocated by the kernel and not mapped as part
+    // of the user address space.
     vm_manager
-        .MapMemoryBlock(Memory::HEAP_VADDR_END - stack_size,
-                        std::make_shared<std::vector<u8>>(stack_size, 0), 0, stack_size,
-                        MemoryState::Heap)
+        .MapMemoryBlock(Memory::STACK_VADDR, std::make_shared<std::vector<u8>>(stack_size, 0), 0,
+                        stack_size, MemoryState::Mapped)
         .Unwrap();
     misc_memory_used += stack_size;
     memory_region->used += stack_size;
@@ -153,9 +154,9 @@ void Process::LoadModule(SharedPtr<CodeSet> module_, VAddr base_addr) {
     };
 
     // Map CodeSet segments
-    MapSegment(module_->code, VMAPermission::ReadExecute, MemoryState::Code);
-    MapSegment(module_->rodata, VMAPermission::Read, MemoryState::Static);
-    MapSegment(module_->data, VMAPermission::ReadWrite, MemoryState::Static);
+    MapSegment(module_->code, VMAPermission::ReadExecute, MemoryState::CodeStatic);
+    MapSegment(module_->rodata, VMAPermission::Read, MemoryState::CodeMutable);
+    MapSegment(module_->data, VMAPermission::ReadWrite, MemoryState::CodeMutable);
 }
 
 VAddr Process::GetLinearHeapAreaAddress() const {
@@ -182,6 +183,8 @@ ResultVal<VAddr> Process::HeapAllocate(VAddr target, u64 size, VMAPermission per
         // Initialize heap
         heap_memory = std::make_shared<std::vector<u8>>();
         heap_start = heap_end = target;
+    } else {
+        vm_manager.UnmapRange(heap_start, heap_end - heap_start);
     }
 
     // If necessary, expand backing vector to cover new heap extents.
@@ -201,7 +204,7 @@ ResultVal<VAddr> Process::HeapAllocate(VAddr target, u64 size, VMAPermission per
                                                        size, MemoryState::Heap));
     vm_manager.Reprotect(vma, perms);
 
-    heap_used += size;
+    heap_used = size;
     memory_region->used += size;
 
     return MakeResult<VAddr>(heap_end - size);
@@ -288,7 +291,7 @@ ResultCode Process::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size) {
 
     CASCADE_RESULT(auto new_vma,
                    vm_manager.MapMemoryBlock(dst_addr, backing_block, backing_block_offset, size,
-                                             vma->second.meminfo_state));
+                                             MemoryState::Mapped));
     // Protect mirror with permissions from old region
     vm_manager.Reprotect(new_vma, vma->second.permissions);
     // Remove permissions from old region

src/core/hle/kernel/svc.cpp

@@ -317,13 +317,13 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
         *result = Core::CurrentProcess()->allowed_thread_priority_mask;
         break;
     case GetInfoType::MapRegionBaseAddr:
-        *result = vm_manager.GetMapRegionBaseAddr();
+        *result = Memory::MAP_REGION_VADDR;
         break;
     case GetInfoType::MapRegionSize:
-        *result = vm_manager.GetAddressSpaceSize();
+        *result = Memory::MAP_REGION_SIZE;
         break;
     case GetInfoType::HeapRegionBaseAddr:
-        *result = vm_manager.GetNewMapRegionBaseAddr() + vm_manager.GetNewMapRegionSize();
+        *result = Memory::HEAP_VADDR;
         break;
     case GetInfoType::HeapRegionSize:
         *result = Memory::HEAP_SIZE;
@@ -347,10 +347,10 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
         *result = vm_manager.GetAddressSpaceSize();
         break;
     case GetInfoType::NewMapRegionBaseAddr:
-        *result = vm_manager.GetNewMapRegionBaseAddr();
+        *result = Memory::NEW_MAP_REGION_VADDR;
         break;
     case GetInfoType::NewMapRegionSize:
-        *result = vm_manager.GetNewMapRegionSize();
+        *result = Memory::NEW_MAP_REGION_SIZE;
         break;
     case GetInfoType::IsVirtualAddressMemoryEnabled:
         *result = Core::CurrentProcess()->is_virtual_address_memory_enabled;
@@ -468,7 +468,7 @@ static ResultCode QueryProcessMemory(MemoryInfo* memory_info, PageInfo* /*page_i
         memory_info->base_address = 0;
         memory_info->permission = static_cast<u32>(VMAPermission::None);
         memory_info->size = 0;
-        memory_info->type = static_cast<u32>(MemoryState::Free);
+        memory_info->type = static_cast<u32>(MemoryState::Unmapped);
     } else {
         memory_info->base_address = vma->second.base;
         memory_info->permission = static_cast<u32>(vma->second.permissions);

src/core/hle/kernel/thread.cpp

@@ -55,16 +55,6 @@ inline static u32 const NewThreadId() {
 Thread::Thread() {}
 Thread::~Thread() {}
 
-/**
- * Check if the specified thread is waiting on the specified address to be arbitrated
- * @param thread The thread to test
- * @param wait_address The address to test against
- * @return True if the thread is waiting, false otherwise
- */
-static bool CheckWait_AddressArbiter(const Thread* thread, VAddr wait_address) {
-    return thread->status == THREADSTATUS_WAIT_ARB && wait_address == thread->wait_address;
-}
-
 void Thread::Stop() {
     // Cancel any outstanding wakeup events for this thread
     CoreTiming::UnscheduleEvent(ThreadWakeupEventType, callback_handle);
@@ -102,12 +92,6 @@ void WaitCurrentThread_Sleep() {
     thread->status = THREADSTATUS_WAIT_SLEEP;
 }
 
-void WaitCurrentThread_ArbitrateAddress(VAddr wait_address) {
-    Thread* thread = GetCurrentThread();
-    thread->wait_address = wait_address;
-    thread->status = THREADSTATUS_WAIT_ARB;
-}
-
 void ExitCurrentThread() {
     Thread* thread = GetCurrentThread();
     thread->Stop();
@@ -129,7 +113,8 @@ static void ThreadWakeupCallback(u64 thread_handle, int cycles_late) {
     bool resume = true;
 
     if (thread->status == THREADSTATUS_WAIT_SYNCH_ANY ||
-        thread->status == THREADSTATUS_WAIT_SYNCH_ALL || thread->status == THREADSTATUS_WAIT_ARB) {
+        thread->status == THREADSTATUS_WAIT_SYNCH_ALL ||
+        thread->status == THREADSTATUS_WAIT_HLE_EVENT) {
 
         // Remove the thread from each of its waiting objects' waitlists
         for (auto& object : thread->wait_objects)
@@ -163,7 +148,7 @@ void Thread::ResumeFromWait() {
     switch (status) {
     case THREADSTATUS_WAIT_SYNCH_ALL:
     case THREADSTATUS_WAIT_SYNCH_ANY:
-    case THREADSTATUS_WAIT_ARB:
+    case THREADSTATUS_WAIT_HLE_EVENT:
     case THREADSTATUS_WAIT_SLEEP:
     case THREADSTATUS_WAIT_IPC:
         break;
@@ -314,7 +299,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point,
         // TODO(Subv): Find the correct MemoryState for this region.
         vm_manager.MapMemoryBlock(Memory::TLS_AREA_VADDR + available_page * Memory::PAGE_SIZE,
                                   linheap_memory, offset, Memory::PAGE_SIZE,
-                                  MemoryState::ThreadLocalStorage);
+                                  MemoryState::ThreadLocal);
     }
 
     // Mark the slot as used
@@ -357,7 +342,7 @@ SharedPtr<Thread> SetupMainThread(VAddr entry_point, u32 priority,
 
     // Initialize new "main" thread
     auto thread_res = Thread::Create("main", entry_point, priority, 0, THREADPROCESSORID_0,
-                                     Memory::HEAP_VADDR_END, owner_process);
+                                     Memory::STACK_VADDR_END, owner_process);
 
     SharedPtr<Thread> thread = std::move(thread_res).Unwrap();
 

src/core/hle/kernel/thread.h

@@ -38,7 +38,7 @@ enum ThreadProcessorId : s32 {
 enum ThreadStatus {
     THREADSTATUS_RUNNING,        ///< Currently running
     THREADSTATUS_READY,          ///< Ready to run
-    THREADSTATUS_WAIT_ARB,       ///< Waiting on an address arbiter
+    THREADSTATUS_WAIT_HLE_EVENT, ///< Waiting for hle event to finish
     THREADSTATUS_WAIT_SLEEP,     ///< Waiting due to a SleepThread SVC
     THREADSTATUS_WAIT_IPC,       ///< Waiting for the reply from an IPC request
     THREADSTATUS_WAIT_SYNCH_ANY, ///< Waiting due to WaitSynch1 or WaitSynchN with wait_all = false

src/core/hle/kernel/vm_manager.cpp

@@ -18,8 +18,26 @@ namespace Kernel {
 
 static const char* GetMemoryStateName(MemoryState state) {
     static const char* names[] = {
-        "Free",   "Reserved",   "IO",      "Static", "Code",      "Private",
-        "Shared", "Continuous", "Aliased", "Alias",  "AliasCode", "Locked",
+        "Unmapped",
+        "Io",
+        "Normal",
+        "CodeStatic",
+        "CodeMutable",
+        "Heap",
+        "Shared",
+        "Unknown1"
+        "ModuleCodeStatic",
+        "ModuleCodeMutable",
+        "IpcBuffer0",
+        "Mapped",
+        "ThreadLocal",
+        "TransferMemoryIsolated",
+        "TransferMemory",
+        "ProcessMemory",
+        "Unknown2"
+        "IpcBuffer1",
+        "IpcBuffer3",
+        "KernelStack",
     };
 
     return names[(int)state];
@@ -142,7 +160,7 @@ VMManager::VMAIter VMManager::Unmap(VMAIter vma_handle) {
     VirtualMemoryArea& vma = vma_handle->second;
     vma.type = VMAType::Free;
     vma.permissions = VMAPermission::None;
-    vma.meminfo_state = MemoryState::Free;
+    vma.meminfo_state = MemoryState::Unmapped;
 
     vma.backing_block = nullptr;
     vma.offset = 0;
@@ -166,6 +184,9 @@ ResultCode VMManager::UnmapRange(VAddr target, u64 size) {
     }
 
     ASSERT(FindVMA(target)->second.size >= size);
+
+    Core::CPU().UnmapMemory(target, size);
+
     return RESULT_SUCCESS;
 }
 
@@ -377,19 +398,4 @@ u64 VMManager::GetAddressSpaceSize() {
     return MAX_ADDRESS;
 }
 
-VAddr VMManager::GetMapRegionBaseAddr() {
-    LOG_WARNING(Kernel, "(STUBBED) called");
-    return Memory::HEAP_VADDR;
-}
-
-VAddr VMManager::GetNewMapRegionBaseAddr() {
-    LOG_WARNING(Kernel, "(STUBBED) called");
-    return 0x8000000;
-}
-
-u64 VMManager::GetNewMapRegionSize() {
-    LOG_WARNING(Kernel, "(STUBBED) called");
-    return 0x8000000;
-}
-
 } // namespace Kernel

src/core/hle/kernel/vm_manager.h

@@ -41,15 +41,24 @@ enum class VMAPermission : u8 {
 
 /// Set of values returned in MemoryInfo.state by svcQueryMemory.
 enum class MemoryState : u32 {
-    Free = 0,
-    IO = 1,
-    Normal = 2,
-    Code = 3,
-    Static = 4,
-    Heap = 5,
-    Shared = 6,
-    Mapped = 6,
-    ThreadLocalStorage = 12,
+    Unmapped = 0x0,
+    Io = 0x1,
+    Normal = 0x2,
+    CodeStatic = 0x3,
+    CodeMutable = 0x4,
+    Heap = 0x5,
+    Shared = 0x6,
+    ModuleCodeStatic = 0x8,
+    ModuleCodeMutable = 0x9,
+    IpcBuffer0 = 0xA,
+    Mapped = 0xB,
+    ThreadLocal = 0xC,
+    TransferMemoryIsolated = 0xD,
+    TransferMemory = 0xE,
+    ProcessMemory = 0xF,
+    IpcBuffer1 = 0x11,
+    IpcBuffer3 = 0x12,
+    KernelStack = 0x13,
 };
 
 /**
@@ -66,7 +75,7 @@ struct VirtualMemoryArea {
     VMAType type = VMAType::Free;
     VMAPermission permissions = VMAPermission::None;
     /// Tag returned by svcQueryMemory. Not otherwise used.
-    MemoryState meminfo_state = MemoryState::Free;
+    MemoryState meminfo_state = MemoryState::Unmapped;
 
     // Settings for type = AllocatedMemoryBlock
     /// Memory block backing this VMA.
@@ -192,15 +201,6 @@ public:
     /// Gets the total address space address size, used by svcGetInfo
     u64 GetAddressSpaceSize();
 
-    /// Gets the map region base address, used by svcGetInfo
-    VAddr GetMapRegionBaseAddr();
-
-    /// Gets the base address for a new memory region, used by svcGetInfo
-    VAddr GetNewMapRegionBaseAddr();
-
-    /// Gets the size for a new memory region, used by svcGetInfo
-    u64 GetNewMapRegionSize();
-
     /// Each VMManager has its own page table, which is set as the main one when the owning process
     /// is scheduled.
     Memory::PageTable page_table;

src/core/hle/kernel/wait_object.cpp

@@ -39,7 +39,8 @@ SharedPtr<Thread> WaitObject::GetHighestPriorityReadyThread() {
     for (const auto& thread : waiting_threads) {
         // The list of waiting threads must not contain threads that are not waiting to be awakened.
         ASSERT_MSG(thread->status == THREADSTATUS_WAIT_SYNCH_ANY ||
-                       thread->status == THREADSTATUS_WAIT_SYNCH_ALL,
+                       thread->status == THREADSTATUS_WAIT_SYNCH_ALL ||
+                       thread->status == THREADSTATUS_WAIT_HLE_EVENT,
                    "Inconsistent thread statuses in waiting_threads");
 
         if (thread->current_priority >= candidate_priority)

src/core/hle/service/nifm/nifm.cpp

@@ -32,7 +32,7 @@ public:
             {0, &IRequest::GetRequestState, "GetRequestState"},
             {1, &IRequest::GetResult, "GetResult"},
             {2, &IRequest::GetSystemEventReadableHandles, "GetSystemEventReadableHandles"},
-            {3, nullptr, "Cancel"},
+            {3, &IRequest::Cancel, "Cancel"},
             {4, nullptr, "Submit"},
             {5, nullptr, "SetRequirement"},
             {6, nullptr, "SetRequirementPreset"},
@@ -80,6 +80,11 @@ private:
         rb.Push(RESULT_SUCCESS);
         rb.PushCopyObjects(event1, event2);
     }
+    void Cancel(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
 
     Kernel::SharedPtr<Kernel::Event> event1, event2;
 };
@@ -96,13 +101,56 @@ public:
     }
 };
 
+class IGeneralService final : public ServiceFramework<IGeneralService> {
+public:
+    IGeneralService();
+
+private:
+    void GetClientId(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        IPC::ResponseBuilder rb{ctx, 4};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u64>(0);
+    }
+    void CreateScanRequest(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+
+        rb.Push(RESULT_SUCCESS);
+        rb.PushIpcInterface<IScanRequest>();
+
+        LOG_DEBUG(Service_NIFM, "called");
+    }
+    void CreateRequest(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+
+        rb.Push(RESULT_SUCCESS);
+        rb.PushIpcInterface<IRequest>();
+
+        LOG_DEBUG(Service_NIFM, "called");
+    }
+    void RemoveNetworkProfile(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+    void CreateTemporaryNetworkProfile(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+
+        rb.Push(RESULT_SUCCESS);
+        rb.PushIpcInterface<INetworkProfile>();
+
+        LOG_DEBUG(Service_NIFM, "called");
+    }
+};
+
 IGeneralService::IGeneralService() : ServiceFramework("IGeneralService") {
     static const FunctionInfo functions[] = {
         {1, &IGeneralService::GetClientId, "GetClientId"},
         {2, &IGeneralService::CreateScanRequest, "CreateScanRequest"},
         {4, &IGeneralService::CreateRequest, "CreateRequest"},
-        {6, nullptr, "GetCurrentNetworkProfile"},
-        {7, nullptr, "EnumerateNetworkInterfaces"},
+        {5, nullptr, "GetCurrentNetworkProfile"},
+        {6, nullptr, "EnumerateNetworkInterfaces"},
+        {7, nullptr, "EnumerateNetworkProfiles"},
         {8, nullptr, "GetNetworkProfile"},
         {9, nullptr, "SetNetworkProfile"},
         {10, &IGeneralService::RemoveNetworkProfile, "RemoveNetworkProfile"},
@@ -137,50 +185,28 @@ IGeneralService::IGeneralService() : ServiceFramework("IGeneralService") {
     RegisterHandlers(functions);
 }
 
-void IGeneralService::GetClientId(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_NIFM, "(STUBBED) called");
-    IPC::ResponseBuilder rb{ctx, 4};
-    rb.Push(RESULT_SUCCESS);
-    rb.Push<u64>(0);
-}
-
-void IGeneralService::CreateScanRequest(Kernel::HLERequestContext& ctx) {
+void Module::Interface::CreateGeneralServiceOld(Kernel::HLERequestContext& ctx) {
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IScanRequest>();
-
+    rb.PushIpcInterface<IGeneralService>();
     LOG_DEBUG(Service_NIFM, "called");
 }
 
-void IGeneralService::CreateRequest(Kernel::HLERequestContext& ctx) {
+void Module::Interface::CreateGeneralService(Kernel::HLERequestContext& ctx) {
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IRequest>();
-
+    rb.PushIpcInterface<IGeneralService>();
     LOG_DEBUG(Service_NIFM, "called");
 }
 
-void IGeneralService::RemoveNetworkProfile(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_NIFM, "(STUBBED) called");
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(RESULT_SUCCESS);
-}
-
-void IGeneralService::CreateTemporaryNetworkProfile(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<INetworkProfile>();
-
-    LOG_DEBUG(Service_NIFM, "called");
-}
+Module::Interface::Interface(std::shared_ptr<Module> module, const char* name)
+    : ServiceFramework(name), module(std::move(module)) {}
 
 void InstallInterfaces(SM::ServiceManager& service_manager) {
-    std::make_shared<NIFM_A>()->InstallAsService(service_manager);
-    std::make_shared<NIFM_S>()->InstallAsService(service_manager);
-    std::make_shared<NIFM_U>()->InstallAsService(service_manager);
+    auto module = std::make_shared<Module>();
+    std::make_shared<NIFM_A>(module)->InstallAsService(service_manager);
+    std::make_shared<NIFM_S>(module)->InstallAsService(service_manager);
+    std::make_shared<NIFM_U>(module)->InstallAsService(service_manager);
 }
 
 } // namespace NIFM

src/core/hle/service/nifm/nifm.h

@@ -9,16 +9,18 @@
 namespace Service {
 namespace NIFM {
 
-class IGeneralService final : public ServiceFramework<IGeneralService> {
+class Module final {
 public:
-    IGeneralService();
+    class Interface : public ServiceFramework<Interface> {
+    public:
+        Interface(std::shared_ptr<Module> module, const char* name);
 
-private:
-    void GetClientId(Kernel::HLERequestContext& ctx);
-    void CreateScanRequest(Kernel::HLERequestContext& ctx);
-    void CreateRequest(Kernel::HLERequestContext& ctx);
-    void RemoveNetworkProfile(Kernel::HLERequestContext& ctx);
-    void CreateTemporaryNetworkProfile(Kernel::HLERequestContext& ctx);
+        void CreateGeneralServiceOld(Kernel::HLERequestContext& ctx);
+        void CreateGeneralService(Kernel::HLERequestContext& ctx);
+
+    protected:
+        std::shared_ptr<Module> module;
+    };
 };
 
 void InstallInterfaces(SM::ServiceManager& service_manager);

src/core/hle/service/nifm/nifm_a.cpp

@@ -2,29 +2,12 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "common/logging/log.h"
-#include "core/hle/ipc_helpers.h"
-#include "core/hle/service/nifm/nifm.h"
 #include "core/hle/service/nifm/nifm_a.h"
 
 namespace Service {
 namespace NIFM {
 
-void NIFM_A::CreateGeneralServiceOld(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IGeneralService>();
-    LOG_DEBUG(Service_NIFM, "called");
-}
-
-void NIFM_A::CreateGeneralService(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IGeneralService>();
-    LOG_DEBUG(Service_NIFM, "called");
-}
-
-NIFM_A::NIFM_A() : ServiceFramework("nifm:a") {
+NIFM_A::NIFM_A(std::shared_ptr<Module> module) : Module::Interface(std::move(module), "nifm:a") {
     static const FunctionInfo functions[] = {
         {4, &NIFM_A::CreateGeneralServiceOld, "CreateGeneralServiceOld"},
         {5, &NIFM_A::CreateGeneralService, "CreateGeneralService"},

src/core/hle/service/nifm/nifm_a.h

@@ -4,20 +4,14 @@
 
 #pragma once
 
-#include "core/hle/kernel/hle_ipc.h"
-#include "core/hle/service/service.h"
+#include "core/hle/service/nifm/nifm.h"
 
 namespace Service {
 namespace NIFM {
 
-class NIFM_A final : public ServiceFramework<NIFM_A> {
+class NIFM_A final : public Module::Interface {
 public:
-    NIFM_A();
-    ~NIFM_A() = default;
-
-private:
-    void CreateGeneralServiceOld(Kernel::HLERequestContext& ctx);
-    void CreateGeneralService(Kernel::HLERequestContext& ctx);
+    explicit NIFM_A(std::shared_ptr<Module> module);
 };
 
 } // namespace NIFM

src/core/hle/service/nifm/nifm_s.cpp

@@ -2,29 +2,12 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "common/logging/log.h"
-#include "core/hle/ipc_helpers.h"
-#include "core/hle/service/nifm/nifm.h"
 #include "core/hle/service/nifm/nifm_s.h"
 
 namespace Service {
 namespace NIFM {
 
-void NIFM_S::CreateGeneralServiceOld(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IGeneralService>();
-    LOG_DEBUG(Service_NIFM, "called");
-}
-
-void NIFM_S::CreateGeneralService(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IGeneralService>();
-    LOG_DEBUG(Service_NIFM, "called");
-}
-
-NIFM_S::NIFM_S() : ServiceFramework("nifm:s") {
+NIFM_S::NIFM_S(std::shared_ptr<Module> module) : Module::Interface(std::move(module), "nifm:s") {
     static const FunctionInfo functions[] = {
         {4, &NIFM_S::CreateGeneralServiceOld, "CreateGeneralServiceOld"},
         {5, &NIFM_S::CreateGeneralService, "CreateGeneralService"},

src/core/hle/service/nifm/nifm_s.h

@@ -4,20 +4,14 @@
 
 #pragma once
 
-#include "core/hle/kernel/hle_ipc.h"
-#include "core/hle/service/service.h"
+#include "core/hle/service/nifm/nifm.h"
 
 namespace Service {
 namespace NIFM {
 
-class NIFM_S final : public ServiceFramework<NIFM_S> {
+class NIFM_S final : public Module::Interface {
 public:
-    NIFM_S();
-    ~NIFM_S() = default;
-
-private:
-    void CreateGeneralServiceOld(Kernel::HLERequestContext& ctx);
-    void CreateGeneralService(Kernel::HLERequestContext& ctx);
+    explicit NIFM_S(std::shared_ptr<Module> module);
 };
 
 } // namespace NIFM

src/core/hle/service/nifm/nifm_u.cpp

@@ -2,29 +2,12 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "common/logging/log.h"
-#include "core/hle/ipc_helpers.h"
-#include "core/hle/service/nifm/nifm.h"
 #include "core/hle/service/nifm/nifm_u.h"
 
 namespace Service {
 namespace NIFM {
 
-void NIFM_U::CreateGeneralServiceOld(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IGeneralService>();
-    LOG_DEBUG(Service_NIFM, "called");
-}
-
-void NIFM_U::CreateGeneralService(Kernel::HLERequestContext& ctx) {
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IGeneralService>();
-    LOG_DEBUG(Service_NIFM, "called");
-}
-
-NIFM_U::NIFM_U() : ServiceFramework("nifm:u") {
+NIFM_U::NIFM_U(std::shared_ptr<Module> module) : Module::Interface(std::move(module), "nifm:u") {
     static const FunctionInfo functions[] = {
         {4, &NIFM_U::CreateGeneralServiceOld, "CreateGeneralServiceOld"},
         {5, &NIFM_U::CreateGeneralService, "CreateGeneralService"},

src/core/hle/service/nifm/nifm_u.h

@@ -4,20 +4,14 @@
 
 #pragma once
 
-#include "core/hle/kernel/hle_ipc.h"
-#include "core/hle/service/service.h"
+#include "core/hle/service/nifm/nifm.h"
 
 namespace Service {
 namespace NIFM {
 
-class NIFM_U final : public ServiceFramework<NIFM_U> {
+class NIFM_U final : public Module::Interface {
 public:
-    NIFM_U();
-    ~NIFM_U() = default;
-
-private:
-    void CreateGeneralServiceOld(Kernel::HLERequestContext& ctx);
-    void CreateGeneralService(Kernel::HLERequestContext& ctx);
+    explicit NIFM_U(std::shared_ptr<Module> module);
 };
 
 } // namespace NIFM

src/core/hle/service/nvflinger/buffer_queue.cpp

@@ -26,26 +26,30 @@ void BufferQueue::SetPreallocatedBuffer(u32 slot, IGBPBuffer& igbp_buffer) {
     LOG_WARNING(Service, "Adding graphics buffer %u", slot);
 
     queue.emplace_back(buffer);
+
+    if (buffer_wait_event) {
+        buffer_wait_event->Signal();
+    }
 }
 
-u32 BufferQueue::DequeueBuffer(u32 pixel_format, u32 width, u32 height) {
+boost::optional<u32> BufferQueue::DequeueBuffer(u32 width, u32 height) {
     auto itr = std::find_if(queue.begin(), queue.end(), [&](const Buffer& buffer) {
         // Only consider free buffers. Buffers become free once again after they've been Acquired
         // and Released by the compositor, see the NVFlinger::Compose method.
-        if (buffer.status != Buffer::Status::Free)
+        if (buffer.status != Buffer::Status::Free) {
             return false;
+        }
 
         // Make sure that the parameters match.
-        auto& igbp_buffer = buffer.igbp_buffer;
-        return igbp_buffer.format == pixel_format && igbp_buffer.width == width &&
-               igbp_buffer.height == height;
+        return buffer.igbp_buffer.width == width && buffer.igbp_buffer.height == height;
     });
+
     if (itr == queue.end()) {
-        LOG_CRITICAL(Service_NVDRV, "no free buffers for pixel_format=%d, width=%d, height=%d",
-                     pixel_format, width, height);
-        itr = queue.begin();
+        return boost::none;
     }
 
+    buffer_wait_event = nullptr;
+
     itr->status = Buffer::Status::Dequeued;
     return itr->slot;
 }
@@ -83,6 +87,10 @@ void BufferQueue::ReleaseBuffer(u32 slot) {
     ASSERT(itr != queue.end());
     ASSERT(itr->status == Buffer::Status::Acquired);
     itr->status = Buffer::Status::Free;
+
+    if (buffer_wait_event) {
+        buffer_wait_event->Signal();
+    }
 }
 
 u32 BufferQueue::Query(QueryType type) {
@@ -98,5 +106,10 @@ u32 BufferQueue::Query(QueryType type) {
     return 0;
 }
 
+void BufferQueue::SetBufferWaitEvent(Kernel::SharedPtr<Kernel::Event>&& wait_event) {
+    ASSERT_MSG(!buffer_wait_event, "buffer_wait_event only supports a single waiting thread!");
+    buffer_wait_event = std::move(wait_event);
+}
+
 } // namespace NVFlinger
 } // namespace Service

src/core/hle/service/nvflinger/buffer_queue.h

@@ -69,12 +69,13 @@ public:
     };
 
     void SetPreallocatedBuffer(u32 slot, IGBPBuffer& buffer);
-    u32 DequeueBuffer(u32 pixel_format, u32 width, u32 height);
+    boost::optional<u32> DequeueBuffer(u32 width, u32 height);
     const IGBPBuffer& RequestBuffer(u32 slot) const;
     void QueueBuffer(u32 slot, BufferTransformFlags transform);
     boost::optional<const Buffer&> AcquireBuffer();
     void ReleaseBuffer(u32 slot);
     u32 Query(QueryType type);
+    void SetBufferWaitEvent(Kernel::SharedPtr<Kernel::Event>&& wait_event);
 
     u32 GetId() const {
         return id;
@@ -90,6 +91,9 @@ private:
 
     std::vector<Buffer> queue;
     Kernel::SharedPtr<Kernel::Event> native_handle;
+
+    /// Used to signal waiting thread when no buffers are available
+    Kernel::SharedPtr<Kernel::Event> buffer_wait_event;
 };
 
 } // namespace NVFlinger

src/core/hle/service/service.cpp

@@ -152,8 +152,7 @@ ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& co
         UNIMPLEMENTED_MSG("command_type=%d", context.GetCommandType());
     }
 
-    u32* cmd_buf = (u32*)Memory::GetPointer(Kernel::GetCurrentThread()->GetTLSAddress());
-    context.WriteToOutgoingCommandBuffer(cmd_buf, *Core::CurrentProcess(), Kernel::g_handle_table);
+    context.WriteToOutgoingCommandBuffer(*Kernel::GetCurrentThread());
 
     return RESULT_SUCCESS;
 }

src/core/hle/service/vi/vi.cpp

@@ -486,12 +486,30 @@ private:
             ctx.WriteBuffer(response.Serialize());
         } else if (transaction == TransactionId::DequeueBuffer) {
             IGBPDequeueBufferRequestParcel request{ctx.ReadBuffer()};
+            const u32 width{request.data.width};
+            const u32 height{request.data.height};
+            boost::optional<u32> slot = buffer_queue->DequeueBuffer(width, height);
 
-            u32 slot = buffer_queue->DequeueBuffer(request.data.pixel_format, request.data.width,
-                                                   request.data.height);
-
-            IGBPDequeueBufferResponseParcel response{slot};
-            ctx.WriteBuffer(response.Serialize());
+            if (slot != boost::none) {
+                // Buffer is available
+                IGBPDequeueBufferResponseParcel response{*slot};
+                ctx.WriteBuffer(response.Serialize());
+            } else {
+                // Wait the current thread until a buffer becomes available
+                auto wait_event = ctx.SleepClientThread(
+                    Kernel::GetCurrentThread(), "IHOSBinderDriver::DequeueBuffer", -1,
+                    [=](Kernel::SharedPtr<Kernel::Thread> thread, Kernel::HLERequestContext& ctx,
+                        ThreadWakeupReason reason) {
+                        // Repeat TransactParcel DequeueBuffer when a buffer is available
+                        auto buffer_queue = nv_flinger->GetBufferQueue(id);
+                        boost::optional<u32> slot = buffer_queue->DequeueBuffer(width, height);
+                        IGBPDequeueBufferResponseParcel response{*slot};
+                        ctx.WriteBuffer(response.Serialize());
+                        IPC::ResponseBuilder rb{ctx, 2};
+                        rb.Push(RESULT_SUCCESS);
+                    });
+                buffer_queue->SetBufferWaitEvent(std::move(wait_event));
+            }
         } else if (transaction == TransactionId::RequestBuffer) {
             IGBPRequestBufferRequestParcel request{ctx.ReadBuffer()};
 

src/core/loader/elf.cpp

@@ -414,7 +414,7 @@ ResultStatus AppLoader_ELF::Load(Kernel::SharedPtr<Kernel::Process>& process) {
     process->resource_limit =
         Kernel::ResourceLimit::GetForCategory(Kernel::ResourceLimitCategory::APPLICATION);
 
-    process->Run(codeset->entrypoint, 48, Kernel::DEFAULT_STACK_SIZE);
+    process->Run(codeset->entrypoint, 48, Memory::STACK_SIZE);
 
     is_loaded = true;
     return ResultStatus::Success;

src/core/loader/nro.cpp

@@ -137,7 +137,7 @@ ResultStatus AppLoader_NRO::Load(Kernel::SharedPtr<Kernel::Process>& process) {
     process->address_mappings = default_address_mappings;
     process->resource_limit =
         Kernel::ResourceLimit::GetForCategory(Kernel::ResourceLimitCategory::APPLICATION);
-    process->Run(base_addr, 48, Kernel::DEFAULT_STACK_SIZE);
+    process->Run(base_addr, 48, Memory::STACK_SIZE);
 
     is_loaded = true;
     return ResultStatus::Success;

src/core/loader/nso.cpp

@@ -165,7 +165,7 @@ ResultStatus AppLoader_NSO::Load(Kernel::SharedPtr<Kernel::Process>& process) {
     process->address_mappings = default_address_mappings;
     process->resource_limit =
         Kernel::ResourceLimit::GetForCategory(Kernel::ResourceLimitCategory::APPLICATION);
-    process->Run(Memory::PROCESS_IMAGE_VADDR, 48, Kernel::DEFAULT_STACK_SIZE);
+    process->Run(Memory::PROCESS_IMAGE_VADDR, 48, Memory::STACK_SIZE);
 
     is_loaded = true;
     return ResultStatus::Success;

src/core/memory.h

@@ -129,21 +129,6 @@ enum : VAddr {
     PROCESS_IMAGE_MAX_SIZE = 0x08000000,
     PROCESS_IMAGE_VADDR_END = PROCESS_IMAGE_VADDR + PROCESS_IMAGE_MAX_SIZE,
 
-    /// Area where IPC buffers are mapped onto.
-    IPC_MAPPING_VADDR = 0x04000000,
-    IPC_MAPPING_SIZE = 0x04000000,
-    IPC_MAPPING_VADDR_END = IPC_MAPPING_VADDR + IPC_MAPPING_SIZE,
-
-    /// Application heap (includes stack).
-    HEAP_VADDR = 0x108000000,
-    HEAP_SIZE = 0xF0000000,
-    HEAP_VADDR_END = HEAP_VADDR + HEAP_SIZE,
-
-    /// Area where shared memory buffers are mapped onto.
-    SHARED_MEMORY_VADDR = 0x10000000,
-    SHARED_MEMORY_SIZE = 0x04000000,
-    SHARED_MEMORY_VADDR_END = SHARED_MEMORY_VADDR + SHARED_MEMORY_SIZE,
-
     /// Maps 1:1 to an offset in FCRAM. Used for HW allocations that need to be linear in physical
     /// memory.
     LINEAR_HEAP_VADDR = 0x14000000,
@@ -176,14 +161,39 @@ enum : VAddr {
     SHARED_PAGE_SIZE = 0x00001000,
     SHARED_PAGE_VADDR_END = SHARED_PAGE_VADDR + SHARED_PAGE_SIZE,
 
-    /// Area where TLS (Thread-Local Storage) buffers are allocated.
-    TLS_AREA_VADDR = 0x228000000,
-    TLS_ENTRY_SIZE = 0x200,
-
     /// Equivalent to LINEAR_HEAP_VADDR, but expanded to cover the extra memory in the New 3DS.
     NEW_LINEAR_HEAP_VADDR = 0x30000000,
     NEW_LINEAR_HEAP_SIZE = 0x10000000,
     NEW_LINEAR_HEAP_VADDR_END = NEW_LINEAR_HEAP_VADDR + NEW_LINEAR_HEAP_SIZE,
+
+    /// Area where TLS (Thread-Local Storage) buffers are allocated.
+    TLS_AREA_VADDR = NEW_LINEAR_HEAP_VADDR_END,
+    TLS_ENTRY_SIZE = 0x200,
+    TLS_AREA_SIZE = 0x10000000,
+    TLS_ADREA_VADDR_END = TLS_AREA_VADDR + TLS_AREA_SIZE,
+
+    /// Application stack
+    STACK_VADDR = TLS_ADREA_VADDR_END,
+    STACK_SIZE = 0x10000,
+    STACK_VADDR_END = STACK_VADDR + STACK_SIZE,
+
+    /// Application heap
+    /// Size is confirmed to be a static value on fw 3.0.0
+    HEAP_VADDR = 0x108000000,
+    HEAP_SIZE = 0x180000000,
+    HEAP_VADDR_END = HEAP_VADDR + HEAP_SIZE,
+
+    /// New map region
+    /// Size is confirmed to be a static value on fw 3.0.0
+    NEW_MAP_REGION_VADDR = HEAP_VADDR_END,
+    NEW_MAP_REGION_SIZE = 0x80000000,
+    NEW_MAP_REGION_VADDR_END = NEW_MAP_REGION_VADDR + NEW_MAP_REGION_SIZE,
+
+    /// Map region
+    /// Size is confirmed to be a static value on fw 3.0.0
+    MAP_REGION_VADDR = NEW_MAP_REGION_VADDR_END,
+    MAP_REGION_SIZE = 0x1000000000,
+    MAP_REGION_VADDR_END = MAP_REGION_VADDR + MAP_REGION_SIZE,
 };
 
 /// Currently active page table

src/video_core/CMakeLists.txt

@@ -7,6 +7,7 @@ add_library(video_core STATIC
     engines/maxwell_3d.h
     engines/maxwell_compute.cpp
     engines/maxwell_compute.h
+    gpu.cpp
     gpu.h
     memory_manager.cpp
     memory_manager.h

src/video_core/command_processor.cpp

@@ -24,12 +24,37 @@ namespace Tegra {
 
 enum class BufferMethods {
     BindObject = 0,
+    SetGraphMacroCode = 0x45,
+    SetGraphMacroCodeArg = 0x46,
+    SetGraphMacroEntry = 0x47,
     CountBufferMethods = 0x100,
 };
 
-void GPU::WriteReg(u32 method, u32 subchannel, u32 value) {
-    LOG_WARNING(HW_GPU, "Processing method %08X on subchannel %u value %08X", method, subchannel,
-                value);
+void GPU::WriteReg(u32 method, u32 subchannel, u32 value, u32 remaining_params) {
+    LOG_WARNING(HW_GPU, "Processing method %08X on subchannel %u value %08X remaining params %u",
+                method, subchannel, value, remaining_params);
+
+    if (method == static_cast<u32>(BufferMethods::SetGraphMacroEntry)) {
+        // Prepare to upload a new macro, reset the upload counter.
+        LOG_DEBUG(HW_GPU, "Uploading GPU macro %08X", value);
+        current_macro_entry = value;
+        current_macro_code.clear();
+        return;
+    }
+
+    if (method == static_cast<u32>(BufferMethods::SetGraphMacroCodeArg)) {
+        // Append a new code word to the current macro.
+        current_macro_code.push_back(value);
+
+        // There are no more params remaining, submit the code to the 3D engine.
+        if (remaining_params == 0) {
+            maxwell_3d->SubmitMacroCode(current_macro_entry, std::move(current_macro_code));
+            current_macro_entry = InvalidGraphMacroEntry;
+            current_macro_code.clear();
+        }
+
+        return;
+    }
 
     if (method == static_cast<u32>(BufferMethods::BindObject)) {
         // Bind the current subchannel to the desired engine id.
@@ -54,7 +79,7 @@ void GPU::WriteReg(u32 method, u32 subchannel, u32 value) {
         fermi_2d->WriteReg(method, value);
         break;
     case EngineID::MAXWELL_B:
-        maxwell_3d->WriteReg(method, value);
+        maxwell_3d->WriteReg(method, value, remaining_params);
         break;
     case EngineID::MAXWELL_COMPUTE_B:
         maxwell_compute->WriteReg(method, value);
@@ -78,7 +103,8 @@ void GPU::ProcessCommandList(GPUVAddr address, u32 size) {
         case SubmissionMode::Increasing: {
             // Increase the method value with each argument.
             for (unsigned i = 0; i < header.arg_count; ++i) {
-                WriteReg(header.method + i, header.subchannel, Memory::Read32(current_addr));
+                WriteReg(header.method + i, header.subchannel, Memory::Read32(current_addr),
+                         header.arg_count - i - 1);
                 current_addr += sizeof(u32);
             }
             break;
@@ -87,27 +113,31 @@ void GPU::ProcessCommandList(GPUVAddr address, u32 size) {
         case SubmissionMode::NonIncreasing: {
             // Use the same method value for all arguments.
             for (unsigned i = 0; i < header.arg_count; ++i) {
-                WriteReg(header.method, header.subchannel, Memory::Read32(current_addr));
+                WriteReg(header.method, header.subchannel, Memory::Read32(current_addr),
+                         header.arg_count - i - 1);
                 current_addr += sizeof(u32);
             }
             break;
         }
         case SubmissionMode::IncreaseOnce: {
             ASSERT(header.arg_count.Value() >= 1);
+
             // Use the original method for the first argument and then the next method for all other
             // arguments.
-            WriteReg(header.method, header.subchannel, Memory::Read32(current_addr));
+            WriteReg(header.method, header.subchannel, Memory::Read32(current_addr),
+                     header.arg_count - 1);
             current_addr += sizeof(u32);
-            // Use the same method value for all arguments.
+
             for (unsigned i = 1; i < header.arg_count; ++i) {
-                WriteReg(header.method + 1, header.subchannel, Memory::Read32(current_addr));
+                WriteReg(header.method + 1, header.subchannel, Memory::Read32(current_addr),
+                         header.arg_count - i - 1);
                 current_addr += sizeof(u32);
             }
             break;
         }
         case SubmissionMode::Inline: {
             // The register value is stored in the bits 16-28 as an immediate
-            WriteReg(header.method, header.subchannel, header.inline_data);
+            WriteReg(header.method, header.subchannel, header.inline_data, 0);
             break;
         }
         default:

src/video_core/command_processor.h

@@ -34,6 +34,4 @@ static_assert(std::is_standard_layout<CommandHeader>::value == true,
               "CommandHeader does not use standard layout");
 static_assert(sizeof(CommandHeader) == sizeof(u32), "CommandHeader has incorrect size!");
 
-void ProcessCommandList(VAddr address, u32 size);
-
 } // namespace Tegra

src/video_core/engines/maxwell_3d.cpp

@@ -8,17 +8,122 @@
 namespace Tegra {
 namespace Engines {
 
+/// First register id that is actually a Macro call.
+constexpr u32 MacroRegistersStart = 0xE00;
+
+const std::unordered_map<u32, Maxwell3D::MethodInfo> Maxwell3D::method_handlers = {
+    {0xE1A, {"BindTextureInfoBuffer", 1, &Maxwell3D::BindTextureInfoBuffer}},
+    {0xE24, {"SetShader", 5, &Maxwell3D::SetShader}},
+    {0xE2A, {"BindStorageBuffer", 1, &Maxwell3D::BindStorageBuffer}},
+};
+
 Maxwell3D::Maxwell3D(MemoryManager& memory_manager) : memory_manager(memory_manager) {}
 
-void Maxwell3D::WriteReg(u32 method, u32 value) {
+void Maxwell3D::SubmitMacroCode(u32 entry, std::vector<u32> code) {
+    uploaded_macros[entry * 2 + MacroRegistersStart] = std::move(code);
+}
+
+void Maxwell3D::CallMacroMethod(u32 method, const std::vector<u32>& parameters) {
+    // TODO(Subv): Write an interpreter for the macros uploaded via registers 0x45 and 0x47
+
+    // The requested macro must have been uploaded already.
+    ASSERT_MSG(uploaded_macros.find(method) != uploaded_macros.end(), "Macro %08X was not uploaded",
+               method);
+
+    auto itr = method_handlers.find(method);
+    ASSERT_MSG(itr != method_handlers.end(), "Unhandled method call %08X", method);
+
+    ASSERT(itr->second.arguments == parameters.size());
+
+    (this->*itr->second.handler)(parameters);
+
+    // Reset the current macro and its parameters.
+    executing_macro = 0;
+    macro_params.clear();
+}
+
+void Maxwell3D::WriteReg(u32 method, u32 value, u32 remaining_params) {
     ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid Maxwell3D register, increase the size of the Regs structure");
 
+    // It is an error to write to a register other than the current macro's ARG register before it
+    // has finished execution.
+    if (executing_macro != 0) {
+        ASSERT(method == executing_macro + 1);
+    }
+
+    // Methods after 0xE00 are special, they're actually triggers for some microcode that was
+    // uploaded to the GPU during initialization.
+    if (method >= MacroRegistersStart) {
+        // We're trying to execute a macro
+        if (executing_macro == 0) {
+            // A macro call must begin by writing the macro method's register, not its argument.
+            ASSERT_MSG((method % 2) == 0,
+                       "Can't start macro execution by writing to the ARGS register");
+            executing_macro = method;
+        }
+
+        macro_params.push_back(value);
+
+        // Call the macro when there are no more parameters in the command buffer
+        if (remaining_params == 0) {
+            CallMacroMethod(executing_macro, macro_params);
+        }
+        return;
+    }
+
     regs.reg_array[method] = value;
 
 #define MAXWELL3D_REG_INDEX(field_name) (offsetof(Regs, field_name) / sizeof(u32))
 
     switch (method) {
+    case MAXWELL3D_REG_INDEX(code_address.code_address_high):
+    case MAXWELL3D_REG_INDEX(code_address.code_address_low): {
+        // Note: For some reason games (like Puyo Puyo Tetris) seem to write 0 to the CODE_ADDRESS
+        // register, we do not currently know if that's intended or a bug, so we assert it lest
+        // stuff breaks in other places (like the shader address calculation).
+        ASSERT_MSG(regs.code_address.CodeAddress() == 0, "Unexpected CODE_ADDRESS register value.");
+        break;
+    }
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[1]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[2]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[3]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[4]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[5]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[6]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[7]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[8]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[9]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[10]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[11]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[12]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[13]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[14]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[15]): {
+        ProcessCBData(value);
+        break;
+    }
+    case MAXWELL3D_REG_INDEX(cb_bind[0].raw_config): {
+        ProcessCBBind(Regs::ShaderStage::Vertex);
+        break;
+    }
+    case MAXWELL3D_REG_INDEX(cb_bind[1].raw_config): {
+        ProcessCBBind(Regs::ShaderStage::TesselationControl);
+        break;
+    }
+    case MAXWELL3D_REG_INDEX(cb_bind[2].raw_config): {
+        ProcessCBBind(Regs::ShaderStage::TesselationEval);
+        break;
+    }
+    case MAXWELL3D_REG_INDEX(cb_bind[3].raw_config): {
+        ProcessCBBind(Regs::ShaderStage::Geometry);
+        break;
+    }
+    case MAXWELL3D_REG_INDEX(cb_bind[4].raw_config): {
+        ProcessCBBind(Regs::ShaderStage::Fragment);
+        break;
+    }
     case MAXWELL3D_REG_INDEX(draw.vertex_end_gl): {
         DrawArrays();
         break;
@@ -56,5 +161,113 @@ void Maxwell3D::DrawArrays() {
     LOG_WARNING(HW_GPU, "Game requested a DrawArrays, ignoring");
 }
 
+void Maxwell3D::BindTextureInfoBuffer(const std::vector<u32>& parameters) {
+    /**
+     * Parameters description:
+     * [0] = Shader stage, usually 4 for FragmentShader
+     */
+
+    u32 stage = parameters[0];
+
+    // Perform the same operations as the real macro code.
+    GPUVAddr address = static_cast<GPUVAddr>(regs.tex_info_buffers.address[stage]) << 8;
+    u32 size = regs.tex_info_buffers.size[stage];
+
+    regs.const_buffer.cb_size = size;
+    regs.const_buffer.cb_address_high = address >> 32;
+    regs.const_buffer.cb_address_low = address & 0xFFFFFFFF;
+}
+
+void Maxwell3D::SetShader(const std::vector<u32>& parameters) {
+    /**
+     * Parameters description:
+     * [0] = Shader Program.
+     * [1] = Unknown, presumably the shader id.
+     * [2] = Offset to the start of the shader, after the 0x30 bytes header.
+     * [3] = Shader Stage.
+     * [4] = Const Buffer Address >> 8.
+     */
+    auto shader_program = static_cast<Regs::ShaderProgram>(parameters[0]);
+    // TODO(Subv): This address is probably an offset from the CODE_ADDRESS register.
+    GPUVAddr address = parameters[2];
+    auto shader_stage = static_cast<Regs::ShaderStage>(parameters[3]);
+    GPUVAddr cb_address = parameters[4] << 8;
+
+    auto& shader = state.shader_programs[static_cast<size_t>(shader_program)];
+    shader.program = shader_program;
+    shader.stage = shader_stage;
+    shader.address = address;
+
+    // Perform the same operations as the real macro code.
+    // TODO(Subv): Early exit if register 0xD1C + shader_program contains the same as params[1].
+    auto& shader_regs = regs.shader_config[static_cast<size_t>(shader_program)];
+    shader_regs.start_id = address;
+    // TODO(Subv): Write params[1] to register 0xD1C + shader_program.
+    // TODO(Subv): Write params[2] to register 0xD22 + shader_program.
+
+    // Note: This value is hardcoded in the macro's code.
+    static constexpr u32 DefaultCBSize = 0x10000;
+    regs.const_buffer.cb_size = DefaultCBSize;
+    regs.const_buffer.cb_address_high = cb_address >> 32;
+    regs.const_buffer.cb_address_low = cb_address & 0xFFFFFFFF;
+
+    // Write a hardcoded 0x11 to CB_BIND, this binds the current const buffer to buffer c1[] in the
+    // shader. It's likely that these are the constants for the shader.
+    regs.cb_bind[static_cast<size_t>(shader_stage)].valid.Assign(1);
+    regs.cb_bind[static_cast<size_t>(shader_stage)].index.Assign(1);
+
+    ProcessCBBind(shader_stage);
+}
+
+void Maxwell3D::BindStorageBuffer(const std::vector<u32>& parameters) {
+    /**
+     * Parameters description:
+     * [0] = Buffer offset >> 2
+     */
+
+    u32 buffer_offset = parameters[0] << 2;
+
+    // Perform the same operations as the real macro code.
+    // Note: This value is hardcoded in the macro's code.
+    static constexpr u32 DefaultCBSize = 0x5F00;
+    regs.const_buffer.cb_size = DefaultCBSize;
+
+    GPUVAddr address = regs.ssbo_info.BufferAddress();
+    regs.const_buffer.cb_address_high = address >> 32;
+    regs.const_buffer.cb_address_low = address & 0xFFFFFFFF;
+
+    regs.const_buffer.cb_pos = buffer_offset;
+}
+
+void Maxwell3D::ProcessCBBind(Regs::ShaderStage stage) {
+    // Bind the buffer currently in CB_ADDRESS to the specified index in the desired shader stage.
+    auto& shader = state.shader_stages[static_cast<size_t>(stage)];
+    auto& bind_data = regs.cb_bind[static_cast<size_t>(stage)];
+
+    auto& buffer = shader.const_buffers[bind_data.index];
+
+    buffer.enabled = bind_data.valid.Value() != 0;
+    buffer.index = bind_data.index;
+    buffer.address = regs.const_buffer.BufferAddress();
+    buffer.size = regs.const_buffer.cb_size;
+}
+
+void Maxwell3D::ProcessCBData(u32 value) {
+    // Write the input value to the current const buffer at the current position.
+    GPUVAddr buffer_address = regs.const_buffer.BufferAddress();
+    ASSERT(buffer_address != 0);
+
+    // Don't allow writing past the end of the buffer.
+    ASSERT(regs.const_buffer.cb_pos + sizeof(u32) <= regs.const_buffer.cb_size);
+
+    VAddr address =
+        memory_manager.PhysicalToVirtualAddress(buffer_address + regs.const_buffer.cb_pos);
+
+    Memory::Write32(address, value);
+
+    // Increment the current buffer position.
+    regs.const_buffer.cb_pos = regs.const_buffer.cb_pos + 4;
+}
+
 } // namespace Engines
 } // namespace Tegra

src/video_core/engines/maxwell_3d.h

@@ -4,6 +4,9 @@
 
 #pragma once
 
+#include <array>
+#include <unordered_map>
+#include <vector>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
@@ -18,21 +21,129 @@ public:
     ~Maxwell3D() = default;
 
     /// Write the value to the register identified by method.
-    void WriteReg(u32 method, u32 value);
+    void WriteReg(u32 method, u32 value, u32 remaining_params);
+
+    /// Uploads the code for a GPU macro program associated with the specified entry.
+    void SubmitMacroCode(u32 entry, std::vector<u32> code);
 
     /// Register structure of the Maxwell3D engine.
     /// TODO(Subv): This structure will need to be made bigger as more registers are discovered.
     struct Regs {
         static constexpr size_t NUM_REGS = 0xE36;
 
+        static constexpr size_t NumRenderTargets = 8;
+        static constexpr size_t NumCBData = 16;
+        static constexpr size_t NumVertexArrays = 32;
+        static constexpr size_t MaxShaderProgram = 6;
+        static constexpr size_t MaxShaderStage = 5;
+        // Maximum number of const buffers per shader stage.
+        static constexpr size_t MaxConstBuffers = 16;
+
         enum class QueryMode : u32 {
             Write = 0,
             Sync = 1,
         };
 
+        enum class ShaderProgram : u32 {
+            VertexA = 0,
+            VertexB = 1,
+            TesselationControl = 2,
+            TesselationEval = 3,
+            Geometry = 4,
+            Fragment = 5,
+        };
+
+        enum class ShaderStage : u32 {
+            Vertex = 0,
+            TesselationControl = 1,
+            TesselationEval = 2,
+            Geometry = 3,
+            Fragment = 4,
+        };
+
         union {
             struct {
-                INSERT_PADDING_WORDS(0x585);
+                INSERT_PADDING_WORDS(0x200);
+
+                struct {
+                    u32 address_high;
+                    u32 address_low;
+                    u32 horiz;
+                    u32 vert;
+                    u32 format;
+                    u32 block_dimensions;
+                    u32 array_mode;
+                    u32 layer_stride;
+                    u32 base_layer;
+                    INSERT_PADDING_WORDS(7);
+
+                    GPUVAddr Address() const {
+                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
+                                                     address_low);
+                    }
+                } rt[NumRenderTargets];
+
+                INSERT_PADDING_WORDS(0x178);
+
+                struct {
+                    u32 address_high;
+                    u32 address_low;
+                    u32 format;
+                    u32 block_dimensions;
+                    u32 layer_stride;
+
+                    GPUVAddr Address() const {
+                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
+                                                     address_low);
+                    }
+                } zeta;
+
+                INSERT_PADDING_WORDS(0x8A);
+
+                struct {
+                    union {
+                        BitField<0, 4, u32> count;
+                    };
+                } rt_control;
+
+                INSERT_PADDING_WORDS(0xCF);
+
+                struct {
+                    u32 tsc_address_high;
+                    u32 tsc_address_low;
+                    u32 tsc_limit;
+
+                    GPUVAddr TSCAddress() const {
+                        return static_cast<GPUVAddr>(
+                            (static_cast<GPUVAddr>(tsc_address_high) << 32) | tsc_address_low);
+                    }
+                } tsc;
+
+                INSERT_PADDING_WORDS(0x3);
+
+                struct {
+                    u32 tic_address_high;
+                    u32 tic_address_low;
+                    u32 tic_limit;
+
+                    GPUVAddr TICAddress() const {
+                        return static_cast<GPUVAddr>(
+                            (static_cast<GPUVAddr>(tic_address_high) << 32) | tic_address_low);
+                    }
+                } tic;
+
+                INSERT_PADDING_WORDS(0x22);
+
+                struct {
+                    u32 code_address_high;
+                    u32 code_address_low;
+
+                    GPUVAddr CodeAddress() const {
+                        return static_cast<GPUVAddr>(
+                            (static_cast<GPUVAddr>(code_address_high) << 32) | code_address_low);
+                    }
+                } code_address;
+                INSERT_PADDING_WORDS(1);
                 struct {
                     u32 vertex_end_gl;
                     u32 vertex_begin_gl;
@@ -54,7 +165,98 @@ public:
                             (static_cast<GPUVAddr>(query_address_high) << 32) | query_address_low);
                     }
                 } query;
-                INSERT_PADDING_WORDS(0x772);
+
+                INSERT_PADDING_WORDS(0x3C);
+
+                struct {
+                    union {
+                        BitField<0, 12, u32> stride;
+                        BitField<12, 1, u32> enable;
+                    };
+                    u32 start_high;
+                    u32 start_low;
+                    u32 divisor;
+
+                    GPUVAddr StartAddress() const {
+                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(start_high) << 32) |
+                                                     start_low);
+                    }
+                } vertex_array[NumVertexArrays];
+
+                INSERT_PADDING_WORDS(0x40);
+
+                struct {
+                    u32 limit_high;
+                    u32 limit_low;
+
+                    GPUVAddr LimitAddress() const {
+                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(limit_high) << 32) |
+                                                     limit_low);
+                    }
+                } vertex_array_limit[NumVertexArrays];
+
+                struct {
+                    union {
+                        BitField<0, 1, u32> enable;
+                        BitField<4, 4, ShaderProgram> program;
+                    };
+                    u32 start_id;
+                    INSERT_PADDING_WORDS(1);
+                    u32 gpr_alloc;
+                    ShaderStage type;
+                    INSERT_PADDING_WORDS(9);
+                } shader_config[MaxShaderProgram];
+
+                INSERT_PADDING_WORDS(0x8C);
+
+                struct {
+                    u32 cb_size;
+                    u32 cb_address_high;
+                    u32 cb_address_low;
+                    u32 cb_pos;
+                    u32 cb_data[NumCBData];
+
+                    GPUVAddr BufferAddress() const {
+                        return static_cast<GPUVAddr>(
+                            (static_cast<GPUVAddr>(cb_address_high) << 32) | cb_address_low);
+                    }
+                } const_buffer;
+
+                INSERT_PADDING_WORDS(0x10);
+
+                struct {
+                    union {
+                        u32 raw_config;
+                        BitField<0, 1, u32> valid;
+                        BitField<4, 5, u32> index;
+                    };
+                    INSERT_PADDING_WORDS(7);
+                } cb_bind[MaxShaderStage];
+
+                INSERT_PADDING_WORDS(0x56);
+
+                u32 tex_cb_index;
+
+                INSERT_PADDING_WORDS(0x395);
+
+                struct {
+                    /// Compressed address of a buffer that holds information about bound SSBOs.
+                    /// This address is usually bound to c0 in the shaders.
+                    u32 buffer_address;
+
+                    GPUVAddr BufferAddress() const {
+                        return static_cast<GPUVAddr>(buffer_address) << 8;
+                    }
+                } ssbo_info;
+
+                INSERT_PADDING_WORDS(0x11);
+
+                struct {
+                    u32 address[MaxShaderStage];
+                    u32 size[MaxShaderStage];
+                } tex_info_buffers;
+
+                INSERT_PADDING_WORDS(0x102);
             };
             std::array<u32, NUM_REGS> reg_array;
         };
@@ -62,21 +264,94 @@ public:
 
     static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32), "Maxwell3D Regs has wrong size");
 
+    struct State {
+        struct ConstBufferInfo {
+            GPUVAddr address;
+            u32 index;
+            u32 size;
+            bool enabled;
+        };
+
+        struct ShaderProgramInfo {
+            Regs::ShaderStage stage;
+            Regs::ShaderProgram program;
+            GPUVAddr address;
+        };
+
+        struct ShaderStageInfo {
+            std::array<ConstBufferInfo, Regs::MaxConstBuffers> const_buffers;
+        };
+
+        std::array<ShaderStageInfo, Regs::MaxShaderStage> shader_stages;
+        std::array<ShaderProgramInfo, Regs::MaxShaderProgram> shader_programs;
+    };
+
+    State state{};
+
 private:
+    MemoryManager& memory_manager;
+
+    std::unordered_map<u32, std::vector<u32>> uploaded_macros;
+
+    /// Macro method that is currently being executed / being fed parameters.
+    u32 executing_macro = 0;
+    /// Parameters that have been submitted to the macro call so far.
+    std::vector<u32> macro_params;
+
+    /**
+     * Call a macro on this engine.
+     * @param method Method to call
+     * @param parameters Arguments to the method call
+     */
+    void CallMacroMethod(u32 method, const std::vector<u32>& parameters);
+
     /// Handles a write to the QUERY_GET register.
     void ProcessQueryGet();
 
+    /// Handles a write to the CB_DATA[i] register.
+    void ProcessCBData(u32 value);
+
+    /// Handles a write to the CB_BIND register.
+    void ProcessCBBind(Regs::ShaderStage stage);
+
     /// Handles a write to the VERTEX_END_GL register, triggering a draw.
     void DrawArrays();
 
-    MemoryManager& memory_manager;
+    /// Method call handlers
+    void BindTextureInfoBuffer(const std::vector<u32>& parameters);
+    void SetShader(const std::vector<u32>& parameters);
+    void BindStorageBuffer(const std::vector<u32>& parameters);
+
+    struct MethodInfo {
+        const char* name;
+        u32 arguments;
+        void (Maxwell3D::*handler)(const std::vector<u32>& parameters);
+    };
+
+    static const std::unordered_map<u32, MethodInfo> method_handlers;
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
     static_assert(offsetof(Maxwell3D::Regs, field_name) == position * 4,                           \
                   "Field " #field_name " has invalid position")
 
+ASSERT_REG_POSITION(rt, 0x200);
+ASSERT_REG_POSITION(zeta, 0x3F8);
+ASSERT_REG_POSITION(rt_control, 0x487);
+ASSERT_REG_POSITION(tsc, 0x557);
+ASSERT_REG_POSITION(tic, 0x55D);
+ASSERT_REG_POSITION(code_address, 0x582);
+ASSERT_REG_POSITION(draw, 0x585);
 ASSERT_REG_POSITION(query, 0x6C0);
+ASSERT_REG_POSITION(vertex_array[0], 0x700);
+ASSERT_REG_POSITION(vertex_array_limit[0], 0x7C0);
+ASSERT_REG_POSITION(shader_config[0], 0x800);
+ASSERT_REG_POSITION(const_buffer, 0x8E0);
+ASSERT_REG_POSITION(cb_bind[0], 0x904);
+ASSERT_REG_POSITION(tex_cb_index, 0x982);
+ASSERT_REG_POSITION(ssbo_info, 0xD18);
+ASSERT_REG_POSITION(tex_info_buffers.address[0], 0xD2A);
+ASSERT_REG_POSITION(tex_info_buffers.size[0], 0xD2F);
 
 #undef ASSERT_REG_POSITION
 

src/video_core/gpu.cpp

@@ -0,0 +1,21 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/engines/fermi_2d.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/engines/maxwell_compute.h"
+#include "video_core/gpu.h"
+
+namespace Tegra {
+
+GPU::GPU() {
+    memory_manager = std::make_unique<MemoryManager>();
+    maxwell_3d = std::make_unique<Engines::Maxwell3D>(*memory_manager);
+    fermi_2d = std::make_unique<Engines::Fermi2D>();
+    maxwell_compute = std::make_unique<Engines::MaxwellCompute>();
+}
+
+GPU::~GPU() = default;
+
+} // namespace Tegra

src/video_core/gpu.h

@@ -6,14 +6,18 @@
 
 #include <memory>
 #include <unordered_map>
+#include <vector>
 #include "common/common_types.h"
-#include "video_core/engines/fermi_2d.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/engines/maxwell_compute.h"
 #include "video_core/memory_manager.h"
 
 namespace Tegra {
 
+namespace Engines {
+class Fermi2D;
+class Maxwell3D;
+class MaxwellCompute;
+} // namespace Engines
+
 enum class EngineID {
     FERMI_TWOD_A = 0x902D, // 2D Engine
     MAXWELL_B = 0xB197,    // 3D Engine
@@ -24,13 +28,8 @@ enum class EngineID {
 
 class GPU final {
 public:
-    GPU() {
-        memory_manager = std::make_unique<MemoryManager>();
-        maxwell_3d = std::make_unique<Engines::Maxwell3D>(*memory_manager);
-        fermi_2d = std::make_unique<Engines::Fermi2D>();
-        maxwell_compute = std::make_unique<Engines::MaxwellCompute>();
-    }
-    ~GPU() = default;
+    GPU();
+    ~GPU();
 
     /// Processes a command list stored at the specified address in GPU memory.
     void ProcessCommandList(GPUVAddr address, u32 size);
@@ -38,8 +37,10 @@ public:
     std::unique_ptr<MemoryManager> memory_manager;
 
 private:
+    static constexpr u32 InvalidGraphMacroEntry = 0xFFFFFFFF;
+
     /// Writes a single register in the engine bound to the specified subchannel
-    void WriteReg(u32 method, u32 subchannel, u32 value);
+    void WriteReg(u32 method, u32 subchannel, u32 value, u32 remaining_params);
 
     /// Mapping of command subchannels to their bound engine ids.
     std::unordered_map<u32, EngineID> bound_engines;
@@ -50,6 +51,11 @@ private:
     std::unique_ptr<Engines::Fermi2D> fermi_2d;
     /// Compute engine
     std::unique_ptr<Engines::MaxwellCompute> maxwell_compute;
+
+    /// Entry of the macro that is currently being uploaded
+    u32 current_macro_entry = InvalidGraphMacroEntry;
+    /// Code being uploaded for the current macro
+    std::vector<u32> current_macro_code;
 };
 
 } // namespace Tegra

src/yuzu/debugger/wait_tree.cpp

@@ -150,8 +150,8 @@ QString WaitTreeThread::GetText() const {
     case THREADSTATUS_READY:
         status = tr("ready");
         break;
-    case THREADSTATUS_WAIT_ARB:
-        status = tr("waiting for address 0x%1").arg(thread.wait_address, 8, 16, QLatin1Char('0'));
+    case THREADSTATUS_WAIT_HLE_EVENT:
+        status = tr("waiting for HLE return");
         break;
     case THREADSTATUS_WAIT_SLEEP:
         status = tr("sleeping");
@@ -180,7 +180,7 @@ QColor WaitTreeThread::GetColor() const {
         return QColor(Qt::GlobalColor::darkGreen);
     case THREADSTATUS_READY:
         return QColor(Qt::GlobalColor::darkBlue);
-    case THREADSTATUS_WAIT_ARB:
+    case THREADSTATUS_WAIT_HLE_EVENT:
         return QColor(Qt::GlobalColor::darkRed);
     case THREADSTATUS_WAIT_SLEEP:
         return QColor(Qt::GlobalColor::darkYellow);