Merge pull request #7118 from ameerj/vc-gpu-impl

gpu: Migrate implementation to the cpp file

src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp

@@ -48,8 +48,9 @@ void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u3
               addr, offset, width, height, stride, format);
 
     const auto pixel_format = static_cast<Tegra::FramebufferConfig::PixelFormat>(format);
-    const Tegra::FramebufferConfig framebuffer{addr,   offset,       width,     height,
-                                               stride, pixel_format, transform, crop_rect};
+    const auto transform_flags = static_cast<Tegra::FramebufferConfig::TransformFlags>(transform);
+    const Tegra::FramebufferConfig framebuffer{addr,   offset,       width,           height,
+                                               stride, pixel_format, transform_flags, crop_rect};
 
     system.GetPerfStats().EndSystemFrame();
     system.GPU().SwapBuffers(&framebuffer);

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

@@ -111,7 +111,6 @@ NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector
         event.event->GetWritableEvent().Signal();
         return NvResult::Success;
     }
-    auto lock = gpu.LockSync();
     const u32 current_syncpoint_value = event.fence.value;
     const s32 diff = current_syncpoint_value - params.threshold;
     if (diff >= 0) {
@@ -132,23 +131,24 @@ NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector
     }
 
     EventState status = events_interface.status[event_id];
-    if (event_id < MaxNvEvents || status == EventState::Free || status == EventState::Registered) {
-        events_interface.SetEventStatus(event_id, EventState::Waiting);
-        events_interface.assigned_syncpt[event_id] = params.syncpt_id;
-        events_interface.assigned_value[event_id] = target_value;
-        if (is_async) {
-            params.value = params.syncpt_id << 4;
-        } else {
-            params.value = ((params.syncpt_id & 0xfff) << 16) | 0x10000000;
-        }
-        params.value |= event_id;
-        event.event->GetWritableEvent().Clear();
-        gpu.RegisterSyncptInterrupt(params.syncpt_id, target_value);
+    const bool bad_parameter = status != EventState::Free && status != EventState::Registered;
+    if (bad_parameter) {
         std::memcpy(output.data(), &params, sizeof(params));
-        return NvResult::Timeout;
+        return NvResult::BadParameter;
     }
+    events_interface.SetEventStatus(event_id, EventState::Waiting);
+    events_interface.assigned_syncpt[event_id] = params.syncpt_id;
+    events_interface.assigned_value[event_id] = target_value;
+    if (is_async) {
+        params.value = params.syncpt_id << 4;
+    } else {
+        params.value = ((params.syncpt_id & 0xfff) << 16) | 0x10000000;
+    }
+    params.value |= event_id;
+    event.event->GetWritableEvent().Clear();
+    gpu.RegisterSyncptInterrupt(params.syncpt_id, target_value);
     std::memcpy(output.data(), &params, sizeof(params));
-    return NvResult::BadParameter;
+    return NvResult::Timeout;
 }
 
 NvResult nvhost_ctrl::IocCtrlEventRegister(const std::vector<u8>& input, std::vector<u8>& output) {

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

@@ -13,6 +13,14 @@
 #include "video_core/memory_manager.h"
 
 namespace Service::Nvidia::Devices {
+namespace {
+Tegra::CommandHeader BuildFenceAction(Tegra::GPU::FenceOperation op, u32 syncpoint_id) {
+    Tegra::GPU::FenceAction result{};
+    result.op.Assign(op);
+    result.syncpoint_id.Assign(syncpoint_id);
+    return {result.raw};
+}
+} // namespace
 
 nvhost_gpu::nvhost_gpu(Core::System& system_, std::shared_ptr<nvmap> nvmap_dev_,
                        SyncpointManager& syncpoint_manager_)
@@ -187,7 +195,7 @@ static std::vector<Tegra::CommandHeader> BuildWaitCommandList(Fence fence) {
         {fence.value},
         Tegra::BuildCommandHeader(Tegra::BufferMethods::FenceAction, 1,
                                   Tegra::SubmissionMode::Increasing),
-        Tegra::GPU::FenceAction::Build(Tegra::GPU::FenceOperation::Acquire, fence.id),
+        BuildFenceAction(Tegra::GPU::FenceOperation::Acquire, fence.id),
     };
 }
 
@@ -200,8 +208,7 @@ static std::vector<Tegra::CommandHeader> BuildIncrementCommandList(Fence fence,
     for (u32 count = 0; count < add_increment; ++count) {
         result.emplace_back(Tegra::BuildCommandHeader(Tegra::BufferMethods::FenceAction, 1,
                                                       Tegra::SubmissionMode::Increasing));
-        result.emplace_back(
-            Tegra::GPU::FenceAction::Build(Tegra::GPU::FenceOperation::Increment, fence.id));
+        result.emplace_back(BuildFenceAction(Tegra::GPU::FenceOperation::Increment, fence.id));
     }
 
     return result;

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

@@ -13,28 +13,20 @@
 #include "common/thread.h"
 #include "core/core.h"
 #include "core/core_timing.h"
-#include "core/core_timing_util.h"
-#include "core/hardware_properties.h"
 #include "core/hle/kernel/k_readable_event.h"
-#include "core/hle/kernel/kernel.h"
 #include "core/hle/service/nvdrv/devices/nvdisp_disp0.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
 #include "core/hle/service/nvflinger/nvflinger.h"
 #include "core/hle/service/vi/display/vi_display.h"
 #include "core/hle/service/vi/layer/vi_layer.h"
-#include "core/perf_stats.h"
-#include "video_core/renderer_base.h"
+#include "video_core/gpu.h"
 
 namespace Service::NVFlinger {
 
 constexpr auto frame_ns = std::chrono::nanoseconds{1000000000 / 60};
 
-void NVFlinger::VSyncThread(NVFlinger& nv_flinger) {
-    nv_flinger.SplitVSync();
-}
-
-void NVFlinger::SplitVSync() {
+void NVFlinger::SplitVSync(std::stop_token stop_token) {
     system.RegisterHostThread();
     std::string name = "yuzu:VSyncThread";
     MicroProfileOnThreadCreate(name.c_str());
@@ -45,7 +37,7 @@ void NVFlinger::SplitVSync() {
     Common::SetCurrentThreadName(name.c_str());
     Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
     s64 delay = 0;
-    while (is_running) {
+    while (!stop_token.stop_requested()) {
         guard->lock();
         const s64 time_start = system.CoreTiming().GetGlobalTimeNs().count();
         Compose();
@@ -55,7 +47,7 @@ void NVFlinger::SplitVSync() {
         const s64 next_time = std::max<s64>(0, ticks - time_passed - delay);
         guard->unlock();
         if (next_time > 0) {
-            wait_event->WaitFor(std::chrono::nanoseconds{next_time});
+            std::this_thread::sleep_for(std::chrono::nanoseconds{next_time});
         }
         delay = (system.CoreTiming().GetGlobalTimeNs().count() - time_end) - next_time;
     }
@@ -84,9 +76,7 @@ NVFlinger::NVFlinger(Core::System& system_)
         });
 
     if (system.IsMulticore()) {
-        is_running = true;
-        wait_event = std::make_unique<Common::Event>();
-        vsync_thread = std::make_unique<std::thread>(VSyncThread, std::ref(*this));
+        vsync_thread = std::jthread([this](std::stop_token token) { SplitVSync(token); });
     } else {
         system.CoreTiming().ScheduleEvent(frame_ns, composition_event);
     }
@@ -96,14 +86,7 @@ NVFlinger::~NVFlinger() {
     for (auto& buffer_queue : buffer_queues) {
         buffer_queue->Disconnect();
     }
-
-    if (system.IsMulticore()) {
-        is_running = false;
-        wait_event->Set();
-        vsync_thread->join();
-        vsync_thread.reset();
-        wait_event.reset();
-    } else {
+    if (!system.IsMulticore()) {
         system.CoreTiming().UnscheduleEvent(composition_event, 0);
     }
 }

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

@@ -4,13 +4,10 @@
 
 #pragma once
 
-#include <atomic>
 #include <list>
 #include <memory>
 #include <mutex>
 #include <optional>
-#include <string>
-#include <string_view>
 #include <thread>
 #include <vector>
 
@@ -109,9 +106,7 @@ private:
     /// Creates a layer with the specified layer ID in the desired display.
     void CreateLayerAtId(VI::Display& display, u64 layer_id);
 
-    static void VSyncThread(NVFlinger& nv_flinger);
-
-    void SplitVSync();
+    void SplitVSync(std::stop_token stop_token);
 
     std::shared_ptr<Nvidia::Module> nvdrv;
 
@@ -133,9 +128,7 @@ private:
 
     Core::System& system;
 
-    std::unique_ptr<std::thread> vsync_thread;
-    std::unique_ptr<Common::Event> wait_event;
-    std::atomic<bool> is_running{};
+    std::jthread vsync_thread;
 
     KernelHelpers::ServiceContext service_context;
 };