38 files changed, 1030 insertions, 239 deletions

diff --git a/src/core/arm/dynarmic/arm_dynarmic.cpp b/src/core/arm/dynarmic/arm_dynarmic.cpp
index 7e978cf7a..0762321a9 100644
--- a/src/core/arm/dynarmic/arm_dynarmic.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic.cpp
@@ -129,7 +129,7 @@ public:
 };
 
 std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit() const {
-    auto& current_process = Core::CurrentProcess();
+    auto* current_process = Core::CurrentProcess();
     auto** const page_table = current_process->VMManager().page_table.pointers.data();
 
     Dynarmic::A64::UserConfig config;
diff --git a/src/core/core.cpp b/src/core/core.cpp
index b6acfb3e4..e2fb9e038 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -136,7 +136,8 @@ struct System::Impl {
         if (virtual_filesystem == nullptr)
             virtual_filesystem = std::make_shared<FileSys::RealVfsFilesystem>();
 
-        kernel.MakeCurrentProcess(Kernel::Process::Create(kernel, "main"));
+        auto main_process = Kernel::Process::Create(kernel, "main");
+        kernel.MakeCurrentProcess(main_process.get());
 
         cpu_barrier = std::make_shared<CpuBarrier>();
         cpu_exclusive_monitor = Cpu::MakeExclusiveMonitor(cpu_cores.size());
@@ -361,11 +362,11 @@ const std::shared_ptr<Kernel::Scheduler>& System::Scheduler(std::size_t core_ind
     return impl->cpu_cores[core_index]->Scheduler();
 }
 
-Kernel::SharedPtr<Kernel::Process>& System::CurrentProcess() {
+Kernel::Process* System::CurrentProcess() {
     return impl->kernel.CurrentProcess();
 }
 
-const Kernel::SharedPtr<Kernel::Process>& System::CurrentProcess() const {
+const Kernel::Process* System::CurrentProcess() const {
     return impl->kernel.CurrentProcess();
 }
 
diff --git a/src/core/core.h b/src/core/core.h
index f9a3e97e3..ea4d53914 100644
--- a/src/core/core.h
+++ b/src/core/core.h
@@ -174,11 +174,11 @@ public:
     /// Gets the scheduler for the CPU core with the specified index
     const std::shared_ptr<Kernel::Scheduler>& Scheduler(std::size_t core_index);
 
-    /// Provides a reference to the current process
-    Kernel::SharedPtr<Kernel::Process>& CurrentProcess();
+    /// Provides a pointer to the current process
+    Kernel::Process* CurrentProcess();
 
-    /// Provides a constant reference to the current process.
-    const Kernel::SharedPtr<Kernel::Process>& CurrentProcess() const;
+    /// Provides a constant pointer to the current process.
+    const Kernel::Process* CurrentProcess() const;
 
     /// Provides a reference to the kernel instance.
     Kernel::KernelCore& Kernel();
@@ -246,7 +246,7 @@ inline TelemetrySession& Telemetry() {
     return System::GetInstance().TelemetrySession();
 }
 
-inline Kernel::SharedPtr<Kernel::Process>& CurrentProcess() {
+inline Kernel::Process* CurrentProcess() {
     return System::GetInstance().CurrentProcess();
 }
 
diff --git a/src/core/file_sys/ips_layer.cpp b/src/core/file_sys/ips_layer.cpp
index 6c072d0a3..554eae9bc 100644
--- a/src/core/file_sys/ips_layer.cpp
+++ b/src/core/file_sys/ips_layer.cpp
@@ -107,12 +107,12 @@ VirtualFile PatchIPS(const VirtualFile& in, const VirtualFile& ips) {
                 return nullptr;
 
             if (real_offset + rle_size > in_data.size())
-                rle_size = in_data.size() - real_offset;
+                rle_size = static_cast<u16>(in_data.size() - real_offset);
             std::memset(in_data.data() + real_offset, data.get(), rle_size);
         } else { // Standard Patch
             auto read = data_size;
             if (real_offset + read > in_data.size())
-                read = in_data.size() - real_offset;
+                read = static_cast<u16>(in_data.size() - real_offset);
             if (ips->Read(in_data.data() + real_offset, read, offset) != data_size)
                 return nullptr;
             offset += data_size;
@@ -265,7 +265,7 @@ void IPSwitchCompiler::Parse() {
                 if (patch_line.length() < 11)
                     break;
                 auto offset = std::stoul(patch_line.substr(0, 8), nullptr, 16);
-                offset += offset_shift;
+                offset += static_cast<unsigned long>(offset_shift);
 
                 std::vector<u8> replace;
                 // 9 - first char of replacement val
diff --git a/src/core/file_sys/patch_manager.cpp b/src/core/file_sys/patch_manager.cpp
index b14d7cb0a..019caebe9 100644
--- a/src/core/file_sys/patch_manager.cpp
+++ b/src/core/file_sys/patch_manager.cpp
@@ -345,23 +345,22 @@ std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNam
     return out;
 }
 
-std::pair<std::shared_ptr<NACP>, VirtualFile> PatchManager::GetControlMetadata() const {
+std::pair<std::unique_ptr<NACP>, VirtualFile> PatchManager::GetControlMetadata() const {
     const auto& installed{Service::FileSystem::GetUnionContents()};
 
     const auto base_control_nca = installed->GetEntry(title_id, ContentRecordType::Control);
     if (base_control_nca == nullptr)
         return {};
 
-    return ParseControlNCA(base_control_nca);
+    return ParseControlNCA(*base_control_nca);
 }
 
-std::pair<std::shared_ptr<NACP>, VirtualFile> PatchManager::ParseControlNCA(
-    const std::shared_ptr<NCA>& nca) const {
-    const auto base_romfs = nca->GetRomFS();
+std::pair<std::unique_ptr<NACP>, VirtualFile> PatchManager::ParseControlNCA(const NCA& nca) const {
+    const auto base_romfs = nca.GetRomFS();
     if (base_romfs == nullptr)
         return {};
 
-    const auto romfs = PatchRomFS(base_romfs, nca->GetBaseIVFCOffset(), ContentRecordType::Control);
+    const auto romfs = PatchRomFS(base_romfs, nca.GetBaseIVFCOffset(), ContentRecordType::Control);
     if (romfs == nullptr)
         return {};
 
@@ -373,7 +372,7 @@ std::pair<std::shared_ptr<NACP>, VirtualFile> PatchManager::ParseControlNCA(
     if (nacp_file == nullptr)
         nacp_file = extracted->GetFile("Control.nacp");
 
-    const auto nacp = nacp_file == nullptr ? nullptr : std::make_shared<NACP>(nacp_file);
+    auto nacp = nacp_file == nullptr ? nullptr : std::make_unique<NACP>(nacp_file);
 
     VirtualFile icon_file;
     for (const auto& language : FileSys::LANGUAGE_NAMES) {
@@ -382,6 +381,6 @@ std::pair<std::shared_ptr<NACP>, VirtualFile> PatchManager::ParseControlNCA(
             break;
     }
 
-    return {nacp, icon_file};
+    return {std::move(nacp), icon_file};
 }
 } // namespace FileSys
diff --git a/src/core/file_sys/patch_manager.h b/src/core/file_sys/patch_manager.h
index eb6fc4607..7d168837f 100644
--- a/src/core/file_sys/patch_manager.h
+++ b/src/core/file_sys/patch_manager.h
@@ -57,11 +57,10 @@ public:
 
     // Given title_id of the program, attempts to get the control data of the update and parse it,
     // falling back to the base control data.
-    std::pair<std::shared_ptr<NACP>, VirtualFile> GetControlMetadata() const;
+    std::pair<std::unique_ptr<NACP>, VirtualFile> GetControlMetadata() const;
 
     // Version of GetControlMetadata that takes an arbitrary NCA
-    std::pair<std::shared_ptr<NACP>, VirtualFile> ParseControlNCA(
-        const std::shared_ptr<NCA>& nca) const;
+    std::pair<std::unique_ptr<NACP>, VirtualFile> ParseControlNCA(const NCA& nca) const;
 
 private:
     u64 title_id;
diff --git a/src/core/hle/kernel/errors.h b/src/core/hle/kernel/errors.h
index e5fa67ae8..885259618 100644
--- a/src/core/hle/kernel/errors.h
+++ b/src/core/hle/kernel/errors.h
@@ -22,6 +22,7 @@ enum {
     HandleTableFull = 105,
     InvalidMemoryState = 106,
     InvalidMemoryPermissions = 108,
+    InvalidMemoryRange = 110,
     InvalidThreadPriority = 112,
     InvalidProcessorId = 113,
     InvalidHandle = 114,
@@ -56,6 +57,7 @@ constexpr ResultCode ERR_INVALID_ADDRESS(ErrorModule::Kernel, ErrCodes::InvalidA
 constexpr ResultCode ERR_INVALID_ADDRESS_STATE(ErrorModule::Kernel, ErrCodes::InvalidMemoryState);
 constexpr ResultCode ERR_INVALID_MEMORY_PERMISSIONS(ErrorModule::Kernel,
                                                     ErrCodes::InvalidMemoryPermissions);
+constexpr ResultCode ERR_INVALID_MEMORY_RANGE(ErrorModule::Kernel, ErrCodes::InvalidMemoryRange);
 constexpr ResultCode ERR_INVALID_HANDLE(ErrorModule::Kernel, ErrCodes::InvalidHandle);
 constexpr ResultCode ERR_INVALID_PROCESSOR_ID(ErrorModule::Kernel, ErrCodes::InvalidProcessorId);
 constexpr ResultCode ERR_INVALID_SIZE(ErrorModule::Kernel, ErrCodes::InvalidSize);
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 98eb74298..bd680adfe 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -116,7 +116,7 @@ struct KernelCore::Impl {
         next_thread_id = 1;
 
         process_list.clear();
-        current_process.reset();
+        current_process = nullptr;
 
         handle_table.Clear();
         resource_limits.fill(nullptr);
@@ -207,7 +207,7 @@ struct KernelCore::Impl {
 
     // Lists all processes that exist in the current session.
     std::vector<SharedPtr<Process>> process_list;
-    SharedPtr<Process> current_process;
+    Process* current_process = nullptr;
 
     Kernel::HandleTable handle_table;
     std::array<SharedPtr<ResourceLimit>, 4> resource_limits;
@@ -266,15 +266,15 @@ void KernelCore::AppendNewProcess(SharedPtr<Process> process) {
     impl->process_list.push_back(std::move(process));
 }
 
-void KernelCore::MakeCurrentProcess(SharedPtr<Process> process) {
-    impl->current_process = std::move(process);
+void KernelCore::MakeCurrentProcess(Process* process) {
+    impl->current_process = process;
 }
 
-SharedPtr<Process>& KernelCore::CurrentProcess() {
+Process* KernelCore::CurrentProcess() {
     return impl->current_process;
 }
 
-const SharedPtr<Process>& KernelCore::CurrentProcess() const {
+const Process* KernelCore::CurrentProcess() const {
     return impl->current_process;
 }
 
diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h
index c0771ecf0..41554821f 100644
--- a/src/core/hle/kernel/kernel.h
+++ b/src/core/hle/kernel/kernel.h
@@ -66,13 +66,13 @@ public:
     void AppendNewProcess(SharedPtr<Process> process);
 
     /// Makes the given process the new current process.
-    void MakeCurrentProcess(SharedPtr<Process> process);
+    void MakeCurrentProcess(Process* process);
 
-    /// Retrieves a reference to the current process.
-    SharedPtr<Process>& CurrentProcess();
+    /// Retrieves a pointer to the current process.
+    Process* CurrentProcess();
 
-    /// Retrieves a const reference to the current process.
-    const SharedPtr<Process>& CurrentProcess() const;
+    /// Retrieves a const pointer to the current process.
+    const Process* CurrentProcess() const;
 
     /// Adds a port to the named port table
     void AddNamedPort(std::string name, SharedPtr<ClientPort> port);
diff --git a/src/core/hle/kernel/scheduler.cpp b/src/core/hle/kernel/scheduler.cpp
index cfd6e1bad..1342c597e 100644
--- a/src/core/hle/kernel/scheduler.cpp
+++ b/src/core/hle/kernel/scheduler.cpp
@@ -9,7 +9,7 @@
 #include "common/logging/log.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
-#include "core/core_timing.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/scheduler.h"
 
@@ -78,16 +78,16 @@ void Scheduler::SwitchContext(Thread* new_thread) {
         // Cancel any outstanding wakeup events for this thread
         new_thread->CancelWakeupTimer();
 
-        auto previous_process = Core::CurrentProcess();
+        auto* const previous_process = Core::CurrentProcess();
 
         current_thread = new_thread;
 
         ready_queue.remove(new_thread->GetPriority(), new_thread);
         new_thread->SetStatus(ThreadStatus::Running);
 
-        const auto thread_owner_process = current_thread->GetOwnerProcess();
+        auto* const thread_owner_process = current_thread->GetOwnerProcess();
         if (previous_process != thread_owner_process) {
-            Core::CurrentProcess() = thread_owner_process;
+            Core::System::GetInstance().Kernel().MakeCurrentProcess(thread_owner_process);
             SetCurrentPageTable(&Core::CurrentProcess()->VMManager().page_table);
         }
 
diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp
index e5dd43275..863ecfa74 100644
--- a/src/core/hle/kernel/svc.cpp
+++ b/src/core/hle/kernel/svc.cpp
@@ -39,6 +39,73 @@ namespace {
 constexpr bool Is4KBAligned(VAddr address) {
     return (address & 0xFFF) == 0;
 }
+
+// Checks if address + size is greater than the given address
+// This can return false if the size causes an overflow of a 64-bit type
+// or if the given size is zero.
+constexpr bool IsValidAddressRange(VAddr address, u64 size) {
+    return address + size > address;
+}
+
+// Checks if a given address range lies within a larger address range.
+constexpr bool IsInsideAddressRange(VAddr address, u64 size, VAddr address_range_begin,
+                                    VAddr address_range_end) {
+    const VAddr end_address = address + size - 1;
+    return address_range_begin <= address && end_address <= address_range_end - 1;
+}
+
+bool IsInsideAddressSpace(const VMManager& vm, VAddr address, u64 size) {
+    return IsInsideAddressRange(address, size, vm.GetAddressSpaceBaseAddress(),
+                                vm.GetAddressSpaceEndAddress());
+}
+
+bool IsInsideNewMapRegion(const VMManager& vm, VAddr address, u64 size) {
+    return IsInsideAddressRange(address, size, vm.GetNewMapRegionBaseAddress(),
+                                vm.GetNewMapRegionEndAddress());
+}
+
+// Helper function that performs the common sanity checks for svcMapMemory
+// and svcUnmapMemory. This is doable, as both functions perform their sanitizing
+// in the same order.
+ResultCode MapUnmapMemorySanityChecks(const VMManager& vm_manager, VAddr dst_addr, VAddr src_addr,
+                                      u64 size) {
+    if (!Is4KBAligned(dst_addr) || !Is4KBAligned(src_addr)) {
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (size == 0 || !Is4KBAligned(size)) {
+        return ERR_INVALID_SIZE;
+    }
+
+    if (!IsValidAddressRange(dst_addr, size)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!IsValidAddressRange(src_addr, size)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!IsInsideAddressSpace(vm_manager, src_addr, size)) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!IsInsideNewMapRegion(vm_manager, dst_addr, size)) {
+        return ERR_INVALID_MEMORY_RANGE;
+    }
+
+    const VAddr dst_end_address = dst_addr + size;
+    if (dst_end_address > vm_manager.GetHeapRegionBaseAddress() &&
+        dst_addr < vm_manager.GetHeapRegionEndAddress()) {
+        return ERR_INVALID_MEMORY_RANGE;
+    }
+
+    if (dst_end_address > vm_manager.GetNewMapRegionBaseAddress() &&
+        dst_addr < vm_manager.GetMapRegionEndAddress()) {
+        return ERR_INVALID_MEMORY_RANGE;
+    }
+
+    return RESULT_SUCCESS;
+}
 } // Anonymous namespace
 
 /// Set the process heap to a given Size. It can both extend and shrink the heap.
@@ -69,15 +136,15 @@ static ResultCode MapMemory(VAddr dst_addr, VAddr src_addr, u64 size) {
     LOG_TRACE(Kernel_SVC, "called, dst_addr=0x{:X}, src_addr=0x{:X}, size=0x{:X}", dst_addr,
               src_addr, size);
 
-    if (!Is4KBAligned(dst_addr) || !Is4KBAligned(src_addr)) {
-        return ERR_INVALID_ADDRESS;
-    }
+    auto* const current_process = Core::CurrentProcess();
+    const auto& vm_manager = current_process->VMManager();
 
-    if (size == 0 || !Is4KBAligned(size)) {
-        return ERR_INVALID_SIZE;
+    const auto result = MapUnmapMemorySanityChecks(vm_manager, dst_addr, src_addr, size);
+    if (result != RESULT_SUCCESS) {
+        return result;
     }
 
-    return Core::CurrentProcess()->MirrorMemory(dst_addr, src_addr, size);
+    return current_process->MirrorMemory(dst_addr, src_addr, size);
 }
 
 /// Unmaps a region that was previously mapped with svcMapMemory
@@ -85,15 +152,15 @@ static ResultCode UnmapMemory(VAddr dst_addr, VAddr src_addr, u64 size) {
     LOG_TRACE(Kernel_SVC, "called, dst_addr=0x{:X}, src_addr=0x{:X}, size=0x{:X}", dst_addr,
               src_addr, size);
 
-    if (!Is4KBAligned(dst_addr) || !Is4KBAligned(src_addr)) {
-        return ERR_INVALID_ADDRESS;
-    }
+    auto* const current_process = Core::CurrentProcess();
+    const auto& vm_manager = current_process->VMManager();
 
-    if (size == 0 || !Is4KBAligned(size)) {
-        return ERR_INVALID_SIZE;
+    const auto result = MapUnmapMemorySanityChecks(vm_manager, dst_addr, src_addr, size);
+    if (result != RESULT_SUCCESS) {
+        return result;
     }
 
-    return Core::CurrentProcess()->UnmapMemory(dst_addr, src_addr, size);
+    return current_process->UnmapMemory(dst_addr, src_addr, size);
 }
 
 /// Connect to an OS service given the port name, returns the handle to the port to out
@@ -341,7 +408,7 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
     LOG_TRACE(Kernel_SVC, "called info_id=0x{:X}, info_sub_id=0x{:X}, handle=0x{:08X}", info_id,
               info_sub_id, handle);
 
-    const auto& current_process = Core::CurrentProcess();
+    const auto* current_process = Core::CurrentProcess();
     const auto& vm_manager = current_process->VMManager();
 
     switch (static_cast<GetInfoType>(info_id)) {
@@ -439,7 +506,7 @@ static ResultCode GetThreadContext(VAddr thread_context, Handle handle) {
         return ERR_INVALID_HANDLE;
     }
 
-    const auto current_process = Core::CurrentProcess();
+    const auto* current_process = Core::CurrentProcess();
     if (thread->GetOwnerProcess() != current_process) {
         return ERR_INVALID_HANDLE;
     }
@@ -531,7 +598,7 @@ static ResultCode MapSharedMemory(Handle shared_memory_handle, VAddr addr, u64 s
         return ERR_INVALID_HANDLE;
     }
 
-    return shared_memory->Map(Core::CurrentProcess().get(), addr, permissions_type,
+    return shared_memory->Map(Core::CurrentProcess(), addr, permissions_type,
                               MemoryPermission::DontCare);
 }
 
@@ -550,7 +617,7 @@ static ResultCode UnmapSharedMemory(Handle shared_memory_handle, VAddr addr, u64
     auto& kernel = Core::System::GetInstance().Kernel();
     auto shared_memory = kernel.HandleTable().Get<SharedMemory>(shared_memory_handle);
 
-    return shared_memory->Unmap(Core::CurrentProcess().get(), addr);
+    return shared_memory->Unmap(Core::CurrentProcess(), addr);
 }
 
 /// Query process memory
@@ -588,7 +655,7 @@ static ResultCode QueryMemory(MemoryInfo* memory_info, PageInfo* page_info, VAdd
 
 /// Exits the current process
 static void ExitProcess() {
-    auto& current_process = Core::CurrentProcess();
+    auto* current_process = Core::CurrentProcess();
 
     LOG_INFO(Kernel_SVC, "Process {} exiting", current_process->GetProcessID());
     ASSERT_MSG(current_process->GetStatus() == ProcessStatus::Running,
@@ -636,7 +703,7 @@ static ResultCode CreateThread(Handle* out_handle, VAddr entry_point, u64 arg, V
     auto& kernel = Core::System::GetInstance().Kernel();
     CASCADE_RESULT(SharedPtr<Thread> thread,
                    Thread::Create(kernel, name, entry_point, priority, arg, processor_id, stack_top,
-                                  Core::CurrentProcess()));
+                                  *Core::CurrentProcess()));
     const auto new_guest_handle = kernel.HandleTable().Create(thread);
     if (new_guest_handle.Failed()) {
         return new_guest_handle.Code();
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index 8e514cf9a..352ce1725 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -183,18 +183,15 @@ void Thread::ResumeFromWait() {
  */
 static void ResetThreadContext(Core::ARM_Interface::ThreadContext& context, VAddr stack_top,
                                VAddr entry_point, u64 arg) {
-    memset(&context, 0, sizeof(Core::ARM_Interface::ThreadContext));
-
+    context = {};
     context.cpu_registers[0] = arg;
     context.pc = entry_point;
     context.sp = stack_top;
-    context.pstate = 0;
-    context.fpcr = 0;
 }
 
 ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name, VAddr entry_point,
                                             u32 priority, u64 arg, s32 processor_id,
-                                            VAddr stack_top, SharedPtr<Process> owner_process) {
+                                            VAddr stack_top, Process& owner_process) {
     // Check if priority is in ranged. Lowest priority -> highest priority id.
     if (priority > THREADPRIO_LOWEST) {
         LOG_ERROR(Kernel_SVC, "Invalid thread priority: {}", priority);
@@ -208,7 +205,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
 
     // TODO(yuriks): Other checks, returning 0xD9001BEA
 
-    if (!Memory::IsValidVirtualAddress(*owner_process, entry_point)) {
+    if (!Memory::IsValidVirtualAddress(owner_process, entry_point)) {
         LOG_ERROR(Kernel_SVC, "(name={}): invalid entry {:016X}", name, entry_point);
         // TODO (bunnei): Find the correct error code to use here
         return ResultCode(-1);
@@ -232,7 +229,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     thread->wait_handle = 0;
     thread->name = std::move(name);
     thread->callback_handle = kernel.ThreadWakeupCallbackHandleTable().Create(thread).Unwrap();
-    thread->owner_process = owner_process;
+    thread->owner_process = &owner_process;
     thread->scheduler = Core::System::GetInstance().Scheduler(processor_id).get();
     thread->scheduler->AddThread(thread, priority);
     thread->tls_address = thread->owner_process->MarkNextAvailableTLSSlotAsUsed(*thread);
@@ -264,7 +261,7 @@ SharedPtr<Thread> SetupMainThread(KernelCore& kernel, VAddr entry_point, u32 pri
     // Initialize new "main" thread
     const VAddr stack_top = owner_process.VMManager().GetTLSIORegionEndAddress();
     auto thread_res = Thread::Create(kernel, "main", entry_point, priority, 0, THREADPROCESSORID_0,
-                                     stack_top, &owner_process);
+                                     stack_top, owner_process);
 
     SharedPtr<Thread> thread = std::move(thread_res).Unwrap();
 
diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h
index c6ffbd28c..f4d7bd235 100644
--- a/src/core/hle/kernel/thread.h
+++ b/src/core/hle/kernel/thread.h
@@ -89,7 +89,7 @@ public:
     static ResultVal<SharedPtr<Thread>> Create(KernelCore& kernel, std::string name,
                                                VAddr entry_point, u32 priority, u64 arg,
                                                s32 processor_id, VAddr stack_top,
-                                               SharedPtr<Process> owner_process);
+                                               Process& owner_process);
 
     std::string GetName() const override {
         return name;
@@ -262,11 +262,11 @@ public:
         return processor_id;
     }
 
-    SharedPtr<Process>& GetOwnerProcess() {
+    Process* GetOwnerProcess() {
         return owner_process;
     }
 
-    const SharedPtr<Process>& GetOwnerProcess() const {
+    const Process* GetOwnerProcess() const {
         return owner_process;
     }
 
@@ -386,7 +386,7 @@ private:
     u64 tpidr_el0 = 0;     ///< TPIDR_EL0 read/write system register.
 
     /// Process that owns this thread
-    SharedPtr<Process> owner_process;
+    Process* owner_process;
 
     /// Objects that the thread is waiting on, in the same order as they were
     /// passed to WaitSynchronization1/N.
diff --git a/src/core/hle/service/audio/hwopus.cpp b/src/core/hle/service/audio/hwopus.cpp
index fc6067e59..7168c6a10 100644
--- a/src/core/hle/service/audio/hwopus.cpp
+++ b/src/core/hle/service/audio/hwopus.cpp
@@ -2,8 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <chrono>
 #include <cstring>
 #include <memory>
+#include <optional>
 #include <vector>
 
 #include <opus.h>
@@ -33,7 +35,8 @@ public:
             {1, nullptr, "SetContext"},
             {2, nullptr, "DecodeInterleavedForMultiStream"},
             {3, nullptr, "SetContextForMultiStream"},
-            {4, nullptr, "Unknown4"},
+            {4, &IHardwareOpusDecoderManager::DecodeInterleavedWithPerformance,
+             "DecodeInterleavedWithPerformance"},
             {5, nullptr, "Unknown5"},
             {6, nullptr, "Unknown6"},
             {7, nullptr, "Unknown7"},
@@ -59,8 +62,31 @@ private:
         ctx.WriteBuffer(samples.data(), samples.size() * sizeof(s16));
     }
 
-    bool Decoder_DecodeInterleaved(u32& consumed, u32& sample_count, const std::vector<u8>& input,
-                                   std::vector<opus_int16>& output) {
+    void DecodeInterleavedWithPerformance(Kernel::HLERequestContext& ctx) {
+        u32 consumed = 0;
+        u32 sample_count = 0;
+        u64 performance = 0;
+        std::vector<opus_int16> samples(ctx.GetWriteBufferSize() / sizeof(opus_int16));
+        if (!Decoder_DecodeInterleaved(consumed, sample_count, ctx.ReadBuffer(), samples,
+                                       performance)) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            // TODO(ogniK): Use correct error code
+            rb.Push(ResultCode(-1));
+            return;
+        }
+        IPC::ResponseBuilder rb{ctx, 6};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u32>(consumed);
+        rb.Push<u64>(performance);
+        rb.Push<u32>(sample_count);
+        ctx.WriteBuffer(samples.data(), samples.size() * sizeof(s16));
+    }
+
+    bool Decoder_DecodeInterleaved(
+        u32& consumed, u32& sample_count, const std::vector<u8>& input,
+        std::vector<opus_int16>& output,
+        std::optional<std::reference_wrapper<u64>> performance_time = std::nullopt) {
+        const auto start_time = std::chrono::high_resolution_clock::now();
         std::size_t raw_output_sz = output.size() * sizeof(opus_int16);
         if (sizeof(OpusHeader) > input.size())
             return false;
@@ -80,8 +106,13 @@ private:
                         (static_cast<int>(raw_output_sz / sizeof(s16) / channel_count)), 0);
         if (out_sample_count < 0)
             return false;
+        const auto end_time = std::chrono::high_resolution_clock::now() - start_time;
         sample_count = out_sample_count;
         consumed = static_cast<u32>(sizeof(OpusHeader) + hdr.sz);
+        if (performance_time.has_value()) {
+            performance_time->get() =
+                std::chrono::duration_cast<std::chrono::milliseconds>(end_time).count();
+        }
         return true;
     }
 
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
index 7555bbe7d..c41ef7058 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
@@ -15,6 +15,11 @@
 #include "video_core/renderer_base.h"
 
 namespace Service::Nvidia::Devices {
+namespace NvErrCodes {
+enum {
+    InvalidNmapHandle = -22,
+};
+}
 
 nvhost_as_gpu::nvhost_as_gpu(std::shared_ptr<nvmap> nvmap_dev) : nvmap_dev(std::move(nvmap_dev)) {}
 nvhost_as_gpu::~nvhost_as_gpu() = default;
@@ -79,14 +84,16 @@ u32 nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& output)
     std::memcpy(entries.data(), input.data(), input.size());
 
     auto& gpu = Core::System::GetInstance().GPU();
-
     for (const auto& entry : entries) {
         LOG_WARNING(Service_NVDRV, "remap entry, offset=0x{:X} handle=0x{:X} pages=0x{:X}",
                     entry.offset, entry.nvmap_handle, entry.pages);
         Tegra::GPUVAddr offset = static_cast<Tegra::GPUVAddr>(entry.offset) << 0x10;
-
         auto object = nvmap_dev->GetObject(entry.nvmap_handle);
-        ASSERT(object);
+        if (!object) {
+            LOG_CRITICAL(Service_NVDRV, "nvmap {} is an invalid handle!", entry.nvmap_handle);
+            std::memcpy(output.data(), entries.data(), output.size());
+            return static_cast<u32>(NvErrCodes::InvalidNmapHandle);
+        }
 
         ASSERT(object->status == nvmap::Object::Status::Allocated);
 
@@ -167,10 +174,11 @@ u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& ou
     auto& system_instance = Core::System::GetInstance();
 
     // Remove this memory region from the rasterizer cache.
-    system_instance.Renderer().Rasterizer().FlushAndInvalidateRegion(params.offset,
-                                                                     itr->second.size);
-
     auto& gpu = system_instance.GPU();
+    auto cpu_addr = gpu.MemoryManager().GpuToCpuAddress(params.offset);
+    ASSERT(cpu_addr);
+    system_instance.Renderer().Rasterizer().FlushAndInvalidateRegion(*cpu_addr, itr->second.size);
+
     params.offset = gpu.MemoryManager().UnmapBuffer(params.offset, itr->second.size);
 
     buffer_mappings.erase(itr->second.offset);
diff --git a/src/core/hle/service/nvdrv/devices/nvmap.cpp b/src/core/hle/service/nvdrv/devices/nvmap.cpp
index a2287cc1b..43651d8a6 100644
--- a/src/core/hle/service/nvdrv/devices/nvmap.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvmap.cpp
@@ -11,6 +11,13 @@
 
 namespace Service::Nvidia::Devices {
 
+namespace NvErrCodes {
+enum {
+    OperationNotPermitted = -1,
+    InvalidValue = -22,
+};
+}
+
 nvmap::nvmap() = default;
 nvmap::~nvmap() = default;
 
@@ -44,7 +51,11 @@ u32 nvmap::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& o
 u32 nvmap::IocCreate(const std::vector<u8>& input, std::vector<u8>& output) {
     IocCreateParams params;
     std::memcpy(&params, input.data(), sizeof(params));
+    LOG_DEBUG(Service_NVDRV, "size=0x{:08X}", params.size);
 
+    if (!params.size) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
     // Create a new nvmap object and obtain a handle to it.
     auto object = std::make_shared<Object>();
     object->id = next_id++;
@@ -55,8 +66,6 @@ u32 nvmap::IocCreate(const std::vector<u8>& input, std::vector<u8>& output) {
     u32 handle = next_handle++;
     handles[handle] = std::move(object);
 
-    LOG_DEBUG(Service_NVDRV, "size=0x{:08X}", params.size);
-
     params.handle = handle;
 
     std::memcpy(output.data(), &params, sizeof(params));
@@ -66,9 +75,29 @@ u32 nvmap::IocCreate(const std::vector<u8>& input, std::vector<u8>& output) {
 u32 nvmap::IocAlloc(const std::vector<u8>& input, std::vector<u8>& output) {
     IocAllocParams params;
     std::memcpy(&params, input.data(), sizeof(params));
+    LOG_DEBUG(Service_NVDRV, "called, addr={:X}", params.addr);
+
+    if (!params.handle) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
+
+    if ((params.align - 1) & params.align) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
+
+    const u32 min_alignment = 0x1000;
+    if (params.align < min_alignment) {
+        params.align = min_alignment;
+    }
 
     auto object = GetObject(params.handle);
-    ASSERT(object);
+    if (!object) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
+
+    if (object->status == Object::Status::Allocated) {
+        return static_cast<u32>(NvErrCodes::OperationNotPermitted);
+    }
 
     object->flags = params.flags;
     object->align = params.align;
@@ -76,8 +105,6 @@ u32 nvmap::IocAlloc(const std::vector<u8>& input, std::vector<u8>& output) {
     object->addr = params.addr;
     object->status = Object::Status::Allocated;
 
-    LOG_DEBUG(Service_NVDRV, "called, addr={:X}", params.addr);
-
     std::memcpy(output.data(), &params, sizeof(params));
     return 0;
 }
@@ -88,8 +115,14 @@ u32 nvmap::IocGetId(const std::vector<u8>& input, std::vector<u8>& output) {
 
     LOG_WARNING(Service_NVDRV, "called");
 
+    if (!params.handle) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
+
     auto object = GetObject(params.handle);
-    ASSERT(object);
+    if (!object) {
+        return static_cast<u32>(NvErrCodes::OperationNotPermitted);
+    }
 
     params.id = object->id;
 
@@ -105,7 +138,14 @@ u32 nvmap::IocFromId(const std::vector<u8>& input, std::vector<u8>& output) {
 
     auto itr = std::find_if(handles.begin(), handles.end(),
                             [&](const auto& entry) { return entry.second->id == params.id; });
-    ASSERT(itr != handles.end());
+    if (itr == handles.end()) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
+
+    auto& object = itr->second;
+    if (object->status != Object::Status::Allocated) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
 
     itr->second->refcount++;
 
@@ -125,8 +165,13 @@ u32 nvmap::IocParam(const std::vector<u8>& input, std::vector<u8>& output) {
     LOG_WARNING(Service_NVDRV, "(STUBBED) called type={}", params.param);
 
     auto object = GetObject(params.handle);
-    ASSERT(object);
-    ASSERT(object->status == Object::Status::Allocated);
+    if (!object) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
+
+    if (object->status != Object::Status::Allocated) {
+        return static_cast<u32>(NvErrCodes::OperationNotPermitted);
+    }
 
     switch (static_cast<ParamTypes>(params.param)) {
     case ParamTypes::Size:
@@ -163,9 +208,12 @@ u32 nvmap::IocFree(const std::vector<u8>& input, std::vector<u8>& output) {
     LOG_WARNING(Service_NVDRV, "(STUBBED) called");
 
     auto itr = handles.find(params.handle);
-    ASSERT(itr != handles.end());
-
-    ASSERT(itr->second->refcount > 0);
+    if (itr == handles.end()) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
+    if (!itr->second->refcount) {
+        return static_cast<u32>(NvErrCodes::InvalidValue);
+    }
 
     itr->second->refcount--;
 
diff --git a/src/core/loader/nsp.cpp b/src/core/loader/nsp.cpp
index 5534ce01c..13e57848d 100644
--- a/src/core/loader/nsp.cpp
+++ b/src/core/loader/nsp.cpp
@@ -35,7 +35,7 @@ AppLoader_NSP::AppLoader_NSP(FileSys::VirtualFile file)
         return;
 
     std::tie(nacp_file, icon_file) =
-        FileSys::PatchManager(nsp->GetProgramTitleID()).ParseControlNCA(control_nca);
+        FileSys::PatchManager(nsp->GetProgramTitleID()).ParseControlNCA(*control_nca);
 }
 
 AppLoader_NSP::~AppLoader_NSP() = default;
diff --git a/src/core/loader/nsp.h b/src/core/loader/nsp.h
index b006594a6..db91cd01e 100644
--- a/src/core/loader/nsp.h
+++ b/src/core/loader/nsp.h
@@ -49,7 +49,7 @@ private:
     std::unique_ptr<AppLoader> secondary_loader;
 
     FileSys::VirtualFile icon_file;
-    std::shared_ptr<FileSys::NACP> nacp_file;
+    std::unique_ptr<FileSys::NACP> nacp_file;
     u64 title_id;
 };
 
diff --git a/src/core/loader/xci.cpp b/src/core/loader/xci.cpp
index ee5452eb9..7a619acb4 100644
--- a/src/core/loader/xci.cpp
+++ b/src/core/loader/xci.cpp
@@ -30,7 +30,7 @@ AppLoader_XCI::AppLoader_XCI(FileSys::VirtualFile file)
         return;
 
     std::tie(nacp_file, icon_file) =
-        FileSys::PatchManager(xci->GetProgramTitleID()).ParseControlNCA(control_nca);
+        FileSys::PatchManager(xci->GetProgramTitleID()).ParseControlNCA(*control_nca);
 }
 
 AppLoader_XCI::~AppLoader_XCI() = default;
diff --git a/src/core/loader/xci.h b/src/core/loader/xci.h
index 770ed1437..46f8dfc9e 100644
--- a/src/core/loader/xci.h
+++ b/src/core/loader/xci.h
@@ -49,7 +49,7 @@ private:
     std::unique_ptr<AppLoader_NCA> nca_loader;
 
     FileSys::VirtualFile icon_file;
-    std::shared_ptr<FileSys::NACP> nacp_file;
+    std::unique_ptr<FileSys::NACP> nacp_file;
 };
 
 } // namespace Loader
diff --git a/src/tests/core/arm/arm_test_common.cpp b/src/tests/core/arm/arm_test_common.cpp
index c0a57e71f..37e15bad0 100644
--- a/src/tests/core/arm/arm_test_common.cpp
+++ b/src/tests/core/arm/arm_test_common.cpp
@@ -15,7 +15,8 @@ namespace ArmTests {
 TestEnvironment::TestEnvironment(bool mutable_memory_)
     : mutable_memory(mutable_memory_), test_memory(std::make_shared<TestMemory>(this)) {
 
-    Core::CurrentProcess() = Kernel::Process::Create(kernel, "");
+    auto process = Kernel::Process::Create(kernel, "");
+    kernel.MakeCurrentProcess(process.get());
     page_table = &Core::CurrentProcess()->VMManager().page_table;
 
     std::fill(page_table->pointers.begin(), page_table->pointers.end(), nullptr);
diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h
index 81d15c62a..2a6e8bbbb 100644
--- a/src/video_core/engines/fermi_2d.h
+++ b/src/video_core/engines/fermi_2d.h
@@ -36,9 +36,9 @@ public:
             RenderTargetFormat format;
             BitField<0, 1, u32> linear;
             union {
-                BitField<0, 4, u32> block_depth;
+                BitField<0, 4, u32> block_width;
                 BitField<4, 4, u32> block_height;
-                BitField<8, 4, u32> block_width;
+                BitField<8, 4, u32> block_depth;
             };
             u32 depth;
             u32 layer;
@@ -53,10 +53,20 @@ public:
                                              address_low);
             }
 
+            u32 BlockWidth() const {
+                // The block width is stored in log2 format.
+                return 1 << block_width;
+            }
+
             u32 BlockHeight() const {
                 // The block height is stored in log2 format.
                 return 1 << block_height;
             }
+
+            u32 BlockDepth() const {
+                // The block depth is stored in log2 format.
+                return 1 << block_depth;
+            }
         };
         static_assert(sizeof(Surface) == 0x28, "Surface has incorrect size");
 
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 4290da33f..c8d1b6478 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -347,6 +347,16 @@ public:
             DecrWrap = 8,
         };
 
+        enum class MemoryLayout : u32 {
+            Linear = 0,
+            BlockLinear = 1,
+        };
+
+        enum class InvMemoryLayout : u32 {
+            BlockLinear = 0,
+            Linear = 1,
+        };
+
         struct Cull {
             enum class FrontFace : u32 {
                 ClockWise = 0x0900,
@@ -432,7 +442,12 @@ public:
             u32 width;
             u32 height;
             Tegra::RenderTargetFormat format;
-            u32 block_dimensions;
+            union {
+                BitField<0, 3, u32> block_width;
+                BitField<4, 3, u32> block_height;
+                BitField<8, 3, u32> block_depth;
+                BitField<12, 1, InvMemoryLayout> type;
+            } memory_layout;
             u32 array_mode;
             u32 layer_stride;
             u32 base_layer;
@@ -532,7 +547,21 @@ public:
                 INSERT_PADDING_WORDS(0x3);
                 s32 clear_stencil;
 
-                INSERT_PADDING_WORDS(0x6C);
+                INSERT_PADDING_WORDS(0x17);
+
+                struct {
+                    u32 enable;
+                    union {
+                        BitField<0, 16, u32> min_x;
+                        BitField<16, 16, u32> max_x;
+                    };
+                    union {
+                        BitField<0, 16, u32> min_y;
+                        BitField<16, 16, u32> max_y;
+                    };
+                } scissor_test;
+
+                INSERT_PADDING_WORDS(0x52);
 
                 s32 stencil_back_func_ref;
                 u32 stencil_back_mask;
@@ -548,7 +577,12 @@ public:
                     u32 address_high;
                     u32 address_low;
                     Tegra::DepthFormat format;
-                    u32 block_dimensions;
+                    union {
+                        BitField<0, 4, u32> block_width;
+                        BitField<4, 4, u32> block_height;
+                        BitField<8, 4, u32> block_depth;
+                        BitField<20, 1, InvMemoryLayout> type;
+                    } memory_layout;
                     u32 layer_stride;
 
                     GPUVAddr Address() const {
@@ -1002,6 +1036,7 @@ ASSERT_REG_POSITION(vertex_buffer, 0x35D);
 ASSERT_REG_POSITION(clear_color[0], 0x360);
 ASSERT_REG_POSITION(clear_depth, 0x364);
 ASSERT_REG_POSITION(clear_stencil, 0x368);
+ASSERT_REG_POSITION(scissor_test, 0x380);
 ASSERT_REG_POSITION(stencil_back_func_ref, 0x3D5);
 ASSERT_REG_POSITION(stencil_back_mask, 0x3D6);
 ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7);
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index b1f137b9c..9a59b65b3 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -214,6 +214,18 @@ enum class IMinMaxExchange : u64 {
     XHi = 3,
 };
 
+enum class VmadType : u64 {
+    Size16_Low = 0,
+    Size16_High = 1,
+    Size32 = 2,
+    Invalid = 3,
+};
+
+enum class VmadShr : u64 {
+    Shr7 = 1,
+    Shr15 = 2,
+};
+
 enum class XmadMode : u64 {
     None = 0,
     CLo = 1,
@@ -314,6 +326,15 @@ enum class TextureMiscMode : u64 {
     PTP,
 };
 
+enum class IsberdMode : u64 {
+    None = 0,
+    Patch = 1,
+    Prim = 2,
+    Attr = 3,
+};
+
+enum class IsberdShift : u64 { None = 0, U16 = 1, B32 = 2 };
+
 enum class IpaInterpMode : u64 {
     Linear = 0,
     Perspective = 1,
@@ -340,6 +361,87 @@ struct IpaMode {
     }
 };
 
+enum class SystemVariable : u64 {
+    LaneId = 0x00,
+    VirtCfg = 0x02,
+    VirtId = 0x03,
+    Pm0 = 0x04,
+    Pm1 = 0x05,
+    Pm2 = 0x06,
+    Pm3 = 0x07,
+    Pm4 = 0x08,
+    Pm5 = 0x09,
+    Pm6 = 0x0a,
+    Pm7 = 0x0b,
+    OrderingTicket = 0x0f,
+    PrimType = 0x10,
+    InvocationId = 0x11,
+    Ydirection = 0x12,
+    ThreadKill = 0x13,
+    ShaderType = 0x14,
+    DirectBeWriteAddressLow = 0x15,
+    DirectBeWriteAddressHigh = 0x16,
+    DirectBeWriteEnabled = 0x17,
+    MachineId0 = 0x18,
+    MachineId1 = 0x19,
+    MachineId2 = 0x1a,
+    MachineId3 = 0x1b,
+    Affinity = 0x1c,
+    InvocationInfo = 0x1d,
+    WscaleFactorXY = 0x1e,
+    WscaleFactorZ = 0x1f,
+    Tid = 0x20,
+    TidX = 0x21,
+    TidY = 0x22,
+    TidZ = 0x23,
+    CtaParam = 0x24,
+    CtaIdX = 0x25,
+    CtaIdY = 0x26,
+    CtaIdZ = 0x27,
+    NtId = 0x28,
+    CirQueueIncrMinusOne = 0x29,
+    Nlatc = 0x2a,
+    SmSpaVersion = 0x2c,
+    MultiPassShaderInfo = 0x2d,
+    LwinHi = 0x2e,
+    SwinHi = 0x2f,
+    SwinLo = 0x30,
+    SwinSz = 0x31,
+    SmemSz = 0x32,
+    SmemBanks = 0x33,
+    LwinLo = 0x34,
+    LwinSz = 0x35,
+    LmemLosz = 0x36,
+    LmemHioff = 0x37,
+    EqMask = 0x38,
+    LtMask = 0x39,
+    LeMask = 0x3a,
+    GtMask = 0x3b,
+    GeMask = 0x3c,
+    RegAlloc = 0x3d,
+    CtxAddr = 0x3e,      // .fmask = F_SM50
+    BarrierAlloc = 0x3e, // .fmask = F_SM60
+    GlobalErrorStatus = 0x40,
+    WarpErrorStatus = 0x42,
+    WarpErrorStatusClear = 0x43,
+    PmHi0 = 0x48,
+    PmHi1 = 0x49,
+    PmHi2 = 0x4a,
+    PmHi3 = 0x4b,
+    PmHi4 = 0x4c,
+    PmHi5 = 0x4d,
+    PmHi6 = 0x4e,
+    PmHi7 = 0x4f,
+    ClockLo = 0x50,
+    ClockHi = 0x51,
+    GlobalTimerLo = 0x52,
+    GlobalTimerHi = 0x53,
+    HwTaskId = 0x60,
+    CircularQueueEntryIndex = 0x61,
+    CircularQueueEntryAddressLow = 0x62,
+    CircularQueueEntryAddressHigh = 0x63,
+};
+
 union Instruction {
     Instruction& operator=(const Instruction& instr) {
         value = instr.value;
@@ -362,6 +464,7 @@ union Instruction {
     BitField<48, 16, u64> opcode;
 
     union {
+        BitField<20, 16, u64> imm20_16;
         BitField<20, 19, u64> imm20_19;
         BitField<20, 32, s64> imm20_32;
         BitField<45, 1, u64> negate_b;
@@ -403,6 +506,10 @@ union Instruction {
             }
         } lop3;
 
+        u16 GetImm20_16() const {
+            return static_cast<u16>(imm20_16);
+        }
+
         u32 GetImm20_19() const {
             u32 imm{static_cast<u32>(imm20_19)};
             imm <<= 12;
@@ -915,6 +1022,35 @@ union Instruction {
     } bra;
 
     union {
+        BitField<39, 1, u64> emit; // EmitVertex
+        BitField<40, 1, u64> cut;  // EndPrimitive
+    } out;
+
+    union {
+        BitField<31, 1, u64> skew;
+        BitField<32, 1, u64> o;
+        BitField<33, 2, IsberdMode> mode;
+        BitField<47, 2, IsberdShift> shift;
+    } isberd;
+
+    union {
+        BitField<48, 1, u64> signed_a;
+        BitField<38, 1, u64> is_byte_chunk_a;
+        BitField<36, 2, VmadType> type_a;
+        BitField<36, 2, u64> byte_height_a;
+
+        BitField<49, 1, u64> signed_b;
+        BitField<50, 1, u64> use_register_b;
+        BitField<30, 1, u64> is_byte_chunk_b;
+        BitField<28, 2, VmadType> type_b;
+        BitField<28, 2, u64> byte_height_b;
+
+        BitField<51, 2, VmadShr> shr;
+        BitField<55, 1, u64> saturate; // Saturates the result (a * b + c)
+        BitField<47, 1, u64> cc;
+    } vmad;
+
+    union {
         BitField<20, 16, u64> imm20_16;
         BitField<36, 1, u64> product_shift_left;
         BitField<37, 1, u64> merge_37;
@@ -936,6 +1072,10 @@ union Instruction {
         BitField<36, 5, u64> index;
     } cbuf36;
 
+    // Unsure about the size of this one.
+    // It's always used with a gpr0, so any size should be fine.
+    BitField<20, 8, SystemVariable> sys20;
+
     BitField<47, 1, u64> generates_cc;
     BitField<61, 1, u64> is_b_imm;
     BitField<60, 1, u64> is_b_gpr;
@@ -975,6 +1115,9 @@ public:
         TMML,   // Texture Mip Map Level
         EXIT,
         IPA,
+        OUT_R, // Emit vertex/primitive
+        ISBERD,
+        VMAD,
         FFMA_IMM, // Fused Multiply and Add
         FFMA_CR,
         FFMA_RC,
@@ -1034,6 +1177,7 @@ public:
         MOV_C,
         MOV_R,
         MOV_IMM,
+        MOV_SYS,
         MOV32_IMM,
         SHL_C,
         SHL_R,
@@ -1209,6 +1353,9 @@ private:
             INST("1101111101011---", Id::TMML, Type::Memory, "TMML"),
             INST("111000110000----", Id::EXIT, Type::Trivial, "EXIT"),
             INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
+            INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),
+            INST("1110111111010---", Id::ISBERD, Type::Trivial, "ISBERD"),
+            INST("01011111--------", Id::VMAD, Type::Trivial, "VMAD"),
             INST("0011001-1-------", Id::FFMA_IMM, Type::Ffma, "FFMA_IMM"),
             INST("010010011-------", Id::FFMA_CR, Type::Ffma, "FFMA_CR"),
             INST("010100011-------", Id::FFMA_RC, Type::Ffma, "FFMA_RC"),
@@ -1255,6 +1402,7 @@ private:
             INST("0100110010011---", Id::MOV_C, Type::Arithmetic, "MOV_C"),
             INST("0101110010011---", Id::MOV_R, Type::Arithmetic, "MOV_R"),
             INST("0011100-10011---", Id::MOV_IMM, Type::Arithmetic, "MOV_IMM"),
+            INST("1111000011001---", Id::MOV_SYS, Type::Trivial, "MOV_SYS"),
             INST("000000010000----", Id::MOV32_IMM, Type::ArithmeticImmediate, "MOV32_IMM"),
             INST("0100110001100---", Id::FMNMX_C, Type::Arithmetic, "FMNMX_C"),
             INST("0101110001100---", Id::FMNMX_R, Type::Arithmetic, "FMNMX_R"),
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 209bdf181..84582c777 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -255,7 +255,7 @@ DrawParameters RasterizerOpenGL::SetupDraw() {
     return params;
 }
 
-void RasterizerOpenGL::SetupShaders() {
+void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     MICROPROFILE_SCOPE(OpenGL_Shader);
     const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
 
@@ -270,6 +270,11 @@ void RasterizerOpenGL::SetupShaders() {
 
         // Skip stages that are not enabled
         if (!gpu.regs.IsShaderConfigEnabled(index)) {
+            switch (program) {
+            case Maxwell::ShaderProgram::Geometry:
+                shader_program_manager->UseTrivialGeometryShader();
+                break;
+            }
             continue;
         }
 
@@ -288,11 +293,18 @@ void RasterizerOpenGL::SetupShaders() {
         switch (program) {
         case Maxwell::ShaderProgram::VertexA:
         case Maxwell::ShaderProgram::VertexB: {
-            shader_program_manager->UseProgrammableVertexShader(shader->GetProgramHandle());
+            shader_program_manager->UseProgrammableVertexShader(
+                shader->GetProgramHandle(primitive_mode));
+            break;
+        }
+        case Maxwell::ShaderProgram::Geometry: {
+            shader_program_manager->UseProgrammableGeometryShader(
+                shader->GetProgramHandle(primitive_mode));
             break;
         }
         case Maxwell::ShaderProgram::Fragment: {
-            shader_program_manager->UseProgrammableFragmentShader(shader->GetProgramHandle());
+            shader_program_manager->UseProgrammableFragmentShader(
+                shader->GetProgramHandle(primitive_mode));
             break;
         }
         default:
@@ -302,12 +314,13 @@ void RasterizerOpenGL::SetupShaders() {
         }
 
         // Configure the const buffers for this shader stage.
-        current_constbuffer_bindpoint = SetupConstBuffers(static_cast<Maxwell::ShaderStage>(stage),
-                                                          shader, current_constbuffer_bindpoint);
+        current_constbuffer_bindpoint =
+            SetupConstBuffers(static_cast<Maxwell::ShaderStage>(stage), shader, primitive_mode,
+                              current_constbuffer_bindpoint);
 
         // Configure the textures for this shader stage.
         current_texture_bindpoint = SetupTextures(static_cast<Maxwell::ShaderStage>(stage), shader,
-                                                  current_texture_bindpoint);
+                                                  primitive_mode, current_texture_bindpoint);
 
         // When VertexA is enabled, we have dual vertex shaders
         if (program == Maxwell::ShaderProgram::VertexA) {
@@ -317,8 +330,6 @@ void RasterizerOpenGL::SetupShaders() {
     }
 
     state.Apply();
-
-    shader_program_manager->UseTrivialGeometryShader();
 }
 
 std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
@@ -541,6 +552,7 @@ void RasterizerOpenGL::DrawArrays() {
     SyncLogicOpState();
     SyncCullMode();
     SyncAlphaTest();
+    SyncScissorTest();
     SyncTransformFeedback();
     SyncPointState();
 
@@ -580,7 +592,7 @@ void RasterizerOpenGL::DrawArrays() {
 
     SetupVertexArrays();
     DrawParameters params = SetupDraw();
-    SetupShaders();
+    SetupShaders(params.primitive_mode);
 
     buffer_cache.Unmap();
 
@@ -719,7 +731,7 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr
 }
 
 u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, Shader& shader,
-                                        u32 current_bindpoint) {
+                                        GLenum primitive_mode, u32 current_bindpoint) {
     MICROPROFILE_SCOPE(OpenGL_UBO);
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
@@ -771,7 +783,7 @@ u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, Shader& shad
             buffer.address, size, static_cast<std::size_t>(uniform_buffer_alignment));
 
         // Now configure the bindpoint of the buffer inside the shader
-        glUniformBlockBinding(shader->GetProgramHandle(),
+        glUniformBlockBinding(shader->GetProgramHandle(primitive_mode),
                               shader->GetProgramResourceIndex(used_buffer),
                               current_bindpoint + bindpoint);
 
@@ -787,7 +799,8 @@ u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, Shader& shad
     return current_bindpoint + static_cast<u32>(entries.size());
 }
 
-u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader, u32 current_unit) {
+u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
+                                    GLenum primitive_mode, u32 current_unit) {
     MICROPROFILE_SCOPE(OpenGL_Texture);
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
@@ -802,8 +815,8 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
 
         // Bind the uniform to the sampler.
 
-        glProgramUniform1i(shader->GetProgramHandle(), shader->GetUniformLocation(entry),
-                           current_bindpoint);
+        glProgramUniform1i(shader->GetProgramHandle(primitive_mode),
+                           shader->GetUniformLocation(entry), current_bindpoint);
 
         const auto texture = maxwell3d.GetStageTexture(entry.GetStage(), entry.GetOffset());
 
@@ -972,6 +985,22 @@ void RasterizerOpenGL::SyncAlphaTest() {
     }
 }
 
+void RasterizerOpenGL::SyncScissorTest() {
+    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+
+    state.scissor.enabled = (regs.scissor_test.enable != 0);
+    // TODO(Blinkhawk): Figure if the hardware supports scissor testing per viewport and how it's
+    // implemented.
+    if (regs.scissor_test.enable != 0) {
+        const u32 width = regs.scissor_test.max_x - regs.scissor_test.min_x;
+        const u32 height = regs.scissor_test.max_y - regs.scissor_test.min_y;
+        state.scissor.x = regs.scissor_test.min_x;
+        state.scissor.y = regs.scissor_test.min_y;
+        state.scissor.width = width;
+        state.scissor.height = height;
+    }
+}
+
 void RasterizerOpenGL::SyncTransformFeedback() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
 
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 0dab2018b..b1f7ccc7e 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -120,7 +120,7 @@ private:
      * @returns The next available bindpoint for use in the next shader stage.
      */
     u32 SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, Shader& shader,
-                          u32 current_bindpoint);
+                          GLenum primitive_mode, u32 current_bindpoint);
 
     /*
      * Configures the current textures to use for the draw command.
@@ -130,7 +130,7 @@ private:
      * @returns The next available bindpoint for use in the next shader stage.
      */
     u32 SetupTextures(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, Shader& shader,
-                      u32 current_unit);
+                      GLenum primitive_mode, u32 current_unit);
 
     /// Syncs the viewport to match the guest state
     void SyncViewport();
@@ -165,6 +165,9 @@ private:
     /// Syncs the alpha test state to match the guest state
     void SyncAlphaTest();
 
+    /// Syncs the scissor test state to match the guest state
+    void SyncScissorTest();
+
     /// Syncs the transform feedback state to match the guest state
     void SyncTransformFeedback();
 
@@ -207,7 +210,7 @@ private:
 
     DrawParameters SetupDraw();
 
-    void SetupShaders();
+    void SetupShaders(GLenum primitive_mode);
 
     enum class AccelDraw { Disabled, Arrays, Indexed };
     AccelDraw accelerate_draw = AccelDraw::Disabled;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 56ff83eff..65a220c41 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -45,7 +45,9 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     SurfaceParams params{};
     params.addr = TryGetCpuAddr(config.tic.Address());
     params.is_tiled = config.tic.IsTiled();
+    params.block_width = params.is_tiled ? config.tic.BlockWidth() : 0,
     params.block_height = params.is_tiled ? config.tic.BlockHeight() : 0,
+    params.block_depth = params.is_tiled ? config.tic.BlockDepth() : 0,
     params.pixel_format =
         PixelFormatFromTextureFormat(config.tic.format, config.tic.r_type.Value());
     params.component_type = ComponentTypeFromTexture(config.tic.r_type.Value());
@@ -97,8 +99,11 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     const auto& config{Core::System::GetInstance().GPU().Maxwell3D().regs.rt[index]};
     SurfaceParams params{};
     params.addr = TryGetCpuAddr(config.Address());
-    params.is_tiled = true;
-    params.block_height = Tegra::Texture::TICEntry::DefaultBlockHeight;
+    params.is_tiled =
+        config.memory_layout.type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
+    params.block_width = 1 << config.memory_layout.block_width;
+    params.block_height = 1 << config.memory_layout.block_height;
+    params.block_depth = 1 << config.memory_layout.block_depth;
     params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
     params.component_type = ComponentTypeFromRenderTarget(config.format);
     params.type = GetFormatType(params.pixel_format);
@@ -120,13 +125,16 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     return params;
 }
 
-/*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(u32 zeta_width, u32 zeta_height,
-                                                             Tegra::GPUVAddr zeta_address,
-                                                             Tegra::DepthFormat format) {
+/*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(
+    u32 zeta_width, u32 zeta_height, Tegra::GPUVAddr zeta_address, Tegra::DepthFormat format,
+    u32 block_width, u32 block_height, u32 block_depth,
+    Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type) {
     SurfaceParams params{};
     params.addr = TryGetCpuAddr(zeta_address);
-    params.is_tiled = true;
-    params.block_height = Tegra::Texture::TICEntry::DefaultBlockHeight;
+    params.is_tiled = type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
+    params.block_width = 1 << std::min(block_width, 5U);
+    params.block_height = 1 << std::min(block_height, 5U);
+    params.block_depth = 1 << std::min(block_depth, 5U);
     params.pixel_format = PixelFormatFromDepthFormat(format);
     params.component_type = ComponentTypeFromDepthFormat(format);
     params.type = GetFormatType(params.pixel_format);
@@ -148,7 +156,9 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     SurfaceParams params{};
     params.addr = TryGetCpuAddr(config.Address());
     params.is_tiled = !config.linear;
-    params.block_height = params.is_tiled ? config.BlockHeight() : 0,
+    params.block_width = params.is_tiled ? std::min(config.BlockWidth(), 32U) : 0,
+    params.block_height = params.is_tiled ? std::min(config.BlockHeight(), 32U) : 0,
+    params.block_depth = params.is_tiled ? std::min(config.BlockDepth(), 32U) : 0,
     params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
     params.component_type = ComponentTypeFromRenderTarget(config.format);
     params.type = GetFormatType(params.pixel_format);
@@ -818,6 +828,11 @@ void CachedSurface::LoadGLBuffer() {
     if (params.is_tiled) {
         gl_buffer.resize(total_size);
 
+        ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
+                   params.block_width, static_cast<u32>(params.target));
+        ASSERT_MSG(params.block_depth == 1, "Block depth is defined as {} on texture type {}",
+                   params.block_depth, static_cast<u32>(params.target));
+
         // TODO(bunnei): This only unswizzles and copies a 2D texture - we do not yet know how to do
         // this for 3D textures, etc.
         switch (params.target) {
@@ -989,7 +1004,9 @@ Surface RasterizerCacheOpenGL::GetDepthBufferSurface(bool preserve_contents) {
     }
 
     SurfaceParams depth_params{SurfaceParams::CreateForDepthBuffer(
-        regs.zeta_width, regs.zeta_height, regs.zeta.Address(), regs.zeta.format)};
+        regs.zeta_width, regs.zeta_height, regs.zeta.Address(), regs.zeta.format,
+        regs.zeta.memory_layout.block_width, regs.zeta.memory_layout.block_height,
+        regs.zeta.memory_layout.block_depth, regs.zeta.memory_layout.type)};
 
     return GetSurface(depth_params, preserve_contents);
 }
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index 0b4940b3c..66d98ad4e 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -716,9 +716,10 @@ struct SurfaceParams {
     static SurfaceParams CreateForFramebuffer(std::size_t index);
 
     /// Creates SurfaceParams for a depth buffer configuration
-    static SurfaceParams CreateForDepthBuffer(u32 zeta_width, u32 zeta_height,
-                                              Tegra::GPUVAddr zeta_address,
-                                              Tegra::DepthFormat format);
+    static SurfaceParams CreateForDepthBuffer(
+        u32 zeta_width, u32 zeta_height, Tegra::GPUVAddr zeta_address, Tegra::DepthFormat format,
+        u32 block_width, u32 block_height, u32 block_depth,
+        Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type);
 
     /// Creates SurfaceParams for a Fermi2D surface copy
     static SurfaceParams CreateForFermiCopySurface(
@@ -733,7 +734,9 @@ struct SurfaceParams {
 
     VAddr addr;
     bool is_tiled;
+    u32 block_width;
     u32 block_height;
+    u32 block_depth;
     PixelFormat pixel_format;
     ComponentType component_type;
     SurfaceType type;
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 7cd8f91e4..1a03a677f 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -68,6 +68,10 @@ CachedShader::CachedShader(VAddr addr, Maxwell::ShaderProgram program_type)
         program_result = GLShader::GenerateVertexShader(setup);
         gl_type = GL_VERTEX_SHADER;
         break;
+    case Maxwell::ShaderProgram::Geometry:
+        program_result = GLShader::GenerateGeometryShader(setup);
+        gl_type = GL_GEOMETRY_SHADER;
+        break;
     case Maxwell::ShaderProgram::Fragment:
         program_result = GLShader::GenerateFragmentShader(setup);
         gl_type = GL_FRAGMENT_SHADER;
@@ -80,11 +84,16 @@ CachedShader::CachedShader(VAddr addr, Maxwell::ShaderProgram program_type)
 
     entries = program_result.second;
 
-    OGLShader shader;
-    shader.Create(program_result.first.c_str(), gl_type);
-    program.Create(true, shader.handle);
-    SetShaderUniformBlockBindings(program.handle);
-    VideoCore::LabelGLObject(GL_PROGRAM, program.handle, addr);
+    if (program_type != Maxwell::ShaderProgram::Geometry) {
+        OGLShader shader;
+        shader.Create(program_result.first.c_str(), gl_type);
+        program.Create(true, shader.handle);
+        SetShaderUniformBlockBindings(program.handle);
+        VideoCore::LabelGLObject(GL_PROGRAM, program.handle, addr);
+    } else {
+        // Store shader's code to lazily build it on draw
+        geometry_programs.code = program_result.first;
+    }
 }
 
 GLuint CachedShader::GetProgramResourceIndex(const GLShader::ConstBufferEntry& buffer) {
@@ -110,6 +119,21 @@ GLint CachedShader::GetUniformLocation(const GLShader::SamplerEntry& sampler) {
     return search->second;
 }
 
+GLuint CachedShader::LazyGeometryProgram(OGLProgram& target_program,
+                                         const std::string& glsl_topology,
+                                         const std::string& debug_name) {
+    if (target_program.handle != 0) {
+        return target_program.handle;
+    }
+    const std::string source{geometry_programs.code + "layout (" + glsl_topology + ") in;\n"};
+    OGLShader shader;
+    shader.Create(source.c_str(), GL_GEOMETRY_SHADER);
+    target_program.Create(true, shader.handle);
+    SetShaderUniformBlockBindings(target_program.handle);
+    VideoCore::LabelGLObject(GL_PROGRAM, target_program.handle, addr, debug_name);
+    return target_program.handle;
+};
+
 Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
     const VAddr program_addr{GetShaderAddress(program)};
 
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index 9bafe43a9..7bb287f56 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -7,6 +7,7 @@
 #include <map>
 #include <memory>
 
+#include "common/assert.h"
 #include "common/common_types.h"
 #include "video_core/rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
@@ -38,8 +39,31 @@ public:
     }
 
     /// Gets the GL program handle for the shader
-    GLuint GetProgramHandle() const {
-        return program.handle;
+    GLuint GetProgramHandle(GLenum primitive_mode) {
+        if (program_type != Maxwell::ShaderProgram::Geometry) {
+            return program.handle;
+        }
+        switch (primitive_mode) {
+        case GL_POINTS:
+            return LazyGeometryProgram(geometry_programs.points, "points", "ShaderPoints");
+        case GL_LINES:
+        case GL_LINE_STRIP:
+            return LazyGeometryProgram(geometry_programs.lines, "lines", "ShaderLines");
+        case GL_LINES_ADJACENCY:
+        case GL_LINE_STRIP_ADJACENCY:
+            return LazyGeometryProgram(geometry_programs.lines_adjacency, "lines_adjacency",
+                                       "ShaderLinesAdjacency");
+        case GL_TRIANGLES:
+        case GL_TRIANGLE_STRIP:
+        case GL_TRIANGLE_FAN:
+            return LazyGeometryProgram(geometry_programs.triangles, "triangles", "ShaderTriangles");
+        case GL_TRIANGLES_ADJACENCY:
+        case GL_TRIANGLE_STRIP_ADJACENCY:
+            return LazyGeometryProgram(geometry_programs.triangles_adjacency, "triangles_adjacency",
+                                       "ShaderLines");
+        default:
+            UNREACHABLE_MSG("Unknown primitive mode.");
+        }
     }
 
     /// Gets the GL program resource location for the specified resource, caching as needed
@@ -49,12 +73,30 @@ public:
     GLint GetUniformLocation(const GLShader::SamplerEntry& sampler);
 
 private:
+    /// Generates a geometry shader or returns one that already exists.
+    GLuint LazyGeometryProgram(OGLProgram& target_program, const std::string& glsl_topology,
+                               const std::string& debug_name);
+
     VAddr addr;
     Maxwell::ShaderProgram program_type;
     GLShader::ShaderSetup setup;
     GLShader::ShaderEntries entries;
+
+    // Non-geometry program.
     OGLProgram program;
 
+    // Geometry programs. These are needed because GLSL needs an input topology but it's not
+    // declared by the hardware. Workaround this issue by generating a different shader per input
+    // topology class.
+    struct {
+        std::string code;
+        OGLProgram points;
+        OGLProgram lines;
+        OGLProgram lines_adjacency;
+        OGLProgram triangles;
+        OGLProgram triangles_adjacency;
+    } geometry_programs;
+
     std::map<u32, GLuint> resource_cache;
     std::map<u32, GLint> uniform_cache;
 };
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 85c668ca1..8dfb49507 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -7,6 +7,7 @@
 #include <string>
 #include <string_view>
 
+#include <boost/optional.hpp>
 #include <fmt/format.h>
 
 #include "common/assert.h"
@@ -29,11 +30,32 @@ using Tegra::Shader::SubOp;
 constexpr u32 PROGRAM_END = MAX_PROGRAM_CODE_LENGTH;
 constexpr u32 PROGRAM_HEADER_SIZE = sizeof(Tegra::Shader::Header);
 
+enum : u32 { POSITION_VARYING_LOCATION = 0, GENERIC_VARYING_START_LOCATION = 1 };
+
+constexpr u32 MAX_GEOMETRY_BUFFERS = 6;
+constexpr u32 MAX_ATTRIBUTES = 0x100; // Size in vec4s, this value is untested
+
 class DecompileFail : public std::runtime_error {
 public:
     using std::runtime_error::runtime_error;
 };
 
+/// Translate topology
+static std::string GetTopologyName(Tegra::Shader::OutputTopology topology) {
+    switch (topology) {
+    case Tegra::Shader::OutputTopology::PointList:
+        return "points";
+    case Tegra::Shader::OutputTopology::LineStrip:
+        return "line_strip";
+    case Tegra::Shader::OutputTopology::TriangleStrip:
+        return "triangle_strip";
+    default:
+        LOG_CRITICAL(Render_OpenGL, "Unknown output topology {}", static_cast<u32>(topology));
+        UNREACHABLE();
+        return "points";
+    }
+}
+
 /// Describes the behaviour of code path of a given entry point and a return point.
 enum class ExitMethod {
     Undetermined, ///< Internal value. Only occur when analyzing JMP loop.
@@ -253,8 +275,9 @@ enum class InternalFlag : u64 {
 class GLSLRegisterManager {
 public:
     GLSLRegisterManager(ShaderWriter& shader, ShaderWriter& declarations,
-                        const Maxwell3D::Regs::ShaderStage& stage, const std::string& suffix)
-        : shader{shader}, declarations{declarations}, stage{stage}, suffix{suffix} {
+                        const Maxwell3D::Regs::ShaderStage& stage, const std::string& suffix,
+                        const Tegra::Shader::Header& header)
+        : shader{shader}, declarations{declarations}, stage{stage}, suffix{suffix}, header{header} {
         BuildRegisterList();
         BuildInputList();
     }
@@ -358,11 +381,13 @@ public:
      * @param reg The destination register to use.
      * @param elem The element to use for the operation.
      * @param attribute The input attribute to use as the source value.
+     * @param vertex The register that decides which vertex to read from (used in GS).
      */
     void SetRegisterToInputAttibute(const Register& reg, u64 elem, Attribute::Index attribute,
-                                    const Tegra::Shader::IpaMode& input_mode) {
+                                    const Tegra::Shader::IpaMode& input_mode,
+                                    boost::optional<Register> vertex = {}) {
         const std::string dest = GetRegisterAsFloat(reg);
-        const std::string src = GetInputAttribute(attribute, input_mode) + GetSwizzle(elem);
+        const std::string src = GetInputAttribute(attribute, input_mode, vertex) + GetSwizzle(elem);
         shader.AddLine(dest + " = " + src + ';');
     }
 
@@ -391,16 +416,29 @@ public:
      * are stored as floats, so this may require conversion.
      * @param attribute The destination output attribute.
      * @param elem The element to use for the operation.
-     * @param reg The register to use as the source value.
+     * @param val_reg The register to use as the source value.
+     * @param buf_reg The register that tells which buffer to write to (used in geometry shaders).
      */
-    void SetOutputAttributeToRegister(Attribute::Index attribute, u64 elem, const Register& reg) {
+    void SetOutputAttributeToRegister(Attribute::Index attribute, u64 elem, const Register& val_reg,
+                                      const Register& buf_reg) {
         const std::string dest = GetOutputAttribute(attribute);
-        const std::string src = GetRegisterAsFloat(reg);
+        const std::string src = GetRegisterAsFloat(val_reg);
 
         if (!dest.empty()) {
             // Can happen with unknown/unimplemented output attributes, in which case we ignore the
             // instruction for now.
-            shader.AddLine(dest + GetSwizzle(elem) + " = " + src + ';');
+            if (stage == Maxwell3D::Regs::ShaderStage::Geometry) {
+                // TODO(Rodrigo): nouveau sets some attributes after setting emitting a geometry
+                // shader. These instructions use a dirty register as buffer index. To avoid some
+                // drivers from complaining for the out of boundary writes, guard them.
+                const std::string buf_index{"min(" + GetRegisterAsInteger(buf_reg) + ", " +
+                                            std::to_string(MAX_GEOMETRY_BUFFERS - 1) + ')'};
+                shader.AddLine("amem[" + buf_index + "][" +
+                               std::to_string(static_cast<u32>(attribute)) + ']' +
+                               GetSwizzle(elem) + " = " + src + ';');
+            } else {
+                shader.AddLine(dest + GetSwizzle(elem) + " = " + src + ';');
+            }
         }
     }
 
@@ -441,41 +479,123 @@ public:
         }
     }
 
-    /// Add declarations for registers
+    /// Add declarations.
     void GenerateDeclarations(const std::string& suffix) {
+        GenerateRegisters(suffix);
+        GenerateInternalFlags();
+        GenerateInputAttrs();
+        GenerateOutputAttrs();
+        GenerateConstBuffers();
+        GenerateSamplers();
+        GenerateGeometry();
+    }
+
+    /// Returns a list of constant buffer declarations.
+    std::vector<ConstBufferEntry> GetConstBuffersDeclarations() const {
+        std::vector<ConstBufferEntry> result;
+        std::copy_if(declr_const_buffers.begin(), declr_const_buffers.end(),
+                     std::back_inserter(result), [](const auto& entry) { return entry.IsUsed(); });
+        return result;
+    }
+
+    /// Returns a list of samplers used in the shader.
+    const std::vector<SamplerEntry>& GetSamplers() const {
+        return used_samplers;
+    }
+
+    /// Returns the GLSL sampler used for the input shader sampler, and creates a new one if
+    /// necessary.
+    std::string AccessSampler(const Sampler& sampler, Tegra::Shader::TextureType type,
+                              bool is_array, bool is_shadow) {
+        const auto offset = static_cast<std::size_t>(sampler.index.Value());
+
+        // If this sampler has already been used, return the existing mapping.
+        const auto itr =
+            std::find_if(used_samplers.begin(), used_samplers.end(),
+                         [&](const SamplerEntry& entry) { return entry.GetOffset() == offset; });
+
+        if (itr != used_samplers.end()) {
+            ASSERT(itr->GetType() == type && itr->IsArray() == is_array &&
+                   itr->IsShadow() == is_shadow);
+            return itr->GetName();
+        }
+
+        // Otherwise create a new mapping for this sampler
+        const std::size_t next_index = used_samplers.size();
+        const SamplerEntry entry{stage, offset, next_index, type, is_array, is_shadow};
+        used_samplers.emplace_back(entry);
+        return entry.GetName();
+    }
+
+private:
+    /// Generates declarations for registers.
+    void GenerateRegisters(const std::string& suffix) {
         for (const auto& reg : regs) {
             declarations.AddLine(GLSLRegister::GetTypeString() + ' ' + reg.GetPrefixString() +
                                  std::to_string(reg.GetIndex()) + '_' + suffix + " = 0;");
         }
         declarations.AddNewLine();
+    }
 
+    /// Generates declarations for internal flags.
+    void GenerateInternalFlags() {
         for (u32 ii = 0; ii < static_cast<u64>(InternalFlag::Amount); ii++) {
             const InternalFlag code = static_cast<InternalFlag>(ii);
             declarations.AddLine("bool " + GetInternalFlag(code) + " = false;");
         }
         declarations.AddNewLine();
+    }
+
+    /// Generates declarations for input attributes.
+    void GenerateInputAttrs() {
+        if (stage != Maxwell3D::Regs::ShaderStage::Vertex) {
+            const std::string attr =
+                stage == Maxwell3D::Regs::ShaderStage::Geometry ? "gs_position[]" : "position";
+            declarations.AddLine("layout (location = " + std::to_string(POSITION_VARYING_LOCATION) +
+                                 ") in vec4 " + attr + ';');
+        }
 
         for (const auto element : declr_input_attribute) {
             // TODO(bunnei): Use proper number of elements for these
             u32 idx =
                 static_cast<u32>(element.first) - static_cast<u32>(Attribute::Index::Attribute_0);
-            declarations.AddLine("layout(location = " + std::to_string(idx) + ")" +
-                                 GetInputFlags(element.first) + "in vec4 " +
-                                 GetInputAttribute(element.first, element.second) + ';');
+            if (stage != Maxwell3D::Regs::ShaderStage::Vertex) {
+                // If inputs are varyings, add an offset
+                idx += GENERIC_VARYING_START_LOCATION;
+            }
+
+            std::string attr{GetInputAttribute(element.first, element.second)};
+            if (stage == Maxwell3D::Regs::ShaderStage::Geometry) {
+                attr = "gs_" + attr + "[]";
+            }
+            declarations.AddLine("layout (location = " + std::to_string(idx) + ") " +
+                                 GetInputFlags(element.first) + "in vec4 " + attr + ';');
         }
+
         declarations.AddNewLine();
+    }
 
+    /// Generates declarations for output attributes.
+    void GenerateOutputAttrs() {
+        if (stage != Maxwell3D::Regs::ShaderStage::Fragment) {
+            declarations.AddLine("layout (location = " + std::to_string(POSITION_VARYING_LOCATION) +
+                                 ") out vec4 position;");
+        }
         for (const auto& index : declr_output_attribute) {
             // TODO(bunnei): Use proper number of elements for these
-            declarations.AddLine("layout(location = " +
-                                 std::to_string(static_cast<u32>(index) -
-                                                static_cast<u32>(Attribute::Index::Attribute_0)) +
-                                 ") out vec4 " + GetOutputAttribute(index) + ';');
+            const u32 idx = static_cast<u32>(index) -
+                            static_cast<u32>(Attribute::Index::Attribute_0) +
+                            GENERIC_VARYING_START_LOCATION;
+            declarations.AddLine("layout (location = " + std::to_string(idx) + ") out vec4 " +
+                                 GetOutputAttribute(index) + ';');
         }
         declarations.AddNewLine();
+    }
 
+    /// Generates declarations for constant buffers.
+    void GenerateConstBuffers() {
         for (const auto& entry : GetConstBuffersDeclarations()) {
-            declarations.AddLine("layout(std140) uniform " + entry.GetName());
+            declarations.AddLine("layout (std140) uniform " + entry.GetName());
             declarations.AddLine('{');
             declarations.AddLine("    vec4 c" + std::to_string(entry.GetIndex()) +
                                  "[MAX_CONSTBUFFER_ELEMENTS];");
@@ -483,7 +603,10 @@ public:
             declarations.AddNewLine();
         }
         declarations.AddNewLine();
+    }
 
+    /// Generates declarations for samplers.
+    void GenerateSamplers() {
         const auto& samplers = GetSamplers();
         for (const auto& sampler : samplers) {
             declarations.AddLine("uniform " + sampler.GetTypeString() + ' ' + sampler.GetName() +
@@ -492,44 +615,42 @@ public:
         declarations.AddNewLine();
     }
 
-    /// Returns a list of constant buffer declarations
-    std::vector<ConstBufferEntry> GetConstBuffersDeclarations() const {
-        std::vector<ConstBufferEntry> result;
-        std::copy_if(declr_const_buffers.begin(), declr_const_buffers.end(),
-                     std::back_inserter(result), [](const auto& entry) { return entry.IsUsed(); });
-        return result;
-    }
-
-    /// Returns a list of samplers used in the shader
-    const std::vector<SamplerEntry>& GetSamplers() const {
-        return used_samplers;
-    }
-
-    /// Returns the GLSL sampler used for the input shader sampler, and creates a new one if
-    /// necessary.
-    std::string AccessSampler(const Sampler& sampler, Tegra::Shader::TextureType type,
-                              bool is_array, bool is_shadow) {
-        const std::size_t offset = static_cast<std::size_t>(sampler.index.Value());
+    /// Generates declarations used for geometry shaders.
+    void GenerateGeometry() {
+        if (stage != Maxwell3D::Regs::ShaderStage::Geometry)
+            return;
 
-        // If this sampler has already been used, return the existing mapping.
-        const auto itr =
-            std::find_if(used_samplers.begin(), used_samplers.end(),
-                         [&](const SamplerEntry& entry) { return entry.GetOffset() == offset; });
+        declarations.AddLine(
+            "layout (" + GetTopologyName(header.common3.output_topology) +
+            ", max_vertices = " + std::to_string(header.common4.max_output_vertices) + ") out;");
+        declarations.AddNewLine();
 
-        if (itr != used_samplers.end()) {
-            ASSERT(itr->GetType() == type && itr->IsArray() == is_array &&
-                   itr->IsShadow() == is_shadow);
-            return itr->GetName();
-        }
+        declarations.AddLine("vec4 amem[" + std::to_string(MAX_GEOMETRY_BUFFERS) + "][" +
+                             std::to_string(MAX_ATTRIBUTES) + "];");
+        declarations.AddNewLine();
 
-        // Otherwise create a new mapping for this sampler
-        const std::size_t next_index = used_samplers.size();
-        const SamplerEntry entry{stage, offset, next_index, type, is_array, is_shadow};
-        used_samplers.emplace_back(entry);
-        return entry.GetName();
+        constexpr char buffer[] = "amem[output_buffer]";
+        declarations.AddLine("void emit_vertex(uint output_buffer) {");
+        ++declarations.scope;
+        for (const auto element : declr_output_attribute) {
+            declarations.AddLine(GetOutputAttribute(element) + " = " + buffer + '[' +
+                                 std::to_string(static_cast<u32>(element)) + "];");
+        }
+
+        declarations.AddLine("position = " + std::string(buffer) + '[' +
+                             std::to_string(static_cast<u32>(Attribute::Index::Position)) + "];");
+
+        // If a geometry shader is attached, it will always flip (it's the last stage before
+        // fragment). For more info about flipping, refer to gl_shader_gen.cpp.
+        declarations.AddLine("position.xy *= viewport_flip.xy;");
+        declarations.AddLine("gl_Position = position;");
+        declarations.AddLine("position.w = 1.0;");
+        declarations.AddLine("EmitVertex();");
+        --declarations.scope;
+        declarations.AddLine('}');
+        declarations.AddNewLine();
     }
 
-private:
     /// Generates code representing a temporary (GPR) register.
     std::string GetRegister(const Register& reg, unsigned elem) {
         if (reg == Register::ZeroIndex) {
@@ -586,11 +707,19 @@ private:
 
     /// Generates code representing an input attribute register.
     std::string GetInputAttribute(Attribute::Index attribute,
-                                  const Tegra::Shader::IpaMode& input_mode) {
+                                  const Tegra::Shader::IpaMode& input_mode,
+                                  boost::optional<Register> vertex = {}) {
+        auto GeometryPass = [&](const std::string& name) {
+            if (stage == Maxwell3D::Regs::ShaderStage::Geometry && vertex) {
+                return "gs_" + name + '[' + GetRegisterAsInteger(vertex.value(), 0, false) + ']';
+            }
+            return name;
+        };
+
         switch (attribute) {
         case Attribute::Index::Position:
             if (stage != Maxwell3D::Regs::ShaderStage::Fragment) {
-                return "position";
+                return GeometryPass("position");
             } else {
                 return "vec4(gl_FragCoord.x, gl_FragCoord.y, gl_FragCoord.z, 1.0)";
             }
@@ -619,7 +748,7 @@ private:
                         UNREACHABLE();
                     }
                 }
-                return "input_attribute_" + std::to_string(index);
+                return GeometryPass("input_attribute_" + std::to_string(index));
             }
 
             LOG_CRITICAL(HW_GPU, "Unhandled input attribute: {}", static_cast<u32>(attribute));
@@ -672,7 +801,7 @@ private:
         return out;
     }
 
-    /// Generates code representing an output attribute register.
+    /// Generates code representing the declaration name of an output attribute register.
     std::string GetOutputAttribute(Attribute::Index attribute) {
         switch (attribute) {
         case Attribute::Index::Position:
@@ -708,6 +837,7 @@ private:
     std::vector<SamplerEntry> used_samplers;
     const Maxwell3D::Regs::ShaderStage& stage;
     const std::string& suffix;
+    const Tegra::Shader::Header& header;
 };
 
 class GLSLGenerator {
@@ -1103,8 +1233,8 @@ private:
             return offset + 1;
         }
 
-        shader.AddLine("// " + std::to_string(offset) + ": " + opcode->GetName() + " (" +
-                       std::to_string(instr.value) + ')');
+        shader.AddLine(
+            fmt::format("// {}: {} (0x{:016x})", offset, opcode->GetName(), instr.value));
 
         using Tegra::Shader::Pred;
         ASSERT_MSG(instr.pred.full_pred != Pred::NeverExecute,
@@ -1826,7 +1956,7 @@ private:
                 const auto LoadNextElement = [&](u32 reg_offset) {
                     regs.SetRegisterToInputAttibute(instr.gpr0.Value() + reg_offset, next_element,
                                                     static_cast<Attribute::Index>(next_index),
-                                                    input_mode);
+                                                    input_mode, instr.gpr39.Value());
 
                     // Load the next attribute element into the following register. If the element
                     // to load goes beyond the vec4 size, load the first element of the next
@@ -1890,8 +2020,8 @@ private:
 
                 const auto StoreNextElement = [&](u32 reg_offset) {
                     regs.SetOutputAttributeToRegister(static_cast<Attribute::Index>(next_index),
-                                                      next_element,
-                                                      instr.gpr0.Value() + reg_offset);
+                                                      next_element, instr.gpr0.Value() + reg_offset,
+                                                      instr.gpr39.Value());
 
                     // Load the next attribute element into the following register. If the element
                     // to load goes beyond the vec4 size, load the first element of the next
@@ -2734,6 +2864,52 @@ private:
 
                 break;
             }
+            case OpCode::Id::OUT_R: {
+                ASSERT(instr.gpr20.Value() == Register::ZeroIndex);
+                ASSERT_MSG(stage == Maxwell3D::Regs::ShaderStage::Geometry,
+                           "OUT is expected to be used in a geometry shader.");
+
+                if (instr.out.emit) {
+                    // gpr0 is used to store the next address. Hardware returns a pointer but
+                    // we just return the next index with a cyclic cap.
+                    const std::string current{regs.GetRegisterAsInteger(instr.gpr8, 0, false)};
+                    const std::string next = "((" + current + " + 1" + ") % " +
+                                             std::to_string(MAX_GEOMETRY_BUFFERS) + ')';
+                    shader.AddLine("emit_vertex(" + current + ");");
+                    regs.SetRegisterToInteger(instr.gpr0, false, 0, next, 1, 1);
+                }
+                if (instr.out.cut) {
+                    shader.AddLine("EndPrimitive();");
+                }
+
+                break;
+            }
+            case OpCode::Id::MOV_SYS: {
+                switch (instr.sys20) {
+                case Tegra::Shader::SystemVariable::InvocationInfo: {
+                    LOG_WARNING(HW_GPU, "MOV_SYS instruction with InvocationInfo is incomplete");
+                    regs.SetRegisterToInteger(instr.gpr0, false, 0, "0u", 1, 1);
+                    break;
+                }
+                default: {
+                    LOG_CRITICAL(HW_GPU, "Unhandled system move: {}",
+                                 static_cast<u32>(instr.sys20.Value()));
+                    UNREACHABLE();
+                }
+                }
+                break;
+            }
+            case OpCode::Id::ISBERD: {
+                ASSERT(instr.isberd.o == 0);
+                ASSERT(instr.isberd.skew == 0);
+                ASSERT(instr.isberd.shift == Tegra::Shader::IsberdShift::None);
+                ASSERT(instr.isberd.mode == Tegra::Shader::IsberdMode::None);
+                ASSERT_MSG(stage == Maxwell3D::Regs::ShaderStage::Geometry,
+                           "ISBERD is expected to be used in a geometry shader.");
+                LOG_WARNING(HW_GPU, "ISBERD instruction is incomplete");
+                regs.SetRegisterToFloat(instr.gpr0, 0, regs.GetRegisterAsFloat(instr.gpr8), 1, 1);
+                break;
+            }
             case OpCode::Id::BRA: {
                 ASSERT_MSG(instr.bra.constant_buffer == 0,
                            "BRA with constant buffers are not implemented");
@@ -2777,6 +2953,88 @@ private:
                 LOG_WARNING(HW_GPU, "DEPBAR instruction is stubbed");
                 break;
             }
+            case OpCode::Id::VMAD: {
+                const bool signed_a = instr.vmad.signed_a == 1;
+                const bool signed_b = instr.vmad.signed_b == 1;
+                const bool result_signed = signed_a || signed_b;
+                boost::optional<std::string> forced_result;
+
+                auto Unpack = [&](const std::string& op, bool is_chunk, bool is_signed,
+                                  Tegra::Shader::VmadType type, u64 byte_height) {
+                    const std::string value = [&]() {
+                        if (!is_chunk) {
+                            const auto offset = static_cast<u32>(byte_height * 8);
+                            return "((" + op + " >> " + std::to_string(offset) + ") & 0xff)";
+                        }
+                        const std::string zero = "0";
+
+                        switch (type) {
+                        case Tegra::Shader::VmadType::Size16_Low:
+                            return '(' + op + " & 0xffff)";
+                        case Tegra::Shader::VmadType::Size16_High:
+                            return '(' + op + " >> 16)";
+                        case Tegra::Shader::VmadType::Size32:
+                            // TODO(Rodrigo): From my hardware tests it becomes a bit "mad" when
+                            // this type is used (1 * 1 + 0 == 0x5b800000). Until a better
+                            // explanation is found: assert.
+                            UNREACHABLE_MSG("Unimplemented");
+                            return zero;
+                        case Tegra::Shader::VmadType::Invalid:
+                            // Note(Rodrigo): This flag is invalid according to nvdisasm. From my
+                            // testing (even though it's invalid) this makes the whole instruction
+                            // assign zero to target register.
+                            forced_result = boost::make_optional(zero);
+                            return zero;
+                        default:
+                            UNREACHABLE();
+                            return zero;
+                        }
+                    }();
+
+                    if (is_signed) {
+                        return "int(" + value + ')';
+                    }
+                    return value;
+                };
+
+                const std::string op_a = Unpack(regs.GetRegisterAsInteger(instr.gpr8, 0, false),
+                                                instr.vmad.is_byte_chunk_a != 0, signed_a,
+                                                instr.vmad.type_a, instr.vmad.byte_height_a);
+
+                std::string op_b;
+                if (instr.vmad.use_register_b) {
+                    op_b = Unpack(regs.GetRegisterAsInteger(instr.gpr20, 0, false),
+                                  instr.vmad.is_byte_chunk_b != 0, signed_b, instr.vmad.type_b,
+                                  instr.vmad.byte_height_b);
+                } else {
+                    op_b = '(' +
+                           std::to_string(signed_b ? static_cast<s16>(instr.alu.GetImm20_16())
+                                                   : instr.alu.GetImm20_16()) +
+                           ')';
+                }
+
+                const std::string op_c = regs.GetRegisterAsInteger(instr.gpr39, 0, result_signed);
+
+                std::string result;
+                if (forced_result) {
+                    result = *forced_result;
+                } else {
+                    result = '(' + op_a + " * " + op_b + " + " + op_c + ')';
+
+                    switch (instr.vmad.shr) {
+                    case Tegra::Shader::VmadShr::Shr7:
+                        result = '(' + result + " >> 7)";
+                        break;
+                    case Tegra::Shader::VmadShr::Shr15:
+                        result = '(' + result + " >> 15)";
+                        break;
+                    }
+                }
+                regs.SetRegisterToInteger(instr.gpr0, result_signed, 1, result, 1, 1,
+                                          instr.vmad.saturate == 1, 0, Register::Size::Word,
+                                          instr.vmad.cc);
+                break;
+            }
             default: {
                 LOG_CRITICAL(HW_GPU, "Unhandled instruction: {}", opcode->GetName());
                 UNREACHABLE();
@@ -2907,7 +3165,7 @@ private:
 
     ShaderWriter shader;
     ShaderWriter declarations;
-    GLSLRegisterManager regs{shader, declarations, stage, suffix};
+    GLSLRegisterManager regs{shader, declarations, stage, suffix, header};
 
     // Declarations
     std::set<std::string> declr_predicates;
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp
index b0466c18f..1e5eb32df 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_gen.cpp
@@ -17,7 +17,18 @@ ProgramResult GenerateVertexShader(const ShaderSetup& setup) {
     std::string out = "#version 430 core\n";
     out += "#extension GL_ARB_separate_shader_objects : enable\n\n";
     out += Decompiler::GetCommonDeclarations();
-    out += "bool exec_vertex();\n";
+
+    out += R"(
+out gl_PerVertex {
+    vec4 gl_Position;
+};
+
+layout(std140) uniform vs_config {
+    vec4 viewport_flip;
+    uvec4 instance_id;
+    uvec4 flip_stage;
+};
+)";
 
     if (setup.IsDualProgram()) {
         out += "bool exec_vertex_b();\n";
@@ -28,18 +39,17 @@ ProgramResult GenerateVertexShader(const ShaderSetup& setup) {
                                      Maxwell3D::Regs::ShaderStage::Vertex, "vertex")
             .get_value_or({});
 
-    out += R"(
-
-out gl_PerVertex {
-    vec4 gl_Position;
-};
+    out += program.first;
 
-out vec4 position;
+    if (setup.IsDualProgram()) {
+        ProgramResult program_b =
+            Decompiler::DecompileProgram(setup.program.code_b, PROGRAM_OFFSET,
+                                         Maxwell3D::Regs::ShaderStage::Vertex, "vertex_b")
+                .get_value_or({});
+        out += program_b.first;
+    }
 
-layout (std140) uniform vs_config {
-    vec4 viewport_flip;
-    uvec4 instance_id;
-};
+    out += R"(
 
 void main() {
     position = vec4(0.0, 0.0, 0.0, 0.0);
@@ -52,27 +62,52 @@ void main() {
 
     out += R"(
 
-    // Viewport can be flipped, which is unsupported by glViewport
-    position.xy *= viewport_flip.xy;
+    // Check if the flip stage is VertexB
+    if (flip_stage[0] == 1) {
+        // Viewport can be flipped, which is unsupported by glViewport
+        position.xy *= viewport_flip.xy;
+    }
     gl_Position = position;
 
     // TODO(bunnei): This is likely a hack, position.w should be interpolated as 1.0
     // For now, this is here to bring order in lieu of proper emulation
-    position.w = 1.0;
+    if (flip_stage[0] == 1) {
+        position.w = 1.0;
+    }
 }
 
 )";
 
-    out += program.first;
+    return {out, program.second};
+}
 
-    if (setup.IsDualProgram()) {
-        ProgramResult program_b =
-            Decompiler::DecompileProgram(setup.program.code_b, PROGRAM_OFFSET,
-                                         Maxwell3D::Regs::ShaderStage::Vertex, "vertex_b")
-                .get_value_or({});
-        out += program_b.first;
-    }
+ProgramResult GenerateGeometryShader(const ShaderSetup& setup) {
+    std::string out = "#version 430 core\n";
+    out += "#extension GL_ARB_separate_shader_objects : enable\n\n";
+    out += Decompiler::GetCommonDeclarations();
+    out += "bool exec_geometry();\n";
+
+    ProgramResult program =
+        Decompiler::DecompileProgram(setup.program.code, PROGRAM_OFFSET,
+                                     Maxwell3D::Regs::ShaderStage::Geometry, "geometry")
+            .get_value_or({});
+    out += R"(
+out gl_PerVertex {
+    vec4 gl_Position;
+};
 
+layout (std140) uniform gs_config {
+    vec4 viewport_flip;
+    uvec4 instance_id;
+    uvec4 flip_stage;
+};
+
+void main() {
+    exec_geometry();
+}
+
+)";
+    out += program.first;
     return {out, program.second};
 }
 
@@ -87,7 +122,6 @@ ProgramResult GenerateFragmentShader(const ShaderSetup& setup) {
                                      Maxwell3D::Regs::ShaderStage::Fragment, "fragment")
             .get_value_or({});
     out += R"(
-in vec4 position;
 layout(location = 0) out vec4 FragColor0;
 layout(location = 1) out vec4 FragColor1;
 layout(location = 2) out vec4 FragColor2;
@@ -100,6 +134,7 @@ layout(location = 7) out vec4 FragColor7;
 layout (std140) uniform fs_config {
     vec4 viewport_flip;
     uvec4 instance_id;
+    uvec4 flip_stage;
 };
 
 void main() {
@@ -110,5 +145,4 @@ void main() {
     out += program.first;
     return {out, program.second};
 }
-
-} // namespace OpenGL::GLShader
+} // namespace OpenGL::GLShader
\ No newline at end of file
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.h b/src/video_core/renderer_opengl/gl_shader_gen.h
index e56f39e78..79596087a 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.h
+++ b/src/video_core/renderer_opengl/gl_shader_gen.h
@@ -196,6 +196,12 @@ private:
 ProgramResult GenerateVertexShader(const ShaderSetup& setup);
 
 /**
+ * Generates the GLSL geometry shader program source code for the given GS program
+ * @returns String of the shader source code
+ */
+ProgramResult GenerateGeometryShader(const ShaderSetup& setup);
+
+/**
  * Generates the GLSL fragment shader program source code for the given FS program
  * @returns String of the shader source code
  */
diff --git a/src/video_core/renderer_opengl/gl_shader_manager.cpp b/src/video_core/renderer_opengl/gl_shader_manager.cpp
index 022d32a86..010857ec6 100644
--- a/src/video_core/renderer_opengl/gl_shader_manager.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_manager.cpp
@@ -18,6 +18,14 @@ void MaxwellUniformData::SetFromRegs(const Maxwell3D::State::ShaderStageInfo& sh
 
     // We only assign the instance to the first component of the vector, the rest is just padding.
     instance_id[0] = state.current_instance;
+
+    // Assign in which stage the position has to be flipped
+    // (the last stage before the fragment shader).
+    if (gpu.regs.shader_config[static_cast<u32>(Maxwell3D::Regs::ShaderProgram::Geometry)].enable) {
+        flip_stage[0] = static_cast<u32>(Maxwell3D::Regs::ShaderProgram::Geometry);
+    } else {
+        flip_stage[0] = static_cast<u32>(Maxwell3D::Regs::ShaderProgram::VertexB);
+    }
 }
 
 } // namespace OpenGL::GLShader
diff --git a/src/video_core/renderer_opengl/gl_shader_manager.h b/src/video_core/renderer_opengl/gl_shader_manager.h
index 3de15ba9b..b3a191cf2 100644
--- a/src/video_core/renderer_opengl/gl_shader_manager.h
+++ b/src/video_core/renderer_opengl/gl_shader_manager.h
@@ -21,8 +21,9 @@ struct MaxwellUniformData {
     void SetFromRegs(const Maxwell3D::State::ShaderStageInfo& shader_stage);
     alignas(16) GLvec4 viewport_flip;
     alignas(16) GLuvec4 instance_id;
+    alignas(16) GLuvec4 flip_stage;
 };
-static_assert(sizeof(MaxwellUniformData) == 32, "MaxwellUniformData structure size is incorrect");
+static_assert(sizeof(MaxwellUniformData) == 48, "MaxwellUniformData structure size is incorrect");
 static_assert(sizeof(MaxwellUniformData) < 16384,
               "MaxwellUniformData structure must be less than 16kb as per the OpenGL spec");
 
@@ -36,6 +37,10 @@ public:
         vs = program;
     }
 
+    void UseProgrammableGeometryShader(GLuint program) {
+        gs = program;
+    }
+
     void UseProgrammableFragmentShader(GLuint program) {
         fs = program;
     }
diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h
index 8f31d825a..58d17abcb 100644
--- a/src/video_core/textures/texture.h
+++ b/src/video_core/textures/texture.h
@@ -161,7 +161,9 @@ struct TICEntry {
         BitField<21, 3, TICHeaderVersion> header_version;
     };
     union {
+        BitField<0, 3, u32> block_width;
         BitField<3, 3, u32> block_height;
+        BitField<6, 3, u32> block_depth;
 
         // High 16 bits of the pitch value
         BitField<0, 16, u32> pitch_high;
@@ -202,13 +204,24 @@ struct TICEntry {
         return depth_minus_1 + 1;
     }
 
+    u32 BlockWidth() const {
+        ASSERT(IsTiled());
+        // The block height is stored in log2 format.
+        return 1 << block_width;
+    }
+
     u32 BlockHeight() const {
-        ASSERT(header_version == TICHeaderVersion::BlockLinear ||
-               header_version == TICHeaderVersion::BlockLinearColorKey);
+        ASSERT(IsTiled());
         // The block height is stored in log2 format.
         return 1 << block_height;
     }
 
+    u32 BlockDepth() const {
+        ASSERT(IsTiled());
+        // The block height is stored in log2 format.
+        return 1 << block_depth;
+    }
+
     bool IsTiled() const {
         return header_version == TICHeaderVersion::BlockLinear ||
                header_version == TICHeaderVersion::BlockLinearColorKey;
diff --git a/src/video_core/utils.h b/src/video_core/utils.h
index 681919ae3..237cc1307 100644
--- a/src/video_core/utils.h
+++ b/src/video_core/utils.h
@@ -169,16 +169,20 @@ static void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr,
     const std::string nice_addr = fmt::format("0x{:016x}", addr);
     std::string object_label;
 
-    switch (identifier) {
-    case GL_TEXTURE:
-        object_label = extra_info + "@" + nice_addr;
-        break;
-    case GL_PROGRAM:
-        object_label = "ShaderProgram@" + nice_addr;
-        break;
-    default:
-        object_label = fmt::format("Object(0x{:x})@{}", identifier, nice_addr);
-        break;
+    if (extra_info.empty()) {
+        switch (identifier) {
+        case GL_TEXTURE:
+            object_label = "Texture@" + nice_addr;
+            break;
+        case GL_PROGRAM:
+            object_label = "Shader@" + nice_addr;
+            break;
+        default:
+            object_label = fmt::format("Object(0x{:x})@{}", identifier, nice_addr);
+            break;
+        }
+    } else {
+        object_label = extra_info + '@' + nice_addr;
     }
     glObjectLabel(identifier, handle, -1, static_cast<const GLchar*>(object_label.c_str()));
 }
diff --git a/src/yuzu/game_list_worker.cpp b/src/yuzu/game_list_worker.cpp
index d2b3de683..8f99a1c78 100644
--- a/src/yuzu/game_list_worker.cpp
+++ b/src/yuzu/game_list_worker.cpp
@@ -27,9 +27,8 @@
 #include "yuzu/ui_settings.h"
 
 namespace {
-void GetMetadataFromControlNCA(const FileSys::PatchManager& patch_manager,
-                               const std::shared_ptr<FileSys::NCA>& nca, std::vector<u8>& icon,
-                               std::string& name) {
+void GetMetadataFromControlNCA(const FileSys::PatchManager& patch_manager, const FileSys::NCA& nca,
+                               std::vector<u8>& icon, std::string& name) {
     auto [nacp, icon_file] = patch_manager.ParseControlNCA(nca);
     if (icon_file != nullptr)
         icon = icon_file->ReadAllBytes();
@@ -110,7 +109,7 @@ void GameListWorker::AddInstalledTitlesToGameList() {
         const FileSys::PatchManager patch{program_id};
         const auto& control = cache->GetEntry(game.title_id, FileSys::ContentRecordType::Control);
         if (control != nullptr)
-            GetMetadataFromControlNCA(patch, control, icon, name);
+            GetMetadataFromControlNCA(patch, *control, icon, name);
 
         auto it = FindMatchingCompatibilityEntry(compatibility_list, program_id);
 
@@ -197,8 +196,8 @@ void GameListWorker::AddFstEntriesToGameList(const std::string& dir_path, unsign
                 res2 == Loader::ResultStatus::Success) {
                 // Use from metadata pool.
                 if (nca_control_map.find(program_id) != nca_control_map.end()) {
-                    const auto nca = nca_control_map[program_id];
-                    GetMetadataFromControlNCA(patch, nca, icon, name);
+                    const auto& nca = nca_control_map[program_id];
+                    GetMetadataFromControlNCA(patch, *nca, icon, name);
                 }
             }