"Merge Tagged PR 9042"

shader: Implement miss attribute layer

src/audio_core/renderer/performance/performance_manager.cpp

@@ -26,6 +26,7 @@ void PerformanceManager::CreateImpl(const size_t version) {
         impl = std::make_unique<
             PerformanceManagerImpl<PerformanceVersion::Version1, PerformanceFrameHeaderVersion1,
                                    PerformanceEntryVersion1, PerformanceDetailVersion1>>();
+        break;
     }
 }
 

src/audio_core/sink/sink_stream.cpp

@@ -265,20 +265,8 @@ void SinkStream::ProcessAudioOutAndRender(std::span<s16> output_buffer, std::siz
     }
 }
 
-void SinkStream::Stall() {
-    if (stalled) {
-        return;
-    }
-    stalled = true;
-    system.StallProcesses();
-}
+void SinkStream::Stall() {}
 
-void SinkStream::Unstall() {
-    if (!stalled) {
-        return;
-    }
-    system.UnstallProcesses();
-    stalled = false;
-}
+void SinkStream::Unstall() {}
 
 } // namespace AudioCore::Sink

src/common/CMakeLists.txt

@@ -34,6 +34,8 @@ add_library(common STATIC
     bit_util.h
     cityhash.cpp
     cityhash.h
+    cache_management.cpp
+    cache_management.h
     common_funcs.h
     common_types.h
     concepts.h

src/common/atomic_helpers.h

@@ -156,6 +156,7 @@ AE_FORCEINLINE void compiler_fence(memory_order order) AE_NO_TSAN {
         break;
     default:
         assert(false);
+        break;
     }
 }
 

src/common/cache_management.cpp

@@ -0,0 +1,60 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <cstring>
+
+#include "alignment.h"
+#include "cache_management.h"
+#include "common_types.h"
+
+namespace Common {
+
+#if defined(ARCHITECTURE_x86_64)
+
+// Most cache operations are no-ops on x86
+
+void DataCacheLineCleanByVAToPoU(void* start, size_t size) {}
+void DataCacheLineCleanAndInvalidateByVAToPoC(void* start, size_t size) {}
+void DataCacheLineCleanByVAToPoC(void* start, size_t size) {}
+void DataCacheZeroByVA(void* start, size_t size) {
+    std::memset(start, 0, size);
+}
+
+#elif defined(ARCHITECTURE_arm64)
+
+// BS/DminLine is log2(cache size in words), we want size in bytes
+#define EXTRACT_DMINLINE(ctr_el0) (1 << ((((ctr_el0) >> 16) & 0xf) + 2))
+#define EXTRACT_BS(dczid_el0) (1 << (((dczid_el0)&0xf) + 2))
+
+#define DEFINE_DC_OP(op_name, function_name)                                                       \
+    void function_name(void* start, size_t size) {                                                 \
+        size_t ctr_el0;                                                                            \
+        asm volatile("mrs %[ctr_el0], ctr_el0\n\t" : [ctr_el0] "=r"(ctr_el0));                     \
+        size_t cacheline_size = EXTRACT_DMINLINE(ctr_el0);                                         \
+        uintptr_t va_start = reinterpret_cast<uintptr_t>(start);                                   \
+        uintptr_t va_end = va_start + size;                                                        \
+        for (uintptr_t va = va_start; va < va_end; va += cacheline_size) {                         \
+            asm volatile("dc " #op_name ", %[va]\n\t" : : [va] "r"(va) : "memory");                \
+        }                                                                                          \
+    }
+
+#define DEFINE_DC_OP_DCZID(op_name, function_name)                                                 \
+    void function_name(void* start, size_t size) {                                                 \
+        size_t dczid_el0;                                                                          \
+        asm volatile("mrs %[dczid_el0], dczid_el0\n\t" : [dczid_el0] "=r"(dczid_el0));             \
+        size_t cacheline_size = EXTRACT_BS(dczid_el0);                                             \
+        uintptr_t va_start = reinterpret_cast<uintptr_t>(start);                                   \
+        uintptr_t va_end = va_start + size;                                                        \
+        for (uintptr_t va = va_start; va < va_end; va += cacheline_size) {                         \
+            asm volatile("dc " #op_name ", %[va]\n\t" : : [va] "r"(va) : "memory");                \
+        }                                                                                          \
+    }
+
+DEFINE_DC_OP(cvau, DataCacheLineCleanByVAToPoU);
+DEFINE_DC_OP(civac, DataCacheLineCleanAndInvalidateByVAToPoC);
+DEFINE_DC_OP(cvac, DataCacheLineCleanByVAToPoC);
+DEFINE_DC_OP_DCZID(zva, DataCacheZeroByVA);
+
+#endif
+
+} // namespace Common

src/common/cache_management.h

@@ -0,0 +1,27 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "stdlib.h"
+
+namespace Common {
+
+// Data cache instructions enabled at EL0 by SCTLR_EL1.UCI.
+// VA = virtual address
+// PoC = point of coherency
+// PoU = point of unification
+
+// dc cvau
+void DataCacheLineCleanByVAToPoU(void* start, size_t size);
+
+// dc civac
+void DataCacheLineCleanAndInvalidateByVAToPoC(void* start, size_t size);
+
+// dc cvac
+void DataCacheLineCleanByVAToPoC(void* start, size_t size);
+
+// dc zva
+void DataCacheZeroByVA(void* start, size_t size);
+
+} // namespace Common

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -348,7 +348,6 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
         if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
             config.unsafe_optimizations = true;
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
             config.fastmem_address_space_bits = 64;
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
         }

src/core/hle/kernel/init/init_slab_setup.cpp

@@ -243,6 +243,7 @@ void InitializeSlabHeaps(Core::System& system, KMemoryLayout& memory_layout) {
             // If we somehow get an invalid type, abort.
         default:
             ASSERT_MSG(false, "Unknown slab type: {}", slab_types[i]);
+            break;
         }
 
         // If we've hit the end of a gap, free it.

src/core/hle/kernel/k_page_table.cpp

@@ -2301,6 +2301,7 @@ Result KPageTable::SetProcessMemoryPermission(VAddr addr, size_t size,
             break;
         default:
             ASSERT(false);
+            break;
         }
     }
 
@@ -2803,6 +2804,7 @@ Result KPageTable::Operate(VAddr addr, size_t num_pages, const KPageGroup& page_
             break;
         default:
             ASSERT(false);
+            break;
         }
 
         addr += size;
@@ -2838,6 +2840,7 @@ Result KPageTable::Operate(VAddr addr, size_t num_pages, KMemoryPermission perm,
         break;
     default:
         ASSERT(false);
+        break;
     }
     R_SUCCEED();
 }

src/core/hle/kernel/k_process.cpp

@@ -395,6 +395,7 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
 
     default:
         ASSERT(false);
+        break;
     }
 
     // Create TLS region

src/core/hle/kernel/svc.cpp

@@ -2701,14 +2701,24 @@ static Result GetThreadList(Core::System& system, u32* out_num_threads, VAddr ou
     return ResultSuccess;
 }
 
-static Result FlushProcessDataCache32([[maybe_unused]] Core::System& system,
-                                      [[maybe_unused]] Handle handle, [[maybe_unused]] u32 address,
-                                      [[maybe_unused]] u32 size) {
-    // Note(Blinkhawk): For emulation purposes of the data cache this is mostly a no-op,
-    // as all emulation is done in the same cache level in host architecture, thus data cache
-    // does not need flushing.
-    LOG_DEBUG(Kernel_SVC, "called");
-    return ResultSuccess;
+static Result FlushProcessDataCache32(Core::System& system, Handle process_handle, u64 address,
+                                      u64 size) {
+    // Validate address/size.
+    R_UNLESS(size > 0, ResultInvalidSize);
+    R_UNLESS(address == static_cast<uintptr_t>(address), ResultInvalidCurrentMemory);
+    R_UNLESS(size == static_cast<size_t>(size), ResultInvalidCurrentMemory);
+
+    // Get the process from its handle.
+    KScopedAutoObject process =
+        system.Kernel().CurrentProcess()->GetHandleTable().GetObject<KProcess>(process_handle);
+    R_UNLESS(process.IsNotNull(), ResultInvalidHandle);
+
+    // Verify the region is within range.
+    auto& page_table = process->PageTable();
+    R_UNLESS(page_table.Contains(address, size), ResultInvalidCurrentMemory);
+
+    // Perform the operation.
+    R_RETURN(system.Memory().FlushDataCache(*process, address, size));
 }
 
 namespace {

src/core/hle/kernel/svc_wrap.h

@@ -722,4 +722,12 @@ void SvcWrap32(Core::System& system) {
     FuncReturn(system, retval);
 }
 
+// Used by Invalidate/Store/FlushProcessDataCache32
+template <Result func(Core::System&, Handle, u64, u64)>
+void SvcWrap32(Core::System& system) {
+    const u64 address = (Param(system, 3) << 32) | Param(system, 2);
+    const u64 size = (Param(system, 4) << 32) | Param(system, 1);
+    FuncReturn32(system, func(system, Param32(system, 0), address, size).raw);
+}
+
 } // namespace Kernel

src/core/hle/service/am/applets/applet_error.cpp

@@ -144,6 +144,7 @@ void Error::Initialize() {
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented LibAppletError mode={:02X}!", mode);
+        break;
     }
 }
 

src/core/hle/service/am/applets/applet_general_backend.cpp

@@ -129,6 +129,7 @@ void Auth::Execute() {
     }
     default:
         unimplemented_log();
+        break;
     }
 }
 
@@ -192,6 +193,7 @@ void PhotoViewer::Execute() {
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented PhotoViewer applet mode={:02X}!", mode);
+        break;
     }
 }
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp

@@ -300,11 +300,10 @@ Kernel::KEvent* nvhost_ctrl_gpu::QueryEvent(u32 event_id) {
         return error_notifier_event;
     case 2:
         return unknown_event;
-    default: {
+    default:
         LOG_CRITICAL(Service_NVDRV, "Unknown Ctrl GPU Event {}", event_id);
+        return nullptr;
     }
-    }
-    return nullptr;
 }
 
 } // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp

@@ -364,11 +364,10 @@ Kernel::KEvent* nvhost_gpu::QueryEvent(u32 event_id) {
         return sm_exception_breakpoint_pause_report_event;
     case 3:
         return error_notifier_event;
-    default: {
+    default:
         LOG_CRITICAL(Service_NVDRV, "Unknown Ctrl GPU Event {}", event_id);
+        return nullptr;
     }
-    }
-    return nullptr;
 }
 
 } // namespace Service::Nvidia::Devices

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

@@ -23,15 +23,17 @@ void BufferQueueCore::NotifyShutdown() {
 }
 
 void BufferQueueCore::SignalDequeueCondition() {
+    dequeue_possible.store(true);
     dequeue_condition.notify_all();
 }
 
-bool BufferQueueCore::WaitForDequeueCondition() {
+bool BufferQueueCore::WaitForDequeueCondition(std::unique_lock<std::mutex>& lk) {
     if (is_shutting_down) {
         return false;
     }
 
-    dequeue_condition.wait(mutex);
+    dequeue_condition.wait(lk, [&] { return dequeue_possible.load(); });
+    dequeue_possible.store(false);
 
     return true;
 }

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

@@ -38,7 +38,7 @@ public:
 
 private:
     void SignalDequeueCondition();
-    bool WaitForDequeueCondition();
+    bool WaitForDequeueCondition(std::unique_lock<std::mutex>& lk);
 
     s32 GetMinUndequeuedBufferCountLocked(bool async) const;
     s32 GetMinMaxBufferCountLocked(bool async) const;
@@ -60,7 +60,8 @@ private:
     BufferQueueDefs::SlotsType slots{};
     std::vector<BufferItem> queue;
     s32 override_max_buffer_count{};
-    mutable std::condition_variable_any dequeue_condition;
+    std::condition_variable dequeue_condition;
+    std::atomic<bool> dequeue_possible{};
     const bool use_async_buffer{}; // This is always disabled on HOS
     bool dequeue_buffer_cannot_block{};
     PixelFormat default_buffer_format{PixelFormat::Rgba8888};

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

@@ -121,8 +121,8 @@ Status BufferQueueProducer::SetBufferCount(s32 buffer_count) {
     return Status::NoError;
 }
 
-Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found,
-                                                      Status* return_flags) const {
+Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found, Status* return_flags,
+                                                      std::unique_lock<std::mutex>& lk) const {
     bool try_again = true;
 
     while (try_again) {
@@ -214,7 +214,7 @@ Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found,
                 return Status::WouldBlock;
             }
 
-            if (!core->WaitForDequeueCondition()) {
+            if (!core->WaitForDequeueCondition(lk)) {
                 // We are no longer running
                 return Status::NoError;
             }
@@ -237,7 +237,7 @@ Status BufferQueueProducer::DequeueBuffer(s32* out_slot, Fence* out_fence, bool
     Status return_flags = Status::NoError;
     bool attached_by_consumer = false;
     {
-        std::scoped_lock lock{core->mutex};
+        std::unique_lock lock{core->mutex};
         core->WaitWhileAllocatingLocked();
 
         if (format == PixelFormat::NoFormat) {
@@ -248,7 +248,7 @@ Status BufferQueueProducer::DequeueBuffer(s32* out_slot, Fence* out_fence, bool
         usage |= core->consumer_usage_bit;
 
         s32 found{};
-        Status status = WaitForFreeSlotThenRelock(async, &found, &return_flags);
+        Status status = WaitForFreeSlotThenRelock(async, &found, &return_flags, lock);
         if (status != Status::NoError) {
             return status;
         }
@@ -400,13 +400,13 @@ Status BufferQueueProducer::AttachBuffer(s32* out_slot,
         return Status::BadValue;
     }
 
-    std::scoped_lock lock{core->mutex};
+    std::unique_lock lock{core->mutex};
     core->WaitWhileAllocatingLocked();
 
     Status return_flags = Status::NoError;
     s32 found{};
 
-    const auto status = WaitForFreeSlotThenRelock(false, &found, &return_flags);
+    const auto status = WaitForFreeSlotThenRelock(false, &found, &return_flags, lock);
     if (status != Status::NoError) {
         return status;
     }

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

@@ -70,7 +70,8 @@ public:
 private:
     BufferQueueProducer(const BufferQueueProducer&) = delete;
 
-    Status WaitForFreeSlotThenRelock(bool async, s32* found, Status* return_flags) const;
+    Status WaitForFreeSlotThenRelock(bool async, s32* found, Status* return_flags,
+                                     std::unique_lock<std::mutex>& lk) const;
 
     Kernel::KEvent* buffer_wait_event{};
     Service::KernelHelpers::ServiceContext& service_context;

src/core/hle/service/service.cpp

@@ -228,6 +228,7 @@ Result ServiceFrameworkBase::HandleSyncRequest(Kernel::KServerSession& session,
         }
 
         UNIMPLEMENTED_MSG("command_type={}", ctx.GetCommandType());
+        break;
     }
 
     // If emulation was shutdown, we are closing service threads, do not write the response back to

src/core/hle/service/time/time_zone_manager.cpp

@@ -280,6 +280,7 @@ static constexpr int TransitionTime(int year, Rule rule, int offset) {
     }
     default:
         ASSERT(false);
+        break;
     }
     return value + rule.transition_time + offset;
 }

src/core/memory.cpp

@@ -6,6 +6,7 @@
 
 #include "common/assert.h"
 #include "common/atomic_ops.h"
+#include "common/cache_management.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "common/page_table.h"
@@ -329,6 +330,55 @@ struct Memory::Impl {
             });
     }
 
+    template <typename Callback>
+    Result PerformCacheOperation(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size,
+                                 Callback&& cb) {
+        class InvalidMemoryException : public std::exception {};
+
+        try {
+            WalkBlock(
+                process, dest_addr, size,
+                [&](const std::size_t block_size, const VAddr current_vaddr) {
+                    LOG_ERROR(HW_Memory, "Unmapped cache maintenance @ {:#018X}", current_vaddr);
+                    throw InvalidMemoryException();
+                },
+                [&](const std::size_t block_size, u8* const host_ptr) { cb(block_size, host_ptr); },
+                [&](const VAddr current_vaddr, const std::size_t block_size, u8* const host_ptr) {
+                    system.GPU().FlushRegion(current_vaddr, block_size);
+                    cb(block_size, host_ptr);
+                },
+                [](const std::size_t block_size) {});
+        } catch (InvalidMemoryException&) {
+            return Kernel::ResultInvalidCurrentMemory;
+        }
+
+        return ResultSuccess;
+    }
+
+    Result InvalidateDataCache(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size) {
+        auto perform = [&](const std::size_t block_size, u8* const host_ptr) {
+            // Do nothing; this operation (dc ivac) cannot be supported
+            // from EL0
+        };
+        return PerformCacheOperation(process, dest_addr, size, perform);
+    }
+
+    Result StoreDataCache(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size) {
+        auto perform = [&](const std::size_t block_size, u8* const host_ptr) {
+            // dc cvac: Store to point of coherency
+            Common::DataCacheLineCleanByVAToPoC(host_ptr, block_size);
+        };
+        return PerformCacheOperation(process, dest_addr, size, perform);
+    }
+
+    Result FlushDataCache(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size) {
+        auto perform = [&](const std::size_t block_size, u8* const host_ptr) {
+            // dc civac: Store to point of coherency, and invalidate from cache
+            Common::DataCacheLineCleanAndInvalidateByVAToPoC(host_ptr, block_size);
+        };
+        return PerformCacheOperation(process, dest_addr, size, perform);
+    }
+
     void MarkRegionDebug(VAddr vaddr, u64 size, bool debug) {
         if (vaddr == 0) {
             return;
@@ -786,6 +836,21 @@ void Memory::ZeroBlock(const Kernel::KProcess& process, VAddr dest_addr, const s
     impl->ZeroBlock(process, dest_addr, size);
 }
 
+Result Memory::InvalidateDataCache(const Kernel::KProcess& process, VAddr dest_addr,
+                                   const std::size_t size) {
+    return impl->InvalidateDataCache(process, dest_addr, size);
+}
+
+Result Memory::StoreDataCache(const Kernel::KProcess& process, VAddr dest_addr,
+                              const std::size_t size) {
+    return impl->StoreDataCache(process, dest_addr, size);
+}
+
+Result Memory::FlushDataCache(const Kernel::KProcess& process, VAddr dest_addr,
+                              const std::size_t size) {
+    return impl->FlushDataCache(process, dest_addr, size);
+}
+
 void Memory::RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
     impl->RasterizerMarkRegionCached(vaddr, size, cached);
 }

src/core/memory.h

@@ -7,6 +7,7 @@
 #include <memory>
 #include <string>
 #include "common/common_types.h"
+#include "core/hle/result.h"
 
 namespace Common {
 struct PageTable;
@@ -449,6 +450,39 @@ public:
      */
     void ZeroBlock(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size);
 
+    /**
+     * Invalidates a range of bytes within the current process' address space at the specified
+     * virtual address.
+     *
+     * @param process   The process that will have data invalidated within its address space.
+     * @param dest_addr The destination virtual address to invalidate the data from.
+     * @param size      The size of the range to invalidate, in bytes.
+     *
+     */
+    Result InvalidateDataCache(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size);
+
+    /**
+     * Stores a range of bytes within the current process' address space at the specified
+     * virtual address.
+     *
+     * @param process   The process that will have data stored within its address space.
+     * @param dest_addr The destination virtual address to store the data from.
+     * @param size      The size of the range to store, in bytes.
+     *
+     */
+    Result StoreDataCache(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size);
+
+    /**
+     * Flushes a range of bytes within the current process' address space at the specified
+     * virtual address.
+     *
+     * @param process   The process that will have data flushed within its address space.
+     * @param dest_addr The destination virtual address to flush the data from.
+     * @param size      The size of the range to flush, in bytes.
+     *
+     */
+    Result FlushDataCache(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size);
+
     /**
      * Marks each page within the specified address range as cached or uncached.
      *

src/shader_recompiler/backend/glasm/emit_glasm.cpp

@@ -320,6 +320,7 @@ void SetupOptions(const IR::Program& program, const Profile& profile,
     }
     if (stage == Stage::Fragment) {
         header += "OPTION ARB_draw_buffers;";
+        header += "OPTION ARB_fragment_layer_viewport;";
     }
 }
 

src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp

@@ -104,6 +104,9 @@ void EmitGetAttribute(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, Scal
     case IR::Attribute::PrimitiveId:
         ctx.Add("MOV.F {}.x,primitive.id;", inst);
         break;
+    case IR::Attribute::Layer:
+        ctx.Add("MOV.F {}.x,fragment.layer;", inst);
+        break;
     case IR::Attribute::PositionX:
     case IR::Attribute::PositionY:
     case IR::Attribute::PositionZ:

src/shader_recompiler/backend/glsl/emit_glsl_context_get_set.cpp

@@ -205,6 +205,9 @@ void EmitGetAttribute(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr,
     case IR::Attribute::PrimitiveId:
         ctx.AddF32("{}=itof(gl_PrimitiveID);", inst);
         break;
+    case IR::Attribute::Layer:
+        ctx.AddF32("{}=itof(gl_Layer);", inst);
+        break;
     case IR::Attribute::PositionX:
     case IR::Attribute::PositionY:
     case IR::Attribute::PositionZ:

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -315,6 +315,8 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
     switch (attr) {
     case IR::Attribute::PrimitiveId:
         return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.primitive_id));
+    case IR::Attribute::Layer:
+        return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.layer));
     case IR::Attribute::PositionX:
     case IR::Attribute::PositionY:
     case IR::Attribute::PositionZ:

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -1359,6 +1359,11 @@ void EmitContext::DefineInputs(const IR::Program& program) {
     if (loads[IR::Attribute::PrimitiveId]) {
         primitive_id = DefineInput(*this, U32[1], false, spv::BuiltIn::PrimitiveId);
     }
+    if (loads[IR::Attribute::Layer]) {
+        AddCapability(spv::Capability::Geometry);
+        layer = DefineInput(*this, U32[1], false, spv::BuiltIn::Layer);
+        Decorate(layer, spv::Decoration::Flat);
+    }
     if (loads.AnyComponent(IR::Attribute::PositionX)) {
         const bool is_fragment{stage != Stage::Fragment};
         const spv::BuiltIn built_in{is_fragment ? spv::BuiltIn::Position : spv::BuiltIn::FragCoord};

src/video_core/engines/maxwell_3d.cpp

@@ -232,7 +232,7 @@ void Maxwell3D::ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argume
         use_topology_override = true;
         return;
     case MAXWELL3D_REG_INDEX(clear_surface):
-        return ProcessClearBuffers();
+        return ProcessClearBuffers(1);
     case MAXWELL3D_REG_INDEX(report_semaphore.query):
         return ProcessQueryGet();
     case MAXWELL3D_REG_INDEX(render_enable.mode):
@@ -596,8 +596,8 @@ u32 Maxwell3D::GetRegisterValue(u32 method) const {
     return regs.reg_array[method];
 }
 
-void Maxwell3D::ProcessClearBuffers() {
-    rasterizer->Clear();
+void Maxwell3D::ProcessClearBuffers(u32 layer_count) {
+    rasterizer->Clear(layer_count);
 }
 
 void Maxwell3D::ProcessDraw(u32 instance_count) {

src/video_core/engines/maxwell_3d.h

@@ -3086,6 +3086,9 @@ public:
 
     std::vector<u8> inline_index_draw_indexes;
 
+    /// Handles a write to the CLEAR_BUFFERS register.
+    void ProcessClearBuffers(u32 layer_count);
+
 private:
     void InitializeRegisterDefaults();
 
@@ -3120,9 +3123,6 @@ private:
     /// Handles firmware blob 4
     void ProcessFirmwareCall4();
 
-    /// Handles a write to the CLEAR_BUFFERS register.
-    void ProcessClearBuffers();
-
     /// Handles a write to the QUERY_GET register.
     void ProcessQueryGet();
 

src/video_core/engines/maxwell_dma.cpp

@@ -314,6 +314,7 @@ void MaxwellDMA::ReleaseSemaphore() {
     }
     default:
         ASSERT_MSG(false, "Unknown semaphore type: {}", static_cast<u32>(type.Value()));
+        break;
     }
 }
 

src/video_core/engines/puller.cpp

@@ -50,6 +50,7 @@ void Puller::ProcessBindMethod(const MethodCall& method_call) {
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented engine {:04X}", engine_id);
+        break;
     }
 }
 
@@ -65,6 +66,7 @@ void Puller::ProcessFenceActionMethod() {
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented operation {}", regs.fence_action.op.Value());
+        break;
     }
 }
 
@@ -228,6 +230,7 @@ void Puller::CallEngineMethod(const MethodCall& method_call) {
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented engine");
+        break;
     }
 }
 
@@ -254,6 +257,7 @@ void Puller::CallEngineMultiMethod(u32 method, u32 subchannel, const u32* base_s
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented engine");
+        break;
     }
 }
 

src/video_core/macro/macro_hle.cpp

@@ -126,11 +126,25 @@ void HLE_3F5E74B9C9A50164(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
     }
 }
 
-constexpr std::array<std::pair<u64, HLEFunction>, 4> hle_funcs{{
+// Multi-layer Clear
+void HLE_EAD26C3E2109B06B(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& parameters) {
+    ASSERT(parameters.size() == 1);
+
+    const Engines::Maxwell3D::Regs::ClearSurface clear_params{parameters[0]};
+    const u32 rt_index = clear_params.RT;
+    const u32 num_layers = maxwell3d.regs.rt[rt_index].depth;
+    ASSERT(clear_params.layer == 0);
+
+    maxwell3d.regs.clear_surface.raw = clear_params.raw;
+    maxwell3d.ProcessClearBuffers(num_layers);
+}
+
+constexpr std::array<std::pair<u64, HLEFunction>, 5> hle_funcs{{
     {0x771BB18C62444DA0, &HLE_771BB18C62444DA0},
     {0x0D61FC9FAAC9FCAD, &HLE_0D61FC9FAAC9FCAD},
     {0x0217920100488FF7, &HLE_0217920100488FF7},
     {0x3F5E74B9C9A50164, &HLE_3F5E74B9C9A50164},
+    {0xEAD26C3E2109B06B, &HLE_EAD26C3E2109B06B},
 }};
 
 class HLEMacroImpl final : public CachedMacro {

src/video_core/macro/macro_interpreter.cpp

@@ -201,6 +201,7 @@ bool MacroInterpreterImpl::Step(bool is_delay_slot) {
     }
     default:
         UNIMPLEMENTED_MSG("Unimplemented macro operation {}", opcode.operation.Value());
+        break;
     }
 
     // An instruction with the Exit flag will not actually
@@ -297,6 +298,7 @@ void MacroInterpreterImpl::ProcessResult(Macro::ResultOperation operation, u32 r
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented result operation {}", operation);
+        break;
     }
 }
 

src/video_core/macro/macro_jit_x64.cpp

@@ -652,6 +652,7 @@ void MacroJITx64Impl::Compile_ProcessResult(Macro::ResultOperation operation, u3
         break;
     default:
         UNIMPLEMENTED_MSG("Unimplemented macro operation {}", operation);
+        break;
     }
 }
 

src/video_core/rasterizer_interface.h

@@ -43,7 +43,7 @@ public:
     virtual void Draw(bool is_indexed, u32 instance_count) = 0;
 
     /// Clear the current framebuffer
-    virtual void Clear() = 0;
+    virtual void Clear(u32 layer_count) = 0;
 
     /// Dispatches a compute shader invocation
     virtual void DispatchCompute() = 0;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -136,7 +136,7 @@ void RasterizerOpenGL::LoadDiskResources(u64 title_id, std::stop_token stop_load
     shader_cache.LoadDiskResources(title_id, stop_loading, callback);
 }
 
-void RasterizerOpenGL::Clear() {
+void RasterizerOpenGL::Clear(u32 layer_count) {
     MICROPROFILE_SCOPE(OpenGL_Clears);
     if (!maxwell3d->ShouldExecute()) {
         return;

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -69,7 +69,7 @@ public:
     ~RasterizerOpenGL() override;
 
     void Draw(bool is_indexed, u32 instance_count) override;
-    void Clear() override;
+    void Clear(u32 layer_count) override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -891,6 +891,7 @@ void Image::CopyBufferToImage(const VideoCommon::BufferImageCopy& copy, size_t b
         break;
     default:
         ASSERT(false);
+        break;
     }
 }
 
@@ -927,6 +928,7 @@ void Image::CopyImageToBuffer(const VideoCommon::BufferImageCopy& copy, size_t b
         break;
     default:
         ASSERT(false);
+        break;
     }
     // Compressed formats don't have a pixel format or type
     const bool is_compressed = gl_format == GL_NONE;

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -340,6 +340,7 @@ void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
         texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV;
         // UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}",
         //                   static_cast<u32>(framebuffer.pixel_format));
+        break;
     }
 
     texture.resource.Release();

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -213,7 +213,7 @@ void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
     EndTransformFeedback();
 }
 
-void RasterizerVulkan::Clear() {
+void RasterizerVulkan::Clear(u32 layer_count) {
     MICROPROFILE_SCOPE(Vulkan_Clearing);
 
     if (!maxwell3d->ShouldExecute()) {
@@ -256,7 +256,7 @@ void RasterizerVulkan::Clear() {
         .rect = regs.clear_control.use_scissor ? GetScissorState(regs, 0, up_scale, down_shift)
                                                : default_scissor,
         .baseArrayLayer = regs.clear_surface.layer,
-        .layerCount = 1,
+        .layerCount = layer_count,
     };
     if (clear_rect.rect.extent.width == 0 || clear_rect.rect.extent.height == 0) {
         return;

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -65,7 +65,7 @@ public:
     ~RasterizerVulkan() override;
 
     void Draw(bool is_indexed, u32 instance_count) override;
-    void Clear() override;
+    void Clear(u32 layer_count) override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;

src/video_core/renderer_vulkan/vk_scheduler.cpp

@@ -221,6 +221,7 @@ void Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_s
             [[fallthrough]];
         default:
             vk::Check(result);
+            break;
         }
     });
     chunk->MarkSubmit();

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -108,6 +108,7 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
             break;
         default:
             ASSERT_MSG(false, "Invalid surface type");
+            break;
         }
     }
     if (info.storage) {

src/video_core/textures/decoders.cpp

@@ -170,6 +170,7 @@ void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixe
 #undef BPP_CASE
     default:
         ASSERT_MSG(false, "Invalid bytes_per_pixel={}", bytes_per_pixel);
+        break;
     }
 }
 
@@ -217,6 +218,7 @@ void SwizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes_p
 #undef BPP_CASE
     default:
         ASSERT_MSG(false, "Invalid bytes_per_pixel={}", bytes_per_pixel);
+        break;
     }
 }
 
@@ -240,6 +242,7 @@ void UnswizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes
 #undef BPP_CASE
     default:
         ASSERT_MSG(false, "Invalid bytes_per_pixel={}", bytes_per_pixel);
+        break;
     }
 }
 

src/yuzu/compatdb.cpp

@@ -126,6 +126,7 @@ void CompatDB::Submit() {
         break;
     default:
         LOG_ERROR(Frontend, "Unexpected page: {}", currentId());
+        break;
     }
 }
 

src/yuzu/main.cpp

@@ -1955,6 +1955,7 @@ void GMainWindow::OnGameListOpenFolder(u64 program_id, GameListOpenTarget target
     }
     default:
         UNIMPLEMENTED();
+        break;
     }
 
     const QString qpath = QString::fromStdString(Common::FS::PathToUTF8String(path));
@@ -3198,6 +3199,7 @@ void GMainWindow::OnToggleGpuAccuracy() {
     case Settings::GPUAccuracy::Extreme:
     default: {
         Settings::values.gpu_accuracy.SetValue(Settings::GPUAccuracy::High);
+        break;
     }
     }
 
@@ -3530,6 +3532,7 @@ void GMainWindow::UpdateGPUAccuracyButton() {
     default: {
         gpu_accuracy_button->setText(tr("GPU ERROR"));
         gpu_accuracy_button->setChecked(true);
+        break;
     }
     }
 }

src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp

@@ -84,6 +84,7 @@ EmuWindow_SDL2_VK::EmuWindow_SDL2_VK(InputCommon::InputSubsystem* input_subsyste
     default:
         LOG_CRITICAL(Frontend, "Window manager subsystem not implemented");
         std::exit(EXIT_FAILURE);
+        break;
     }
 
     OnResize();

src/yuzu_cmd/yuzu.cpp

@@ -351,6 +351,7 @@ int main(int argc, char** argv) {
                          "additional help.\n\nError Code: {:04X}-{:04X}\nError Description: {}",
                          loader_id, error_id, static_cast<Loader::ResultStatus>(error_id));
         }
+        break;
     }
 
     system.TelemetrySession().AddField(Common::Telemetry::FieldType::App, "Frontend", "SDL");