Revert

- [x] I have read and followed the [Contribution Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/CONTRIBUTING.md#code-contributions).
- [x] I have read and followed the [AI Policy](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/AI.md)
- [x] I have read and followed the [Coding Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/Coding.md) to the best of my ability.

-------------------

This one reverts part of the PR 4319 which was aimed to allow child titles to inherit bundle settings.
The fix seems good so far, but by applying it to save folders caused a trap:
Now child's save folders move to bundle folder, hence saves will vanish, and the android/qt save tools are still pointing to old folders.
Since the save part was only a compliance fix, instead of fiddle in master i'd rather revert this part now, then only touch this again if any bug show up.
No worries. I'll be right here waiting.

Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4462
Reviewed-by: lizzie <lizzie@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>

CMakeLists.txt

@@ -82,8 +82,8 @@ cmake_dependent_option(YUZU_USE_BUNDLED_QT "Download bundled Qt binaries" "${MSV
 option(ENABLE_DEBUG_TOOLS "Enable debugging tools (maxwell disassembler, SPIRV translator, etc)" OFF)
 option(ENABLE_WERROR "Enable -Werror diagnostics" ON)
 
-# Lossless Scaling frame generation.
-option(ENABLE_LSFG "Enable Lossless Scaling frame generation" ON)
+# Lossless Scaling frame generation. Only Android.
+cmake_dependent_option(ENABLE_LSFG "Enable Lossless Scaling frame generation" ON "ANDROID" OFF)
 
 option(ENABLE_RESHADE "Enable ReShade FX post-processing effects" ON)
 

src/android/app/src/main/java/org/yuzu/yuzu_emu/adapters/GameAdapter.kt

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 package org.yuzu.yuzu_emu.adapters

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

@@ -20,21 +20,18 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
     EMULATE_BGR565("emulate_bgr565"),
     RESCALE_HACK("rescale_hack"),
     CPUOPT_UNSAFE_HOST_MMU("cpuopt_unsafe_host_mmu"),
-    RELOCATE_36BIT_ADDRESS_SPACE("relocate_36bit_address_space"),
     USE_DOCKED_MODE("use_docked_mode"),
     USE_AUTO_STUB("use_auto_stub"),
     RENDERER_USE_DISK_SHADER_CACHE("use_disk_shader_cache"),
     RENDERER_FORCE_MAX_CLOCK("force_max_clock"),
     RENDERER_ASYNCHRONOUS_GPU_EMULATION("use_asynchronous_gpu_emulation"),
+    RENDERER_ASYNC_PRESENTATION("async_presentation"),
     RENDERER_ASYNCHRONOUS_SHADERS("use_asynchronous_shaders"),
     RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
-    RENDERER_BARRIER_FEEDBACK_LOOPS("barrier_feedback_loops"),
     ENABLE_BUFFER_HISTORY("enable_buffer_history"),
     USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"),
     ENABLE_GPU_BUFFER_READBACK("enable_gpu_buffer_readback"),
-    USE_UNIFIED_MEMORY("use_unified_memory"),
     SYNC_MEMORY_OPERATIONS("sync_memory_operations"),
-    STALL_ON_GPU_FENCE_WAIT("stall_on_gpu_fence_wait"),
     BUFFER_REORDER_DISABLE("disable_buffer_reorder"),
     RENDERER_DEBUG("debug"),
     RENDERER_PATCH_OLD_QCOM_DRIVERS("patch_old_qcom_drivers"),
@@ -42,6 +39,7 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
     RENDERER_SAMPLE_SHADING("sample_shading"),
     RENDERER_FRAME_GEN("frame_gen"),
     RENDERER_FRAME_GEN_FLOW_SCALE_AUTO("frame_gen_flow_scale_auto"),
+    GPU_UNSWIZZLE_ENABLED("gpu_unswizzle_enabled"),
     PICTURE_IN_PICTURE("picture_in_picture"),
     USE_CUSTOM_RTC("custom_rtc_enabled"),
     BLACK_BACKGROUNDS("black_backgrounds"),

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

@@ -32,6 +32,7 @@ enum class IntSetting(override val key: String) : AbstractIntSetting {
     RENDERER_DYNA_STATE("dyna_state"),
     DMA_ACCURACY("dma_accuracy"),
     GPU_FENCE_BEHAVIOR("gpu_fence_behavior"),
+    FRAME_PACING_MODE("frame_pacing_mode"),
     AUDIO_OUTPUT_ENGINE("output_engine"),
     MAX_ANISOTROPY("max_anisotropy"),
     THEME("theme"),
@@ -50,6 +51,9 @@ enum class IntSetting(override val key: String) : AbstractIntSetting {
     FAST_CPU_TIME("fast_cpu_time"),
     CPU_TICKS("cpu_ticks"),
     FAST_GPU_TIME("fast_gpu_time"),
+    GPU_UNSWIZZLE_TEXTURE_SIZE("gpu_unswizzle_texture_size"),
+    GPU_UNSWIZZLE_STREAM_SIZE("gpu_unswizzle_stream_size"),
+    GPU_UNSWIZZLE_CHUNK_SIZE("gpu_unswizzle_chunk_size"),
     BAT_TEMPERATURE_UNIT("bat_temperature_unit"),
     CABINET_APPLET("cabinet_applet_mode"),
     CONTROLLER_APPLET("controller_applet_mode"),

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

@@ -0,0 +1,89 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+package org.yuzu.yuzu_emu.features.settings.model.view
+
+import androidx.annotation.ArrayRes
+import androidx.annotation.StringRes
+import org.yuzu.yuzu_emu.features.settings.model.AbstractSetting
+import org.yuzu.yuzu_emu.features.settings.model.BooleanSetting
+import org.yuzu.yuzu_emu.features.settings.model.IntSetting
+
+class GpuUnswizzleSetting(
+    @StringRes titleId: Int = 0,
+    titleString: String = "",
+    @StringRes descriptionId: Int = 0,
+    descriptionString: String = "",
+    @ArrayRes val textureSizeChoicesId: Int,
+    @ArrayRes val textureSizeValuesId: Int,
+    @ArrayRes val streamSizeChoicesId: Int,
+    @ArrayRes val streamSizeValuesId: Int,
+    @ArrayRes val chunkSizeChoicesId: Int,
+    @ArrayRes val chunkSizeValuesId: Int
+) : SettingsItem(
+    object : AbstractSetting {
+        override val key: String = SettingsItem.GPU_UNSWIZZLE_COMBINED
+        override val defaultValue: Any = false
+        override val isSaveable = true
+        override val isRuntimeModifiable = true
+        override val isSwitchable = true
+        override val pairedSettingKey: String = ""
+        override var global: Boolean
+            get() {
+                return BooleanSetting.GPU_UNSWIZZLE_ENABLED.global &&
+                    IntSetting.GPU_UNSWIZZLE_TEXTURE_SIZE.global &&
+                    IntSetting.GPU_UNSWIZZLE_STREAM_SIZE.global &&
+                    IntSetting.GPU_UNSWIZZLE_CHUNK_SIZE.global
+            }
+            set(value) {
+                BooleanSetting.GPU_UNSWIZZLE_ENABLED.global = value
+                IntSetting.GPU_UNSWIZZLE_TEXTURE_SIZE.global = value
+                IntSetting.GPU_UNSWIZZLE_STREAM_SIZE.global = value
+                IntSetting.GPU_UNSWIZZLE_CHUNK_SIZE.global = value
+            }
+        override fun getValueAsString(needsGlobal: Boolean): String = "combined"
+        override fun reset() {
+            BooleanSetting.GPU_UNSWIZZLE_ENABLED.reset()
+            IntSetting.GPU_UNSWIZZLE_TEXTURE_SIZE.reset()
+            IntSetting.GPU_UNSWIZZLE_STREAM_SIZE.reset()
+            IntSetting.GPU_UNSWIZZLE_CHUNK_SIZE.reset()
+        }
+    },
+    titleId,
+    titleString,
+    descriptionId,
+    descriptionString
+) {
+    override val type = SettingsItem.TYPE_GPU_UNSWIZZLE
+
+    // Check if GPU unswizzle is enabled via the dedicated boolean setting
+    fun isEnabled(needsGlobal: Boolean = false): Boolean =
+        BooleanSetting.GPU_UNSWIZZLE_ENABLED.getBoolean(needsGlobal)
+
+    fun setEnabled(value: Boolean) =
+        BooleanSetting.GPU_UNSWIZZLE_ENABLED.setBoolean(value)
+
+    fun enable() = setEnabled(true)
+
+    fun disable() = setEnabled(false)
+
+    fun getTextureSize(needsGlobal: Boolean = false): Int =
+        IntSetting.GPU_UNSWIZZLE_TEXTURE_SIZE.getInt(needsGlobal)
+
+    fun setTextureSize(value: Int) =
+        IntSetting.GPU_UNSWIZZLE_TEXTURE_SIZE.setInt(value)
+
+    fun getStreamSize(needsGlobal: Boolean = false): Int =
+        IntSetting.GPU_UNSWIZZLE_STREAM_SIZE.getInt(needsGlobal)
+
+    fun setStreamSize(value: Int) =
+        IntSetting.GPU_UNSWIZZLE_STREAM_SIZE.setInt(value)
+
+    fun getChunkSize(needsGlobal: Boolean = false): Int =
+        IntSetting.GPU_UNSWIZZLE_CHUNK_SIZE.getInt(needsGlobal)
+
+    fun setChunkSize(value: Int) =
+        IntSetting.GPU_UNSWIZZLE_CHUNK_SIZE.setInt(value)
+
+    fun reset() = setting.reset()
+}

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

@@ -141,12 +141,14 @@ abstract class SettingsItem(
         const val TYPE_SPINBOX = 12
         const val TYPE_LAUNCHABLE = 13
         const val TYPE_PATH = 14
+        const val TYPE_GPU_UNSWIZZLE = 15
         const val TYPE_FX_TOOLBAR = 16
         const val TYPE_FX_PRESET = 17
         const val TYPE_FX_SHADER = 18
         const val TYPE_FX_BUTTON = 19
 
         const val FASTMEM_COMBINED = "fastmem_combined"
+        const val GPU_UNSWIZZLE_COMBINED = "gpu_unswizzle_combined"
 
         val emptySetting = object : AbstractSetting {
             override val key: String = ""
@@ -751,6 +753,13 @@ abstract class SettingsItem(
                     descriptionId = R.string.renderer_asynchronous_gpu_emulation_description
                 )
             )
+            put(
+                SwitchSetting(
+                    BooleanSetting.RENDERER_ASYNC_PRESENTATION,
+                    titleId = R.string.renderer_async_presentation,
+                    descriptionId = R.string.renderer_async_presentation_description
+                )
+            )
             put(
                 SingleChoiceSetting(
                     IntSetting.DMA_ACCURACY,
@@ -785,6 +794,45 @@ abstract class SettingsItem(
                     valuesId = R.array.gpuValues
                 )
             )
+            put(
+                SingleChoiceSetting(
+                    IntSetting.GPU_UNSWIZZLE_TEXTURE_SIZE,
+                    titleId = R.string.gpu_unswizzle_texture_size,
+                    descriptionId = R.string.gpu_unswizzle_texture_size_description,
+                    choicesId = R.array.gpuTextureSizeSwizzleEntries,
+                    valuesId = R.array.gpuTextureSizeSwizzleValues
+                )
+            )
+            put(
+                SingleChoiceSetting(
+                    IntSetting.GPU_UNSWIZZLE_STREAM_SIZE,
+                    titleId = R.string.gpu_unswizzle_stream_size,
+                    descriptionId = R.string.gpu_unswizzle_stream_size_description,
+                    choicesId = R.array.gpuSwizzleEntries,
+                    valuesId = R.array.gpuSwizzleValues
+                )
+            )
+            put(
+                SingleChoiceSetting(
+                    IntSetting.GPU_UNSWIZZLE_CHUNK_SIZE,
+                    titleId = R.string.gpu_unswizzle_chunk_size,
+                    descriptionId = R.string.gpu_unswizzle_chunk_size_description,
+                    choicesId = R.array.gpuSwizzleChunkEntries,
+                    valuesId = R.array.gpuSwizzleChunkValues
+                )
+            )
+            put(
+                GpuUnswizzleSetting(
+                    titleId = R.string.gpu_unswizzle_settings,
+                    descriptionId = R.string.gpu_unswizzle_settings_description,
+                    textureSizeChoicesId = R.array.gpuTextureSizeSwizzleEntries,
+                    textureSizeValuesId = R.array.gpuTextureSizeSwizzleValues,
+                    streamSizeChoicesId = R.array.gpuSwizzleEntries,
+                    streamSizeValuesId = R.array.gpuSwizzleValues,
+                    chunkSizeChoicesId = R.array.gpuSwizzleChunkEntries,
+                    chunkSizeValuesId = R.array.gpuSwizzleChunkValues
+                )
+            )
             put(
                 SingleChoiceSetting(
                     IntSetting.FAST_CPU_TIME,
@@ -846,13 +894,6 @@ abstract class SettingsItem(
                     descriptionId = R.string.cpuopt_unsafe_host_mmu_description
                 )
             )
-            put(
-                SwitchSetting(
-                    BooleanSetting.RELOCATE_36BIT_ADDRESS_SPACE,
-                    titleId = R.string.relocate_36bit_address_space,
-                    descriptionId = R.string.relocate_36bit_address_space_description
-                )
-            )
             put(
                 SwitchSetting(
                     BooleanSetting.RENDERER_REACTIVE_FLUSHING,
@@ -860,13 +901,6 @@ abstract class SettingsItem(
                     descriptionId = R.string.renderer_reactive_flushing_description
                 )
             )
-            put(
-                SwitchSetting(
-                    BooleanSetting.RENDERER_BARRIER_FEEDBACK_LOOPS,
-                    titleId = R.string.renderer_barrier_feedback_loops,
-                    descriptionId = R.string.renderer_barrier_feedback_loops_description
-                )
-            )
             put(
                 SwitchSetting(
                     BooleanSetting.ENABLE_BUFFER_HISTORY,
@@ -881,13 +915,6 @@ abstract class SettingsItem(
                     descriptionId = R.string.enable_gpu_buffer_readback_description
                 )
             )
-            put(
-                SwitchSetting(
-                    BooleanSetting.USE_UNIFIED_MEMORY,
-                    titleId = R.string.use_unified_memory,
-                    descriptionId = R.string.use_unified_memory_description
-                )
-            )
             put(
                 SwitchSetting(
                     BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS,
@@ -902,13 +929,6 @@ abstract class SettingsItem(
                     descriptionId = R.string.sync_memory_operations_description
                 )
             )
-            put(
-                SwitchSetting(
-                    BooleanSetting.STALL_ON_GPU_FENCE_WAIT,
-                    titleId = R.string.stall_on_gpu_fence_wait,
-                    descriptionId = R.string.stall_on_gpu_fence_wait_description
-                )
-            )
             put(
                 SwitchSetting(
                     BooleanSetting.BUFFER_REORDER_DISABLE,

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

@@ -0,0 +1,206 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+package org.yuzu.yuzu_emu.features.settings.ui
+
+import android.app.Dialog
+import android.content.DialogInterface
+import android.os.Bundle
+import android.view.LayoutInflater
+import android.widget.ArrayAdapter
+import androidx.fragment.app.DialogFragment
+import androidx.fragment.app.activityViewModels
+import com.google.android.material.dialog.MaterialAlertDialogBuilder
+import org.yuzu.yuzu_emu.R
+import org.yuzu.yuzu_emu.databinding.DialogGpuUnswizzleBinding
+import org.yuzu.yuzu_emu.features.settings.model.view.GpuUnswizzleSetting
+
+class GpuUnswizzleDialogFragment : DialogFragment() {
+    private var position = 0
+    private val settingsViewModel: SettingsViewModel by activityViewModels()
+    private lateinit var binding: DialogGpuUnswizzleBinding
+
+    override fun onCreate(savedInstanceState: Bundle?) {
+        super.onCreate(savedInstanceState)
+        position = requireArguments().getInt(POSITION)
+
+        if (settingsViewModel.clickedItem == null) dismiss()
+    }
+
+    override fun onCreateDialog(savedInstanceState: Bundle?): Dialog {
+        binding = DialogGpuUnswizzleBinding.inflate(LayoutInflater.from(requireContext()))
+        val item = settingsViewModel.clickedItem as GpuUnswizzleSetting
+
+        // Setup texture size dropdown
+        val textureSizeEntries = resources.getStringArray(item.textureSizeChoicesId)
+        val textureSizeValues = resources.getIntArray(item.textureSizeValuesId)
+        val textureSizeAdapter = ArrayAdapter(
+            requireContext(),
+            android.R.layout.simple_dropdown_item_1line,
+            textureSizeEntries.toMutableList()
+        )
+        binding.dropdownTextureSize.setAdapter(textureSizeAdapter)
+
+        // Setup stream size dropdown
+        val streamSizeEntries = resources.getStringArray(item.streamSizeChoicesId)
+        val streamSizeValues = resources.getIntArray(item.streamSizeValuesId)
+        val streamSizeAdapter = ArrayAdapter(
+            requireContext(),
+            android.R.layout.simple_dropdown_item_1line,
+            streamSizeEntries.toMutableList()
+        )
+        binding.dropdownStreamSize.setAdapter(streamSizeAdapter)
+
+        // Setup chunk size dropdown
+        val chunkSizeEntries = resources.getStringArray(item.chunkSizeChoicesId)
+        val chunkSizeValues = resources.getIntArray(item.chunkSizeValuesId)
+        val chunkSizeAdapter = ArrayAdapter(
+            requireContext(),
+            android.R.layout.simple_dropdown_item_1line,
+            chunkSizeEntries.toMutableList()
+        )
+        binding.dropdownChunkSize.setAdapter(chunkSizeAdapter)
+
+        // Load current values
+        val isEnabled = item.isEnabled()
+        binding.switchEnable.isChecked = isEnabled
+
+        if (isEnabled) {
+            val textureSizeIndex = textureSizeValues.indexOf(item.getTextureSize())
+            if (textureSizeIndex >= 0) {
+                binding.dropdownTextureSize.setText(textureSizeEntries[textureSizeIndex], false)
+            }
+
+            val streamSizeIndex = streamSizeValues.indexOf(item.getStreamSize())
+            if (streamSizeIndex >= 0) {
+                binding.dropdownStreamSize.setText(streamSizeEntries[streamSizeIndex], false)
+            }
+
+            val chunkSizeIndex = chunkSizeValues.indexOf(item.getChunkSize())
+            if (chunkSizeIndex >= 0) {
+                binding.dropdownChunkSize.setText(chunkSizeEntries[chunkSizeIndex], false)
+            }
+        } else {
+            // Set default/recommended values when disabling
+            binding.dropdownTextureSize.setText(textureSizeEntries[3], false)
+            binding.dropdownStreamSize.setText(streamSizeEntries[3], false)
+            binding.dropdownChunkSize.setText(chunkSizeEntries[3], false)
+        }
+
+        // Clear adapter filters after setText to fix rotation bug
+        textureSizeAdapter.filter.filter(null)
+        streamSizeAdapter.filter.filter(null)
+        chunkSizeAdapter.filter.filter(null)
+
+        // Enable/disable dropdowns based on switch state
+        updateDropdownsState(isEnabled)
+        binding.switchEnable.setOnCheckedChangeListener { _, checked ->
+            updateDropdownsState(checked)
+        }
+
+        val dialog = MaterialAlertDialogBuilder(requireContext())
+            .setTitle(item.title)
+            .setView(binding.root)
+            .create()
+
+        // Setup button listeners
+        binding.btnDefault.setOnClickListener {
+            // Reset to defaults
+            item.reset()
+            // Refresh values with adapters reset
+            val textureSizeIndex = textureSizeValues.indexOf(item.getTextureSize())
+            if (textureSizeIndex >= 0) {
+                binding.dropdownTextureSize.setText(textureSizeEntries[textureSizeIndex], false)
+            }
+            val streamSizeIndex = streamSizeValues.indexOf(item.getStreamSize())
+            if (streamSizeIndex >= 0) {
+                binding.dropdownStreamSize.setText(streamSizeEntries[streamSizeIndex], false)
+            }
+            val chunkSizeIndex = chunkSizeValues.indexOf(item.getChunkSize())
+            if (chunkSizeIndex >= 0) {
+                binding.dropdownChunkSize.setText(chunkSizeEntries[chunkSizeIndex], false)
+            }
+            // Clear filters
+            textureSizeAdapter.filter.filter(null)
+            streamSizeAdapter.filter.filter(null)
+            chunkSizeAdapter.filter.filter(null)
+
+            settingsViewModel.setAdapterItemChanged(position)
+            settingsViewModel.setShouldReloadSettingsList(true)
+        }
+
+        binding.btnCancel.setOnClickListener {
+            dialog.dismiss()
+        }
+
+        binding.btnOk.setOnClickListener {
+            if (binding.switchEnable.isChecked) {
+                item.enable()
+                // Save the selected values
+                val selectedTextureIndex = textureSizeEntries.indexOf(
+                    binding.dropdownTextureSize.text.toString()
+                )
+                if (selectedTextureIndex >= 0) {
+                    item.setTextureSize(textureSizeValues[selectedTextureIndex])
+                }
+
+                val selectedStreamIndex = streamSizeEntries.indexOf(
+                    binding.dropdownStreamSize.text.toString()
+                )
+                if (selectedStreamIndex >= 0) {
+                    item.setStreamSize(streamSizeValues[selectedStreamIndex])
+                }
+
+                val selectedChunkIndex = chunkSizeEntries.indexOf(
+                    binding.dropdownChunkSize.text.toString()
+                )
+                if (selectedChunkIndex >= 0) {
+                    item.setChunkSize(chunkSizeValues[selectedChunkIndex])
+                }
+            } else {
+                // Disable GPU unswizzle
+                item.disable()
+            }
+
+            settingsViewModel.setAdapterItemChanged(position)
+            settingsViewModel.setShouldReloadSettingsList(true)
+            dialog.dismiss()
+        }
+
+        // Ensure filters are cleared after dialog is shown
+        binding.root.post {
+            textureSizeAdapter.filter.filter(null)
+            streamSizeAdapter.filter.filter(null)
+            chunkSizeAdapter.filter.filter(null)
+        }
+
+        return dialog
+    }
+
+    private fun updateDropdownsState(enabled: Boolean) {
+        binding.layoutTextureSize.isEnabled = enabled
+        binding.dropdownTextureSize.isEnabled = enabled
+        binding.layoutStreamSize.isEnabled = enabled
+        binding.dropdownStreamSize.isEnabled = enabled
+        binding.layoutChunkSize.isEnabled = enabled
+        binding.dropdownChunkSize.isEnabled = enabled
+    }
+
+    companion object {
+        const val TAG = "GpuUnswizzleDialogFragment"
+        const val POSITION = "Position"
+
+        fun newInstance(
+            settingsViewModel: SettingsViewModel,
+            item: GpuUnswizzleSetting,
+            position: Int
+        ): GpuUnswizzleDialogFragment {
+            val dialog = GpuUnswizzleDialogFragment()
+            val args = Bundle()
+            args.putInt(POSITION, position)
+            dialog.arguments = args
+            settingsViewModel.clickedItem = item
+            return dialog
+        }
+    }
+}

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

@@ -106,6 +106,10 @@ class SettingsAdapter(
                 PathViewHolder(ListItemSettingBinding.inflate(inflater), this)
             }
 
+            SettingsItem.TYPE_GPU_UNSWIZZLE -> {
+                GpuUnswizzleViewHolder(ListItemSettingBinding.inflate(inflater), this)
+            }
+
             SettingsItem.TYPE_FX_TOOLBAR -> {
                 FxToolbarViewHolder(
                     ListItemSettingFxToolbarBinding.inflate(inflater, parent, false),
@@ -507,6 +511,14 @@ class SettingsAdapter(
         settingsViewModel.setShouldShowPathResetDialog(true)
     }
 
+    fun onGpuUnswizzleClick(item: GpuUnswizzleSetting, position: Int) {
+        GpuUnswizzleDialogFragment.newInstance(
+            settingsViewModel,
+            item,
+            position
+        ).show(fragment.childFragmentManager, GpuUnswizzleDialogFragment.TAG)
+    }
+
     private class DiffCallback : DiffUtil.ItemCallback<SettingsItem>() {
         override fun areItemsTheSame(oldItem: SettingsItem, newItem: SettingsItem): Boolean {
             return oldItem.setting.key == newItem.setting.key

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

@@ -545,14 +545,11 @@ class SettingsFragmentPresenter(
             add(IntSetting.RENDERER_NVDEC_EMULATION.key)
 
             add(BooleanSetting.SYNC_MEMORY_OPERATIONS.key)
-            add(BooleanSetting.STALL_ON_GPU_FENCE_WAIT.key)
             add(BooleanSetting.RENDERER_USE_DISK_SHADER_CACHE.key)
             add(BooleanSetting.RENDERER_FORCE_MAX_CLOCK.key)
             add(BooleanSetting.RENDERER_REACTIVE_FLUSHING.key)
-            add(BooleanSetting.RENDERER_BARRIER_FEEDBACK_LOOPS.key)
             add(BooleanSetting.ENABLE_BUFFER_HISTORY.key)
             add(BooleanSetting.ENABLE_GPU_BUFFER_READBACK.key)
-            add(BooleanSetting.USE_UNIFIED_MEMORY.key)
             add(BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS.key)
 
             add(HeaderSetting(R.string.hacks))
@@ -563,6 +560,8 @@ class SettingsFragmentPresenter(
             add(BooleanSetting.RENDERER_ASYNCHRONOUS_SHADERS.key)
             add(IntSetting.ANDROID_PIPELINE_WORKERS.key)
             add(BooleanSetting.RENDERER_ASYNCHRONOUS_GPU_EMULATION.key)
+            add(BooleanSetting.RENDERER_ASYNC_PRESENTATION.key)
+            add(SettingsItem.GPU_UNSWIZZLE_COMBINED)
 
             add(HeaderSetting(R.string.extensions))
 
@@ -1528,7 +1527,6 @@ class SettingsFragmentPresenter(
             add(HeaderSetting(R.string.cpu))
 
             add(IntSetting.CPU_BACKEND.key)
-            add(BooleanSetting.RELOCATE_36BIT_ADDRESS_SPACE.key)
             add(IntSetting.CPU_ACCURACY.key)
             add(BooleanSetting.USE_AUTO_STUB.key)
             add(SettingsItem.FASTMEM_COMBINED)

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

@@ -0,0 +1,71 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+package org.yuzu.yuzu_emu.features.settings.ui.viewholder
+
+import android.view.View
+import org.yuzu.yuzu_emu.R
+import org.yuzu.yuzu_emu.databinding.ListItemSettingBinding
+import org.yuzu.yuzu_emu.features.settings.model.view.GpuUnswizzleSetting
+import org.yuzu.yuzu_emu.features.settings.model.view.SettingsItem
+import org.yuzu.yuzu_emu.features.settings.ui.SettingsAdapter
+import org.yuzu.yuzu_emu.utils.ViewUtils.setVisible
+
+class GpuUnswizzleViewHolder(val binding: ListItemSettingBinding, adapter: SettingsAdapter) :
+    SettingViewHolder(binding.root, adapter) {
+    private lateinit var setting: GpuUnswizzleSetting
+
+    override fun bind(item: SettingsItem) {
+        setting = item as GpuUnswizzleSetting
+        binding.textSettingName.text = setting.title
+        binding.textSettingDescription.setVisible(item.description.isNotEmpty())
+        binding.textSettingDescription.text = item.description
+
+        binding.textSettingValue.setVisible(true)
+        val resMgr = binding.root.context.resources
+
+        if (setting.isEnabled()) {
+            // Show a summary of current settings
+            val textureSizeEntries = resMgr.getStringArray(setting.textureSizeChoicesId)
+            val textureSizeValues = resMgr.getIntArray(setting.textureSizeValuesId)
+            val textureSizeIndex = textureSizeValues.indexOf(setting.getTextureSize())
+            val textureSizeLabel = if (textureSizeIndex >= 0) textureSizeEntries[textureSizeIndex] else "?"
+
+            val streamSizeEntries = resMgr.getStringArray(setting.streamSizeChoicesId)
+            val streamSizeValues = resMgr.getIntArray(setting.streamSizeValuesId)
+            val streamSizeIndex = streamSizeValues.indexOf(setting.getStreamSize())
+            val streamSizeLabel = if (streamSizeIndex >= 0) streamSizeEntries[streamSizeIndex] else "?"
+
+            val chunkSizeEntries = resMgr.getStringArray(setting.chunkSizeChoicesId)
+            val chunkSizeValues = resMgr.getIntArray(setting.chunkSizeValuesId)
+            val chunkSizeIndex = chunkSizeValues.indexOf(setting.getChunkSize())
+            val chunkSizeLabel = if (chunkSizeIndex >= 0) chunkSizeEntries[chunkSizeIndex] else "?"
+
+            binding.textSettingValue.text = "$textureSizeLabel ⋅ $streamSizeLabel ⋅ $chunkSizeLabel"
+        } else {
+            binding.textSettingValue.text = resMgr.getString(R.string.gpu_unswizzle_disabled)
+        }
+
+        binding.buttonClear.setVisible(setting.clearable)
+        binding.buttonClear.setOnClickListener {
+            adapter.onClearClick(setting, bindingAdapterPosition)
+        }
+
+        setStyle(setting.isEditable, binding)
+    }
+
+    override fun onClick(clicked: View) {
+        if (!setting.isEditable) {
+            return
+        }
+
+        adapter.onGpuUnswizzleClick(setting, bindingAdapterPosition)
+    }
+
+    override fun onLongClick(clicked: View): Boolean {
+        if (setting.isEditable) {
+            return adapter.onLongClick(setting, bindingAdapterPosition)
+        }
+        return false
+    }
+}

src/android/app/src/main/java/org/yuzu/yuzu_emu/model/GameProperties.kt

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: 2023 yuzu Emulator Project

src/android/app/src/main/jni/emu_window/emu_window.cpp

@@ -64,8 +64,8 @@ void EmuWindow_Android::OnTouchReleased(int id) {
     EmulationSession::GetInstance().GetInputSubsystem().GetTouchScreen()->TouchReleased(id);
 }
 
-void EmuWindow_Android::OnFrameDisplayed(u32 presented_frames) {
-    UpdateObservedFrameRate(presented_frames);
+void EmuWindow_Android::OnFrameDisplayed() {
+    UpdateObservedFrameRate();
     UpdateFrameRateHint();
 
     if (!m_first_frame) {
@@ -75,16 +75,15 @@ void EmuWindow_Android::OnFrameDisplayed(u32 presented_frames) {
     }
 }
 
-void EmuWindow_Android::UpdateObservedFrameRate(u32 presented_frames) {
-    m_presented_frames = static_cast<float>((std::max)(presented_frames, 1u));
+void EmuWindow_Android::UpdateObservedFrameRate() {
     const auto now = Clock::now();
     if (m_last_frame_display_time.time_since_epoch().count() != 0) {
         const auto frame_time = std::chrono::duration<float>(now - m_last_frame_display_time);
         const float seconds = frame_time.count();
         if (seconds > 0.0f) {
-            const float instantaneous_rate = m_presented_frames / seconds;
+            const float instantaneous_rate = 1.0f / seconds;
             if (std::isfinite(instantaneous_rate) && instantaneous_rate >= 1.0f &&
-                instantaneous_rate <= 240.0f * m_presented_frames) {
+                instantaneous_rate <= 240.0f) {
                 constexpr float SmoothingFactor = 0.15f;
                 if (m_smoothed_present_rate <= 0.0f) {
                     m_smoothed_present_rate = instantaneous_rate;
@@ -125,9 +124,18 @@ float EmuWindow_Android::GetFrameTimeVerifiedHint() const {
     return QuantizeFrameRateHint(verified_rate);
 }
 
+float EmuWindow_Android::GetPresentedFrameMultiplier() {
+    if (!Settings::values.frame_gen.GetValue()) {
+        return 1.0f;
+    }
+    return static_cast<float>(std::clamp<u32>(Settings::values.frame_gen_multiplier.GetValue(), 2, 4));
+}
+
 float EmuWindow_Android::GetFrameRateHint() const {
-    const float observed_rate = std::clamp(m_smoothed_present_rate, 0.0f, 240.0f);
-    const float frame_time_verified_hint = GetFrameTimeVerifiedHint() * m_presented_frames;
+    const float presented_multiplier = GetPresentedFrameMultiplier();
+    const float observed_rate =
+        std::clamp(m_smoothed_present_rate * presented_multiplier, 0.0f, 240.0f);
+    const float frame_time_verified_hint = GetFrameTimeVerifiedHint() * presented_multiplier;
 
     if (m_last_frame_rate_hint > 0.0f && observed_rate > 0.0f) {
         const float tolerance = std::max(m_last_frame_rate_hint * 0.12f, 4.0f);
@@ -151,7 +159,7 @@ float EmuWindow_Android::GetFrameRateHint() const {
         return frame_time_verified_hint;
     }
 
-    const float nominal_rate = 60.0f * m_presented_frames;
+    const float nominal_rate = 60.0f * presented_multiplier;
     if (!Settings::values.use_speed_limit.GetValue()) {
         return QuantizeFrameRateHint(nominal_rate);
     }

src/android/app/src/main/jni/emu_window/emu_window.h

@@ -41,7 +41,7 @@ public:
     ~EmuWindow_Android() = default;
 
     void OnSurfaceChanged(ANativeWindow* surface);
-    void OnFrameDisplayed(u32 presented_frames) override;
+    void OnFrameDisplayed() override;
 
     void OnTouchPressed(int id, float x, float y);
     void OnTouchMoved(int id, float x, float y);
@@ -58,9 +58,10 @@ private:
     using Clock = std::chrono::steady_clock;
 
     void UpdateFrameRateHint();
-    void UpdateObservedFrameRate(u32 presented_frames);
+    void UpdateObservedFrameRate();
     [[nodiscard]] float GetFrameRateHint() const;
     [[nodiscard]] float GetFrameTimeVerifiedHint() const;
+    [[nodiscard]] static float GetPresentedFrameMultiplier();
     [[nodiscard]] static float QuantizeFrameRateHint(float frame_rate);
 
     float m_window_width{};
@@ -72,7 +73,6 @@ private:
     float m_last_frame_rate_hint = -1.0f;
     float m_pending_frame_rate_hint = -1.0f;
     float m_smoothed_present_rate = 0.0f;
-    float m_presented_frames = 1.0f;
     Clock::time_point m_last_frame_display_time{};
     Clock::time_point m_pending_frame_rate_since{};
     std::uint32_t m_pending_frame_rate_hint_votes = 0;

src/android/app/src/main/jni/native.cpp

@@ -738,35 +738,6 @@ const char* fallback_cpu_detection() {
     return s_result.c_str();
 }
 
-#ifdef ARCHITECTURE_arm64
-struct SystemDriverInfo {
-    VkPhysicalDeviceProperties properties{};
-    VkPhysicalDeviceDriverProperties driver{};
-};
-
-SystemDriverInfo QuerySystemDriverInfo(JNIEnv* env, jstring j_hook_lib_dir) {
-    const std::string hook_lib_dir = Common::Android::GetJString(env, j_hook_lib_dir);
-    const Common::DynamicLibrary library{adrenotools_open_libvulkan(
-        RTLD_NOW, 0, nullptr, hook_lib_dir.c_str(), nullptr, nullptr, nullptr, nullptr)};
-    Vulkan::vk::InstanceDispatch dld;
-    const Vulkan::vk::Instance instance = Vulkan::CreateInstance(library, dld, VK_API_VERSION_1_1);
-    const std::vector<VkPhysicalDevice> devices = instance.EnumeratePhysicalDevices();
-    if (devices.empty()) {
-        throw Vulkan::vk::Exception(VK_ERROR_INITIALIZATION_FAILED);
-    }
-    SystemDriverInfo info{};
-    info.driver.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES;
-    VkPhysicalDeviceProperties2 properties2{
-        .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2,
-        .pNext = &info.driver,
-        .properties = {},
-    };
-    Vulkan::vk::PhysicalDevice(devices[0], dld).GetProperties2(properties2);
-    info.properties = properties2.properties;
-    return info;
-}
-#endif
-
 } // namespace
 
 extern "C" {
@@ -886,18 +857,34 @@ jboolean JNICALL Java_org_yuzu_yuzu_1emu_utils_GpuDriverHelper_supportsCustomDri
 
 jobjectArray Java_org_yuzu_yuzu_1emu_utils_GpuDriverHelper_getSystemDriverInfo(
     JNIEnv* env, jobject j_obj, jobject j_surf, jstring j_hook_lib_dir) {
-    std::string version_string{"1.1.0"};
-    std::string driver_name{"generic"};
 #ifdef ARCHITECTURE_arm64
-    try {
-        const SystemDriverInfo info = QuerySystemDriverInfo(env, j_hook_lib_dir);
-        const u32 driver_version = info.properties.driverVersion;
-        version_string =
-            fmt::format("{}.{}.{}", VK_API_VERSION_MAJOR(driver_version),
-                        VK_API_VERSION_MINOR(driver_version), VK_API_VERSION_PATCH(driver_version));
-        driver_name = Vulkan::vk::GetDriverName(info.driver);
-    } catch (...) {
-    }
+    const char* file_redirect_dir_{};
+    int featureFlags{};
+    std::string hook_lib_dir = Common::Android::GetJString(env, j_hook_lib_dir);
+    auto handle = adrenotools_open_libvulkan(RTLD_NOW, featureFlags, nullptr, hook_lib_dir.c_str(),
+                                             nullptr, nullptr, file_redirect_dir_, nullptr);
+    auto driver_library = std::make_shared<Common::DynamicLibrary>(handle);
+    InputCommon::InputSubsystem input_subsystem;
+    auto window =
+        std::make_unique<EmuWindow_Android>(ANativeWindow_fromSurface(env, j_surf), driver_library);
+
+    Vulkan::vk::InstanceDispatch dld;
+    Vulkan::vk::Instance vk_instance = Vulkan::CreateInstance(
+        *driver_library, dld, VK_API_VERSION_1_1, Core::Frontend::WindowSystemType::Android);
+
+    auto surface = Vulkan::CreateSurface(vk_instance, window->GetWindowInfo());
+
+    auto device = Vulkan::CreateDevice(vk_instance, dld, *surface);
+
+    auto driver_version = device.GetDriverVersion();
+    auto version_string =
+        fmt::format("{}.{}.{}", VK_API_VERSION_MAJOR(driver_version),
+                    VK_API_VERSION_MINOR(driver_version), VK_API_VERSION_PATCH(driver_version));
+    auto driver_name = device.GetDriverName();
+#else
+    auto driver_version = "1.0.0";
+    auto version_string = "1.1.0"; //Assume lowest Vulkan level
+    auto driver_name = "generic";
 #endif
     jobjectArray j_driver_info = env->NewObjectArray(2, Common::Android::GetStringClass(), Common::Android::ToJString(env, version_string));
     env->SetObjectArrayElement(j_driver_info, 1, Common::Android::ToJString(env, driver_name));
@@ -906,12 +893,32 @@ jobjectArray Java_org_yuzu_yuzu_1emu_utils_GpuDriverHelper_getSystemDriverInfo(
 
 jstring Java_org_yuzu_yuzu_1emu_utils_GpuDriverHelper_getGpuModel(JNIEnv *env, jobject j_obj, jobject j_surf, jstring j_hook_lib_dir) {
 #ifdef ARCHITECTURE_arm64
-    try {
-        return Common::Android::ToJString(env, QuerySystemDriverInfo(env, j_hook_lib_dir).properties.deviceName);
-    } catch (...) {
-    }
-#endif
+    const char* file_redirect_dir_{};
+    int featureFlags{};
+    std::string hook_lib_dir = Common::Android::GetJString(env, j_hook_lib_dir);
+    auto handle = adrenotools_open_libvulkan(RTLD_NOW, featureFlags, nullptr, hook_lib_dir.c_str(),
+                                             nullptr, nullptr, file_redirect_dir_, nullptr);
+    auto driver_library = std::make_shared<Common::DynamicLibrary>(handle);
+    InputCommon::InputSubsystem input_subsystem;
+    auto window =
+            std::make_unique<EmuWindow_Android>(ANativeWindow_fromSurface(env, j_surf), driver_library);
+
+    Vulkan::vk::InstanceDispatch dld;
+    Vulkan::vk::Instance vk_instance = Vulkan::CreateInstance(
+            *driver_library, dld, VK_API_VERSION_1_1, Core::Frontend::WindowSystemType::Android);
+
+    auto surface = Vulkan::CreateSurface(vk_instance, window->GetWindowInfo());
+
+    auto device = Vulkan::CreateDevice(vk_instance, dld, *surface);
+
+    const std::string model_name{device.GetModelName()};
+
+    window.release();
+
+    return Common::Android::ToJString(env, model_name);
+#else
     return Common::Android::ToJString(env, "no-info");
+#endif
 }
 
 jboolean Java_org_yuzu_yuzu_1emu_NativeLibrary_reloadKeys(JNIEnv* env, jclass clazz) {

src/android/app/src/main/res/layout/dialog_gpu_unswizzle.xml

@@ -0,0 +1,96 @@
+<?xml version="1.0" encoding="utf-8"?>
+<ScrollView xmlns:android="http://schemas.android.com/apk/res/android"
+    xmlns:app="http://schemas.android.com/apk/res-auto"
+    android:layout_width="match_parent"
+    android:layout_height="wrap_content"
+    android:scrollbars="vertical">
+
+    <LinearLayout
+        android:layout_width="match_parent"
+        android:layout_height="wrap_content"
+        android:orientation="vertical"
+        android:padding="24dp">
+
+        <com.google.android.material.switchmaterial.SwitchMaterial
+            android:id="@+id/switch_enable"
+            android:layout_width="match_parent"
+            android:layout_height="wrap_content"
+            android:layout_marginBottom="16dp"
+            android:text="@string/gpu_unswizzle_enable" />
+
+        <com.google.android.material.textfield.TextInputLayout
+        android:id="@+id/layout_texture_size"
+        style="@style/Widget.Material3.TextInputLayout.OutlinedBox.ExposedDropdownMenu"
+        android:layout_width="match_parent"
+        android:layout_height="wrap_content"
+        android:layout_marginBottom="12dp"
+        android:hint="@string/gpu_unswizzle_texture_size">
+
+        <com.google.android.material.textfield.MaterialAutoCompleteTextView
+            android:id="@+id/dropdown_texture_size"
+            android:layout_width="match_parent"
+            android:layout_height="wrap_content"
+            android:inputType="none" />
+        </com.google.android.material.textfield.TextInputLayout>
+
+        <com.google.android.material.textfield.TextInputLayout
+            android:id="@+id/layout_stream_size"
+            style="@style/Widget.Material3.TextInputLayout.OutlinedBox.ExposedDropdownMenu"
+            android:layout_width="match_parent"
+            android:layout_height="wrap_content"
+            android:layout_marginBottom="12dp"
+            android:hint="@string/gpu_unswizzle_stream_size">
+
+            <com.google.android.material.textfield.MaterialAutoCompleteTextView
+                android:id="@+id/dropdown_stream_size"
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:inputType="none" />
+        </com.google.android.material.textfield.TextInputLayout>
+
+        <com.google.android.material.textfield.TextInputLayout
+            android:id="@+id/layout_chunk_size"
+            style="@style/Widget.Material3.TextInputLayout.OutlinedBox.ExposedDropdownMenu"
+            android:layout_width="match_parent"
+            android:layout_height="wrap_content"
+            android:hint="@string/gpu_unswizzle_chunk_size">
+
+            <com.google.android.material.textfield.MaterialAutoCompleteTextView
+                android:id="@+id/dropdown_chunk_size"
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:inputType="none" />
+        </com.google.android.material.textfield.TextInputLayout>
+
+        <LinearLayout
+            android:layout_width="match_parent"
+            android:layout_height="wrap_content"
+            android:layout_marginTop="24dp"
+            android:orientation="horizontal"
+            android:gravity="end"
+            android:spacing="8dp">
+
+            <com.google.android.material.button.MaterialButton
+                android:id="@+id/btn_default"
+                style="@style/Widget.Material3.Button.TextButton"
+                android:layout_width="wrap_content"
+                android:layout_height="wrap_content"
+                android:text="@string/gpu_unswizzle_default_button" />
+
+            <com.google.android.material.button.MaterialButton
+                android:id="@+id/btn_cancel"
+                style="@style/Widget.Material3.Button.TextButton"
+                android:layout_width="wrap_content"
+                android:layout_height="wrap_content"
+                android:text="@android:string/cancel" />
+
+            <com.google.android.material.button.MaterialButton
+                android:id="@+id/btn_ok"
+                style="@style/Widget.Material3.Button.TextButton"
+                android:layout_width="wrap_content"
+                android:layout_height="wrap_content"
+                android:text="@android:string/ok" />
+        </LinearLayout>
+
+    </LinearLayout>
+</ScrollView>

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

@@ -574,6 +574,8 @@
     <string name="renderer_force_max_clock_description">يجبر وحدة معالجة الرسومات على العمل بأقصى سرعة ممكنة (سيظل يتم تطبيق القيود الحرارية).</string>
     <string name="renderer_asynchronous_gpu_emulation">محاكاة غير متزامنة لوحدة معالجة الرسومات</string>
     <string name="renderer_asynchronous_gpu_emulation_description">يمكن لهذه الحيلة أن تزيد الأداء عن طريق تشغيل محاكاة وحدة معالجة الرسومات بشكل غير متزامن على حساب مشاكل الرسومات وزيادة معدلات الأعطال بسبب العمليات المتعلقة بالتوقيت.</string>
+    <string name="renderer_async_presentation">عرض غير متزامن</string>
+    <string name="renderer_async_presentation_description">يمكن لهذه الحيلة أن تزيد من الأداء عن طريق نقل عملية العرض إلى خيط معالجة منفصل على حساب مشاكل الرسوميات.</string>
     <string name="renderer_reactive_flushing">استخدم التنظيف التفاعلي</string>
     <string name="renderer_reactive_flushing_description">يحسن دقة العرض في بعض الألعاب على حساب الأداء.</string>
     <string name="enable_buffer_history">تمكين سجل التخزين المؤقت</string>
@@ -597,6 +599,17 @@
     <string name="rescale_hack_description">يُمكّن هذا الخيار من التعامل مع عملية إعادة تحجيم الألعاب بطريقة تقليدية باستخدام مسار إعادة التحجيم السريع</string>
     <string name="renderer_asynchronous_shaders">استخدم تظليل غير متزامن</string>
     <string name="renderer_asynchronous_shaders_description">يقوم بتجميع التظليل بشكل غير متزامن. قد يقلل ذلك من التقطعات ولكنه قد يؤدي أيضًا إلى حدوث أخطاء.</string>
+    <string name="gpu_unswizzle_settings">إعدادات إلغاء ترتيب بيانات وحدة معالجة الرسومات</string>
+    <string name="gpu_unswizzle_settings_description">قم بضبط معلمات فكّ تشابك النسيج المستندة إلى وحدة معالجة الرسومات أو تعطيلها تمامًا. اضبط هذه الإعدادات لتحقيق التوازن بين الأداء وجودة تحميل النسيج.</string>
+    <string name="gpu_unswizzle_enable">تفعيل إلغاء ترتيب بيانات وحدة معالجة الرسومات</string>
+    <string name="gpu_unswizzle_disabled">تعطيل</string>
+    <string name="gpu_unswizzle_texture_size">الحد الأقصى لحجم النسيج في وحدة معالجة الرسومات بعد إعادة ترتيب البيانات</string>
+    <string name="gpu_unswizzle_texture_size_description">يُحدد هذا الخيار الحد الأقصى لحجم (ميغابايت) معالجة الصور باستخدام وحدة معالجة الرسومات. مع أن وحدة معالجة الرسومات أسرع في معالجة الصور المتوسطة والكبيرة، إلا أن وحدة المعالجة المركزية قد تكون أكثر كفاءة في معالجة الصور الصغيرة جدًا. اضبط هذا الخيار لتحقيق التوازن الأمثل بين سرعة معالجة الرسومات واستهلاك وحدة المعالجة المركزية.</string>
+    <string name="gpu_unswizzle_stream_size">حجم تدفق إلغاء ترتيب بيانات وحدة معالجة الرسومات</string>
+    <string name="gpu_unswizzle_stream_size_description">يحدد هذا الخيار حد البيانات لكل إطار لمعالجة الصور الكبيرة. القيم الأعلى تُسرّع تحميل الصور على حساب زيادة زمن استجابة الإطارات؛ أما القيم الأقل فتُقلل من الحمل الزائد على وحدة معالجة الرسومات، ولكنها قد تتسبب في ظهور الصور بشكل مفاجئ.</string>
+    <string name="gpu_unswizzle_chunk_size">حجم كتلة إلغاء ترتيب بيانات وحدة معالجة الرسومات</string>
+    <string name="gpu_unswizzle_chunk_size_description">يُحدد هذا الخيار عدد شرائح العمق التي تتم معالجتها لكل دفعة من الصور ثلاثية الأبعاد (3D). زيادة هذا العدد تُحسّن كفاءة الإنتاجية على وحدات معالجة الرسومات القوية، ولكنها قد تُسبب تقطعًا أو انقطاعًا في عمل برنامج التشغيل على الأجهزة ذات المواصفات الأقل قوة.</string>
+    <string name="gpu_unswizzle_default_button">افتراضي</string>
 
 
     <string name="extensions">إضافات</string>
@@ -1037,10 +1050,25 @@
     <string name="fast_gpu_high">كسر السرعة</string>
 
     <!-- GPU swizzle texture size -->
+    <string name="gpu_texturesizeswizzle_verysmall">صغير جدًا (16 ميغابايت)</string>
+    <string name="gpu_texturesizeswizzle_small">صغير (32 ميغابايت)</string>
+    <string name="gpu_texturesizeswizzle_normal">قياسي (128 ميغابايت)</string>
+    <string name="gpu_texturesizeswizzle_large">كبير (256 ميغابايت)</string>
+    <string name="gpu_texturesizeswizzle_verylarge">كبير جدًا (512 ميغابايت)</string>
 
     <!-- GPU swizzle streams -->
+    <string name="gpu_swizzle_verylow">منخفض جدًا (4 ميغابايت)</string>
+    <string name="gpu_swizzle_low">منخفض (8 ميغابايت)</string>
+    <string name="gpu_swizzle_normal">قياسي (16 ميغابايت)</string>
+    <string name="gpu_swizzle_medium">متوسط (32 ميغابايت)</string>
+    <string name="gpu_swizzle_high">عالي (64 ميغابايت)</string>
 
     <!-- GPU swizzle chunks -->
+    <string name="gpu_swizzlechunk_verylow">منخفض جدًا (32)</string>
+    <string name="gpu_swizzlechunk_low">منخفض (64)</string>
+    <string name="gpu_swizzlechunk_normal">قياسي (128)</string>
+    <string name="gpu_swizzlechunk_medium">متوسط (256)</string>
+    <string name="gpu_swizzlechunk_high">عالي (512)</string>
 
     <!-- Temperature Units -->
     <string name="temperature_celsius">مئوية</string>

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

@@ -508,6 +508,7 @@ Wird der Handheld-Modus verwendet, verringert es die Auflösung und erhöht die
     <string name="skip_cpu_inner_invalidation_description">Überspringt bestimmte Cache-Invalidierungen auf CPU-Seite während Speicherupdates, reduziert die CPU-Auslastung und verbessert die Leistung. Kann in einigen Spielen zu Fehlern oder Abstürzen führen.</string>
     <string name="renderer_asynchronous_shaders">Asynchrone Shader</string>
     <string name="renderer_asynchronous_shaders_description">Kompiliert Shader asynchron. Dies kann Ruckler reduzieren, aber auch Grafikfehler verursachen.</string>
+    <string name="gpu_unswizzle_default_button">Standard</string>
 
 
     <string name="extensions">Erweiterungen</string>
@@ -917,10 +918,25 @@ Wirklich fortfahren?</string>
     <string name="fast_gpu_high">Übertaktung</string>
 
     <!-- GPU swizzle texture size -->
+    <string name="gpu_texturesizeswizzle_verysmall">Sehr klein (16 MB)</string>
+    <string name="gpu_texturesizeswizzle_small">Klein (32 MB)</string>
+    <string name="gpu_texturesizeswizzle_normal">Normal (128 MB)</string>
+    <string name="gpu_texturesizeswizzle_large">Groß (256 MB)</string>
+    <string name="gpu_texturesizeswizzle_verylarge">Sehr groß (512 MB)</string>
 
     <!-- GPU swizzle streams -->
+    <string name="gpu_swizzle_verylow">Sehr niedrig (4 MB)</string>
+    <string name="gpu_swizzle_low">Niedrig (8 MB)</string>
+    <string name="gpu_swizzle_normal">Normal (16 MB)</string>
+    <string name="gpu_swizzle_medium">Mittel (32 MB)</string>
+    <string name="gpu_swizzle_high">Hoch (64 MB)</string>
 
     <!-- GPU swizzle chunks -->
+    <string name="gpu_swizzlechunk_verylow">Sehr niedrig (32)</string>
+    <string name="gpu_swizzlechunk_low">Niedrig (64)</string>
+    <string name="gpu_swizzlechunk_normal">Normal (128)</string>
+    <string name="gpu_swizzlechunk_medium">Mittel (256)</string>
+    <string name="gpu_swizzlechunk_high">Hoch (512)</string>
 
     <!-- Temperature Units -->
     <string name="temperature_celsius">Celsius</string>

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

@@ -517,6 +517,8 @@
     <string name="renderer_force_max_clock_description">Fuerza a la GPU a ejecutarse a la velocidad máxima de reloj posible (se seguirán aplicando restricciones térmicas).</string>
     <string name="renderer_asynchronous_gpu_emulation">Emulación de GPU asíncrona</string>
     <string name="renderer_asynchronous_gpu_emulation_description">Este hack puede aumentar el rendimiento ejecutando la emulación de la GPU de forma asíncrona, a costa de problemas gráficos y un aumento en la tasa de fallos debido a operaciones relacionadas con la sincronización.</string>
+    <string name="renderer_async_presentation">Presentación asíncrona</string>
+    <string name="renderer_async_presentation_description">Este hack puede aumentar el rendimiento al mover la presentación a un hilo independiente de la CPU a costa de problemas gráficos.</string>
     <string name="renderer_reactive_flushing">Usar limpieza reactiva</string>
     <string name="renderer_reactive_flushing_description">Mejora la precisión de renderizado en algunos juegos, pero reduce el rendimiento.</string>
     <string name="enable_buffer_history">Activar el historial del búfer</string>
@@ -539,6 +541,17 @@
     <string name="rescale_hack_description">Permite el manejo de versiones anteriores para el paso de configuración de reescalado para juegos mediante el uso de una ruta de reescalado rápida.</string>
     <string name="renderer_asynchronous_shaders">Usar sombreadores asíncronos</string>
     <string name="renderer_asynchronous_shaders_description">Compila los sombreadores de forma asíncrona. Esto puede reducir los tirones, pero también puede introducir errores gráficos.</string>
+    <string name="gpu_unswizzle_settings">Ajustes de desentrelazado de la GPU</string>
+    <string name="gpu_unswizzle_settings_description">Configura los parámetros de desentrelazado de texturas basadas ​​en la GPU o desactívelos por completo. Modifique estos ajustes para equilibrar el rendimiento y la calidad de las texturas cargadas.</string>
+    <string name="gpu_unswizzle_enable">Activar desentrelazado de la GPU</string>
+    <string name="gpu_unswizzle_disabled">Desactivado</string>
+    <string name="gpu_unswizzle_texture_size">Tamaño máximo de textura de desentrelazado de la GPU</string>
+    <string name="gpu_unswizzle_texture_size_description">Establece el tamaño máximo (en MB) para el desentrelazado de texturas basada en GPU. Aunque la GPU es más rápida para texturas medianas y grandes, la CPU puede ser más eficiente para texturas muy pequeñas. Ajuste este valor para encontrar el equilibrio entre la aceleración de la GPU y la sobrecarga de la CPU.</string>
+    <string name="gpu_unswizzle_stream_size">Tamaño del flujo de desentrelazado de la GPU</string>
+    <string name="gpu_unswizzle_stream_size_description">Establece el límite de datos por fotograma para desentrelazar texturas grandes. Los valores altos aceleran la carga de texturas, a coste de una mayor latencia por fotograma; los valores bajos reducen la carga de la GPU, pero pueden causar parpadeos visibles en las texturas.</string>
+    <string name="gpu_unswizzle_chunk_size">Tamaño del trozo de desentrelazado de la GPU</string>
+    <string name="gpu_unswizzle_chunk_size_description">Determina la cantidad de cortes de profundidad procesados ​​en un solo envío de texturas 3D. Aumentar este valor puede mejorar el rendimiento en una GPU de gama alta, pero puede causar tirones y problemas en los tiempos de respuesta en hardware más modesto.</string>
+    <string name="gpu_unswizzle_default_button">Por defecto</string>
 
 
     <string name="extensions">Extensiones</string>
@@ -975,10 +988,25 @@
     <string name="fast_gpu_high">Overclock</string>
 
     <!-- GPU swizzle texture size -->
+    <string name="gpu_texturesizeswizzle_verysmall">Muy pequeño (16 MB)</string>
+    <string name="gpu_texturesizeswizzle_small">Pequeño (32 MB)</string>
+    <string name="gpu_texturesizeswizzle_normal">Normal (128 MB)</string>
+    <string name="gpu_texturesizeswizzle_large">Grande (256 MB)</string>
+    <string name="gpu_texturesizeswizzle_verylarge">Muy grande (512 MB)</string>
 
     <!-- GPU swizzle streams -->
+    <string name="gpu_swizzle_verylow">Muy bajo (4 MB)</string>
+    <string name="gpu_swizzle_low">Bajo (8 MB)</string>
+    <string name="gpu_swizzle_normal">Normal (16 MB)</string>
+    <string name="gpu_swizzle_medium">Medio (32 MB)</string>
+    <string name="gpu_swizzle_high">Alto (64 MB)</string>
 
     <!-- GPU swizzle chunks -->
+    <string name="gpu_swizzlechunk_verylow">Muy bajo (32)</string>
+    <string name="gpu_swizzlechunk_low">Bajo (64)</string>
+    <string name="gpu_swizzlechunk_normal">Normal (128)</string>
+    <string name="gpu_swizzlechunk_medium">Medio (256)</string>
+    <string name="gpu_swizzlechunk_high">Alto (512)</string>
 
     <!-- Temperature Units -->
     <string name="temperature_celsius">Celsius</string>

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

@@ -482,6 +482,8 @@
     <string name="emulate_bgr565">Emuler BGR565</string>
     <string name="renderer_asynchronous_shaders">Utiliser les shaders asynchrones</string>
     <string name="renderer_asynchronous_shaders_description">Compile les shaders de manière asynchrone. Cela peut réduire les saccades mais peut aussi provoquer des problèmes graphiques.</string>
+    <string name="gpu_unswizzle_disabled">Désactivé</string>
+    <string name="gpu_unswizzle_default_button">Par défaut</string>
 
 
     <string name="extensions">Extensions</string>
@@ -882,10 +884,25 @@
     <string name="memory_8gb">8 Go (Dangereux)</string>
 
     <!-- GPU swizzle texture size -->
+    <string name="gpu_texturesizeswizzle_verysmall">Très petit (16 Mo)</string>
+    <string name="gpu_texturesizeswizzle_small">Petit (32 Mo)</string>
+    <string name="gpu_texturesizeswizzle_normal">Normal (128 Mo)</string>
+    <string name="gpu_texturesizeswizzle_large">Large (256 Mo)</string>
+    <string name="gpu_texturesizeswizzle_verylarge">Très large (512 Mo)</string>
 
     <!-- GPU swizzle streams -->
+    <string name="gpu_swizzle_verylow">Très faible (4 Mo)</string>
+    <string name="gpu_swizzle_low">Faible (8 Mo)</string>
+    <string name="gpu_swizzle_normal">Normal (16 Mo)</string>
+    <string name="gpu_swizzle_medium">Moyen (32 Mo)</string>
+    <string name="gpu_swizzle_high">Élevé (64 Mo)</string>
 
     <!-- GPU swizzle chunks -->
+    <string name="gpu_swizzlechunk_verylow">Très faible (32)</string>
+    <string name="gpu_swizzlechunk_low">Faible (64)</string>
+    <string name="gpu_swizzlechunk_normal">Normal (128)</string>
+    <string name="gpu_swizzlechunk_medium">Moyen (256)</string>
+    <string name="gpu_swizzlechunk_high">Élevé (512)</string>
 
     <!-- Temperature Units -->
     <string name="temperature_celsius">Celsius</string>

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

@@ -554,6 +554,8 @@
     <string name="renderer_force_max_clock_description">Заставляет ГПУ работать на максимально возможных тактовых частотах (тепловые ограничения все равно будут применяться).</string>
     <string name="renderer_asynchronous_gpu_emulation">Асинхронная эмуляция ГПУ</string>
     <string name="renderer_asynchronous_gpu_emulation_description">Может повысить производительность за счёт асинхронного запуска эмуляции ГПУ, но ценой появления графических ошибок и увеличения частоты вылетов из-за операций, зависящих от синхронизации.</string>
+    <string name="renderer_async_presentation">Асинхронная презентация</string>
+    <string name="renderer_async_presentation_description">Может повысить производительность за счёт перемещения вывода кадров в отдельный поток ЦП, но ценой возникновения графических проблем.</string>
     <string name="renderer_reactive_flushing">Реактивная очистка</string>
     <string name="renderer_reactive_flushing_description">Повышение точности рендеринга в некоторых играх за счет снижения производительности.</string>
     <string name="enable_buffer_history">Включить историю буфера</string>
@@ -577,6 +579,17 @@
     <string name="rescale_hack_description">Включает старый метод обработки этапа перенастройки масштабирования для игр за счёт использования быстрого алгоритма перемасштабирования.</string>
     <string name="renderer_asynchronous_shaders">Использовать асинхронные шейдеры</string>
     <string name="renderer_asynchronous_shaders_description">Компилирует шейдеры асинхронно. Это может уменьшить подтормаживания, но также может вызвать графические артефакты.</string>
+    <string name="gpu_unswizzle_settings">Настройки распаковки текстур (Unswizzle)</string>
+    <string name="gpu_unswizzle_settings_description">Настройте параметры распаковки текстур на стороне ГПУ либо полностью отключите эту функцию. Изменение этих параметров позволяет найти баланс между производительностью и качеством загрузки текстур.</string>
+    <string name="gpu_unswizzle_enable">Включить распаковку текстур (Unswizzle)</string>
+    <string name="gpu_unswizzle_disabled">Отключено</string>
+    <string name="gpu_unswizzle_texture_size">Макс. размер текстуры Unswizzle</string>
+    <string name="gpu_unswizzle_texture_size_description">Задает максимальный размер (в МБ) текстур для преобразования формата (unswizzle) на ГПУ. Хотя ГПУ быстрее работает со средними и большими текстурами, ЦП может быть эффективнее для очень маленьких. Настройте это значение, чтобы найти баланс между ускорением на ГПУ и нагрузкой на ЦП.</string>
+    <string name="gpu_unswizzle_stream_size">Размер потока Unswizzle</string>
+    <string name="gpu_unswizzle_stream_size_description">Задает лимит данных на кадр для преобразования крупных текстур (unswizzle). Высокие значения ускоряют загрузку текстур ценой увеличения задержки кадра; низкие значения снижают нагрузку на ГПУ, но могут вызывать заметную постепенную подгрузку текстур.</string>
+    <string name="gpu_unswizzle_chunk_size">Размер блока Unswizzle</string>
+    <string name="gpu_unswizzle_chunk_size_description">Задает количество слоёв глубины, обрабатываемых за одну пачку для 3D-текстур. Увеличение этого значения улучшает пропускную способность на мощных ГПУ, но может вызывать подтормаживания или таймауты драйвера на слабом железе.</string>
+    <string name="gpu_unswizzle_default_button">По умолчанию</string>
 
 
     <string name="extensions">Расширения</string>
@@ -1017,10 +1030,25 @@
     <string name="fast_gpu_high">Разгон</string>
 
     <!-- GPU swizzle texture size -->
+    <string name="gpu_texturesizeswizzle_verysmall">Очень малый (16 МБ)</string>
+    <string name="gpu_texturesizeswizzle_small">Малый (32 МБ)</string>
+    <string name="gpu_texturesizeswizzle_normal">Обычный (128 МБ)</string>
+    <string name="gpu_texturesizeswizzle_large">Большой (256 МБ)</string>
+    <string name="gpu_texturesizeswizzle_verylarge">Очень большой (512 МБ)</string>
 
     <!-- GPU swizzle streams -->
+    <string name="gpu_swizzle_verylow">Очень низкий (4 МБ)</string>
+    <string name="gpu_swizzle_low">Низкий (8 МБ)</string>
+    <string name="gpu_swizzle_normal">Обычный (16 МБ)</string>
+    <string name="gpu_swizzle_medium">Средний (32 МБ)</string>
+    <string name="gpu_swizzle_high">Высокий (64 МБ)</string>
 
     <!-- GPU swizzle chunks -->
+    <string name="gpu_swizzlechunk_verylow">Очень малый (32)</string>
+    <string name="gpu_swizzlechunk_low">Малый (64)</string>
+    <string name="gpu_swizzlechunk_normal">Обычный (128)</string>
+    <string name="gpu_swizzlechunk_medium">Средний (256)</string>
+    <string name="gpu_swizzlechunk_high">Большой (512)</string>
 
     <!-- Temperature Units -->
     <string name="temperature_celsius">Цельсий</string>

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

@@ -481,6 +481,8 @@
     <string name="renderer_force_max_clock_description">Змушує GPU працювати на максимальній тактовій частоті.</string>
     <string name="renderer_asynchronous_gpu_emulation">Асинхронна емуляція ГП</string>
     <string name="renderer_asynchronous_gpu_emulation_description">Це обхідне рішення може покращити продуктивність завдяки асинхронному виконанню емуляції ГП, але спричинить проблеми з графікою та збільшить частоту збоїв чутливих до таймінгів операцій.</string>
+    <string name="renderer_async_presentation">Асинхронне подання</string>
+    <string name="renderer_async_presentation_description">Це обхідне рішення може покращити продуктивність завдяки переміщенню подання на окремий потік ЦП, але спричинить проблеми з графікою.</string>
     <string name="renderer_reactive_flushing">Реактивне очищення</string>
     <string name="renderer_reactive_flushing_description">Покращує точність рендерингу в деяких іграх.</string>
     <string name="enable_buffer_history">Увімкнути історію буфера</string>
@@ -500,6 +502,17 @@
     <string name="rescale_hack_description">Вмикає застарілу обробку масштабування для ігор, використовуючи швидкий шлях масштабування</string>
     <string name="renderer_asynchronous_shaders">Асинхронні шейдери</string>
     <string name="renderer_asynchronous_shaders_description">Компілює шейдери асинхронно. Це може зменшити затримки, але також може спричинити графічні баги.</string>
+    <string name="gpu_unswizzle_settings">Налаштування розпакування за допомогою ГП</string>
+    <string name="gpu_unswizzle_settings_description">Налаштуйте розпакування текстур за допомогою ГП або повністю вимкнути його. Відкоригуйте ці налаштування, щоб урівноважити продуктивність і якість завантаження текстур.</string>
+    <string name="gpu_unswizzle_enable">Увімкнути розпакування за допомогою ГП</string>
+    <string name="gpu_unswizzle_disabled">Вимкнено</string>
+    <string name="gpu_unswizzle_texture_size">Максимальний розмір текстур для розпакування ГП за допомогою ГП</string>
+    <string name="gpu_unswizzle_texture_size_description">Встановлює максимальний розмір (МБ) для розпакування текстур за допомогою ГП. ГП швидше справляється з текстурами середніх і великих розмірів, а ЦП ефективніший для дуже маленьких. Налаштуйте, щоб збалансувати ГП-прискоренням і навантаженням на ЦП.</string>
+    <string name="gpu_unswizzle_stream_size">Розмір потоку розпакування за допомогою ГП</string>
+    <string name="gpu_unswizzle_stream_size_description">Встановлює обмеження даних на кадр для розпакування великих текстур. Вищі значення пришвидшують завантаження текстур за рахунок більших кадрових затримок; менші значення зменшують перевантаження ГП але може спричинити помітні появи текстур.</string>
+    <string name="gpu_unswizzle_chunk_size">Розмір блоків розпакування за допомогою ГП</string>
+    <string name="gpu_unswizzle_chunk_size_description">Визначає кількість зрізів глибини, оброблених за партію 3D-текстур. Збільшення здатне покращити пропускну здатність на потужних ГП, але може призвести до затримок або затримок драйвера зі слабшим устаткуванням.</string>
+    <string name="gpu_unswizzle_default_button">Стандартно</string>
 
 
     <string name="extensions">Розширення</string>
@@ -916,10 +929,25 @@
     <string name="memory_8gb">8 ГБ (Небезпечно)</string>
 
     <!-- GPU swizzle texture size -->
+    <string name="gpu_texturesizeswizzle_verysmall">Дуже малий (16 МБ)</string>
+    <string name="gpu_texturesizeswizzle_small">Малий (32 МБ)</string>
+    <string name="gpu_texturesizeswizzle_normal">Нормальний (128 МБ)</string>
+    <string name="gpu_texturesizeswizzle_large">Великий (256 МБ)</string>
+    <string name="gpu_texturesizeswizzle_verylarge">Дуже великий (512 МБ)</string>
 
     <!-- GPU swizzle streams -->
+    <string name="gpu_swizzle_verylow">Дуже низький (4 МБ)</string>
+    <string name="gpu_swizzle_low">Низький (8 МБ)</string>
+    <string name="gpu_swizzle_normal">Нормальний (16 МБ)</string>
+    <string name="gpu_swizzle_medium">Середній (32 МБ)</string>
+    <string name="gpu_swizzle_high">Високий (64 МБ)</string>
 
     <!-- GPU swizzle chunks -->
+    <string name="gpu_swizzlechunk_verylow">Дуже низький (32)</string>
+    <string name="gpu_swizzlechunk_low">Низький (64)</string>
+    <string name="gpu_swizzlechunk_normal">Нормальний (128)</string>
+    <string name="gpu_swizzlechunk_medium">Середній (256)</string>
+    <string name="gpu_swizzlechunk_high">Високий (512)</string>
 
     <!-- Temperature Units -->
     <string name="temperature_celsius">Цельсій</string>

src/android/app/src/main/res/values-zh-rCN/strings.xml

@@ -535,6 +535,8 @@
     <string name="renderer_force_max_clock_description">强制 GPU 以最大时钟运行 (温控依然生效)。</string>
     <string name="renderer_asynchronous_gpu_emulation">GPU 异步模拟</string>
     <string name="renderer_asynchronous_gpu_emulation_description">此技巧可通过异步运行 GPU 模拟来提升性能，但在执行与时序相关的操作时，可能带来图形显示问题以及增加崩溃概率。</string>
+    <string name="renderer_async_presentation">异步呈现</string>
+    <string name="renderer_async_presentation_description">此技巧通过将图形呈现移至独立的 CPU 线程来提升性能，但可能会带来图形显示问题。</string>
     <string name="renderer_reactive_flushing">启用反应性刷新</string>
     <string name="renderer_reactive_flushing_description">通过牺牲性能来提升某些游戏的渲染精度。</string>
     <string name="enable_buffer_history">启用缓冲区历史</string>
@@ -558,6 +560,17 @@
     <string name="rescale_hack_description">启用通过使用快速缩放路径，来为游戏提供缩放配置处理的传统处理方式</string>
     <string name="renderer_asynchronous_shaders">使用异步着色器</string>
     <string name="renderer_asynchronous_shaders_description">以异步方式编译着色器。采用此方式或可减少卡顿，但也可能引入故障点。</string>
+    <string name="gpu_unswizzle_settings">GPU Unswizzle 设置</string>
+    <string name="gpu_unswizzle_settings_description">配置基于 GPU 的纹理 unswizzling 参数，或完全禁用该功能。通过调整这些设置，以尝试在性能与纹理加载质量之间取得平衡。</string>
+    <string name="gpu_unswizzle_enable">启用 GPU Unswizzle</string>
+    <string name="gpu_unswizzle_disabled">禁用</string>
+    <string name="gpu_unswizzle_texture_size">GPU Unswizzle 最大纹理尺寸</string>
+    <string name="gpu_unswizzle_texture_size_description">设置基于 GPU 的纹理 unswizzling 的最大尺寸（MB）。虽然 GPU 处理中等和大型纹理的速度更快，但对于非常小的纹理，CPU 可能更为高效。通过调节此项设置，以平衡GPU 加速与 CPU 开销。</string>
+    <string name="gpu_unswizzle_stream_size">GPU Unswizzle 流大小</string>
+    <string name="gpu_unswizzle_stream_size_description">设置用于 unswizzling 大型纹理时的每帧数据限制。较高的数值可以加速纹理的加载过程，但会带来更高的帧延迟。而较低的数值则可以降低 GPU 的开销，但也可能会导致可见的纹理闪现。</string>
+    <string name="gpu_unswizzle_chunk_size">GPU Unswizzle 块大小</string>
+    <string name="gpu_unswizzle_chunk_size_description">定义了 3D 纹理每批次处理的深度切片数量。增加此数值可在高性能 GPU 上提升吞吐效率，但在性能较弱的硬件上可能会导致卡顿或驱动超时。</string>
+    <string name="gpu_unswizzle_default_button">默认</string>
 
 
     <string name="extensions">扩展</string>
@@ -991,10 +1004,25 @@
     <string name="fast_gpu_high">超频</string>
 
     <!-- GPU swizzle texture size -->
+    <string name="gpu_texturesizeswizzle_verysmall">极小 (16 MB)</string>
+    <string name="gpu_texturesizeswizzle_small">较小 (32 MB)</string>
+    <string name="gpu_texturesizeswizzle_normal">正常 (128 MB)</string>
+    <string name="gpu_texturesizeswizzle_large">较大 (256 MB)</string>
+    <string name="gpu_texturesizeswizzle_verylarge">极大 (512 MB)</string>
 
     <!-- GPU swizzle streams -->
+    <string name="gpu_swizzle_verylow">极低 (4 MB)</string>
+    <string name="gpu_swizzle_low">低 (8 MB)</string>
+    <string name="gpu_swizzle_normal">正常 (16 MB)</string>
+    <string name="gpu_swizzle_medium">中 (32 MB)</string>
+    <string name="gpu_swizzle_high">高 (64 MB)</string>
 
     <!-- GPU swizzle chunks -->
+    <string name="gpu_swizzlechunk_verylow">极低 (32)</string>
+    <string name="gpu_swizzlechunk_low">低 (64)</string>
+    <string name="gpu_swizzlechunk_normal">正常 (128)</string>
+    <string name="gpu_swizzlechunk_medium">中 (256)</string>
+    <string name="gpu_swizzlechunk_high">高 (512)</string>
 
     <!-- Temperature Units -->
     <string name="temperature_celsius">摄氏度</string>

src/android/app/src/main/res/values-zh-rTW/strings.xml

@@ -555,6 +555,8 @@
     <string name="renderer_force_max_clock_description">強制 GPU 以可能的最大時脈執行 (熱溫限制仍會被套用)</string>
     <string name="renderer_asynchronous_gpu_emulation">GPU 非同步模擬</string>
     <string name="renderer_asynchronous_gpu_emulation_description">此修改可透過使用 GPU 非同步模擬來提升性能，不過可能會導致圖形問題且會提高在進行時序相關操作時當機的機率</string>
+    <string name="renderer_async_presentation">非同步呈現</string>
+    <string name="renderer_async_presentation_description">此修改可以透過將渲染移至單獨的 CPU 執行緒來提升性能，但可能會導致圖形問題</string>
     <string name="renderer_reactive_flushing">使用重新啟用排清</string>
     <string name="renderer_reactive_flushing_description">犧牲效能，以改善部分遊戲的轉譯準確度</string>
     <string name="enable_buffer_history">啟用緩衝區歷史</string>
@@ -578,6 +580,17 @@
     <string name="rescale_hack_description">透過快速重新縮放處理來啟用遊戲重新縮放設定階段的舊版處理方式</string>
     <string name="renderer_asynchronous_shaders">使用非同步著色器</string>
     <string name="renderer_asynchronous_shaders_description">使用非同步編譯著色器。這可能會減少卡頓，但也可能導致圖形錯誤。</string>
+    <string name="gpu_unswizzle_settings">GPU Unswizzle 設定</string>
+    <string name="gpu_unswizzle_settings_description">設定基於 GPU 的 Unswizzling 參數或完全停用此功能。調整此設定以在性能與紋理載入品質中取得平衡</string>
+    <string name="gpu_unswizzle_enable">啟用 GPU Unswizzle</string>
+    <string name="gpu_unswizzle_disabled">停用</string>
+    <string name="gpu_unswizzle_texture_size">GPU Unswizzle 最大紋理尺寸</string>
+    <string name="gpu_unswizzle_texture_size_description">設定 GPU 紋理 Unswizzling 的最大值 （MB）。雖然 GPU 處理中型和大型紋理的速度較快，但 CPU 處理非常小的紋理可能更有效率。整此設定以在 GPU 加速與 CPU 負載之間取得平衡</string>
+    <string name="gpu_unswizzle_stream_size">GPU Unswizzle 流大小</string>
+    <string name="gpu_unswizzle_stream_size_description">設定每個影格處理大型紋理 Unswizzling 的資料上限。數值越高載入紋理的速度越快，但會增加畫面延遲；數值較低則能降低 GPU 負載，不過有可能導致至紋理載入不完全甚至突然出現</string>
+    <string name="gpu_unswizzle_chunk_size">GPU Unswizzle 塊大小</string>
+    <string name="gpu_unswizzle_chunk_size_description">設定 3D 紋理每批次處理的深度切片數量。增加此數值可提升規格較佳 GPU 的處理效率，但在較低規格的硬體上可能導致卡頓或驅動程式逾時</string>
+    <string name="gpu_unswizzle_default_button">預設</string>
 
 
     <string name="extensions">擴充功能</string>

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

@@ -591,6 +591,54 @@
         <item>2</item>
     </integer-array>
 
+    <string-array name="gpuTextureSizeSwizzleEntries">
+        <item>@string/gpu_texturesizeswizzle_verysmall</item>
+        <item>@string/gpu_texturesizeswizzle_small</item>
+        <item>@string/gpu_texturesizeswizzle_normal</item>
+        <item>@string/gpu_texturesizeswizzle_large</item>
+        <item>@string/gpu_texturesizeswizzle_verylarge</item>
+    </string-array>
+
+    <integer-array name="gpuTextureSizeSwizzleValues">
+        <item>0</item>
+        <item>1</item>
+        <item>2</item>
+        <item>3</item>
+        <item>4</item>
+    </integer-array>
+
+    <string-array name="gpuSwizzleEntries">
+        <item>@string/gpu_swizzle_verylow</item>
+        <item>@string/gpu_swizzle_low</item>
+        <item>@string/gpu_swizzle_normal</item>
+        <item>@string/gpu_swizzle_medium</item>
+        <item>@string/gpu_swizzle_high</item>
+    </string-array>
+
+    <integer-array name="gpuSwizzleValues">
+        <item>0</item>
+        <item>1</item>
+        <item>2</item>
+        <item>3</item>
+        <item>4</item>
+    </integer-array>
+    
+    <string-array name="gpuSwizzleChunkEntries">
+        <item>@string/gpu_swizzlechunk_verylow</item>
+        <item>@string/gpu_swizzlechunk_low</item>
+        <item>@string/gpu_swizzlechunk_normal</item>
+        <item>@string/gpu_swizzlechunk_medium</item>
+        <item>@string/gpu_swizzlechunk_high</item>
+    </string-array>
+
+    <integer-array name="gpuSwizzleChunkValues">
+        <item>0</item>
+        <item>1</item>
+        <item>2</item>
+        <item>3</item>
+        <item>4</item>
+    </integer-array>
+
     <string-array name="temperatureUnitEntries">
         <item>@string/temperature_celsius</item>
         <item>@string/temperature_fahrenheit</item>

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

@@ -105,8 +105,6 @@
     <string name="use_sync_core_description">Synchronize the core tick speed to the maximum speed percentage to improve performance without altering the game\'s actual speed.</string>
     <string name="cpuopt_unsafe_host_mmu">Enable Host MMU Emulation</string>
     <string name="cpuopt_unsafe_host_mmu_description">This optimization speeds up memory accesses by the guest program. Enabling it causes guest memory reads/writes to be done directly into memory and make use of Host\'s MMU. Disabling this forces all memory accesses to use Software MMU Emulation.</string>
-    <string name="relocate_36bit_address_space">Run 36-bit games natively</string>
-    <string name="relocate_36bit_address_space_description">Gives games built for the 36-bit address space the 39-bit layout so they can run with NCE instead of the JIT. Experimental: a game that relies on 36-bit addresses can crash.</string>
     <string name="debug_knobs">Debug knobs</string>
     <string name="debug_knobs_description">For development use only.</string>
     <string name="debug_knobs_hint">0 to 65535</string>
@@ -348,7 +346,7 @@
     <string name="frame_gen_flow_scale_auto">Match motion estimation to the game</string>
     <string name="frame_gen_flow_scale_auto_description">Estimate motion at the resolution the game actually renders instead of the upscaled output. Costs nothing in accuracy, since upscaling adds no motion detail.</string>
     <string name="frame_gen_flow_scale">Motion estimation resolution</string>
-    <string name="frame_gen_flow_scale_description">Resolution of the optical flow pass, as a fraction of the game\'s render resolution. Lowering it is the cheapest way to reclaim performance.</string>
+    <string name="frame_gen_flow_scale_description">Resolution of the optical flow pass, as a fraction of the output. Lowering it is the cheapest way to reclaim performance.</string>
     <string name="frame_gen_unsupported">Frame generation unavailable</string>
     <string name="frame_gen_unsupported_description">This GPU driver lacks the Vulkan memory model or half precision (float16) support that the Lossless Scaling shaders require.</string>
     <string name="lossless_scaling_setup_description">Optional. Provide your own Lossless.dll to enable frame generation later</string>
@@ -575,8 +573,6 @@
     <string name="accelerate_astc_description">Pick how ASTC-compressed textures are decoded for rendering: CPU (slow, safe), GPU (fast, recommended), or CPU Async (no stutters, may cause issues)</string>
 
     <string name="sync_memory_operations">Sync Memory Operations</string>
-    <string name="stall_on_gpu_fence_wait">Pause CPU on GPU fence timeouts</string>
-    <string name="stall_on_gpu_fence_wait_description">When a game keeps timing out while waiting for the GPU, pauses every CPU thread until the GPU catches up. Disabling it keeps the other threads running, which can reduce stutter but may break games that rely on it.</string>
     <string name="sync_memory_operations_description">Ensures data consistency between compute and memory operations. This option should fix issues in some games, but may also reduce performance in some cases. Unreal Engine 4 games often see the most significant changes thereof.</string>
     <string name="use_disk_shader_cache">Disk shader cache</string>
     <string name="use_disk_shader_cache_description">Reduces stuttering by locally storing and loading generated shaders.</string>
@@ -584,15 +580,13 @@
     <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_asynchronous_gpu_emulation">GPU async emulation</string>
     <string name="renderer_asynchronous_gpu_emulation_description">This hack can increase performance by running GPU emulation asynchronously at the cost of graphical issues and increased crash rates by timing-related operations.</string>
+    <string name="renderer_async_presentation">Asynchronous presentation</string>
+    <string name="renderer_async_presentation_description">This hack can increase performance by moving presentation to a separate CPU thread at the cost of graphical issues.</string>
     <string name="renderer_reactive_flushing">Use reactive flushing</string>
     <string name="renderer_reactive_flushing_description">Improves rendering accuracy in some games at the cost of performance.</string>
-    <string name="renderer_barrier_feedback_loops">Barrier feedback loops</string>
-    <string name="renderer_barrier_feedback_loops_description">Improves rendering of transparency effects in specific games at the cost of performance.</string>
     <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="enable_gpu_buffer_readback">Enable GPU Buffer Readback</string>
-    <string name="use_unified_memory">Unified Memory</string>
-    <string name="use_unified_memory_description">Backs part of the emulated memory with GPU-shared buffers so textures are unswizzled directly from it, skipping a CPU copy. Needs 12 GB of RAM or more and a GPU with I/O coherency; it turns itself off otherwise. Takes effect on the next game launch.</string>
     <string name="enable_gpu_buffer_readback_description">Preserves GPU-modified buffer data by reading it back before uploads. Some games require this to render certain effects properly. May cause issues if the hardware cannot handle the additional workload.</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/ QCOM drivers. Will crash on older Turnip drivers (25.3 and below).</string>
@@ -611,6 +605,17 @@
     <string name="rescale_hack_description">Enables a legacy handling for the rescale configuration pass for games by using a quick rescale path</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>
+    <string name="gpu_unswizzle_settings_description">Configure GPU-based texture unswizzling parameters or disable it entirely. Adjust these settings to balance performance and texture loading quality.</string>
+    <string name="gpu_unswizzle_enable">Enable GPU Unswizzle</string>
+    <string name="gpu_unswizzle_disabled">Disabled</string>
+    <string name="gpu_unswizzle_texture_size">GPU Unswizzle Max Texture Size</string>
+    <string name="gpu_unswizzle_texture_size_description">Sets the maximum size (MB) for GPU-based texture unswizzling. While the GPU is faster for medium and large textures, the CPU may be more efficient for very small ones. Adjust this to find the balance between GPU acceleration and CPU overhead.</string>
+    <string name="gpu_unswizzle_stream_size">GPU Unswizzle Stream Size</string>
+    <string name="gpu_unswizzle_stream_size_description">Sets the data limit per frame for unswizzling large textures. Higher values speed up texture loading at the cost of higher frame latency; lower values reduce GPU overhead but may cause visible texture pop-in.</string>
+    <string name="gpu_unswizzle_chunk_size">GPU Unswizzle Chunk Size</string>
+    <string name="gpu_unswizzle_chunk_size_description">Defines the number of depth slices processed per batch for 3D textures. Increasing this improves throughput efficiency on powerful GPUs but may cause stuttering or driver timeouts on weaker hardware.</string>
+    <string name="gpu_unswizzle_default_button">Default</string>
 
 
     <string name="extensions">Extensions</string>
@@ -1060,10 +1065,25 @@
     <string name="fast_gpu_high">Overclock</string>
 
     <!-- GPU swizzle texture size -->
+    <string name="gpu_texturesizeswizzle_verysmall">Very Small (16 MB)</string>
+    <string name="gpu_texturesizeswizzle_small">Small (32 MB)</string>
+    <string name="gpu_texturesizeswizzle_normal">Normal (128 MB)</string>
+    <string name="gpu_texturesizeswizzle_large">Large (256 MB)</string>
+    <string name="gpu_texturesizeswizzle_verylarge">Very Large (512 MB)</string>
 
     <!-- GPU swizzle streams -->
+    <string name="gpu_swizzle_verylow">Very Low (4 MB)</string>
+    <string name="gpu_swizzle_low">Low (8 MB)</string>
+    <string name="gpu_swizzle_normal">Normal (16 MB)</string>
+    <string name="gpu_swizzle_medium">Medium (32 MB)</string>
+    <string name="gpu_swizzle_high">High (64 MB)</string>
 
     <!-- GPU swizzle chunks -->
+    <string name="gpu_swizzlechunk_verylow">Very Low (32)</string>
+    <string name="gpu_swizzlechunk_low">Low (64)</string>
+    <string name="gpu_swizzlechunk_normal">Normal (128)</string>
+    <string name="gpu_swizzlechunk_medium">Medium (256)</string>
+    <string name="gpu_swizzlechunk_high">High (512)</string>
 
     <!-- Temperature Units -->
     <string name="temperature_celsius">Celsius</string>

src/common/cpu_features.cpp

@@ -244,7 +244,8 @@ WallClock::WallClock(bool invariant_, u64 rdtsc_frequency_) noexcept
     , ns_rdtsc_factor{invariant_ ? GetFixedPoint64Factor(NsRatio::den, rdtsc_frequency_) : 0}
     , us_rdtsc_factor{invariant_ ? GetFixedPoint64Factor(UsRatio::den, rdtsc_frequency_) : 0}
     , ms_rdtsc_factor{invariant_ ? GetFixedPoint64Factor(MsRatio::den, rdtsc_frequency_) : 0}
-    , rdtsc_ns_factor{invariant_ ? GetFixedPoint64Factor(rdtsc_frequency_, NsRatio::den) : 1}
+    , rdtsc_ns_integer{invariant_ ? rdtsc_frequency_ / NsRatio::den : 1}
+    , rdtsc_ns_factor{invariant_ ? GetFixedPoint64Factor(rdtsc_frequency_ % NsRatio::den, NsRatio::den) : 0}
     , cntpct_rdtsc_factor{invariant_ ? GetFixedPoint64Factor(CNTFRQ, rdtsc_frequency_) : 0}
     , gputick_rdtsc_factor{invariant_ ? GetFixedPoint64Factor(GPUTickFreq, rdtsc_frequency_) : 0}
     , invariant{invariant_}
@@ -291,7 +292,7 @@ bool WallClock::IsNative() const {
 }
 
 u64 WallClock::NsToTicks(std::chrono::nanoseconds ns) const {
-    return invariant ? MultiplyHigh(ns.count(), rdtsc_ns_factor) : ns.count();
+    return ns.count() * rdtsc_ns_integer + MultiplyHigh(ns.count(), rdtsc_ns_factor);
 }
 #elif defined(HAS_NCE)
 namespace {
@@ -416,7 +417,7 @@ u64 WallClock::NsToTicks(std::chrono::nanoseconds ns) const {
 const WallClock g_wall_clock = [] {
 #if defined(ARCHITECTURE_x86_64)
     auto const& caps = Common::g_cpu_caps;
-    return WallClock(caps.invariant_tsc && caps.tsc_frequency >= std::nano::den, caps.tsc_frequency);
+    return WallClock(caps.invariant_tsc && caps.tsc_frequency > std::nano::den, caps.tsc_frequency);
 #elif defined(HAS_NCE)
     return WallClock(false, 1);
 #else

src/common/cpu_features.h

@@ -96,6 +96,7 @@ public:
     u64 ns_rdtsc_factor;
     u64 us_rdtsc_factor;
     u64 ms_rdtsc_factor;
+    u64 rdtsc_ns_integer;
     u64 rdtsc_ns_factor;
     u64 cntpct_rdtsc_factor;
     u64 gputick_rdtsc_factor;

src/common/error.cpp

@@ -17,35 +17,33 @@
 
 namespace Common {
 
+// glibc, mlibc, musl, and newlib all define their own variants of strerror_r
+// We don't need to use the preprocessor, we can just select depending on return type
+template<typename T> std::string HandleStrerrorR(T r, char *err_str);
+template<> std::string HandleStrerrorR(char* r, char *) { return std::string{r}; }
+template<> std::string HandleStrerrorR(const char* r, char *) { return std::string{r}; }
+template<> std::string HandleStrerrorR(int r, char *err_str) {
+    return std::string{r != 0
+        ? "(strerror_r failed to format error)"
+        : err_str};
+}
+
 std::string NativeErrorToString(int e) {
 #ifdef _WIN32
     LPSTR err_str;
-
     DWORD res = FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_ALLOCATE_BUFFER |
                                    FORMAT_MESSAGE_IGNORE_INSERTS,
                                nullptr, e, MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT),
-                               reinterpret_cast<LPSTR>(&err_str), 1, nullptr);
-    if (!res) {
-        return "(FormatMessageA failed to format error)";
+                               LPSTR(&err_str), 1, nullptr);
+    if (res) {
+        std::string ret(err_str);
+        LocalFree(err_str);
+        return ret;
     }
-    std::string ret(err_str);
-    LocalFree(err_str);
-    return ret;
+    return "(FormatMessageA failed to format error)";
 #else
     char err_str[255];
-#if defined(__ANDROID__) ||                                                                            \
-    (defined(__GLIBC__) && (_GNU_SOURCE || (_POSIX_C_SOURCE < 200112L && _XOPEN_SOURCE < 600)))
-    // Thread safe (GNU-specific)
-    const char* str = strerror_r(e, err_str, sizeof(err_str));
-    return std::string(str);
-#else
-    // Thread safe (XSI-compliant)
-    int second_err = strerror_r(e, err_str, sizeof(err_str));
-    if (second_err != 0) {
-        return "(strerror_r failed to format error)";
-    }
-    return std::string(err_str);
-#endif // GLIBC etc.
+    return HandleStrerrorR(strerror_r(e, err_str, sizeof(err_str)), err_str);
 #endif // _WIN32
 }
 

src/common/fs/path_util.cpp

@@ -5,6 +5,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <algorithm>
+#include <cctype>
 #include <iostream>
 #include <sstream>
 #include "common/container/unordered_map.h"
@@ -480,28 +481,80 @@ std::string SanitizePath(std::string_view path_, DirectorySeparator directory_se
                            [type2](char c1, char c2) { return c1 == type2 && c2 == type2; }),
                path.end());
 
-    const bool absolute = !path.empty() && path[0] == type2;
-    std::vector<std::string_view> parts;
+    std::string root;
+    std::string_view components{path};
+    bool drive_relative = false;
 
-    for (const auto part : SplitPathComponents(path))
-    {
-        if (part.empty() || part == ".")
-            continue;
-        if (part == ".." && !parts.empty() && parts.back() != "..")
-            parts.pop_back();
-        else if (part != "..") parts.push_back(part);
+#ifdef _WIN32
+    const bool network = path.size() > 1 && path[0] == type2 && path[1] == type2;
+    const bool drive =
+        path.size() > 1 && std::isalpha(static_cast<unsigned char>(path[0])) && path[1] == ':';
+
+    if (network) {
+        root.assign(2, type2);
+        components.remove_prefix(2);
+    } else if (drive) {
+        root.assign(path.data(), 2);
+        components.remove_prefix(2);
+        if (!components.empty() && components.front() == type2) {
+            root += type2;
+            components.remove_prefix(1);
+        } else {
+            drive_relative = true;
+        }
+    }
+#endif
+
+    if (root.empty() && !components.empty() && components.front() == type2) {
+        root += type2;
+        components.remove_prefix(1);
     }
 
-    std::string resolved = absolute ? std::string(1, type2) : std::string{};
-    for (std::size_t i = 0; i < parts.size(); ++i)
-    {
-        if (i != 0)
+    const auto path_parts = SplitPathComponents(components);
+    std::size_t root_component_count = 0;
+#ifdef _WIN32
+    if (network) {
+        root_component_count = 2;
+
+        const auto is_unc = [](std::string_view part) {
+            return part.size() == 3 && (part[0] == 'U' || part[0] == 'u') &&
+                   (part[1] == 'N' || part[1] == 'n') && (part[2] == 'C' || part[2] == 'c');
+        };
+        if (path_parts.size() >= 2 && path_parts[0] == "?" && is_unc(path_parts[1])) {
+            root_component_count = 4;
+        }
+    }
+#endif
+
+    std::vector<std::string_view> parts;
+    for (std::size_t i = 0; i < path_parts.size(); ++i) {
+        const auto part = path_parts[i];
+        if (i < root_component_count) {
+            parts.push_back(part);
+        } else if (part.empty() || part == ".") {
+            continue;
+        } else if (part == "..") {
+            if (parts.size() > root_component_count) {
+                parts.pop_back();
+            }
+        } else {
+            parts.push_back(part);
+        }
+    }
+
+    const std::size_t root_length = root.size();
+    std::string resolved = std::move(root);
+    for (std::size_t i = 0; i < parts.size(); ++i) {
+        if (i != 0 || (!resolved.empty() && resolved.back() != type2 && !drive_relative))
             resolved += type2;
         resolved.append(parts[i].data(), parts[i].size());
     }
 
     path = std::move(resolved);
 
+    if (!path.empty() && path.size() == root_length) {
+        return path;
+    }
     return std::string(RemoveTrailingSlash(path));
 }
 

src/common/fs/path_util.h

@@ -347,8 +347,9 @@ enum class DirectorySeparator {
 // i.e. "C:\Users\Yuzu\Documents\save.bin" becomes {"C:", "Users", "Yuzu", "Documents", "save.bin" }
 [[nodiscard]] std::vector<std::string> SplitPathComponentsCopy(std::string_view filename);
 
-// Removes trailing slash, makes all '\\' into '/', and removes duplicate '/'. Makes '/' into '\\'
-// depending if directory_separator is BackwardSlash or PlatformDefault and running on windows
+// Normalizes directory separators, removes duplicate and non-root trailing separators, and resolves
+// '.' and '..' components without traversing above the path root. Windows drive and UNC roots are
+// preserved.
 [[nodiscard]] std::string SanitizePath(
     std::string_view path,
     DirectorySeparator directory_separator = DirectorySeparator::ForwardSlash);

src/common/page_table.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project

src/common/settings.cpp

@@ -48,6 +48,7 @@ SWITCHABLE(AnisotropyMode, true);
 SWITCHABLE(AntiAliasing, false);
 SWITCHABLE(AspectRatio, true);
 SWITCHABLE(AstcDecodeMode, true);
+SWITCHABLE(AstcRecompression, true);
 SWITCHABLE(AudioMode, true);
 SWITCHABLE(CpuBackend, true);
 SWITCHABLE(CpuAccuracy, true);

src/common/settings.h

@@ -65,6 +65,7 @@ SWITCHABLE(AnisotropyMode, true);
 SWITCHABLE(AntiAliasing, false);
 SWITCHABLE(AspectRatio, true);
 SWITCHABLE(AstcDecodeMode, true);
+SWITCHABLE(AstcRecompression, true);
 SWITCHABLE(AudioMode, true);
 SWITCHABLE(CpuBackend, true);
 SWITCHABLE(CpuAccuracy, true);
@@ -259,8 +260,6 @@ struct Values {
                                                     Category::Cpu};
     SwitchableSetting<CpuAccuracy, true> cpu_accuracy{linkage, CpuAccuracy::Auto,
                                                       "cpu_accuracy", Category::Cpu};
-    SwitchableSetting<bool> relocate_36bit_address_space{
-        linkage, false, "relocate_36bit_address_space", Category::Cpu};
     SwitchableSetting<CpuClock> cpu_clock{linkage,
                                               CpuClock::Normal,
                                               "fast_cpu_time",
@@ -554,6 +553,22 @@ struct Values {
                                                             "accelerate_astc",
                                                             Category::RendererAdvanced};
 
+    SwitchableSetting<FramePacingMode, true> frame_pacing_mode{linkage,
+                                                               FramePacingMode::Target_Auto,
+                                                               FramePacingMode::Target_Auto,
+                                                               FramePacingMode::Target_120,
+                                                               "frame_pacing_mode",
+                                                               Category::RendererAdvanced,
+                                                               Specialization::Default,
+                                                               true,
+                                                               true};
+
+    SwitchableSetting<AstcRecompression, true> astc_recompression{linkage,
+                                                                  AstcRecompression::Uncompressed,
+                                                                  "astc_recompression",
+                                                                  Category::RendererAdvanced};
+
+
     SwitchableSetting<bool> sync_memory_operations{linkage,
                                                    false,
                                                    "sync_memory_operations",
@@ -561,8 +576,6 @@ struct Values {
                                                    Specialization::Default,
                                                    true,
                                                    true};
-    SwitchableSetting<bool> stall_on_gpu_fence_wait{linkage, true, "stall_on_gpu_fence_wait",
-                                                    Category::RendererAdvanced};
 
     SwitchableSetting<bool> renderer_force_max_clock{linkage, false, "force_max_clock",
                                                      Category::RendererAdvanced};
@@ -589,13 +602,8 @@ struct Values {
                                                   "use_reactive_flushing",
                                                   Category::RendererAdvanced};
 
-    SwitchableSetting<bool> barrier_feedback_loops{linkage,
-                                                   true,
-                                                   "barrier_feedback_loops",
-                                                   Category::RendererAdvanced,
-                                                   Specialization::Default,
-                                                   true,
-                                                   true};
+    SwitchableSetting<bool> barrier_feedback_loops{linkage, true, "barrier_feedback_loops",
+                                                   Category::RendererAdvanced};
 
     SwitchableSetting<bool> enable_buffer_history{linkage,
                                                   false,
@@ -633,7 +641,7 @@ struct Values {
                                                         true};
     SwitchableSetting<bool> async_presentation{linkage,
 #ifdef __ANDROID__
-                                               true,
+                                               false,
 #else
                                                false,
 #endif
@@ -658,8 +666,26 @@ struct Values {
     SwitchableSetting<bool> use_asynchronous_shaders{linkage, false, "use_asynchronous_shaders",
                                                      Category::RendererHacks};
 
-    SwitchableSetting<bool> use_unified_memory{linkage, false, "use_unified_memory",
-                                               Category::RendererHacks};
+    SwitchableSetting<GpuUnswizzleSize> gpu_unswizzle_texture_size{linkage,
+                                                  GpuUnswizzleSize::Large,
+                                                  "gpu_unswizzle_texture_size",
+                                                  Category::RendererHacks,
+                                                  Specialization::Default};
+
+    SwitchableSetting<GpuUnswizzle> gpu_unswizzle_stream_size{linkage,
+                                                  GpuUnswizzle::Medium,
+                                                  "gpu_unswizzle_stream_size",
+                                                  Category::RendererHacks,
+                                                  Specialization::Default};
+
+    SwitchableSetting<GpuUnswizzleChunk> gpu_unswizzle_chunk_size{linkage,
+                                                  GpuUnswizzleChunk::Medium,
+                                                  "gpu_unswizzle_chunk_size",
+                                                  Category::RendererHacks,
+                                                  Specialization::Default};
+
+    SwitchableSetting<bool> gpu_unswizzle_enabled{linkage, false, "gpu_unswizzle_enabled",
+                                                  Category::RendererHacks};
 
     SwitchableSetting<ExtendedDynamicState> dyna_state{linkage,
 #if defined(__ANDROID__)

src/common/settings_enums.h

@@ -130,6 +130,8 @@ ENUM(TimeZone, Auto, Default, Cet, Cst6Cdt, Cuba, Eet, Egypt, Eire, Est, Est5Edt
     Roc, Rok, Singapore, Turkey, Uct, Universal, Utc, WSu, Wet, Zulu);
 ENUM(AnisotropyMode, Automatic, Default, X2, X4, X8, X16);
 ENUM(AstcDecodeMode, Cpu, Gpu, CpuAsynchronous);
+ENUM(AstcRecompression, Uncompressed, Bc1, Bc3);
+ENUM(FramePacingMode, Target_Auto, Target_30, Target_60, Target_90, Target_120);
 ENUM(VSyncMode, Immediate, Mailbox, Fifo, FifoRelaxed);
 ENUM(VramUsageMode, Conservative, Aggressive);
 ENUM(RendererBackend, OpenGL_GLSL, Vulkan, Null, OpenGL_GLASM, OpenGL_SPIRV);
@@ -151,6 +153,9 @@ ENUM(ConsoleMode, Handheld, Docked);
 ENUM(AppletMode, HLE, LLE);
 ENUM(SpirvOptimizeMode, Never, OnLoad, Always);
 ENUM(GpuClock, Normal, Boost, Overclock)
+ENUM(GpuUnswizzleSize, VerySmall, Small, Normal, Large, VeryLarge)
+ENUM(GpuUnswizzle, VeryLow, Low, Normal, Medium, High)
+ENUM(GpuUnswizzleChunk, VeryLow, Low, Normal, Medium, High)
 ENUM(TemperatureUnits, Celsius, Fahrenheit)
 ENUM(ExtendedDynamicState, Disabled, EDS1, EDS2, EDS3);
 ENUM(GpuLogLevel, Off, Errors, Standard, Verbose, All)

src/common/spin_lock.h

@@ -1,50 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#pragma once
-
-#ifdef _MSC_VER
-#include <intrin.h>
-#elif defined(ARCHITECTURE_x86_64)
-#include <xmmintrin.h>
-#endif
-#include <atomic>
-
-namespace Common {
-
-/// @brief A lock similar to mutex that forces a thread to spin wait instead calling the
-/// supervisor. Should be used on short sequences of code.
-struct SpinLock {
-    SpinLock() noexcept = default;
-    SpinLock(const SpinLock&) noexcept = delete;
-    SpinLock& operator=(const SpinLock&) noexcept = delete;
-    SpinLock(SpinLock&&) noexcept = delete;
-    SpinLock& operator=(SpinLock&&) noexcept = delete;
-
-    inline void lock() noexcept {
-        while (lck.test_and_set(std::memory_order_acquire)) {
-#if defined(ARCHITECTURE_x86_64)
-            _mm_pause();
-#elif defined(ARCHITECTURE_arm64) && defined(_MSC_VER)
-            __yield();
-#elif defined(ARCHITECTURE_arm64)
-            asm("yield");
-#endif
-        }
-    }
-
-    inline void unlock() noexcept {
-        lck.clear(std::memory_order_release);
-    }
-
-    [[nodiscard]] inline bool try_lock() noexcept {
-        return !lck.test_and_set(std::memory_order_acquire);
-    }
-
-    std::atomic_flag lck = ATOMIC_FLAG_INIT;
-};
-
-} // namespace Common

src/common/uuid.cpp

@@ -208,9 +208,9 @@ UUID UUID::MakeRandomRFC4122V4() {
     return uuid;
 }
 
-UUID UUID::MakeRFC4122V5(std::span<u8, 20> sha1) {
+UUID UUID::MakeRFC4122V5(std::span<u8, 16> sha1) {
     UUID uuid{};
-    std::memcpy(&uuid.uuid, sha1.data(), sizeof(UUID));
+    std::memcpy(&uuid.uuid, sha1.data(), sha1.size());
     uuid.uuid[8] = 0x80 | (uuid.uuid[8] & 0x3F);
     uuid.uuid[6] = 0x50 | (uuid.uuid[6] & 0xF);
     return uuid;

src/common/uuid.h

@@ -104,7 +104,7 @@ struct UUID {
     /// @returns A random UUID that is RFC 4122 Version 4 compliant.
     [[nodiscard]] static UUID MakeRandomRFC4122V4();
 
-    [[nodiscard]] static UUID MakeRFC4122V5(std::span<u8, 20> sha1);
+    [[nodiscard]] static UUID MakeRFC4122V5(std::span<u8, 16> sha1);
 
     friend constexpr bool operator==(const UUID& lhs, const UUID& rhs) = default;
 };

src/core/CMakeLists.txt

@@ -796,6 +796,8 @@ add_library(core STATIC
     hle/service/ns/application_manager_interface.h
     hle/service/ns/application_version_interface.cpp
     hle/service/ns/application_version_interface.h
+    hle/service/ns/async_result.cpp
+    hle/service/ns/async_result.h
     hle/service/ns/content_management_interface.cpp
     hle/service/ns/content_management_interface.h
     hle/service/ns/develop_interface.cpp
@@ -810,8 +812,6 @@ add_library(core STATIC
     hle/service/ns/ecommerce_interface.h
     hle/service/ns/factory_reset_interface.cpp
     hle/service/ns/factory_reset_interface.h
-    hle/service/ns/i_async_result.cpp
-    hle/service/ns/i_async_result.h
     hle/service/ns/language.cpp
     hle/service/ns/language.h
     hle/service/ns/ns.cpp
@@ -1238,7 +1238,8 @@ if (HAS_NCE)
         arm/nce/interpreter_visitor.cpp
         arm/nce/interpreter_visitor.h
         arm/nce/patcher.cpp
-        arm/nce/patcher.h)
+        arm/nce/patcher.h
+        arm/nce/visitor_base.h)
     target_link_libraries(core PRIVATE merry::oaknut)
 endif()
 

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -22,21 +22,15 @@ DynarmicCallbacks32::DynarmicCallbacks32(ArmDynarmic32& parent, Kernel::KProcess
     , m_check_memory_access{m_debugger_enabled || !Settings::values.cpuopt_ignore_memory_aborts.GetValue()}
 {}
 
-u8 DynarmicCallbacks32::MemoryRead8(u32 vaddr) {
-    CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read8(vaddr);
-}
-u16 DynarmicCallbacks32::MemoryRead16(u32 vaddr) {
-    CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read16(vaddr);
-}
-u32 DynarmicCallbacks32::MemoryRead32(u32 vaddr) {
-    CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read32(vaddr);
-}
-u64 DynarmicCallbacks32::MemoryRead64(u32 vaddr) {
-    CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read64(vaddr);
+u64 DynarmicCallbacks32::MemoryRead(u32 vaddr, size_t size) {
+    CheckMemoryAccess(vaddr, size, Kernel::DebugWatchpointType::Read);
+    switch (size) {
+    case sizeof(u64): return m_memory.Read64(vaddr);
+    case sizeof(u32): return m_memory.Read32(vaddr);
+    case sizeof(u16): return m_memory.Read16(vaddr);
+    case sizeof(u8): return m_memory.Read8(vaddr);
+    default: UNREACHABLE();
+    }
 }
 
 std::optional<u32> DynarmicCallbacks32::MemoryReadCode(u32 vaddr) {
@@ -50,27 +44,17 @@ std::optional<u32> DynarmicCallbacks32::MemoryReadCode(u32 vaddr) {
     return cached_code_page.inst[(vaddr & Core::Memory::YUZU_PAGEMASK) / sizeof(u32)];
 }
 
-void DynarmicCallbacks32::MemoryWrite8(u32 vaddr, u8 value) {
-    if (CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write8(vaddr, value);
+void DynarmicCallbacks32::MemoryWrite(Dynarmic::A32::VAddr vaddr, u64 value, size_t size) {
+    if (CheckMemoryAccess(vaddr, size, Kernel::DebugWatchpointType::Write)) {
+        switch (size) {
+        case sizeof(u64): return m_memory.Write64(vaddr, value);
+        case sizeof(u32): return m_memory.Write32(vaddr, u32(value));
+        case sizeof(u16): return m_memory.Write16(vaddr, u16(value));
+        case sizeof(u8): return m_memory.Write8(vaddr, u8(value));
+        default: UNREACHABLE();
+        }
     }
 }
-void DynarmicCallbacks32::MemoryWrite16(u32 vaddr, u16 value) {
-    if (CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write16(vaddr, value);
-    }
-}
-void DynarmicCallbacks32::MemoryWrite32(u32 vaddr, u32 value) {
-    if (CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write32(vaddr, value);
-    }
-}
-void DynarmicCallbacks32::MemoryWrite64(u32 vaddr, u64 value) {
-    if (CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write64(vaddr, value);
-    }
-}
-
 bool DynarmicCallbacks32::MemoryWriteExclusive8(u32 vaddr, u8 value, u8 expected) {
     return CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write) &&
             m_memory.WriteExclusive8(vaddr, value, expected);

src/core/arm/dynarmic/arm_dynarmic_32.h

@@ -30,18 +30,12 @@ class System;
 class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
 public:
     explicit DynarmicCallbacks32(ArmDynarmic32& parent, Kernel::KProcess* process);
-    u8 MemoryRead8(u32 vaddr) override;
-    u16 MemoryRead16(u32 vaddr) override;
-    u32 MemoryRead32(u32 vaddr) override;
-    u64 MemoryRead64(u32 vaddr) override;
+    u64 MemoryRead(u32 vaddr, size_t size) override;
     std::optional<u32> MemoryReadCode(u32 vaddr) override;
     void InstructionSynchronizationBarrierRaised() override {
         last_code_addr = u64(-1); //reset back, force refetch
     }
-    void MemoryWrite8(u32 vaddr, u8 value) override;
-    void MemoryWrite16(u32 vaddr, u16 value) override;
-    void MemoryWrite32(u32 vaddr, u32 value) override;
-    void MemoryWrite64(u32 vaddr, u64 value) override;
+    void MemoryWrite(Dynarmic::A32::VAddr vaddr, u64 value, size_t size) override;
     bool MemoryWriteExclusive8(u32 vaddr, u8 value, u8 expected) override;
     bool MemoryWriteExclusive16(u32 vaddr, u16 value, u16 expected) override;
     bool MemoryWriteExclusive32(u32 vaddr, u32 value, u32 expected) override;

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -10,6 +10,7 @@
 #include "core/arm/dynarmic/dynarmic_exclusive_monitor.h"
 #include "core/core_timing.h"
 #include "core/hle/kernel/k_process.h"
+#include "dynarmic/interface/A64/config.h"
 
 namespace Core {
 
@@ -21,21 +22,15 @@ DynarmicCallbacks64::DynarmicCallbacks64(ArmDynarmic64& parent, Kernel::KProcess
     , m_check_memory_access{m_debugger_enabled || !Settings::values.cpuopt_ignore_memory_aborts.GetValue()}
 {}
 
-u8 DynarmicCallbacks64::MemoryRead8(u64 vaddr) {
-    CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read8(vaddr);
-}
-u16 DynarmicCallbacks64::MemoryRead16(u64 vaddr) {
-    CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read16(vaddr);
-}
-u32 DynarmicCallbacks64::MemoryRead32(u64 vaddr) {
-    CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read32(vaddr);
-}
-u64 DynarmicCallbacks64::MemoryRead64(u64 vaddr) {
-    CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read64(vaddr);
+u64 DynarmicCallbacks64::MemoryRead(u64 vaddr, size_t size) {
+    CheckMemoryAccess(vaddr, size, Kernel::DebugWatchpointType::Read);
+    switch (size) {
+    case sizeof(u64): return m_memory.Read64(vaddr);
+    case sizeof(u32): return m_memory.Read32(vaddr);
+    case sizeof(u16): return m_memory.Read16(vaddr);
+    case sizeof(u8): return m_memory.Read8(vaddr);
+    default: UNREACHABLE();
+    }
 }
 Dynarmic::A64::Vector DynarmicCallbacks64::MemoryRead128(u64 vaddr) {
     CheckMemoryAccess(vaddr, 16, Kernel::DebugWatchpointType::Read);
@@ -53,24 +48,15 @@ std::optional<u32> DynarmicCallbacks64::MemoryReadCode(u64 vaddr) {
     return cached_code_page.inst[(vaddr & Core::Memory::YUZU_PAGEMASK) / sizeof(u32)];
 }
 
-void DynarmicCallbacks64::MemoryWrite8(u64 vaddr, u8 value) {
-    if (CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write8(vaddr, value);
-    }
-}
-void DynarmicCallbacks64::MemoryWrite16(u64 vaddr, u16 value) {
-    if (CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write16(vaddr, value);
-    }
-}
-void DynarmicCallbacks64::MemoryWrite32(u64 vaddr, u32 value) {
-    if (CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write32(vaddr, value);
-    }
-}
-void DynarmicCallbacks64::MemoryWrite64(u64 vaddr, u64 value) {
-    if (CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write64(vaddr, value);
+void DynarmicCallbacks64::MemoryWrite(Dynarmic::A64::VAddr vaddr, u64 value, std::size_t size) {
+    if (CheckMemoryAccess(vaddr, size, Kernel::DebugWatchpointType::Write)) {
+        switch (size) {
+        case sizeof(u64): return m_memory.Write64(vaddr, u64(value));
+        case sizeof(u32): return m_memory.Write32(vaddr, u32(value));
+        case sizeof(u16): return m_memory.Write16(vaddr, u16(value));
+        case sizeof(u8): return m_memory.Write8(vaddr, u8(value));
+        default: UNREACHABLE();
+        }
     }
 }
 void DynarmicCallbacks64::MemoryWrite128(u64 vaddr, Dynarmic::A64::Vector value) {

src/core/arm/dynarmic/arm_dynarmic_64.h

@@ -16,6 +16,7 @@
 #include "common/hash.h"
 #include "core/arm/arm_interface.h"
 #include "core/arm/dynarmic/dynarmic_exclusive_monitor.h"
+#include "dynarmic/interface/A64/config.h"
 
 namespace Core::Memory {
 class Memory;
@@ -36,19 +37,13 @@ class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
     explicit DynarmicCallbacks64(ArmDynarmic64& parent, Kernel::KProcess* process);
 
-    u8 MemoryRead8(u64 vaddr) override;
-    u16 MemoryRead16(u64 vaddr) override;
-    u32 MemoryRead32(u64 vaddr) override;
-    u64 MemoryRead64(u64 vaddr) override;
+    u64 MemoryRead(u64 vaddr, size_t size) override;
     Dynarmic::A64::Vector MemoryRead128(u64 vaddr) override;
     std::optional<u32> MemoryReadCode(u64 vaddr) override;
     void InstructionSynchronizationBarrierRaised() override {
         last_code_addr = u64(-1); //reset back, force refetch
     }
-    void MemoryWrite8(u64 vaddr, u8 value) override;
-    void MemoryWrite16(u64 vaddr, u16 value) override;
-    void MemoryWrite32(u64 vaddr, u32 value) override;
-    void MemoryWrite64(u64 vaddr, u64 value) override;
+    void MemoryWrite(Dynarmic::A64::VAddr vaddr, u64 value, std::size_t size) override;
     void MemoryWrite128(u64 vaddr, Dynarmic::A64::Vector value) override;
     bool MemoryWriteExclusive8(u64 vaddr, std::uint8_t value, std::uint8_t expected) override;
     bool MemoryWriteExclusive16(u64 vaddr, std::uint16_t value, std::uint16_t expected) override;

src/core/arm/nce/interpreter_visitor.h

@@ -7,77 +7,97 @@
 
 #pragma once
 
-#include <optional>
-#include <span>
+#include <atomic>
 #include <signal.h>
+#include <unistd.h>
+#include <span>
 
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wshadow"
-#include <dynarmic/frontend/A64/a64_types.h>
-#include <dynarmic/frontend/imm.h>
-#pragma GCC diagnostic pop
-
+#include "core/hle/kernel/k_thread.h"
 #include "core/memory.h"
+#include "common/logging.h"
+#include "core/arm/nce/visitor_base.h"
 
 namespace Core {
 
-class InterpreterVisitor {
+namespace Memory {
+class Memory;
+}
+
+class InterpreterVisitor final : public VisitorBase {
 public:
     explicit InterpreterVisitor(Core::Memory::Memory& memory, std::span<u64, 31> regs,
                                 std::span<u128, 32> fpsimd_regs, u64& sp, const u64& pc)
         : m_memory(memory), m_regs(regs), m_fpsimd_regs(fpsimd_regs), m_sp(sp), m_pc(pc) {}
+    ~InterpreterVisitor() override = default;
 
-    bool Execute(u32 inst);
+    enum class MemOp {
+        Load,
+        Store,
+        Prefetch,
+    };
+
+    u128 GetVec(Vec v);
+    u64 GetReg(Reg r);
+    u64 GetSp();
+    u64 GetPc();
+
+    void SetVec(Vec v, u128 value);
+    void SetReg(Reg r, u64 value);
+    void SetSp(u64 value);
+
+    u64 ExtendReg(size_t bitsize, Reg reg, Imm<3> option, u8 shift);
+
+    // Loads and stores - Load/Store Exclusive
+    bool Ordered(size_t size, bool L, bool o0, Reg Rn, Reg Rt);
+    bool STLLR(Imm<2> size, Reg Rn, Reg Rt) override;
+    bool STLR(Imm<2> size, Reg Rn, Reg Rt) override;
+    bool LDLAR(Imm<2> size, Reg Rn, Reg Rt) override;
+    bool LDAR(Imm<2> size, Reg Rn, Reg Rt) override;
+
+    // Loads and stores - Load register (literal)
+    bool LDR_lit_gen(bool opc_0, Imm<19> imm19, Reg Rt) override;
+    bool LDR_lit_fpsimd(Imm<2> opc, Imm<19> imm19, Vec Vt) override;
+
+    // Loads and stores - Load/Store register pair
+    bool STP_LDP_gen(Imm<2> opc, bool not_postindex, bool wback, Imm<1> L, Imm<7> imm7, Reg Rt2,
+                     Reg Rn, Reg Rt) override;
+    bool STP_LDP_fpsimd(Imm<2> opc, bool not_postindex, bool wback, Imm<1> L, Imm<7> imm7, Vec Vt2,
+                        Reg Rn, Vec Vt) override;
+
+    // Loads and stores - Load/Store register (immediate)
+    bool RegisterImmediate(bool wback, bool postindex, size_t scale, u64 offset, Imm<2> size,
+                           Imm<2> opc, Reg Rn, Reg Rt);
+    bool STRx_LDRx_imm_1(Imm<2> size, Imm<2> opc, Imm<9> imm9, bool not_postindex, Reg Rn,
+                         Reg Rt) override;
+    bool STRx_LDRx_imm_2(Imm<2> size, Imm<2> opc, Imm<12> imm12, Reg Rn, Reg Rt) override;
+    bool STURx_LDURx(Imm<2> size, Imm<2> opc, Imm<9> imm9, Reg Rn, Reg Rt) override;
+
+    bool SIMDImmediate(bool wback, bool postindex, size_t scale, u64 offset, MemOp memop, Reg Rn,
+                       Vec Vt);
+    bool STR_imm_fpsimd_1(Imm<2> size, Imm<1> opc_1, Imm<9> imm9, bool not_postindex, Reg Rn,
+                          Vec Vt) override;
+    bool STR_imm_fpsimd_2(Imm<2> size, Imm<1> opc_1, Imm<12> imm12, Reg Rn, Vec Vt) override;
+    bool LDR_imm_fpsimd_1(Imm<2> size, Imm<1> opc_1, Imm<9> imm9, bool not_postindex, Reg Rn,
+                          Vec Vt) override;
+    bool LDR_imm_fpsimd_2(Imm<2> size, Imm<1> opc_1, Imm<12> imm12, Reg Rn, Vec Vt) override;
+    bool STUR_fpsimd(Imm<2> size, Imm<1> opc_1, Imm<9> imm9, Reg Rn, Vec Vt) override;
+    bool LDUR_fpsimd(Imm<2> size, Imm<1> opc_1, Imm<9> imm9, Reg Rn, Vec Vt) override;
+
+    // Loads and stores - Load/Store register (register offset)
+    bool RegisterOffset(size_t scale, u8 shift, Imm<2> size, Imm<1> opc_1, Imm<1> opc_0, Reg Rm,
+                        Imm<3> option, Reg Rn, Reg Rt);
+    bool STRx_reg(Imm<2> size, Imm<1> opc_1, Reg Rm, Imm<3> option, bool S, Reg Rn,
+                  Reg Rt) override;
+    bool LDRx_reg(Imm<2> size, Imm<1> opc_1, Reg Rm, Imm<3> option, bool S, Reg Rn,
+                  Reg Rt) override;
+
+    bool SIMDOffset(size_t scale, u8 shift, Imm<1> opc_0, Reg Rm, Imm<3> option, Reg Rn, Vec Vt);
+    bool STR_reg_fpsimd(Imm<2> size, Imm<1> opc_1, Reg Rm, Imm<3> option, bool S, Reg Rn,
+                        Vec Vt) override;
+    bool LDR_reg_fpsimd(Imm<2> size, Imm<1> opc_1, Reg Rm, Imm<3> option, bool S, Reg Rn,
+                        Vec Vt) override;
 
 private:
-    template <size_t BitSize>
-    using Imm = Dynarmic::Imm<BitSize>;
-    using Reg = Dynarmic::A64::Reg;
-    using Vec = Dynarmic::A64::Vec;
-
-    u128 GetVec(Vec v) const {
-        return m_fpsimd_regs[static_cast<u32>(v)];
-    }
-    void SetVec(Vec v, u128 value) {
-        m_fpsimd_regs[static_cast<u32>(v)] = value;
-    }
-    u64 GetReg(Reg r) const {
-        return m_regs[static_cast<u32>(r)];
-    }
-    void SetReg(Reg r, u64 value) {
-        m_regs[static_cast<u32>(r)] = value;
-    }
-    u64 GetRegSp(Reg r) const {
-        if (r == Reg::SP) {
-            return m_sp;
-        }
-        return m_regs[static_cast<u32>(r)];
-    }
-    void SetRegSp(Reg r, u64 value) {
-        if (r == Reg::SP) {
-            m_sp = value;
-            return;
-        }
-        m_regs[static_cast<u32>(r)] = value;
-    }
-
-    u64 ExtendReg(Reg reg, Imm<3> option, u8 shift);
-
-    bool Ordered(size_t size, bool load, Reg Rn, Reg Rt);
-    bool LoadLiteral(bool wide, Imm<19> imm19, Reg Rt);
-    bool LoadLiteralSimd(Imm<2> opc, Imm<19> imm19, Vec Vt);
-    bool Pair(Imm<2> opc, bool not_postindex, bool wback, bool load, Imm<7> imm7, Reg Rt2, Reg Rn,
-              Reg Rt);
-    bool PairSimd(Imm<2> opc, bool not_postindex, bool wback, bool load, Imm<7> imm7, Vec Vt2,
-                  Reg Rn, Vec Vt);
-    bool RegisterImmediate(bool wback, bool postindex, u64 offset, Imm<2> size, Imm<2> opc, Reg Rn,
-                           Reg Rt);
-    bool RegisterOffset(bool S, Imm<2> size, Imm<1> opc_1, Imm<1> opc_0, Reg Rm, Imm<3> option,
-                        Reg Rn, Reg Rt);
-    bool SimdImmediate(bool wback, bool postindex, size_t scale, u64 offset, bool load, Reg Rn,
-                       Vec Vt);
-    bool SimdOffset(size_t scale, bool S, bool load, Reg Rm, Imm<3> option, Reg Rn, Vec Vt);
-
     Core::Memory::Memory& m_memory;
     std::span<u64, 31> m_regs;
     std::span<u128, 32> m_fpsimd_regs;

src/core/arm/nce/patcher.cpp

@@ -371,9 +371,14 @@ size_t Patcher::GetPreSectionSize() const noexcept {
 }
 
 void Patcher::WriteLoadContext(oaknut::VectorCodeGenerator& cg) {
+    // This function was called, which modifies X30, so use that as a scratch register.
+    // SP contains the guest X30, so save our return X30 to SP + 8, since we have allocated 16 bytes
+    // of stack.
     cg.STR(X30, SP, 8);
-    cg.LDR(X30, SP, 0);
+    cg.MRS(X30, oaknut::SystemReg::TPIDR_EL0);
+    cg.LDR(X30, X30, offsetof(NativeExecutionParameters, native_context));
 
+    // Load system registers.
     cg.LDR(W0, X30, offsetof(GuestContext, fpsr));
     cg.MSR(oaknut::SystemReg::FPSR, X0);
     cg.LDR(W0, X30, offsetof(GuestContext, fpcr));
@@ -381,9 +386,7 @@ void Patcher::WriteLoadContext(oaknut::VectorCodeGenerator& cg) {
     cg.LDR(W0, X30, offsetof(GuestContext, nzcv));
     cg.MSR(oaknut::SystemReg::NZCV, X0);
 
-    cg.LDR(X0, X30, 8 * 30);
-    cg.STR(X0, SP, 0);
-
+    // Load all vector registers.
     static constexpr size_t VEC_OFF = offsetof(GuestContext, vector_registers);
     for (int i = 0; i <= 30; i += 2) {
         cg.LDP(oaknut::QReg{i}, oaknut::QReg{i + 1}, X30, VEC_OFF + 16 * i);
@@ -433,7 +436,7 @@ void Patcher::WriteSaveContext(oaknut::VectorCodeGenerator& cg) {
     cg.STR(W0, X30, offsetof(GuestContext, nzcv));
     cg.LDR(X0, SP, POST_INDEXED, 16);
 
-    cg.MOV(X1, X30);
+    // Reload our return X30 from the stack, and return.
     cg.LDR(X30, SP, 8);
     cg.RET();
 }
@@ -454,6 +457,8 @@ void Patcher::WriteSvcTrampoline(ModuleDestLabel module_dest, u32 svc_id, oaknut
     // Now that we've saved all registers, we can use any registers as scratch.
     // Store PC + 4 to arm interface, since we know the instruction offset from the entry point.
     oaknut::Label pc_after_svc;
+    cg.MRS(X1, oaknut::SystemReg::TPIDR_EL0);
+    cg.LDR(X1, X1, offsetof(NativeExecutionParameters, native_context));
     cg.LDR(X2, pc_after_svc);
     cg.STR(X2, X1, offsetof(GuestContext, pc));
 
@@ -509,13 +514,24 @@ void Patcher::WriteSvcTrampoline(ModuleDestLabel module_dest, u32 svc_id, oaknut
 
     // Host called this location. Save the return address so we can
     // unwind the stack properly when jumping back.
-    cg.ADD(X0, X1, offsetof(GuestContext, host_ctx));
+    cg.MRS(X2, oaknut::SystemReg::TPIDR_EL0);
+    cg.LDR(X2, X2, offsetof(NativeExecutionParameters, native_context));
+    cg.ADD(X0, X2, offsetof(GuestContext, host_ctx));
     cg.STR(X30, X0, offsetof(HostContext, host_saved_regs) + 11 * sizeof(u64));
 
-    cg.STR(X1, SP, PRE_INDEXED, -16);
+    // Reload all guest registers except X30 and PC.
+    // The function also expects 16 bytes of stack already allocated.
+    cg.STR(X30, SP, PRE_INDEXED, -16);
     cg.BL(load_ctx);
     cg.LDR(X30, SP, POST_INDEXED, 16);
 
+    // Use X1 as a scratch register to restore X30.
+    cg.STR(X1, SP, PRE_INDEXED, -16);
+    cg.MRS(X1, oaknut::SystemReg::TPIDR_EL0);
+    cg.LDR(X1, X1, offsetof(NativeExecutionParameters, native_context));
+    cg.LDR(X30, X1, offsetof(GuestContext, cpu_registers) + sizeof(u64) * 30);
+    cg.LDR(X1, SP, POST_INDEXED, 16);
+
     // Unlock the context.
     this->UnlockContext(cg);
 

src/core/device_memory.h

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 

src/core/device_memory_manager.h

@@ -20,10 +20,6 @@
 #include "common/scratch_buffer.h"
 #include "common/virtual_buffer.h"
 
-namespace Common {
-class HostMemory;
-}
-
 namespace Core {
 
 constexpr size_t DEVICE_PAGEBITS = 12ULL;
@@ -130,10 +126,6 @@ public:
     // New batch API to update multiple ranges with a single lock acquisition.
     void UpdatePagesCachedBatch(std::span<const std::pair<DAddr, size_t>> ranges, s32 delta);
 
-    const Common::HostMemory& GetHostMemory() const noexcept {
-        return host_memory;
-    }
-
 private:
     struct TranslationEntry {
         DAddr guest_page{};
@@ -179,7 +171,6 @@ private:
     std::unique_ptr<DeviceMemoryManagerAllocator<Traits>> impl;
 
     const uintptr_t physical_base;
-    const Common::HostMemory& host_memory;
     DeviceInterface* device_inter;
 
     struct TrackedEntry {

src/core/device_memory_manager.inc

@@ -171,7 +171,6 @@ struct DeviceMemoryManagerAllocator {
 template <typename Traits>
 DeviceMemoryManager<Traits>::DeviceMemoryManager(const DeviceMemory& device_memory_)
     : physical_base{uintptr_t(device_memory_.buffer.BackingBasePointer())}
-    , host_memory{device_memory_.buffer}
     , device_inter{nullptr}
     , compressed_device_addr(1ULL << ((Settings::values.memory_layout_mode.GetValue() == Settings::MemoryLayout::Memory_4Gb ? physical_min_bits : physical_max_bits) - Memory::YUZU_PAGEBITS))
     , tracked_entries(device_as_size >> Memory::YUZU_PAGEBITS)

src/core/file_sys/ips_layer.cpp

@@ -100,8 +100,7 @@ VirtualFile PatchIPS(const VirtualFile& in, const VirtualFile& ips) {
 
 
 struct IPSwitchRecord {
-    std::array<uint8_t, 256 - sizeof(size_t)> data;
-    size_t count;
+    std::vector<uint8_t> data;
 };
 struct IPSwitchCompiler::IPSwitchPatch {
     ::Common::unordered_map<u32, IPSwitchRecord> records;
@@ -122,22 +121,23 @@ static IPSwitchRecord EscapeStringSequences(std::string_view sv) {
     IPSwitchRecord r{};
     for (auto it = sv.cbegin(); it < sv.cend(); ) {
         if (*it == '\\' && it + 1 < sv.cend()) {
-            switch (it[1]) {
-            case 'a': r.data[r.count] = '\a'; break;
-            case 'b': r.data[r.count] = '\b'; break;
-            case 'e': r.data[r.count] = '\e'; break;
-            case 'f': r.data[r.count] = '\f'; break;
-            case 'n': r.data[r.count] = '\n'; break;
-            case 'r': r.data[r.count] = '\r'; break;
-            case 't': r.data[r.count] = '\t'; break;
-            case 'v': r.data[r.count] = '\v'; break;
-            case '?': r.data[r.count] = '\?'; break;
-            default: r.data[r.count] = it[1]; break;
-            }
-            ++r.count;
+            r.data.push_back([it]() {
+                switch (it[1]) {
+                case 'a': return '\a';
+                case 'b': return '\b';
+                case 'e': return '\e';
+                case 'f': return '\f';
+                case 'n': return '\n';
+                case 'r': return '\r';
+                case 't': return '\t';
+                case 'v': return '\v';
+                case '?': return '\?';
+                default: return it[1];
+                }
+            }());
             it += 2;
         } else {
-            ++r.count;
+            r.data.push_back(*it);
             ++it;
         }
     }
@@ -223,8 +223,8 @@ void IPSwitchCompiler::Parse(std::span<u8 const> bytes) {
                 if (start <= line.cend() && end <= line.cend()) {
                     // Actually IPS wants ordering from {lsb, ..., msb} -- so LE and BE are inverted, fun!
                     auto const hs = Common::HexStringToVector({start, end}, is_little_endian);
+                    r.data.resize(hs.size());
                     std::memcpy(r.data.data(), hs.data(), hs.size());
-                    r.count = hs.size();
                     LOG_INFO(Loader, "[H] value @ {:#08X}", offset);
                     patches.back().records.insert_or_assign(u32(offset), std::move(r));
                 } else {
@@ -293,7 +293,7 @@ VirtualFile IPSwitchCompiler::Apply(const VirtualFile& in) const {
         if (patch.enabled) {
             for (const auto& record : patch.records) {
                 if (record.first < in_data.size()) {
-                    auto replace_size = record.second.count;
+                    auto replace_size = record.second.data.size();
                     if (record.first + replace_size > in_data.size())
                         replace_size = in_data.size() - record.first;
                     std::memcpy(in_data.data() + record.first, record.second.data.data(), replace_size);

src/core/file_sys/patch_manager.cpp

@@ -9,6 +9,7 @@
 #include <cstddef>
 #include <cstring>
 
+#include "common/assert.h"
 #include "common/hex_util.h"
 #include "common/logging.h"
 #include "common/settings.h"
@@ -156,12 +157,15 @@ std::string GetUpdateVersionStringFromSlot(const ContentProvider* provider, u64
     NACP nacp{nacp_file};
     return nacp.GetVersionString();
 }
+YUZU_NO_INLINE std::optional<u64> GetParentApplicationId(const ContentProvider* provider, u64 title_id) {
+    return provider == nullptr ? std::nullopt : provider->GetParentApplicationId(title_id);
+}
 } // Anonymous namespace
 
 PatchManager::PatchManager(u64 title_id_,
                            const Service::FileSystem::FileSystemController& fs_controller_,
                            const ContentProvider& content_provider_)
-    : title_id{title_id_}, fs_controller{fs_controller_}, content_provider{content_provider_} {}
+    : title_id{title_id_}, parent_title_id{GetParentApplicationId(std::addressof(content_provider_), title_id_)}, fs_controller{fs_controller_}, content_provider{content_provider_} {}
 
 PatchManager::~PatchManager() = default;
 
@@ -169,13 +173,35 @@ u64 PatchManager::GetTitleID() const {
     return title_id;
 }
 
+VirtualDir PatchManager::GetModificationLoadRoot(bool sdmc) const {
+    const auto get_root = [&](u64 id) {
+        return sdmc ? fs_controller.GetSDMCModificationLoadRoot(id) : fs_controller.GetModificationLoadRoot(id);
+    };
+    auto root = get_root(title_id);
+    if (!parent_title_id) {
+        return root;
+    }
+    std::vector<VirtualDir> roots{std::move(root), get_root(*parent_title_id)};
+    std::erase(roots, nullptr);
+    return LayeredVfsDirectory::MakeLayeredDirectory(std::move(roots));
+}
+
+std::vector<std::string> PatchManager::GetDisabledAddons() const {
+    auto disabled = Settings::values.disabled_addons[title_id];
+    if (parent_title_id) {
+        const auto& shared = Settings::values.disabled_addons[*parent_title_id];
+        disabled.insert(disabled.end(), shared.begin(), shared.end());
+    }
+    return disabled;
+}
+
 VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
     LOG_INFO(Loader, "Patching ExeFS for title_id={:016X}", title_id);
 
     if (exefs == nullptr)
         return exefs;
 
-    const auto& disabled = Settings::values.disabled_addons[title_id];
+    const auto disabled = GetDisabledAddons();
 
     bool update_disabled = true;
     std::optional<u32> enabled_version;
@@ -303,8 +329,8 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
     }
 
     // LayeredExeFS
-    const auto load_dir = fs_controller.GetModificationLoadRoot(title_id);
-    const auto sdmc_load_dir = fs_controller.GetSDMCModificationLoadRoot(title_id);
+    const auto load_dir = GetModificationLoadRoot();
+    const auto sdmc_load_dir = GetModificationLoadRoot(true);
 
     std::vector<VirtualDir> patch_dirs = {sdmc_load_dir};
     if (load_dir != nullptr) {
@@ -346,7 +372,7 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
 }
 
 std::vector<VirtualFile> PatchManager::CollectPatches(const std::vector<VirtualDir>& patch_dirs, const std::string& build_id) const {
-    const auto& disabled = Settings::values.disabled_addons[title_id];
+    const auto disabled = GetDisabledAddons();
     const auto nso_build_id = fmt::format("{:0<64}", build_id);
 
     std::vector<VirtualFile> out;
@@ -405,7 +431,7 @@ std::vector<u8> PatchManager::PatchNSO(const std::vector<u8>& nso, const std::st
 
     LOG_INFO(Loader, "Patching NSO for name={}, build_id={}", name, build_id);
 
-    const auto load_dir = fs_controller.GetModificationLoadRoot(title_id);
+    const auto load_dir = GetModificationLoadRoot();
     if (load_dir == nullptr) {
         LOG_ERROR(Loader, "Cannot load mods for invalid title_id={:016X}", title_id);
         return nso;
@@ -448,7 +474,7 @@ bool PatchManager::HasNSOPatch(const BuildID& build_id_, std::string_view name)
 
     LOG_INFO(Loader, "Querying NSO patch existence for build_id={}, name={}", build_id, name);
 
-    const auto load_dir = fs_controller.GetModificationLoadRoot(title_id);
+    const auto load_dir = GetModificationLoadRoot();
     if (load_dir == nullptr) {
         LOG_ERROR(Loader, "Cannot load mods for invalid title_id={:016X}", title_id);
         return false;
@@ -462,13 +488,13 @@ bool PatchManager::HasNSOPatch(const BuildID& build_id_, std::string_view name)
 }
 
 std::vector<Core::Memory::CheatEntry> PatchManager::CreateCheatList(const BuildID& build_id_) const {
-    const auto load_dir = fs_controller.GetModificationLoadRoot(title_id);
+    const auto load_dir = GetModificationLoadRoot();
     if (load_dir == nullptr) {
         LOG_ERROR(Loader, "Cannot load mods for invalid title_id={:016X}", title_id);
         return {};
     }
 
-    const auto& disabled = Settings::values.disabled_addons[title_id];
+    const auto disabled = GetDisabledAddons();
     auto patch_dirs = load_dir->GetSubdirectories();
     std::sort(patch_dirs.begin(), patch_dirs.end(), [](auto const& l, auto const& r) { return l->GetName() < r->GetName(); });
 
@@ -502,17 +528,16 @@ std::vector<Core::Memory::CheatEntry> PatchManager::CreateCheatList(const BuildI
     return out;
 }
 
-static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType type,
-                           const Service::FileSystem::FileSystemController& fs_controller) {
-    const auto load_dir = fs_controller.GetModificationLoadRoot(title_id);
-    const auto sdmc_load_dir = fs_controller.GetSDMCModificationLoadRoot(title_id);
+void PatchManager::ApplyLayeredFS(VirtualFile& romfs, ContentRecordType type) const {
+    const auto load_dir = GetModificationLoadRoot();
+    const auto sdmc_load_dir = GetModificationLoadRoot(true);
     if ((type != ContentRecordType::Program && type != ContentRecordType::Data &&
          type != ContentRecordType::HtmlDocument) ||
         (load_dir == nullptr && sdmc_load_dir == nullptr)) {
         return;
     }
 
-    const auto& disabled = Settings::values.disabled_addons[title_id];
+    const auto disabled = GetDisabledAddons();
     std::vector<VirtualDir> patch_dirs = load_dir->GetSubdirectories();
     if (std::find(disabled.cbegin(), disabled.cend(), "SDMC") == disabled.cend()) {
         patch_dirs.push_back(sdmc_load_dir);
@@ -591,7 +616,7 @@ VirtualFile PatchManager::PatchRomFS(const NCA* base_nca, VirtualFile base_romfs
 
     // Game Updates
     const auto update_tid = GetUpdateTitleID(title_id);
-    const auto& disabled = Settings::values.disabled_addons[title_id];
+    const auto disabled = GetDisabledAddons();
 
     bool update_disabled = true;
     std::optional<u32> enabled_version;
@@ -698,7 +723,7 @@ VirtualFile PatchManager::PatchRomFS(const NCA* base_nca, VirtualFile base_romfs
 
     // LayeredFS
     if (apply_layeredfs) {
-        ApplyLayeredFS(romfs, title_id, type, fs_controller);
+        ApplyLayeredFS(romfs, type);
     }
 
     return romfs;
@@ -1072,15 +1097,22 @@ std::vector<Patch> PatchManager::GetPatches(VirtualFile update_raw) const {
 std::optional<u32> PatchManager::GetGameVersion() const {
     const auto update_tid = GetUpdateTitleID(title_id);
     if (content_provider.HasEntry(update_tid, ContentRecordType::Program)) {
-        return content_provider.GetEntryVersion(update_tid);
+        const auto version = content_provider.GetEntryVersion(update_tid);
+        return version || !parent_title_id ? version : content_provider.GetEntryVersion(GetUpdateTitleID(*parent_title_id));
     }
 
-    return content_provider.GetEntryVersion(title_id);
+    const auto version = content_provider.GetEntryVersion(title_id);
+    return version || !parent_title_id ? version : content_provider.GetEntryVersion(*parent_title_id);
 }
 
 PatchManager::Metadata PatchManager::GetControlMetadata() const {
     const auto base_control_nca = content_provider.GetEntry(title_id, ContentRecordType::Control);
     if (base_control_nca == nullptr) {
+        if (parent_title_id) {
+            const auto control_id = content_provider.HasEntry(*parent_title_id, ContentRecordType::Control) ? *parent_title_id : GetUpdateTitleID(*parent_title_id);
+            const PatchManager parent{control_id, fs_controller, content_provider};
+            return parent.GetControlMetadata();
+        }
         return {};
     }
 

src/core/file_sys/patch_manager.h

@@ -109,10 +109,14 @@ public:
     [[nodiscard]] static PatchManager::Metadata GetMetadataFromBaseOrUpdate(Core::System& system, u64 application_id) noexcept;
 
 private:
+    [[nodiscard]] VirtualDir GetModificationLoadRoot(bool sdmc = false) const;
+    [[nodiscard]] std::vector<std::string> GetDisabledAddons() const;
+    void ApplyLayeredFS(VirtualFile& romfs, ContentRecordType type) const;
     [[nodiscard]] std::vector<VirtualFile> CollectPatches(const std::vector<VirtualDir>& patch_dirs,
                                                           const std::string& build_id) const;
 
     u64 title_id;
+    std::optional<u64> parent_title_id;
     const Service::FileSystem::FileSystemController& fs_controller;
     const ContentProvider& content_provider;
 };

src/core/file_sys/program_metadata.cpp

@@ -147,10 +147,6 @@ ProgramAddressSpaceType ProgramMetadata::GetAddressSpaceType() const {
     return npdm_header.address_space_type;
 }
 
-void ProgramMetadata::SetAddressSpaceType(ProgramAddressSpaceType address_space) {
-    npdm_header.address_space_type.Assign(address_space);
-}
-
 u8 ProgramMetadata::GetMainThreadPriority() const {
     return npdm_header.main_thread_priority;
 }

src/core/file_sys/program_metadata.h

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -76,7 +73,6 @@ public:
 
     bool Is64BitProgram() const;
     ProgramAddressSpaceType GetAddressSpaceType() const;
-    void SetAddressSpaceType(ProgramAddressSpaceType address_space);
     u8 GetMainThreadPriority() const;
     u8 GetMainThreadCore() const;
     u32 GetMainThreadStackSize() const;

src/core/file_sys/registered_cache.cpp

@@ -5,6 +5,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <algorithm>
+#include <limits>
 #include <random>
 #include <regex>
 #include <openssl/evp.h>
@@ -348,6 +349,22 @@ std::vector<ContentProviderEntry> ContentProvider::ListEntries() const {
     return ListEntriesFilter(std::nullopt, std::nullopt, std::nullopt);
 }
 
+std::optional<u64> ContentProvider::GetParentApplicationId(u64 program_id) const {
+    const auto application_id = GetBaseTitleID(program_id);
+    const auto program_index = program_id - application_id;
+    if (program_index == 0 || program_index > std::numeric_limits<u8>::max()) {
+        return std::nullopt;
+    }
+    if (!ListEntriesFilter(TitleType::Application, ContentRecordType::Meta, program_id).empty()) {
+        return std::nullopt;
+    }
+    if ((!ListEntriesFilter(TitleType::Application, ContentRecordType::Meta, application_id).empty() && HasEntry(program_id, ContentRecordType::Program))
+     || (!ListEntriesFilter(TitleType::Update, ContentRecordType::Meta, GetUpdateTitleID(application_id)).empty() && HasEntry(GetUpdateTitleID(program_id), ContentRecordType::Program))) {
+        return application_id;
+    }
+    return std::nullopt;
+}
+
 PlaceholderCache::PlaceholderCache(VirtualDir dir_) : dir(std::move(dir_)) {}
 
 bool PlaceholderCache::Create(const NcaID& id, u64 size) const {

src/core/file_sys/registered_cache.h

@@ -100,6 +100,8 @@ public:
         std::optional<TitleType> title_type = {}, std::optional<ContentRecordType> record_type = {},
         std::optional<u64> title_id = {}) const = 0;
 
+    [[nodiscard]] std::optional<u64> GetParentApplicationId(u64 program_id) const;
+
 protected:
     // A single instance of KeyManager to be used by GetEntry()
     Core::Crypto::KeyManager& keys = Core::Crypto::KeyManager::Instance();

src/core/file_sys/savedata_factory.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 

src/core/frontend/emu_window.h

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: 2014 Citra Emulator Project
@@ -74,7 +74,7 @@ public:
     };
 
     /// Called from GPU thread when a frame is displayed.
-    virtual void OnFrameDisplayed([[maybe_unused]] u32 presented_frames) {}
+    virtual void OnFrameDisplayed() {}
 
     /**
      * Returns a GraphicsContext that the frontend provides to be used for rendering.

src/core/hle/kernel/k_scheduler.cpp

@@ -465,6 +465,10 @@ void KScheduler::ScheduleImplFiber(KernelCore& kernel) {
         // Check if we need scheduling. If we do, then we can't complete the switch and should
         // retry.
         if (m_state.needs_scheduling.load(std::memory_order_seq_cst)) {
+            // Some libc++ lazily init mutex
+            [[maybe_unused]] auto const can_lock = highest_priority_thread->m_context_guard.try_lock();
+            DEBUG_ASSERT(!can_lock);
+
             // Our switch failed.
             // We should unlock the thread context, and then retry.
             highest_priority_thread->m_context_guard.unlock();
@@ -496,6 +500,10 @@ void KScheduler::Unload(KernelCore& kernel, KThread* thread) {
 
     // Check if the thread is terminated by checking the DPC flags.
     if ((thread->GetStackParameters().dpc_flags & static_cast<u32>(DpcFlag::Terminated)) == 0) {
+        // Some libc++ lazily init mutex
+        [[maybe_unused]] auto const can_lock = thread->m_context_guard.try_lock();
+        DEBUG_ASSERT(!can_lock);
+
         // The thread isn't terminated, so we want to unlock it.
         thread->m_context_guard.unlock();
     }

src/core/hle/kernel/k_slab_heap.h

@@ -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 2020 yuzu Emulator Project
@@ -12,7 +12,6 @@
 #include "common/atomic_ops.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
-#include "common/spin_lock.h"
 
 namespace Kernel {
 
@@ -30,7 +29,7 @@ public:
     };
 
 public:
-    constexpr KSlabHeapImpl() = default;
+    KSlabHeapImpl() = default;
 
     void Initialize() {
         ASSERT(m_head == nullptr);
@@ -68,7 +67,7 @@ public:
 
 private:
     std::atomic<Node*> m_head{};
-    Common::SpinLock m_lock;
+    std::mutex m_lock;
 };
 
 } // namespace impl

src/core/hle/kernel/k_thread.h

@@ -19,7 +19,6 @@
 
 #include "common/intrusive_red_black_tree.h"
 #include "common/scratch_buffer.h"
-#include "common/spin_lock.h"
 #include "core/arm/arm_interface.h"
 #include "core/hle/kernel/k_affinity_mask.h"
 #include "core/hle/kernel/k_light_lock.h"
@@ -920,7 +919,7 @@ private:
     bool m_resource_limit_release_hint{};
     bool m_is_kernel_address_key{};
     StackParameters m_stack_parameters{};
-    Common::SpinLock m_context_guard{};
+    std::mutex m_context_guard{};
 
     // For emulation
     std::shared_ptr<Common::Fiber> m_host_context{};

src/core/hle/service/acc/profile_manager.h

@@ -23,6 +23,12 @@ constexpr std::size_t profile_username_size{32};
 using ProfileUsername = std::array<u8, profile_username_size>;
 using UserIDArray = std::array<Common::UUID, MAX_USERS>;
 
+// This is nn::account::Uid
+struct Uid {
+    std::array<u8, 0x10> unk0;
+};
+static_assert(sizeof(Uid) == 0x10);
+
 /// Contains extra data related to a user.
 /// TODO: RE this structure
 struct UserData {

src/core/hle/service/am/frontend/applets.cpp

@@ -29,6 +29,7 @@
 #include "core/hle/service/am/frontend/applet_web_browser.h"
 #include "core/hle/service/am/frontend/applets.h"
 #include "core/hle/service/am/service/storage.h"
+#include "core/hle/service/am/window_system.h"
 #include "core/hle/service/sm/sm.h"
 
 namespace Service::AM::Frontend {
@@ -72,10 +73,16 @@ void FrontendApplet::PushInteractiveOutData(std::shared_ptr<IStorage> storage) {
 
 void FrontendApplet::Exit() {
     auto applet_ = applet.lock();
-
-    std::scoped_lock lk{applet_->lock};
-    applet_->is_completed = true;
-    applet_->state_changed_event.Signal(system.Kernel());
+    {
+        std::scoped_lock lk{applet_->lock};
+        applet_->is_completed = true;
+        applet_->state_changed_event.Signal(system.Kernel());
+    }
+    if (auto caller_applet = applet_->caller_applet.lock()) {
+        std::scoped_lock lk{caller_applet->lock};
+        std::erase(caller_applet->child_applets, applet_);
+    }
+    if (auto* window_system = system.GetAppletManager().GetWindowSystem()) window_system->RequestUpdate();
 }
 
 FrontendAppletSet::FrontendAppletSet() = default;

src/core/hle/service/am/lifecycle_manager.h

@@ -92,6 +92,13 @@ public:
         }
     }
 
+    void RequestFocusStateChangedNotification(Kernel::KernelCore& kernel) {
+        if (m_focus_state_changed_notification_enabled) {
+            m_has_focus_state_changed = true;
+            this->SignalSystemEventIfNeeded(kernel);
+        }
+    }
+
     void OnOperationAndPerformanceModeChanged(Kernel::KernelCore& kernel);
 
 public:

src/core/hle/service/am/service/application_functions.cpp

@@ -347,21 +347,22 @@ Result IApplicationFunctions::NotifyRunning(Out<bool> out_became_running) {
 
 Result IApplicationFunctions::GetPseudoDeviceId(Out<Common::UUID> out_pseudo_device_id) {
     LOG_WARNING(Service_AM, "(stubbed)");
-    R_UNLESS(out_pseudo_device_id, ResultUnknown);
+    R_UNLESS(out_pseudo_device_id != nullptr, ResultUnknown);
 
     // This should be hashed with the device specific hash
     // for now this will do
     const auto res = FileSys::PatchManager::GetMetadataFromBaseOrUpdate(system, m_applet->program_id);
-    u8 hash[EVP_MAX_MD_SIZE];
+    R_UNLESS(res.first != nullptr, ResultUnknown);
+    std::array<u8, EVP_MAX_MD_SIZE> hash;
     unsigned int hash_len = 0;
     auto const seed = res.first->raw.seed_for_pseudo_device_id;
     EVP_MD_CTX *ctx = EVP_MD_CTX_new();
     auto const algorithm = EVP_sha1();
     EVP_DigestInit_ex(ctx, algorithm, nullptr);
     EVP_DigestUpdate(ctx, &seed, sizeof(seed));
-    EVP_DigestFinal_ex(ctx, hash, &hash_len);
+    EVP_DigestFinal_ex(ctx, hash.data(), &hash_len);
     EVP_MD_CTX_free(ctx);
-    *out_pseudo_device_id = Common::UUID::MakeRFC4122V5(std::span<u8, 20>{hash, std::size(hash)});
+    *out_pseudo_device_id = Common::UUID::MakeRFC4122V5(std::span<u8, 16>{hash.begin(), hash.begin() + 16});
     R_SUCCEED();
 }
 

src/core/hle/service/am/service/library_applet_accessor.cpp

@@ -164,14 +164,14 @@ Result ILibraryAppletAccessor::PushInData(SharedPointer<IStorage> storage) {
 Result ILibraryAppletAccessor::PopOutData(Out<SharedPointer<IStorage>> out_storage) {
     LOG_DEBUG(Service_AM, "called");
 
-    if (auto caller_applet = m_applet->caller_applet.lock(); caller_applet) {
-        caller_applet->lifecycle_manager.GetSystemEvent().Signal(system.Kernel());
-        caller_applet->lifecycle_manager.RequestResumeNotification();
-        caller_applet->lifecycle_manager.GetSystemEvent().Clear(system.Kernel());
-        caller_applet->lifecycle_manager.UpdateRequestedFocusState();
-    }
-
     R_TRY(m_broker->GetOutData().Pop(system.Kernel(), out_storage.Get()));
+    if (auto caller_applet = m_applet->caller_applet.lock(); caller_applet) {
+        std::scoped_lock lk{caller_applet->lock};
+        const bool focus_state_changed = caller_applet->lifecycle_manager.UpdateRequestedFocusState();
+        const bool is_front_app = m_applet->frontend && caller_applet->lifecycle_manager.IsApplication();
+        if (focus_state_changed) caller_applet->lifecycle_manager.SignalSystemEventIfNeeded(system.Kernel());
+        else if (is_front_app) caller_applet->lifecycle_manager.RequestFocusStateChangedNotification(system.Kernel());
+    }
 
     if (m_applet->applet_id == AppletId::ProfileSelect && *out_storage) {
         auto impl = (*out_storage)->GetImpl();

src/core/hle/service/am/service/library_applet_creator.cpp

@@ -122,7 +122,10 @@ std::shared_ptr<ILibraryAppletAccessor> CreateGuestApplet(Core::System& system,
         auto broker = std::make_shared<AppletDataBroker>(system);
         applet->caller_applet = caller_applet;
         applet->caller_applet_broker = broker;
-        caller_applet->child_applets.push_back(applet);
+        {
+            std::scoped_lock lk{caller_applet->lock};
+            caller_applet->child_applets.push_back(applet);
+        }
         window_system.TrackApplet(applet, false);
         return std::make_shared<ILibraryAppletAccessor>(system, broker, applet);
     }
@@ -148,10 +151,10 @@ std::shared_ptr<ILibraryAppletAccessor> CreateFrontendApplet(Core::System& syste
     applet->caller_applet = caller_applet;
     applet->caller_applet_broker = storage;
     applet->frontend = system.GetFrontendAppletHolder().GetApplet(applet, applet_id, mode);
-    caller_applet->child_applets.push_back(applet);
-
-    window_system.TrackApplet(applet, false);
-
+    {
+        std::scoped_lock lk{caller_applet->lock};
+        caller_applet->child_applets.push_back(applet);
+    }
     return std::make_shared<ILibraryAppletAccessor>(system, storage, applet);
 }
 

src/core/hle/service/filesystem/fsp/fsp_srv.cpp

@@ -154,6 +154,9 @@ FSP_SRV::FSP_SRV(Core::System& system_)
         {720, nullptr, "AbandonAccessFailure"},
         {800, nullptr, "GetAndClearFileSystemProxyErrorInfo"},
         {810, nullptr, "RegisterProgramIndexMapInfo"},
+        {820, nullptr, "GetContentStorageInfoIndex"},
+        {830, nullptr, "EncryptStreamPlaySaveData"},
+        {831, nullptr, "DecryptStreamPlaySaveData"},
         {1000, nullptr, "SetBisRootForHost"},
         {1001, nullptr, "SetSaveDataSize"},
         {1002, nullptr, "SetSaveDataRootPath"},

src/core/hle/service/hid/hid_system_server.cpp

@@ -121,7 +121,7 @@ IHidSystemServer::IHidSystemServer(Core::System& system_, std::shared_ptr<Resour
         {547, nullptr, "GetAllowedBluetoothLinksCount"},
         {548, &IHidSystemServer::GetRegisteredDevices, "GetRegisteredDevices"},
         {549, nullptr, "GetConnectableRegisteredDevices"},
-        {551, nullptr, "GetRegisteredDevicesForControllerSupport"}, //20.0.0+
+        {551, &IHidSystemServer::GetRegisteredDevices, "GetRegisteredDevicesForControllerSupport"}, //20.0.0+ //mocked via 548 for Diablo 3 (at least)
         {700, nullptr, "ActivateUniquePad"},
         {702, &IHidSystemServer::AcquireUniquePadConnectionEventHandle, "AcquireUniquePadConnectionEventHandle"},
         {703, &IHidSystemServer::GetUniquePadIds, "GetUniquePadIds"},
@@ -758,7 +758,7 @@ void IHidSystemServer::AcquireDeviceRegisteredEventForControllerSupport(HLEReque
 }
 
 void IHidSystemServer::GetRegisteredDevices(HLERequestContext& ctx) {
-    LOG_WARNING(Service_HID, "(STUBBED) called");
+    LOG_WARNING(Service_HID, "(STUBBED) called, command={}", ctx.GetCommand()); //548 or 551
 
     struct RegisterData {
         std::array<u8, 0x68> data;

src/core/hle/service/ipc_helpers.h

@@ -22,6 +22,7 @@
 
 namespace IPC {
 
+constexpr Result ResultNotSupported{ErrorModule::HIPC, 1};
 constexpr Result ResultSessionClosed{ErrorModule::HIPC, 301};
 
 struct ResponseBuilder {

src/core/hle/service/jit/jit_context.cpp

@@ -71,17 +71,14 @@ public:
     void InstructionSynchronizationBarrierRaised() override {
         last_code_addr = u64(-1); //reset back, force refetch
     }
-    u8 MemoryRead8(u64 vaddr) override {
-        return ReadMemory<u8>(vaddr);
-    }
-    u16 MemoryRead16(u64 vaddr) override {
-        return ReadMemory<u16>(vaddr);
-    }
-    u32 MemoryRead32(u64 vaddr) override {
-        return ReadMemory<u32>(vaddr);
-    }
-    u64 MemoryRead64(u64 vaddr) override {
-        return ReadMemory<u64>(vaddr);
+    u64 MemoryRead(u64 vaddr, size_t size) override {
+        switch (size) {
+        case sizeof(u64): return ReadMemory<u64>(vaddr);
+        case sizeof(u32): return ReadMemory<u32>(vaddr);
+        case sizeof(u16): return ReadMemory<u16>(vaddr);
+        case sizeof(u8): return ReadMemory<u8>(vaddr);
+        default: UNREACHABLE();
+        }
     }
     u128 MemoryRead128(u64 vaddr) override {
         return ReadMemory<u128>(vaddr);
@@ -89,22 +86,19 @@ public:
     std::string MemoryReadCString(u64 vaddr) {
         std::string result{};
         u8 next;
-        while ((next = MemoryRead8(vaddr++)) != 0)
+        while ((next = u8(MemoryRead(vaddr++, sizeof(u8)))) != 0)
             result += char(next);
         return result;
     }
 
-    void MemoryWrite8(u64 vaddr, u8 value) override {
-        WriteMemory<u8>(vaddr, value);
-    }
-    void MemoryWrite16(u64 vaddr, u16 value) override {
-        WriteMemory<u16>(vaddr, value);
-    }
-    void MemoryWrite32(u64 vaddr, u32 value) override {
-        WriteMemory<u32>(vaddr, value);
-    }
-    void MemoryWrite64(u64 vaddr, u64 value) override {
-        WriteMemory<u64>(vaddr, value);
+    void MemoryWrite(u64 vaddr, u64 value, size_t size) override {
+        switch (size) {
+        case sizeof(u64): WriteMemory<u64>(vaddr, u64(value)); break;
+        case sizeof(u32): WriteMemory<u32>(vaddr, u32(value)); break;
+        case sizeof(u16): WriteMemory<u16>(vaddr, u16(value)); break;
+        case sizeof(u8): WriteMemory<u8>(vaddr, u8(value)); break;
+        default: UNREACHABLE();
+        }
     }
     void MemoryWrite128(u64 vaddr, u128 value) override {
         WriteMemory<u128>(vaddr, value);
@@ -193,14 +187,14 @@ public:
         // The loaded NRO file has ELF relocations that must be processed before it can run.
         // Normally this would be processed by RTLD, but in HLE context, we don't have
         // the linker available, so we have to do it ourselves.
-        const VAddr mod_offset{callbacks->MemoryRead32(4)};
-        if (callbacks->MemoryRead32(mod_offset) != Common::MakeMagic('M', 'O', 'D', '0'))
+        const VAddr mod_offset{callbacks->MemoryRead(4, sizeof(u32))};
+        if (callbacks->MemoryRead(mod_offset, sizeof(u32)) != Common::MakeMagic('M', 'O', 'D', '0'))
             return false;
 
         // For more info about dynamic entries, see the ELF ABI specification:
         // https://refspecs.linuxbase.org/elf/gabi4+/ch5.dynamic.html
         // https://refspecs.linuxbase.org/elf/gabi4+/ch4.reloc.html
-        VAddr dynamic_offset{mod_offset + callbacks->MemoryRead32(mod_offset + 4)};
+        VAddr dynamic_offset{mod_offset + callbacks->MemoryRead(mod_offset + 4, sizeof(u32))};
         VAddr rela_dyn = 0, relr_dyn = 0;
         size_t num_rela = 0, num_relr = 0;
         while (true) {
@@ -222,8 +216,8 @@ public:
         for (size_t i = 0; i < num_rela; i++) {
             const auto rela{callbacks->ReadMemory<Elf64_Rela>(rela_dyn + i * sizeof(Elf64_Rela))};
             if (Elf64RelType(rela.r_info) == ElfAArch64Relative) {
-                const VAddr contents{callbacks->MemoryRead64(rela.r_offset)};
-                callbacks->MemoryWrite64(rela.r_offset, contents + rela.r_addend);
+                const VAddr contents{callbacks->MemoryRead(rela.r_offset, sizeof(u64))};
+                callbacks->MemoryWrite(rela.r_offset, contents + rela.r_addend, sizeof(u64));
             }
         }
 
@@ -231,7 +225,7 @@ public:
         for (size_t i = 0; i < num_relr; i++) {
             const auto relr = callbacks->ReadMemory<Elf64_Relr>(relr_dyn + i * sizeof(Elf64_Relr));
             const auto incr = [&](VAddr where) {
-                callbacks->MemoryWrite64(where, callbacks->MemoryRead64(where) + relocbase);
+                callbacks->MemoryWrite(where, callbacks->MemoryRead(where, sizeof(u64)) + relocbase, sizeof(u64));
             };
             if ((relr & 1) == 0) {
                 // where pointer
@@ -294,7 +288,7 @@ public:
         if (argument_stack.size() > 8) {
             const VAddr new_sp = Common::AlignDown(top_of_stack - (argument_stack.size() - 8) * sizeof(u64), STACK_ALIGN);
             for (size_t i = 8; i < argument_stack.size(); i++)
-                callbacks->MemoryWrite64(new_sp + (i - 8) * sizeof(u64), argument_stack[i]);
+                callbacks->MemoryWrite(new_sp + (i - 8) * sizeof(u64), argument_stack[i], sizeof(u64));
             jit->SetSP(new_sp);
         }
         // Reset the call state for the next invocation
@@ -385,17 +379,17 @@ void DynarmicCallbacks64::CallSVC(u32 swi) {
 
         if (dest < src) {
             for (size_t i = 0; i < n; i++)
-                MemoryWrite8(dest + i, MemoryRead8(src + i));
+                MemoryWrite(dest + i, u8(MemoryRead(src + i, sizeof(u8))), sizeof(u8));
         } else {
             for (size_t i = n; i > 0; i--)
-                MemoryWrite8(dest + i - 1, MemoryRead8(src + i - 1));
+                MemoryWrite(dest + i - 1, u8(MemoryRead(src + i - 1, sizeof(u8))), sizeof(u8));
         }
     } else if (pc == parent.helpers[size_t(HelperFn::Memset)]) {
         const VAddr dest{parent.jit->GetRegister(0)};
         const u64 c{parent.jit->GetRegister(1)};
         const size_t n{parent.jit->GetRegister(2)};
         for (size_t i = 0; i < n; i++)
-            MemoryWrite8(dest + i, u8(c));
+            MemoryWrite(dest + i, u8(c), sizeof(u8));
     } else if (pc == parent.helpers[size_t(HelperFn::Resolve)]) {
         // X0 contains a char* for a symbol to resolve
         const auto name{MemoryReadCString(parent.jit->GetRegister(0))};
@@ -422,7 +416,7 @@ void DynarmicCallbacks64::CallSVC(u32 swi) {
 }
 
 void DynarmicCallbacks64::ExceptionRaised(u64 pc, Dynarmic::A64::Exception exception) {
-    auto const inst = MemoryRead32(pc);
+    auto const inst = MemoryRead(pc, sizeof(u32));
     LOG_CRITICAL(Service_JIT, "{} PC @ {:08x}, data = {:08x}", exception, pc, inst);
     parent.jit->HaltExecution();
 }

src/core/hle/service/ngc/ngc.cpp

@@ -6,10 +6,16 @@
 
 #include "common/string_util.h"
 #include "core/core.h"
+#include "core/hle/kernel/k_client_session.h"
+#include "core/hle/result.h"
+#include "core/hle/service/acc/profile_manager.h"
+#include "core/hle/service/cmif_types.h"
+#include "core/hle/service/cmif_serialization.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/hle/service/ngc/ngc.h"
 #include "core/hle/service/server_manager.h"
 #include "core/hle/service/service.h"
+#include "frontend_common/firmware_manager.h"
 
 namespace Service::NGC {
 
@@ -166,12 +172,123 @@ public:
     }
 };
 
+struct SaveDataHandle {
+    u64 unk0;
+};
+static_assert(sizeof(SaveDataHandle) == 0x08);
+
+class IUserShimScopedObject final : public ServiceFramework<IUserShimScopedObject> {
+public:
+    explicit IUserShimScopedObject(Core::System& system_) : ServiceFramework(system_, "IUserShimScopedObject") {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {450, nullptr, "InitializeForSaveData"},
+            {451, nullptr, "FinalizeForSaveData"},
+            {452, D<&IUserShimScopedObject::OpenSaveData>, "OpenSaveData"},
+            {453, nullptr, "CloseSaveData"},
+            {454, D<&IUserShimScopedObject::ReadSaveSlot>, "ReadSaveSlot"},
+            {455, D<&IUserShimScopedObject::WriteSaveSlot>, "WriteSaveSlot"},
+            {456, nullptr, "FlushSaveSlot"},
+            {457, nullptr, "CommitSaveData"},
+        };
+        // clang-format on
+        RegisterHandlers(functions);
+    }
+
+    Result OpenSaveData(Account::Uid unk0, Out<SaveDataHandle> unk1) {
+        LOG_WARNING(Service_NGC, "stubbed");
+        R_THROW(IPC::ResultNotSupported);
+    }
+
+    Result ReadSaveSlot(s32 offset, SaveDataHandle handle, OutBuffer<BufferAttr_HipcAutoSelect> out_data, Out<u32> out_size) {
+        LOG_WARNING(Service_NGC, "stubbed");
+        R_THROW(IPC::ResultNotSupported);
+    }
+
+    Result WriteSaveSlot(s32 offset, SaveDataHandle handle, InBuffer<BufferAttr_HipcAutoSelect> out_data) {
+        LOG_WARNING(Service_NGC, "stubbed");
+        // to implement
+        R_SUCCEED();
+    }
+};
+
+class IUserService final : public ServiceFramework<IUserService> {
+public:
+    explicit IUserService(Core::System& system_) : ServiceFramework(system_, "stpl:u") {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0 , D<&IUserService::Cmd0>, "Cmd0"},
+        };
+        // clang-format on
+        RegisterHandlers(functions);
+    }
+    Result Cmd0(u32 unk0, OutInterface<IUserShimScopedObject> out_interface) {
+        LOG_WARNING(Service_NGC, "stubbed");
+        *out_interface = std::make_shared<IUserShimScopedObject>(system);
+        R_SUCCEED();
+    }
+};
+
+class ISystemShimScopedObject final : public ServiceFramework<ISystemShimScopedObject> {
+public:
+    explicit ISystemShimScopedObject(Core::System& system_) : ServiceFramework(system_, "ISystemShimScopedObject") {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {106, nullptr, "Cmd106"},
+            {107, nullptr, "Cmd107"},
+            {108, D<&ISystemShimScopedObject::Cmd108>, "Cmd108"},
+            {207, nullptr, "Cmd207"},
+            {208, D<&ISystemShimScopedObject::Cmd208>, "Cmd208"},
+            {209, nullptr, "Cmd209"},
+            {210, nullptr, "Cmd210"},
+            {211, nullptr, "Cmd211"},
+            {212, nullptr, "Cmd212"},
+        };
+        // clang-format on
+        RegisterHandlers(functions);
+    }
+
+    Result Cmd108() {
+        LOG_WARNING(Service_NGC, "stubbed");
+        R_THROW(IPC::ResultNotSupported);
+    }
+
+    Result Cmd208(Out<std::array<u8, 0x20>> unk0) {
+        LOG_WARNING(Service_NGC, "stubbed");
+        R_THROW(IPC::ResultNotSupported);
+    }
+};
+
+class ISystemService final : public ServiceFramework<ISystemService> {
+public:
+    explicit ISystemService(Core::System& system_) : ServiceFramework(system_, "stpl:sys") {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0 , D<&ISystemService::Cmd0>, "Cmd0"},
+        };
+        // clang-format on
+        RegisterHandlers(functions);
+    }
+    Result Cmd0(OutInterface<ISystemShimScopedObject> out_interface) {
+        LOG_WARNING(Service_NGC, "stubbed");
+        *out_interface = std::make_shared<ISystemShimScopedObject>(system);
+        R_SUCCEED();
+    }
+};
+
 void LoopProcess(Core::System& system) {
     auto server_manager = std::make_unique<ServerManager>(system);
 
     server_manager->RegisterNamedService("ngct:u", std::make_shared<IService>(system), 4);
     server_manager->RegisterNamedService("ngct:s", std::make_shared<IServiceWithManagementApi>(system), 4);
     server_manager->RegisterNamedService("ngc:u", std::make_shared<NgcServiceImpl>(system), 4);
+
+    // +23.0.0
+    if (FirmwareManager::GetFirmwareVersion(system).first.major >= 23) {
+        server_manager->RegisterNamedService("stpl:u", std::make_shared<IUserService>(system), 4);
+        server_manager->RegisterNamedService("stpl:sys", std::make_shared<ISystemService>(system), 4);
+    }
+
     ServerManager::RunServer(std::move(server_manager));
 }
 

src/core/hle/service/ns/application_manager_interface.h

@@ -7,7 +7,7 @@
 #pragma once
 
 #include "core/hle/service/cmif_types.h"
-#include "core/hle/service/ns/i_async_result.h"
+#include "core/hle/service/ns/async_result.h"
 #include "core/hle/service/ns/language.h"
 #include "core/hle/service/ns/ns_types.h"
 #include "core/hle/service/os/event.h"

src/core/hle/service/ns/async_result.cpp

@@ -2,7 +2,7 @@
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 #include "core/hle/service/cmif_serialization.h"
-#include "core/hle/service/ns/i_async_result.h"
+#include "core/hle/service/ns/async_result.h"
 
 #include <cstring>
 
@@ -32,4 +32,4 @@ Result IAsyncResult::Cancel() {
     R_SUCCEED();
 }
 
-} // namespace Service::NS
+} // namespace Service::NS

src/core/hle/service/ns/read_only_application_control_data_interface.cpp

@@ -311,27 +311,30 @@ void IReadOnlyApplicationControlDataInterface::ListApplicationIcon(HLERequestCon
         // u64 - app count
         memory.WriteBlock(t_mem_address + out_length, &app_count, sizeof(u64));
         out_length += sizeof(u64);
+        ASSERT(out_length <= t_mem->GetSize());
         // [list of u64] - size of icons
         for (size_t i = 0; i < app_count; ++i) {
             const u64 app_id = app_ids_buffer[i];
             const FileSys::PatchManager pm{app_id, system.GetFileSystemController(), system.GetContentProvider()};
-            const auto control = pm.GetControlMetadata();
-            u64 full_size = control.second->GetSize();
-            memory.WriteBlock(t_mem_address + out_length, &full_size, sizeof(u64));
+            if (const auto control = pm.GetControlMetadata(); control.second) {
+                u64 full_size = control.second->GetSize();
+                memory.WriteBlock(t_mem_address + out_length, &full_size, sizeof(u64));
+            }
             out_length += sizeof(u64);
+            ASSERT(out_length <= t_mem->GetSize());
         }
         // [list of raw icon data]
-        std::vector<u8> full_icon_data;
         for (size_t i = 0; i < app_count; ++i) {
             const u64 app_id = app_ids_buffer[i];
             const FileSys::PatchManager pm{app_id, system.GetFileSystemController(), system.GetContentProvider()};
-            const auto control = pm.GetControlMetadata();
-            auto const full_size = control.second->GetSize();
-            if (full_size > 0) {
-                full_icon_data.resize(full_size);
-                control.second->Read(full_icon_data.data(), full_size, 0);
-                memory.WriteBlock(t_mem_address + out_length, full_icon_data.data(), full_size);
-                out_length += full_size;
+            if (const auto control = pm.GetControlMetadata(); control.second) {
+                if (auto const full_size = control.second->GetSize(); full_size > 0) {
+                    std::vector<u8> full_icon_data(full_size);
+                    control.second->Read(full_icon_data.data(), full_size, 0);
+                    memory.WriteBlock(t_mem_address + out_length, full_icon_data.data(), full_size);
+                    out_length += full_size;
+                    ASSERT(out_length <= t_mem->GetSize());
+                }
             }
         }
     }
@@ -345,6 +348,12 @@ void IReadOnlyApplicationControlDataInterface::ListApplicationIcon(HLERequestCon
 void IReadOnlyApplicationControlDataInterface::ListApplicationTitle(HLERequestContext& ctx) {
     const auto app_ids_buffer = ctx.ReadBuffer();
     const size_t app_count = app_ids_buffer.size() / sizeof(u64);
+
+    std::vector<u64> application_ids(app_count);
+    if (app_count > 0) {
+        std::memcpy(application_ids.data(), app_ids_buffer.data(), app_count * sizeof(u64));
+    }
+
     auto t_mem_obj = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(ctx.GetCopyHandle(0));
     auto* t_mem = t_mem_obj.GetPointerUnsafe();
     constexpr size_t title_entry_size = sizeof(FileSys::LanguageEntry);
@@ -354,8 +363,9 @@ void IReadOnlyApplicationControlDataInterface::ListApplicationTitle(HLERequestCo
         auto& memory = system.ApplicationMemory();
         const auto t_mem_address = t_mem->GetSourceAddress();
         for (size_t i = 0; i < app_count; ++i) {
-            const u64 app_id = app_ids_buffer[i];
-            const FileSys::PatchManager pm{app_id, system.GetFileSystemController(), system.GetContentProvider()};
+            const u64 app_id = application_ids[i];
+            const FileSys::PatchManager pm{app_id, system.GetFileSystemController(),
+                                           system.GetContentProvider()};
             const auto control = pm.GetControlMetadata();
             FileSys::LanguageEntry entry{};
             if (control.first != nullptr) {

src/core/hle/service/nvdrv/core/nvmap.cpp

@@ -211,19 +211,14 @@ DAddr NvMap::PinHandle(NvMap::Handle::Id handle, bool low_area_pin) {
             while ((address = smmu.Allocate(aligned_up)) == 0) {
                 // Free handles until the allocation succeeds
                 std::scoped_lock queueLock(unmap_queue_lock);
-                if (unmap_queue.empty()) {
-                    LOG_CRITICAL(Service_NVDRV, "Ran out of SMMU address space!");
-                    return 0;
-                }
-                // Handles in the unmap queue are guaranteed not to be pinned so don't bother
-                // checking if they are before unmapping
-                const std::shared_ptr<Handle> freeHandleDesc = unmap_queue.front();
-                std::scoped_lock freeLock(freeHandleDesc->mutex);
-                if (freeHandleDesc->d_address) {
-                    UnmapHandle(*freeHandleDesc);
+                if (auto freeHandleDesc{unmap_queue.front()}) {
+                    // Handles in the unmap queue are guaranteed not to be pinned so don't bother
+                    // checking if they are before unmapping
+                    std::scoped_lock freeLock(freeHandleDesc->mutex);
+                    if (handle_description->d_address)
+                        UnmapHandle(*freeHandleDesc);
                 } else {
-                    unmap_queue.pop_front();
-                    freeHandleDesc->unmap_queue_entry.reset();
+                    LOG_CRITICAL(Service_NVDRV, "Ran out of SMMU address space!");
                 }
             }
 

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp

@@ -233,9 +233,9 @@ NvResult nvhost_as_gpu::FreeSpace(IoctlFreeSpace& params) {
         if (allocation.page_size != params.page_size || allocation.size != (u64(params.pages) * params.page_size))
             return NvResult::BadValue;
 
-        for (const auto mapping_offset : allocation.mappings) {
-            void(FreeMappingLocked(mapping_offset));
-        }
+        for (const auto mapping_offset : allocation.mappings)
+            if (!FreeMappingLocked(mapping_offset))
+                return NvResult::BadValue;
 
         // Unset sparse flag if required
         if (allocation.sparse)
@@ -375,16 +375,39 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
         mapping_map.insert_or_assign(params.offset, Mapping(params.handle, device_address, params.offset, size, false, big_page, false));
     }
 
+    map_buffer_offsets.insert(params.offset);
+
     return NvResult::Success;
 }
 
 NvResult nvhost_as_gpu::UnmapBuffer(IoctlUnmapBuffer& params) {
-    LOG_DEBUG(Service_NVDRV, "called, offset={:#x}", params.offset);
     std::scoped_lock lock(mutex);
-    if (!vm.initialised) {
-        return NvResult::BadValue;
+    if (auto const offset_it = map_buffer_offsets.find(params.offset); offset_it != map_buffer_offsets.end()) {
+        LOG_DEBUG(Service_NVDRV, "called, offset={:#x}", params.offset);
+        if (!vm.initialised) {
+            return NvResult::BadValue;
+        }
+
+        auto const it = mapping_map.find(params.offset);
+        auto const mapping = it->second;
+        if (!mapping.fixed) {
+            auto& allocator{mapping.big_page ? *vm.big_page_allocator : *vm.small_page_allocator};
+            u32 page_size_bits{mapping.big_page ? vm.big_page_size_bits : VM::PAGE_SIZE_BITS};
+            allocator.Free(u32(mapping.offset >> page_size_bits), u32(mapping.size >> page_size_bits));
+        }
+
+        // Sparse mappings shouldn't be fully unmapped, just returned to their sparse state
+        // Only FreeSpace can unmap them fully
+        if (mapping.sparse_alloc) {
+            gmmu->MapSparse(params.offset, mapping.size, mapping.big_page);
+        } else {
+            gmmu->Unmap(params.offset, mapping.size);
+        }
+
+        nvmap.UnpinHandle(mapping.handle);
+        mapping_map.erase(params.offset);
+        map_buffer_offsets.erase(params.offset);
     }
-    void(FreeMappingLocked(params.offset));
     return NvResult::Success;
 }
 

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h

@@ -113,6 +113,8 @@ private:
     };
     static_assert(sizeof(IoctlRemapEntry) == 20, "IoctlRemapEntry is incorrect size");
 
+    ::Common::unordered_set<s64_le> map_buffer_offsets{};
+
     struct IoctlMapBufferEx {
         MappingFlags flags{}; // bit0: fixed_offset, bit2: cacheable
         u32_le kind{};        // -1 is default

src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp

@@ -13,7 +13,6 @@
 #include "common/assert.h"
 #include "common/logging.h"
 #include "common/scope_exit.h"
-#include "common/settings.h"
 #include "core/core.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/service/nvdrv/core/container.h"
@@ -148,12 +147,9 @@ NvResult nvhost_ctrl::IocCtrlEventWait(IocCtrlEventWaitParams& params, bool is_a
 
     const auto check_failing = [&]() {
         if (events[slot].fails > 2) {
-            std::unique_lock<std::mutex> stall;
-            if (Settings::values.stall_on_gpu_fence_wait.GetValue()) {
-                stall = system.StallApplication();
-            }
-            host1x_syncpoint_manager.WaitHost(fence_id, target_value);
-            if (stall.owns_lock()) {
+            {
+                auto lk = system.StallApplication();
+                host1x_syncpoint_manager.WaitHost(fence_id, target_value);
                 system.UnstallApplication();
             }
             params.value.raw = target_value;

src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp

@@ -312,17 +312,19 @@ static boost::container::small_vector<Tegra::CommandHeader, 512> BuildWaitComman
 
 static boost::container::small_vector<Tegra::CommandHeader, 512> BuildIncrementCommandList(
     NvFence fence) {
-    const Tegra::CommandHeader increment =
-        BuildFenceAction(Tegra::Engines::Puller::FenceOperation::Increment, fence.id);
-    return {
+    boost::container::small_vector<Tegra::CommandHeader, 512> result{
         Tegra::BuildCommandHeader(Tegra::BufferMethods::SyncpointPayload, 1,
                                   Tegra::SubmissionMode::Increasing),
-        {},
-        Tegra::BuildCommandHeader(Tegra::BufferMethods::SyncpointOperation, 2,
-                                  Tegra::SubmissionMode::NonIncreasing),
-        increment,
-        increment,
-    };
+        {}};
+
+    for (u32 count = 0; count < 2; ++count) {
+        result.push_back(Tegra::BuildCommandHeader(Tegra::BufferMethods::SyncpointOperation, 1,
+                                                   Tegra::SubmissionMode::Increasing));
+        result.push_back(
+            BuildFenceAction(Tegra::Engines::Puller::FenceOperation::Increment, fence.id));
+    }
+
+    return result;
 }
 
 static boost::container::small_vector<Tegra::CommandHeader, 512> BuildIncrementWithWfiCommandList(
@@ -368,14 +370,17 @@ NvResult nvhost_gpu::SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, Tegra::CommandL
     u32 increment{(flags.fence_increment.Value() != 0 ? 2 : 0) +
                   (flags.increment_value.Value() != 0 ? params.fence.value : 0)};
     params.fence.value = syncpoint_manager.IncrementSyncpointMaxExt(channel_syncpoint, increment);
+    gpu.PushGPUEntries(bind_id, std::move(entries));
+
     if (flags.fence_increment.Value()) {
         if (flags.suppress_wfi.Value()) {
-            entries.prefetch_command_list = BuildIncrementCommandList(params.fence);
+            gpu.PushGPUEntries(bind_id,
+                               Tegra::CommandList{BuildIncrementCommandList(params.fence)});
         } else {
-            entries.prefetch_command_list = BuildIncrementWithWfiCommandList(params.fence);
+            gpu.PushGPUEntries(bind_id,
+                               Tegra::CommandList{BuildIncrementWithWfiCommandList(params.fence)});
         }
     }
-    gpu.PushGPUEntries(bind_id, std::move(entries));
 
     flags.raw = 0;
 

src/core/hle/service/nvdrv/nvdrv.cpp

@@ -165,22 +165,13 @@ Module::Module(Core::System& system)
 
 Module::~Module() {}
 
-std::shared_ptr<Devices::nvdevice> Module::FindDevice(DeviceFD fd) const {
-    std::scoped_lock lock(open_files_mutex);
-    const auto itr = open_files.find(fd);
-    if (itr == open_files.end()) {
-        return nullptr;
-    }
-    return itr->second;
-}
-
 NvResult Module::VerifyFD(DeviceFD fd) const {
     if (fd < 0) {
         LOG_ERROR(Service_NVDRV, "Invalid DeviceFD={}!", fd);
         return NvResult::InvalidState;
     }
 
-    if (!FindDevice(fd)) {
+    if (open_files.find(fd) == open_files.end()) {
         LOG_ERROR(Service_NVDRV, "Could not find DeviceFD={}!", fd);
         return NvResult::NotImplemented;
     }
@@ -196,11 +187,8 @@ DeviceFD Module::Open(const std::string& device_name, NvCore::SessionId session_
     }
 
     const DeviceFD fd = next_fd++;
-    std::shared_ptr<Devices::nvdevice> device;
-    {
-        std::scoped_lock lock(open_files_mutex);
-        device = it->second(fd)->second;
-    }
+    auto& builder = it->second;
+    auto device = builder(fd)->second;
 
     device->OnOpen(session_id, fd);
 
@@ -214,13 +202,14 @@ NvResult Module::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> input,
         return NvResult::InvalidState;
     }
 
-    const auto device = FindDevice(fd);
-    if (!device) {
+    const auto itr = open_files.find(fd);
+
+    if (itr == open_files.end()) {
         LOG_ERROR(Service_NVDRV, "Could not find DeviceFD={}!", fd);
         return NvResult::NotImplemented;
     }
 
-    return device->Ioctl1(fd, command, input, output);
+    return itr->second->Ioctl1(fd, command, input, output);
 }
 
 NvResult Module::Ioctl2(DeviceFD fd, Ioctl command, std::span<const u8> input,
@@ -230,13 +219,14 @@ NvResult Module::Ioctl2(DeviceFD fd, Ioctl command, std::span<const u8> input,
         return NvResult::InvalidState;
     }
 
-    const auto device = FindDevice(fd);
-    if (!device) {
+    const auto itr = open_files.find(fd);
+
+    if (itr == open_files.end()) {
         LOG_ERROR(Service_NVDRV, "Could not find DeviceFD={}!", fd);
         return NvResult::NotImplemented;
     }
 
-    return device->Ioctl2(fd, command, input, inline_input, output);
+    return itr->second->Ioctl2(fd, command, input, inline_input, output);
 }
 
 NvResult Module::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, std::span<u8> output,
@@ -246,13 +236,14 @@ NvResult Module::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> input, s
         return NvResult::InvalidState;
     }
 
-    const auto device = FindDevice(fd);
-    if (!device) {
+    const auto itr = open_files.find(fd);
+
+    if (itr == open_files.end()) {
         LOG_ERROR(Service_NVDRV, "Could not find DeviceFD={}!", fd);
         return NvResult::NotImplemented;
     }
 
-    return device->Ioctl3(fd, command, input, output, inline_output);
+    return itr->second->Ioctl3(fd, command, input, output, inline_output);
 }
 
 NvResult Module::Close(DeviceFD fd) {
@@ -261,17 +252,16 @@ NvResult Module::Close(DeviceFD fd) {
         return NvResult::InvalidState;
     }
 
-    const auto device = FindDevice(fd);
-    if (!device) {
+    const auto itr = open_files.find(fd);
+
+    if (itr == open_files.end()) {
         LOG_ERROR(Service_NVDRV, "Could not find DeviceFD={}!", fd);
         return NvResult::NotImplemented;
     }
 
-    {
-        std::scoped_lock lock(open_files_mutex);
-        open_files.erase(fd);
-    }
-    device->OnClose(fd);
+    itr->second->OnClose(fd);
+
+    open_files.erase(itr);
 
     return NvResult::Success;
 }
@@ -282,13 +272,14 @@ NvResult Module::QueryEvent(DeviceFD fd, u32 event_id, Kernel::KEvent*& event) {
         return NvResult::InvalidState;
     }
 
-    const auto device = FindDevice(fd);
-    if (!device) {
+    const auto itr = open_files.find(fd);
+
+    if (itr == open_files.end()) {
         LOG_ERROR(Service_NVDRV, "Could not find DeviceFD={}!", fd);
         return NvResult::NotImplemented;
     }
 
-    event = device->QueryEvent(event_id);
+    event = itr->second->QueryEvent(event_id);
     if (!event) {
         return NvResult::BadParameter;
     }

src/core/hle/service/nvdrv/nvdrv.h

@@ -65,7 +65,10 @@ public:
     /// Returns a pointer to one of the available devices, identified by its name.
     template <typename T>
     std::shared_ptr<T> GetDevice(DeviceFD fd) {
-        return std::static_pointer_cast<T>(FindDevice(fd));
+        auto itr = open_files.find(fd);
+        if (itr == open_files.end())
+            return nullptr;
+        return std::static_pointer_cast<T>(itr->second);
     }
 
     NvResult VerifyFD(DeviceFD fd) const;
@@ -94,8 +97,6 @@ public:
 private:
     friend class EventInterface;
 
-    std::shared_ptr<Devices::nvdevice> FindDevice(DeviceFD fd) const;
-
     /// Manages syncpoints on the host
     NvCore::Container container;
 
@@ -105,7 +106,6 @@ private:
     using FilesContainerType = ::Common::unordered_map<DeviceFD, std::shared_ptr<Devices::nvdevice>>;
     /// Mapping of file descriptors to the devices they reference.
     FilesContainerType open_files;
-    mutable std::mutex open_files_mutex;
 
     KernelHelpers::ServiceContext service_context;
 

src/core/hle/service/ssl/ssl.cpp

@@ -48,6 +48,12 @@ enum class OptionType : u32 {
     EnableAlpn = 3,
 };
 
+// This is nn::ssl::sf::RenegotiationMode
+enum RenegotiationMode : u32 {
+    None = 0, ///< None
+    Secure = 1, ///< Secure
+};
+
 // This is nn::ssl::sf::SslVersion
 struct SslVersion {
     union {
@@ -75,34 +81,34 @@ public:
           shared_data{shared_data_in}, backend{std::move(backend_in)} {
         // clang-format off
         static const FunctionInfo functions[] = {
-            {0, &ISslConnection::SetSocketDescriptor, "SetSocketDescriptor"},
-            {1, &ISslConnection::SetHostName, "SetHostName"},
-            {2, &ISslConnection::SetVerifyOption, "SetVerifyOption"},
-            {3, &ISslConnection::SetIoMode, "SetIoMode"},
-            {4, nullptr, "GetSocketDescriptor"},
-            {5, nullptr, "GetHostName"},
+            {0, D<&ISslConnection::SetSocketDescriptor>, "SetSocketDescriptor"},
+            {1, D<&ISslConnection::SetHostName>, "SetHostName"},
+            {2, D<&ISslConnection::SetVerifyOption>, "SetVerifyOption"},
+            {3, D<&ISslConnection::SetIoMode>, "SetIoMode"},
+            {4, D<&ISslConnection::GetSocketDescriptor>, "GetSocketDescriptor"},
+            {5, D<&ISslConnection::GetHostName>, "GetHostName"},
             {6, nullptr, "GetVerifyOption"},
-            {7, nullptr, "GetIoMode"},
-            {8, &ISslConnection::DoHandshake, "DoHandshake"},
+            {7, D<&ISslConnection::GetIoMode>, "GetIoMode"},
+            {8, D<&ISslConnection::DoHandshake>, "DoHandshake"},
             {9, &ISslConnection::DoHandshakeGetServerCert, "DoHandshakeGetServerCert"},
-            {10, &ISslConnection::Read, "Read"},
-            {11, &ISslConnection::Write, "Write"},
-            {12, &ISslConnection::Pending, "Pending"},
-            {13, nullptr, "Peek"},
-            {14, nullptr, "Poll"},
-            {15, nullptr, "GetVerifyCertError"},
-            {16, nullptr, "GetNeededServerCertBufferSize"},
-            {17, &ISslConnection::SetSessionCacheMode, "SetSessionCacheMode"},
-            {18, nullptr, "GetSessionCacheMode"},
-            {19, nullptr, "FlushSessionCache"},
-            {20, nullptr, "SetRenegotiationMode"},
-            {21, nullptr, "GetRenegotiationMode"},
-            {22, &ISslConnection::SetOption, "SetOption"},
-            {23, &ISslConnection::GetOption, "GetOption"},
+            {10, D<&ISslConnection::Read>, "Read"},
+            {11, D<&ISslConnection::Write>, "Write"},
+            {12, D<&ISslConnection::Pending>, "Pending"},
+            {13, D<&ISslConnection::Peek>, "Peek"},
+            {14, D<&ISslConnection::Poll>, "Poll"},
+            {15, D<&ISslConnection::GetVerifyCertError>, "GetVerifyCertError"},
+            {16, D<&ISslConnection::GetNeededServerCertBufferSize>, "GetNeededServerCertBufferSize"},
+            {17, D<&ISslConnection::SetSessionCacheMode>, "SetSessionCacheMode"},
+            {18, D<&ISslConnection::GetSessionCacheMode>, "GetSessionCacheMode"},
+            {19, D<&ISslConnection::FlushSessionCache>, "FlushSessionCache"},
+            {20, D<&ISslConnection::SetRenegotiationMode>, "SetRenegotiationMode"},
+            {21, D<&ISslConnection::GetRenegotiationMode>, "GetRenegotiationMode"},
+            {22, D<&ISslConnection::SetOption>, "SetOption"},
+            {23, D<&ISslConnection::GetOption>, "GetOption"},
             {24, nullptr, "GetVerifyCertErrors"},
             {25, nullptr, "GetCipherInfo"},
-            {26, &ISslConnection::SetNextAlpnProto, "SetNextAlpnProto"},
-            {27, &ISslConnection::GetNextAlpnProto, "GetNextAlpnProto"},
+            {26, D<&ISslConnection::SetNextAlpnProto>, "SetNextAlpnProto"},
+            {27, D<&ISslConnection::GetNextAlpnProto>, "GetNextAlpnProto"},
             {28, nullptr, "SetDtlsSocketDescriptor"},
             {29, nullptr, "GetDtlsHandshakeTimeout"},
             {30, nullptr, "SetPrivateOption"},
@@ -141,80 +147,6 @@ public:
     }
 
 private:
-    SslVersion ssl_version;
-    std::shared_ptr<SslContextSharedData> shared_data;
-    std::unique_ptr<SSLConnectionBackend> backend;
-    std::optional<int> fd_to_close;
-    bool do_not_close_socket = false;
-    bool get_server_cert_chain = false;
-    bool skip_default_verify = false;
-    bool enable_alpn = false;
-    std::shared_ptr<Network::SocketBase> socket;
-    std::vector<u8> next_alpn_proto;
-    bool did_handshake = false;
-    u32 verify_option = 0;
-
-    Result SetSocketDescriptorImpl(s32* out_fd, s32 fd) {
-        LOG_DEBUG(Service_SSL, "called, fd={}", fd);
-        ASSERT(!did_handshake);
-        auto bsd = system.ServiceManager().GetService<Service::Sockets::BSD_USA>("bsd:u");
-        ASSERT_OR_EXECUTE(bsd, { return ResultInternalError; });
-
-        auto const res_v = bsd->DuplicateSocketImpl(fd);
-        if (auto *res = std::get_if<s32>(&res_v)) {
-            const s32 duplicated_fd = *res;
-            if (do_not_close_socket) {
-                *out_fd = duplicated_fd;
-            } else {
-                *out_fd = -1;
-                fd_to_close = duplicated_fd;
-            }
-            std::optional<std::shared_ptr<Network::SocketBase>> sock = bsd->GetSocket(duplicated_fd);
-            if (!sock.has_value()) {
-                LOG_ERROR(Service_SSL, "invalid socket fd {} after duplication", duplicated_fd);
-                return ResultInvalidSocket;
-            }
-            socket = std::move(*sock);
-            backend->SetSocket(socket);
-            return ResultSuccess;
-        }
-        LOG_ERROR(Service_SSL, "Failed to duplicate socket with fd {}", fd);
-        return ResultInvalidSocket;
-    }
-
-    Result SetHostNameImpl(const std::string& hostname) {
-        LOG_DEBUG(Service_SSL, "called. hostname={}", hostname);
-        ASSERT(!did_handshake);
-        return backend->SetHostName(hostname);
-    }
-
-    Result SetVerifyOptionImpl(u32 option) {
-        ASSERT(!did_handshake);
-        LOG_DEBUG(Service_SSL, "called. option={} (forcing 0)", option);
-        verify_option = 0;
-        backend->SetVerifyOption(0);
-        return ResultSuccess;
-    }
-
-    Result SetIoModeImpl(u32 input_mode) {
-        auto mode = static_cast<IoMode>(input_mode);
-        ASSERT(mode == IoMode::Blocking || mode == IoMode::NonBlocking);
-        ASSERT_OR_EXECUTE(socket, { return ResultNoSocket; });
-
-        const bool non_block = mode == IoMode::NonBlocking;
-        const Network::Errno error = socket->SetNonBlock(non_block);
-        if (error != Network::Errno::SUCCESS) {
-            LOG_ERROR(Service_SSL, "Failed to set native socket non-block flag to {}", non_block);
-        }
-        return ResultSuccess;
-    }
-
-    Result SetSessionCacheModeImpl(u32 mode) {
-        ASSERT(!did_handshake);
-        LOG_WARNING(Service_SSL, "(STUBBED) called. value={}", mode);
-        return ResultSuccess;
-    }
-
     Result DoHandshakeImpl() {
         ASSERT_OR_EXECUTE(!did_handshake && socket, { return ResultNoSocket; });
         Result res = backend->DoHandshake();
@@ -234,19 +166,17 @@ private:
         };
         if (!get_server_cert_chain) {
             // Just return the first one, unencoded.
-            ASSERT_OR_EXECUTE_MSG(
-                !certs.empty(), { return {}; }, "Should be at least one server cert");
+            ASSERT_OR_EXECUTE_MSG(!certs.empty(), { return {}; }, "Should be at least one server cert");
             return certs[0];
         }
         std::vector<u8> ret;
-        Header header{0x4E4D684374726543, static_cast<u32>(certs.size()), 0};
+        Header header{0x4E4D684374726543, u32(certs.size()), 0};
         ret.insert(ret.end(), reinterpret_cast<u8*>(&header), reinterpret_cast<u8*>(&header + 1));
         size_t data_offset = sizeof(Header) + certs.size() * sizeof(EntryHeader);
         for (auto& cert : certs) {
-            EntryHeader entry_header{static_cast<u32>(cert.size()), static_cast<u32>(data_offset)};
+            EntryHeader entry_header{u32(cert.size()), u32(data_offset)};
             data_offset += cert.size();
-            ret.insert(ret.end(), reinterpret_cast<u8*>(&entry_header),
-                       reinterpret_cast<u8*>(&entry_header + 1));
+            ret.insert(ret.end(), reinterpret_cast<u8*>(&entry_header), reinterpret_cast<u8*>(&entry_header + 1));
         }
         for (auto& cert : certs) {
             ret.insert(ret.end(), cert.begin(), cert.end());
@@ -254,65 +184,77 @@ private:
         return ret;
     }
 
-    Result ReadImpl(std::vector<u8>* out_data) {
-        ASSERT_OR_EXECUTE(did_handshake, { return ResultInternalError; });
-        size_t actual_size{};
-        Result res = backend->Read(&actual_size, *out_data);
-        if (res != ResultSuccess) {
-            return res;
+    Result SetSocketDescriptor(s32 in_fd, Out<s32> out_fd) {
+        LOG_DEBUG(Service_SSL, "called, fd={}", in_fd);
+        ASSERT(!did_handshake);
+        auto bsd = system.ServiceManager().GetService<Service::Sockets::BSD_USA>("bsd:u");
+        ASSERT_OR_EXECUTE(bsd, { return ResultInternalError; });
+
+        auto const res_v = bsd->DuplicateSocketImpl(in_fd);
+        if (auto *res = std::get_if<s32>(&res_v)) {
+            const s32 dup_fd = *res;
+            *out_fd = do_not_close_socket ? dup_fd : -1;
+            if (!do_not_close_socket)
+                fd_to_close = dup_fd;
+            auto const sock = bsd->GetSocket(dup_fd);
+            if (!sock.has_value()) {
+                LOG_ERROR(Service_SSL, "invalid socket fd {} after duplication", dup_fd);
+                return ResultInvalidSocket;
+            }
+            socket = std::move(*sock);
+            backend->SetSocket(std::move(socket));
+            return ResultSuccess;
         }
-        out_data->resize(actual_size);
-        return res;
+        LOG_ERROR(Service_SSL, "Failed to duplicate socket with fd {}", in_fd);
+        return ResultInvalidSocket;
     }
 
-    Result WriteImpl(size_t* out_size, std::span<const u8> data) {
-        ASSERT_OR_EXECUTE(did_handshake, { return ResultInternalError; });
-        return backend->Write(out_size, data);
+    Result SetHostName(InBuffer<BufferAttr_HipcMapAlias> buf) {
+        auto const hostname = Common::StringFromBuffer(buf);
+        LOG_DEBUG(Service_SSL, "called. hostname={}", hostname);
+        ASSERT(!did_handshake);
+        return backend->SetHostName(hostname.c_str());
     }
 
-    Result PendingImpl(s32* out_pending) {
-        LOG_WARNING(Service_SSL, "(STUBBED) called.");
-        *out_pending = 0;
-        return ResultSuccess;
+    Result SetVerifyOption(u32 option) {
+        LOG_DEBUG(Service_SSL, "called. option={} (forcing 0)", option);
+        ASSERT(!did_handshake);
+        verify_option = 0;
+        backend->SetVerifyOption(0);
+        R_SUCCEED();
     }
 
-    void SetSocketDescriptor(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const s32 in_fd = rp.Pop<s32>();
-        s32 out_fd{-1};
-        const Result res = SetSocketDescriptorImpl(&out_fd, in_fd);
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(res);
-        rb.Push<s32>(out_fd);
+    Result SetIoMode(u32 input_mode) {
+        auto mode = IoMode(input_mode);
+        ASSERT(mode == IoMode::Blocking || mode == IoMode::NonBlocking);
+        R_UNLESS(socket, ResultNoSocket);
+        const bool non_block = mode == IoMode::NonBlocking;
+        const Network::Errno error = socket->SetNonBlock(non_block);
+        if (error != Network::Errno::SUCCESS) {
+            LOG_ERROR(Service_SSL, "Failed to set native socket non-block flag to {}", non_block);
+        }
+        R_SUCCEED();
     }
 
-    void SetHostName(HLERequestContext& ctx) {
-        const std::string hostname = Common::StringFromBuffer(ctx.ReadBuffer());
-        const Result res = SetHostNameImpl(hostname);
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(res);
+    Result GetSocketDescriptor(Out<u32> out_fd) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        *out_fd = uint32_t(socket->GetFD());
+        R_SUCCEED();
     }
 
-    void SetVerifyOption(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const u32 option = rp.Pop<u32>();
-        const Result res = SetVerifyOptionImpl(option);
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(res);
+    Result GetHostName(OutBuffer<BufferAttr_HipcMapAlias> data, Out<u32> out_size) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        ASSERT(!did_handshake);
+        return backend->GetHostName(std::span<u8>{data.begin(), data.end()}, out_size);
     }
 
-    void SetIoMode(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const u32 mode = rp.Pop<u32>();
-        const Result res = SetIoModeImpl(mode);
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(res);
+    Result GetIoMode(Out<u32> out_mode) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        R_SUCCEED();
     }
 
-    void DoHandshake(HLERequestContext& ctx) {
-        const Result res = DoHandshakeImpl();
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(res);
+    Result DoHandshake() {
+        return DoHandshakeImpl();
     }
 
     void DoHandshakeGetServerCert(HLERequestContext& ctx) {
@@ -351,131 +293,158 @@ private:
         rb.PushRaw(out);
     }
 
-    void Read(HLERequestContext& ctx) {
-        std::vector<u8> output_bytes(ctx.GetWriteBufferSize());
-        const Result res = ReadImpl(&output_bytes);
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(res);
-        if (res == ResultSuccess) {
-            rb.Push(static_cast<u32>(output_bytes.size()));
-            ctx.WriteBuffer(output_bytes);
-        } else {
-            rb.Push(static_cast<u32>(0));
-        }
+    Result Read(OutBuffer<BufferAttr_HipcMapAlias> data, Out<u32> out_size) {
+        R_UNLESS(did_handshake, ResultInternalError);
+        size_t tmp{};
+        auto const res = backend->Read(&tmp, data);
+        *out_size = u32(tmp);
+        return res;
     }
 
-    void Write(HLERequestContext& ctx) {
-        size_t write_size{0};
-        const Result res = WriteImpl(&write_size, ctx.ReadBuffer());
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(res);
-        rb.Push(static_cast<u32>(write_size));
+    Result Write(InBuffer<BufferAttr_HipcMapAlias> data, Out<u32> out_size) {
+        R_UNLESS(did_handshake, ResultInternalError);
+        size_t tmp{};
+        auto const res = backend->Write(&tmp, data);
+        *out_size = u32(tmp);
+        return res;
     }
 
-    void Pending(HLERequestContext& ctx) {
-        s32 pending_size{0};
-        const Result res = PendingImpl(&pending_size);
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(res);
-        rb.Push<s32>(pending_size);
+    Result Pending(Out<s32> out_pending_size) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        *out_pending_size = s32(backend->Pending());
+        R_SUCCEED();
     }
 
-    void SetSessionCacheMode(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const u32 mode = rp.Pop<u32>();
-        const Result res = SetSessionCacheModeImpl(mode);
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(res);
+    Result Peek(OutBuffer<BufferAttr_HipcMapAlias> data, Out<u32> out_size) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        size_t tmp{};
+        auto const res = backend->Peek(&tmp, data);
+        *out_size = u32(tmp);
+        return res;
     }
 
-    void SetOption(HLERequestContext& ctx) {
-        struct Parameters {
-            OptionType option;
-            s32 value;
-        };
-        static_assert(sizeof(Parameters) == 0x8, "Parameters is an invalid size");
-
-        IPC::RequestParser rp{ctx};
-        const auto parameters = rp.PopRaw<Parameters>();
-
-        switch (parameters.option) {
-        case OptionType::DoNotCloseSocket:
-            do_not_close_socket = static_cast<bool>(parameters.value);
-            break;
-        case OptionType::GetServerCertChain:
-            get_server_cert_chain = static_cast<bool>(parameters.value);
-            break;
-        case OptionType::SkipDefaultVerify:
-            skip_default_verify = static_cast<bool>(parameters.value);
-            break;
-        case OptionType::EnableAlpn:
-            enable_alpn = static_cast<bool>(parameters.value);
-            break;
-        default:
-            LOG_WARNING(Service_SSL, "Unknown option={}, value={}", parameters.option,
-                        parameters.value);
-        }
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultSuccess);
+    Result Poll(u32 in_pollevent, u32 timer, Out<u32> out_pollevent) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        R_SUCCEED();
     }
 
-    void GetOption(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto option = rp.PopRaw<OptionType>();
+    Result GetVerifyCertError() {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        R_SUCCEED();
+    }
 
-        u8 value = 0;
+    Result GetNeededServerCertBufferSize(Out<u32> out_needed_buffer_size) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        R_SUCCEED();
+    }
 
+    Result SetSessionCacheMode(u32 mode) {
+        LOG_WARNING(Service_SSL, "(STUBBED) called. value={}", mode);
+        R_UNLESS(!did_handshake, ResultInternalError);
+        R_SUCCEED();
+    }
+
+    Result GetSessionCacheMode(Out<u32> mode) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        R_UNLESS(!did_handshake, ResultInternalError);
+        R_SUCCEED();
+    }
+
+    Result FlushSessionCache() {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        R_UNLESS(!did_handshake, ResultInternalError);
+        R_SUCCEED();
+    }
+
+    Result SetRenegotiationMode(RenegotiationMode mode) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        backend->SetRenegotiationMode(u32(mode));
+        R_SUCCEED();
+    }
+
+    Result GetRenegotiationMode(Out<RenegotiationMode> mode) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        u32 tmp{};
+        auto const res = backend->GetRenegotiationMode(&tmp);
+        *mode = RenegotiationMode(tmp);
+        return res;
+    }
+
+    Result SetOption(OptionType option, s32 value) {
         switch (option) {
         case OptionType::DoNotCloseSocket:
-            value = static_cast<u8>(do_not_close_socket);
+            do_not_close_socket = bool(value);
             break;
         case OptionType::GetServerCertChain:
-            value = static_cast<u8>(get_server_cert_chain);
+            get_server_cert_chain = bool(value);
             break;
         case OptionType::SkipDefaultVerify:
-            value = static_cast<u8>(skip_default_verify);
+            skip_default_verify = bool(value);
             break;
         case OptionType::EnableAlpn:
-            value = static_cast<u8>(enable_alpn);
+            enable_alpn = bool(value);
             break;
         default:
-            LOG_WARNING(Service_SSL, "Unknown option={}", option);
-            value = 0;
+            LOG_WARNING(Service_SSL, "Unknown option={}, value={}", option, value);
+        }
+        R_SUCCEED();
+    }
+
+    Result GetOption(OptionType option, Out<u8> value) {
+        switch (option) {
+        case OptionType::DoNotCloseSocket:
+            *value = u8(do_not_close_socket);
+            break;
+        case OptionType::GetServerCertChain:
+            *value = u8(get_server_cert_chain);
+            break;
+        case OptionType::SkipDefaultVerify:
+            *value = u8(skip_default_verify);
+            break;
+        case OptionType::EnableAlpn:
+            *value = u8(enable_alpn);
+            break;
+        default:
+            LOG_WARNING(Service_SSL, "Unknown option={}", u32(option));
+            *value = 0;
             break;
         }
-
-        LOG_DEBUG(Service_SSL, "GetOption called, option={}, ret value={}", option, value);
-
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(ResultSuccess);
-        rb.Push<u8>(value);
+        LOG_DEBUG(Service_SSL, "GetOption called, option={}, ret value={}", u32(option), *value);
+        R_SUCCEED();
     }
 
-    void SetNextAlpnProto(HLERequestContext& ctx) {
-        const auto data = ctx.ReadBuffer(0);
-        next_alpn_proto.assign(data.begin(), data.end());
-
+    Result SetNextAlpnProto(InBuffer<BufferAttr_HipcMapAlias> data) {
+        auto const to_write = u32((std::min)(next_alpn_proto.size(), data.size()));
+        next_alpn_proto.assign(data.begin(), data.begin() + to_write);
         LOG_DEBUG(Service_SSL, "SetNextAlpnProto called, size={}", next_alpn_proto.size());
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultSuccess);
+        R_SUCCEED();
     }
 
-    void GetNextAlpnProto(HLERequestContext& ctx) {
-        const size_t writable = ctx.GetWriteBufferSize();
-        const size_t to_write = (std::min)(next_alpn_proto.size(), writable);
-
-        if (to_write != 0) {
-            ctx.WriteBuffer(std::span<const u8>(next_alpn_proto.data(), to_write));
-        }
-
-        LOG_DEBUG(Service_SSL, "GetNextAlpnProto called, size={}", to_write);
-
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(ResultSuccess);
-        rb.Push<u32>(static_cast<u32>(to_write));
+    Result GetNextAlpnProto(OutBuffer<BufferAttr_HipcMapAlias> data, Out<u32> to_write) {
+        *to_write = u32((std::min)(next_alpn_proto.size(), data.size()));
+        next_alpn_proto.assign(data.begin(), data.begin() + *to_write);
+        LOG_DEBUG(Service_SSL, "GetNextAlpnProto called, size={}", *to_write);
+        R_SUCCEED();
     }
+
+    Result GetVerifyCertErrors(OutBuffer<BufferAttr_HipcMapAlias> unk0, Out<u32> unk1, Out<u32> unk2) {
+        LOG_WARNING(Service_SSL, "(STUBBED)");
+        R_SUCCEED();
+    }
+
+    SslVersion ssl_version;
+    std::shared_ptr<SslContextSharedData> shared_data;
+    std::unique_ptr<SSLConnectionBackend> backend;
+    std::optional<int> fd_to_close;
+    std::shared_ptr<Network::SocketBase> socket;
+    std::vector<u8> next_alpn_proto;
+    u32 verify_option = 0;
+
+    bool do_not_close_socket = false;
+    bool get_server_cert_chain = false;
+    bool skip_default_verify = false;
+    bool enable_alpn = false;
+    bool did_handshake = false;
 };
 
 class ISslContext final : public ServiceFramework<ISslContext> {
@@ -492,7 +461,7 @@ public:
             {5, &ISslContext::ImportClientPki, "ImportClientPki"},
             {6, nullptr, "RemoveServerPki"},
             {7, nullptr, "RemoveClientPki"},
-            {8, nullptr, "RegisterInternalPki"},
+            {8, D<&ISslContext::RegisterInternalPki>, "RegisterInternalPki"},
             {9, nullptr, "AddPolicyOid"},
             {10, nullptr, "ImportCrl"},
             {11, nullptr, "RemoveCrl"},
@@ -587,6 +556,11 @@ private:
         rb.Push(ResultSuccess);
         rb.Push(client_id);
     }
+
+    Result RegisterInternalPki() {
+        LOG_WARNING(Service_SSL, "(STUBBED) called");
+        R_SUCCEED();
+    }
 };
 
 class ISslService final : public ServiceFramework<ISslService> {

src/core/hle/service/ssl/ssl_backend.h

@@ -36,12 +36,17 @@ class SSLConnectionBackend {
 public:
     virtual ~SSLConnectionBackend() {}
     virtual void SetSocket(std::shared_ptr<Network::SocketBase> socket) = 0;
-    virtual Result SetHostName(const std::string& hostname) = 0;
     virtual void SetVerifyOption(u32 option) = 0;
     virtual Result DoHandshake() = 0;
     virtual Result Read(size_t* out_size, std::span<u8> data) = 0;
+    virtual Result Peek(size_t* out_size, std::span<u8> data) = 0;
     virtual Result Write(size_t* out_size, std::span<const u8> data) = 0;
     virtual Result GetServerCerts(std::vector<std::vector<u8>>* out_certs) = 0;
+    virtual Result SetHostName(const char* hostname) = 0;
+    virtual Result GetHostName(std::span<u8> hostname, u32* out_size) = 0;
+    virtual int Pending() = 0;
+    virtual Result SetRenegotiationMode(u32 mode) = 0;
+    virtual Result GetRenegotiationMode(u32* mode) = 0;
 };
 
 Result CreateSSLConnectionBackend(std::unique_ptr<SSLConnectionBackend>* out_backend);

src/core/hle/service/ssl/ssl_backend_openssl.cpp

@@ -157,18 +157,30 @@ public:
         socket = std::move(socket_in);
     }
 
-    Result SetHostName(const std::string& hostname) override {
+    Result SetHostName(const char* hostname) override {
         if (!skip_cert_verification) {
-            if (!SSL_set1_host(ssl, hostname.c_str())) {
+            if (!SSL_set1_host(ssl, hostname)) {
                 LOG_ERROR(Service_SSL, "SSL_set1_host({}) failed", hostname);
                 return CheckOpenSSLErrors();
             }
         }
-        if (!SSL_set_tlsext_host_name(ssl, hostname.c_str())) { // hostname for SNI
+        if (!SSL_set_tlsext_host_name(ssl, hostname)) { // hostname for SNI
             LOG_ERROR(Service_SSL, "SSL_set_tlsext_host_name({}) failed", hostname);
             return CheckOpenSSLErrors();
         }
-        return ResultSuccess;
+        R_SUCCEED();
+    }
+
+    Result GetHostName(std::span<u8> data, u32* out_size) override {
+        auto const peer_name = SSL_get0_peername(ssl);
+        if (peer_name == nullptr) {
+            LOG_ERROR(Service_SSL, "SSL_get0_peername()");
+            return CheckOpenSSLErrors();
+        }
+        auto const s = std::string{peer_name};
+        *out_size = u32(s.size());
+        std::memcpy(data.data(), s.data(), (std::min)(s.size(), data.size()));
+        R_SUCCEED();
     }
 
     void SetVerifyOption(u32 option) override {
@@ -213,6 +225,13 @@ public:
         return HandleReturn("SSL_read_ex", out_size, ret);
     }
 
+    Result Peek(size_t* out_size, std::span<u8> data) override {
+        auto const n = (std::min)(data.size(), *out_size);
+        const int ret = SSL_peek(ssl, data.data(), int(n));
+        *out_size = n;
+        return HandleReturn("SSL_write_ex", out_size, ret);
+    }
+
     Result Write(size_t* out_size, std::span<const u8> data) override {
         const int ret = SSL_write_ex(ssl, data.data(), data.size(), out_size);
         return HandleReturn("SSL_write_ex", out_size, ret);
@@ -263,7 +282,25 @@ public:
             out_certs->emplace_back(buf, buf + len);
             OPENSSL_free(buf);
         }
-        return ResultSuccess;
+        R_SUCCEED();
+    }
+
+    int Pending() override {
+        return SSL_pending(ssl);
+    }
+
+    Result SetRenegotiationMode(u32 mode) override {
+        if (mode == 0) {
+            SSL_CTX_set_options(ssl_ctx, SSL_OP_NO_RENEGOTIATION);
+        } else {
+            SSL_CTX_set_options(ssl_ctx, SSL_OP_ALLOW_CLIENT_RENEGOTIATION);
+        }
+        R_SUCCEED();
+    }
+
+    Result GetRenegotiationMode(u32* mode) override {
+        *mode = SSL_get_secure_renegotiation_support(ssl) ? 1 : 0;
+        R_SUCCEED();
     }
 
     ~SSLConnectionBackendOpenSSL() {

src/core/loader/deconstructed_rom_directory.cpp

@@ -182,12 +182,6 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
     }
     metadata.Print();
 
-    if (Settings::values.relocate_36bit_address_space.GetValue() &&
-        Settings::values.cpu_backend.GetValue() == Settings::CpuBackend::Nce &&
-        metadata.GetAddressSpaceType() == FileSys::ProgramAddressSpaceType::Is36Bit) {
-        metadata.SetAddressSpaceType(FileSys::ProgramAddressSpaceType::Is39Bit);
-    }
-
     // Enable NCE only for applications with 39-bit address space.
     const bool is_39bit =
         metadata.GetAddressSpaceType() == FileSys::ProgramAddressSpaceType::Is39Bit;

src/core/tools/freezer.cpp

@@ -33,18 +33,10 @@ u64 MemoryReadWidth(Core::Memory::Memory& memory, u32 width, VAddr addr) {
 
 void MemoryWriteWidth(Core::Memory::Memory& memory, u32 width, VAddr addr, u64 value) {
     switch (width) {
-    case 1:
-        memory.Write8(addr, static_cast<u8>(value));
-        break;
-    case 2:
-        memory.Write16(addr, static_cast<u16>(value));
-        break;
-    case 4:
-        memory.Write32(addr, static_cast<u32>(value));
-        break;
-    case 8:
-        memory.Write64(addr, value);
-        break;
+    case sizeof(u64): return memory.Write64(addr, value);
+    case sizeof(u32): return memory.Write32(addr, u32(value));
+    case sizeof(u16): return memory.Write16(addr, u16(value));
+    case sizeof(u8): return memory.Write8(addr, u8(value));
     default:
         UNREACHABLE();
     }

src/dynarmic/src/dynarmic/backend/arm64/a32_address_space.cpp

@@ -24,29 +24,65 @@
 namespace Dynarmic::Backend::Arm64 {
 
 template<auto mfp, typename T>
-static void* EmitCallTrampoline(oaknut::CodeGenerator& code, T* this_) {
+static void* EmitCallReadTrampoline(oaknut::CodeGenerator& code, T* this_, size_t bitsize) {
     using namespace oaknut::util;
-
     const auto info = Devirtualize<mfp>(this_);
-
     oaknut::Label l_addr, l_this;
-
     void* target = code.xptr<void*>();
+    // params = { this, vaddr, bitsize }
     code.LDR(X0, l_this);
+    // X1 = vaddr
+    code.MOV(X2, bitsize);
     code.LDR(Xscratch0, l_addr);
     code.BR(Xscratch0);
-
     code.align(8);
     code.l(l_this);
     code.dx(info.this_ptr);
     code.l(l_addr);
     code.dx(info.fn_ptr);
-
     return target;
 }
 
 template<auto mfp, typename T>
-static void* EmitWrappedReadCallTrampoline(oaknut::CodeGenerator& code, T* this_) {
+static void* EmitCallWriteTrampoline(oaknut::CodeGenerator& code, T* this_, size_t bitsize) {
+    using namespace oaknut::util;
+    const auto info = Devirtualize<mfp>(this_);
+    oaknut::Label l_addr, l_this;
+    void* target = code.xptr<void*>();
+    // params = { this, vaddr, value, bitsize }
+    code.LDR(X0, l_this);
+    // X1 = vaddr
+    // X2 = value
+    code.MOV(X3, bitsize);
+    code.LDR(Xscratch0, l_addr);
+    code.BR(Xscratch0);
+    code.align(8);
+    code.l(l_this);
+    code.dx(info.this_ptr);
+    code.l(l_addr);
+    code.dx(info.fn_ptr);
+    return target;
+}
+
+template<auto mfp, typename T>
+static void* EmitCallTrampoline(oaknut::CodeGenerator& code, T* this_) {
+    using namespace oaknut::util;
+    const auto info = Devirtualize<mfp>(this_);
+    oaknut::Label l_addr, l_this;
+    void* target = code.xptr<void*>();
+    code.LDR(X0, l_this);
+    code.LDR(Xscratch0, l_addr);
+    code.BR(Xscratch0);
+    code.align(8);
+    code.l(l_this);
+    code.dx(info.this_ptr);
+    code.l(l_addr);
+    code.dx(info.fn_ptr);
+    return target;
+}
+
+template<auto mfp, typename T>
+static void* EmitWrappedReadCallTrampoline(oaknut::CodeGenerator& code, T* this_, size_t bitsize) {
     using namespace oaknut::util;
 
     const auto info = Devirtualize<mfp>(this_);
@@ -59,6 +95,7 @@ static void* EmitWrappedReadCallTrampoline(oaknut::CodeGenerator& code, T* this_
     ABI_PushRegisters(code, save_regs, 0);
     code.LDR(X0, l_this);
     code.MOV(X1, Xscratch0);
+    code.MOV(X2, bitsize);
     code.LDR(Xscratch0, l_addr);
     code.BLR(Xscratch0);
     code.MOV(Xscratch0, X0);
@@ -82,7 +119,7 @@ static void* EmitExclusiveReadCallTrampoline(oaknut::CodeGenerator& code, const
 
     auto fn = [](const A32::UserConfig& conf, A32::VAddr vaddr) -> T {
         return conf.global_monitor->ReadAndMark<T>(conf.processor_id, vaddr, [&]() -> T {
-            return (conf.callbacks->*callback)(vaddr);
+            return (conf.callbacks->*callback)(vaddr, sizeof(T));
         });
     };
 
@@ -101,7 +138,7 @@ static void* EmitExclusiveReadCallTrampoline(oaknut::CodeGenerator& code, const
 }
 
 template<auto mfp, typename T>
-static void* EmitWrappedWriteCallTrampoline(oaknut::CodeGenerator& code, T* this_) {
+static void* EmitWrappedWriteCallTrampoline(oaknut::CodeGenerator& code, T* this_, size_t bitsize) {
     using namespace oaknut::util;
 
     const auto info = Devirtualize<mfp>(this_);
@@ -115,6 +152,7 @@ static void* EmitWrappedWriteCallTrampoline(oaknut::CodeGenerator& code, T* this
     code.LDR(X0, l_this);
     code.MOV(X1, Xscratch0);
     code.MOV(X2, Xscratch1);
+    code.MOV(X3, bitsize);
     code.LDR(Xscratch0, l_addr);
     code.BLR(Xscratch0);
     ABI_PopRegisters(code, save_regs, 0);
@@ -136,12 +174,9 @@ static void* EmitExclusiveWriteCallTrampoline(oaknut::CodeGenerator& code, const
     oaknut::Label l_addr, l_this;
 
     auto fn = [](const A32::UserConfig& conf, A32::VAddr vaddr, T value) -> u32 {
-        return conf.global_monitor->DoExclusiveOperation<T>(conf.processor_id, vaddr,
-                                                            [&](T expected) -> bool {
-                                                                return (conf.callbacks->*callback)(vaddr, value, expected);
-                                                            })
-                 ? 0
-                 : 1;
+        return conf.global_monitor->DoExclusiveOperation<T>(conf.processor_id, vaddr, [&](T expected) -> bool {
+            return (conf.callbacks->*callback)(vaddr, value, expected);
+        }) ? 0 : 1;
     };
 
     void* target = code.xptr<void*>();
@@ -179,26 +214,27 @@ void A32AddressSpace::EmitPrelude() {
 
     UnprotectCodeMemory();
 
-    prelude_info.read_memory_8 = EmitCallTrampoline<&A32::UserCallbacks::MemoryRead8>(code, conf.callbacks);
-    prelude_info.read_memory_16 = EmitCallTrampoline<&A32::UserCallbacks::MemoryRead16>(code, conf.callbacks);
-    prelude_info.read_memory_32 = EmitCallTrampoline<&A32::UserCallbacks::MemoryRead32>(code, conf.callbacks);
-    prelude_info.read_memory_64 = EmitCallTrampoline<&A32::UserCallbacks::MemoryRead64>(code, conf.callbacks);
-    prelude_info.wrapped_read_memory_8 = EmitWrappedReadCallTrampoline<&A32::UserCallbacks::MemoryRead8>(code, conf.callbacks);
-    prelude_info.wrapped_read_memory_16 = EmitWrappedReadCallTrampoline<&A32::UserCallbacks::MemoryRead16>(code, conf.callbacks);
-    prelude_info.wrapped_read_memory_32 = EmitWrappedReadCallTrampoline<&A32::UserCallbacks::MemoryRead32>(code, conf.callbacks);
-    prelude_info.wrapped_read_memory_64 = EmitWrappedReadCallTrampoline<&A32::UserCallbacks::MemoryRead64>(code, conf.callbacks);
-    prelude_info.exclusive_read_memory_8 = EmitExclusiveReadCallTrampoline<&A32::UserCallbacks::MemoryRead8, u8>(code, conf);
-    prelude_info.exclusive_read_memory_16 = EmitExclusiveReadCallTrampoline<&A32::UserCallbacks::MemoryRead16, u16>(code, conf);
-    prelude_info.exclusive_read_memory_32 = EmitExclusiveReadCallTrampoline<&A32::UserCallbacks::MemoryRead32, u32>(code, conf);
-    prelude_info.exclusive_read_memory_64 = EmitExclusiveReadCallTrampoline<&A32::UserCallbacks::MemoryRead64, u64>(code, conf);
-    prelude_info.write_memory_8 = EmitCallTrampoline<&A32::UserCallbacks::MemoryWrite8>(code, conf.callbacks);
-    prelude_info.write_memory_16 = EmitCallTrampoline<&A32::UserCallbacks::MemoryWrite16>(code, conf.callbacks);
-    prelude_info.write_memory_32 = EmitCallTrampoline<&A32::UserCallbacks::MemoryWrite32>(code, conf.callbacks);
-    prelude_info.write_memory_64 = EmitCallTrampoline<&A32::UserCallbacks::MemoryWrite64>(code, conf.callbacks);
-    prelude_info.wrapped_write_memory_8 = EmitWrappedWriteCallTrampoline<&A32::UserCallbacks::MemoryWrite8>(code, conf.callbacks);
-    prelude_info.wrapped_write_memory_16 = EmitWrappedWriteCallTrampoline<&A32::UserCallbacks::MemoryWrite16>(code, conf.callbacks);
-    prelude_info.wrapped_write_memory_32 = EmitWrappedWriteCallTrampoline<&A32::UserCallbacks::MemoryWrite32>(code, conf.callbacks);
-    prelude_info.wrapped_write_memory_64 = EmitWrappedWriteCallTrampoline<&A32::UserCallbacks::MemoryWrite64>(code, conf.callbacks);
+
+    prelude_info.read_memory_8 = EmitCallReadTrampoline<&A32::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u8));
+    prelude_info.read_memory_16 = EmitCallReadTrampoline<&A32::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u16));
+    prelude_info.read_memory_32 = EmitCallReadTrampoline<&A32::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u32));
+    prelude_info.read_memory_64 = EmitCallReadTrampoline<&A32::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u64));
+    prelude_info.wrapped_read_memory_8 = EmitWrappedReadCallTrampoline<&A32::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u8));
+    prelude_info.wrapped_read_memory_16 = EmitWrappedReadCallTrampoline<&A32::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u16));
+    prelude_info.wrapped_read_memory_32 = EmitWrappedReadCallTrampoline<&A32::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u32));
+    prelude_info.wrapped_read_memory_64 = EmitWrappedReadCallTrampoline<&A32::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u64));
+    prelude_info.exclusive_read_memory_8 = EmitExclusiveReadCallTrampoline<&A32::UserCallbacks::MemoryRead, u8>(code, conf);
+    prelude_info.exclusive_read_memory_16 = EmitExclusiveReadCallTrampoline<&A32::UserCallbacks::MemoryRead, u16>(code, conf);
+    prelude_info.exclusive_read_memory_32 = EmitExclusiveReadCallTrampoline<&A32::UserCallbacks::MemoryRead, u32>(code, conf);
+    prelude_info.exclusive_read_memory_64 = EmitExclusiveReadCallTrampoline<&A32::UserCallbacks::MemoryRead, u64>(code, conf);
+    prelude_info.write_memory_8 = EmitCallWriteTrampoline<&A32::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u8));
+    prelude_info.write_memory_16 = EmitCallWriteTrampoline<&A32::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u16));
+    prelude_info.write_memory_32 = EmitCallWriteTrampoline<&A32::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u32));
+    prelude_info.write_memory_64 = EmitCallWriteTrampoline<&A32::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u64));
+    prelude_info.wrapped_write_memory_8 = EmitWrappedWriteCallTrampoline<&A32::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u8));
+    prelude_info.wrapped_write_memory_16 = EmitWrappedWriteCallTrampoline<&A32::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u16));
+    prelude_info.wrapped_write_memory_32 = EmitWrappedWriteCallTrampoline<&A32::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u32));
+    prelude_info.wrapped_write_memory_64 = EmitWrappedWriteCallTrampoline<&A32::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u64));
     prelude_info.exclusive_write_memory_8 = EmitExclusiveWriteCallTrampoline<&A32::UserCallbacks::MemoryWriteExclusive8, u8>(code, conf);
     prelude_info.exclusive_write_memory_16 = EmitExclusiveWriteCallTrampoline<&A32::UserCallbacks::MemoryWriteExclusive16, u16>(code, conf);
     prelude_info.exclusive_write_memory_32 = EmitExclusiveWriteCallTrampoline<&A32::UserCallbacks::MemoryWriteExclusive32, u32>(code, conf);

src/dynarmic/src/dynarmic/backend/arm64/a64_address_space.cpp

@@ -23,29 +23,65 @@
 namespace Dynarmic::Backend::Arm64 {
 
 template<auto mfp, typename T>
-static void* EmitCallTrampoline(oaknut::CodeGenerator& code, T* this_) {
+static void* EmitCallReadTrampoline(oaknut::CodeGenerator& code, T* this_, size_t bitsize) {
     using namespace oaknut::util;
-
     const auto info = Devirtualize<mfp>(this_);
-
     oaknut::Label l_addr, l_this;
-
     void* target = code.xptr<void*>();
+    // params = { this, vaddr, bitsize }
     code.LDR(X0, l_this);
+    // X1 = vaddr
+    code.MOV(X2, bitsize);
     code.LDR(Xscratch0, l_addr);
     code.BR(Xscratch0);
-
     code.align(8);
     code.l(l_this);
     code.dx(info.this_ptr);
     code.l(l_addr);
     code.dx(info.fn_ptr);
-
     return target;
 }
 
 template<auto mfp, typename T>
-static void* EmitWrappedReadCallTrampoline(oaknut::CodeGenerator& code, T* this_) {
+static void* EmitCallWriteTrampoline(oaknut::CodeGenerator& code, T* this_, size_t bitsize) {
+    using namespace oaknut::util;
+    const auto info = Devirtualize<mfp>(this_);
+    oaknut::Label l_addr, l_this;
+    void* target = code.xptr<void*>();
+    // params = { this, vaddr, value, bitsize }
+    code.LDR(X0, l_this);
+    // X1 = vaddr
+    // X2 = value
+    code.MOV(X3, bitsize);
+    code.LDR(Xscratch0, l_addr);
+    code.BR(Xscratch0);
+    code.align(8);
+    code.l(l_this);
+    code.dx(info.this_ptr);
+    code.l(l_addr);
+    code.dx(info.fn_ptr);
+    return target;
+}
+
+template<auto mfp, typename T>
+static void* EmitCallTrampoline(oaknut::CodeGenerator& code, T* this_) {
+    using namespace oaknut::util;
+    const auto info = Devirtualize<mfp>(this_);
+    oaknut::Label l_addr, l_this;
+    void* target = code.xptr<void*>();
+    code.LDR(X0, l_this);
+    code.LDR(Xscratch0, l_addr);
+    code.BR(Xscratch0);
+    code.align(8);
+    code.l(l_this);
+    code.dx(info.this_ptr);
+    code.l(l_addr);
+    code.dx(info.fn_ptr);
+    return target;
+}
+
+template<auto mfp, typename T>
+static void* EmitWrappedReadCallTrampoline(oaknut::CodeGenerator& code, T* this_, size_t bitsize) {
     using namespace oaknut::util;
 
     const auto info = Devirtualize<mfp>(this_);
@@ -58,6 +94,7 @@ static void* EmitWrappedReadCallTrampoline(oaknut::CodeGenerator& code, T* this_
     ABI_PushRegisters(code, save_regs, 0);
     code.LDR(X0, l_this);
     code.MOV(X1, Xscratch0);
+    code.MOV(X2, bitsize);
     code.LDR(Xscratch0, l_addr);
     code.BLR(Xscratch0);
     code.MOV(Xscratch0, X0);
@@ -81,7 +118,7 @@ static void* EmitExclusiveReadCallTrampoline(oaknut::CodeGenerator& code, const
 
     auto fn = [](const A64::UserConfig& conf, A64::VAddr vaddr) -> T {
         return conf.global_monitor->ReadAndMark<T>(conf.processor_id, vaddr, [&]() -> T {
-            return (conf.callbacks->*callback)(vaddr);
+            return (conf.callbacks->*callback)(vaddr, sizeof(T));
         });
     };
 
@@ -100,7 +137,7 @@ static void* EmitExclusiveReadCallTrampoline(oaknut::CodeGenerator& code, const
 }
 
 template<auto mfp, typename T>
-static void* EmitWrappedWriteCallTrampoline(oaknut::CodeGenerator& code, T* this_) {
+static void* EmitWrappedWriteCallTrampoline(oaknut::CodeGenerator& code, T* this_, size_t bitsize) {
     using namespace oaknut::util;
 
     const auto info = Devirtualize<mfp>(this_);
@@ -114,6 +151,7 @@ static void* EmitWrappedWriteCallTrampoline(oaknut::CodeGenerator& code, T* this
     code.LDR(X0, l_this);
     code.MOV(X1, Xscratch0);
     code.MOV(X2, Xscratch1);
+    code.MOV(X3, bitsize);
     code.LDR(Xscratch0, l_addr);
     code.BLR(Xscratch0);
     ABI_PopRegisters(code, save_regs, 0);
@@ -133,14 +171,10 @@ static void* EmitExclusiveWriteCallTrampoline(oaknut::CodeGenerator& code, const
     using namespace oaknut::util;
 
     oaknut::Label l_addr, l_this;
-
     auto fn = [](const A64::UserConfig& conf, A64::VAddr vaddr, T value) -> u32 {
-        return conf.global_monitor->DoExclusiveOperation<T>(conf.processor_id, vaddr,
-                                                            [&](T expected) -> bool {
-                                                                return (conf.callbacks->*callback)(vaddr, value, expected);
-                                                            })
-                 ? 0
-                 : 1;
+        return conf.global_monitor->DoExclusiveOperation<T>(conf.processor_id, vaddr, [&](T expected) -> bool {
+            return (conf.callbacks->*callback)(vaddr, value, expected);
+        }) ? 0 : 1;
     };
 
     void* target = code.xptr<void*>();
@@ -346,30 +380,30 @@ void A64AddressSpace::EmitPrelude() {
 
     UnprotectCodeMemory();
 
-    prelude_info.read_memory_8 = EmitCallTrampoline<&A64::UserCallbacks::MemoryRead8>(code, conf.callbacks);
-    prelude_info.read_memory_16 = EmitCallTrampoline<&A64::UserCallbacks::MemoryRead16>(code, conf.callbacks);
-    prelude_info.read_memory_32 = EmitCallTrampoline<&A64::UserCallbacks::MemoryRead32>(code, conf.callbacks);
-    prelude_info.read_memory_64 = EmitCallTrampoline<&A64::UserCallbacks::MemoryRead64>(code, conf.callbacks);
+    prelude_info.read_memory_8 = EmitCallReadTrampoline<&A64::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u8));
+    prelude_info.read_memory_16 = EmitCallReadTrampoline<&A64::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u16));
+    prelude_info.read_memory_32 = EmitCallReadTrampoline<&A64::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u32));
+    prelude_info.read_memory_64 = EmitCallReadTrampoline<&A64::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u64));
     prelude_info.read_memory_128 = EmitRead128CallTrampoline(code, conf.callbacks);
-    prelude_info.wrapped_read_memory_8 = EmitWrappedReadCallTrampoline<&A64::UserCallbacks::MemoryRead8>(code, conf.callbacks);
-    prelude_info.wrapped_read_memory_16 = EmitWrappedReadCallTrampoline<&A64::UserCallbacks::MemoryRead16>(code, conf.callbacks);
-    prelude_info.wrapped_read_memory_32 = EmitWrappedReadCallTrampoline<&A64::UserCallbacks::MemoryRead32>(code, conf.callbacks);
-    prelude_info.wrapped_read_memory_64 = EmitWrappedReadCallTrampoline<&A64::UserCallbacks::MemoryRead64>(code, conf.callbacks);
+    prelude_info.wrapped_read_memory_8 = EmitWrappedReadCallTrampoline<&A64::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u8));
+    prelude_info.wrapped_read_memory_16 = EmitWrappedReadCallTrampoline<&A64::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u16));
+    prelude_info.wrapped_read_memory_32 = EmitWrappedReadCallTrampoline<&A64::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u32));
+    prelude_info.wrapped_read_memory_64 = EmitWrappedReadCallTrampoline<&A64::UserCallbacks::MemoryRead>(code, conf.callbacks, sizeof(u64));
     prelude_info.wrapped_read_memory_128 = EmitWrappedRead128CallTrampoline(code, conf.callbacks);
-    prelude_info.exclusive_read_memory_8 = EmitExclusiveReadCallTrampoline<&A64::UserCallbacks::MemoryRead8, u8>(code, conf);
-    prelude_info.exclusive_read_memory_16 = EmitExclusiveReadCallTrampoline<&A64::UserCallbacks::MemoryRead16, u16>(code, conf);
-    prelude_info.exclusive_read_memory_32 = EmitExclusiveReadCallTrampoline<&A64::UserCallbacks::MemoryRead32, u32>(code, conf);
-    prelude_info.exclusive_read_memory_64 = EmitExclusiveReadCallTrampoline<&A64::UserCallbacks::MemoryRead64, u64>(code, conf);
+    prelude_info.exclusive_read_memory_8 = EmitExclusiveReadCallTrampoline<&A64::UserCallbacks::MemoryRead, u8>(code, conf);
+    prelude_info.exclusive_read_memory_16 = EmitExclusiveReadCallTrampoline<&A64::UserCallbacks::MemoryRead, u16>(code, conf);
+    prelude_info.exclusive_read_memory_32 = EmitExclusiveReadCallTrampoline<&A64::UserCallbacks::MemoryRead, u32>(code, conf);
+    prelude_info.exclusive_read_memory_64 = EmitExclusiveReadCallTrampoline<&A64::UserCallbacks::MemoryRead, u64>(code, conf);
     prelude_info.exclusive_read_memory_128 = EmitExclusiveRead128CallTrampoline(code, conf);
-    prelude_info.write_memory_8 = EmitCallTrampoline<&A64::UserCallbacks::MemoryWrite8>(code, conf.callbacks);
-    prelude_info.write_memory_16 = EmitCallTrampoline<&A64::UserCallbacks::MemoryWrite16>(code, conf.callbacks);
-    prelude_info.write_memory_32 = EmitCallTrampoline<&A64::UserCallbacks::MemoryWrite32>(code, conf.callbacks);
-    prelude_info.write_memory_64 = EmitCallTrampoline<&A64::UserCallbacks::MemoryWrite64>(code, conf.callbacks);
+    prelude_info.write_memory_8 = EmitCallWriteTrampoline<&A64::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u8));
+    prelude_info.write_memory_16 = EmitCallWriteTrampoline<&A64::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u16));
+    prelude_info.write_memory_32 = EmitCallWriteTrampoline<&A64::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u32));
+    prelude_info.write_memory_64 = EmitCallWriteTrampoline<&A64::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u64));
     prelude_info.write_memory_128 = EmitWrite128CallTrampoline(code, conf.callbacks);
-    prelude_info.wrapped_write_memory_8 = EmitWrappedWriteCallTrampoline<&A64::UserCallbacks::MemoryWrite8>(code, conf.callbacks);
-    prelude_info.wrapped_write_memory_16 = EmitWrappedWriteCallTrampoline<&A64::UserCallbacks::MemoryWrite16>(code, conf.callbacks);
-    prelude_info.wrapped_write_memory_32 = EmitWrappedWriteCallTrampoline<&A64::UserCallbacks::MemoryWrite32>(code, conf.callbacks);
-    prelude_info.wrapped_write_memory_64 = EmitWrappedWriteCallTrampoline<&A64::UserCallbacks::MemoryWrite64>(code, conf.callbacks);
+    prelude_info.wrapped_write_memory_8 = EmitWrappedWriteCallTrampoline<&A64::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u8));
+    prelude_info.wrapped_write_memory_16 = EmitWrappedWriteCallTrampoline<&A64::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u16));
+    prelude_info.wrapped_write_memory_32 = EmitWrappedWriteCallTrampoline<&A64::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u32));
+    prelude_info.wrapped_write_memory_64 = EmitWrappedWriteCallTrampoline<&A64::UserCallbacks::MemoryWrite>(code, conf.callbacks, sizeof(u64));
     prelude_info.wrapped_write_memory_128 = EmitWrappedWrite128CallTrampoline(code, conf.callbacks);
     prelude_info.exclusive_write_memory_8 = EmitExclusiveWriteCallTrampoline<&A64::UserCallbacks::MemoryWriteExclusive8, u8>(code, conf);
     prelude_info.exclusive_write_memory_16 = EmitExclusiveWriteCallTrampoline<&A64::UserCallbacks::MemoryWriteExclusive16, u16>(code, conf);

src/dynarmic/src/dynarmic/backend/arm64/address_space.cpp

@@ -145,6 +145,7 @@ void AddressSpace::Link(EmittedBlockInfo& block_info) {
         case LinkTarget::ReturnFromRunCode:
             c.B(prelude_info.return_from_run_code);
             break;
+        // { this, vaddr, size }
         case LinkTarget::ReadMemory8:
             c.BL(prelude_info.read_memory_8);
             break;
@@ -160,6 +161,7 @@ void AddressSpace::Link(EmittedBlockInfo& block_info) {
         case LinkTarget::ReadMemory128:
             c.BL(prelude_info.read_memory_128);
             break;
+        // { this, vaddr, size }
         case LinkTarget::WrappedReadMemory8:
             c.BL(prelude_info.wrapped_read_memory_8);
             break;
@@ -175,6 +177,7 @@ void AddressSpace::Link(EmittedBlockInfo& block_info) {
         case LinkTarget::WrappedReadMemory128:
             c.BL(prelude_info.wrapped_read_memory_128);
             break;
+        //
         case LinkTarget::ExclusiveReadMemory8:
             c.BL(prelude_info.exclusive_read_memory_8);
             break;
@@ -190,6 +193,7 @@ void AddressSpace::Link(EmittedBlockInfo& block_info) {
         case LinkTarget::ExclusiveReadMemory128:
             c.BL(prelude_info.exclusive_read_memory_128);
             break;
+        // { this, vaddr, value, size }
         case LinkTarget::WriteMemory8:
             c.BL(prelude_info.write_memory_8);
             break;
@@ -205,6 +209,7 @@ void AddressSpace::Link(EmittedBlockInfo& block_info) {
         case LinkTarget::WriteMemory128:
             c.BL(prelude_info.write_memory_128);
             break;
+        //
         case LinkTarget::WrappedWriteMemory8:
             c.BL(prelude_info.wrapped_write_memory_8);
             break;
@@ -220,6 +225,7 @@ void AddressSpace::Link(EmittedBlockInfo& block_info) {
         case LinkTarget::WrappedWriteMemory128:
             c.BL(prelude_info.wrapped_write_memory_128);
             break;
+        //
         case LinkTarget::ExclusiveWriteMemory8:
             c.BL(prelude_info.exclusive_write_memory_8);
             break;

src/dynarmic/src/dynarmic/backend/x64/a32_emit_x64_memory.cpp

@@ -31,16 +31,16 @@ using namespace Xbyak::util;
 void A32EmitX64::GenFastmemFallbacks() {
     const std::initializer_list<int> idxes{0, 1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14};
     const std::array<std::pair<size_t, ArgCallback>, 4> read_callbacks{{
-        {8, Devirtualize<&A32::UserCallbacks::MemoryRead8>(conf.callbacks)},
-        {16, Devirtualize<&A32::UserCallbacks::MemoryRead16>(conf.callbacks)},
-        {32, Devirtualize<&A32::UserCallbacks::MemoryRead32>(conf.callbacks)},
-        {64, Devirtualize<&A32::UserCallbacks::MemoryRead64>(conf.callbacks)},
+        {8, Devirtualize<&A32::UserCallbacks::MemoryRead>(conf.callbacks)},
+        {16, Devirtualize<&A32::UserCallbacks::MemoryRead>(conf.callbacks)},
+        {32, Devirtualize<&A32::UserCallbacks::MemoryRead>(conf.callbacks)},
+        {64, Devirtualize<&A32::UserCallbacks::MemoryRead>(conf.callbacks)},
     }};
     const std::array<std::pair<size_t, ArgCallback>, 4> write_callbacks{{
-        {8, Devirtualize<&A32::UserCallbacks::MemoryWrite8>(conf.callbacks)},
-        {16, Devirtualize<&A32::UserCallbacks::MemoryWrite16>(conf.callbacks)},
-        {32, Devirtualize<&A32::UserCallbacks::MemoryWrite32>(conf.callbacks)},
-        {64, Devirtualize<&A32::UserCallbacks::MemoryWrite64>(conf.callbacks)},
+        {8, Devirtualize<&A32::UserCallbacks::MemoryWrite>(conf.callbacks)},
+        {16, Devirtualize<&A32::UserCallbacks::MemoryWrite>(conf.callbacks)},
+        {32, Devirtualize<&A32::UserCallbacks::MemoryWrite>(conf.callbacks)},
+        {64, Devirtualize<&A32::UserCallbacks::MemoryWrite>(conf.callbacks)},
     }};
     const std::array<std::pair<size_t, ArgCallback>, 4> exclusive_write_callbacks{{
         {8, Devirtualize<&A32::UserCallbacks::MemoryWriteExclusive8>(conf.callbacks)},
@@ -56,12 +56,12 @@ void A32EmitX64::GenFastmemFallbacks() {
                     code.align();
                     read_fallbacks[std::make_tuple(ordered, bitsize, vaddr_idx, value_idx)] = code.getCurr<void (*)()>();
                     ABI_PushCallerSaveRegistersAndAdjustStackExcept(code, HostLocRegIdx(value_idx));
+                    // params = { this, vaddr, size }
                     if (vaddr_idx != code.ABI_PARAM2.getIdx()) {
                         code.mov(code.ABI_PARAM2, Xbyak::Reg64{vaddr_idx});
                     }
-                    if (ordered) {
-                        code.mfence();
-                    }
+                    code.mov(code.ABI_PARAM3, bitsize / CHAR_BIT);
+                    if (ordered) code.mfence();
                     callback.EmitCall(code);
                     if (value_idx != code.ABI_RETURN.getIdx()) {
                         code.mov(Xbyak::Reg64{value_idx}, code.ABI_RETURN);
@@ -76,6 +76,7 @@ void A32EmitX64::GenFastmemFallbacks() {
                     code.align();
                     write_fallbacks[std::make_tuple(ordered, bitsize, vaddr_idx, value_idx)] = code.getCurr<void (*)()>();
                     ABI_PushCallerSaveRegistersAndAdjustStack(code);
+                    // params = { this, vaddr, value, size }
                     if (vaddr_idx == code.ABI_PARAM3.getIdx() && value_idx == code.ABI_PARAM2.getIdx()) {
                         code.xchg(code.ABI_PARAM2, code.ABI_PARAM3);
                     } else if (vaddr_idx == code.ABI_PARAM3.getIdx()) {
@@ -92,10 +93,9 @@ void A32EmitX64::GenFastmemFallbacks() {
                         }
                     }
                     code.ZeroExtendFrom(bitsize, code.ABI_PARAM3);
+                    code.mov(code.ABI_PARAM4, bitsize / CHAR_BIT);
                     callback.EmitCall(code);
-                    if (ordered) {
-                        code.mfence();
-                    }
+                    if (ordered) code.mfence();
                     ABI_PopCallerSaveRegistersAndAdjustStack(code);
                     code.ret();
                     PerfMapRegister(write_fallbacks[std::make_tuple(ordered, bitsize, vaddr_idx, value_idx)], code.getCurr(), fmt::format("a32_write_fallback_{}", bitsize));
@@ -138,35 +138,35 @@ void A32EmitX64::GenFastmemFallbacks() {
 #undef Axx
 
 void A32EmitX64::EmitA32ReadMemory8(A32EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryRead<8, &A32::UserCallbacks::MemoryRead8>(ctx, inst);
+    EmitMemoryRead<8, &A32::UserCallbacks::MemoryRead>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32ReadMemory16(A32EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryRead<16, &A32::UserCallbacks::MemoryRead16>(ctx, inst);
+    EmitMemoryRead<16, &A32::UserCallbacks::MemoryRead>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32ReadMemory32(A32EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryRead<32, &A32::UserCallbacks::MemoryRead32>(ctx, inst);
+    EmitMemoryRead<32, &A32::UserCallbacks::MemoryRead>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32ReadMemory64(A32EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryRead<64, &A32::UserCallbacks::MemoryRead64>(ctx, inst);
+    EmitMemoryRead<64, &A32::UserCallbacks::MemoryRead>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32WriteMemory8(A32EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<8, &A32::UserCallbacks::MemoryWrite8>(ctx, inst);
+    EmitMemoryWrite<8, &A32::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32WriteMemory16(A32EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<16, &A32::UserCallbacks::MemoryWrite16>(ctx, inst);
+    EmitMemoryWrite<16, &A32::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32WriteMemory32(A32EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<32, &A32::UserCallbacks::MemoryWrite32>(ctx, inst);
+    EmitMemoryWrite<32, &A32::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32WriteMemory64(A32EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<64, &A32::UserCallbacks::MemoryWrite64>(ctx, inst);
+    EmitMemoryWrite<64, &A32::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32ClearExclusive(A32EmitContext&, IR::Inst*) {
@@ -175,33 +175,33 @@ void A32EmitX64::EmitA32ClearExclusive(A32EmitContext&, IR::Inst*) {
 
 void A32EmitX64::EmitA32ExclusiveReadMemory8(A32EmitContext& ctx, IR::Inst* inst) {
     if (conf.fastmem_exclusive_access) {
-        EmitExclusiveReadMemoryInline<8, &A32::UserCallbacks::MemoryRead8>(ctx, inst);
+        EmitExclusiveReadMemoryInline<8, &A32::UserCallbacks::MemoryRead>(ctx, inst);
     } else {
-        EmitExclusiveReadMemory<8, &A32::UserCallbacks::MemoryRead8>(ctx, inst);
+        EmitExclusiveReadMemory<8, &A32::UserCallbacks::MemoryRead>(ctx, inst);
     }
 }
 
 void A32EmitX64::EmitA32ExclusiveReadMemory16(A32EmitContext& ctx, IR::Inst* inst) {
     if (conf.fastmem_exclusive_access) {
-        EmitExclusiveReadMemoryInline<16, &A32::UserCallbacks::MemoryRead16>(ctx, inst);
+        EmitExclusiveReadMemoryInline<16, &A32::UserCallbacks::MemoryRead>(ctx, inst);
     } else {
-        EmitExclusiveReadMemory<16, &A32::UserCallbacks::MemoryRead16>(ctx, inst);
+        EmitExclusiveReadMemory<16, &A32::UserCallbacks::MemoryRead>(ctx, inst);
     }
 }
 
 void A32EmitX64::EmitA32ExclusiveReadMemory32(A32EmitContext& ctx, IR::Inst* inst) {
     if (conf.fastmem_exclusive_access) {
-        EmitExclusiveReadMemoryInline<32, &A32::UserCallbacks::MemoryRead32>(ctx, inst);
+        EmitExclusiveReadMemoryInline<32, &A32::UserCallbacks::MemoryRead>(ctx, inst);
     } else {
-        EmitExclusiveReadMemory<32, &A32::UserCallbacks::MemoryRead32>(ctx, inst);
+        EmitExclusiveReadMemory<32, &A32::UserCallbacks::MemoryRead>(ctx, inst);
     }
 }
 
 void A32EmitX64::EmitA32ExclusiveReadMemory64(A32EmitContext& ctx, IR::Inst* inst) {
     if (conf.fastmem_exclusive_access) {
-        EmitExclusiveReadMemoryInline<64, &A32::UserCallbacks::MemoryRead64>(ctx, inst);
+        EmitExclusiveReadMemoryInline<64, &A32::UserCallbacks::MemoryRead>(ctx, inst);
     } else {
-        EmitExclusiveReadMemory<64, &A32::UserCallbacks::MemoryRead64>(ctx, inst);
+        EmitExclusiveReadMemory<64, &A32::UserCallbacks::MemoryRead>(ctx, inst);
     }
 }
 

src/dynarmic/src/dynarmic/backend/x64/a64_emit_x64_memory.cpp

@@ -115,16 +115,16 @@ void A64EmitX64::GenMemory128Accessors() {
 void A64EmitX64::GenFastmemFallbacks() {
     const std::initializer_list<int> idxes{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};
     const std::array<std::pair<size_t, ArgCallback>, 4> read_callbacks{{
-        {8, Devirtualize<&A64::UserCallbacks::MemoryRead8>(conf.callbacks)},
-        {16, Devirtualize<&A64::UserCallbacks::MemoryRead16>(conf.callbacks)},
-        {32, Devirtualize<&A64::UserCallbacks::MemoryRead32>(conf.callbacks)},
-        {64, Devirtualize<&A64::UserCallbacks::MemoryRead64>(conf.callbacks)},
+        {8, Devirtualize<&A64::UserCallbacks::MemoryRead>(conf.callbacks)},
+        {16, Devirtualize<&A64::UserCallbacks::MemoryRead>(conf.callbacks)},
+        {32, Devirtualize<&A64::UserCallbacks::MemoryRead>(conf.callbacks)},
+        {64, Devirtualize<&A64::UserCallbacks::MemoryRead>(conf.callbacks)},
     }};
     const std::array<std::pair<size_t, ArgCallback>, 4> write_callbacks{{
-        {8, Devirtualize<&A64::UserCallbacks::MemoryWrite8>(conf.callbacks)},
-        {16, Devirtualize<&A64::UserCallbacks::MemoryWrite16>(conf.callbacks)},
-        {32, Devirtualize<&A64::UserCallbacks::MemoryWrite32>(conf.callbacks)},
-        {64, Devirtualize<&A64::UserCallbacks::MemoryWrite64>(conf.callbacks)},
+        {8, Devirtualize<&A64::UserCallbacks::MemoryWrite>(conf.callbacks)},
+        {16, Devirtualize<&A64::UserCallbacks::MemoryWrite>(conf.callbacks)},
+        {32, Devirtualize<&A64::UserCallbacks::MemoryWrite>(conf.callbacks)},
+        {64, Devirtualize<&A64::UserCallbacks::MemoryWrite>(conf.callbacks)},
     }};
     const std::array<std::pair<size_t, ArgCallback>, 4> exclusive_write_callbacks{{
         {8, Devirtualize<&A64::UserCallbacks::MemoryWriteExclusive8>(conf.callbacks)},
@@ -204,12 +204,12 @@ void A64EmitX64::GenFastmemFallbacks() {
                     code.align();
                     read_fallbacks[std::make_tuple(ordered, bitsize, vaddr_idx, value_idx)] = code.getCurr<void (*)()>();
                     ABI_PushCallerSaveRegistersAndAdjustStackExcept(code, HostLocRegIdx(value_idx));
+                    // params = { this, vaddr, size }
                     if (vaddr_idx != code.ABI_PARAM2.getIdx()) {
                         code.mov(code.ABI_PARAM2, Xbyak::Reg64{vaddr_idx});
                     }
-                    if (ordered) {
-                        code.mfence();
-                    }
+                    code.mov(code.ABI_PARAM3, bitsize / CHAR_BIT);
+                    if (ordered) code.mfence();
                     callback.EmitCall(code);
                     if (value_idx != code.ABI_RETURN.getIdx()) {
                         code.mov(Xbyak::Reg64{value_idx}, code.ABI_RETURN);
@@ -224,6 +224,7 @@ void A64EmitX64::GenFastmemFallbacks() {
                     code.align();
                     write_fallbacks[std::make_tuple(ordered, bitsize, vaddr_idx, value_idx)] = code.getCurr<void (*)()>();
                     ABI_PushCallerSaveRegistersAndAdjustStack(code);
+                    // params = { this, vaddr, value, size }
                     if (vaddr_idx == code.ABI_PARAM3.getIdx() && value_idx == code.ABI_PARAM2.getIdx()) {
                         code.xchg(code.ABI_PARAM2, code.ABI_PARAM3);
                     } else if (vaddr_idx == code.ABI_PARAM3.getIdx()) {
@@ -240,10 +241,9 @@ void A64EmitX64::GenFastmemFallbacks() {
                         }
                     }
                     code.ZeroExtendFrom(bitsize, code.ABI_PARAM3);
+                    code.mov(code.ABI_PARAM4, bitsize / CHAR_BIT);
                     callback.EmitCall(code);
-                    if (ordered) {
-                        code.mfence();
-                    }
+                    if (ordered) code.mfence();
                     ABI_PopCallerSaveRegistersAndAdjustStack(code);
                     code.ret();
                     PerfMapRegister(write_fallbacks[std::make_tuple(ordered, bitsize, vaddr_idx, value_idx)], code.getCurr(), fmt::format("a64_write_fallback_{}", bitsize));
@@ -286,19 +286,19 @@ void A64EmitX64::GenFastmemFallbacks() {
 #undef Axx
 
 void A64EmitX64::EmitA64ReadMemory8(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryRead<8, &A64::UserCallbacks::MemoryRead8>(ctx, inst);
+    EmitMemoryRead<8, &A64::UserCallbacks::MemoryRead>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64ReadMemory16(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryRead<16, &A64::UserCallbacks::MemoryRead16>(ctx, inst);
+    EmitMemoryRead<16, &A64::UserCallbacks::MemoryRead>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64ReadMemory32(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryRead<32, &A64::UserCallbacks::MemoryRead32>(ctx, inst);
+    EmitMemoryRead<32, &A64::UserCallbacks::MemoryRead>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64ReadMemory64(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryRead<64, &A64::UserCallbacks::MemoryRead64>(ctx, inst);
+    EmitMemoryRead<64, &A64::UserCallbacks::MemoryRead>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64ReadMemory128(A64EmitContext& ctx, IR::Inst* inst) {
@@ -306,23 +306,23 @@ void A64EmitX64::EmitA64ReadMemory128(A64EmitContext& ctx, IR::Inst* inst) {
 }
 
 void A64EmitX64::EmitA64WriteMemory8(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<8, &A64::UserCallbacks::MemoryWrite8>(ctx, inst);
+    EmitMemoryWrite<8, &A64::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64WriteMemory16(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<16, &A64::UserCallbacks::MemoryWrite16>(ctx, inst);
+    EmitMemoryWrite<16, &A64::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64WriteMemory32(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<32, &A64::UserCallbacks::MemoryWrite32>(ctx, inst);
+    EmitMemoryWrite<32, &A64::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64WriteMemory64(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<64, &A64::UserCallbacks::MemoryWrite64>(ctx, inst);
+    EmitMemoryWrite<64, &A64::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64WriteMemory128(A64EmitContext& ctx, IR::Inst* inst) {
-    EmitMemoryWrite<128, &A64::UserCallbacks::MemoryWrite64>(ctx, inst);
+    EmitMemoryWrite<128, &A64::UserCallbacks::MemoryWrite>(ctx, inst);
 }
 
 void A64EmitX64::EmitA64ClearExclusive(A64EmitContext&, IR::Inst*) {
@@ -331,33 +331,33 @@ void A64EmitX64::EmitA64ClearExclusive(A64EmitContext&, IR::Inst*) {
 
 void A64EmitX64::EmitA64ExclusiveReadMemory8(A64EmitContext& ctx, IR::Inst* inst) {
     if (conf.fastmem_exclusive_access) {
-        EmitExclusiveReadMemoryInline<8, &A64::UserCallbacks::MemoryRead8>(ctx, inst);
+        EmitExclusiveReadMemoryInline<8, &A64::UserCallbacks::MemoryRead>(ctx, inst);
     } else {
-        EmitExclusiveReadMemory<8, &A64::UserCallbacks::MemoryRead8>(ctx, inst);
+        EmitExclusiveReadMemory<8, &A64::UserCallbacks::MemoryRead>(ctx, inst);
     }
 }
 
 void A64EmitX64::EmitA64ExclusiveReadMemory16(A64EmitContext& ctx, IR::Inst* inst) {
     if (conf.fastmem_exclusive_access) {
-        EmitExclusiveReadMemoryInline<16, &A64::UserCallbacks::MemoryRead16>(ctx, inst);
+        EmitExclusiveReadMemoryInline<16, &A64::UserCallbacks::MemoryRead>(ctx, inst);
     } else {
-        EmitExclusiveReadMemory<16, &A64::UserCallbacks::MemoryRead16>(ctx, inst);
+        EmitExclusiveReadMemory<16, &A64::UserCallbacks::MemoryRead>(ctx, inst);
     }
 }
 
 void A64EmitX64::EmitA64ExclusiveReadMemory32(A64EmitContext& ctx, IR::Inst* inst) {
     if (conf.fastmem_exclusive_access) {
-        EmitExclusiveReadMemoryInline<32, &A64::UserCallbacks::MemoryRead32>(ctx, inst);
+        EmitExclusiveReadMemoryInline<32, &A64::UserCallbacks::MemoryRead>(ctx, inst);
     } else {
-        EmitExclusiveReadMemory<32, &A64::UserCallbacks::MemoryRead32>(ctx, inst);
+        EmitExclusiveReadMemory<32, &A64::UserCallbacks::MemoryRead>(ctx, inst);
     }
 }
 
 void A64EmitX64::EmitA64ExclusiveReadMemory64(A64EmitContext& ctx, IR::Inst* inst) {
     if (conf.fastmem_exclusive_access) {
-        EmitExclusiveReadMemoryInline<64, &A64::UserCallbacks::MemoryRead64>(ctx, inst);
+        EmitExclusiveReadMemoryInline<64, &A64::UserCallbacks::MemoryRead>(ctx, inst);
     } else {
-        EmitExclusiveReadMemory<64, &A64::UserCallbacks::MemoryRead64>(ctx, inst);
+        EmitExclusiveReadMemory<64, &A64::UserCallbacks::MemoryRead>(ctx, inst);
     }
 }
 

src/dynarmic/src/dynarmic/backend/x64/emit_x64_memory.cpp.inc

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 /* This file is part of the dynarmic project.
  * Copyright (c) 2022 MerryMage
  * SPDX-License-Identifier: 0BSD
@@ -56,16 +59,13 @@ void AxxEmitX64::EmitMemoryRead(AxxEmitContext& ctx, IR::Inst* inst) {
         // Neither fastmem nor page table: Use callbacks
         if constexpr (bitsize == 128) {
             ctx.reg_alloc.HostCall(code, nullptr, {}, args[1]);
-            if (ordered) {
-                code.mfence();
-            }
+            if (ordered) code.mfence();
             code.CallFunction(memory_read_128);
             ctx.reg_alloc.DefineValue(code, inst, xmm1);
         } else {
             ctx.reg_alloc.HostCall(code, inst, {}, args[1]);
-            if (ordered) {
-                code.mfence();
-            }
+            code.mov(code.ABI_PARAM3.cvt32(), bitsize / CHAR_BIT);
+            if (ordered) code.mfence();
             Devirtualize<callback>(conf.callbacks).EmitCall(code);
             code.ZeroExtendFrom(bitsize, code.ABI_RETURN);
         }
@@ -148,12 +148,12 @@ void AxxEmitX64::EmitMemoryWrite(AxxEmitContext& ctx, IR::Inst* inst) {
             ctx.reg_alloc.HostCall(code, nullptr);
             code.CallFunction(memory_write_128);
         } else {
+            // { this, vaddr, value, size }
             ctx.reg_alloc.HostCall(code, nullptr, {}, args[1], args[2]);
+            code.mov(code.ABI_PARAM4.cvt32(), bitsize / CHAR_BIT);
             Devirtualize<callback>(conf.callbacks).EmitCall(code);
         }
-        if (ordered) {
-            code.mfence();
-        }
+        if (ordered) code.mfence();
         EmitCheckMemoryAbort(ctx, inst);
         return;
     }
@@ -230,12 +230,11 @@ void AxxEmitX64::EmitExclusiveReadMemory(AxxEmitContext& ctx, IR::Inst* inst) {
         if (ordered) {
             code.mfence();
         }
-        code.CallLambda(
-            [](AxxUserConfig& conf, Axx::VAddr vaddr) -> T {
-                return conf.global_monitor->ReadAndMark<T>(conf.processor_id, vaddr, [&]() -> T {
-                    return (conf.callbacks->*callback)(vaddr);
-                });
+        code.CallLambda([](AxxUserConfig& conf, Axx::VAddr vaddr) -> T {
+            return conf.global_monitor->ReadAndMark<T>(conf.processor_id, vaddr, [&]() -> T {
+                return (conf.callbacks->*callback)(vaddr, bitsize / CHAR_BIT);
             });
+        });
         code.ZeroExtendFrom(bitsize, code.ABI_RETURN);
     } else {
         const Xbyak::Xmm result = ctx.reg_alloc.ScratchXmm(code);
@@ -250,12 +249,11 @@ void AxxEmitX64::EmitExclusiveReadMemory(AxxEmitContext& ctx, IR::Inst* inst) {
         if (ordered) {
             code.mfence();
         }
-        code.CallLambda(
-            [](AxxUserConfig& conf, Axx::VAddr vaddr, Vector& ret) {
-                ret = conf.global_monitor->ReadAndMark<Vector>(conf.processor_id, vaddr, [&]() -> Vector {
-                    return (conf.callbacks->*callback)(vaddr);
-                });
+        code.CallLambda([](AxxUserConfig& conf, Axx::VAddr vaddr, Vector& ret) {
+            ret = conf.global_monitor->ReadAndMark<Vector>(conf.processor_id, vaddr, [&]() -> Vector {
+                return (conf.callbacks->*callback)(vaddr);
             });
+        });
         code.movups(result, xword[rsp + ABI_SHADOW_SPACE]);
         ctx.reg_alloc.ReleaseStackSpace(code, 16 + ABI_SHADOW_SPACE);
 

src/dynarmic/src/dynarmic/interface/A32/config.h

@@ -66,7 +66,9 @@ struct UserCallbacks : public TranslateCallbacks {
 
     // All reads through this callback are 4-byte aligned.
     // Memory must be interpreted as little endian.
-    std::optional<std::uint32_t> MemoryReadCode(VAddr vaddr) override { return MemoryRead32(vaddr); }
+    std::optional<std::uint32_t> MemoryReadCode(VAddr vaddr) override {
+        return std::uint32_t(MemoryRead(vaddr, sizeof(std::uint32_t)));
+    }
 
     // This function is called before the instruction at pc is read.
     // IR code can be emitted by the callee prior to instruction handling.
@@ -80,16 +82,10 @@ struct UserCallbacks : public TranslateCallbacks {
 
     // Reads through these callbacks may not be aligned.
     // Memory must be interpreted as if ENDIANSTATE == 0, endianness will be corrected by the JIT.
-    virtual std::uint8_t MemoryRead8(VAddr vaddr) = 0;
-    virtual std::uint16_t MemoryRead16(VAddr vaddr) = 0;
-    virtual std::uint32_t MemoryRead32(VAddr vaddr) = 0;
-    virtual std::uint64_t MemoryRead64(VAddr vaddr) = 0;
+    virtual std::uint64_t MemoryRead(VAddr vaddr, std::size_t size) = 0;
 
     // Writes through these callbacks may not be aligned.
-    virtual void MemoryWrite8(VAddr vaddr, std::uint8_t value) = 0;
-    virtual void MemoryWrite16(VAddr vaddr, std::uint16_t value) = 0;
-    virtual void MemoryWrite32(VAddr vaddr, std::uint32_t value) = 0;
-    virtual void MemoryWrite64(VAddr vaddr, std::uint64_t value) = 0;
+    virtual void MemoryWrite(VAddr vaddr, std::uint64_t value, std::size_t size) = 0;
 
     // Writes through these callbacks may not be aligned.
     virtual bool MemoryWriteExclusive8(VAddr /*vaddr*/, std::uint8_t /*value*/, std::uint8_t /*expected*/) { return false; }

src/dynarmic/src/dynarmic/interface/A64/config.h

@@ -89,20 +89,16 @@ struct UserCallbacks {
 
     // All reads through this callback are 4-byte aligned.
     // Memory must be interpreted as little endian.
-    virtual std::optional<std::uint32_t> MemoryReadCode(VAddr vaddr) { return MemoryRead32(vaddr); }
+    virtual std::optional<std::uint32_t> MemoryReadCode(VAddr vaddr) {
+        return std::uint32_t(MemoryRead(vaddr, sizeof(std::uint32_t)));
+    }
 
     // Reads through these callbacks may not be aligned.
-    virtual std::uint8_t MemoryRead8(VAddr vaddr) = 0;
-    virtual std::uint16_t MemoryRead16(VAddr vaddr) = 0;
-    virtual std::uint32_t MemoryRead32(VAddr vaddr) = 0;
-    virtual std::uint64_t MemoryRead64(VAddr vaddr) = 0;
+    virtual std::uint64_t MemoryRead(VAddr vaddr, std::size_t size) = 0;
     virtual Vector MemoryRead128(VAddr vaddr) = 0;
 
     // Writes through these callbacks may not be aligned.
-    virtual void MemoryWrite8(VAddr vaddr, std::uint8_t value) = 0;
-    virtual void MemoryWrite16(VAddr vaddr, std::uint16_t value) = 0;
-    virtual void MemoryWrite32(VAddr vaddr, std::uint32_t value) = 0;
-    virtual void MemoryWrite64(VAddr vaddr, std::uint64_t value) = 0;
+    virtual void MemoryWrite(VAddr vaddr, std::uint64_t value, std::size_t size) = 0;
     virtual void MemoryWrite128(VAddr vaddr, Vector value) = 0;
 
     // Writes through these callbacks may not be aligned.

src/dynarmic/src/dynarmic/ir/opt_passes.cpp

@@ -40,7 +40,7 @@ static void ConstantMemoryReads(IR::Block& block, A32::UserCallbacks* cb) {
             if (inst.AreAllArgsImmediates()) {
                 const u32 vaddr = inst.GetArg(1).GetU32();
                 if (cb->IsReadOnlyMemory(vaddr)) {
-                    const u8 value_from_memory = cb->MemoryRead8(vaddr);
+                    const u8 value_from_memory = u8(cb->MemoryRead(vaddr, sizeof(u8)));
                     inst.ReplaceUsesWith(IR::Value{value_from_memory});
                 }
             }
@@ -51,7 +51,7 @@ static void ConstantMemoryReads(IR::Block& block, A32::UserCallbacks* cb) {
             if (inst.AreAllArgsImmediates()) {
                 const u32 vaddr = inst.GetArg(1).GetU32();
                 if (cb->IsReadOnlyMemory(vaddr)) {
-                    const u16 value_from_memory = cb->MemoryRead16(vaddr);
+                    const u16 value_from_memory = u16(cb->MemoryRead(vaddr, sizeof(u16)));
                     inst.ReplaceUsesWith(IR::Value{value_from_memory});
                 }
             }
@@ -62,7 +62,7 @@ static void ConstantMemoryReads(IR::Block& block, A32::UserCallbacks* cb) {
             if (inst.AreAllArgsImmediates()) {
                 const u32 vaddr = inst.GetArg(1).GetU32();
                 if (cb->IsReadOnlyMemory(vaddr)) {
-                    const u32 value_from_memory = cb->MemoryRead32(vaddr);
+                    const u32 value_from_memory = u32(cb->MemoryRead(vaddr, sizeof(u32)));
                     inst.ReplaceUsesWith(IR::Value{value_from_memory});
                 }
             }
@@ -73,7 +73,7 @@ static void ConstantMemoryReads(IR::Block& block, A32::UserCallbacks* cb) {
             if (inst.AreAllArgsImmediates()) {
                 const u32 vaddr = inst.GetArg(1).GetU32();
                 if (cb->IsReadOnlyMemory(vaddr)) {
-                    const u64 value_from_memory = cb->MemoryRead64(vaddr);
+                    const u64 value_from_memory = u64(cb->MemoryRead(vaddr, sizeof(u64)));
                     inst.ReplaceUsesWith(IR::Value{value_from_memory});
                 }
             }

src/dynarmic/tests/A32/fuzz_thumb.cpp

@@ -503,7 +503,7 @@ TEST_CASE("Fuzz Thumb32 instructions set", "[JitX64][Thumb][Thumb32]") {
     }
 }
 
-TEST_CASE("Verify fix for off by one error in MemoryRead32 worked", "[Thumb][Thumb16]") {
+TEST_CASE("Verify fix for off by one error in MemoryRead<32> worked", "[Thumb][Thumb16]") {
     ThumbTestEnv test_env;
 
     // Prepare test subjects

src/dynarmic/tests/A32/test_arm_instructions.cpp

@@ -553,9 +553,9 @@ TEST_CASE("arm: Memory access (fastmem)", "[arm][A32]") {
     memset(backing_memory, 0, memory_size);
     memcpy(backing_memory + 0x100, "Lorem ipsum dolor sit amet, consectetur adipiscing elit.", 57);
 
-    env.MemoryWrite32(0, 0xE5904000);  // LDR R4, [R0]
-    env.MemoryWrite32(4, 0xE5814000);  // STR R4, [R1]
-    env.MemoryWrite32(8, 0xEAFFFFFE);  // B .
+    env.MemoryWrite(0, 0xE5904000, sizeof(u32));  // LDR R4, [R0]
+    env.MemoryWrite(4, 0xE5814000, sizeof(u32));  // STR R4, [R1]
+    env.MemoryWrite(8, 0xEAFFFFFE, sizeof(u32));  // B .
     jit.Regs()[0] = 0x100;
     jit.Regs()[1] = 0x1F0;
     jit.Regs()[15] = 0;       // PC = 0

src/dynarmic/tests/A32/testenv.h

@@ -16,6 +16,7 @@
 
 #include "common/assert.h"
 #include "common/common_types.h"
+#include "dynarmic/frontend/A32/translate/translate_callbacks.h"
 #include "dynarmic/interface/A32/a32.h"
 
 template<typename InstructionType_, u32 infinite_loop_u32>
@@ -59,42 +60,51 @@ public:
         return infinite_loop_u32;  // B .
     }
 
-    std::uint8_t MemoryRead8(u32 vaddr) override {
-        if (IsInCodeMem(vaddr)) {
-            return reinterpret_cast<u8*>(code_mem.data())[vaddr];
+    u64 MemoryRead(u32 vaddr, size_t size) override {
+        switch (size) {
+        case sizeof(u64):
+            return MemoryRead(vaddr, sizeof(u32))
+                | MemoryRead(vaddr + sizeof(u32), sizeof(u32)) << 32;
+        case sizeof(u32):
+            return MemoryRead(vaddr, sizeof(u16))
+                | MemoryRead(vaddr + sizeof(u16), sizeof(u16)) << 16;
+        case sizeof(u16):
+            return MemoryRead(vaddr, sizeof(u8))
+                | MemoryRead(vaddr + sizeof(u8), sizeof(u8)) << 8;
+        case sizeof(u8): {
+            if (IsInCodeMem(vaddr))
+                return reinterpret_cast<u8*>(code_mem.data())[vaddr];
+            if (auto iter = modified_memory.find(vaddr); iter != modified_memory.end())
+                return iter->second;
+            return u8(vaddr);
         }
-        if (auto iter = modified_memory.find(vaddr); iter != modified_memory.end()) {
-            return iter->second;
+        default:
+            std::abort();
         }
-        return static_cast<u8>(vaddr);
-    }
-    std::uint16_t MemoryRead16(u32 vaddr) override {
-        return u16(MemoryRead8(vaddr)) | u16(MemoryRead8(vaddr + 1)) << 8;
-    }
-    std::uint32_t MemoryRead32(u32 vaddr) override {
-        return u32(MemoryRead16(vaddr)) | u32(MemoryRead16(vaddr + 2)) << 16;
-    }
-    std::uint64_t MemoryRead64(u32 vaddr) override {
-        return u64(MemoryRead32(vaddr)) | u64(MemoryRead32(vaddr + 4)) << 32;
     }
 
-    void MemoryWrite8(u32 vaddr, std::uint8_t value) override {
-        if (vaddr < code_mem.size() * sizeof(u32)) {
-            code_mem_modified_by_guest = true;
+    void MemoryWrite(Dynarmic::A32::VAddr vaddr, u64 value, size_t size) override {
+        switch (size) {
+        case sizeof(u64):
+            MemoryWrite(vaddr, u32(value), sizeof(u32));
+            MemoryWrite(vaddr + 4, u32(value >> 32), sizeof(u32));
+            break;
+        case sizeof(u32):
+            MemoryWrite(vaddr, u16(value), sizeof(u16));
+            MemoryWrite(vaddr + 2, u16(value >> 16), sizeof(u16));
+            break;
+        case sizeof(u16):
+            MemoryWrite(vaddr, u8(value), sizeof(u8));
+            MemoryWrite(vaddr + 1, u8(value >> 8), sizeof(u8));
+            break;
+        case sizeof(u8):
+            if (vaddr < code_mem.size() * sizeof(u32))
+                code_mem_modified_by_guest = true;
+            modified_memory[vaddr] = value;
+            break;
+        default:
+            std::abort();
         }
-        modified_memory[vaddr] = value;
-    }
-    void MemoryWrite16(u32 vaddr, std::uint16_t value) override {
-        MemoryWrite8(vaddr, static_cast<u8>(value));
-        MemoryWrite8(vaddr + 1, static_cast<u8>(value >> 8));
-    }
-    void MemoryWrite32(u32 vaddr, std::uint32_t value) override {
-        MemoryWrite16(vaddr, static_cast<u16>(value));
-        MemoryWrite16(vaddr + 2, static_cast<u16>(value >> 16));
-    }
-    void MemoryWrite64(u32 vaddr, std::uint64_t value) override {
-        MemoryWrite32(vaddr, static_cast<u32>(value));
-        MemoryWrite32(vaddr + 4, static_cast<u32>(value >> 32));
     }
 
     void CallSVC(std::uint32_t swi) override {
@@ -143,46 +153,41 @@ public:
         return read<std::uint32_t>(vaddr);
     }
 
-    std::uint8_t MemoryRead8(std::uint32_t vaddr) override {
-        return read<std::uint8_t>(vaddr);
-    }
-    std::uint16_t MemoryRead16(std::uint32_t vaddr) override {
-        return read<std::uint16_t>(vaddr);
-    }
-    std::uint32_t MemoryRead32(std::uint32_t vaddr) override {
-        return read<std::uint32_t>(vaddr);
-    }
-    std::uint64_t MemoryRead64(std::uint32_t vaddr) override {
-        return read<std::uint64_t>(vaddr);
+    u64 MemoryRead(u32 vaddr, size_t size) override {
+        switch (size) {
+        case sizeof(u64): return read<u64>(vaddr);
+        case sizeof(u32): return read<u32>(vaddr);
+        case sizeof(u16): return read<u16>(vaddr);
+        case sizeof(u8): return read<u8>(vaddr);
+        default:
+            std::abort();
+        }
     }
 
-    void MemoryWrite8(std::uint32_t vaddr, std::uint8_t value) override {
-        write(vaddr, value);
-    }
-    void MemoryWrite16(std::uint32_t vaddr, std::uint16_t value) override {
-        write(vaddr, value);
-    }
-    void MemoryWrite32(std::uint32_t vaddr, std::uint32_t value) override {
-        write(vaddr, value);
-    }
-    void MemoryWrite64(std::uint32_t vaddr, std::uint64_t value) override {
-        write(vaddr, value);
+    void MemoryWrite(Dynarmic::A32::VAddr vaddr, std::uint64_t value, size_t size) override {
+        switch (size) {
+        case sizeof(u64): return write<u64>(vaddr, u64(value));
+        case sizeof(u32): return write<u32>(vaddr, u32(value));
+        case sizeof(u16): return write<u16>(vaddr, u16(value));
+        case sizeof(u8): return write<u8>(vaddr, u8(value));
+        default: std::abort();
+        }
     }
 
-    bool MemoryWriteExclusive8(std::uint32_t vaddr, std::uint8_t value, [[maybe_unused]] std::uint8_t expected) override {
-        MemoryWrite8(vaddr, value);
+    bool MemoryWriteExclusive8(Dynarmic::A32::VAddr vaddr, std::uint8_t value, [[maybe_unused]] std::uint8_t expected) override {
+        MemoryWrite(vaddr, value, sizeof(u8));
         return true;
     }
-    bool MemoryWriteExclusive16(std::uint32_t vaddr, std::uint16_t value, [[maybe_unused]] std::uint16_t expected) override {
-        MemoryWrite16(vaddr, value);
+    bool MemoryWriteExclusive16(Dynarmic::A32::VAddr vaddr, std::uint16_t value, [[maybe_unused]] std::uint16_t expected) override {
+        MemoryWrite(vaddr, value, sizeof(u16));
         return true;
     }
-    bool MemoryWriteExclusive32(std::uint32_t vaddr, std::uint32_t value, [[maybe_unused]] std::uint32_t expected) override {
-        MemoryWrite32(vaddr, value);
+    bool MemoryWriteExclusive32(Dynarmic::A32::VAddr vaddr, std::uint32_t value, [[maybe_unused]] std::uint32_t expected) override {
+        MemoryWrite(vaddr, value, sizeof(u32));
         return true;
     }
-    bool MemoryWriteExclusive64(std::uint32_t vaddr, std::uint64_t value, [[maybe_unused]] std::uint64_t expected) override {
-        MemoryWrite64(vaddr, value);
+    bool MemoryWriteExclusive64(Dynarmic::A32::VAddr vaddr, std::uint64_t value, [[maybe_unused]] std::uint64_t expected) override {
+        MemoryWrite(vaddr, value, sizeof(u64));
         return true;
     }