rebuild

- [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 is a follow up for the biomutant fix. By the time i knew it was shallow, coz i was running patchtext routine even for svc and modules, causing a routine which were needed to run 14 times to run 4000+ instead.
Since some perf loss was reported in 30 sep nightly, and it caused ac3r to not boot, we had added a temp toggle, which is now being removed as optimizations + fix are in.

Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4523
Reviewed-by: lizzie <lizzie@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>

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

@@ -17,7 +17,6 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
     USE_CUSTOM_CPU_TICKS("use_custom_cpu_ticks"),
     SKIP_CPU_INNER_INVALIDATION("skip_cpu_inner_invalidation"),
     NCE_INVALIDATION_GPU_READBACK("nce_invalidation_gpu_readback"),
-    NCE_RUNTIME_NRO_PATCH("nce_runtime_nro_patch"),
     FIX_BLOOM_EFFECTS("fix_bloom_effects"),
     EMULATE_BGR565("emulate_bgr565"),
     RESCALE_HACK("rescale_hack"),
@@ -95,7 +94,7 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
 
     ENABLE_QUICK_SETTINGS("enable_quick_settings");
 
-//  external fun isFrameSkippingEnabled(): Boolean
+    //external fun isFrameSkippingEnabled(): Boolean
     external fun isFrameInterpolationEnabled(): Boolean
 
     override fun getBoolean(needsGlobal: Boolean): Boolean =

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

@@ -866,13 +866,6 @@ abstract class SettingsItem(
                     descriptionId = R.string.nce_invalidation_gpu_readback_description
                 )
             )
-            put(
-                SwitchSetting(
-                    BooleanSetting.NCE_RUNTIME_NRO_PATCH,
-                    titleId = R.string.nce_runtime_nro_patch,
-                    descriptionId = R.string.nce_runtime_nro_patch_description
-                )
-            )
             put(
                 SwitchSetting(
                     BooleanSetting.FIX_BLOOM_EFFECTS,

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

@@ -555,7 +555,6 @@ class SettingsFragmentPresenter(
 
             add(BooleanSetting.SKIP_CPU_INNER_INVALIDATION.key)
             add(BooleanSetting.NCE_INVALIDATION_GPU_READBACK.key)
-            add(BooleanSetting.NCE_RUNTIME_NRO_PATCH.key)
             add(BooleanSetting.FIX_BLOOM_EFFECTS.key)
             add(BooleanSetting.EMULATE_BGR565.key)
             add(BooleanSetting.RENDERER_ASYNCHRONOUS_SHADERS.key)

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

@@ -589,8 +589,6 @@
     <string name="skip_cpu_inner_invalidation_description">يتخطى بعض عمليات إبطال ذاكرة التخزين المؤقتة من جانب وحدة المعالجة المركزية أثناء تحديثات الذاكرة، مما يقلل من استخدام وحدة المعالجة المركزية ويحسن أداءها. قد يتسبب ذلك في حدوث أعطال أو تعطل في بعض الألعاب.</string>
     <string name="nce_invalidation_gpu_readback">إعادة قراءة وحدة معالجة الرسومات قبل إبطال صلاحية NCE</string>
     <string name="nce_invalidation_gpu_readback_description">إصلاح مؤقت. يعالج مشكلة تعطل لعبة Ender Magnolia (وغيرها من ألعاب UE). قد يؤدي إلى انخفاض في الأداء. يجري العمل على تحسينه.</string>
-    <string name="nce_runtime_nro_patch">تصحيح حالات NRO في وقت التشغيل لـ NCE</string>
-    <string name="nce_runtime_nro_patch_description">حل مؤقت. إصلاح مشكلة تشغيل لعبة Biomutant. قد يؤثر ذلك على الأداء. يجري العمل على تحسينه.</string>
     <string name="fix_bloom_effects">إصلاح تأثيرات التوهج</string>
     <string name="fix_bloom_effects_description">يقلل من ضبابية التوهج في LA/EOW  (Adreno A6XX - A7XX/ Turnip)، ويزيل التوهج في Burnout. تحذير: قد يسبب تشوهات رسومية في ألعاب أخرى.</string>
     <string name="emulate_bgr565">محاكاة BGR565</string>

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

@@ -579,8 +579,6 @@
     <string name="skip_cpu_inner_invalidation_description">在記憶體更新期間跳過某些 CPU 端快取的失效處理，減少 CPU 使用率並提高其性能。可能會導致某些遊戲出現故障或當機</string>
     <string name="nce_invalidation_gpu_readback">在 NCE 無效化前進行 GPU 回讀</string>
     <string name="nce_invalidation_gpu_readback_description">（暫時性功能）修復 \"終結者瑪格諾莉亞：霧中綻放\" 及其它使用虛幻引擎之遊戲中的當機問題，可能會降低效能。此功能正在改善中</string>
-    <string name="nce_runtime_nro_patch">對 NCE Runtime 的 NRO 檔案進行修補</string>
-    <string name="nce_runtime_nro_patch_description">（暫時性功能）修復 \"Biomutant\" 的啟動問題，可能會對效能造成影響。此功能正在改善中</string>
     <string name="fix_bloom_effects">修正光暈特效</string>
     <string name="fix_bloom_effects_description">在 \"薩爾達傳說：織夢島\" 和 \"薩爾達傳說：智慧的再現\" 中減少光暈模糊的問題並移除 \"橫衝直撞：狂飆樂園重製版\" 中的光暈效果（適用於 Adreno A6XX - A7XX/ Turnip）。警告：可能會在其它遊戲中導致圖形異常</string>
     <string name="emulate_bgr565">模擬 BGR565</string>
@@ -1197,4 +1195,4 @@
     <!-- Licenses screen strings -->
     <string name="licenses">授權</string>
     <string name="license_fidelityfx_fsr_description">來自 AMD 的升級圖像品質</string>
-    </resources>
+    </resources>

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

@@ -595,8 +595,6 @@
     <string name="skip_cpu_inner_invalidation_description">Skips certain CPU-side cache invalidations during memory updates, reducing CPU usage and improving it\'s performance. This may cause glitches or crashes on some games.</string>
     <string name="nce_invalidation_gpu_readback">GPU Readback Before NCE Invalidation</string>
     <string name="nce_invalidation_gpu_readback_description">Temporary. Fix Ender Magnolia (and other UE titles) crashes. May reduce performance. Improvement in progress.</string>
-    <string name="nce_runtime_nro_patch">Patch Runtime NROs for NCE</string>
-    <string name="nce_runtime_nro_patch_description">Temporary. Fix Biomutant boot. May impact performance. Improvement in progress.</string>
     <string name="fix_bloom_effects">Fix Bloom Effects</string>
     <string name="fix_bloom_effects_description">Reduces bloom blur in LA/EOW (Adreno A6XX - A7XX/ Turnip), removes bloom in Burnout. Warning: may cause graphical artifacts in other games.</string>
     <string name="emulate_bgr565">Emulate BGR565</string>

src/common/settings.h

@@ -641,8 +641,6 @@ struct Values {
                                                         true};
     SwitchableSetting<bool> nce_invalidation_gpu_readback{
         linkage, false, "nce_invalidation_gpu_readback", Category::RendererHacks};
-    SwitchableSetting<bool> nce_runtime_nro_patch{
-        linkage, false, "nce_runtime_nro_patch", Category::RendererHacks};
     SwitchableSetting<bool> async_presentation{linkage,
 #ifdef __ANDROID__
                                                true,

src/core/arm/nce/patcher.cpp

@@ -45,6 +45,23 @@ Patcher::Patcher() : c(m_patch_instructions), c_pre(m_patch_instructions_pre) {
     WriteLoadContext(c_pre);
 }
 
+Patcher::Patcher(std::span<const u8> text) : c(m_patch_instructions), c_pre(m_patch_instructions_pre) {
+    modules.emplace_back();
+    curr_patch = &modules.back();
+    const auto words = std::span<const u32>{reinterpret_cast<const u32*>(text.data()), text.size() / sizeof(u32)};
+    for (u32 i = ModuleCodeIndex; i < static_cast<u32>(words.size()); ++i) {
+        const uintptr_t offset = i * sizeof(u32);
+        if (auto mrs = MRS{words[i]}; mrs.Verify() && mrs.GetRt() != 31 && (mrs.GetSystemReg() == TpidrEl0 || mrs.GetSystemReg() == TpidrroEl0)) {
+            BranchToPatch(offset);
+            const auto src_reg = mrs.GetSystemReg() == TpidrroEl0 ? oaknut::SystemReg::TPIDRRO_EL0 : oaknut::SystemReg::TPIDR_EL0;
+            WriteMrsHandler(offset + sizeof(u32), oaknut::XReg{static_cast<int>(mrs.GetRt())}, src_reg);
+        } else if (auto msr = MSR{words[i]}; msr.Verify() && msr.GetSystemReg() == TpidrEl0) {
+            BranchToPatch(offset);
+            WriteMsrHandler(offset + sizeof(u32), oaknut::XReg{static_cast<int>(msr.GetRt())});
+        }
+    }
+}
+
 Patcher::~Patcher() = default;
 
 bool Patcher::PatchText(std::span<const u8> program_image, const Kernel::CodeSet::Segment& code) {
@@ -182,6 +199,31 @@ bool Patcher::PatchText(std::span<const u8> program_image, const Kernel::CodeSet
     return true;
 }
 
+std::span<const u32> Patcher::RelocateRuntimeTls(
+    s64 patch_offset, std::vector<InstructionPatch>& out_patches) {
+    const auto& patch = modules.front();
+    const auto can_branch = [](s64 offset) {
+        return offset % 4 == 0 && offset >= -static_cast<s64>(MaxRelativeBranch) && offset < static_cast<s64>(MaxRelativeBranch);
+    };
+    out_patches.clear();
+    out_patches.reserve(patch.m_branch_to_patch_relocations.size());
+    for (const auto& rel : patch.m_branch_to_patch_relocations) {
+        const s64 offset = patch_offset + rel.patch_offset - static_cast<s64>(rel.module_offset);
+        if (!can_branch(offset)) {return {};}
+        u32 replacement;
+        oaknut::CodeGenerator rc{&replacement};
+        rc.B(offset);
+        out_patches.push_back({rel.module_offset, replacement});
+    }
+    for (const auto& rel : patch.m_branch_to_module_relocations) {
+        const s64 offset = static_cast<s64>(rel.module_offset) - patch_offset - rel.patch_offset;
+        if (!can_branch(offset)) {return {};}
+        oaknut::CodeGenerator rc{m_patch_instructions.data() + rel.patch_offset / sizeof(u32)};
+        rc.B(offset);
+    }
+    return m_patch_instructions;
+}
+
 bool Patcher::RelocateAndCopy(Common::ProcessAddress load_base, const Kernel::CodeSet::Segment& code, std::vector<u8>& program_image, EntryTrampolines* out_trampolines) {
     const size_t patch_size = GetSectionSize();
     const size_t pre_patch_size = GetPreSectionSize();

src/core/arm/nce/patcher.h

@@ -48,15 +48,22 @@ using ModuleTextAddress = u64;
 using PatchTextAddress = u64;
 using EntryTrampolines = ::Common::unordered_map<ModuleTextAddress, PatchTextAddress>;
 
+struct InstructionPatch {
+    u64 offset;
+    u32 replacement;
+};
+
 class Patcher {
 public:
     void SetModuleID(const ModuleID& id) {
         module_id = id;
     }
     explicit Patcher();
+    explicit Patcher(std::span<const u8> text);
     ~Patcher();
     bool PatchText(std::span<const u8> program_image, const Kernel::CodeSet::Segment& code);
     bool RelocateAndCopy(Common::ProcessAddress load_base, const Kernel::CodeSet::Segment& code, std::vector<u8>& program_image, EntryTrampolines* out_trampolines);
+    std::span<const u32> RelocateRuntimeTls(s64 patch_offset, std::vector<InstructionPatch>& out_patches);
     size_t GetSectionSize() const noexcept;
     size_t GetPreSectionSize() const noexcept;
 

src/core/hle/service/ro/ro.cpp

@@ -7,6 +7,7 @@
 #include <openssl/err.h>
 #include <openssl/evp.h>
 
+#include "common/scope_exit.h"
 #include "core/hle/kernel/k_process.h"
 
 #include "core/hle/service/cmif_serialization.h"
@@ -425,36 +426,51 @@ public:
         R_TRY(context->ValidateNro(std::addressof(nro_info->module_id), std::addressof(rx_size),
                                    std::addressof(ro_size), std::addressof(rw_size),
                                    nro_info->base_address, nro_size, bss_size));
+        R_TRY(SetNroPerms(context->GetProcess(), nro_info->base_address, rx_size, ro_size, rw_size + bss_size));
 #ifdef HAS_NCE
-        if (Settings::values.nce_runtime_nro_patch.GetValue()) {
-            if (Settings::IsNceEnabled()) {
-                auto* process = context->GetProcess();
-                auto& memory = process->GetMemory();
-
-                std::vector<u8> image(total_size);
-                memory.ReadBlock(nro_info->base_address, image.data(), rx_size);
-
-                Kernel::CodeSet::Segment code{.size = static_cast<u32>(rx_size)};
-                Core::NCE::Patcher patch;
-                patch.PatchText(image, code);
-                patch.RelocateAndCopy(nro_info->base_address, code, image, nullptr);
-
-                const u64 patch_address = nro_info->base_address + total_size;
-                const size_t patch_size = patch.GetSectionSize();
+        if (Settings::IsNceEnabled()) {
+            auto* process = context->GetProcess();
+            auto& memory = process->GetMemory();
+            std::vector<u8> text(rx_size);
+            memory.ReadBlock(nro_info->base_address, text.data(), text.size());
+            Core::NCE::Patcher patch{text};
+            const size_t patch_size = patch.GetSectionSize();
+            if (patch_size != 0) {
                 constexpr auto permission = Kernel::Svc::MemoryPermission::ReadExecute;
-
                 auto* patch_memory = Kernel::KSharedMemory::Create(kernel);
+                SCOPE_EXIT { patch_memory->Close(kernel); };
                 R_TRY(patch_memory->Initialize(kernel, kernel.System().DeviceMemory(), process, permission, permission, patch_size));
-                std::memcpy(patch_memory->GetPointer(), image.data() + total_size, patch_size);
-                R_TRY(process->AddSharedMemory(kernel, patch_memory, patch_address, patch_size));
-                R_TRY(patch_memory->Map(*process, patch_address, patch_size, permission));
-                memory.WriteBlock(nro_info->base_address, image.data(), rx_size);
+                Kernel::KSharedMemory::Register(kernel, patch_memory);
+                const std::array<s64, 2> offsets{static_cast<s64>(total_size + Kernel::PageSize),-static_cast<s64>(patch_size + Kernel::PageSize)};
+                std::vector<Core::NCE::InstructionPatch> tls_patches;
+                std::span<const u32> instructions;
+                u64 patch_address = 0;
+                for (const s64 offset : offsets) {
+                    instructions = patch.RelocateRuntimeTls(offset, tls_patches);
+                    if (instructions.empty()) {continue;}
+                    const u64 address = static_cast<u64>(static_cast<s64>(nro_info->base_address) + offset);
+                    if (R_FAILED(patch_memory->Map(*process, address, patch_size, permission))) {continue;}
+                    patch_address = address;
+                    break;
+                }
+                R_UNLESS(patch_address != 0, RO::ResultOutOfAddressSpace);
+                const Result add_shared_memory_result = process->AddSharedMemory(kernel, patch_memory, patch_address, patch_size);
+                if (R_FAILED(add_shared_memory_result)) {
+                    R_ASSERT(patch_memory->Unmap(*process, patch_address, patch_size));
+                }
+                R_TRY(add_shared_memory_result);
+                std::memcpy(patch_memory->GetPointer(), instructions.data(), instructions.size_bytes());
+                auto* ptr = reinterpret_cast<char*>(patch_address);
+                __builtin___clear_cache(ptr, ptr + instructions.size_bytes());
+                for (const auto& instruction : tls_patches) {
+                    const u64 address = nro_info->base_address + instruction.offset;
+                    memory.Write32(address, instruction.replacement);
+                    ptr = reinterpret_cast<char*>(address);
+                    __builtin___clear_cache(ptr, ptr + sizeof(u32));
+                }
             }
         }
 #endif
-        // Set NRO perms.
-        R_TRY(SetNroPerms(context->GetProcess(), nro_info->base_address, rx_size, ro_size,
-                          rw_size + bss_size));
 
         context->SetNroInfoInUse(nro_info, true);
         nro_info->code_size = rx_size + ro_size;

src/qt_common/config/shared_translation.cpp

@@ -198,10 +198,6 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QObject* parent) {
            tr("GPU Readback Before NCE Invalidation"),
            tr("Flushes GPU buffer data back to guest memory before NCE invalidates the whole "
               "guest page, preserving it while the CPU store proceeds."));
-    INSERT(Settings, nce_runtime_nro_patch,
-           tr("Patch Runtime NROs for NCE"),
-           tr("Applies NCE code patches to runtime-loaded NRO modules."
-              "This is required for some games to run correctly."));
     INSERT(Settings, vsync_mode, tr("VSync Mode:"),
            tr("FIFO (VSync) does not drop frames or exhibit tearing but is limited by the screen "
               "refresh rate.\nFIFO Relaxed allows tearing as it recovers from a slow down.\n"

src/video_core/buffer_cache/buffer_cache.h

@@ -1738,64 +1738,36 @@ void BufferCache<P>::TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept {
 
 template <class P>
 bool BufferCache<P>::SynchronizeBuffer(Buffer& buffer, DAddr device_addr, u32 size) {
-#ifdef __ANDROID__
-    const bool use_optimized_vertex_buffers = Settings::values.use_optimized_vertex_buffers.GetValue();
-#else
-    constexpr bool use_optimized_vertex_buffers = true;
-#endif
-
-    if (use_optimized_vertex_buffers) {
-        upload_copies.clear();
-        u64 total_size_bytes = 0;
-        u64 largest_copy = 0;
-        const DAddr buffer_start = buffer.cpu_addr_cached;
-        const auto add_upload = [&](DAddr start, DAddr end) {
-            if (start == end) return;
-            const u64 range_size = end - start;
-            upload_copies.push_back(BufferCopy{
-                .src_offset = total_size_bytes,
-                .dst_offset = start - buffer_start,
-                .size = range_size,
-            });
-            total_size_bytes += range_size;
-            largest_copy = (std::max)(largest_copy, range_size);
-        };
-        memory_tracker.ForEachUploadRange(device_addr, size, [&](u64 device_addr_out, u64 range_size) {
-            DAddr upload_start = device_addr_out;
-            gpu_modified_ranges.ForEachInRange(device_addr_out, range_size, [&](DAddr gpu_start, DAddr gpu_end) {
-                add_upload(upload_start, gpu_start);
-                upload_start = gpu_end;
-            });
-            add_upload(upload_start, device_addr_out + range_size);
+    upload_copies.clear();
+    u64 total_size_bytes = 0;
+    u64 largest_copy = 0;
+    const DAddr buffer_start = buffer.cpu_addr_cached;
+    const auto add_upload = [&](DAddr start, DAddr end) {
+        if (start == end) return;
+        const u64 range_size = end - start;
+        upload_copies.push_back(BufferCopy{
+            .src_offset = total_size_bytes,
+            .dst_offset = start - buffer_start,
+            .size = range_size,
         });
-        if (total_size_bytes == 0) {
-            return true;
-        }
-        const std::span<BufferCopy> copies_span(upload_copies.data(), upload_copies.size());
-        UploadMemory(buffer, total_size_bytes, largest_copy, copies_span);
-        any_buffer_uploaded = true;
-        return false;
-    } else {
-        boost::container::small_vector<BufferCopy, 4> copies;
-        u64 total_size_bytes = 0;
-        u64 largest_copy = 0;
-        DAddr buffer_start = buffer.CpuAddr();
-        memory_tracker.ForEachUploadRange(device_addr, size, [&](u64 device_addr_out, u64 range_size) {
-            copies.push_back(BufferCopy{
-                .src_offset = total_size_bytes,
-                .dst_offset = device_addr_out - buffer_start,
-                .size = range_size,
-            });
-            total_size_bytes += range_size;
-            largest_copy = (std::max)(largest_copy, range_size);
+        total_size_bytes += range_size;
+        largest_copy = (std::max)(largest_copy, range_size);
+    };
+    memory_tracker.ForEachUploadRange(device_addr, size, [&](u64 device_addr_out, u64 range_size) {
+        DAddr upload_start = device_addr_out;
+        gpu_modified_ranges.ForEachInRange(device_addr_out, range_size, [&](DAddr gpu_start, DAddr gpu_end) {
+            add_upload(upload_start, gpu_start);
+            upload_start = gpu_end;
         });
-        if (total_size_bytes == 0) {
-            return true;
-        }
-        const std::span<BufferCopy> copies_span(copies.data(), copies.size());
-        UploadMemory(buffer, total_size_bytes, largest_copy, copies_span);
-        return false;
+        add_upload(upload_start, device_addr_out + range_size);
+    });
+    if (total_size_bytes == 0) {
+        return true;
     }
+    const std::span<BufferCopy> copies_span(upload_copies.data(), upload_copies.size());
+    UploadMemory(buffer, total_size_bytes, largest_copy, copies_span);
+    any_buffer_uploaded = true;
+    return false;
 }
 
 template <class P>