[vulkan, android] Add BGR565 emulated toggle

This reverts commit 470faa62551294e8073478552ccec437b0d4a9c9.

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt

@@ -17,6 +17,7 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
     USE_CUSTOM_CPU_TICKS("use_custom_cpu_ticks"),
     SKIP_CPU_INNER_INVALIDATION("skip_cpu_inner_invalidation"),
     FIX_BLOOM_EFFECTS("fix_bloom_effects"),
+    EMULATE_BGR565("emulate_bgr565"),
     CPUOPT_UNSAFE_HOST_MMU("cpuopt_unsafe_host_mmu"),
     USE_DOCKED_MODE("use_docked_mode"),
     USE_AUTO_STUB("use_auto_stub"),

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt

@@ -756,6 +756,13 @@ abstract class SettingsItem(
                     descriptionId = R.string.fix_bloom_effects_description
                 )
             )
+            put(
+                SwitchSetting(
+                    BooleanSetting.EMULATE_BGR565,
+                    titleId = R.string.emulate_bgr565,
+                    descriptionId = R.string.emulate_bgr565_description
+                )
+            )
             put(
                 SwitchSetting(
                     BooleanSetting.CPUOPT_UNSAFE_HOST_MMU,

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/ui/SettingsFragmentPresenter.kt

@@ -293,6 +293,7 @@ class SettingsFragmentPresenter(
             add(IntSetting.FAST_GPU_TIME.key)
             add(BooleanSetting.SKIP_CPU_INNER_INVALIDATION.key)
             add(BooleanSetting.FIX_BLOOM_EFFECTS.key)
+            add(BooleanSetting.EMULATE_BGR565.key)
             add(BooleanSetting.RENDERER_ASYNCHRONOUS_SHADERS.key)
             add(BooleanSetting.RENDERER_ASYNCHRONOUS_GPU_EMULATION.key)
             add(BooleanSetting.RENDERER_ASYNC_PRESENTATION.key)

src/android/app/src/main/res/values/strings.xml

@@ -501,7 +501,7 @@
     <string name="enable_buffer_history">Enable buffer history</string>
     <string name="enable_buffer_history_description">Enables access to previous buffer states. This option may improve rendering quality and performance consistency in some games.</string>
     <string name="use_optimized_vertex_buffers">Optimized Vertex Buffers</string>
-    <string name="use_optimized_vertex_buffers_description">Enables optimized vertex buffer binding for improved performance. Requires Mesa 26.0+ Turnip drivers. Will crash on older drivers.</string>
+    <string name="use_optimized_vertex_buffers_description">Enables optimized vertex buffer binding for improved performance. Requires Mesa 26.0+ Turnip drivers/ QCOM drivers. Will crash on older Turnip drivers.</string>
 
     <string name="hacks">Hacks</string>
 
@@ -510,7 +510,9 @@
     <string name="skip_cpu_inner_invalidation">Skip CPU Inner Invalidation</string>
     <string name="skip_cpu_inner_invalidation_description">Skips certain CPU-side cache invalidations during memory updates, reducing CPU usage and improving it\'s performance. This may cause glitches or crashes on some games.</string>
     <string name="fix_bloom_effects">Fix Bloom Effects</string>
-    <string name="fix_bloom_effects_description">Reduces bloom blur in LA/EOW (Adreno 700), removes bloom in Burnout. Warning: may cause graphical artifacts in other games.</string>
+    <string name="fix_bloom_effects_description">Reduces bloom blur in LA/EOW (Adreno A6XX - A7XX/ Turnip), removes bloom in Burnout. Warning: may cause graphical artifacts in other games.</string>
+    <string name="emulate_bgr565">Emulate BGR565</string>
+    <string name="emulate_bgr565_description">Fixes problems with inverted colors in games or strange artifacts or strange shadows.</string>
     <string name="renderer_asynchronous_shaders">Use asynchronous shaders</string>
     <string name="renderer_asynchronous_shaders_description">Compiles shaders asynchronously. This may reduce stutters but may also introduce glitches.</string>
     <string name="gpu_unswizzle_settings">GPU Unswizzle Settings</string>

src/common/settings.h

@@ -555,6 +555,9 @@ struct Values {
     SwitchableSetting<bool> fix_bloom_effects{linkage, false, "fix_bloom_effects",
                                                      Category::RendererHacks};
 
+    SwitchableSetting<bool> emulate_bgr565{linkage, false, "emulate_bgr565",
+                                            Category::RendererHacks};
+
     SwitchableSetting<bool> rescale_hack{linkage, false, "rescale_hack",
                                                      Category::RendererHacks};
 
@@ -584,7 +587,7 @@ struct Values {
 
     SwitchableSetting<ExtendedDynamicState> dyna_state{linkage,
 #if defined(ANDROID)
-                                           ExtendedDynamicState::EDS1,
+                                           ExtendedDynamicState::Disabled,
 #elif defined(__APPLE__)
                                            ExtendedDynamicState::Disabled,
 #else

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -178,6 +178,9 @@ void DefineGenericOutput(EmitContext& ctx, size_t index, std::optional<u32> invo
             ctx.Decorate(id, spv::Decoration::XfbBuffer, xfb_varying->buffer);
             ctx.Decorate(id, spv::Decoration::XfbStride, xfb_varying->stride);
             ctx.Decorate(id, spv::Decoration::Offset, xfb_varying->offset);
+            if (ctx.stage == Stage::Geometry && xfb_varying->stream != 0) {
+                ctx.Decorate(id, spv::Decoration::Stream, xfb_varying->stream);
+            }
         }
         if (num_components < 4 || element > 0) {
             const std::string_view subswizzle{swizzle.substr(element, num_components)};

src/shader_recompiler/runtime_info.h

@@ -76,6 +76,7 @@ enum class TessSpacing {
 
 struct TransformFeedbackVarying {
     u32 buffer{};
+    u32 stream{};
     u32 stride{};
     u32 offset{};
     u32 components{};

src/video_core/buffer_cache/buffer_cache.h

@@ -1067,26 +1067,29 @@ void BufferCache<P>::BindHostTransformFeedbackBuffers() {
     HostBindings<typename P::Buffer> host_bindings;
     for (u32 index = 0; index < NUM_TRANSFORM_FEEDBACK_BUFFERS; ++index) {
         const Binding& binding = channel_state->transform_feedback_buffers[index];
-        if (maxwell3d->regs.transform_feedback.controls[index].varying_count == 0 &&
-            maxwell3d->regs.transform_feedback.controls[index].stride == 0) {
-            break;
+        const auto& control = maxwell3d->regs.transform_feedback.controls[index];
+        const bool has_layout = control.varying_count != 0 || control.stride != 0;
+
+        Buffer* host_buffer = &slot_buffers[NULL_BUFFER_ID];
+        u32 offset = 0;
+        u32 size = 0;
+
+        if (has_layout && binding.buffer_id != NULL_BUFFER_ID && binding.size != 0) {
+            Buffer& buffer = slot_buffers[binding.buffer_id];
+            TouchBuffer(buffer, binding.buffer_id);
+            size = binding.size;
+            SynchronizeBuffer(buffer, binding.device_addr, size);
+            MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size);
+            offset = buffer.Offset(binding.device_addr);
+            buffer.MarkUsage(offset, size);
+            host_buffer = &buffer;
         }
-        Buffer& buffer = slot_buffers[binding.buffer_id];
-        TouchBuffer(buffer, binding.buffer_id);
-        const u32 size = binding.size;
-        SynchronizeBuffer(buffer, binding.device_addr, size);
 
-        MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size);
-
-        const u32 offset = buffer.Offset(binding.device_addr);
-        buffer.MarkUsage(offset, size);
-        host_bindings.buffers.push_back(&buffer);
+        host_bindings.buffers.push_back(host_buffer);
         host_bindings.offsets.push_back(offset);
         host_bindings.sizes.push_back(size);
     }
-    if (host_bindings.buffers.size() > 0) {
-        runtime.BindTransformFeedbackBuffers(host_bindings);
-    }
+    runtime.BindTransformFeedbackBuffers(host_bindings);
 }
 
 template <class P>

src/video_core/gpu_thread.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
@@ -92,7 +92,16 @@ void ThreadManager::InvalidateRegion(DAddr addr, u64 size) {
 }
 
 void ThreadManager::FlushAndInvalidateRegion(DAddr addr, u64 size) {
-    // Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important
+    if (Settings::IsGPULevelHigh()) {
+        if (!is_async) {
+            PushCommand(FlushRegionCommand(addr, size));
+        } else {
+            auto& gpu = system.GPU();
+            const u64 fence = gpu.RequestFlush(addr, size);
+            TickGPU();
+            gpu.WaitForSyncOperation(fence);
+        }
+    }
     rasterizer->OnCacheInvalidation(addr, size);
 }
 

src/video_core/host_shaders/resolve_conditional_render.comp

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
@@ -5,16 +8,22 @@
 
 layout(local_size_x = 1) in;
 
-layout(std430, binding = 0) buffer Query {
-    uvec2 initial;
-    uvec2 unknown;
-    uvec2 current;
+layout(std430, binding = 0) readonly buffer Query {
+    uint data[];
 };
 
-layout(std430, binding = 1) buffer Result {
+layout(std430, binding = 1) writeonly buffer Result {
     uint result;
 };
 
+layout(push_constant) uniform PushConstants {
+    uint compare_to_zero;
+};
+
 void main() {
-    result = all(equal(initial, current)) ? 1 : 0;
+    if (compare_to_zero != 0u) {
+        result = (data[0] != 0u && data[1] != 0u) ? 1u : 0u;
+    } else {
+        result = (data[0] == data[4] && data[1] == data[5]) ? 1u : 0u;
+    }
 }

src/video_core/macro.cpp

@@ -285,11 +285,11 @@ void HLE_MultiDrawIndexedIndirectCount::Fallback(Engines::Maxwell3D& maxwell3d,
 }
 void HLE_DrawIndirectByteCount::Execute(Engines::Maxwell3D& maxwell3d, std::span<const u32> parameters, [[maybe_unused]] u32 method) {
     const bool force = maxwell3d.Rasterizer().HasDrawTransformFeedback();
-    auto topology = Maxwell3D::Regs::PrimitiveTopology(parameters[0] & 0xFFFFU);
-    if (!force && (!maxwell3d.AnyParametersDirty() || !IsTopologySafe(topology))) {
+    if (!force) {
         Fallback(maxwell3d, parameters);
         return;
     }
+    auto topology = Maxwell3D::Regs::PrimitiveTopology(parameters[0] & 0xFFFFU);
     auto& params = maxwell3d.draw_manager->GetIndirectParams();
     params.is_byte_count = true;
     params.is_indexed = false;

src/video_core/query_cache/query_cache.h

@@ -412,6 +412,7 @@ bool QueryCacheBase<Traits>::AccelerateHostConditionalRendering() {
                     .found_query = nullptr,
                 };
             }
+            it_current = it_current_2;
         }
         auto* query = impl->ObtainQuery(it_current->second);
         qc_dirty |= True(query->flags & QueryFlagBits::IsHostManaged) &&

src/video_core/query_cache/query_stream.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
@@ -75,7 +78,7 @@ public:
     }
 
     u64 GetDependentMask() const {
-        return dependence_mask;
+        return dependent_mask;
     }
 
     u64 GetAmendValue() const {

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -629,6 +629,9 @@ void RasterizerOpenGL::ReleaseFences(bool force) {
 
 void RasterizerOpenGL::FlushAndInvalidateRegion(DAddr addr, u64 size,
                                                 VideoCommon::CacheType which) {
+    if (Settings::IsGPULevelHigh()) {
+        FlushRegion(addr, size, which);
+    }
     InvalidateRegion(addr, size, which);
 }
 

src/video_core/renderer_vulkan/fixed_pipeline_state.cpp

@@ -190,9 +190,7 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFe
             }
         }
     }
-    if (!extended_dynamic_state_3_enables) {
-        dynamic_state.Refresh3(regs);
-    }
+    dynamic_state.Refresh3(regs, features);
     if (xfb_enabled) {
         RefreshXfbState(xfb_state, regs);
     }
@@ -295,16 +293,22 @@ void FixedPipelineState::DynamicState::Refresh2(const Maxwell& regs,
     depth_bias_enable.Assign(enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]] != 0 ? 1 : 0);
 }
 
-void FixedPipelineState::DynamicState::Refresh3(const Maxwell& regs) {
-    logic_op_enable.Assign(regs.logic_op.enable != 0 ? 1 : 0);
-    depth_clamp_disabled.Assign(regs.viewport_clip_control.geometry_clip ==
-                                    Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
-                                regs.viewport_clip_control.geometry_clip ==
-                                    Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
-                                regs.viewport_clip_control.geometry_clip ==
-                                    Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
-
-    line_stipple_enable.Assign(regs.line_stipple_enable);
+void FixedPipelineState::DynamicState::Refresh3(const Maxwell& regs,
+                                                const DynamicFeatures& features) {
+    if (!features.has_dynamic_state3_logic_op_enable) {
+        logic_op_enable.Assign(regs.logic_op.enable != 0 ? 1 : 0);
+    }
+    if (!features.has_dynamic_state3_depth_clamp_enable) {
+        depth_clamp_disabled.Assign(regs.viewport_clip_control.geometry_clip ==
+                                        Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
+                                    regs.viewport_clip_control.geometry_clip ==
+                                        Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
+                                    regs.viewport_clip_control.geometry_clip ==
+                                        Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
+    }
+    if (!features.has_dynamic_state3_line_stipple_enable) {
+        line_stipple_enable.Assign(regs.line_stipple_enable);
+    }
 }
 
 size_t FixedPipelineState::Hash() const noexcept {

src/video_core/renderer_vulkan/fixed_pipeline_state.h

@@ -27,6 +27,9 @@ struct DynamicFeatures {
     bool has_extended_dynamic_state_2_patch_control_points;
     bool has_extended_dynamic_state_3_blend;
     bool has_extended_dynamic_state_3_enables;
+    bool has_dynamic_state3_depth_clamp_enable;
+    bool has_dynamic_state3_logic_op_enable;
+    bool has_dynamic_state3_line_stipple_enable;
     bool has_dynamic_vertex_input;
     bool has_provoking_vertex;
     bool has_provoking_vertex_first_mode;
@@ -175,7 +178,7 @@ struct FixedPipelineState {
         void Refresh(const Maxwell& regs);
         void Refresh2(const Maxwell& regs, Maxwell::PrimitiveTopology topology,
                       bool base_features_supported);
-        void Refresh3(const Maxwell& regs);
+        void Refresh3(const Maxwell& regs, const DynamicFeatures& features);
 
         Maxwell::ComparisonOp DepthTestFunc() const noexcept {
             return UnpackComparisonOp(depth_test_func);
@@ -265,8 +268,7 @@ struct FixedPipelineState {
             return sizeof(*this);
         }
         if (dynamic_vertex_input && extended_dynamic_state_3_blend) {
-            // Exclude dynamic state and attributes
-            return offsetof(FixedPipelineState, dynamic_state);
+            return offsetof(FixedPipelineState, attachments);
         }
         if (dynamic_vertex_input) {
             // Exclude dynamic state

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -8,6 +8,7 @@
 #include <array>
 #include <cstring>
 #include <memory>
+#include <mutex>
 #include <optional>
 #include <string>
 #include <vector>
@@ -171,7 +172,11 @@ try
 
 RendererVulkan::~RendererVulkan() {
     scheduler.RegisterOnSubmit([] {});
-    void(device.GetLogical().WaitIdle());
+    scheduler.Finish();
+    {
+        std::scoped_lock lock{scheduler.submit_mutex};
+        void(device.GetLogical().WaitIdle());
+    }
 }
 
 void RendererVulkan::Composite(std::span<const Tegra::FramebufferConfig> framebuffers) {

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -8,6 +8,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <vulkan/vulkan_core.h>
+#include <mutex>
 #include "video_core/framebuffer_config.h"
 #include "video_core/present.h"
 #include "video_core/renderer_vulkan/present/filters.h"
@@ -31,7 +32,10 @@ BlitScreen::~BlitScreen() = default;
 void BlitScreen::WaitIdle() {
     present_manager.WaitPresent();
     scheduler.Finish();
-    device.GetLogical().WaitIdle();
+    {
+        std::scoped_lock lock{scheduler.submit_mutex};
+        device.GetLogical().WaitIdle();
+    }
 }
 
 void BlitScreen::SetWindowAdaptPass() {

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -637,12 +637,10 @@ void BufferCacheRuntime::BindTransformFeedbackBuffers(VideoCommon::HostBindings<
     for (u32 i = 0; i < bindings.buffers.size(); ++i) {
         auto handle = bindings.buffers[i]->Handle();
         if (handle == VK_NULL_HANDLE) {
+            ReserveNullBuffer();
+            handle = *null_buffer;
             bindings.offsets[i] = 0;
-            bindings.sizes[i] = VK_WHOLE_SIZE;
-            if (!device.HasNullDescriptor()) {
-                ReserveNullBuffer();
-                handle = *null_buffer;
-            }
+            bindings.sizes[i] = 0;
         }
         buffer_handles[i] = handle;
     }

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -228,6 +228,10 @@ struct QueriesPrefixScanPushConstants {
     u32 accumulation_limit;
     u32 buffer_offset;
 };
+
+struct ConditionalRenderingResolvePushConstants {
+    u32 compare_to_zero;
+};
 } // Anonymous namespace
 
 ComputePass::ComputePass(const Device& device_, DescriptorPool& descriptor_pool,
@@ -413,7 +417,8 @@ ConditionalRenderingResolvePass::ConditionalRenderingResolvePass(
     const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
     ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
     : ComputePass(device_, descriptor_pool_, INPUT_OUTPUT_DESCRIPTOR_SET_BINDINGS,
-                  INPUT_OUTPUT_DESCRIPTOR_UPDATE_TEMPLATE, INPUT_OUTPUT_BANK_INFO, nullptr,
+                  INPUT_OUTPUT_DESCRIPTOR_UPDATE_TEMPLATE, INPUT_OUTPUT_BANK_INFO,
+                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(ConditionalRenderingResolvePushConstants)>,
                   RESOLVE_CONDITIONAL_RENDER_COMP_SPV),
       scheduler{scheduler_}, compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
 
@@ -430,7 +435,7 @@ void ConditionalRenderingResolvePass::Resolve(VkBuffer dst_buffer, VkBuffer src_
     const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([this, descriptor_data](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([this, descriptor_data, compare_to_zero](vk::CommandBuffer cmdbuf) {
         static constexpr VkMemoryBarrier read_barrier{
             .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
             .pNext = nullptr,
@@ -443,6 +448,9 @@ void ConditionalRenderingResolvePass::Resolve(VkBuffer dst_buffer, VkBuffer src_
             .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
             .dstAccessMask = VK_ACCESS_CONDITIONAL_RENDERING_READ_BIT_EXT,
         };
+        const ConditionalRenderingResolvePushConstants uniforms{
+            .compare_to_zero = compare_to_zero ? 1U : 0U,
+        };
         const VkDescriptorSet set = descriptor_allocator.Commit();
         device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
 
@@ -450,9 +458,11 @@ void ConditionalRenderingResolvePass::Resolve(VkBuffer dst_buffer, VkBuffer src_
                                VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, read_barrier);
         cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
         cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
+        cmdbuf.PushConstants(*layout, VK_SHADER_STAGE_COMPUTE_BIT, uniforms);
         cmdbuf.Dispatch(1, 1, 1);
         cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
-                               VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, write_barrier);
+                               VK_PIPELINE_STAGE_CONDITIONAL_RENDERING_BIT_EXT, 0,
+                               write_barrier);
     });
 }
 
@@ -520,7 +530,7 @@ void QueriesPrefixScanPass::Run(VkBuffer accumulation_buffer, VkBuffer dst_buffe
             const VkDescriptorSet set = descriptor_allocator.Commit();
             device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
 
-            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
+            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
                                    VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, read_barrier);
             cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
             cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});