"Merge Tagged PR 12173"

src/audio_core/sink/cubeb_sink.cpp

@@ -253,8 +253,9 @@ CubebSink::~CubebSink() {
 #endif
 }
 
-SinkStream* CubebSink::AcquireSinkStream(Core::System& system, u32 system_channels,
+SinkStream* CubebSink::AcquireSinkStream(Core::System& system, u32 system_channels_,
                                          const std::string& name, StreamType type) {
+    system_channels = system_channels_;
     SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<CubebSinkStream>(
         ctx, device_channels, system_channels, output_device, input_device, name, type, system));
 

src/audio_core/sink/sdl2_sink.cpp

@@ -168,8 +168,9 @@ SDLSink::SDLSink(std::string_view target_device_name) {
 
 SDLSink::~SDLSink() = default;
 
-SinkStream* SDLSink::AcquireSinkStream(Core::System& system, u32 system_channels,
+SinkStream* SDLSink::AcquireSinkStream(Core::System& system, u32 system_channels_,
                                        const std::string&, StreamType type) {
+    system_channels = system_channels_;
     SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<SDLSinkStream>(
         device_channels, system_channels, output_device, input_device, type, system));
     return stream.get();

src/audio_core/sink/sink.h

@@ -85,9 +85,21 @@ public:
      */
     virtual void SetSystemVolume(f32 volume) = 0;
 
+    /**
+     * Get the number of channels the game has set, can be different to the host hardware's support.
+     * Either 2 or 6.
+     *
+     * @return Number of device channels.
+     */
+    u32 GetSystemChannels() const {
+        return system_channels;
+    }
+
 protected:
     /// Number of device channels supported by the hardware
     u32 device_channels{2};
+    /// Number of channels the game is sending
+    u32 system_channels{2};
 };
 
 using SinkPtr = std::unique_ptr<Sink>;

src/audio_core/sink/sink_stream.cpp

@@ -40,29 +40,36 @@ void SinkStream::AppendBuffer(SinkBuffer& buffer, std::span<s16> samples) {
 
     if (system_channels == 6 && device_channels == 2) {
         // We're given 6 channels, but our device only outputs 2, so downmix.
-        static constexpr std::array<f32, 4> down_mix_coeff{1.0f, 0.707f, 0.251f, 0.707f};
+        // Front = 1.0
+        // Center = 0.596
+        // LFE = 0.354
+        // Back = 0.707
+        static constexpr std::array<f32, 4> down_mix_coeff{1.0, 0.596f, 0.354f, 0.707f};
 
         for (u32 read_index = 0, write_index = 0; read_index < samples.size();
              read_index += system_channels, write_index += device_channels) {
+            const auto fl =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::FrontLeft)]);
+            const auto fr =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::FrontRight)]);
+            const auto c =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::Center)]);
+            const auto lfe =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::LFE)]);
+            const auto bl =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::BackLeft)]);
+            const auto br =
+                static_cast<f32>(samples[read_index + static_cast<u32>(Channels::BackRight)]);
+
             const auto left_sample{
-                ((Common::FixedPoint<49, 15>(
-                      samples[read_index + static_cast<u32>(Channels::FrontLeft)]) *
-                      down_mix_coeff[0] +
-                  samples[read_index + static_cast<u32>(Channels::Center)] * down_mix_coeff[1] +
-                  samples[read_index + static_cast<u32>(Channels::LFE)] * down_mix_coeff[2] +
-                  samples[read_index + static_cast<u32>(Channels::BackLeft)] * down_mix_coeff[3]) *
-                 volume)
-                    .to_int()};
+                static_cast<s32>((fl * down_mix_coeff[0] + c * down_mix_coeff[1] +
+                                  lfe * down_mix_coeff[2] + bl * down_mix_coeff[3]) *
+                                 volume)};
 
             const auto right_sample{
-                ((Common::FixedPoint<49, 15>(
-                      samples[read_index + static_cast<u32>(Channels::FrontRight)]) *
-                      down_mix_coeff[0] +
-                  samples[read_index + static_cast<u32>(Channels::Center)] * down_mix_coeff[1] +
-                  samples[read_index + static_cast<u32>(Channels::LFE)] * down_mix_coeff[2] +
-                  samples[read_index + static_cast<u32>(Channels::BackRight)] * down_mix_coeff[3]) *
-                 volume)
-                    .to_int()};
+                static_cast<s32>((fr * down_mix_coeff[0] + c * down_mix_coeff[1] +
+                                  lfe * down_mix_coeff[2] + br * down_mix_coeff[3]) *
+                                 volume)};
 
             samples[write_index + static_cast<u32>(Channels::FrontLeft)] =
                 static_cast<s16>(std::clamp(left_sample, min, max));

src/core/hle/service/audio/audren_u.cpp

@@ -359,7 +359,7 @@ private:
 
     void GetActiveChannelCount(HLERequestContext& ctx) {
         const auto& sink{system.AudioCore().GetOutputSink()};
-        u32 channel_count{sink.GetDeviceChannels()};
+        u32 channel_count{sink.GetSystemChannels()};
 
         LOG_DEBUG(Service_Audio, "(STUBBED) called. Channels={}", channel_count);
 

src/core/hle/service/nvnflinger/buffer_queue_consumer.cpp

@@ -90,6 +90,18 @@ Status BufferQueueConsumer::AcquireBuffer(BufferItem* out_buffer,
 
     LOG_DEBUG(Service_Nvnflinger, "acquiring slot={}", slot);
 
+    // If the front buffer is still being tracked, update its slot state
+    if (core->StillTracking(*front)) {
+        slots[slot].acquire_called = true;
+        slots[slot].needs_cleanup_on_release = false;
+        slots[slot].buffer_state = BufferState::Acquired;
+
+        // TODO: for now, avoid resetting the fence, so that when we next return this
+        // slot to the producer, it will wait for the fence to pass. We should fix this
+        // by properly waiting for the fence in the BufferItemConsumer.
+        // slots[slot].fence = Fence::NoFence();
+    }
+
     // If the buffer has previously been acquired by the consumer, set graphic_buffer to nullptr to
     // avoid unnecessarily remapping this buffer on the consumer side.
     if (out_buffer->acquire_called) {
@@ -132,11 +144,28 @@ Status BufferQueueConsumer::ReleaseBuffer(s32 slot, u64 frame_number, const Fenc
             ++current;
         }
 
-        slots[slot].buffer_state = BufferState::Free;
+        if (slots[slot].buffer_state == BufferState::Acquired) {
+            // TODO: for now, avoid resetting the fence, so that when we next return this
+            // slot to the producer, it can wait for its own fence to pass. We should fix this
+            // by properly waiting for the fence in the BufferItemConsumer.
+            // slots[slot].fence = release_fence;
+            slots[slot].buffer_state = BufferState::Free;
 
-        listener = core->connected_producer_listener;
+            listener = core->connected_producer_listener;
 
-        LOG_DEBUG(Service_Nvnflinger, "releasing slot {}", slot);
+            LOG_DEBUG(Service_Nvnflinger, "releasing slot {}", slot);
+        } else if (slots[slot].needs_cleanup_on_release) {
+            LOG_DEBUG(Service_Nvnflinger, "releasing a stale buffer slot {} (state = {})", slot,
+                      slots[slot].buffer_state);
+            slots[slot].needs_cleanup_on_release = false;
+            return Status::StaleBufferSlot;
+        } else {
+            LOG_ERROR(Service_Nvnflinger,
+                      "attempted to release buffer slot {} but its state was {}", slot,
+                      slots[slot].buffer_state);
+
+            return Status::BadValue;
+        }
 
         core->SignalDequeueCondition();
     }

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

@@ -74,6 +74,10 @@ void BufferQueueCore::FreeBufferLocked(s32 slot) {
 
     slots[slot].graphic_buffer.reset();
 
+    if (slots[slot].buffer_state == BufferState::Acquired) {
+        slots[slot].needs_cleanup_on_release = true;
+    }
+
     slots[slot].buffer_state = BufferState::Free;
     slots[slot].frame_number = UINT32_MAX;
     slots[slot].acquire_called = false;

src/core/hle/service/nvnflinger/buffer_slot.h

@@ -31,6 +31,7 @@ struct BufferSlot final {
     u64 frame_number{};
     Fence fence;
     bool acquire_called{};
+    bool needs_cleanup_on_release{};
     bool attached_by_consumer{};
     bool is_preallocated{};
 };

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -1039,37 +1039,16 @@ 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;
-    bool force_unorm = ([&] {
-        if (!is_d24 || device.SupportsD24DepthBuffer()) {
-            return false;
-        }
-        if (device.IsExtDepthBiasControlSupported()) {
-            return true;
-        }
-        if (!Settings::values.renderer_amdvlk_depth_bias_workaround) {
-            return false;
-        }
+    if (is_d24 && !device.SupportsD24DepthBuffer() &&
+        Settings::values.renderer_amdvlk_depth_bias_workaround) {
         // 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 =
             static_cast<double>(1ULL << (32 - 24)) / (static_cast<double>(0x1.ep+127));
         units = static_cast<float>(static_cast<double>(units) * rescale_factor);
-        return false;
-    })();
+    }
     scheduler.Record([constant = units, clamp = regs.depth_bias_clamp,
-                      factor = regs.slope_scale_depth_bias, force_unorm,
-                      precise = device.HasExactDepthBiasControl()](vk::CommandBuffer cmdbuf) {
-        if (force_unorm) {
-            VkDepthBiasRepresentationInfoEXT info{
-                .sType = VK_STRUCTURE_TYPE_DEPTH_BIAS_REPRESENTATION_INFO_EXT,
-                .pNext = nullptr,
-                .depthBiasRepresentation =
-                    VK_DEPTH_BIAS_REPRESENTATION_LEAST_REPRESENTABLE_VALUE_FORCE_UNORM_EXT,
-                .depthBiasExact = precise ? VK_TRUE : VK_FALSE,
-            };
-            cmdbuf.SetDepthBias(constant, clamp, factor, &info);
-            return;
-        }
+                      factor = regs.slope_scale_depth_bias](vk::CommandBuffer cmdbuf) {
         cmdbuf.SetDepthBias(constant, clamp, factor);
     });
 }

src/video_core/vulkan_common/vulkan_device.cpp

@@ -1130,13 +1130,6 @@ 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,

src/video_core/vulkan_common/vulkan_device.h

@@ -41,7 +41,6 @@ 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)         \
@@ -97,7 +96,6 @@ 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)                                 \
@@ -150,9 +148,6 @@ 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)                                                 \
@@ -479,11 +474,6 @@ 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;
@@ -649,10 +639,6 @@ public:
         return features.robustness2.nullDescriptor;
     }
 
-    bool HasExactDepthBiasControl() const {
-        return features.depth_bias_control.depthBiasExact;
-    }
-
     u32 GetMaxVertexInputAttributes() const {
         return properties.properties.limits.maxVertexInputAttributes;
     }