Add option to only optimize SPIRV during load

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

@@ -18,6 +18,7 @@ enum class IntSetting(override val key: String) : AbstractIntSetting {
     RENDERER_ANTI_ALIASING("anti_aliasing"),
     RENDERER_SCREEN_LAYOUT("screen_layout"),
     RENDERER_ASPECT_RATIO("aspect_ratio"),
+    RENDERER_OPTIMIZE_SPIRV_OUTPUT("optimize_spirv_output"),
     AUDIO_OUTPUT_ENGINE("output_engine"),
     MAX_ANISOTROPY("max_anisotropy"),
     THEME("theme"),

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

@@ -117,4 +117,15 @@ object Settings {
                 entries.firstOrNull { it.int == int } ?: Center
         }
     }
+
+    enum class OptimizeSpirvOutput(val int: Int) {
+        Never(0),
+        OnLoad(1),
+        Always(2);
+
+        companion object {
+            fun from(int: Int): OptimizeSpirvOutput =
+                entries.firstOrNull { it.int == int } ?: OnLoad
+        }
+    }
 }

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

@@ -291,6 +291,15 @@ abstract class SettingsItem(
                     descriptionId = R.string.renderer_force_max_clock_description
                 )
             )
+            put(
+                SingleChoiceSetting(
+                    IntSetting.RENDERER_OPTIMIZE_SPIRV_OUTPUT,
+                    titleId = R.string.renderer_optimize_spirv_output,
+                    descriptionId = 0,
+                    choicesId = R.array.optimizeSpirvOutputEntries,
+                    valuesId = R.array.optimizeSpirvOutputValues
+                )
+            )
             put(
                 SwitchSetting(
                     BooleanSetting.RENDERER_ASYNCHRONOUS_SHADERS,

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

@@ -177,6 +177,7 @@ class SettingsFragmentPresenter(
             add(IntSetting.RENDERER_SCREEN_LAYOUT.key)
             add(IntSetting.RENDERER_ASPECT_RATIO.key)
             add(IntSetting.VERTICAL_ALIGNMENT.key)
+            add(IntSetting.RENDERER_OPTIMIZE_SPIRV_OUTPUT.key)
             add(BooleanSetting.PICTURE_IN_PICTURE.key)
             add(BooleanSetting.RENDERER_USE_DISK_SHADER_CACHE.key)
             add(BooleanSetting.RENDERER_FORCE_MAX_CLOCK.key)

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

@@ -303,4 +303,15 @@
         <item>2</item>
     </integer-array>
 
+    <string-array name="optimizeSpirvOutputEntries">
+        <item>@string/never</item>
+        <item>@string/on_load</item>
+        <item>@string/always</item>
+    </string-array>
+    <integer-array name="optimizeSpirvOutputValues">
+        <item>0</item>
+        <item>1</item>
+        <item>2</item>
+    </integer-array>
+
 </resources>

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

@@ -234,6 +234,7 @@
     <string name="renderer_anti_aliasing">Anti-aliasing method</string>
     <string name="renderer_force_max_clock">Force maximum clocks (Adreno only)</string>
     <string name="renderer_force_max_clock_description">Forces the GPU to run at the maximum possible clocks (thermal constraints will still be applied).</string>
+    <string name="renderer_optimize_spirv_output">Optimize SPIRV output</string>
     <string name="renderer_asynchronous_shaders">Use asynchronous shaders</string>
     <string name="renderer_asynchronous_shaders_description">Compiles shaders asynchronously, reducing stutter but may introduce glitches.</string>
     <string name="renderer_reactive_flushing">Use reactive flushing</string>
@@ -662,6 +663,11 @@
     <string name="center">Center</string>
     <string name="bottom">Bottom</string>
 
+    <!-- Optimize SPIRV output -->
+    <string name="never">Never</string>
+    <string name="on_load">On Load</string>
+    <string name="always">Always</string>
+
     <!-- Licenses screen strings -->
     <string name="licenses">Licenses</string>
     <string name="license_fidelityfx_fsr" translatable="false">FidelityFX-FSR</string>

src/common/settings.cpp

@@ -52,6 +52,7 @@ SWITCHABLE(NvdecEmulation, false);
 SWITCHABLE(Region, true);
 SWITCHABLE(RendererBackend, true);
 SWITCHABLE(ScalingFilter, false);
+SWITCHABLE(SpirvOptimizeMode, true);
 SWITCHABLE(ShaderBackend, true);
 SWITCHABLE(TimeZone, true);
 SETTING(VSyncMode, true);

src/common/settings.h

@@ -73,6 +73,7 @@ SWITCHABLE(NvdecEmulation, false);
 SWITCHABLE(Region, true);
 SWITCHABLE(RendererBackend, true);
 SWITCHABLE(ScalingFilter, false);
+SWITCHABLE(SpirvOptimizeMode, true);
 SWITCHABLE(ShaderBackend, true);
 SWITCHABLE(TimeZone, true);
 SETTING(VSyncMode, true);
@@ -278,8 +279,12 @@ struct Values {
 
     SwitchableSetting<bool> use_disk_shader_cache{linkage, true, "use_disk_shader_cache",
                                                   Category::Renderer};
-    SwitchableSetting<bool> optimize_spirv_output{linkage, false, "optimize_spirv_output",
+    SwitchableSetting<SpirvOptimizeMode, true> optimize_spirv_output{linkage,
-                                                  Category::Renderer};
+                                                                     SpirvOptimizeMode::OnLoad,
+                                                                     SpirvOptimizeMode::Never,
+                                                                     SpirvOptimizeMode::Always,
+                                                                     "optimize_spirv_output",
+                                                                     Category::Renderer};
     SwitchableSetting<bool> use_asynchronous_gpu_emulation{
         linkage, true, "use_asynchronous_gpu_emulation", Category::Renderer};
     SwitchableSetting<AstcDecodeMode, true> accelerate_astc{linkage,

src/common/settings_enums.h

@@ -155,6 +155,8 @@ ENUM(ConsoleMode, Handheld, Docked);
 
 ENUM(AppletMode, HLE, LLE);
 
+ENUM(SpirvOptimizeMode, Never, OnLoad, Always)
+
 template <typename Type>
 inline std::string CanonicalizeEnum(Type id) {
     const auto group = EnumMetadata<Type>::Canonicalizations();

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -178,6 +178,7 @@ ShaderCache::ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
       state_tracker{state_tracker_}, shader_notify{shader_notify_},
       use_asynchronous_shaders{device.UseAsynchronousShaders()},
       strict_context_required{device.StrictContextRequired()},
+      optimize_spirv_output{Settings::values.optimize_spirv_output.GetValue() != Settings::SpirvOptimizeMode::Never},
       profile{
           .supported_spirv = 0x00010000,
 
@@ -344,6 +345,10 @@ void ShaderCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading,
     if (!use_asynchronous_shaders) {
         workers.reset();
     }
+
+    if (Settings::values.optimize_spirv_output.GetValue() != Settings::SpirvOptimizeMode::Always) {
+        this->optimize_spirv_output = false;
+    }
 }
 
 GraphicsPipeline* ShaderCache::CurrentGraphicsPipeline() {
@@ -538,8 +543,7 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
         case Settings::ShaderBackend::SpirV:
             ConvertLegacyToGeneric(program, runtime_info);
             sources_spirv[stage_index] =
-                EmitSPIRV(profile, runtime_info, program, binding,
+                EmitSPIRV(profile, runtime_info, program, binding, this->optimize_spirv_output);
-                          Settings::values.optimize_spirv_output.GetValue());
             break;
         }
         previous_program = &program;
@@ -598,7 +602,7 @@ std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline(
         code = EmitGLASM(profile, info, program);
         break;
     case Settings::ShaderBackend::SpirV:
-        code_spirv = EmitSPIRV(profile, program, Settings::values.optimize_spirv_output.GetValue());
+        code_spirv = EmitSPIRV(profile, program, this->optimize_spirv_output);
         break;
     }
 

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -73,6 +73,7 @@ private:
     VideoCore::ShaderNotify& shader_notify;
     const bool use_asynchronous_shaders;
     const bool strict_context_required;
+    bool optimize_spirv_output{};
 
     GraphicsPipelineKey graphics_key{};
     GraphicsPipeline* current_pipeline{};

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -310,6 +310,7 @@ PipelineCache::PipelineCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
       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()},
+      optimize_spirv_output{Settings::values.optimize_spirv_output.GetValue() != Settings::SpirvOptimizeMode::Never},
       workers(device.HasBrokenParallelShaderCompiling() ? 1ULL : GetTotalPipelineWorkers(),
               "VkPipelineBuilder"),
       serialization_thread(1, "VkPipelineSerialization") {
@@ -564,6 +565,10 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
     if (state.statistics) {
         state.statistics->Report();
     }
+
+    if (Settings::values.optimize_spirv_output.GetValue() != Settings::SpirvOptimizeMode::Always) {
+        this->optimize_spirv_output = false;
+    }
 }
 
 GraphicsPipeline* PipelineCache::CurrentGraphicsPipelineSlowPath() {
@@ -673,8 +678,7 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
 
         const auto runtime_info{MakeRuntimeInfo(programs, key, program, previous_stage)};
         ConvertLegacyToGeneric(program, runtime_info);
-        const std::vector<u32> code{EmitSPIRV(profile, runtime_info, program, binding,
+        const std::vector<u32> code{EmitSPIRV(profile, runtime_info, program, binding, this->optimize_spirv_output)};
-                                              Settings::values.optimize_spirv_output.GetValue())};
         device.SaveShader(code);
         modules[stage_index] = BuildShader(device, code);
         if (device.HasDebuggingToolAttached()) {
@@ -768,8 +772,7 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
     }
 
     auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
-    const std::vector<u32> code{
+    const std::vector<u32> code{EmitSPIRV(profile, program, this->optimize_spirv_output)};
-        EmitSPIRV(profile, program, Settings::values.optimize_spirv_output.GetValue())};
     device.SaveShader(code);
     vk::ShaderModule spv_module{BuildShader(device, code)};
     if (device.HasDebuggingToolAttached()) {

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -150,6 +150,7 @@ private:
     VideoCore::ShaderNotify& shader_notify;
     bool use_asynchronous_shaders{};
     bool use_vulkan_pipeline_cache{};
+    bool optimize_spirv_output{};
 
     GraphicsPipelineCacheKey graphics_key{};
     GraphicsPipeline* current_pipeline{};

src/yuzu/configuration/shared_translation.cpp

@@ -319,6 +319,12 @@ std::unique_ptr<ComboboxTranslationMap> ComboboxEnumeration(QWidget* parent) {
                               PAIR(AppletMode, LLE, tr("Real applet")),
                           }});
 
+    translations->insert({Settings::EnumMetadata<Settings::SpirvOptimizeMode>::Index(),
+                          {
+                              PAIR(SpirvOptimizeMode, Never, tr("Never")),
+                              PAIR(SpirvOptimizeMode, OnLoad, tr("On Load")),
+                              PAIR(SpirvOptimizeMode, Always, tr("Always")),
+                          }});
     translations->insert({Settings::EnumMetadata<Settings::AstcDecodeMode>::Index(),
                           {
                               PAIR(AstcDecodeMode, Cpu, tr("CPU")),

src/yuzu/uisettings.h

@@ -277,3 +277,4 @@ Q_DECLARE_METATYPE(Settings::RendererBackend);
 Q_DECLARE_METATYPE(Settings::ShaderBackend);
 Q_DECLARE_METATYPE(Settings::AstcRecompression);
 Q_DECLARE_METATYPE(Settings::AstcDecodeMode);
+Q_DECLARE_METATYPE(Settings::SpirvOptimizeMode);