early-access version 3982

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 3981.
+This is the source code for early-access 3982.
 
 ## Legal Notice
 

src/frontend_common/config.cpp

@@ -856,7 +856,15 @@ std::string Config::AdjustKey(const std::string& key) {
 std::string Config::AdjustOutputString(const std::string& string) {
     std::string adjusted_string(string);
     boost::replace_all(adjusted_string, "\\", "/");
+
+    // Windows requires that two forward slashes are used at the start of a path for unmapped
+    // network drives so we have to watch for that here
+    if (string.substr(0, 2) == "//") {
         boost::replace_all(adjusted_string, "//", "/");
+        adjusted_string.insert(0, "/");
+    } else {
+        boost::replace_all(adjusted_string, "//", "/");
+    }
 
     // Needed for backwards compatibility with QSettings deserialization
     for (const auto& special_character : special_characters) {

src/shader_recompiler/backend/spirv/emit_spirv.cpp

@@ -407,7 +407,7 @@ void SetupCapabilities(const Profile& profile, const Info& info, EmitContext& ct
         }
         ctx.AddCapability(spv::Capability::DemoteToHelperInvocation);
     }
-    if (info.stores[IR::Attribute::ViewportIndex]) {
+    if (info.stores[IR::Attribute::ViewportIndex] && profile.support_multi_viewport) {
         ctx.AddCapability(spv::Capability::MultiViewport);
     }
     if (info.stores[IR::Attribute::ViewportMask] && profile.support_viewport_mask) {

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -84,6 +84,10 @@ std::optional<OutAttr> OutputAttrPointer(EmitContext& ctx, IR::Attribute attr) {
         }
         return std::nullopt;
     case IR::Attribute::ViewportIndex:
+        if (!ctx.profile.support_multi_viewport) {
+            LOG_WARNING(Shader, "Ignoring viewport index store on non-supporting driver");
+            return std::nullopt;
+        }
         if (ctx.profile.support_viewport_index_layer_non_geometry ||
             ctx.stage == Shader::Stage::Geometry) {
             return OutAttr{ctx.viewport_index, ctx.U32[1]};

src/shader_recompiler/profile.h

@@ -43,6 +43,7 @@ struct Profile {
     bool support_gl_sparse_textures{};
     bool support_gl_derivative_control{};
     bool support_scaled_attributes{};
+    bool support_multi_viewport{};
 
     bool warp_size_potentially_larger_than_guest{};
 

src/video_core/buffer_cache/buffer_cache.h

@@ -1209,11 +1209,6 @@ void BufferCache<P>::UpdateDrawIndirect() {
             .size = static_cast<u32>(size),
             .buffer_id = FindBuffer(*cpu_addr, static_cast<u32>(size)),
         };
-        VAddr cpu_addr_start = Common::AlignDown(*cpu_addr, 64);
-        VAddr cpu_addr_end = Common::AlignUp(*cpu_addr + size, 64);
-        IntervalType interval{cpu_addr_start, cpu_addr_end};
-        ClearDownload(interval);
-        common_ranges.subtract(interval);
     };
     if (current_draw_indirect->include_count) {
         update(current_draw_indirect->count_start_address, sizeof(u32),

src/video_core/engines/maxwell_3d.cpp

@@ -268,7 +268,7 @@ size_t Maxwell3D::EstimateIndexBufferSize() {
                                                         std::numeric_limits<u32>::max()};
     const size_t byte_size = regs.index_buffer.FormatSizeInBytes();
     const size_t log2_byte_size = Common::Log2Ceil64(byte_size);
-    const size_t cap{GetMaxCurrentVertices() * 3 * byte_size};
+    const size_t cap{GetMaxCurrentVertices() * 4 * byte_size};
     const size_t lower_cap =
         std::min<size_t>(static_cast<size_t>(end_address - start_address), cap);
     return std::min<size_t>(

src/video_core/renderer_opengl/gl_graphics_pipeline.cpp

@@ -559,7 +559,9 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
 }
 
 void GraphicsPipeline::ConfigureTransformFeedbackImpl() const {
-    glTransformFeedbackAttribsNV(num_xfb_attribs, xfb_attribs.data(), GL_SEPARATE_ATTRIBS);
+    const GLenum buffer_mode =
+        num_xfb_buffers_active == 1 ? GL_INTERLEAVED_ATTRIBS : GL_SEPARATE_ATTRIBS;
+    glTransformFeedbackAttribsNV(num_xfb_attribs, xfb_attribs.data(), buffer_mode);
 }
 
 void GraphicsPipeline::GenerateTransformFeedbackState() {
@@ -567,12 +569,14 @@ void GraphicsPipeline::GenerateTransformFeedbackState() {
     // when this is required.
     GLint* cursor{xfb_attribs.data()};
 
+    num_xfb_buffers_active = 0;
     for (size_t feedback = 0; feedback < Maxwell::NumTransformFeedbackBuffers; ++feedback) {
         const auto& layout = key.xfb_state.layouts[feedback];
         UNIMPLEMENTED_IF_MSG(layout.stride != layout.varying_count * 4, "Stride padding");
         if (layout.varying_count == 0) {
             continue;
         }
+        num_xfb_buffers_active++;
 
         const auto& locations = key.xfb_state.varyings[feedback];
         std::optional<u32> current_index;

src/video_core/renderer_opengl/gl_graphics_pipeline.h

@@ -154,6 +154,7 @@ private:
 
     static constexpr std::size_t XFB_ENTRY_STRIDE = 3;
     GLsizei num_xfb_attribs{};
+    u32 num_xfb_buffers_active{};
     std::array<GLint, 128 * XFB_ENTRY_STRIDE * Maxwell::NumTransformFeedbackBuffers> xfb_attribs{};
 
     std::mutex built_mutex;

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -263,6 +263,22 @@ Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> program
     info.y_negate = key.state.y_negate != 0;
     return info;
 }
+
+size_t GetTotalPipelineWorkers() {
+    const size_t max_core_threads =
+        std::max<size_t>(static_cast<size_t>(std::thread::hardware_concurrency()), 2ULL) - 1ULL;
+#ifdef ANDROID
+    // Leave at least a few cores free in android
+    constexpr size_t free_cores = 3ULL;
+    if (max_core_threads <= free_cores) {
+        return 1ULL;
+    }
+    return max_core_threads - free_cores;
+#else
+    return max_core_threads;
+#endif
+}
+
 } // Anonymous namespace
 
 size_t ComputePipelineCacheKey::Hash() const noexcept {
@@ -294,11 +310,8 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
       texture_cache{texture_cache_}, shader_notify{shader_notify_},
       use_asynchronous_shaders{Settings::values.use_asynchronous_shaders.GetValue()},
       use_vulkan_pipeline_cache{Settings::values.use_vulkan_driver_pipeline_cache.GetValue()},
-#ifdef ANDROID
+      workers(device.HasBrokenParallelShaderCompiling() ? 1ULL : GetTotalPipelineWorkers(),
-      workers(1, "VkPipelineBuilder"),
+              "VkPipelineBuilder"),
-#else
-      workers(std::max(std::thread::hardware_concurrency(), 2U) - 1, "VkPipelineBuilder"),
-#endif
       serialization_thread(1, "VkPipelineSerialization") {
     const auto& float_control{device.FloatControlProperties()};
     const VkDriverId driver_id{device.GetDriverID()};
@@ -338,6 +351,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .support_geometry_shader_passthrough = device.IsNvGeometryShaderPassthroughSupported(),
         .support_native_ndc = device.IsExtDepthClipControlSupported(),
         .support_scaled_attributes = !device.MustEmulateScaledFormats(),
+        .support_multi_viewport = device.SupportsMultiViewport(),
 
         .warp_size_potentially_larger_than_guest = device.IsWarpSizePotentiallyBiggerThanGuest(),
 

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -211,6 +211,13 @@ public:
             return;
         }
         PauseCounter();
+        const auto driver_id = device.GetDriverID();
+        if (driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY ||
+            driver_id == VK_DRIVER_ID_ARM_PROPRIETARY || driver_id == VK_DRIVER_ID_MESA_TURNIP) {
+            pending_sync.clear();
+            sync_values_stash.clear();
+            return;
+        }
         sync_values_stash.clear();
         sync_values_stash.emplace_back();
         std::vector<HostSyncValues>* sync_values = &sync_values_stash.back();
@@ -1378,6 +1385,12 @@ bool QueryCacheRuntime::HostConditionalRenderingCompareValues(VideoCommon::Looku
         return true;
     }
 
+    auto driver_id = impl->device.GetDriverID();
+    if (driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY ||
+        driver_id == VK_DRIVER_ID_ARM_PROPRIETARY || driver_id == VK_DRIVER_ID_MESA_TURNIP) {
+        return true;
+    }
+
     for (size_t i = 0; i < 2; i++) {
         is_null[i] = !is_in_ac[i] && check_value(objects[i]->address);
     }

src/video_core/vulkan_common/vulkan_device.cpp

@@ -635,6 +635,12 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
             has_broken_cube_compatibility = true;
         }
     }
+    if (is_qualcomm) {
+        const u32 version = (properties.properties.driverVersion << 3) >> 3;
+        if (version < VK_MAKE_API_VERSION(0, 255, 615, 512)) {
+            has_broken_parallel_compiling = true;
+        }
+    }
     if (extensions.sampler_filter_minmax && is_amd) {
         // Disable ext_sampler_filter_minmax on AMD GCN4 and lower as it is broken.
         if (!features.shader_float16_int8.shaderFloat16) {

src/video_core/vulkan_common/vulkan_device.h

@@ -102,6 +102,7 @@ VK_DEFINE_HANDLE(VmaAllocator)
     EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME)                                   \
     EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME)                                 \
     EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME)                                 \
+    EXTENSION_NAME(VK_EXT_EXTERNAL_MEMORY_HOST_EXTENSION_NAME)                                     \
     EXTENSION_NAME(VK_EXT_4444_FORMATS_EXTENSION_NAME)                                             \
     EXTENSION_NAME(VK_EXT_LINE_RASTERIZATION_EXTENSION_NAME)                                       \
     EXTENSION_NAME(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME)                                             \
@@ -599,6 +600,11 @@ public:
         return has_broken_cube_compatibility;
     }
 
+    /// Returns true if parallel shader compiling has issues with the current driver.
+    bool HasBrokenParallelShaderCompiling() const {
+        return has_broken_parallel_compiling;
+    }
+
     /// Returns the vendor name reported from Vulkan.
     std::string_view GetVendorName() const {
         return properties.driver.driverName;
@@ -663,6 +669,10 @@ public:
         return supports_conditional_barriers;
     }
 
+    bool SupportsMultiViewport() const {
+        return features2.features.multiViewport;
+    }
+
     [[nodiscard]] static constexpr bool CheckBrokenCompute(VkDriverId driver_id,
                                                            u32 driver_version) {
         if (driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS) {
@@ -794,6 +804,7 @@ private:
     bool is_non_gpu{};                         ///< Is SoftwareRasterizer, FPGA, non-GPU device.
     bool has_broken_compute{};                 ///< Compute shaders can cause crashes
     bool has_broken_cube_compatibility{};      ///< Has broken cube compatibility bit
+    bool has_broken_parallel_compiling{};      ///< Has broken parallel shader compiling.
     bool has_renderdoc{};                      ///< Has RenderDoc attached
     bool has_nsight_graphics{};                ///< Has Nsight Graphics attached
     bool supports_d24_depth{};                 ///< Supports D24 depth buffers.

src/yuzu/util/util.cpp

@@ -4,7 +4,10 @@
 #include <array>
 #include <cmath>
 #include <QPainter>
+
+#include "common/logging/log.h"
 #include "yuzu/util/util.h"
+
 #ifdef _WIN32
 #include <windows.h>
 #include "common/fs/file.h"