early-access version 4174EA-4174

17 files changed, 66 insertions, 30 deletions

diff --git a/README.md b/README.md
index 0bcbbc551..08db4df06 100755
--- a/README.md
+++ b/README.md
@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 4173.
+This is the source code for early-access 4174.
 
 ## Legal Notice
 
diff --git a/src/common/settings.h b/src/common/settings.h
index 4e4071189..0c7d22073 100755
--- a/src/common/settings.h
+++ b/src/common/settings.h
@@ -435,8 +435,6 @@ struct Values {
         linkage, false, "disable_shader_loop_safety_checks", Category::RendererDebug};

     Setting<bool> enable_renderdoc_hotkey{linkage, false, "renderdoc_hotkey",

                                           Category::RendererDebug};

-    // TODO: remove this once AMDVLK supports VK_EXT_depth_bias_control

-    bool renderer_amdvlk_depth_bias_workaround{};

     Setting<bool> disable_buffer_reorder{linkage, false, "disable_buffer_reorder",

                                          Category::RendererDebug};

 

diff --git a/src/core/core.cpp b/src/core/core.cpp
index 6ea4c7849..ee06615d5 100755
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -394,11 +394,6 @@ struct System::Impl {
             room_member->SendGameInfo(game_info);

         }

 

-        // Workarounds:

-        // Activate this in Super Smash Brothers Ultimate, it only affects AMD cards using AMDVLK

-        Settings::values.renderer_amdvlk_depth_bias_workaround =

-            params.program_id == 0x1006A800016E000ULL;

-

         status = SystemResultStatus::Success;

         return status;

     }

@@ -457,9 +452,6 @@ struct System::Impl {
             room_member->SendGameInfo(game_info);

         }

 

-        // Workarounds

-        Settings::values.renderer_amdvlk_depth_bias_workaround = false;

-

         // Reset all glue registrations

         arp_manager.ResetAll();

 

diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
index e71c58aa9..616c387e9 100755
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
@@ -5,6 +5,7 @@
 #include "common/assert.h"

 #include "common/logging/log.h"

 #include "core/core.h"

+#include "core/hle/kernel/k_process.h"

 #include "core/hle/service/nvdrv/core/container.h"

 #include "core/hle/service/nvdrv/core/nvmap.h"

 #include "core/hle/service/nvdrv/core/syncpoint_manager.h"

@@ -75,7 +76,7 @@ NvResult nvhost_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
         case 0xd:

             return WrapFixed(this, &nvhost_gpu::SetChannelPriority, input, output);

         case 0x1a:

-            return WrapFixed(this, &nvhost_gpu::AllocGPFIFOEx2, input, output);

+            return WrapFixed(this, &nvhost_gpu::AllocGPFIFOEx2, input, output, fd);

         case 0x1b:

             return WrapFixedVariable(this, &nvhost_gpu::SubmitGPFIFOBase1, input, output, true);

         case 0x1d:

@@ -120,8 +121,13 @@ NvResult nvhost_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     return NvResult::NotImplemented;

 }

 

-void nvhost_gpu::OnOpen(NvCore::SessionId session_id, DeviceFD fd) {}

-void nvhost_gpu::OnClose(DeviceFD fd) {}

+void nvhost_gpu::OnOpen(NvCore::SessionId session_id, DeviceFD fd) {

+    sessions[fd] = session_id;

+}

+

+void nvhost_gpu::OnClose(DeviceFD fd) {

+    sessions.erase(fd);

+}

 

 NvResult nvhost_gpu::SetNVMAPfd(IoctlSetNvmapFD& params) {

     LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);

@@ -161,7 +167,7 @@ NvResult nvhost_gpu::SetChannelPriority(IoctlChannelSetPriority& params) {
     return NvResult::Success;

 }

 

-NvResult nvhost_gpu::AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params) {

+NvResult nvhost_gpu::AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params, DeviceFD fd) {

     LOG_WARNING(Service_NVDRV,

                 "(STUBBED) called, num_entries={:X}, flags={:X}, unk0={:X}, "

                 "unk1={:X}, unk2={:X}, unk3={:X}",

@@ -173,7 +179,12 @@ NvResult nvhost_gpu::AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params) {
         return NvResult::AlreadyAllocated;

     }

 

-    system.GPU().InitChannel(*channel_state);

+    u64 program_id{};

+    if (auto* const session = core.GetSession(sessions[fd]); session != nullptr) {

+        program_id = session->process->GetProgramId();

+    }

+

+    system.GPU().InitChannel(*channel_state, program_id);

 

     params.fence_out = syncpoint_manager.GetSyncpointFence(channel_syncpoint);

 

diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
index 40c2291b1..cde3418ad 100755
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
@@ -192,7 +192,7 @@ private:
     NvResult ZCullBind(IoctlZCullBind& params);

     NvResult SetErrorNotifier(IoctlSetErrorNotifier& params);

     NvResult SetChannelPriority(IoctlChannelSetPriority& params);

-    NvResult AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params);

+    NvResult AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params, DeviceFD fd);

     NvResult AllocateObjectContext(IoctlAllocObjCtx& params);

 

     NvResult SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, Tegra::CommandList&& entries);

@@ -210,6 +210,7 @@ private:
     NvCore::SyncpointManager& syncpoint_manager;

     NvCore::NvMap& nvmap;

     std::shared_ptr<Tegra::Control::ChannelState> channel_state;

+    std::unordered_map<DeviceFD, NvCore::SessionId> sessions;

     u32 channel_syncpoint;

     std::mutex channel_mutex;

 

diff --git a/src/video_core/buffer_cache/usage_tracker.h b/src/video_core/buffer_cache/usage_tracker.h
index 5f8688d31..ae511ccb6 100755
--- a/src/video_core/buffer_cache/usage_tracker.h
+++ b/src/video_core/buffer_cache/usage_tracker.h
@@ -26,6 +26,9 @@ public:
     void Track(u64 offset, u64 size) noexcept {
         const size_t page = offset >> PAGE_SHIFT;
         const size_t page_end = (offset + size) >> PAGE_SHIFT;
+        if (page_end < page || page_end >= pages.size()) {
+            return;
+        }
         TrackPage(page, offset, size);
         if (page == page_end) {
             return;
@@ -41,6 +44,9 @@ public:
     [[nodiscard]] bool IsUsed(u64 offset, u64 size) const noexcept {
         const size_t page = offset >> PAGE_SHIFT;
         const size_t page_end = (offset + size) >> PAGE_SHIFT;
+        if (page_end < page || page_end >= pages.size()) {
+            return false;
+        }
         if (IsPageUsed(page, offset, size)) {
             return true;
         }
diff --git a/src/video_core/control/channel_state.cpp b/src/video_core/control/channel_state.cpp
index d182bad57..a3780a046 100755
--- a/src/video_core/control/channel_state.cpp
+++ b/src/video_core/control/channel_state.cpp
@@ -16,8 +16,9 @@ namespace Tegra::Control {
 

 ChannelState::ChannelState(s32 bind_id_) : bind_id{bind_id_}, initialized{} {}

 

-void ChannelState::Init(Core::System& system, GPU& gpu) {

+void ChannelState::Init(Core::System& system, GPU& gpu, u64 program_id_) {

     ASSERT(memory_manager);

+    program_id = program_id_;

     dma_pusher = std::make_unique<Tegra::DmaPusher>(system, gpu, *memory_manager, *this);

     maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, *memory_manager);

     fermi_2d = std::make_unique<Engines::Fermi2D>(*memory_manager);

diff --git a/src/video_core/control/channel_state.h b/src/video_core/control/channel_state.h
index ceaa92647..12fb84811 100755
--- a/src/video_core/control/channel_state.h
+++ b/src/video_core/control/channel_state.h
@@ -40,11 +40,12 @@ struct ChannelState {
     ChannelState(ChannelState&& other) noexcept = default;

     ChannelState& operator=(ChannelState&& other) noexcept = default;

 

-    void Init(Core::System& system, GPU& gpu);

+    void Init(Core::System& system, GPU& gpu, u64 program_id);

 

     void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);

 

     s32 bind_id = -1;

+    u64 program_id = 0;

     /// 3D engine

     std::unique_ptr<Engines::Maxwell3D> maxwell_3d;

     /// 2D engine

diff --git a/src/video_core/control/channel_state_cache.cpp b/src/video_core/control/channel_state_cache.cpp
index eb59d65c5..d225de7ac 100755
--- a/src/video_core/control/channel_state_cache.cpp
+++ b/src/video_core/control/channel_state_cache.cpp
@@ -7,7 +7,7 @@ namespace VideoCommon {
 

 ChannelInfo::ChannelInfo(Tegra::Control::ChannelState& channel_state)

     : maxwell3d{*channel_state.maxwell_3d}, kepler_compute{*channel_state.kepler_compute},

-      gpu_memory{*channel_state.memory_manager} {}

+      gpu_memory{*channel_state.memory_manager}, program_id{channel_state.program_id} {}

 

 template class VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo>;

 

diff --git a/src/video_core/control/channel_state_cache.h b/src/video_core/control/channel_state_cache.h
index efc992d83..5d7c411b9 100755
--- a/src/video_core/control/channel_state_cache.h
+++ b/src/video_core/control/channel_state_cache.h
@@ -39,6 +39,7 @@ public:
     Tegra::Engines::Maxwell3D& maxwell3d;

     Tegra::Engines::KeplerCompute& kepler_compute;

     Tegra::MemoryManager& gpu_memory;

+    u64 program_id;

 };

 

 template <class P>

@@ -77,9 +78,10 @@ protected:
     P* channel_state;

     size_t current_channel_id{UNSET_CHANNEL};

     size_t current_address_space{};

-    Tegra::Engines::Maxwell3D* maxwell3d;

-    Tegra::Engines::KeplerCompute* kepler_compute;

-    Tegra::MemoryManager* gpu_memory;

+    Tegra::Engines::Maxwell3D* maxwell3d{};

+    Tegra::Engines::KeplerCompute* kepler_compute{};

+    Tegra::MemoryManager* gpu_memory{};

+    u64 program_id{};

 

     std::deque<P> channel_storage;

     std::deque<size_t> free_channel_ids;

diff --git a/src/video_core/control/channel_state_cache.inc b/src/video_core/control/channel_state_cache.inc
index 5a6ab0052..e3d2b4447 100755
--- a/src/video_core/control/channel_state_cache.inc
+++ b/src/video_core/control/channel_state_cache.inc
@@ -58,6 +58,7 @@ void ChannelSetupCaches<P>::BindToChannel(s32 id) {
     maxwell3d = &channel_state->maxwell3d;

     kepler_compute = &channel_state->kepler_compute;

     gpu_memory = &channel_state->gpu_memory;

+    program_id = channel_state->program_id;

     current_address_space = gpu_memory->GetID();

 }

 

@@ -76,6 +77,7 @@ void ChannelSetupCaches<P>::EraseChannel(s32 id) {
         maxwell3d = nullptr;

         kepler_compute = nullptr;

         gpu_memory = nullptr;

+        program_id = 0;

     } else if (current_channel_id != UNSET_CHANNEL) {

         channel_state = &channel_storage[current_channel_id];

     }

diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 59356015b..d325608c9 100755
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -67,8 +67,8 @@ struct GPU::Impl {
         return CreateChannel(new_channel_id++);

     }

 

-    void InitChannel(Control::ChannelState& to_init) {

-        to_init.Init(system, gpu);

+    void InitChannel(Control::ChannelState& to_init, u64 program_id) {

+        to_init.Init(system, gpu, program_id);

         to_init.BindRasterizer(rasterizer);

         rasterizer->InitializeChannel(to_init);

     }

@@ -387,8 +387,8 @@ std::shared_ptr<Control::ChannelState> GPU::AllocateChannel() {
     return impl->AllocateChannel();

 }

 

-void GPU::InitChannel(Control::ChannelState& to_init) {

-    impl->InitChannel(to_init);

+void GPU::InitChannel(Control::ChannelState& to_init, u64 program_id) {

+    impl->InitChannel(to_init, program_id);

 }

 

 void GPU::BindChannel(s32 channel_id) {

diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 25c75a109..a4726ce13 100755
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -149,7 +149,7 @@ public:
 

     std::shared_ptr<Control::ChannelState> AllocateChannel();

 

-    void InitChannel(Control::ChannelState& to_init);

+    void InitChannel(Control::ChannelState& to_init, u64 program_id);

 

     void BindChannel(s32 channel_id);

 

diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 5ada85685..c3530c79c 100755
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -215,6 +215,7 @@ ShaderCache::ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
           .support_gl_variable_aoffi = device.HasVariableAoffi(),

           .support_gl_sparse_textures = device.HasSparseTexture2(),

           .support_gl_derivative_control = device.HasDerivativeControl(),

+          .support_geometry_streams = true,

 

           .warp_size_potentially_larger_than_guest = device.IsWarpSizePotentiallyLargerThanGuest(),

 

diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index d9b43379c..cc47eb345 100755
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -1082,8 +1082,8 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {
                         regs.zeta.format == Tegra::DepthFormat::X8Z24_UNORM ||

                         regs.zeta.format == Tegra::DepthFormat::S8Z24_UNORM ||

                         regs.zeta.format == Tegra::DepthFormat::V8Z24_UNORM;

-    if (is_d24 && !device.SupportsD24DepthBuffer() &&

-        Settings::values.renderer_amdvlk_depth_bias_workaround) {

+    if (is_d24 && !device.SupportsD24DepthBuffer() && program_id == 0x1006A800016E000ULL) {

+        // Only activate this in Super Smash Brothers Ultimate

         // the base formulas can be obtained from here:

         //   https://docs.microsoft.com/en-us/windows/win32/direct3d11/d3d10-graphics-programming-guide-output-merger-stage-depth-bias

         const double rescale_factor =

diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp
index 7f183bacc..f11d0ba16 100755
--- a/src/video_core/vulkan_common/vulkan_device.cpp
+++ b/src/video_core/vulkan_common/vulkan_device.cpp
@@ -1137,6 +1137,13 @@ void Device::RemoveUnsuitableExtensions() {
     RemoveExtensionFeatureIfUnsuitable(extensions.custom_border_color, features.custom_border_color,

                                        VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME);

 

+    // VK_EXT_depth_bias_control

+    extensions.depth_bias_control =

+        features.depth_bias_control.depthBiasControl &&

+        features.depth_bias_control.leastRepresentableValueForceUnormRepresentation;

+    RemoveExtensionFeatureIfUnsuitable(extensions.depth_bias_control, features.depth_bias_control,

+                                       VK_EXT_DEPTH_BIAS_CONTROL_EXTENSION_NAME);

+

     // VK_EXT_depth_clip_control

     extensions.depth_clip_control = features.depth_clip_control.depthClipControl;

     RemoveExtensionFeatureIfUnsuitable(extensions.depth_clip_control, features.depth_clip_control,

diff --git a/src/video_core/vulkan_common/vulkan_device.h b/src/video_core/vulkan_common/vulkan_device.h
index f0cb2b2dc..e6e4a3089 100755
--- a/src/video_core/vulkan_common/vulkan_device.h
+++ b/src/video_core/vulkan_common/vulkan_device.h
@@ -41,6 +41,7 @@ VK_DEFINE_HANDLE(VmaAllocator)
 // Define all features which may be used by the implementation and require an extension here.

 #define FOR_EACH_VK_FEATURE_EXT(FEATURE)                                                           \

     FEATURE(EXT, CustomBorderColor, CUSTOM_BORDER_COLOR, custom_border_color)                      \

+    FEATURE(EXT, DepthBiasControl, DEPTH_BIAS_CONTROL, depth_bias_control)                         \

     FEATURE(EXT, DepthClipControl, DEPTH_CLIP_CONTROL, depth_clip_control)                         \

     FEATURE(EXT, ExtendedDynamicState, EXTENDED_DYNAMIC_STATE, extended_dynamic_state)             \

     FEATURE(EXT, ExtendedDynamicState2, EXTENDED_DYNAMIC_STATE_2, extended_dynamic_state2)         \

@@ -96,6 +97,7 @@ VK_DEFINE_HANDLE(VmaAllocator)
 #define FOR_EACH_VK_RECOMMENDED_EXTENSION(EXTENSION_NAME)                                          \

     EXTENSION_NAME(VK_EXT_CONDITIONAL_RENDERING_EXTENSION_NAME)                                    \

     EXTENSION_NAME(VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME)                               \

+    EXTENSION_NAME(VK_EXT_DEPTH_BIAS_CONTROL_EXTENSION_NAME)                                       \

     EXTENSION_NAME(VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME)                                 \

     EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME)                                   \

     EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME)                                 \

@@ -148,6 +150,9 @@ VK_DEFINE_HANDLE(VmaAllocator)
 // Define features where the absence of the feature may result in a degraded experience.

 #define FOR_EACH_VK_RECOMMENDED_FEATURE(FEATURE_NAME)                                              \

     FEATURE_NAME(custom_border_color, customBorderColors)                                          \

+    FEATURE_NAME(depth_bias_control, depthBiasControl)                                             \

+    FEATURE_NAME(depth_bias_control, leastRepresentableValueForceUnormRepresentation)              \

+    FEATURE_NAME(depth_bias_control, depthBiasExact)                                               \

     FEATURE_NAME(extended_dynamic_state, extendedDynamicState)                                     \

     FEATURE_NAME(format_a4b4g4r4, formatA4B4G4R4)                                                  \

     FEATURE_NAME(index_type_uint8, indexTypeUint8)                                                 \

@@ -474,6 +479,11 @@ public:
         return extensions.depth_clip_control;

     }

 

+    /// Returns true if the device supports VK_EXT_depth_bias_control.

+    bool IsExtDepthBiasControlSupported() const {

+        return extensions.depth_bias_control;

+    }

+

     /// Returns true if the device supports VK_EXT_shader_viewport_index_layer.

     bool IsExtShaderViewportIndexLayerSupported() const {

         return extensions.shader_viewport_index_layer;

@@ -644,6 +654,10 @@ public:
         return features.robustness2.nullDescriptor;

     }

 

+    bool HasExactDepthBiasControl() const {

+        return features.depth_bias_control.depthBiasExact;

+    }

+

     u32 GetMaxVertexInputAttributes() const {

         return properties.properties.limits.maxVertexInputAttributes;

     }