common_func: Use std::array for INSERT_PADDING_* macros.

- Zero initialization here is useful for determinism.

.gitmodules

@@ -26,11 +26,11 @@
     path = externals/mbedtls
     url = https://github.com/DarkLordZach/mbedtls
 [submodule "opus"]
-    path = externals/opus
-    url = https://github.com/ogniK5377/opus.git
+    path = externals/opus/opus
+    url = https://github.com/xiph/opus.git
 [submodule "soundtouch"]
-	path = externals/soundtouch
-	url = https://github.com/citra-emu/ext-soundtouch.git
+    path = externals/soundtouch
+    url = https://github.com/citra-emu/ext-soundtouch.git
 [submodule "libressl"]
     path = externals/libressl
     url = https://github.com/citra-emu/ext-libressl-portable.git

CMakeModules/GenerateSCMRev.cmake

@@ -85,10 +85,12 @@ set(HASH_FILES
     "${VIDEO_CORE}/shader/decode/xmad.cpp"
     "${VIDEO_CORE}/shader/ast.cpp"
     "${VIDEO_CORE}/shader/ast.h"
-    "${VIDEO_CORE}/shader/control_flow.cpp"
-    "${VIDEO_CORE}/shader/control_flow.h"
     "${VIDEO_CORE}/shader/compiler_settings.cpp"
     "${VIDEO_CORE}/shader/compiler_settings.h"
+    "${VIDEO_CORE}/shader/const_buffer_locker.cpp"
+    "${VIDEO_CORE}/shader/const_buffer_locker.h"
+    "${VIDEO_CORE}/shader/control_flow.cpp"
+    "${VIDEO_CORE}/shader/control_flow.h"
     "${VIDEO_CORE}/shader/decode.cpp"
     "${VIDEO_CORE}/shader/expr.cpp"
     "${VIDEO_CORE}/shader/expr.h"

CONTRIBUTING.md

@@ -1 +1 @@
-**The Contributor's Guide has moved to [the Citra wiki](https://github.com/citra-emu/citra/wiki/Contributing).**
+**The Contributor's Guide has moved to [the Yuzu wiki](https://github.com/yuzu-emu/yuzu/wiki/Contributing).**

externals/CMakeLists.txt

@@ -42,9 +42,6 @@ target_include_directories(mbedtls PUBLIC ./mbedtls/include)
 add_library(microprofile INTERFACE)
 target_include_directories(microprofile INTERFACE ./microprofile)
 
-# Open Source Archives
-add_subdirectory(open_source_archives EXCLUDE_FROM_ALL)
-
 # Unicorn
 add_library(unicorn-headers INTERFACE)
 target_include_directories(unicorn-headers INTERFACE ./unicorn/include)

externals/open_source_archives/CMakeLists.txt

@@ -1,16 +0,0 @@
-add_library(open_source_archives
-    src/FontChineseSimplified.cpp
-    src/FontChineseTraditional.cpp
-    src/FontExtendedChineseSimplified.cpp
-    src/FontKorean.cpp
-    src/FontNintendoExtended.cpp
-    src/FontStandard.cpp
-    include/FontChineseSimplified.h
-    include/FontChineseTraditional.h
-    include/FontExtendedChineseSimplified.h
-    include/FontKorean.h
-    include/FontNintendoExtended.h
-    include/FontStandard.h
-)
-
-target_include_directories(open_source_archives PUBLIC include)

externals/open_source_archives/Readme.md

@@ -1,4 +0,0 @@
-These files were generated by https://github.com/FearlessTobi/yuzu_system_archives at git commit 0a24b0c9f38d71fb2c4bba5645a39029e539a5ec. To generate the files use the run.sh inside that repository.  
-
-The follwing system archives are currently included:
- - JPN/EUR/USA System Font

externals/open_source_archives/include/FontChineseSimplified.h

@@ -1,6 +0,0 @@
-#pragma once
-
-#include <array>
-
-extern const std::array<unsigned char, 217276> FontChineseSimplified;
-

externals/open_source_archives/include/FontChineseTraditional.h

@@ -1,6 +0,0 @@
-#pragma once
-
-#include <array>
-
-extern const std::array<unsigned char, 222236> FontChineseTraditional;
-

externals/open_source_archives/include/FontExtendedChineseSimplified.h

@@ -1,6 +0,0 @@
-#pragma once
-
-#include <array>
-
-extern const std::array<unsigned char, 293516> FontExtendedChineseSimplified;
-

externals/open_source_archives/include/FontKorean.h

@@ -1,6 +0,0 @@
-#pragma once
-
-#include <array>
-
-extern const std::array<unsigned char, 217276> FontKorean;
-

externals/open_source_archives/include/FontNintendoExtended.h

@@ -1,6 +0,0 @@
-#pragma once
-
-#include <array>
-
-extern const std::array<unsigned char, 172064> FontNintendoExtended;
-

externals/open_source_archives/include/FontStandard.h

@@ -1,6 +0,0 @@
-#pragma once
-
-#include <array>
-
-extern const std::array<unsigned char, 217276> FontStandard;
-

externals/opus/CMakeLists.txt

@@ -0,0 +1,250 @@
+cmake_minimum_required(VERSION 3.8)
+
+project(opus)
+
+option(OPUS_STACK_PROTECTOR "Use stack protection" OFF)
+option(OPUS_USE_ALLOCA "Use alloca for stack arrays (on non-C99 compilers)" OFF)
+option(OPUS_CUSTOM_MODES "Enable non-Opus modes, e.g. 44.1 kHz & 2^n frames" OFF)
+option(OPUS_FIXED_POINT "Compile as fixed-point (for machines without a fast enough FPU)" OFF)
+option(OPUS_ENABLE_FLOAT_API "Compile with the floating point API (for machines with float library" ON)
+
+include(opus/opus_functions.cmake)
+
+if(OPUS_STACK_PROTECTOR)
+    if(NOT MSVC) # GC on by default on MSVC
+        check_and_set_flag(STACK_PROTECTION_STRONG -fstack-protector-strong)
+    endif()
+else()
+    if(MSVC)
+        check_and_set_flag(BUFFER_SECURITY_CHECK /GS-)
+    endif()
+endif()
+
+add_library(opus STATIC
+    # CELT sources
+    opus/celt/bands.c
+    opus/celt/celt.c
+    opus/celt/celt_decoder.c
+    opus/celt/celt_encoder.c
+    opus/celt/celt_lpc.c
+    opus/celt/cwrs.c
+    opus/celt/entcode.c
+    opus/celt/entdec.c
+    opus/celt/entenc.c
+    opus/celt/kiss_fft.c
+    opus/celt/laplace.c
+    opus/celt/mathops.c
+    opus/celt/mdct.c
+    opus/celt/modes.c
+    opus/celt/pitch.c
+    opus/celt/quant_bands.c
+    opus/celt/rate.c
+    opus/celt/vq.c
+
+    # SILK sources
+    opus/silk/A2NLSF.c
+    opus/silk/CNG.c
+    opus/silk/HP_variable_cutoff.c
+    opus/silk/LPC_analysis_filter.c
+    opus/silk/LPC_fit.c
+    opus/silk/LPC_inv_pred_gain.c
+    opus/silk/LP_variable_cutoff.c
+    opus/silk/NLSF2A.c
+    opus/silk/NLSF_VQ.c
+    opus/silk/NLSF_VQ_weights_laroia.c
+    opus/silk/NLSF_decode.c
+    opus/silk/NLSF_del_dec_quant.c
+    opus/silk/NLSF_encode.c
+    opus/silk/NLSF_stabilize.c
+    opus/silk/NLSF_unpack.c
+    opus/silk/NSQ.c
+    opus/silk/NSQ_del_dec.c
+    opus/silk/PLC.c
+    opus/silk/VAD.c
+    opus/silk/VQ_WMat_EC.c
+    opus/silk/ana_filt_bank_1.c
+    opus/silk/biquad_alt.c
+    opus/silk/bwexpander.c
+    opus/silk/bwexpander_32.c
+    opus/silk/check_control_input.c
+    opus/silk/code_signs.c
+    opus/silk/control_SNR.c
+    opus/silk/control_audio_bandwidth.c
+    opus/silk/control_codec.c
+    opus/silk/dec_API.c
+    opus/silk/decode_core.c
+    opus/silk/decode_frame.c
+    opus/silk/decode_indices.c
+    opus/silk/decode_parameters.c
+    opus/silk/decode_pitch.c
+    opus/silk/decode_pulses.c
+    opus/silk/decoder_set_fs.c
+    opus/silk/enc_API.c
+    opus/silk/encode_indices.c
+    opus/silk/encode_pulses.c
+    opus/silk/gain_quant.c
+    opus/silk/init_decoder.c
+    opus/silk/init_encoder.c
+    opus/silk/inner_prod_aligned.c
+    opus/silk/interpolate.c
+    opus/silk/lin2log.c
+    opus/silk/log2lin.c
+    opus/silk/pitch_est_tables.c
+    opus/silk/process_NLSFs.c
+    opus/silk/quant_LTP_gains.c
+    opus/silk/resampler.c
+    opus/silk/resampler_down2.c
+    opus/silk/resampler_down2_3.c
+    opus/silk/resampler_private_AR2.c
+    opus/silk/resampler_private_IIR_FIR.c
+    opus/silk/resampler_private_down_FIR.c
+    opus/silk/resampler_private_up2_HQ.c
+    opus/silk/resampler_rom.c
+    opus/silk/shell_coder.c
+    opus/silk/sigm_Q15.c
+    opus/silk/sort.c
+    opus/silk/stereo_LR_to_MS.c
+    opus/silk/stereo_MS_to_LR.c
+    opus/silk/stereo_decode_pred.c
+    opus/silk/stereo_encode_pred.c
+    opus/silk/stereo_find_predictor.c
+    opus/silk/stereo_quant_pred.c
+    opus/silk/sum_sqr_shift.c
+    opus/silk/table_LSF_cos.c
+    opus/silk/tables_LTP.c
+    opus/silk/tables_NLSF_CB_NB_MB.c
+    opus/silk/tables_NLSF_CB_WB.c
+    opus/silk/tables_gain.c
+    opus/silk/tables_other.c
+    opus/silk/tables_pitch_lag.c
+    opus/silk/tables_pulses_per_block.c
+
+    # Opus sources
+    opus/src/analysis.c
+    opus/src/mapping_matrix.c
+    opus/src/mlp.c
+    opus/src/mlp_data.c
+    opus/src/opus.c
+    opus/src/opus_decoder.c
+    opus/src/opus_encoder.c
+    opus/src/opus_multistream.c
+    opus/src/opus_multistream_decoder.c
+    opus/src/opus_multistream_encoder.c
+    opus/src/opus_projection_decoder.c
+    opus/src/opus_projection_encoder.c
+    opus/src/repacketizer.c
+)
+
+if (DEBUG)
+    target_sources(opus PRIVATE opus/silk/debug.c)
+endif()
+
+if (OPUS_FIXED_POINT)
+    target_sources(opus PRIVATE
+        opus/silk/fixed/LTP_analysis_filter_FIX.c
+        opus/silk/fixed/LTP_scale_ctrl_FIX.c
+        opus/silk/fixed/apply_sine_window_FIX.c
+        opus/silk/fixed/autocorr_FIX.c
+        opus/silk/fixed/burg_modified_FIX.c
+        opus/silk/fixed/corrMatrix_FIX.c
+        opus/silk/fixed/encode_frame_FIX.c
+        opus/silk/fixed/find_LPC_FIX.c
+        opus/silk/fixed/find_LTP_FIX.c
+        opus/silk/fixed/find_pitch_lags_FIX.c
+        opus/silk/fixed/find_pred_coefs_FIX.c
+        opus/silk/fixed/k2a_FIX.c
+        opus/silk/fixed/k2a_Q16_FIX.c
+        opus/silk/fixed/noise_shape_analysis_FIX.c
+        opus/silk/fixed/pitch_analysis_core_FIX.c
+        opus/silk/fixed/prefilter_FIX.c
+        opus/silk/fixed/process_gains_FIX.c
+        opus/silk/fixed/regularize_correlations_FIX.c
+        opus/silk/fixed/residual_energy16_FIX.c
+        opus/silk/fixed/residual_energy_FIX.c
+        opus/silk/fixed/schur64_FIX.c
+        opus/silk/fixed/schur_FIX.c
+        opus/silk/fixed/solve_LS_FIX.c
+        opus/silk/fixed/vector_ops_FIX.c
+        opus/silk/fixed/warped_autocorrelation_FIX.c
+    )
+else()
+    target_sources(opus PRIVATE
+        opus/silk/float/LPC_analysis_filter_FLP.c
+        opus/silk/float/LPC_inv_pred_gain_FLP.c
+        opus/silk/float/LTP_analysis_filter_FLP.c
+        opus/silk/float/LTP_scale_ctrl_FLP.c
+        opus/silk/float/apply_sine_window_FLP.c
+        opus/silk/float/autocorrelation_FLP.c
+        opus/silk/float/burg_modified_FLP.c
+        opus/silk/float/bwexpander_FLP.c
+        opus/silk/float/corrMatrix_FLP.c
+        opus/silk/float/encode_frame_FLP.c
+        opus/silk/float/energy_FLP.c
+        opus/silk/float/find_LPC_FLP.c
+        opus/silk/float/find_LTP_FLP.c
+        opus/silk/float/find_pitch_lags_FLP.c
+        opus/silk/float/find_pred_coefs_FLP.c
+        opus/silk/float/inner_product_FLP.c
+        opus/silk/float/k2a_FLP.c
+        opus/silk/float/noise_shape_analysis_FLP.c
+        opus/silk/float/pitch_analysis_core_FLP.c
+        opus/silk/float/process_gains_FLP.c
+        opus/silk/float/regularize_correlations_FLP.c
+        opus/silk/float/residual_energy_FLP.c
+        opus/silk/float/scale_copy_vector_FLP.c
+        opus/silk/float/scale_vector_FLP.c
+        opus/silk/float/schur_FLP.c
+        opus/silk/float/sort_FLP.c
+        opus/silk/float/warped_autocorrelation_FLP.c
+        opus/silk/float/wrappers_FLP.c
+    )
+endif()
+
+target_compile_definitions(opus PRIVATE OPUS_BUILD ENABLE_HARDENING)
+
+if(NOT MSVC)
+    target_compile_definitions(opus PRIVATE _FORTIFY_SOURCE=2)
+endif()
+
+# It is strongly recommended to uncomment one of these VAR_ARRAYS: Use C99
+# variable-length arrays for stack allocation USE_ALLOCA: Use alloca() for stack
+# allocation If none is defined, then the fallback is a non-threadsafe global
+# array
+if(OPUS_USE_ALLOCA OR MSVC)
+    target_compile_definitions(opus PRIVATE USE_ALLOCA)
+else()
+    target_compile_definitions(opus PRIVATE VAR_ARRAYS)
+endif()
+
+if(OPUS_CUSTOM_MODES)
+    target_compile_definitions(opus PRIVATE CUSTOM_MODES)
+endif()
+
+if(NOT OPUS_ENABLE_FLOAT_API)
+    target_compile_definitions(opus PRIVATE DISABLE_FLOAT_API)
+endif()
+
+target_compile_definitions(opus
+PUBLIC
+    -DOPUS_VERSION="\\"1.3.1\\""
+
+PRIVATE
+    # Use C99 intrinsics to speed up float-to-int conversion
+    HAVE_LRINTF
+)
+
+if (FIXED_POINT)
+    target_compile_definitions(opus PRIVATE -DFIXED_POINT=1 -DDISABLE_FLOAT_API)
+endif()
+
+target_include_directories(opus
+PUBLIC
+    opus/include
+
+PRIVATE
+    opus/celt
+    opus/silk
+    opus/silk/fixed
+    opus/silk/float
+    opus/src
+)

src/common/CMakeLists.txt

@@ -74,10 +74,12 @@ add_custom_command(OUTPUT scm_rev.cpp
       "${VIDEO_CORE}/shader/decode/xmad.cpp"
       "${VIDEO_CORE}/shader/ast.cpp"
       "${VIDEO_CORE}/shader/ast.h"
-      "${VIDEO_CORE}/shader/control_flow.cpp"
-      "${VIDEO_CORE}/shader/control_flow.h"
       "${VIDEO_CORE}/shader/compiler_settings.cpp"
       "${VIDEO_CORE}/shader/compiler_settings.h"
+      "${VIDEO_CORE}/shader/const_buffer_locker.cpp"
+      "${VIDEO_CORE}/shader/const_buffer_locker.h"
+      "${VIDEO_CORE}/shader/control_flow.cpp"
+      "${VIDEO_CORE}/shader/control_flow.h"
       "${VIDEO_CORE}/shader/decode.cpp"
       "${VIDEO_CORE}/shader/expr.cpp"
       "${VIDEO_CORE}/shader/expr.h"
@@ -95,11 +97,11 @@ add_custom_command(OUTPUT scm_rev.cpp
 )
 
 add_library(common STATIC
+    algorithm.h
     alignment.h
     assert.h
     detached_tasks.cpp
     detached_tasks.h
-    binary_find.h
     bit_field.h
     bit_util.h
     cityhash.cpp

src/common/algorithm.h

@@ -5,6 +5,12 @@
 #pragma once
 
 #include <algorithm>
+#include <functional>
+
+// Algorithms that operate on iterators, much like the <algorithm> header.
+//
+// Note: If the algorithm is not general-purpose and/or doesn't operate on iterators,
+//       it should probably not be placed within this header.
 
 namespace Common {
 

src/common/common_funcs.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <algorithm>
+#include <array>
 #include <string>
 
 #if !defined(ARCHITECTURE_x86_64)
@@ -19,8 +20,8 @@
 // helper macro to properly align structure members.
 // Calling INSERT_PADDING_BYTES will add a new member variable with a name like "pad121",
 // depending on the current source line to make sure variable names are unique.
-#define INSERT_PADDING_BYTES(num_bytes) u8 CONCAT2(pad, __LINE__)[(num_bytes)]
-#define INSERT_PADDING_WORDS(num_words) u32 CONCAT2(pad, __LINE__)[(num_words)]
+#define INSERT_PADDING_BYTES(num_bytes) std::array<u8, num_bytes> CONCAT2(pad, __LINE__)
+#define INSERT_PADDING_WORDS(num_words) std::array<u32, num_words> CONCAT2(pad, __LINE__)
 
 // Inlining
 #ifdef _WIN32

src/common/hash.h

@@ -6,6 +6,8 @@
 
 #include <cstddef>
 #include <cstring>
+#include <utility>
+#include <boost/functional/hash.hpp>
 #include "common/cityhash.h"
 #include "common/common_types.h"
 
@@ -68,4 +70,13 @@ struct HashableStruct {
     }
 };
 
+struct PairHash {
+    template <class T1, class T2>
+    std::size_t operator()(const std::pair<T1, T2>& pair) const noexcept {
+        std::size_t seed = std::hash<T1>()(pair.first);
+        boost::hash_combine(seed, std::hash<T2>()(pair.second));
+        return seed;
+    }
+};
+
 } // namespace Common

src/common/multi_level_queue.h

@@ -304,6 +304,13 @@ public:
         return levels[priority == Depth ? 63 : priority].back();
     }
 
+    void clear() {
+        used_priorities = 0;
+        for (std::size_t i = 0; i < Depth; i++) {
+            levels[i].clear();
+        }
+    }
+
 private:
     using const_list_iterator = typename std::list<T>::const_iterator;
 

src/core/CMakeLists.txt

@@ -74,10 +74,24 @@ add_library(core STATIC
     file_sys/sdmc_factory.h
     file_sys/submission_package.cpp
     file_sys/submission_package.h
+    file_sys/system_archive/data/font_chinese_simplified.cpp
+    file_sys/system_archive/data/font_chinese_simplified.h
+    file_sys/system_archive/data/font_chinese_traditional.cpp
+    file_sys/system_archive/data/font_chinese_traditional.h
+    file_sys/system_archive/data/font_extended_chinese_simplified.cpp
+    file_sys/system_archive/data/font_extended_chinese_simplified.h
+    file_sys/system_archive/data/font_korean.cpp
+    file_sys/system_archive/data/font_korean.h
+    file_sys/system_archive/data/font_nintendo_extended.cpp
+    file_sys/system_archive/data/font_nintendo_extended.h
+    file_sys/system_archive/data/font_standard.cpp
+    file_sys/system_archive/data/font_standard.h
     file_sys/system_archive/mii_model.cpp
     file_sys/system_archive/mii_model.h
     file_sys/system_archive/ng_word.cpp
     file_sys/system_archive/ng_word.h
+    file_sys/system_archive/shared_font.cpp
+    file_sys/system_archive/shared_font.h
     file_sys/system_archive/system_archive.cpp
     file_sys/system_archive/system_archive.h
     file_sys/system_archive/system_version.cpp
@@ -511,7 +525,7 @@ add_library(core STATIC
 create_target_directory_groups(core)
 
 target_link_libraries(core PUBLIC common PRIVATE audio_core video_core)
-target_link_libraries(core PUBLIC Boost::boost PRIVATE fmt json-headers mbedtls opus unicorn open_source_archives)
+target_link_libraries(core PUBLIC Boost::boost PRIVATE fmt json-headers mbedtls opus unicorn)
 
 if (YUZU_ENABLE_BOXCAT)
     get_directory_property(OPENSSL_LIBS

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -116,7 +116,7 @@ public:
         num_interpreted_instructions = 0;
     }
     u64 GetTicksRemaining() override {
-        return std::max(parent.system.CoreTiming().GetDowncount(), 0);
+        return std::max(parent.system.CoreTiming().GetDowncount(), s64{0});
     }
     u64 GetCNTPCT() override {
         return Timing::CpuCyclesToClockCycles(parent.system.CoreTiming().GetTicks());

src/core/arm/unicorn/arm_unicorn.cpp

@@ -156,7 +156,7 @@ void ARM_Unicorn::Run() {
     if (GDBStub::IsServerEnabled()) {
         ExecuteInstructions(std::max(4000000, 0));
     } else {
-        ExecuteInstructions(std::max(system.CoreTiming().GetDowncount(), 0));
+        ExecuteInstructions(std::max(system.CoreTiming().GetDowncount(), s64{0}));
     }
 }
 

src/core/core.cpp

@@ -112,8 +112,8 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
 }
 struct System::Impl {
     explicit Impl(System& system)
-        : kernel{system}, fs_controller{system}, cpu_core_manager{system},
-          applet_manager{system}, reporter{system} {}
+        : kernel{system}, fs_controller{system}, cpu_core_manager{system}, reporter{system},
+          applet_manager{system} {}
 
     Cpu& CurrentCpuCore() {
         return cpu_core_manager.GetCurrentCore();
@@ -240,22 +240,27 @@ struct System::Impl {
     }
 
     void Shutdown() {
-        // Log last frame performance stats
-        const auto perf_results = GetAndResetPerfStats();
-        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_EmulationSpeed",
-                                    perf_results.emulation_speed * 100.0);
-        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Framerate",
-                                    perf_results.game_fps);
-        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Frametime",
-                                    perf_results.frametime * 1000.0);
-        telemetry_session->AddField(Telemetry::FieldType::Performance, "Mean_Frametime_MS",
-                                    perf_stats->GetMeanFrametime());
+        // Log last frame performance stats if game was loded
+        if (perf_stats) {
+            const auto perf_results = GetAndResetPerfStats();
+            telemetry_session->AddField(Telemetry::FieldType::Performance,
+                                        "Shutdown_EmulationSpeed",
+                                        perf_results.emulation_speed * 100.0);
+            telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Framerate",
+                                        perf_results.game_fps);
+            telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Frametime",
+                                        perf_results.frametime * 1000.0);
+            telemetry_session->AddField(Telemetry::FieldType::Performance, "Mean_Frametime_MS",
+                                        perf_stats->GetMeanFrametime());
+        }
 
         lm_manager.Flush();
 
         is_powered_on = false;
         exit_lock = false;
 
+        gpu_core->WaitIdle();
+
         // Shutdown emulation session
         renderer.reset();
         GDBStub::Shutdown();
@@ -404,6 +409,12 @@ void System::PrepareReschedule() {
     CurrentCpuCore().PrepareReschedule();
 }
 
+void System::PrepareReschedule(const u32 core_index) {
+    if (core_index < GlobalScheduler().CpuCoresCount()) {
+        CpuCore(core_index).PrepareReschedule();
+    }
+}
+
 PerfStatsResults System::GetAndResetPerfStats() {
     return impl->GetAndResetPerfStats();
 }
@@ -444,6 +455,16 @@ const Kernel::Scheduler& System::Scheduler(std::size_t core_index) const {
     return CpuCore(core_index).Scheduler();
 }
 
+/// Gets the global scheduler
+Kernel::GlobalScheduler& System::GlobalScheduler() {
+    return impl->kernel.GlobalScheduler();
+}
+
+/// Gets the global scheduler
+const Kernel::GlobalScheduler& System::GlobalScheduler() const {
+    return impl->kernel.GlobalScheduler();
+}
+
 Kernel::Process* System::CurrentProcess() {
     return impl->kernel.CurrentProcess();
 }

src/core/core.h

@@ -24,6 +24,7 @@ class VfsFilesystem;
 } // namespace FileSys
 
 namespace Kernel {
+class GlobalScheduler;
 class KernelCore;
 class Process;
 class Scheduler;
@@ -184,6 +185,9 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule();
 
+    /// Prepare the core emulation for a reschedule
+    void PrepareReschedule(u32 core_index);
+
     /// Gets and resets core performance statistics
     PerfStatsResults GetAndResetPerfStats();
 
@@ -238,6 +242,12 @@ public:
     /// Gets the scheduler for the CPU core with the specified index
     const Kernel::Scheduler& Scheduler(std::size_t core_index) const;
 
+    /// Gets the global scheduler
+    Kernel::GlobalScheduler& GlobalScheduler();
+
+    /// Gets the global scheduler
+    const Kernel::GlobalScheduler& GlobalScheduler() const;
+
     /// Provides a pointer to the current process
     Kernel::Process* CurrentProcess();
 

src/core/core_cpu.cpp

@@ -52,7 +52,8 @@ bool CpuBarrier::Rendezvous() {
 
 Cpu::Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
          std::size_t core_index)
-    : cpu_barrier{cpu_barrier}, core_timing{system.CoreTiming()}, core_index{core_index} {
+    : cpu_barrier{cpu_barrier}, global_scheduler{system.GlobalScheduler()},
+      core_timing{system.CoreTiming()}, core_index{core_index} {
 #ifdef ARCHITECTURE_x86_64
     arm_interface = std::make_unique<ARM_Dynarmic>(system, exclusive_monitor, core_index);
 #else
@@ -60,7 +61,7 @@ Cpu::Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_ba
     LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
 
-    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface);
+    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface, core_index);
 }
 
 Cpu::~Cpu() = default;
@@ -81,29 +82,21 @@ void Cpu::RunLoop(bool tight_loop) {
         return;
     }
 
+    Reschedule();
+
     // If we don't have a currently active thread then don't execute instructions,
     // instead advance to the next event and try to yield to the next thread
     if (Kernel::GetCurrentThread() == nullptr) {
         LOG_TRACE(Core, "Core-{} idling", core_index);
-
-        if (IsMainCore()) {
-            // TODO(Subv): Only let CoreTiming idle if all 4 cores are idling.
-            core_timing.Idle();
-            core_timing.Advance();
-        }
-
-        PrepareReschedule();
+        core_timing.Idle();
     } else {
-        if (IsMainCore()) {
-            core_timing.Advance();
-        }
-
         if (tight_loop) {
             arm_interface->Run();
         } else {
             arm_interface->Step();
         }
     }
+    core_timing.Advance();
 
     Reschedule();
 }
@@ -114,18 +107,18 @@ void Cpu::SingleStep() {
 
 void Cpu::PrepareReschedule() {
     arm_interface->PrepareReschedule();
-    reschedule_pending = true;
 }
 
 void Cpu::Reschedule() {
-    if (!reschedule_pending) {
-        return;
-    }
-
-    reschedule_pending = false;
     // Lock the global kernel mutex when we manipulate the HLE state
-    std::lock_guard lock{HLE::g_hle_lock};
-    scheduler->Reschedule();
+    std::lock_guard lock(HLE::g_hle_lock);
+
+    global_scheduler.SelectThread(core_index);
+    scheduler->TryDoContextSwitch();
+}
+
+void Cpu::Shutdown() {
+    scheduler->Shutdown();
 }
 
 } // namespace Core

src/core/core_cpu.h

@@ -12,8 +12,9 @@
 #include "common/common_types.h"
 
 namespace Kernel {
+class GlobalScheduler;
 class Scheduler;
-}
+} // namespace Kernel
 
 namespace Core {
 class System;
@@ -83,6 +84,8 @@ public:
         return core_index;
     }
 
+    void Shutdown();
+
     static std::unique_ptr<ExclusiveMonitor> MakeExclusiveMonitor(std::size_t num_cores);
 
 private:
@@ -90,6 +93,7 @@ private:
 
     std::unique_ptr<ARM_Interface> arm_interface;
     CpuBarrier& cpu_barrier;
+    Kernel::GlobalScheduler& global_scheduler;
     std::unique_ptr<Kernel::Scheduler> scheduler;
     Timing::CoreTiming& core_timing;
 

src/core/core_timing.cpp

@@ -15,7 +15,7 @@
 
 namespace Core::Timing {
 
-constexpr int MAX_SLICE_LENGTH = 20000;
+constexpr int MAX_SLICE_LENGTH = 10000;
 
 struct CoreTiming::Event {
     s64 time;
@@ -38,10 +38,12 @@ CoreTiming::CoreTiming() = default;
 CoreTiming::~CoreTiming() = default;
 
 void CoreTiming::Initialize() {
-    downcount = MAX_SLICE_LENGTH;
+    downcounts.fill(MAX_SLICE_LENGTH);
+    time_slice.fill(MAX_SLICE_LENGTH);
     slice_length = MAX_SLICE_LENGTH;
     global_timer = 0;
     idled_cycles = 0;
+    current_context = 0;
 
     // The time between CoreTiming being initialized and the first call to Advance() is considered
     // the slice boundary between slice -1 and slice 0. Dispatcher loops must call Advance() before
@@ -110,7 +112,7 @@ void CoreTiming::UnscheduleEvent(const EventType* event_type, u64 userdata) {
 u64 CoreTiming::GetTicks() const {
     u64 ticks = static_cast<u64>(global_timer);
     if (!is_global_timer_sane) {
-        ticks += slice_length - downcount;
+        ticks += accumulated_ticks;
     }
     return ticks;
 }
@@ -120,7 +122,8 @@ u64 CoreTiming::GetIdleTicks() const {
 }
 
 void CoreTiming::AddTicks(u64 ticks) {
-    downcount -= static_cast<int>(ticks);
+    accumulated_ticks += ticks;
+    downcounts[current_context] -= static_cast<s64>(ticks);
 }
 
 void CoreTiming::ClearPendingEvents() {
@@ -141,22 +144,35 @@ void CoreTiming::RemoveEvent(const EventType* event_type) {
 
 void CoreTiming::ForceExceptionCheck(s64 cycles) {
     cycles = std::max<s64>(0, cycles);
-    if (downcount <= cycles) {
+    if (downcounts[current_context] <= cycles) {
         return;
     }
 
     // downcount is always (much) smaller than MAX_INT so we can safely cast cycles to an int
     // here. Account for cycles already executed by adjusting the g.slice_length
-    slice_length -= downcount - static_cast<int>(cycles);
-    downcount = static_cast<int>(cycles);
+    downcounts[current_context] = static_cast<int>(cycles);
+}
+
+std::optional<u64> CoreTiming::NextAvailableCore(const s64 needed_ticks) const {
+    const u64 original_context = current_context;
+    u64 next_context = (original_context + 1) % num_cpu_cores;
+    while (next_context != original_context) {
+        if (time_slice[next_context] >= needed_ticks) {
+            return {next_context};
+        } else if (time_slice[next_context] >= 0) {
+            return std::nullopt;
+        }
+        next_context = (next_context + 1) % num_cpu_cores;
+    }
+    return std::nullopt;
 }
 
 void CoreTiming::Advance() {
     std::unique_lock<std::mutex> guard(inner_mutex);
 
-    const int cycles_executed = slice_length - downcount;
+    const u64 cycles_executed = accumulated_ticks;
+    time_slice[current_context] = std::max<s64>(0, time_slice[current_context] - accumulated_ticks);
     global_timer += cycles_executed;
-    slice_length = MAX_SLICE_LENGTH;
 
     is_global_timer_sane = true;
 
@@ -173,24 +189,46 @@ void CoreTiming::Advance() {
 
     // Still events left (scheduled in the future)
     if (!event_queue.empty()) {
-        slice_length = static_cast<int>(
-            std::min<s64>(event_queue.front().time - global_timer, MAX_SLICE_LENGTH));
+        const s64 needed_ticks =
+            std::min<s64>(event_queue.front().time - global_timer, MAX_SLICE_LENGTH);
+        const auto next_core = NextAvailableCore(needed_ticks);
+        if (next_core) {
+            downcounts[*next_core] = needed_ticks;
+        }
     }
 
-    downcount = slice_length;
+    accumulated_ticks = 0;
+
+    downcounts[current_context] = time_slice[current_context];
+}
+
+void CoreTiming::ResetRun() {
+    downcounts.fill(MAX_SLICE_LENGTH);
+    time_slice.fill(MAX_SLICE_LENGTH);
+    current_context = 0;
+    // Still events left (scheduled in the future)
+    if (!event_queue.empty()) {
+        const s64 needed_ticks =
+            std::min<s64>(event_queue.front().time - global_timer, MAX_SLICE_LENGTH);
+        downcounts[current_context] = needed_ticks;
+    }
+
+    is_global_timer_sane = false;
+    accumulated_ticks = 0;
 }
 
 void CoreTiming::Idle() {
-    idled_cycles += downcount;
-    downcount = 0;
+    accumulated_ticks += downcounts[current_context];
+    idled_cycles += downcounts[current_context];
+    downcounts[current_context] = 0;
 }
 
 std::chrono::microseconds CoreTiming::GetGlobalTimeUs() const {
     return std::chrono::microseconds{GetTicks() * 1000000 / BASE_CLOCK_RATE};
 }
 
-int CoreTiming::GetDowncount() const {
-    return downcount;
+s64 CoreTiming::GetDowncount() const {
+    return downcounts[current_context];
 }
 
 } // namespace Core::Timing

src/core/core_timing.h

@@ -7,6 +7,7 @@
 #include <chrono>
 #include <functional>
 #include <mutex>
+#include <optional>
 #include <string>
 #include <unordered_map>
 #include <vector>
@@ -104,7 +105,19 @@ public:
 
     std::chrono::microseconds GetGlobalTimeUs() const;
 
-    int GetDowncount() const;
+    void ResetRun();
+
+    s64 GetDowncount() const;
+
+    void SwitchContext(u64 new_context) {
+        current_context = new_context;
+    }
+
+    bool CanCurrentContextRun() const {
+        return time_slice[current_context] > 0;
+    }
+
+    std::optional<u64> NextAvailableCore(const s64 needed_ticks) const;
 
 private:
     struct Event;
@@ -112,10 +125,16 @@ private:
     /// Clear all pending events. This should ONLY be done on exit.
     void ClearPendingEvents();
 
+    static constexpr u64 num_cpu_cores = 4;
+
     s64 global_timer = 0;
     s64 idled_cycles = 0;
-    int slice_length = 0;
-    int downcount = 0;
+    s64 slice_length = 0;
+    u64 accumulated_ticks = 0;
+    std::array<s64, num_cpu_cores> downcounts{};
+    // Slice of time assigned to each core per run.
+    std::array<s64, num_cpu_cores> time_slice{};
+    u64 current_context = 0;
 
     // Are we in a function that has been called from Advance()
     // If events are scheduled from a function that gets called from Advance(),

src/core/cpu_core_manager.cpp

@@ -6,6 +6,7 @@
 #include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
 #include "core/core_cpu.h"
+#include "core/core_timing.h"
 #include "core/cpu_core_manager.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/settings.h"
@@ -57,6 +58,7 @@ void CpuCoreManager::Shutdown() {
 
     thread_to_cpu.clear();
     for (auto& cpu_core : cores) {
+        cpu_core->Shutdown();
         cpu_core.reset();
     }
 
@@ -122,13 +124,19 @@ void CpuCoreManager::RunLoop(bool tight_loop) {
         }
     }
 
-    for (active_core = 0; active_core < NUM_CPU_CORES; ++active_core) {
-        cores[active_core]->RunLoop(tight_loop);
-        if (Settings::values.use_multi_core) {
-            // Cores 1-3 are run on other threads in this mode
-            break;
+    auto& core_timing = system.CoreTiming();
+    core_timing.ResetRun();
+    bool keep_running{};
+    do {
+        keep_running = false;
+        for (active_core = 0; active_core < NUM_CPU_CORES; ++active_core) {
+            core_timing.SwitchContext(active_core);
+            if (core_timing.CanCurrentContextRun()) {
+                cores[active_core]->RunLoop(tight_loop);
+            }
+            keep_running |= core_timing.CanCurrentContextRun();
         }
-    }
+    } while (keep_running);
 
     if (GDBStub::IsServerEnabled()) {
         GDBStub::SetCpuStepFlag(false);

src/core/file_sys/card_image.cpp

@@ -31,7 +31,7 @@ constexpr std::array partition_names{
 
 XCI::XCI(VirtualFile file_)
     : file(std::move(file_)), program_nca_status{Loader::ResultStatus::ErrorXCIMissingProgramNCA},
-      partitions(partition_names.size()) {
+      partitions(partition_names.size()), partitions_raw(partition_names.size()) {
     if (file->ReadObject(&header) != sizeof(GamecardHeader)) {
         status = Loader::ResultStatus::ErrorBadXCIHeader;
         return;
@@ -42,8 +42,10 @@ XCI::XCI(VirtualFile file_)
         return;
     }
 
-    PartitionFilesystem main_hfs(
-        std::make_shared<OffsetVfsFile>(file, header.hfs_size, header.hfs_offset));
+    PartitionFilesystem main_hfs(std::make_shared<OffsetVfsFile>(
+        file, file->GetSize() - header.hfs_offset, header.hfs_offset));
+
+    update_normal_partition_end = main_hfs.GetFileOffsets()["secure"];
 
     if (main_hfs.GetStatus() != Loader::ResultStatus::Success) {
         status = main_hfs.GetStatus();
@@ -55,9 +57,7 @@ XCI::XCI(VirtualFile file_)
         const auto partition_idx = static_cast<std::size_t>(partition);
         auto raw = main_hfs.GetFile(partition_names[partition_idx]);
 
-        if (raw != nullptr) {
-            partitions[partition_idx] = std::make_shared<PartitionFilesystem>(std::move(raw));
-        }
+        partitions_raw[static_cast<std::size_t>(partition)] = std::move(raw);
     }
 
     secure_partition = std::make_shared<NSP>(
@@ -71,13 +71,7 @@ XCI::XCI(VirtualFile file_)
         program_nca_status = Loader::ResultStatus::ErrorXCIMissingProgramNCA;
     }
 
-    auto result = AddNCAFromPartition(XCIPartition::Update);
-    if (result != Loader::ResultStatus::Success) {
-        status = result;
-        return;
-    }
-
-    result = AddNCAFromPartition(XCIPartition::Normal);
+    auto result = AddNCAFromPartition(XCIPartition::Normal);
     if (result != Loader::ResultStatus::Success) {
         status = result;
         return;
@@ -104,34 +98,114 @@ Loader::ResultStatus XCI::GetProgramNCAStatus() const {
     return program_nca_status;
 }
 
-VirtualDir XCI::GetPartition(XCIPartition partition) const {
+VirtualDir XCI::GetPartition(XCIPartition partition) {
+    const auto id = static_cast<std::size_t>(partition);
+    if (partitions[id] == nullptr && partitions_raw[id] != nullptr) {
+        partitions[id] = std::make_shared<PartitionFilesystem>(partitions_raw[id]);
+    }
+
     return partitions[static_cast<std::size_t>(partition)];
 }
 
+std::vector<VirtualDir> XCI::GetPartitions() {
+    std::vector<VirtualDir> out;
+    for (const auto& id :
+         {XCIPartition::Update, XCIPartition::Normal, XCIPartition::Secure, XCIPartition::Logo}) {
+        const auto part = GetPartition(id);
+        if (part != nullptr) {
+            out.push_back(part);
+        }
+    }
+    return out;
+}
+
 std::shared_ptr<NSP> XCI::GetSecurePartitionNSP() const {
     return secure_partition;
 }
 
-VirtualDir XCI::GetSecurePartition() const {
+VirtualDir XCI::GetSecurePartition() {
     return GetPartition(XCIPartition::Secure);
 }
 
-VirtualDir XCI::GetNormalPartition() const {
+VirtualDir XCI::GetNormalPartition() {
     return GetPartition(XCIPartition::Normal);
 }
 
-VirtualDir XCI::GetUpdatePartition() const {
+VirtualDir XCI::GetUpdatePartition() {
     return GetPartition(XCIPartition::Update);
 }
 
-VirtualDir XCI::GetLogoPartition() const {
+VirtualDir XCI::GetLogoPartition() {
     return GetPartition(XCIPartition::Logo);
 }
 
+VirtualFile XCI::GetPartitionRaw(XCIPartition partition) const {
+    return partitions_raw[static_cast<std::size_t>(partition)];
+}
+
+VirtualFile XCI::GetSecurePartitionRaw() const {
+    return GetPartitionRaw(XCIPartition::Secure);
+}
+
+VirtualFile XCI::GetStoragePartition0() const {
+    return std::make_shared<OffsetVfsFile>(file, update_normal_partition_end, 0, "partition0");
+}
+
+VirtualFile XCI::GetStoragePartition1() const {
+    return std::make_shared<OffsetVfsFile>(file, file->GetSize() - update_normal_partition_end,
+                                           update_normal_partition_end, "partition1");
+}
+
+VirtualFile XCI::GetNormalPartitionRaw() const {
+    return GetPartitionRaw(XCIPartition::Normal);
+}
+
+VirtualFile XCI::GetUpdatePartitionRaw() const {
+    return GetPartitionRaw(XCIPartition::Update);
+}
+
+VirtualFile XCI::GetLogoPartitionRaw() const {
+    return GetPartitionRaw(XCIPartition::Logo);
+}
+
 u64 XCI::GetProgramTitleID() const {
     return secure_partition->GetProgramTitleID();
 }
 
+u32 XCI::GetSystemUpdateVersion() {
+    const auto update = GetPartition(XCIPartition::Update);
+    if (update == nullptr)
+        return 0;
+
+    for (const auto& file : update->GetFiles()) {
+        NCA nca{file, nullptr, 0, keys};
+
+        if (nca.GetStatus() != Loader::ResultStatus::Success)
+            continue;
+
+        if (nca.GetType() == NCAContentType::Meta && nca.GetTitleId() == 0x0100000000000816) {
+            const auto dir = nca.GetSubdirectories()[0];
+            const auto cnmt = dir->GetFile("SystemUpdate_0100000000000816.cnmt");
+            if (cnmt == nullptr)
+                continue;
+
+            CNMT cnmt_data{cnmt};
+
+            const auto metas = cnmt_data.GetMetaRecords();
+            if (metas.empty())
+                continue;
+
+            return metas[0].title_version;
+        }
+    }
+
+    return 0;
+}
+
+u64 XCI::GetSystemUpdateTitleID() const {
+    return 0x0100000000000816;
+}
+
 bool XCI::HasProgramNCA() const {
     return program != nullptr;
 }
@@ -201,7 +275,7 @@ std::array<u8, 0x200> XCI::GetCertificate() const {
 
 Loader::ResultStatus XCI::AddNCAFromPartition(XCIPartition part) {
     const auto partition_index = static_cast<std::size_t>(part);
-    const auto& partition = partitions[partition_index];
+    const auto partition = GetPartition(part);
 
     if (partition == nullptr) {
         return Loader::ResultStatus::ErrorXCIMissingPartition;
@@ -232,7 +306,7 @@ Loader::ResultStatus XCI::AddNCAFromPartition(XCIPartition part) {
     return Loader::ResultStatus::Success;
 }
 
-u8 XCI::GetFormatVersion() const {
+u8 XCI::GetFormatVersion() {
     return GetLogoPartition() == nullptr ? 0x1 : 0x2;
 }
 } // namespace FileSys

src/core/file_sys/card_image.h

@@ -81,14 +81,24 @@ public:
     Loader::ResultStatus GetStatus() const;
     Loader::ResultStatus GetProgramNCAStatus() const;
 
-    u8 GetFormatVersion() const;
+    u8 GetFormatVersion();
+
+    VirtualDir GetPartition(XCIPartition partition);
+    std::vector<VirtualDir> GetPartitions();
 
-    VirtualDir GetPartition(XCIPartition partition) const;
     std::shared_ptr<NSP> GetSecurePartitionNSP() const;
-    VirtualDir GetSecurePartition() const;
-    VirtualDir GetNormalPartition() const;
-    VirtualDir GetUpdatePartition() const;
-    VirtualDir GetLogoPartition() const;
+    VirtualDir GetSecurePartition();
+    VirtualDir GetNormalPartition();
+    VirtualDir GetUpdatePartition();
+    VirtualDir GetLogoPartition();
+
+    VirtualFile GetPartitionRaw(XCIPartition partition) const;
+    VirtualFile GetSecurePartitionRaw() const;
+    VirtualFile GetStoragePartition0() const;
+    VirtualFile GetStoragePartition1() const;
+    VirtualFile GetNormalPartitionRaw() const;
+    VirtualFile GetUpdatePartitionRaw() const;
+    VirtualFile GetLogoPartitionRaw() const;
 
     u64 GetProgramTitleID() const;
     u32 GetSystemUpdateVersion();
@@ -123,6 +133,7 @@ private:
     Loader::ResultStatus program_nca_status;
 
     std::vector<VirtualDir> partitions;
+    std::vector<VirtualFile> partitions_raw;
     std::shared_ptr<NSP> secure_partition;
     std::shared_ptr<NCA> program;
     std::vector<std::shared_ptr<NCA>> ncas;

src/core/file_sys/partition_filesystem.cpp

@@ -65,6 +65,9 @@ PartitionFilesystem::PartitionFilesystem(std::shared_ptr<VfsFile> file) {
         std::string name(
             reinterpret_cast<const char*>(&file_data[strtab_offset + entry.strtab_offset]));
 
+        offsets.insert_or_assign(name, content_offset + entry.offset);
+        sizes.insert_or_assign(name, entry.size);
+
         pfs_files.emplace_back(std::make_shared<OffsetVfsFile>(
             file, entry.size, content_offset + entry.offset, std::move(name)));
     }
@@ -78,6 +81,14 @@ Loader::ResultStatus PartitionFilesystem::GetStatus() const {
     return status;
 }
 
+std::map<std::string, u64> PartitionFilesystem::GetFileOffsets() const {
+    return offsets;
+}
+
+std::map<std::string, u64> PartitionFilesystem::GetFileSizes() const {
+    return sizes;
+}
+
 std::vector<std::shared_ptr<VfsFile>> PartitionFilesystem::GetFiles() const {
     return pfs_files;
 }

src/core/file_sys/partition_filesystem.h

@@ -29,6 +29,9 @@ public:
 
     Loader::ResultStatus GetStatus() const;
 
+    std::map<std::string, u64> GetFileOffsets() const;
+    std::map<std::string, u64> GetFileSizes() const;
+
     std::vector<std::shared_ptr<VfsFile>> GetFiles() const override;
     std::vector<std::shared_ptr<VfsDirectory>> GetSubdirectories() const override;
     std::string GetName() const override;
@@ -80,6 +83,9 @@ private:
     bool is_hfs = false;
     std::size_t content_offset = 0;
 
+    std::map<std::string, u64> offsets;
+    std::map<std::string, u64> sizes;
+
     std::vector<VirtualFile> pfs_files;
 };
 

src/core/file_sys/savedata_factory.cpp

@@ -16,6 +16,7 @@ namespace FileSys {
 constexpr char SAVE_DATA_SIZE_FILENAME[] = ".yuzu_save_size";
 
 namespace {
+
 void PrintSaveDataDescriptorWarnings(SaveDataDescriptor meta) {
     if (meta.type == SaveDataType::SystemSaveData || meta.type == SaveDataType::SaveData) {
         if (meta.zero_1 != 0) {
@@ -52,6 +53,13 @@ void PrintSaveDataDescriptorWarnings(SaveDataDescriptor meta) {
                     meta.user_id[1], meta.user_id[0]);
     }
 }
+
+bool ShouldSaveDataBeAutomaticallyCreated(SaveDataSpaceId space, const SaveDataDescriptor& desc) {
+    return desc.type == SaveDataType::CacheStorage || desc.type == SaveDataType::TemporaryStorage ||
+           (space == SaveDataSpaceId::NandUser && ///< Normal Save Data -- Current Title & User
+            desc.type == SaveDataType::SaveData && desc.title_id == 0 && desc.save_id == 0);
+}
+
 } // Anonymous namespace
 
 std::string SaveDataDescriptor::DebugInfo() const {
@@ -96,6 +104,10 @@ ResultVal<VirtualDir> SaveDataFactory::Open(SaveDataSpaceId space,
 
     auto out = dir->GetDirectoryRelative(save_directory);
 
+    if (out == nullptr && ShouldSaveDataBeAutomaticallyCreated(space, meta)) {
+        return Create(space, meta);
+    }
+
     // Return an error if the save data doesn't actually exist.
     if (out == nullptr) {
         // TODO(Subv): Find out correct error code.

src/core/file_sys/system_archive/data/font_chinese_simplified.h

@@ -0,0 +1,13 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+namespace FileSys::SystemArchive::SharedFontData {
+
+extern const std::array<unsigned char, 217276> FONT_CHINESE_SIMPLIFIED;
+
+} // namespace FileSys::SystemArchive::SharedFontData

src/core/file_sys/system_archive/data/font_chinese_traditional.h

@@ -0,0 +1,13 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+namespace FileSys::SystemArchive::SharedFontData {
+
+extern const std::array<unsigned char, 222236> FONT_CHINESE_TRADITIONAL;
+
+} // namespace FileSys::SystemArchive::SharedFontData

src/core/file_sys/system_archive/data/font_extended_chinese_simplified.h

@@ -0,0 +1,13 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+namespace FileSys::SystemArchive::SharedFontData {
+
+extern const std::array<unsigned char, 293516> FONT_EXTENDED_CHINESE_SIMPLIFIED;
+
+} // namespace FileSys::SystemArchive::SharedFontData

src/core/file_sys/system_archive/data/font_korean.h

@@ -0,0 +1,13 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+namespace FileSys::SystemArchive::SharedFontData {
+
+extern const std::array<unsigned char, 217276> FONT_KOREAN;
+
+} // namespace FileSys::SystemArchive::SharedFontData

src/core/file_sys/system_archive/data/font_nintendo_extended.cpp

@@ -0,0 +1,196 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/file_sys/system_archive/data/font_nintendo_extended.h"
+
+namespace FileSys::SystemArchive::SharedFontData {
+
+const std::array<unsigned char, 2932> FONT_NINTENDO_EXTENDED{{
+    0x00, 0x01, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x80, 0x00, 0x03, 0x00, 0x70, 0x44, 0x53, 0x49, 0x47,
+    0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x0b, 0x6c, 0x00, 0x00, 0x00, 0x08, 0x4f, 0x53, 0x2f, 0x32,
+    0x33, 0x86, 0x1d, 0x9b, 0x00, 0x00, 0x01, 0x78, 0x00, 0x00, 0x00, 0x60, 0x63, 0x6d, 0x61, 0x70,
+    0xc2, 0x06, 0x20, 0xde, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0xa0, 0x63, 0x76, 0x74, 0x20,
+    0x00, 0x14, 0x00, 0x00, 0x00, 0x00, 0x04, 0x2c, 0x00, 0x00, 0x00, 0x06, 0x66, 0x70, 0x67, 0x6d,
+    0x06, 0x59, 0x9c, 0x37, 0x00, 0x00, 0x02, 0xa0, 0x00, 0x00, 0x01, 0x73, 0x67, 0x61, 0x73, 0x70,
+    0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x0b, 0x64, 0x00, 0x00, 0x00, 0x08, 0x67, 0x6c, 0x79, 0x66,
+    0x10, 0x31, 0x88, 0x00, 0x00, 0x00, 0x04, 0x34, 0x00, 0x00, 0x04, 0x64, 0x68, 0x65, 0x61, 0x64,
+    0x15, 0x9d, 0xef, 0x91, 0x00, 0x00, 0x00, 0xfc, 0x00, 0x00, 0x00, 0x36, 0x68, 0x68, 0x65, 0x61,
+    0x09, 0x60, 0x03, 0x71, 0x00, 0x00, 0x01, 0x34, 0x00, 0x00, 0x00, 0x24, 0x68, 0x6d, 0x74, 0x78,
+    0x0d, 0x2e, 0x03, 0xa7, 0x00, 0x00, 0x01, 0xd8, 0x00, 0x00, 0x00, 0x26, 0x6c, 0x6f, 0x63, 0x61,
+    0x05, 0xc0, 0x04, 0x6c, 0x00, 0x00, 0x08, 0x98, 0x00, 0x00, 0x00, 0x1e, 0x6d, 0x61, 0x78, 0x70,
+    0x02, 0x1c, 0x00, 0x5f, 0x00, 0x00, 0x01, 0x58, 0x00, 0x00, 0x00, 0x20, 0x6e, 0x61, 0x6d, 0x65,
+    0x7c, 0xe0, 0x84, 0x5c, 0x00, 0x00, 0x08, 0xb8, 0x00, 0x00, 0x02, 0x09, 0x70, 0x6f, 0x73, 0x74,
+    0x47, 0x4e, 0x74, 0x19, 0x00, 0x00, 0x0a, 0xc4, 0x00, 0x00, 0x00, 0x9e, 0x70, 0x72, 0x65, 0x70,
+    0x1c, 0xfc, 0x7d, 0x9c, 0x00, 0x00, 0x04, 0x14, 0x00, 0x00, 0x00, 0x16, 0x00, 0x01, 0x00, 0x00,
+    0x00, 0x01, 0x00, 0x00, 0x7c, 0xc7, 0xb1, 0x63, 0x5f, 0x0f, 0x3c, 0xf5, 0x00, 0x1b, 0x03, 0xe8,
+    0x00, 0x00, 0x00, 0x00, 0xd9, 0x44, 0x2f, 0x5d, 0x00, 0x00, 0x00, 0x00, 0xd9, 0x45, 0x7b, 0x69,
+    0x00, 0x00, 0x00, 0x00, 0x03, 0xe6, 0x03, 0xe8, 0x00, 0x00, 0x00, 0x06, 0x00, 0x02, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x03, 0x84, 0xff, 0x83, 0x01, 0xf4, 0x03, 0xe8,
+    0x00, 0x00, 0x00, 0x00, 0x03, 0xe6, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x5e,
+    0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x03, 0x74, 0x01, 0x90, 0x00, 0x05,
+    0x00, 0x04, 0x00, 0xcd, 0x00, 0xcd, 0x00, 0x00, 0x01, 0x1f, 0x00, 0xcd, 0x00, 0xcd, 0x00, 0x00,
+    0x03, 0xc3, 0x00, 0x66, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x20, 0x20, 0x20, 0x20, 0x00, 0xc0, 0x00, 0x00, 0xe0, 0xe9, 0x03, 0x84, 0xff, 0x83,
+    0x01, 0xf4, 0x02, 0xee, 0x00, 0xfa, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x03, 0xe8,
+    0x02, 0xbc, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x03, 0xe8, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0xfa, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x00, 0x03, 0xe8, 0x00, 0xeb, 0x01, 0x21, 0x00, 0xff,
+    0x00, 0xff, 0x01, 0x3d, 0x01, 0x17, 0x00, 0x42, 0x00, 0x1c, 0x00, 0x3e, 0x00, 0x17, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x68, 0x00, 0x01, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x1c, 0x00, 0x03, 0x00, 0x01, 0x00, 0x00, 0x00, 0x68, 0x00, 0x06, 0x00, 0x4c,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x21, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x04, 0x00, 0x38, 0x00, 0x00, 0x00, 0x0a,
+    0x00, 0x08, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x20, 0xe0, 0xe9, 0xff, 0xff,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x20, 0xe0, 0xe0, 0xff, 0xff, 0x00, 0x01, 0xff, 0xf5,
+    0xff, 0xe3, 0x1f, 0x24, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0xb8, 0x00, 0x00, 0x2c, 0x4b, 0xb8, 0x00, 0x09, 0x50, 0x58, 0xb1, 0x01, 0x01, 0x8e, 0x59, 0xb8,
+    0x01, 0xff, 0x85, 0xb8, 0x00, 0x44, 0x1d, 0xb9, 0x00, 0x09, 0x00, 0x03, 0x5f, 0x5e, 0x2d, 0xb8,
+    0x00, 0x01, 0x2c, 0x20, 0x20, 0x45, 0x69, 0x44, 0xb0, 0x01, 0x60, 0x2d, 0xb8, 0x00, 0x02, 0x2c,
+    0xb8, 0x00, 0x01, 0x2a, 0x21, 0x2d, 0xb8, 0x00, 0x03, 0x2c, 0x20, 0x46, 0xb0, 0x03, 0x25, 0x46,
+    0x52, 0x58, 0x23, 0x59, 0x20, 0x8a, 0x20, 0x8a, 0x49, 0x64, 0x8a, 0x20, 0x46, 0x20, 0x68, 0x61,
+    0x64, 0xb0, 0x04, 0x25, 0x46, 0x20, 0x68, 0x61, 0x64, 0x52, 0x58, 0x23, 0x65, 0x8a, 0x59, 0x2f,
+    0x20, 0xb0, 0x00, 0x53, 0x58, 0x69, 0x20, 0xb0, 0x00, 0x54, 0x58, 0x21, 0xb0, 0x40, 0x59, 0x1b,
+    0x69, 0x20, 0xb0, 0x00, 0x54, 0x58, 0x21, 0xb0, 0x40, 0x65, 0x59, 0x59, 0x3a, 0x2d, 0xb8, 0x00,
+    0x04, 0x2c, 0x20, 0x46, 0xb0, 0x04, 0x25, 0x46, 0x52, 0x58, 0x23, 0x8a, 0x59, 0x20, 0x46, 0x20,
+    0x6a, 0x61, 0x64, 0xb0, 0x04, 0x25, 0x46, 0x20, 0x6a, 0x61, 0x64, 0x52, 0x58, 0x23, 0x8a, 0x59,
+    0x2f, 0xfd, 0x2d, 0xb8, 0x00, 0x05, 0x2c, 0x4b, 0x20, 0xb0, 0x03, 0x26, 0x50, 0x58, 0x51, 0x58,
+    0xb0, 0x80, 0x44, 0x1b, 0xb0, 0x40, 0x44, 0x59, 0x1b, 0x21, 0x21, 0x20, 0x45, 0xb0, 0xc0, 0x50,
+    0x58, 0xb0, 0xc0, 0x44, 0x1b, 0x21, 0x59, 0x59, 0x2d, 0xb8, 0x00, 0x06, 0x2c, 0x20, 0x20, 0x45,
+    0x69, 0x44, 0xb0, 0x01, 0x60, 0x20, 0x20, 0x45, 0x7d, 0x69, 0x18, 0x44, 0xb0, 0x01, 0x60, 0x2d,
+    0xb8, 0x00, 0x07, 0x2c, 0xb8, 0x00, 0x06, 0x2a, 0x2d, 0xb8, 0x00, 0x08, 0x2c, 0x4b, 0x20, 0xb0,
+    0x03, 0x26, 0x53, 0x58, 0xb0, 0x40, 0x1b, 0xb0, 0x00, 0x59, 0x8a, 0x8a, 0x20, 0xb0, 0x03, 0x26,
+    0x53, 0x58, 0x23, 0x21, 0xb0, 0x80, 0x8a, 0x8a, 0x1b, 0x8a, 0x23, 0x59, 0x20, 0xb0, 0x03, 0x26,
+    0x53, 0x58, 0x23, 0x21, 0xb8, 0x00, 0xc0, 0x8a, 0x8a, 0x1b, 0x8a, 0x23, 0x59, 0x20, 0xb0, 0x03,
+    0x26, 0x53, 0x58, 0x23, 0x21, 0xb8, 0x01, 0x00, 0x8a, 0x8a, 0x1b, 0x8a, 0x23, 0x59, 0x20, 0xb0,
+    0x03, 0x26, 0x53, 0x58, 0x23, 0x21, 0xb8, 0x01, 0x40, 0x8a, 0x8a, 0x1b, 0x8a, 0x23, 0x59, 0x20,
+    0xb8, 0x00, 0x03, 0x26, 0x53, 0x58, 0xb0, 0x03, 0x25, 0x45, 0xb8, 0x01, 0x80, 0x50, 0x58, 0x23,
+    0x21, 0xb8, 0x01, 0x80, 0x23, 0x21, 0x1b, 0xb0, 0x03, 0x25, 0x45, 0x23, 0x21, 0x23, 0x21, 0x59,
+    0x1b, 0x21, 0x59, 0x44, 0x2d, 0xb8, 0x00, 0x09, 0x2c, 0x4b, 0x53, 0x58, 0x45, 0x44, 0x1b, 0x21,
+    0x21, 0x59, 0x2d, 0x00, 0xb8, 0x00, 0x00, 0x2b, 0x00, 0xba, 0x00, 0x01, 0x00, 0x01, 0x00, 0x07,
+    0x2b, 0xb8, 0x00, 0x00, 0x20, 0x45, 0x7d, 0x69, 0x18, 0x44, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x02, 0x03, 0xe6, 0x03, 0xe8, 0x00, 0x06,
+    0x00, 0x00, 0x35, 0x01, 0x33, 0x15, 0x01, 0x23, 0x35, 0x03, 0x52, 0x94, 0xfc, 0xa6, 0x8c, 0x90,
+    0x03, 0x58, 0x86, 0xfc, 0xa0, 0x8e, 0x00, 0x00, 0x00, 0x02, 0x00, 0xeb, 0x00, 0xcc, 0x02, 0xfb,
+    0x03, 0x1e, 0x00, 0x08, 0x00, 0x0f, 0x00, 0x00, 0x01, 0x33, 0x13, 0x23, 0x27, 0x23, 0x07, 0x23,
+    0x13, 0x17, 0x07, 0x06, 0x15, 0x33, 0x27, 0x07, 0x01, 0xbc, 0x6d, 0xd2, 0x7c, 0x26, 0xcc, 0x26,
+    0x7c, 0xd1, 0x35, 0x40, 0x02, 0x89, 0x45, 0x02, 0x03, 0x1e, 0xfd, 0xae, 0x77, 0x77, 0x02, 0x52,
+    0x9b, 0xcc, 0x08, 0x04, 0xda, 0x02, 0x00, 0x00, 0x00, 0x03, 0x01, 0x21, 0x00, 0xcc, 0x02, 0xc5,
+    0x03, 0x1e, 0x00, 0x15, 0x00, 0x1f, 0x00, 0x2b, 0x00, 0x00, 0x25, 0x11, 0x33, 0x32, 0x1e, 0x02,
+    0x15, 0x14, 0x0e, 0x02, 0x07, 0x1e, 0x01, 0x15, 0x14, 0x0e, 0x02, 0x2b, 0x01, 0x13, 0x33, 0x32,
+    0x36, 0x35, 0x34, 0x26, 0x2b, 0x01, 0x1d, 0x01, 0x33, 0x32, 0x3e, 0x02, 0x35, 0x34, 0x26, 0x2b,
+    0x01, 0x15, 0x01, 0x21, 0xea, 0x25, 0x3f, 0x2e, 0x1a, 0x0e, 0x15, 0x1b, 0x0e, 0x2d, 0x2d, 0x1a,
+    0x2e, 0x3f, 0x25, 0xf8, 0x76, 0x62, 0x20, 0x2a, 0x28, 0x22, 0x62, 0x76, 0x10, 0x18, 0x11, 0x09,
+    0x22, 0x22, 0x74, 0xcc, 0x02, 0x52, 0x18, 0x2b, 0x3c, 0x24, 0x1d, 0x1f, 0x17, 0x17, 0x14, 0x0f,
+    0x48, 0x2f, 0x24, 0x3f, 0x2e, 0x1a, 0x01, 0x5b, 0x29, 0x20, 0x20, 0x2b, 0x94, 0xf8, 0x0e, 0x16,
+    0x1c, 0x0e, 0x1f, 0x31, 0x9e, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0xff, 0x00, 0xcc, 0x02, 0xe7,
+    0x03, 0x1e, 0x00, 0x0c, 0x00, 0x00, 0x01, 0x33, 0x17, 0x37, 0x33, 0x03, 0x13, 0x23, 0x27, 0x07,
+    0x23, 0x13, 0x03, 0x01, 0x04, 0x86, 0x69, 0x69, 0x86, 0xa3, 0xa8, 0x88, 0x6c, 0x6c, 0x88, 0xa8,
+    0xa3, 0x03, 0x1e, 0xcb, 0xcb, 0xfe, 0xda, 0xfe, 0xd4, 0xcf, 0xcf, 0x01, 0x2c, 0x01, 0x26, 0x00,
+    0x00, 0x01, 0x00, 0xff, 0x00, 0xcc, 0x02, 0xe7, 0x03, 0x1e, 0x00, 0x0f, 0x00, 0x00, 0x01, 0x03,
+    0x33, 0x17, 0x32, 0x15, 0x1e, 0x01, 0x15, 0x1b, 0x01, 0x33, 0x03, 0x15, 0x23, 0x35, 0x01, 0xb8,
+    0xb9, 0x7e, 0x01, 0x01, 0x01, 0x03, 0x70, 0x75, 0x7f, 0xb9, 0x76, 0x01, 0xa3, 0x01, 0x7b, 0x01,
+    0x01, 0x01, 0x05, 0x02, 0xff, 0x00, 0x01, 0x0a, 0xfe, 0x85, 0xd7, 0xd7, 0x00, 0x01, 0x01, 0x3d,
+    0x00, 0xcc, 0x02, 0xa9, 0x03, 0x1e, 0x00, 0x06, 0x00, 0x00, 0x25, 0x11, 0x33, 0x11, 0x33, 0x15,
+    0x21, 0x01, 0x3d, 0x75, 0xf7, 0xfe, 0x94, 0xcc, 0x02, 0x52, 0xfe, 0x10, 0x62, 0x00, 0x00, 0x00,
+    0x00, 0x02, 0x01, 0x17, 0x00, 0xbc, 0x02, 0xcf, 0x03, 0x0e, 0x00, 0x15, 0x00, 0x21, 0x00, 0x00,
+    0x25, 0x11, 0x33, 0x32, 0x1e, 0x02, 0x1d, 0x01, 0x0e, 0x03, 0x1d, 0x01, 0x17, 0x15, 0x23, 0x27,
+    0x23, 0x15, 0x23, 0x13, 0x33, 0x32, 0x3e, 0x02, 0x35, 0x34, 0x26, 0x2b, 0x01, 0x15, 0x01, 0x17,
+    0xf4, 0x27, 0x40, 0x2e, 0x19, 0x01, 0x1f, 0x24, 0x1e, 0x78, 0x7d, 0x6a, 0x5c, 0x75, 0x76, 0x72,
+    0x12, 0x19, 0x11, 0x08, 0x26, 0x26, 0x6a, 0xbc, 0x02, 0x52, 0x1d, 0x31, 0x42, 0x25, 0x16, 0x18,
+    0x32, 0x2a, 0x1b, 0x02, 0x01, 0xef, 0x06, 0xd7, 0xd7, 0x01, 0x3f, 0x10, 0x1a, 0x1e, 0x0f, 0x23,
+    0x36, 0xb0, 0x00, 0x00, 0x00, 0x02, 0x00, 0x42, 0x00, 0xbc, 0x03, 0xa4, 0x03, 0x0e, 0x00, 0x0a,
+    0x00, 0x11, 0x00, 0x00, 0x13, 0x35, 0x21, 0x15, 0x01, 0x21, 0x15, 0x21, 0x35, 0x01, 0x21, 0x01,
+    0x11, 0x33, 0x11, 0x33, 0x15, 0x21, 0x42, 0x01, 0xa7, 0xfe, 0xeb, 0x01, 0x1b, 0xfe, 0x53, 0x01,
+    0x15, 0xfe, 0xeb, 0x01, 0xf7, 0x75, 0xf6, 0xfe, 0x95, 0x02, 0xac, 0x62, 0x45, 0xfe, 0x55, 0x62,
+    0x47, 0x01, 0xa9, 0xfe, 0x10, 0x02, 0x52, 0xfe, 0x10, 0x62, 0x00, 0x00, 0x00, 0x03, 0x00, 0x1c,
+    0x00, 0xbc, 0x03, 0xca, 0x03, 0x0e, 0x00, 0x0a, 0x00, 0x21, 0x00, 0x2f, 0x00, 0x00, 0x13, 0x35,
+    0x21, 0x15, 0x01, 0x21, 0x15, 0x21, 0x35, 0x01, 0x21, 0x01, 0x11, 0x33, 0x32, 0x1e, 0x02, 0x15,
+    0x14, 0x06, 0x07, 0x0e, 0x03, 0x15, 0x17, 0x15, 0x23, 0x27, 0x23, 0x15, 0x23, 0x13, 0x33, 0x32,
+    0x3e, 0x02, 0x35, 0x34, 0x2e, 0x02, 0x2b, 0x01, 0x15, 0x1c, 0x01, 0xa7, 0xfe, 0xeb, 0x01, 0x1b,
+    0xfe, 0x53, 0x01, 0x15, 0xfe, 0xeb, 0x01, 0xf7, 0xf3, 0x27, 0x41, 0x2d, 0x19, 0x1c, 0x20, 0x01,
+    0x0d, 0x0e, 0x0a, 0x78, 0x7d, 0x69, 0x5c, 0x75, 0x76, 0x71, 0x11, 0x1a, 0x12, 0x09, 0x0a, 0x14,
+    0x1d, 0x13, 0x69, 0x02, 0xac, 0x62, 0x45, 0xfe, 0x55, 0x62, 0x47, 0x01, 0xa9, 0xfe, 0x10, 0x02,
+    0x52, 0x1d, 0x31, 0x42, 0x25, 0x2b, 0x44, 0x1d, 0x01, 0x08, 0x09, 0x07, 0x01, 0xf1, 0x06, 0xd7,
+    0xd7, 0x01, 0x3f, 0x11, 0x19, 0x1f, 0x0e, 0x11, 0x20, 0x19, 0x0f, 0xb0, 0x00, 0x02, 0x00, 0x3e,
+    0x00, 0xb3, 0x03, 0xa8, 0x03, 0x17, 0x00, 0x3a, 0x00, 0x41, 0x00, 0x00, 0x13, 0x34, 0x3e, 0x02,
+    0x33, 0x32, 0x1e, 0x02, 0x15, 0x23, 0x27, 0x34, 0x27, 0x2e, 0x01, 0x23, 0x22, 0x0e, 0x02, 0x15,
+    0x14, 0x16, 0x15, 0x1e, 0x05, 0x15, 0x14, 0x0e, 0x02, 0x23, 0x22, 0x2e, 0x02, 0x35, 0x33, 0x1e,
+    0x01, 0x33, 0x32, 0x3e, 0x02, 0x35, 0x34, 0x2e, 0x04, 0x35, 0x01, 0x11, 0x33, 0x11, 0x33, 0x15,
+    0x21, 0x50, 0x24, 0x3b, 0x4a, 0x27, 0x28, 0x4b, 0x39, 0x22, 0x73, 0x01, 0x01, 0x08, 0x2b, 0x29,
+    0x10, 0x20, 0x19, 0x0f, 0x01, 0x0b, 0x35, 0x41, 0x46, 0x3b, 0x25, 0x23, 0x3a, 0x4b, 0x27, 0x2b,
+    0x50, 0x3f, 0x26, 0x74, 0x05, 0x34, 0x33, 0x10, 0x20, 0x1a, 0x11, 0x2c, 0x42, 0x4d, 0x42, 0x2c,
+    0x01, 0xef, 0x73, 0xf6, 0xfe, 0x97, 0x02, 0x70, 0x2a, 0x3f, 0x2a, 0x14, 0x18, 0x2e, 0x44, 0x2c,
+    0x02, 0x03, 0x01, 0x27, 0x27, 0x07, 0x10, 0x1a, 0x12, 0x02, 0x0b, 0x02, 0x1f, 0x22, 0x19, 0x17,
+    0x27, 0x3f, 0x34, 0x2c, 0x3e, 0x28, 0x13, 0x1a, 0x32, 0x48, 0x2e, 0x30, 0x30, 0x06, 0x0f, 0x1a,
+    0x13, 0x21, 0x27, 0x1e, 0x1b, 0x29, 0x3e, 0x31, 0xfe, 0x4c, 0x02, 0x53, 0xfe, 0x10, 0x63, 0x00,
+    0x00, 0x03, 0x00, 0x17, 0x00, 0xb3, 0x03, 0xce, 0x03, 0x17, 0x00, 0x38, 0x00, 0x4f, 0x00, 0x5d,
+    0x00, 0x00, 0x13, 0x34, 0x3e, 0x02, 0x33, 0x32, 0x1e, 0x02, 0x15, 0x23, 0x27, 0x34, 0x23, 0x2e,
+    0x01, 0x23, 0x22, 0x0e, 0x02, 0x15, 0x14, 0x1e, 0x04, 0x15, 0x14, 0x0e, 0x02, 0x23, 0x22, 0x2e,
+    0x02, 0x35, 0x33, 0x1e, 0x01, 0x33, 0x32, 0x3e, 0x02, 0x35, 0x34, 0x26, 0x27, 0x2e, 0x03, 0x35,
+    0x01, 0x11, 0x33, 0x32, 0x1e, 0x02, 0x15, 0x14, 0x06, 0x07, 0x30, 0x0e, 0x02, 0x31, 0x17, 0x15,
+    0x23, 0x27, 0x23, 0x15, 0x23, 0x13, 0x33, 0x32, 0x3e, 0x02, 0x35, 0x34, 0x2e, 0x02, 0x2b, 0x01,
+    0x15, 0x2a, 0x24, 0x3a, 0x4a, 0x26, 0x29, 0x4b, 0x39, 0x23, 0x73, 0x01, 0x01, 0x08, 0x2a, 0x2a,
+    0x10, 0x1f, 0x1a, 0x10, 0x2c, 0x42, 0x4d, 0x42, 0x2c, 0x23, 0x39, 0x4b, 0x27, 0x2b, 0x51, 0x3f,
+    0x27, 0x75, 0x05, 0x34, 0x33, 0x10, 0x20, 0x1a, 0x10, 0x1f, 0x1c, 0x25, 0x53, 0x47, 0x2e, 0x01,
+    0xed, 0xf3, 0x27, 0x41, 0x2d, 0x19, 0x1c, 0x20, 0x0c, 0x0e, 0x0c, 0x78, 0x7d, 0x68, 0x5d, 0x75,
+    0x76, 0x71, 0x11, 0x1a, 0x12, 0x09, 0x0a, 0x14, 0x1d, 0x13, 0x69, 0x02, 0x71, 0x2a, 0x3e, 0x2a,
+    0x14, 0x18, 0x2e, 0x44, 0x2c, 0x02, 0x02, 0x27, 0x29, 0x07, 0x11, 0x1a, 0x12, 0x1d, 0x24, 0x1c,
+    0x1d, 0x2b, 0x40, 0x32, 0x2c, 0x3f, 0x29, 0x13, 0x1a, 0x31, 0x49, 0x2e, 0x30, 0x30, 0x06, 0x0f,
+    0x19, 0x13, 0x1e, 0x22, 0x0b, 0x0e, 0x20, 0x2f, 0x43, 0x30, 0xfe, 0x4b, 0x02, 0x52, 0x1d, 0x32,
+    0x42, 0x25, 0x2c, 0x42, 0x1d, 0x08, 0x0a, 0x08, 0xf1, 0x06, 0xd7, 0xd7, 0x01, 0x3f, 0x11, 0x19,
+    0x1f, 0x0e, 0x11, 0x20, 0x19, 0x0f, 0xb0, 0x00, 0x00, 0x00, 0x00, 0x12, 0x00, 0x12, 0x00, 0x12,
+    0x00, 0x12, 0x00, 0x32, 0x00, 0x72, 0x00, 0x8e, 0x00, 0xac, 0x00, 0xbe, 0x00, 0xf0, 0x01, 0x14,
+    0x01, 0x5c, 0x01, 0xb6, 0x02, 0x32, 0x00, 0x00, 0x00, 0x00, 0x00, 0x0d, 0x00, 0xa2, 0x00, 0x01,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x10, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x02, 0x00, 0x07, 0x00, 0x10, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x2f,
+    0x00, 0x17, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x12, 0x00, 0x46, 0x00, 0x01,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x0d, 0x00, 0x58, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x06, 0x00, 0x12, 0x00, 0x65, 0x00, 0x03, 0x00, 0x01, 0x04, 0x09, 0x00, 0x01, 0x00, 0x20,
+    0x00, 0x77, 0x00, 0x03, 0x00, 0x01, 0x04, 0x09, 0x00, 0x02, 0x00, 0x0e, 0x00, 0x97, 0x00, 0x03,
+    0x00, 0x01, 0x04, 0x09, 0x00, 0x03, 0x00, 0x5e, 0x00, 0xa5, 0x00, 0x03, 0x00, 0x01, 0x04, 0x09,
+    0x00, 0x04, 0x00, 0x24, 0x01, 0x03, 0x00, 0x03, 0x00, 0x01, 0x04, 0x09, 0x00, 0x05, 0x00, 0x1a,
+    0x01, 0x27, 0x00, 0x03, 0x00, 0x01, 0x04, 0x09, 0x00, 0x06, 0x00, 0x24, 0x01, 0x41, 0x00, 0x03,
+    0x00, 0x01, 0x04, 0x09, 0x00, 0x11, 0x00, 0x02, 0x01, 0x65, 0x59, 0x75, 0x7a, 0x75, 0x4f, 0x53,
+    0x53, 0x45, 0x78, 0x74, 0x65, 0x6e, 0x73, 0x69, 0x6f, 0x6e, 0x52, 0x65, 0x67, 0x75, 0x6c, 0x61,
+    0x72, 0x56, 0x65, 0x72, 0x73, 0x69, 0x6f, 0x6e, 0x20, 0x31, 0x2e, 0x30, 0x30, 0x30, 0x3b, 0x3b,
+    0x59, 0x75, 0x7a, 0x75, 0x4f, 0x53, 0x53, 0x45, 0x78, 0x74, 0x65, 0x6e, 0x73, 0x69, 0x6f, 0x6e,
+    0x2d, 0x52, 0x3b, 0x32, 0x30, 0x31, 0x39, 0x3b, 0x46, 0x4c, 0x56, 0x49, 0x2d, 0x36, 0x31, 0x34,
+    0x59, 0x75, 0x7a, 0x75, 0x4f, 0x53, 0x53, 0x45, 0x78, 0x74, 0x65, 0x6e, 0x73, 0x69, 0x6f, 0x6e,
+    0x20, 0x52, 0x56, 0x65, 0x72, 0x73, 0x69, 0x6f, 0x6e, 0x20, 0x31, 0x2e, 0x30, 0x30, 0x30, 0x59,
+    0x75, 0x7a, 0x75, 0x4f, 0x53, 0x53, 0x45, 0x78, 0x74, 0x65, 0x6e, 0x73, 0x69, 0x6f, 0x6e, 0x2d,
+    0x52, 0x00, 0x59, 0x00, 0x75, 0x00, 0x7a, 0x00, 0x75, 0x00, 0x4f, 0x00, 0x53, 0x00, 0x53, 0x00,
+    0x45, 0x00, 0x78, 0x00, 0x74, 0x00, 0x65, 0x00, 0x6e, 0x00, 0x73, 0x00, 0x69, 0x00, 0x6f, 0x00,
+    0x6e, 0x00, 0x52, 0x00, 0x65, 0x00, 0x67, 0x00, 0x75, 0x00, 0x6c, 0x00, 0x61, 0x00, 0x72, 0x00,
+    0x56, 0x00, 0x65, 0x00, 0x72, 0x00, 0x73, 0x00, 0x69, 0x00, 0x6f, 0x00, 0x6e, 0x00, 0x20, 0x00,
+    0x31, 0x00, 0x2e, 0x00, 0x30, 0x00, 0x30, 0x00, 0x30, 0x00, 0x3b, 0x00, 0x3b, 0x00, 0x59, 0x00,
+    0x75, 0x00, 0x7a, 0x00, 0x75, 0x00, 0x4f, 0x00, 0x53, 0x00, 0x53, 0x00, 0x45, 0x00, 0x78, 0x00,
+    0x74, 0x00, 0x65, 0x00, 0x6e, 0x00, 0x73, 0x00, 0x69, 0x00, 0x6f, 0x00, 0x6e, 0x00, 0x2d, 0x00,
+    0x52, 0x00, 0x3b, 0x00, 0x32, 0x00, 0x30, 0x00, 0x31, 0x00, 0x39, 0x00, 0x3b, 0x00, 0x46, 0x00,
+    0x4c, 0x00, 0x56, 0x00, 0x49, 0x00, 0x2d, 0x00, 0x36, 0x00, 0x31, 0x00, 0x34, 0x00, 0x59, 0x00,
+    0x75, 0x00, 0x7a, 0x00, 0x75, 0x00, 0x4f, 0x00, 0x53, 0x00, 0x53, 0x00, 0x45, 0x00, 0x78, 0x00,
+    0x74, 0x00, 0x65, 0x00, 0x6e, 0x00, 0x73, 0x00, 0x69, 0x00, 0x6f, 0x00, 0x6e, 0x00, 0x20, 0x00,
+    0x52, 0x00, 0x56, 0x00, 0x65, 0x00, 0x72, 0x00, 0x73, 0x00, 0x69, 0x00, 0x6f, 0x00, 0x6e, 0x00,
+    0x20, 0x00, 0x31, 0x00, 0x2e, 0x00, 0x30, 0x00, 0x30, 0x00, 0x30, 0x00, 0x59, 0x00, 0x75, 0x00,
+    0x7a, 0x00, 0x75, 0x00, 0x4f, 0x00, 0x53, 0x00, 0x53, 0x00, 0x45, 0x00, 0x78, 0x00, 0x74, 0x00,
+    0x65, 0x00, 0x6e, 0x00, 0x73, 0x00, 0x69, 0x00, 0x6f, 0x00, 0x6e, 0x00, 0x2d, 0x00, 0x52, 0x00,
+    0x52, 0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xff, 0x9c, 0x00, 0x32,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x00, 0x01, 0x02, 0x01, 0x03, 0x00, 0x03, 0x01, 0x04,
+    0x01, 0x05, 0x01, 0x06, 0x01, 0x07, 0x01, 0x08, 0x01, 0x09, 0x01, 0x0a, 0x01, 0x0b, 0x01, 0x0c,
+    0x01, 0x0d, 0x07, 0x75, 0x6e, 0x69, 0x30, 0x30, 0x30, 0x30, 0x07, 0x75, 0x6e, 0x69, 0x30, 0x30,
+    0x30, 0x44, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30, 0x45, 0x30, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30,
+    0x45, 0x31, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30, 0x45, 0x32, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30,
+    0x45, 0x33, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30, 0x45, 0x34, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30,
+    0x45, 0x35, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30, 0x45, 0x36, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30,
+    0x45, 0x37, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30, 0x45, 0x38, 0x07, 0x75, 0x6e, 0x69, 0x45, 0x30,
+    0x45, 0x39, 0x00, 0x00, 0x00, 0x01, 0x00, 0x01, 0xff, 0xff, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x01,
+    0x00, 0x00, 0x00, 0x00,
+}};
+
+} // namespace FileSys::SystemArchive::SharedFontData

src/core/file_sys/system_archive/data/font_nintendo_extended.h

@@ -0,0 +1,13 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+namespace FileSys::SystemArchive::SharedFontData {
+
+extern const std::array<unsigned char, 2932> FONT_NINTENDO_EXTENDED;
+
+} // namespace FileSys::SystemArchive::SharedFontData

src/core/file_sys/system_archive/data/font_standard.h

@@ -0,0 +1,13 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+namespace FileSys::SystemArchive::SharedFontData {
+
+extern const std::array<unsigned char, 217276> FONT_STANDARD;
+
+} // namespace FileSys::SystemArchive::SharedFontData

src/core/file_sys/system_archive/shared_font.cpp

@@ -0,0 +1,78 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/file_sys/system_archive/data/font_chinese_simplified.h"
+#include "core/file_sys/system_archive/data/font_chinese_traditional.h"
+#include "core/file_sys/system_archive/data/font_extended_chinese_simplified.h"
+#include "core/file_sys/system_archive/data/font_korean.h"
+#include "core/file_sys/system_archive/data/font_nintendo_extended.h"
+#include "core/file_sys/system_archive/data/font_standard.h"
+#include "core/file_sys/system_archive/shared_font.h"
+#include "core/file_sys/vfs_vector.h"
+#include "core/hle/service/ns/pl_u.h"
+
+namespace FileSys::SystemArchive {
+
+namespace {
+
+template <std::size_t Size>
+VirtualFile PackBFTTF(const std::array<u8, Size>& data, const std::string& name) {
+    std::vector<u32> vec(Size / sizeof(u32));
+    std::memcpy(vec.data(), data.data(), vec.size() * sizeof(u32));
+
+    std::vector<u8> bfttf(Size + sizeof(u64));
+
+    u64 offset = 0;
+    Service::NS::EncryptSharedFont(vec, bfttf, offset);
+    return std::make_shared<VectorVfsFile>(std::move(bfttf), name);
+}
+
+} // Anonymous namespace
+
+VirtualDir FontNintendoExtension() {
+    return std::make_shared<VectorVfsDirectory>(
+        std::vector<VirtualFile>{
+            PackBFTTF(SharedFontData::FONT_NINTENDO_EXTENDED, "nintendo_ext_003.bfttf"),
+            PackBFTTF(SharedFontData::FONT_NINTENDO_EXTENDED, "nintendo_ext2_003.bfttf"),
+        },
+        std::vector<VirtualDir>{});
+}
+
+VirtualDir FontStandard() {
+    return std::make_shared<VectorVfsDirectory>(
+        std::vector<VirtualFile>{
+            PackBFTTF(SharedFontData::FONT_STANDARD, "nintendo_udsg-r_std_003.bfttf"),
+        },
+        std::vector<VirtualDir>{});
+}
+
+VirtualDir FontKorean() {
+    return std::make_shared<VectorVfsDirectory>(
+        std::vector<VirtualFile>{
+            PackBFTTF(SharedFontData::FONT_KOREAN, "nintendo_udsg-r_ko_003.bfttf"),
+        },
+        std::vector<VirtualDir>{});
+}
+
+VirtualDir FontChineseTraditional() {
+    return std::make_shared<VectorVfsDirectory>(
+        std::vector<VirtualFile>{
+            PackBFTTF(SharedFontData::FONT_CHINESE_TRADITIONAL,
+                      "nintendo_udjxh-db_zh-tw_003.bfttf"),
+        },
+        std::vector<VirtualDir>{});
+}
+
+VirtualDir FontChineseSimple() {
+    return std::make_shared<VectorVfsDirectory>(
+        std::vector<VirtualFile>{
+            PackBFTTF(SharedFontData::FONT_CHINESE_SIMPLIFIED,
+                      "nintendo_udsg-r_org_zh-cn_003.bfttf"),
+            PackBFTTF(SharedFontData::FONT_EXTENDED_CHINESE_SIMPLIFIED,
+                      "nintendo_udsg-r_ext_zh-cn_003.bfttf"),
+        },
+        std::vector<VirtualDir>{});
+}
+
+} // namespace FileSys::SystemArchive

src/core/file_sys/system_archive/shared_font.h

@@ -0,0 +1,17 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/file_sys/vfs_types.h"
+
+namespace FileSys::SystemArchive {
+
+VirtualDir FontNintendoExtension();
+VirtualDir FontStandard();
+VirtualDir FontKorean();
+VirtualDir FontChineseTraditional();
+VirtualDir FontChineseSimple();
+
+} // namespace FileSys::SystemArchive

src/core/file_sys/system_archive/system_archive.cpp

@@ -6,6 +6,7 @@
 #include "core/file_sys/romfs.h"
 #include "core/file_sys/system_archive/mii_model.h"
 #include "core/file_sys/system_archive/ng_word.h"
+#include "core/file_sys/system_archive/shared_font.h"
 #include "core/file_sys/system_archive/system_archive.h"
 #include "core/file_sys/system_archive/system_version.h"
 
@@ -39,11 +40,11 @@ constexpr std::array<SystemArchiveDescriptor, SYSTEM_ARCHIVE_COUNT> SYSTEM_ARCHI
     {0x010000000000080D, "UrlBlackList", nullptr},
     {0x010000000000080E, "TimeZoneBinary", nullptr},
     {0x010000000000080F, "CertStoreCruiser", nullptr},
-    {0x0100000000000810, "FontNintendoExtension", nullptr},
-    {0x0100000000000811, "FontStandard", nullptr},
-    {0x0100000000000812, "FontKorean", nullptr},
-    {0x0100000000000813, "FontChineseTraditional", nullptr},
-    {0x0100000000000814, "FontChineseSimple", nullptr},
+    {0x0100000000000810, "FontNintendoExtension", &FontNintendoExtension},
+    {0x0100000000000811, "FontStandard", &FontStandard},
+    {0x0100000000000812, "FontKorean", &FontKorean},
+    {0x0100000000000813, "FontChineseTraditional", &FontChineseTraditional},
+    {0x0100000000000814, "FontChineseSimple", &FontChineseSimple},
     {0x0100000000000815, "FontBfcpx", nullptr},
     {0x0100000000000816, "SystemUpdate", nullptr},
     {0x0100000000000817, "0100000000000817", nullptr},

src/core/gdbstub/gdbstub.cpp

@@ -202,13 +202,11 @@ void RegisterModule(std::string name, VAddr beg, VAddr end, bool add_elf_ext) {
 }
 
 static Kernel::Thread* FindThreadById(s64 id) {
-    for (u32 core = 0; core < Core::NUM_CPU_CORES; core++) {
-        const auto& threads = Core::System::GetInstance().Scheduler(core).GetThreadList();
-        for (auto& thread : threads) {
-            if (thread->GetThreadID() == static_cast<u64>(id)) {
-                current_core = core;
-                return thread.get();
-            }
+    const auto& threads = Core::System::GetInstance().GlobalScheduler().GetThreadList();
+    for (auto& thread : threads) {
+        if (thread->GetThreadID() == static_cast<u64>(id)) {
+            current_core = thread->GetProcessorID();
+            return thread.get();
         }
     }
     return nullptr;
@@ -647,11 +645,9 @@ static void HandleQuery() {
         SendReply(buffer.c_str());
     } else if (strncmp(query, "fThreadInfo", strlen("fThreadInfo")) == 0) {
         std::string val = "m";
-        for (u32 core = 0; core < Core::NUM_CPU_CORES; core++) {
-            const auto& threads = Core::System::GetInstance().Scheduler(core).GetThreadList();
-            for (const auto& thread : threads) {
-                val += fmt::format("{:x},", thread->GetThreadID());
-            }
+        const auto& threads = Core::System::GetInstance().GlobalScheduler().GetThreadList();
+        for (const auto& thread : threads) {
+            val += fmt::format("{:x},", thread->GetThreadID());
         }
         val.pop_back();
         SendReply(val.c_str());
@@ -661,13 +657,11 @@ static void HandleQuery() {
         std::string buffer;
         buffer += "l<?xml version=\"1.0\"?>";
         buffer += "<threads>";
-        for (u32 core = 0; core < Core::NUM_CPU_CORES; core++) {
-            const auto& threads = Core::System::GetInstance().Scheduler(core).GetThreadList();
-            for (const auto& thread : threads) {
-                buffer +=
-                    fmt::format(R"*(<thread id="{:x}" core="{:d}" name="Thread {:x}"></thread>)*",
-                                thread->GetThreadID(), core, thread->GetThreadID());
-            }
+        const auto& threads = Core::System::GetInstance().GlobalScheduler().GetThreadList();
+        for (const auto& thread : threads) {
+            buffer +=
+                fmt::format(R"*(<thread id="{:x}" core="{:d}" name="Thread {:x}"></thread>)*",
+                            thread->GetThreadID(), thread->GetProcessorID(), thread->GetThreadID());
         }
         buffer += "</threads>";
         SendReply(buffer.c_str());

src/core/hle/kernel/address_arbiter.cpp

@@ -22,6 +22,7 @@ namespace Kernel {
 namespace {
 // Wake up num_to_wake (or all) threads in a vector.
 void WakeThreads(const std::vector<SharedPtr<Thread>>& waiting_threads, s32 num_to_wake) {
+    auto& system = Core::System::GetInstance();
     // Only process up to 'target' threads, unless 'target' is <= 0, in which case process
     // them all.
     std::size_t last = waiting_threads.size();
@@ -35,6 +36,7 @@ void WakeThreads(const std::vector<SharedPtr<Thread>>& waiting_threads, s32 num_
         waiting_threads[i]->SetWaitSynchronizationResult(RESULT_SUCCESS);
         waiting_threads[i]->SetArbiterWaitAddress(0);
         waiting_threads[i]->ResumeFromWait();
+        system.PrepareReschedule(waiting_threads[i]->GetProcessorID());
     }
 }
 } // Anonymous namespace
@@ -89,12 +91,20 @@ ResultCode AddressArbiter::ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr a
 
     // Determine the modified value depending on the waiting count.
     s32 updated_value;
-    if (waiting_threads.empty()) {
-        updated_value = value + 1;
-    } else if (num_to_wake <= 0 || waiting_threads.size() <= static_cast<u32>(num_to_wake)) {
-        updated_value = value - 1;
+    if (num_to_wake <= 0) {
+        if (waiting_threads.empty()) {
+            updated_value = value + 1;
+        } else {
+            updated_value = value - 1;
+        }
     } else {
-        updated_value = value;
+        if (waiting_threads.empty()) {
+            updated_value = value + 1;
+        } else if (waiting_threads.size() <= static_cast<u32>(num_to_wake)) {
+            updated_value = value - 1;
+        } else {
+            updated_value = value;
+        }
     }
 
     if (static_cast<s32>(Memory::Read32(address)) != value) {
@@ -169,30 +179,22 @@ ResultCode AddressArbiter::WaitForAddressImpl(VAddr address, s64 timeout) {
 
     current_thread->WakeAfterDelay(timeout);
 
-    system.CpuCore(current_thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(current_thread->GetProcessorID());
     return RESULT_TIMEOUT;
 }
 
 std::vector<SharedPtr<Thread>> AddressArbiter::GetThreadsWaitingOnAddress(VAddr address) const {
-    const auto RetrieveWaitingThreads = [this](std::size_t core_index,
-                                               std::vector<SharedPtr<Thread>>& waiting_threads,
-                                               VAddr arb_addr) {
-        const auto& scheduler = system.Scheduler(core_index);
-        const auto& thread_list = scheduler.GetThreadList();
-
-        for (const auto& thread : thread_list) {
-            if (thread->GetArbiterWaitAddress() == arb_addr) {
-                waiting_threads.push_back(thread);
-            }
-        }
-    };
 
     // Retrieve all threads that are waiting for this address.
     std::vector<SharedPtr<Thread>> threads;
-    RetrieveWaitingThreads(0, threads, address);
-    RetrieveWaitingThreads(1, threads, address);
-    RetrieveWaitingThreads(2, threads, address);
-    RetrieveWaitingThreads(3, threads, address);
+    const auto& scheduler = system.GlobalScheduler();
+    const auto& thread_list = scheduler.GetThreadList();
+
+    for (const auto& thread : thread_list) {
+        if (thread->GetArbiterWaitAddress() == address) {
+            threads.push_back(thread);
+        }
+    }
 
     // Sort them by priority, such that the highest priority ones come first.
     std::sort(threads.begin(), threads.end(),

src/core/hle/kernel/kernel.cpp

@@ -12,12 +12,15 @@
 
 #include "core/core.h"
 #include "core/core_timing.h"
+#include "core/core_timing_util.h"
 #include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/client_port.h"
+#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
+#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/lock.h"
 #include "core/hle/result.h"
@@ -58,12 +61,8 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] s64 cycles_
         if (thread->HasWakeupCallback()) {
             resume = thread->InvokeWakeupCallback(ThreadWakeupReason::Timeout, thread, nullptr, 0);
         }
-    }
-
-    if (thread->GetMutexWaitAddress() != 0 || thread->GetCondVarWaitAddress() != 0 ||
-        thread->GetWaitHandle() != 0) {
-        ASSERT(thread->GetStatus() == ThreadStatus::WaitMutex ||
-               thread->GetStatus() == ThreadStatus::WaitCondVar);
+    } else if (thread->GetStatus() == ThreadStatus::WaitMutex ||
+               thread->GetStatus() == ThreadStatus::WaitCondVar) {
         thread->SetMutexWaitAddress(0);
         thread->SetCondVarWaitAddress(0);
         thread->SetWaitHandle(0);
@@ -83,18 +82,23 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] s64 cycles_
     }
 
     if (resume) {
+        if (thread->GetStatus() == ThreadStatus::WaitCondVar ||
+            thread->GetStatus() == ThreadStatus::WaitArb) {
+            thread->SetWaitSynchronizationResult(RESULT_TIMEOUT);
+        }
         thread->ResumeFromWait();
     }
 }
 
 struct KernelCore::Impl {
-    explicit Impl(Core::System& system) : system{system} {}
+    explicit Impl(Core::System& system) : system{system}, global_scheduler{system} {}
 
     void Initialize(KernelCore& kernel) {
         Shutdown();
 
         InitializeSystemResourceLimit(kernel);
         InitializeThreads();
+        InitializePreemption();
     }
 
     void Shutdown() {
@@ -110,6 +114,9 @@ struct KernelCore::Impl {
 
         thread_wakeup_callback_handle_table.Clear();
         thread_wakeup_event_type = nullptr;
+        preemption_event = nullptr;
+
+        global_scheduler.Shutdown();
 
         named_ports.clear();
     }
@@ -132,6 +139,18 @@ struct KernelCore::Impl {
             system.CoreTiming().RegisterEvent("ThreadWakeupCallback", ThreadWakeupCallback);
     }
 
+    void InitializePreemption() {
+        preemption_event = system.CoreTiming().RegisterEvent(
+            "PreemptionCallback", [this](u64 userdata, s64 cycles_late) {
+                global_scheduler.PreemptThreads();
+                s64 time_interval = Core::Timing::msToCycles(std::chrono::milliseconds(10));
+                system.CoreTiming().ScheduleEvent(time_interval, preemption_event);
+            });
+
+        s64 time_interval = Core::Timing::msToCycles(std::chrono::milliseconds(10));
+        system.CoreTiming().ScheduleEvent(time_interval, preemption_event);
+    }
+
     std::atomic<u32> next_object_id{0};
     std::atomic<u64> next_kernel_process_id{Process::InitialKIPIDMin};
     std::atomic<u64> next_user_process_id{Process::ProcessIDMin};
@@ -140,10 +159,12 @@ struct KernelCore::Impl {
     // Lists all processes that exist in the current session.
     std::vector<SharedPtr<Process>> process_list;
     Process* current_process = nullptr;
+    Kernel::GlobalScheduler global_scheduler;
 
     SharedPtr<ResourceLimit> system_resource_limit;
 
     Core::Timing::EventType* thread_wakeup_event_type = nullptr;
+    Core::Timing::EventType* preemption_event = nullptr;
     // TODO(yuriks): This can be removed if Thread objects are explicitly pooled in the future,
     // allowing us to simply use a pool index or similar.
     Kernel::HandleTable thread_wakeup_callback_handle_table;
@@ -203,6 +224,14 @@ const std::vector<SharedPtr<Process>>& KernelCore::GetProcessList() const {
     return impl->process_list;
 }
 
+Kernel::GlobalScheduler& KernelCore::GlobalScheduler() {
+    return impl->global_scheduler;
+}
+
+const Kernel::GlobalScheduler& KernelCore::GlobalScheduler() const {
+    return impl->global_scheduler;
+}
+
 void KernelCore::AddNamedPort(std::string name, SharedPtr<ClientPort> port) {
     impl->named_ports.emplace(std::move(name), std::move(port));
 }

src/core/hle/kernel/kernel.h

@@ -21,6 +21,7 @@ namespace Kernel {
 
 class AddressArbiter;
 class ClientPort;
+class GlobalScheduler;
 class HandleTable;
 class Process;
 class ResourceLimit;
@@ -75,6 +76,12 @@ public:
     /// Retrieves the list of processes.
     const std::vector<SharedPtr<Process>>& GetProcessList() const;
 
+    /// Gets the sole instance of the global scheduler
+    Kernel::GlobalScheduler& GlobalScheduler();
+
+    /// Gets the sole instance of the global scheduler
+    const Kernel::GlobalScheduler& GlobalScheduler() const;
+
     /// Adds a port to the named port table
     void AddNamedPort(std::string name, SharedPtr<ClientPort> port);
 

src/core/hle/kernel/mutex.cpp

@@ -139,6 +139,9 @@ ResultCode Mutex::Release(VAddr address) {
     thread->SetCondVarWaitAddress(0);
     thread->SetMutexWaitAddress(0);
     thread->SetWaitHandle(0);
+    thread->SetWaitSynchronizationResult(RESULT_SUCCESS);
+
+    system.PrepareReschedule();
 
     return RESULT_SUCCESS;
 }

src/core/hle/kernel/process.cpp

@@ -213,10 +213,7 @@ void Process::PrepareForTermination() {
         }
     };
 
-    stop_threads(system.Scheduler(0).GetThreadList());
-    stop_threads(system.Scheduler(1).GetThreadList());
-    stop_threads(system.Scheduler(2).GetThreadList());
-    stop_threads(system.Scheduler(3).GetThreadList());
+    stop_threads(system.GlobalScheduler().GetThreadList());
 
     FreeTLSRegion(tls_region_address);
     tls_region_address = 0;

src/core/hle/kernel/scheduler.cpp

@@ -1,8 +1,13 @@
 // Copyright 2018 yuzu emulator team
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
+//
+// SelectThreads, Yield functions originally by TuxSH.
+// licensed under GPLv2 or later under exception provided by the author.
 
 #include <algorithm>
+#include <set>
+#include <unordered_set>
 #include <utility>
 
 #include "common/assert.h"
@@ -17,56 +22,374 @@
 
 namespace Kernel {
 
-std::mutex Scheduler::scheduler_mutex;
+GlobalScheduler::GlobalScheduler(Core::System& system) : system{system} {}
 
-Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core)
-    : cpu_core{cpu_core}, system{system} {}
+GlobalScheduler::~GlobalScheduler() = default;
 
-Scheduler::~Scheduler() {
-    for (auto& thread : thread_list) {
-        thread->Stop();
+void GlobalScheduler::AddThread(SharedPtr<Thread> thread) {
+    thread_list.push_back(std::move(thread));
+}
+
+void GlobalScheduler::RemoveThread(const Thread* thread) {
+    thread_list.erase(std::remove(thread_list.begin(), thread_list.end(), thread),
+                      thread_list.end());
+}
+
+void GlobalScheduler::UnloadThread(s32 core) {
+    Scheduler& sched = system.Scheduler(core);
+    sched.UnloadThread();
+}
+
+void GlobalScheduler::SelectThread(u32 core) {
+    const auto update_thread = [](Thread* thread, Scheduler& sched) {
+        if (thread != sched.selected_thread) {
+            if (thread == nullptr) {
+                ++sched.idle_selection_count;
+            }
+            sched.selected_thread = thread;
+        }
+        sched.is_context_switch_pending = sched.selected_thread != sched.current_thread;
+        std::atomic_thread_fence(std::memory_order_seq_cst);
+    };
+    Scheduler& sched = system.Scheduler(core);
+    Thread* current_thread = nullptr;
+    // Step 1: Get top thread in schedule queue.
+    current_thread = scheduled_queue[core].empty() ? nullptr : scheduled_queue[core].front();
+    if (current_thread) {
+        update_thread(current_thread, sched);
+        return;
+    }
+    // Step 2: Try selecting a suggested thread.
+    Thread* winner = nullptr;
+    std::set<s32> sug_cores;
+    for (auto thread : suggested_queue[core]) {
+        s32 this_core = thread->GetProcessorID();
+        Thread* thread_on_core = nullptr;
+        if (this_core >= 0) {
+            thread_on_core = scheduled_queue[this_core].front();
+        }
+        if (this_core < 0 || thread != thread_on_core) {
+            winner = thread;
+            break;
+        }
+        sug_cores.insert(this_core);
+    }
+    // if we got a suggested thread, select it, else do a second pass.
+    if (winner && winner->GetPriority() > 2) {
+        if (winner->IsRunning()) {
+            UnloadThread(winner->GetProcessorID());
+        }
+        TransferToCore(winner->GetPriority(), core, winner);
+        update_thread(winner, sched);
+        return;
+    }
+    // Step 3: Select a suggested thread from another core
+    for (auto& src_core : sug_cores) {
+        auto it = scheduled_queue[src_core].begin();
+        it++;
+        if (it != scheduled_queue[src_core].end()) {
+            Thread* thread_on_core = scheduled_queue[src_core].front();
+            Thread* to_change = *it;
+            if (thread_on_core->IsRunning() || to_change->IsRunning()) {
+                UnloadThread(src_core);
+            }
+            TransferToCore(thread_on_core->GetPriority(), core, thread_on_core);
+            current_thread = thread_on_core;
+            break;
+        }
+    }
+    update_thread(current_thread, sched);
+}
+
+bool GlobalScheduler::YieldThread(Thread* yielding_thread) {
+    // Note: caller should use critical section, etc.
+    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
+    const u32 priority = yielding_thread->GetPriority();
+
+    // Yield the thread
+    const Thread* const winner = scheduled_queue[core_id].front(priority);
+    ASSERT_MSG(yielding_thread == winner, "Thread yielding without being in front");
+    scheduled_queue[core_id].yield(priority);
+
+    return AskForReselectionOrMarkRedundant(yielding_thread, winner);
+}
+
+bool GlobalScheduler::YieldThreadAndBalanceLoad(Thread* yielding_thread) {
+    // Note: caller should check if !thread.IsSchedulerOperationRedundant and use critical section,
+    // etc.
+    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
+    const u32 priority = yielding_thread->GetPriority();
+
+    // Yield the thread
+    ASSERT_MSG(yielding_thread == scheduled_queue[core_id].front(priority),
+               "Thread yielding without being in front");
+    scheduled_queue[core_id].yield(priority);
+
+    std::array<Thread*, NUM_CPU_CORES> current_threads;
+    for (u32 i = 0; i < NUM_CPU_CORES; i++) {
+        current_threads[i] = scheduled_queue[i].empty() ? nullptr : scheduled_queue[i].front();
+    }
+
+    Thread* next_thread = scheduled_queue[core_id].front(priority);
+    Thread* winner = nullptr;
+    for (auto& thread : suggested_queue[core_id]) {
+        const s32 source_core = thread->GetProcessorID();
+        if (source_core >= 0) {
+            if (current_threads[source_core] != nullptr) {
+                if (thread == current_threads[source_core] ||
+                    current_threads[source_core]->GetPriority() < min_regular_priority) {
+                    continue;
+                }
+            }
+        }
+        if (next_thread->GetLastRunningTicks() >= thread->GetLastRunningTicks() ||
+            next_thread->GetPriority() < thread->GetPriority()) {
+            if (thread->GetPriority() <= priority) {
+                winner = thread;
+                break;
+            }
+        }
+    }
+
+    if (winner != nullptr) {
+        if (winner != yielding_thread) {
+            if (winner->IsRunning()) {
+                UnloadThread(winner->GetProcessorID());
+            }
+            TransferToCore(winner->GetPriority(), core_id, winner);
+        }
+    } else {
+        winner = next_thread;
+    }
+
+    return AskForReselectionOrMarkRedundant(yielding_thread, winner);
+}
+
+bool GlobalScheduler::YieldThreadAndWaitForLoadBalancing(Thread* yielding_thread) {
+    // Note: caller should check if !thread.IsSchedulerOperationRedundant and use critical section,
+    // etc.
+    Thread* winner = nullptr;
+    const u32 core_id = static_cast<u32>(yielding_thread->GetProcessorID());
+
+    // Remove the thread from its scheduled mlq, put it on the corresponding "suggested" one instead
+    TransferToCore(yielding_thread->GetPriority(), -1, yielding_thread);
+
+    // If the core is idle, perform load balancing, excluding the threads that have just used this
+    // function...
+    if (scheduled_queue[core_id].empty()) {
+        // Here, "current_threads" is calculated after the ""yield"", unlike yield -1
+        std::array<Thread*, NUM_CPU_CORES> current_threads;
+        for (u32 i = 0; i < NUM_CPU_CORES; i++) {
+            current_threads[i] = scheduled_queue[i].empty() ? nullptr : scheduled_queue[i].front();
+        }
+        for (auto& thread : suggested_queue[core_id]) {
+            const s32 source_core = thread->GetProcessorID();
+            if (source_core < 0 || thread == current_threads[source_core]) {
+                continue;
+            }
+            if (current_threads[source_core] == nullptr ||
+                current_threads[source_core]->GetPriority() >= min_regular_priority) {
+                winner = thread;
+            }
+            break;
+        }
+        if (winner != nullptr) {
+            if (winner != yielding_thread) {
+                if (winner->IsRunning()) {
+                    UnloadThread(winner->GetProcessorID());
+                }
+                TransferToCore(winner->GetPriority(), core_id, winner);
+            }
+        } else {
+            winner = yielding_thread;
+        }
+    }
+
+    return AskForReselectionOrMarkRedundant(yielding_thread, winner);
+}
+
+void GlobalScheduler::PreemptThreads() {
+    for (std::size_t core_id = 0; core_id < NUM_CPU_CORES; core_id++) {
+        const u32 priority = preemption_priorities[core_id];
+
+        if (scheduled_queue[core_id].size(priority) > 0) {
+            scheduled_queue[core_id].front(priority)->IncrementYieldCount();
+            scheduled_queue[core_id].yield(priority);
+            if (scheduled_queue[core_id].size(priority) > 1) {
+                scheduled_queue[core_id].front(priority)->IncrementYieldCount();
+            }
+        }
+
+        Thread* current_thread =
+            scheduled_queue[core_id].empty() ? nullptr : scheduled_queue[core_id].front();
+        Thread* winner = nullptr;
+        for (auto& thread : suggested_queue[core_id]) {
+            const s32 source_core = thread->GetProcessorID();
+            if (thread->GetPriority() != priority) {
+                continue;
+            }
+            if (source_core >= 0) {
+                Thread* next_thread = scheduled_queue[source_core].empty()
+                                          ? nullptr
+                                          : scheduled_queue[source_core].front();
+                if (next_thread != nullptr && next_thread->GetPriority() < 2) {
+                    break;
+                }
+                if (next_thread == thread) {
+                    continue;
+                }
+            }
+            if (current_thread != nullptr &&
+                current_thread->GetLastRunningTicks() >= thread->GetLastRunningTicks()) {
+                winner = thread;
+                break;
+            }
+        }
+
+        if (winner != nullptr) {
+            if (winner->IsRunning()) {
+                UnloadThread(winner->GetProcessorID());
+            }
+            TransferToCore(winner->GetPriority(), s32(core_id), winner);
+            current_thread =
+                winner->GetPriority() <= current_thread->GetPriority() ? winner : current_thread;
+        }
+
+        if (current_thread != nullptr && current_thread->GetPriority() > priority) {
+            for (auto& thread : suggested_queue[core_id]) {
+                const s32 source_core = thread->GetProcessorID();
+                if (thread->GetPriority() < priority) {
+                    continue;
+                }
+                if (source_core >= 0) {
+                    Thread* next_thread = scheduled_queue[source_core].empty()
+                                              ? nullptr
+                                              : scheduled_queue[source_core].front();
+                    if (next_thread != nullptr && next_thread->GetPriority() < 2) {
+                        break;
+                    }
+                    if (next_thread == thread) {
+                        continue;
+                    }
+                }
+                if (current_thread != nullptr &&
+                    current_thread->GetLastRunningTicks() >= thread->GetLastRunningTicks()) {
+                    winner = thread;
+                    break;
+                }
+            }
+
+            if (winner != nullptr) {
+                if (winner->IsRunning()) {
+                    UnloadThread(winner->GetProcessorID());
+                }
+                TransferToCore(winner->GetPriority(), s32(core_id), winner);
+                current_thread = winner;
+            }
+        }
+
+        is_reselection_pending.store(true, std::memory_order_release);
     }
 }
 
+void GlobalScheduler::Suggest(u32 priority, u32 core, Thread* thread) {
+    suggested_queue[core].add(thread, priority);
+}
+
+void GlobalScheduler::Unsuggest(u32 priority, u32 core, Thread* thread) {
+    suggested_queue[core].remove(thread, priority);
+}
+
+void GlobalScheduler::Schedule(u32 priority, u32 core, Thread* thread) {
+    ASSERT_MSG(thread->GetProcessorID() == s32(core), "Thread must be assigned to this core.");
+    scheduled_queue[core].add(thread, priority);
+}
+
+void GlobalScheduler::SchedulePrepend(u32 priority, u32 core, Thread* thread) {
+    ASSERT_MSG(thread->GetProcessorID() == s32(core), "Thread must be assigned to this core.");
+    scheduled_queue[core].add(thread, priority, false);
+}
+
+void GlobalScheduler::Reschedule(u32 priority, u32 core, Thread* thread) {
+    scheduled_queue[core].remove(thread, priority);
+    scheduled_queue[core].add(thread, priority);
+}
+
+void GlobalScheduler::Unschedule(u32 priority, u32 core, Thread* thread) {
+    scheduled_queue[core].remove(thread, priority);
+}
+
+void GlobalScheduler::TransferToCore(u32 priority, s32 destination_core, Thread* thread) {
+    const bool schedulable = thread->GetPriority() < THREADPRIO_COUNT;
+    const s32 source_core = thread->GetProcessorID();
+    if (source_core == destination_core || !schedulable) {
+        return;
+    }
+    thread->SetProcessorID(destination_core);
+    if (source_core >= 0) {
+        Unschedule(priority, source_core, thread);
+    }
+    if (destination_core >= 0) {
+        Unsuggest(priority, destination_core, thread);
+        Schedule(priority, destination_core, thread);
+    }
+    if (source_core >= 0) {
+        Suggest(priority, source_core, thread);
+    }
+}
+
+bool GlobalScheduler::AskForReselectionOrMarkRedundant(Thread* current_thread,
+                                                       const Thread* winner) {
+    if (current_thread == winner) {
+        current_thread->IncrementYieldCount();
+        return true;
+    } else {
+        is_reselection_pending.store(true, std::memory_order_release);
+        return false;
+    }
+}
+
+void GlobalScheduler::Shutdown() {
+    for (std::size_t core = 0; core < NUM_CPU_CORES; core++) {
+        scheduled_queue[core].clear();
+        suggested_queue[core].clear();
+    }
+    thread_list.clear();
+}
+
+Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, u32 core_id)
+    : system(system), cpu_core(cpu_core), core_id(core_id) {}
+
+Scheduler::~Scheduler() = default;
+
 bool Scheduler::HaveReadyThreads() const {
-    std::lock_guard lock{scheduler_mutex};
-    return !ready_queue.empty();
+    return system.GlobalScheduler().HaveReadyThreads(core_id);
 }
 
 Thread* Scheduler::GetCurrentThread() const {
     return current_thread.get();
 }
 
+Thread* Scheduler::GetSelectedThread() const {
+    return selected_thread.get();
+}
+
+void Scheduler::SelectThreads() {
+    system.GlobalScheduler().SelectThread(core_id);
+}
+
 u64 Scheduler::GetLastContextSwitchTicks() const {
     return last_context_switch_time;
 }
 
-Thread* Scheduler::PopNextReadyThread() {
-    Thread* next = nullptr;
-    Thread* thread = GetCurrentThread();
-
-    if (thread && thread->GetStatus() == ThreadStatus::Running) {
-        if (ready_queue.empty()) {
-            return thread;
-        }
-        // We have to do better than the current thread.
-        // This call returns null when that's not possible.
-        next = ready_queue.front();
-        if (next == nullptr || next->GetPriority() >= thread->GetPriority()) {
-            next = thread;
-        }
-    } else {
-        if (ready_queue.empty()) {
-            return nullptr;
-        }
-        next = ready_queue.front();
+void Scheduler::TryDoContextSwitch() {
+    if (is_context_switch_pending) {
+        SwitchContext();
     }
-
-    return next;
 }
 
-void Scheduler::SwitchContext(Thread* new_thread) {
-    Thread* previous_thread = GetCurrentThread();
+void Scheduler::UnloadThread() {
+    Thread* const previous_thread = GetCurrentThread();
     Process* const previous_process = system.Kernel().CurrentProcess();
 
     UpdateLastContextSwitchTime(previous_thread, previous_process);
@@ -80,23 +403,52 @@ void Scheduler::SwitchContext(Thread* new_thread) {
         if (previous_thread->GetStatus() == ThreadStatus::Running) {
             // This is only the case when a reschedule is triggered without the current thread
             // yielding execution (i.e. an event triggered, system core time-sliced, etc)
-            ready_queue.add(previous_thread, previous_thread->GetPriority(), false);
             previous_thread->SetStatus(ThreadStatus::Ready);
         }
+        previous_thread->SetIsRunning(false);
+    }
+    current_thread = nullptr;
+}
+
+void Scheduler::SwitchContext() {
+    Thread* const previous_thread = GetCurrentThread();
+    Thread* const new_thread = GetSelectedThread();
+
+    is_context_switch_pending = false;
+    if (new_thread == previous_thread) {
+        return;
+    }
+
+    Process* const previous_process = system.Kernel().CurrentProcess();
+
+    UpdateLastContextSwitchTime(previous_thread, previous_process);
+
+    // Save context for previous thread
+    if (previous_thread) {
+        cpu_core.SaveContext(previous_thread->GetContext());
+        // Save the TPIDR_EL0 system register in case it was modified.
+        previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+
+        if (previous_thread->GetStatus() == ThreadStatus::Running) {
+            // This is only the case when a reschedule is triggered without the current thread
+            // yielding execution (i.e. an event triggered, system core time-sliced, etc)
+            previous_thread->SetStatus(ThreadStatus::Ready);
+        }
+        previous_thread->SetIsRunning(false);
     }
 
     // Load context of new thread
     if (new_thread) {
+        ASSERT_MSG(new_thread->GetProcessorID() == s32(this->core_id),
+                   "Thread must be assigned to this core.");
         ASSERT_MSG(new_thread->GetStatus() == ThreadStatus::Ready,
                    "Thread must be ready to become running.");
 
         // Cancel any outstanding wakeup events for this thread
         new_thread->CancelWakeupTimer();
-
         current_thread = new_thread;
-
-        ready_queue.remove(new_thread, new_thread->GetPriority());
         new_thread->SetStatus(ThreadStatus::Running);
+        new_thread->SetIsRunning(true);
 
         auto* const thread_owner_process = current_thread->GetOwnerProcess();
         if (previous_process != thread_owner_process) {
@@ -130,124 +482,9 @@ void Scheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
     last_context_switch_time = most_recent_switch_ticks;
 }
 
-void Scheduler::Reschedule() {
-    std::lock_guard lock{scheduler_mutex};
-
-    Thread* cur = GetCurrentThread();
-    Thread* next = PopNextReadyThread();
-
-    if (cur && next) {
-        LOG_TRACE(Kernel, "context switch {} -> {}", cur->GetObjectId(), next->GetObjectId());
-    } else if (cur) {
-        LOG_TRACE(Kernel, "context switch {} -> idle", cur->GetObjectId());
-    } else if (next) {
-        LOG_TRACE(Kernel, "context switch idle -> {}", next->GetObjectId());
-    }
-
-    SwitchContext(next);
-}
-
-void Scheduler::AddThread(SharedPtr<Thread> thread) {
-    std::lock_guard lock{scheduler_mutex};
-
-    thread_list.push_back(std::move(thread));
-}
-
-void Scheduler::RemoveThread(Thread* thread) {
-    std::lock_guard lock{scheduler_mutex};
-
-    thread_list.erase(std::remove(thread_list.begin(), thread_list.end(), thread),
-                      thread_list.end());
-}
-
-void Scheduler::ScheduleThread(Thread* thread, u32 priority) {
-    std::lock_guard lock{scheduler_mutex};
-
-    ASSERT(thread->GetStatus() == ThreadStatus::Ready);
-    ready_queue.add(thread, priority);
-}
-
-void Scheduler::UnscheduleThread(Thread* thread, u32 priority) {
-    std::lock_guard lock{scheduler_mutex};
-
-    ASSERT(thread->GetStatus() == ThreadStatus::Ready);
-    ready_queue.remove(thread, priority);
-}
-
-void Scheduler::SetThreadPriority(Thread* thread, u32 priority) {
-    std::lock_guard lock{scheduler_mutex};
-    if (thread->GetPriority() == priority) {
-        return;
-    }
-
-    // If thread was ready, adjust queues
-    if (thread->GetStatus() == ThreadStatus::Ready)
-        ready_queue.adjust(thread, thread->GetPriority(), priority);
-}
-
-Thread* Scheduler::GetNextSuggestedThread(u32 core, u32 maximum_priority) const {
-    std::lock_guard lock{scheduler_mutex};
-
-    const u32 mask = 1U << core;
-    for (auto* thread : ready_queue) {
-        if ((thread->GetAffinityMask() & mask) != 0 && thread->GetPriority() < maximum_priority) {
-            return thread;
-        }
-    }
-    return nullptr;
-}
-
-void Scheduler::YieldWithoutLoadBalancing(Thread* thread) {
-    ASSERT(thread != nullptr);
-    // Avoid yielding if the thread isn't even running.
-    ASSERT(thread->GetStatus() == ThreadStatus::Running);
-
-    // Sanity check that the priority is valid
-    ASSERT(thread->GetPriority() < THREADPRIO_COUNT);
-
-    // Yield this thread -- sleep for zero time and force reschedule to different thread
-    GetCurrentThread()->Sleep(0);
-}
-
-void Scheduler::YieldWithLoadBalancing(Thread* thread) {
-    ASSERT(thread != nullptr);
-    const auto priority = thread->GetPriority();
-    const auto core = static_cast<u32>(thread->GetProcessorID());
-
-    // Avoid yielding if the thread isn't even running.
-    ASSERT(thread->GetStatus() == ThreadStatus::Running);
-
-    // Sanity check that the priority is valid
-    ASSERT(priority < THREADPRIO_COUNT);
-
-    // Sleep for zero time to be able to force reschedule to different thread
-    GetCurrentThread()->Sleep(0);
-
-    Thread* suggested_thread = nullptr;
-
-    // Search through all of the cpu cores (except this one) for a suggested thread.
-    // Take the first non-nullptr one
-    for (unsigned cur_core = 0; cur_core < Core::NUM_CPU_CORES; ++cur_core) {
-        const auto res =
-            system.CpuCore(cur_core).Scheduler().GetNextSuggestedThread(core, priority);
-
-        // If scheduler provides a suggested thread
-        if (res != nullptr) {
-            // And its better than the current suggested thread (or is the first valid one)
-            if (suggested_thread == nullptr ||
-                suggested_thread->GetPriority() > res->GetPriority()) {
-                suggested_thread = res;
-            }
-        }
-    }
-
-    // If a suggested thread was found, queue that for this core
-    if (suggested_thread != nullptr)
-        suggested_thread->ChangeCore(core, suggested_thread->GetAffinityMask());
-}
-
-void Scheduler::YieldAndWaitForLoadBalancing(Thread* thread) {
-    UNIMPLEMENTED_MSG("Wait for load balancing thread yield type is not implemented!");
+void Scheduler::Shutdown() {
+    current_thread = nullptr;
+    selected_thread = nullptr;
 }
 
 } // namespace Kernel

src/core/hle/kernel/scheduler.h

@@ -20,124 +20,185 @@ namespace Kernel {
 
 class Process;
 
-class Scheduler final {
+class GlobalScheduler final {
 public:
-    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core);
-    ~Scheduler();
+    static constexpr u32 NUM_CPU_CORES = 4;
 
-    /// Returns whether there are any threads that are ready to run.
-    bool HaveReadyThreads() const;
-
-    /// Reschedules to the next available thread (call after current thread is suspended)
-    void Reschedule();
-
-    /// Gets the current running thread
-    Thread* GetCurrentThread() const;
-
-    /// Gets the timestamp for the last context switch in ticks.
-    u64 GetLastContextSwitchTicks() const;
+    explicit GlobalScheduler(Core::System& system);
+    ~GlobalScheduler();
 
     /// Adds a new thread to the scheduler
     void AddThread(SharedPtr<Thread> thread);
 
     /// Removes a thread from the scheduler
-    void RemoveThread(Thread* thread);
-
-    /// Schedules a thread that has become "ready"
-    void ScheduleThread(Thread* thread, u32 priority);
-
-    /// Unschedules a thread that was already scheduled
-    void UnscheduleThread(Thread* thread, u32 priority);
-
-    /// Sets the priority of a thread in the scheduler
-    void SetThreadPriority(Thread* thread, u32 priority);
-
-    /// Gets the next suggested thread for load balancing
-    Thread* GetNextSuggestedThread(u32 core, u32 minimum_priority) const;
-
-    /**
-     * YieldWithoutLoadBalancing -- analogous to normal yield on a system
-     * Moves the thread to the end of the ready queue for its priority, and then reschedules the
-     * system to the new head of the queue.
-     *
-     * Example (Single Core -- but can be extrapolated to multi):
-     * ready_queue[prio=0]: ThreadA, ThreadB, ThreadC (->exec order->)
-     * Currently Running: ThreadR
-     *
-     * ThreadR calls YieldWithoutLoadBalancing
-     *
-     * ThreadR is moved to the end of ready_queue[prio=0]:
-     * ready_queue[prio=0]: ThreadA, ThreadB, ThreadC, ThreadR (->exec order->)
-     * Currently Running: Nothing
-     *
-     * System is rescheduled (ThreadA is popped off of queue):
-     * ready_queue[prio=0]: ThreadB, ThreadC, ThreadR (->exec order->)
-     * Currently Running: ThreadA
-     *
-     * If the queue is empty at time of call, no yielding occurs. This does not cross between cores
-     * or priorities at all.
-     */
-    void YieldWithoutLoadBalancing(Thread* thread);
-
-    /**
-     * YieldWithLoadBalancing -- yield but with better selection of the new running thread
-     * Moves the current thread to the end of the ready queue for its priority, then selects a
-     * 'suggested thread' (a thread on a different core that could run on this core) from the
-     * scheduler, changes its core, and reschedules the current core to that thread.
-     *
-     * Example (Dual Core -- can be extrapolated to Quad Core, this is just normal yield if it were
-     * single core):
-     * ready_queue[core=0][prio=0]: ThreadA, ThreadB (affinities not pictured as irrelevant
-     * ready_queue[core=1][prio=0]: ThreadC[affinity=both], ThreadD[affinity=core1only]
-     * Currently Running: ThreadQ on Core 0 || ThreadP on Core 1
-     *
-     * ThreadQ calls YieldWithLoadBalancing
-     *
-     * ThreadQ is moved to the end of ready_queue[core=0][prio=0]:
-     * ready_queue[core=0][prio=0]: ThreadA, ThreadB
-     * ready_queue[core=1][prio=0]: ThreadC[affinity=both], ThreadD[affinity=core1only]
-     * Currently Running: ThreadQ on Core 0 || ThreadP on Core 1
-     *
-     * A list of suggested threads for each core is compiled
-     * Suggested Threads: {ThreadC on Core 1}
-     * If this were quad core (as the switch is), there could be between 0 and 3 threads in this
-     * list. If there are more than one, the thread is selected by highest prio.
-     *
-     * ThreadC is core changed to Core 0:
-     * ready_queue[core=0][prio=0]: ThreadC, ThreadA, ThreadB, ThreadQ
-     * ready_queue[core=1][prio=0]: ThreadD
-     * Currently Running: None on Core 0 || ThreadP on Core 1
-     *
-     * System is rescheduled (ThreadC is popped off of queue):
-     * ready_queue[core=0][prio=0]: ThreadA, ThreadB, ThreadQ
-     * ready_queue[core=1][prio=0]: ThreadD
-     * Currently Running: ThreadC on Core 0 || ThreadP on Core 1
-     *
-     * If no suggested threads can be found this will behave just as normal yield. If there are
-     * multiple candidates for the suggested thread on a core, the highest prio is taken.
-     */
-    void YieldWithLoadBalancing(Thread* thread);
-
-    /// Currently unknown -- asserts as unimplemented on call
-    void YieldAndWaitForLoadBalancing(Thread* thread);
+    void RemoveThread(const Thread* thread);
 
     /// Returns a list of all threads managed by the scheduler
     const std::vector<SharedPtr<Thread>>& GetThreadList() const {
         return thread_list;
     }
 
-private:
     /**
-     * Pops and returns the next thread from the thread queue
-     * @return A pointer to the next ready thread
+     * Add a thread to the suggested queue of a cpu core. Suggested threads may be
+     * picked if no thread is scheduled to run on the core.
      */
-    Thread* PopNextReadyThread();
+    void Suggest(u32 priority, u32 core, Thread* thread);
 
     /**
-     * Switches the CPU's active thread context to that of the specified thread
-     * @param new_thread The thread to switch to
+     * Remove a thread to the suggested queue of a cpu core. Suggested threads may be
+     * picked if no thread is scheduled to run on the core.
      */
-    void SwitchContext(Thread* new_thread);
+    void Unsuggest(u32 priority, u32 core, Thread* thread);
+
+    /**
+     * Add a thread to the scheduling queue of a cpu core. The thread is added at the
+     * back the queue in its priority level.
+     */
+    void Schedule(u32 priority, u32 core, Thread* thread);
+
+    /**
+     * Add a thread to the scheduling queue of a cpu core. The thread is added at the
+     * front the queue in its priority level.
+     */
+    void SchedulePrepend(u32 priority, u32 core, Thread* thread);
+
+    /// Reschedule an already scheduled thread based on a new priority
+    void Reschedule(u32 priority, u32 core, Thread* thread);
+
+    /// Unschedules a thread.
+    void Unschedule(u32 priority, u32 core, Thread* thread);
+
+    /**
+     * Transfers a thread into an specific core. If the destination_core is -1
+     * it will be unscheduled from its source code and added into its suggested
+     * queue.
+     */
+    void TransferToCore(u32 priority, s32 destination_core, Thread* thread);
+
+    /// Selects a core and forces it to unload its current thread's context
+    void UnloadThread(s32 core);
+
+    /**
+     * Takes care of selecting the new scheduled thread in three steps:
+     *
+     * 1. First a thread is selected from the top of the priority queue. If no thread
+     *    is obtained then we move to step two, else we are done.
+     *
+     * 2. Second we try to get a suggested thread that's not assigned to any core or
+     *    that is not the top thread in that core.
+     *
+     * 3. Third is no suggested thread is found, we do a second pass and pick a running
+     *    thread in another core and swap it with its current thread.
+     */
+    void SelectThread(u32 core);
+
+    bool HaveReadyThreads(u32 core_id) const {
+        return !scheduled_queue[core_id].empty();
+    }
+
+    /**
+     * Takes a thread and moves it to the back of the it's priority list.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    bool YieldThread(Thread* thread);
+
+    /**
+     * Takes a thread and moves it to the back of the it's priority list.
+     * Afterwards, tries to pick a suggested thread from the suggested queue that has worse time or
+     * a better priority than the next thread in the core.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    bool YieldThreadAndBalanceLoad(Thread* thread);
+
+    /**
+     * Takes a thread and moves it out of the scheduling queue.
+     * and into the suggested queue. If no thread can be scheduled afterwards in that core,
+     * a suggested thread is obtained instead.
+     *
+     * @note This operation can be redundant and no scheduling is changed if marked as so.
+     */
+    bool YieldThreadAndWaitForLoadBalancing(Thread* thread);
+
+    /**
+     * Rotates the scheduling queues of threads at a preemption priority and then does
+     * some core rebalancing. Preemption priorities can be found in the array
+     * 'preemption_priorities'.
+     *
+     * @note This operation happens every 10ms.
+     */
+    void PreemptThreads();
+
+    u32 CpuCoresCount() const {
+        return NUM_CPU_CORES;
+    }
+
+    void SetReselectionPending() {
+        is_reselection_pending.store(true, std::memory_order_release);
+    }
+
+    bool IsReselectionPending() const {
+        return is_reselection_pending.load(std::memory_order_acquire);
+    }
+
+    void Shutdown();
+
+private:
+    bool AskForReselectionOrMarkRedundant(Thread* current_thread, const Thread* winner);
+
+    static constexpr u32 min_regular_priority = 2;
+    std::array<Common::MultiLevelQueue<Thread*, THREADPRIO_COUNT>, NUM_CPU_CORES> scheduled_queue;
+    std::array<Common::MultiLevelQueue<Thread*, THREADPRIO_COUNT>, NUM_CPU_CORES> suggested_queue;
+    std::atomic<bool> is_reselection_pending{false};
+
+    // The priority levels at which the global scheduler preempts threads every 10 ms. They are
+    // ordered from Core 0 to Core 3.
+    std::array<u32, NUM_CPU_CORES> preemption_priorities = {59, 59, 59, 62};
+
+    /// Lists all thread ids that aren't deleted/etc.
+    std::vector<SharedPtr<Thread>> thread_list;
+    Core::System& system;
+};
+
+class Scheduler final {
+public:
+    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, u32 core_id);
+    ~Scheduler();
+
+    /// Returns whether there are any threads that are ready to run.
+    bool HaveReadyThreads() const;
+
+    /// Reschedules to the next available thread (call after current thread is suspended)
+    void TryDoContextSwitch();
+
+    /// Unloads currently running thread
+    void UnloadThread();
+
+    /// Select the threads in top of the scheduling multilist.
+    void SelectThreads();
+
+    /// Gets the current running thread
+    Thread* GetCurrentThread() const;
+
+    /// Gets the currently selected thread from the top of the multilevel queue
+    Thread* GetSelectedThread() const;
+
+    /// Gets the timestamp for the last context switch in ticks.
+    u64 GetLastContextSwitchTicks() const;
+
+    bool ContextSwitchPending() const {
+        return is_context_switch_pending;
+    }
+
+    /// Shutdowns the scheduler.
+    void Shutdown();
+
+private:
+    friend class GlobalScheduler;
+
+    /// Switches the CPU's active thread context to that of the specified thread
+    void SwitchContext();
 
     /**
      * Called on every context switch to update the internal timestamp
@@ -152,19 +213,16 @@ private:
      */
     void UpdateLastContextSwitchTime(Thread* thread, Process* process);
 
-    /// Lists all thread ids that aren't deleted/etc.
-    std::vector<SharedPtr<Thread>> thread_list;
-
-    /// Lists only ready thread ids.
-    Common::MultiLevelQueue<Thread*, THREADPRIO_LOWEST + 1> ready_queue;
-
     SharedPtr<Thread> current_thread = nullptr;
-
-    Core::ARM_Interface& cpu_core;
-    u64 last_context_switch_time = 0;
+    SharedPtr<Thread> selected_thread = nullptr;
 
     Core::System& system;
-    static std::mutex scheduler_mutex;
+    Core::ARM_Interface& cpu_core;
+    u64 last_context_switch_time = 0;
+    u64 idle_selection_count = 0;
+    const u32 core_id;
+
+    bool is_context_switch_pending = false;
 };
 
 } // namespace Kernel

src/core/hle/kernel/svc.cpp

@@ -516,7 +516,7 @@ static ResultCode WaitSynchronization(Core::System& system, Handle* index, VAddr
     thread->WakeAfterDelay(nano_seconds);
     thread->SetWakeupCallback(DefaultThreadWakeupCallback);
 
-    system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(thread->GetProcessorID());
 
     return RESULT_TIMEOUT;
 }
@@ -534,6 +534,7 @@ static ResultCode CancelSynchronization(Core::System& system, Handle thread_hand
     }
 
     thread->CancelWait();
+    system.PrepareReschedule(thread->GetProcessorID());
     return RESULT_SUCCESS;
 }
 
@@ -1066,6 +1067,8 @@ static ResultCode SetThreadActivity(Core::System& system, Handle handle, u32 act
     }
 
     thread->SetActivity(static_cast<ThreadActivity>(activity));
+
+    system.PrepareReschedule(thread->GetProcessorID());
     return RESULT_SUCCESS;
 }
 
@@ -1147,7 +1150,7 @@ static ResultCode SetThreadPriority(Core::System& system, Handle handle, u32 pri
 
     thread->SetPriority(priority);
 
-    system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(thread->GetProcessorID());
     return RESULT_SUCCESS;
 }
 
@@ -1503,7 +1506,7 @@ static ResultCode CreateThread(Core::System& system, Handle* out_handle, VAddr e
     thread->SetName(
         fmt::format("thread[entry_point={:X}, handle={:X}]", entry_point, *new_thread_handle));
 
-    system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(thread->GetProcessorID());
 
     return RESULT_SUCCESS;
 }
@@ -1525,7 +1528,7 @@ static ResultCode StartThread(Core::System& system, Handle thread_handle) {
     thread->ResumeFromWait();
 
     if (thread->GetStatus() == ThreadStatus::Ready) {
-        system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+        system.PrepareReschedule(thread->GetProcessorID());
     }
 
     return RESULT_SUCCESS;
@@ -1537,7 +1540,7 @@ static void ExitThread(Core::System& system) {
 
     auto* const current_thread = system.CurrentScheduler().GetCurrentThread();
     current_thread->Stop();
-    system.CurrentScheduler().RemoveThread(current_thread);
+    system.GlobalScheduler().RemoveThread(current_thread);
     system.PrepareReschedule();
 }
 
@@ -1553,17 +1556,18 @@ static void SleepThread(Core::System& system, s64 nanoseconds) {
 
     auto& scheduler = system.CurrentScheduler();
     auto* const current_thread = scheduler.GetCurrentThread();
+    bool is_redundant = false;
 
     if (nanoseconds <= 0) {
         switch (static_cast<SleepType>(nanoseconds)) {
         case SleepType::YieldWithoutLoadBalancing:
-            scheduler.YieldWithoutLoadBalancing(current_thread);
+            is_redundant = current_thread->YieldSimple();
             break;
         case SleepType::YieldWithLoadBalancing:
-            scheduler.YieldWithLoadBalancing(current_thread);
+            is_redundant = current_thread->YieldAndBalanceLoad();
             break;
         case SleepType::YieldAndWaitForLoadBalancing:
-            scheduler.YieldAndWaitForLoadBalancing(current_thread);
+            is_redundant = current_thread->YieldAndWaitForLoadBalancing();
             break;
         default:
             UNREACHABLE_MSG("Unimplemented sleep yield type '{:016X}'!", nanoseconds);
@@ -1572,10 +1576,13 @@ static void SleepThread(Core::System& system, s64 nanoseconds) {
         current_thread->Sleep(nanoseconds);
     }
 
-    // Reschedule all CPU cores
-    for (std::size_t i = 0; i < Core::NUM_CPU_CORES; ++i) {
-        system.CpuCore(i).PrepareReschedule();
+    if (is_redundant) {
+        // If it's redundant, the core is pretty much idle. Some games keep idling
+        // a core while it's doing nothing, we advance timing to avoid costly continuous
+        // calls.
+        system.CoreTiming().AddTicks(2000);
     }
+    system.PrepareReschedule(current_thread->GetProcessorID());
 }
 
 /// Wait process wide key atomic
@@ -1601,6 +1608,8 @@ static ResultCode WaitProcessWideKeyAtomic(Core::System& system, VAddr mutex_add
         return ERR_INVALID_ADDRESS;
     }
 
+    ASSERT(condition_variable_addr == Common::AlignDown(condition_variable_addr, 4));
+
     auto* const current_process = system.Kernel().CurrentProcess();
     const auto& handle_table = current_process->GetHandleTable();
     SharedPtr<Thread> thread = handle_table.Get<Thread>(thread_handle);
@@ -1622,7 +1631,7 @@ static ResultCode WaitProcessWideKeyAtomic(Core::System& system, VAddr mutex_add
 
     // Note: Deliberately don't attempt to inherit the lock owner's priority.
 
-    system.CpuCore(current_thread->GetProcessorID()).PrepareReschedule();
+    system.PrepareReschedule(current_thread->GetProcessorID());
     return RESULT_SUCCESS;
 }
 
@@ -1632,24 +1641,19 @@ static ResultCode SignalProcessWideKey(Core::System& system, VAddr condition_var
     LOG_TRACE(Kernel_SVC, "called, condition_variable_addr=0x{:X}, target=0x{:08X}",
               condition_variable_addr, target);
 
-    const auto RetrieveWaitingThreads = [&system](std::size_t core_index,
-                                                  std::vector<SharedPtr<Thread>>& waiting_threads,
-                                                  VAddr condvar_addr) {
-        const auto& scheduler = system.Scheduler(core_index);
-        const auto& thread_list = scheduler.GetThreadList();
-
-        for (const auto& thread : thread_list) {
-            if (thread->GetCondVarWaitAddress() == condvar_addr)
-                waiting_threads.push_back(thread);
-        }
-    };
+    ASSERT(condition_variable_addr == Common::AlignDown(condition_variable_addr, 4));
 
     // Retrieve a list of all threads that are waiting for this condition variable.
     std::vector<SharedPtr<Thread>> waiting_threads;
-    RetrieveWaitingThreads(0, waiting_threads, condition_variable_addr);
-    RetrieveWaitingThreads(1, waiting_threads, condition_variable_addr);
-    RetrieveWaitingThreads(2, waiting_threads, condition_variable_addr);
-    RetrieveWaitingThreads(3, waiting_threads, condition_variable_addr);
+    const auto& scheduler = system.GlobalScheduler();
+    const auto& thread_list = scheduler.GetThreadList();
+
+    for (const auto& thread : thread_list) {
+        if (thread->GetCondVarWaitAddress() == condition_variable_addr) {
+            waiting_threads.push_back(thread);
+        }
+    }
+
     // Sort them by priority, such that the highest priority ones come first.
     std::sort(waiting_threads.begin(), waiting_threads.end(),
               [](const SharedPtr<Thread>& lhs, const SharedPtr<Thread>& rhs) {
@@ -1679,18 +1683,20 @@ static ResultCode SignalProcessWideKey(Core::System& system, VAddr condition_var
 
         // Atomically read the value of the mutex.
         u32 mutex_val = 0;
+        u32 update_val = 0;
+        const VAddr mutex_address = thread->GetMutexWaitAddress();
         do {
-            monitor.SetExclusive(current_core, thread->GetMutexWaitAddress());
+            monitor.SetExclusive(current_core, mutex_address);
 
             // If the mutex is not yet acquired, acquire it.
-            mutex_val = Memory::Read32(thread->GetMutexWaitAddress());
+            mutex_val = Memory::Read32(mutex_address);
 
             if (mutex_val != 0) {
-                monitor.ClearExclusive();
-                break;
+                update_val = mutex_val | Mutex::MutexHasWaitersFlag;
+            } else {
+                update_val = thread->GetWaitHandle();
             }
-        } while (!monitor.ExclusiveWrite32(current_core, thread->GetMutexWaitAddress(),
-                                           thread->GetWaitHandle()));
+        } while (!monitor.ExclusiveWrite32(current_core, mutex_address, update_val));
         if (mutex_val == 0) {
             // We were able to acquire the mutex, resume this thread.
             ASSERT(thread->GetStatus() == ThreadStatus::WaitCondVar);
@@ -1704,20 +1710,9 @@ static ResultCode SignalProcessWideKey(Core::System& system, VAddr condition_var
             thread->SetLockOwner(nullptr);
             thread->SetMutexWaitAddress(0);
             thread->SetWaitHandle(0);
-            system.CpuCore(thread->GetProcessorID()).PrepareReschedule();
+            thread->SetWaitSynchronizationResult(RESULT_SUCCESS);
+            system.PrepareReschedule(thread->GetProcessorID());
         } else {
-            // Atomically signal that the mutex now has a waiting thread.
-            do {
-                monitor.SetExclusive(current_core, thread->GetMutexWaitAddress());
-
-                // Ensure that the mutex value is still what we expect.
-                u32 value = Memory::Read32(thread->GetMutexWaitAddress());
-                // TODO(Subv): When this happens, the kernel just clears the exclusive state and
-                // retries the initial read for this thread.
-                ASSERT_MSG(mutex_val == value, "Unhandled synchronization primitive case");
-            } while (!monitor.ExclusiveWrite32(current_core, thread->GetMutexWaitAddress(),
-                                               mutex_val | Mutex::MutexHasWaitersFlag));
-
             // The mutex is already owned by some other thread, make this thread wait on it.
             const Handle owner_handle = static_cast<Handle>(mutex_val & Mutex::MutexOwnerMask);
             const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
@@ -1728,6 +1723,7 @@ static ResultCode SignalProcessWideKey(Core::System& system, VAddr condition_var
             thread->SetStatus(ThreadStatus::WaitMutex);
 
             owner->AddMutexWaiter(thread);
+            system.PrepareReschedule(thread->GetProcessorID());
         }
     }
 
@@ -1754,7 +1750,12 @@ static ResultCode WaitForAddress(Core::System& system, VAddr address, u32 type,
 
     const auto arbitration_type = static_cast<AddressArbiter::ArbitrationType>(type);
     auto& address_arbiter = system.Kernel().CurrentProcess()->GetAddressArbiter();
-    return address_arbiter.WaitForAddress(address, arbitration_type, value, timeout);
+    const ResultCode result =
+        address_arbiter.WaitForAddress(address, arbitration_type, value, timeout);
+    if (result == RESULT_SUCCESS) {
+        system.PrepareReschedule();
+    }
+    return result;
 }
 
 // Signals to an address (via Address Arbiter)
@@ -2040,7 +2041,10 @@ static ResultCode SetThreadCoreMask(Core::System& system, Handle thread_handle,
         return ERR_INVALID_HANDLE;
     }
 
+    system.PrepareReschedule(thread->GetProcessorID());
     thread->ChangeCore(core, affinity_mask);
+    system.PrepareReschedule(thread->GetProcessorID());
+
     return RESULT_SUCCESS;
 }
 
@@ -2151,6 +2155,7 @@ static ResultCode SignalEvent(Core::System& system, Handle handle) {
     }
 
     writable_event->Signal();
+    system.PrepareReschedule();
     return RESULT_SUCCESS;
 }
 

src/core/hle/kernel/thread.cpp

@@ -45,15 +45,7 @@ void Thread::Stop() {
                                                              callback_handle);
     kernel.ThreadWakeupCallbackHandleTable().Close(callback_handle);
     callback_handle = 0;
-
-    // Clean up thread from ready queue
-    // This is only needed when the thread is terminated forcefully (SVC TerminateProcess)
-    if (status == ThreadStatus::Ready || status == ThreadStatus::Paused) {
-        scheduler->UnscheduleThread(this, current_priority);
-    }
-
-    status = ThreadStatus::Dead;
-
+    SetStatus(ThreadStatus::Dead);
     WakeupAllWaitingThreads();
 
     // Clean up any dangling references in objects that this thread was waiting for
@@ -132,17 +124,16 @@ void Thread::ResumeFromWait() {
     wakeup_callback = nullptr;
 
     if (activity == ThreadActivity::Paused) {
-        status = ThreadStatus::Paused;
+        SetStatus(ThreadStatus::Paused);
         return;
     }
 
-    status = ThreadStatus::Ready;
-
-    ChangeScheduler();
+    SetStatus(ThreadStatus::Ready);
 }
 
 void Thread::CancelWait() {
     ASSERT(GetStatus() == ThreadStatus::WaitSynch);
+    ClearWaitObjects();
     SetWaitSynchronizationResult(ERR_SYNCHRONIZATION_CANCELED);
     ResumeFromWait();
 }
@@ -205,9 +196,9 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     thread->name = std::move(name);
     thread->callback_handle = kernel.ThreadWakeupCallbackHandleTable().Create(thread).Unwrap();
     thread->owner_process = &owner_process;
+    auto& scheduler = kernel.GlobalScheduler();
+    scheduler.AddThread(thread);
     thread->tls_address = thread->owner_process->CreateTLSRegion();
-    thread->scheduler = &system.Scheduler(processor_id);
-    thread->scheduler->AddThread(thread);
 
     thread->owner_process->RegisterThread(thread.get());
 
@@ -250,6 +241,22 @@ void Thread::SetStatus(ThreadStatus new_status) {
         return;
     }
 
+    switch (new_status) {
+    case ThreadStatus::Ready:
+    case ThreadStatus::Running:
+        SetSchedulingStatus(ThreadSchedStatus::Runnable);
+        break;
+    case ThreadStatus::Dormant:
+        SetSchedulingStatus(ThreadSchedStatus::None);
+        break;
+    case ThreadStatus::Dead:
+        SetSchedulingStatus(ThreadSchedStatus::Exited);
+        break;
+    default:
+        SetSchedulingStatus(ThreadSchedStatus::Paused);
+        break;
+    }
+
     if (status == ThreadStatus::Running) {
         last_running_ticks = Core::System::GetInstance().CoreTiming().GetTicks();
     }
@@ -311,8 +318,7 @@ void Thread::UpdatePriority() {
         return;
     }
 
-    scheduler->SetThreadPriority(this, new_priority);
-    current_priority = new_priority;
+    SetCurrentPriority(new_priority);
 
     if (!lock_owner) {
         return;
@@ -328,47 +334,7 @@ void Thread::UpdatePriority() {
 }
 
 void Thread::ChangeCore(u32 core, u64 mask) {
-    ideal_core = core;
-    affinity_mask = mask;
-    ChangeScheduler();
-}
-
-void Thread::ChangeScheduler() {
-    if (status != ThreadStatus::Ready) {
-        return;
-    }
-
-    auto& system = Core::System::GetInstance();
-    std::optional<s32> new_processor_id{GetNextProcessorId(affinity_mask)};
-
-    if (!new_processor_id) {
-        new_processor_id = processor_id;
-    }
-    if (ideal_core != -1 && system.Scheduler(ideal_core).GetCurrentThread() == nullptr) {
-        new_processor_id = ideal_core;
-    }
-
-    ASSERT(*new_processor_id < 4);
-
-    // Add thread to new core's scheduler
-    auto& next_scheduler = system.Scheduler(*new_processor_id);
-
-    if (*new_processor_id != processor_id) {
-        // Remove thread from previous core's scheduler
-        scheduler->RemoveThread(this);
-        next_scheduler.AddThread(this);
-    }
-
-    processor_id = *new_processor_id;
-
-    // If the thread was ready, unschedule from the previous core and schedule on the new core
-    scheduler->UnscheduleThread(this, current_priority);
-    next_scheduler.ScheduleThread(this, current_priority);
-
-    // Change thread's scheduler
-    scheduler = &next_scheduler;
-
-    system.CpuCore(processor_id).PrepareReschedule();
+    SetCoreAndAffinityMask(core, mask);
 }
 
 bool Thread::AllWaitObjectsReady() const {
@@ -388,10 +354,8 @@ void Thread::SetActivity(ThreadActivity value) {
 
     if (value == ThreadActivity::Paused) {
         // Set status if not waiting
-        if (status == ThreadStatus::Ready) {
-            status = ThreadStatus::Paused;
-        } else if (status == ThreadStatus::Running) {
-            status = ThreadStatus::Paused;
+        if (status == ThreadStatus::Ready || status == ThreadStatus::Running) {
+            SetStatus(ThreadStatus::Paused);
             Core::System::GetInstance().CpuCore(processor_id).PrepareReschedule();
         }
     } else if (status == ThreadStatus::Paused) {
@@ -408,6 +372,170 @@ void Thread::Sleep(s64 nanoseconds) {
     WakeAfterDelay(nanoseconds);
 }
 
+bool Thread::YieldSimple() {
+    auto& scheduler = kernel.GlobalScheduler();
+    return scheduler.YieldThread(this);
+}
+
+bool Thread::YieldAndBalanceLoad() {
+    auto& scheduler = kernel.GlobalScheduler();
+    return scheduler.YieldThreadAndBalanceLoad(this);
+}
+
+bool Thread::YieldAndWaitForLoadBalancing() {
+    auto& scheduler = kernel.GlobalScheduler();
+    return scheduler.YieldThreadAndWaitForLoadBalancing(this);
+}
+
+void Thread::SetSchedulingStatus(ThreadSchedStatus new_status) {
+    const u32 old_flags = scheduling_state;
+    scheduling_state = (scheduling_state & static_cast<u32>(ThreadSchedMasks::HighMask)) |
+                       static_cast<u32>(new_status);
+    AdjustSchedulingOnStatus(old_flags);
+}
+
+void Thread::SetCurrentPriority(u32 new_priority) {
+    const u32 old_priority = std::exchange(current_priority, new_priority);
+    AdjustSchedulingOnPriority(old_priority);
+}
+
+ResultCode Thread::SetCoreAndAffinityMask(s32 new_core, u64 new_affinity_mask) {
+    const auto HighestSetCore = [](u64 mask, u32 max_cores) {
+        for (s32 core = max_cores - 1; core >= 0; core--) {
+            if (((mask >> core) & 1) != 0) {
+                return core;
+            }
+        }
+        return -1;
+    };
+
+    const bool use_override = affinity_override_count != 0;
+    if (new_core == THREADPROCESSORID_DONT_UPDATE) {
+        new_core = use_override ? ideal_core_override : ideal_core;
+        if ((new_affinity_mask & (1ULL << new_core)) == 0) {
+            return ERR_INVALID_COMBINATION;
+        }
+    }
+    if (use_override) {
+        ideal_core_override = new_core;
+        affinity_mask_override = new_affinity_mask;
+    } else {
+        const u64 old_affinity_mask = std::exchange(affinity_mask, new_affinity_mask);
+        ideal_core = new_core;
+        if (old_affinity_mask != new_affinity_mask) {
+            const s32 old_core = processor_id;
+            if (processor_id >= 0 && ((affinity_mask >> processor_id) & 1) == 0) {
+                if (ideal_core < 0) {
+                    processor_id = HighestSetCore(affinity_mask, GlobalScheduler::NUM_CPU_CORES);
+                } else {
+                    processor_id = ideal_core;
+                }
+            }
+            AdjustSchedulingOnAffinity(old_affinity_mask, old_core);
+        }
+    }
+    return RESULT_SUCCESS;
+}
+
+void Thread::AdjustSchedulingOnStatus(u32 old_flags) {
+    if (old_flags == scheduling_state) {
+        return;
+    }
+
+    auto& scheduler = kernel.GlobalScheduler();
+    if (static_cast<ThreadSchedStatus>(old_flags & static_cast<u32>(ThreadSchedMasks::LowMask)) ==
+        ThreadSchedStatus::Runnable) {
+        // In this case the thread was running, now it's pausing/exitting
+        if (processor_id >= 0) {
+            scheduler.Unschedule(current_priority, processor_id, this);
+        }
+
+        for (s32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+            if (core != processor_id && ((affinity_mask >> core) & 1) != 0) {
+                scheduler.Unsuggest(current_priority, static_cast<u32>(core), this);
+            }
+        }
+    } else if (GetSchedulingStatus() == ThreadSchedStatus::Runnable) {
+        // The thread is now set to running from being stopped
+        if (processor_id >= 0) {
+            scheduler.Schedule(current_priority, processor_id, this);
+        }
+
+        for (s32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+            if (core != processor_id && ((affinity_mask >> core) & 1) != 0) {
+                scheduler.Suggest(current_priority, static_cast<u32>(core), this);
+            }
+        }
+    }
+
+    scheduler.SetReselectionPending();
+}
+
+void Thread::AdjustSchedulingOnPriority(u32 old_priority) {
+    if (GetSchedulingStatus() != ThreadSchedStatus::Runnable) {
+        return;
+    }
+    auto& scheduler = Core::System::GetInstance().GlobalScheduler();
+    if (processor_id >= 0) {
+        scheduler.Unschedule(old_priority, processor_id, this);
+    }
+
+    for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+        if (core != processor_id && ((affinity_mask >> core) & 1) != 0) {
+            scheduler.Unsuggest(old_priority, core, this);
+        }
+    }
+
+    // Add thread to the new priority queues.
+    Thread* current_thread = GetCurrentThread();
+
+    if (processor_id >= 0) {
+        if (current_thread == this) {
+            scheduler.SchedulePrepend(current_priority, processor_id, this);
+        } else {
+            scheduler.Schedule(current_priority, processor_id, this);
+        }
+    }
+
+    for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+        if (core != processor_id && ((affinity_mask >> core) & 1) != 0) {
+            scheduler.Suggest(current_priority, core, this);
+        }
+    }
+
+    scheduler.SetReselectionPending();
+}
+
+void Thread::AdjustSchedulingOnAffinity(u64 old_affinity_mask, s32 old_core) {
+    auto& scheduler = Core::System::GetInstance().GlobalScheduler();
+    if (GetSchedulingStatus() != ThreadSchedStatus::Runnable ||
+        current_priority >= THREADPRIO_COUNT) {
+        return;
+    }
+
+    for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+        if (((old_affinity_mask >> core) & 1) != 0) {
+            if (core == old_core) {
+                scheduler.Unschedule(current_priority, core, this);
+            } else {
+                scheduler.Unsuggest(current_priority, core, this);
+            }
+        }
+    }
+
+    for (u32 core = 0; core < GlobalScheduler::NUM_CPU_CORES; core++) {
+        if (((affinity_mask >> core) & 1) != 0) {
+            if (core == processor_id) {
+                scheduler.Schedule(current_priority, core, this);
+            } else {
+                scheduler.Suggest(current_priority, core, this);
+            }
+        }
+    }
+
+    scheduler.SetReselectionPending();
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
 /**

src/core/hle/kernel/thread.h

@@ -75,6 +75,26 @@ enum class ThreadActivity : u32 {
     Paused = 1,
 };
 
+enum class ThreadSchedStatus : u32 {
+    None = 0,
+    Paused = 1,
+    Runnable = 2,
+    Exited = 3,
+};
+
+enum class ThreadSchedFlags : u32 {
+    ProcessPauseFlag = 1 << 4,
+    ThreadPauseFlag = 1 << 5,
+    ProcessDebugPauseFlag = 1 << 6,
+    KernelInitPauseFlag = 1 << 8,
+};
+
+enum class ThreadSchedMasks : u32 {
+    LowMask = 0x000f,
+    HighMask = 0xfff0,
+    ForcePauseMask = 0x0070,
+};
+
 class Thread final : public WaitObject {
 public:
     using MutexWaitingThreads = std::vector<SharedPtr<Thread>>;
@@ -278,6 +298,10 @@ public:
         return processor_id;
     }
 
+    void SetProcessorID(s32 new_core) {
+        processor_id = new_core;
+    }
+
     Process* GetOwnerProcess() {
         return owner_process;
     }
@@ -295,6 +319,9 @@ public:
     }
 
     void ClearWaitObjects() {
+        for (const auto& waiting_object : wait_objects) {
+            waiting_object->RemoveWaitingThread(this);
+        }
         wait_objects.clear();
     }
 
@@ -383,11 +410,47 @@ public:
     /// Sleeps this thread for the given amount of nanoseconds.
     void Sleep(s64 nanoseconds);
 
+    /// Yields this thread without rebalancing loads.
+    bool YieldSimple();
+
+    /// Yields this thread and does a load rebalancing.
+    bool YieldAndBalanceLoad();
+
+    /// Yields this thread and if the core is left idle, loads are rebalanced
+    bool YieldAndWaitForLoadBalancing();
+
+    void IncrementYieldCount() {
+        yield_count++;
+    }
+
+    u64 GetYieldCount() const {
+        return yield_count;
+    }
+
+    ThreadSchedStatus GetSchedulingStatus() const {
+        return static_cast<ThreadSchedStatus>(scheduling_state &
+                                              static_cast<u32>(ThreadSchedMasks::LowMask));
+    }
+
+    bool IsRunning() const {
+        return is_running;
+    }
+
+    void SetIsRunning(bool value) {
+        is_running = value;
+    }
+
 private:
     explicit Thread(KernelCore& kernel);
     ~Thread() override;
 
-    void ChangeScheduler();
+    void SetSchedulingStatus(ThreadSchedStatus new_status);
+    void SetCurrentPriority(u32 new_priority);
+    ResultCode SetCoreAndAffinityMask(s32 new_core, u64 new_affinity_mask);
+
+    void AdjustSchedulingOnStatus(u32 old_flags);
+    void AdjustSchedulingOnPriority(u32 old_priority);
+    void AdjustSchedulingOnAffinity(u64 old_affinity_mask, s32 old_core);
 
     Core::ARM_Interface::ThreadContext context{};
 
@@ -409,6 +472,8 @@ private:
 
     u64 total_cpu_time_ticks = 0; ///< Total CPU running ticks.
     u64 last_running_ticks = 0;   ///< CPU tick when thread was last running
+    u64 yield_count = 0;          ///< Number of redundant yields carried by this thread.
+                                  ///< a redundant yield is one where no scheduling is changed
 
     s32 processor_id = 0;
 
@@ -453,6 +518,13 @@ private:
 
     ThreadActivity activity = ThreadActivity::Normal;
 
+    s32 ideal_core_override = -1;
+    u64 affinity_mask_override = 0x1;
+    u32 affinity_override_count = 0;
+
+    u32 scheduling_state = 0;
+    bool is_running = false;
+
     std::string name;
 };
 

src/core/hle/kernel/wait_object.cpp

@@ -6,6 +6,9 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
+#include "core/core.h"
+#include "core/core_cpu.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/thread.h"
@@ -82,9 +85,6 @@ void WaitObject::WakeupWaitingThread(SharedPtr<Thread> thread) {
 
     const std::size_t index = thread->GetWaitObjectIndex(this);
 
-    for (const auto& object : thread->GetWaitObjects()) {
-        object->RemoveWaitingThread(thread.get());
-    }
     thread->ClearWaitObjects();
 
     thread->CancelWakeupTimer();
@@ -95,6 +95,7 @@ void WaitObject::WakeupWaitingThread(SharedPtr<Thread> thread) {
     }
     if (resume) {
         thread->ResumeFromWait();
+        Core::System::GetInstance().PrepareReschedule(thread->GetProcessorID());
     }
 }
 

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

@@ -1140,8 +1140,9 @@ void IApplicationFunctions::PopLaunchParameter(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called, kind={:08X}", static_cast<u8>(kind));
 
     if (kind == LaunchParameterKind::ApplicationSpecific && !launch_popped_application_specific) {
-        const auto backend = BCAT::CreateBackendFromSettings(
-            [this](u64 tid) { return system.GetFileSystemController().GetBCATDirectory(tid); });
+        const auto backend = BCAT::CreateBackendFromSettings(system, [this](u64 tid) {
+            return system.GetFileSystemController().GetBCATDirectory(tid);
+        });
         const auto build_id_full = system.GetCurrentProcessBuildID();
         u64 build_id{};
         std::memcpy(&build_id, build_id_full.data(), sizeof(u64));

src/core/hle/service/apm/controller.cpp

@@ -2,6 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
+#include <array>
+#include <utility>
+
 #include "common/logging/log.h"
 #include "core/core_timing.h"
 #include "core/hle/service/apm/controller.h"
@@ -9,8 +13,7 @@
 
 namespace Service::APM {
 
-constexpr PerformanceConfiguration DEFAULT_PERFORMANCE_CONFIGURATION =
-    PerformanceConfiguration::Config7;
+constexpr auto DEFAULT_PERFORMANCE_CONFIGURATION = PerformanceConfiguration::Config7;
 
 Controller::Controller(Core::Timing::CoreTiming& core_timing)
     : core_timing{core_timing}, configs{
@@ -22,18 +25,35 @@ Controller::~Controller() = default;
 
 void Controller::SetPerformanceConfiguration(PerformanceMode mode,
                                              PerformanceConfiguration config) {
-    static const std::map<PerformanceConfiguration, u32> PCONFIG_TO_SPEED_MAP{
-        {PerformanceConfiguration::Config1, 1020},  {PerformanceConfiguration::Config2, 1020},
-        {PerformanceConfiguration::Config3, 1224},  {PerformanceConfiguration::Config4, 1020},
-        {PerformanceConfiguration::Config5, 1020},  {PerformanceConfiguration::Config6, 1224},
-        {PerformanceConfiguration::Config7, 1020},  {PerformanceConfiguration::Config8, 1020},
-        {PerformanceConfiguration::Config9, 1020},  {PerformanceConfiguration::Config10, 1020},
-        {PerformanceConfiguration::Config11, 1020}, {PerformanceConfiguration::Config12, 1020},
-        {PerformanceConfiguration::Config13, 1785}, {PerformanceConfiguration::Config14, 1785},
-        {PerformanceConfiguration::Config15, 1020}, {PerformanceConfiguration::Config16, 1020},
-    };
+    static constexpr std::array<std::pair<PerformanceConfiguration, u32>, 16> config_to_speed{{
+        {PerformanceConfiguration::Config1, 1020},
+        {PerformanceConfiguration::Config2, 1020},
+        {PerformanceConfiguration::Config3, 1224},
+        {PerformanceConfiguration::Config4, 1020},
+        {PerformanceConfiguration::Config5, 1020},
+        {PerformanceConfiguration::Config6, 1224},
+        {PerformanceConfiguration::Config7, 1020},
+        {PerformanceConfiguration::Config8, 1020},
+        {PerformanceConfiguration::Config9, 1020},
+        {PerformanceConfiguration::Config10, 1020},
+        {PerformanceConfiguration::Config11, 1020},
+        {PerformanceConfiguration::Config12, 1020},
+        {PerformanceConfiguration::Config13, 1785},
+        {PerformanceConfiguration::Config14, 1785},
+        {PerformanceConfiguration::Config15, 1020},
+        {PerformanceConfiguration::Config16, 1020},
+    }};
 
-    SetClockSpeed(PCONFIG_TO_SPEED_MAP.find(config)->second);
+    const auto iter = std::find_if(config_to_speed.cbegin(), config_to_speed.cend(),
+                                   [config](const auto& entry) { return entry.first == config; });
+
+    if (iter == config_to_speed.cend()) {
+        LOG_ERROR(Service_APM, "Invalid performance configuration value provided: {}",
+                  static_cast<u32>(config));
+        return;
+    }
+
+    SetClockSpeed(iter->second);
     configs.insert_or_assign(mode, config);
 }
 
@@ -48,7 +68,7 @@ void Controller::SetFromCpuBoostMode(CpuBoostMode mode) {
                                 BOOST_MODE_TO_CONFIG_MAP.at(static_cast<u32>(mode)));
 }
 
-PerformanceMode Controller::GetCurrentPerformanceMode() {
+PerformanceMode Controller::GetCurrentPerformanceMode() const {
     return Settings::values.use_docked_mode ? PerformanceMode::Docked : PerformanceMode::Handheld;
 }
 

src/core/hle/service/apm/controller.h

@@ -56,7 +56,7 @@ public:
     void SetPerformanceConfiguration(PerformanceMode mode, PerformanceConfiguration config);
     void SetFromCpuBoostMode(CpuBoostMode mode);
 
-    PerformanceMode GetCurrentPerformanceMode();
+    PerformanceMode GetCurrentPerformanceMode() const;
     PerformanceConfiguration GetCurrentPerformanceConfiguration(PerformanceMode mode);
 
 private:

src/core/hle/service/bcat/backend/backend.cpp

@@ -10,8 +10,8 @@
 
 namespace Service::BCAT {
 
-ProgressServiceBackend::ProgressServiceBackend(std::string_view event_name) {
-    auto& kernel{Core::System::GetInstance().Kernel()};
+ProgressServiceBackend::ProgressServiceBackend(Kernel::KernelCore& kernel,
+                                               std::string_view event_name) {
     event = Kernel::WritableEvent::CreateEventPair(
         kernel, Kernel::ResetType::Automatic,
         std::string("ProgressServiceBackend:UpdateEvent:").append(event_name));

src/core/hle/service/bcat/backend/backend.h

@@ -15,6 +15,14 @@
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/result.h"
 
+namespace Core {
+class System;
+}
+
+namespace Kernel {
+class KernelCore;
+}
+
 namespace Service::BCAT {
 
 struct DeliveryCacheProgressImpl;
@@ -88,7 +96,7 @@ public:
     void FinishDownload(ResultCode result);
 
 private:
-    explicit ProgressServiceBackend(std::string_view event_name);
+    explicit ProgressServiceBackend(Kernel::KernelCore& kernel, std::string_view event_name);
 
     Kernel::SharedPtr<Kernel::ReadableEvent> GetEvent() const;
     DeliveryCacheProgressImpl& GetImpl();
@@ -145,6 +153,6 @@ public:
     std::optional<std::vector<u8>> GetLaunchParameter(TitleIDVersion title) override;
 };
 
-std::unique_ptr<Backend> CreateBackendFromSettings(DirectoryGetter getter);
+std::unique_ptr<Backend> CreateBackendFromSettings(Core::System& system, DirectoryGetter getter);
 
 } // namespace Service::BCAT

src/core/hle/service/bcat/backend/boxcat.cpp

@@ -104,14 +104,15 @@ std::string GetZIPFilePath(u64 title_id) {
 
 // If the error is something the user should know about (build ID mismatch, bad client version),
 // display an error.
-void HandleDownloadDisplayResult(DownloadResult res) {
+void HandleDownloadDisplayResult(const AM::Applets::AppletManager& applet_manager,
+                                 DownloadResult res) {
     if (res == DownloadResult::Success || res == DownloadResult::NoResponse ||
         res == DownloadResult::GeneralWebError || res == DownloadResult::GeneralFSError ||
         res == DownloadResult::NoMatchTitleId || res == DownloadResult::InvalidContentType) {
         return;
     }
 
-    const auto& frontend{Core::System::GetInstance().GetAppletManager().GetAppletFrontendSet()};
+    const auto& frontend{applet_manager.GetAppletFrontendSet()};
     frontend.error->ShowCustomErrorText(
         ResultCode(-1), "There was an error while attempting to use Boxcat.",
         DOWNLOAD_RESULT_LOG_MESSAGES[static_cast<std::size_t>(res)], [] {});
@@ -264,12 +265,13 @@ private:
     u64 build_id;
 };
 
-Boxcat::Boxcat(DirectoryGetter getter) : Backend(std::move(getter)) {}
+Boxcat::Boxcat(AM::Applets::AppletManager& applet_manager_, DirectoryGetter getter)
+    : Backend(std::move(getter)), applet_manager{applet_manager_} {}
 
 Boxcat::~Boxcat() = default;
 
-void SynchronizeInternal(DirectoryGetter dir_getter, TitleIDVersion title,
-                         ProgressServiceBackend& progress,
+void SynchronizeInternal(AM::Applets::AppletManager& applet_manager, DirectoryGetter dir_getter,
+                         TitleIDVersion title, ProgressServiceBackend& progress,
                          std::optional<std::string> dir_name = {}) {
     progress.SetNeedHLELock(true);
 
@@ -295,7 +297,7 @@ void SynchronizeInternal(DirectoryGetter dir_getter, TitleIDVersion title,
             FileUtil::Delete(zip_path);
         }
 
-        HandleDownloadDisplayResult(res);
+        HandleDownloadDisplayResult(applet_manager, res);
         progress.FinishDownload(ERROR_GENERAL_BCAT_FAILURE);
         return;
     }
@@ -364,17 +366,24 @@ void SynchronizeInternal(DirectoryGetter dir_getter, TitleIDVersion title,
 
 bool Boxcat::Synchronize(TitleIDVersion title, ProgressServiceBackend& progress) {
     is_syncing.exchange(true);
-    std::thread([this, title, &progress] { SynchronizeInternal(dir_getter, title, progress); })
+
+    std::thread([this, title, &progress] {
+        SynchronizeInternal(applet_manager, dir_getter, title, progress);
+    })
         .detach();
+
     return true;
 }
 
 bool Boxcat::SynchronizeDirectory(TitleIDVersion title, std::string name,
                                   ProgressServiceBackend& progress) {
     is_syncing.exchange(true);
-    std::thread(
-        [this, title, name, &progress] { SynchronizeInternal(dir_getter, title, progress, name); })
+
+    std::thread([this, title, name, &progress] {
+        SynchronizeInternal(applet_manager, dir_getter, title, progress, name);
+    })
         .detach();
+
     return true;
 }
 
@@ -420,7 +429,7 @@ std::optional<std::vector<u8>> Boxcat::GetLaunchParameter(TitleIDVersion title)
                 FileUtil::Delete(path);
             }
 
-            HandleDownloadDisplayResult(res);
+            HandleDownloadDisplayResult(applet_manager, res);
             return std::nullopt;
         }
     }

src/core/hle/service/bcat/backend/boxcat.h

@@ -9,6 +9,10 @@
 #include <optional>
 #include "core/hle/service/bcat/backend/backend.h"
 
+namespace Service::AM::Applets {
+class AppletManager;
+}
+
 namespace Service::BCAT {
 
 struct EventStatus {
@@ -20,12 +24,13 @@ struct EventStatus {
 /// Boxcat is yuzu's custom backend implementation of Nintendo's BCAT service. It is free to use and
 /// doesn't require a switch or nintendo account. The content is controlled by the yuzu team.
 class Boxcat final : public Backend {
-    friend void SynchronizeInternal(DirectoryGetter dir_getter, TitleIDVersion title,
+    friend void SynchronizeInternal(AM::Applets::AppletManager& applet_manager,
+                                    DirectoryGetter dir_getter, TitleIDVersion title,
                                     ProgressServiceBackend& progress,
                                     std::optional<std::string> dir_name);
 
 public:
-    explicit Boxcat(DirectoryGetter getter);
+    explicit Boxcat(AM::Applets::AppletManager& applet_manager_, DirectoryGetter getter);
     ~Boxcat() override;
 
     bool Synchronize(TitleIDVersion title, ProgressServiceBackend& progress) override;
@@ -53,6 +58,7 @@ private:
 
     class Client;
     std::unique_ptr<Client> client;
+    AM::Applets::AppletManager& applet_manager;
 };
 
 } // namespace Service::BCAT

src/core/hle/service/bcat/module.cpp

@@ -125,7 +125,11 @@ private:
 class IBcatService final : public ServiceFramework<IBcatService> {
 public:
     explicit IBcatService(Core::System& system_, Backend& backend_)
-        : ServiceFramework("IBcatService"), system{system_}, backend{backend_} {
+        : ServiceFramework("IBcatService"), system{system_}, backend{backend_},
+          progress{{
+              ProgressServiceBackend{system_.Kernel(), "Normal"},
+              ProgressServiceBackend{system_.Kernel(), "Directory"},
+          }} {
         // clang-format off
         static const FunctionInfo functions[] = {
             {10100, &IBcatService::RequestSyncDeliveryCache, "RequestSyncDeliveryCache"},
@@ -249,10 +253,7 @@ private:
     Core::System& system;
     Backend& backend;
 
-    std::array<ProgressServiceBackend, static_cast<std::size_t>(SyncType::Count)> progress{
-        ProgressServiceBackend{"Normal"},
-        ProgressServiceBackend{"Directory"},
-    };
+    std::array<ProgressServiceBackend, static_cast<std::size_t>(SyncType::Count)> progress;
 };
 
 void Module::Interface::CreateBcatService(Kernel::HLERequestContext& ctx) {
@@ -557,12 +558,12 @@ void Module::Interface::CreateDeliveryCacheStorageServiceWithApplicationId(
     rb.PushIpcInterface<IDeliveryCacheStorageService>(fsc.GetBCATDirectory(title_id));
 }
 
-std::unique_ptr<Backend> CreateBackendFromSettings(DirectoryGetter getter) {
-    const auto backend = Settings::values.bcat_backend;
-
+std::unique_ptr<Backend> CreateBackendFromSettings([[maybe_unused]] Core::System& system,
+                                                   DirectoryGetter getter) {
 #ifdef YUZU_ENABLE_BOXCAT
-    if (backend == "boxcat")
-        return std::make_unique<Boxcat>(std::move(getter));
+    if (Settings::values.bcat_backend == "boxcat") {
+        return std::make_unique<Boxcat>(system.GetAppletManager(), std::move(getter));
+    }
 #endif
 
     return std::make_unique<NullBackend>(std::move(getter));
@@ -571,7 +572,8 @@ std::unique_ptr<Backend> CreateBackendFromSettings(DirectoryGetter getter) {
 Module::Interface::Interface(Core::System& system_, std::shared_ptr<Module> module_,
                              FileSystem::FileSystemController& fsc_, const char* name)
     : ServiceFramework(name), fsc{fsc_}, module{std::move(module_)},
-      backend{CreateBackendFromSettings([&fsc_](u64 tid) { return fsc_.GetBCATDirectory(tid); })},
+      backend{CreateBackendFromSettings(system_,
+                                        [&fsc_](u64 tid) { return fsc_.GetBCATDirectory(tid); })},
       system{system_} {}
 
 Module::Interface::~Interface() = default;

src/core/hle/service/es/es.cpp

@@ -128,7 +128,7 @@ private:
     void CountCommonTicket(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_ETicket, "called");
 
-        const auto count = keys.GetCommonTickets().size();
+        const u32 count = static_cast<u32>(keys.GetCommonTickets().size());
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
@@ -138,7 +138,7 @@ private:
     void CountPersonalizedTicket(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_ETicket, "called");
 
-        const auto count = keys.GetPersonalizedTickets().size();
+        const u32 count = static_cast<u32>(keys.GetPersonalizedTickets().size());
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
@@ -150,7 +150,7 @@ private:
         if (keys.GetCommonTickets().empty())
             out_entries = 0;
         else
-            out_entries = ctx.GetWriteBufferSize() / sizeof(u128);
+            out_entries = static_cast<u32>(ctx.GetWriteBufferSize() / sizeof(u128));
 
         LOG_DEBUG(Service_ETicket, "called, entries={:016X}", out_entries);
 
@@ -160,7 +160,7 @@ private:
         std::transform(tickets.begin(), tickets.end(), std::back_inserter(ids),
                        [](const auto& pair) { return pair.first; });
 
-        out_entries = std::min<u32>(ids.size(), out_entries);
+        out_entries = static_cast<u32>(std::min<std::size_t>(ids.size(), out_entries));
         ctx.WriteBuffer(ids.data(), out_entries * sizeof(u128));
 
         IPC::ResponseBuilder rb{ctx, 3};
@@ -173,7 +173,7 @@ private:
         if (keys.GetPersonalizedTickets().empty())
             out_entries = 0;
         else
-            out_entries = ctx.GetWriteBufferSize() / sizeof(u128);
+            out_entries = static_cast<u32>(ctx.GetWriteBufferSize() / sizeof(u128));
 
         LOG_DEBUG(Service_ETicket, "called, entries={:016X}", out_entries);
 
@@ -183,7 +183,7 @@ private:
         std::transform(tickets.begin(), tickets.end(), std::back_inserter(ids),
                        [](const auto& pair) { return pair.first; });
 
-        out_entries = std::min<u32>(ids.size(), out_entries);
+        out_entries = static_cast<u32>(std::min<std::size_t>(ids.size(), out_entries));
         ctx.WriteBuffer(ids.data(), out_entries * sizeof(u128));
 
         IPC::ResponseBuilder rb{ctx, 3};

src/core/hle/service/hid/controllers/npad.cpp

@@ -583,36 +583,6 @@ bool Controller_NPad::SwapNpadAssignment(u32 npad_id_1, u32 npad_id_2) {
     return true;
 }
 
-bool Controller_NPad::IsControllerSupported(NPadControllerType controller) {
-    if (controller == NPadControllerType::Handheld) {
-        // Handheld is not even a supported type, lets stop here
-        if (std::find(supported_npad_id_types.begin(), supported_npad_id_types.end(),
-                      NPAD_HANDHELD) == supported_npad_id_types.end()) {
-            return false;
-        }
-        // Handheld should not be supported in docked mode
-        if (Settings::values.use_docked_mode) {
-            return false;
-        }
-    }
-    switch (controller) {
-    case NPadControllerType::ProController:
-        return style.pro_controller;
-    case NPadControllerType::Handheld:
-        return style.handheld;
-    case NPadControllerType::JoyDual:
-        return style.joycon_dual;
-    case NPadControllerType::JoyLeft:
-        return style.joycon_left;
-    case NPadControllerType::JoyRight:
-        return style.joycon_right;
-    case NPadControllerType::Pokeball:
-        return style.pokeball;
-    default:
-        return false;
-    }
-}
-
 Controller_NPad::LedPattern Controller_NPad::GetLedPattern(u32 npad_id) {
     if (npad_id == npad_id_list.back() || npad_id == npad_id_list[npad_id_list.size() - 2]) {
         // These are controllers without led patterns
@@ -659,25 +629,24 @@ void Controller_NPad::ClearAllConnectedControllers() {
 }
 
 void Controller_NPad::DisconnectAllConnectedControllers() {
-    std::for_each(connected_controllers.begin(), connected_controllers.end(),
-                  [](ControllerHolder& controller) { controller.is_connected = false; });
+    for (ControllerHolder& controller : connected_controllers) {
+        controller.is_connected = false;
+    }
 }
 
 void Controller_NPad::ConnectAllDisconnectedControllers() {
-    std::for_each(connected_controllers.begin(), connected_controllers.end(),
-                  [](ControllerHolder& controller) {
-                      if (controller.type != NPadControllerType::None && !controller.is_connected) {
-                          controller.is_connected = false;
-                      }
-                  });
+    for (ControllerHolder& controller : connected_controllers) {
+        if (controller.type != NPadControllerType::None && !controller.is_connected) {
+            controller.is_connected = true;
+        }
+    }
 }
 
 void Controller_NPad::ClearAllControllers() {
-    std::for_each(connected_controllers.begin(), connected_controllers.end(),
-                  [](ControllerHolder& controller) {
-                      controller.type = NPadControllerType::None;
-                      controller.is_connected = false;
-                  });
+    for (ControllerHolder& controller : connected_controllers) {
+        controller.type = NPadControllerType::None;
+        controller.is_connected = false;
+    }
 }
 
 u32 Controller_NPad::GetAndResetPressState() {
@@ -685,10 +654,10 @@ u32 Controller_NPad::GetAndResetPressState() {
 }
 
 bool Controller_NPad::IsControllerSupported(NPadControllerType controller) const {
-    const bool support_handheld =
-        std::find(supported_npad_id_types.begin(), supported_npad_id_types.end(), NPAD_HANDHELD) !=
-        supported_npad_id_types.end();
     if (controller == NPadControllerType::Handheld) {
+        const bool support_handheld =
+            std::find(supported_npad_id_types.begin(), supported_npad_id_types.end(),
+                      NPAD_HANDHELD) != supported_npad_id_types.end();
         // Handheld is not even a supported type, lets stop here
         if (!support_handheld) {
             return false;
@@ -700,6 +669,7 @@ bool Controller_NPad::IsControllerSupported(NPadControllerType controller) const
 
         return true;
     }
+
     if (std::any_of(supported_npad_id_types.begin(), supported_npad_id_types.end(),
                     [](u32 npad_id) { return npad_id <= MAX_NPAD_ID; })) {
         switch (controller) {
@@ -717,6 +687,7 @@ bool Controller_NPad::IsControllerSupported(NPadControllerType controller) const
             return false;
         }
     }
+
     return false;
 }
 
@@ -795,6 +766,7 @@ Controller_NPad::NPadControllerType Controller_NPad::DecideBestController(
         priority_list.push_back(NPadControllerType::JoyLeft);
         priority_list.push_back(NPadControllerType::JoyRight);
         priority_list.push_back(NPadControllerType::JoyDual);
+        break;
     }
 
     const auto iter = std::find_if(priority_list.begin(), priority_list.end(),

src/core/hle/service/hid/controllers/npad.h

@@ -301,6 +301,11 @@ private:
         bool is_connected;
     };
 
+    void InitNewlyAddedControler(std::size_t controller_idx);
+    bool IsControllerSupported(NPadControllerType controller) const;
+    NPadControllerType DecideBestController(NPadControllerType priority) const;
+    void RequestPadStateUpdate(u32 npad_id);
+
     u32 press_state{};
 
     NPadType style{};
@@ -321,12 +326,7 @@ private:
     std::array<ControllerHolder, 10> connected_controllers{};
     bool can_controllers_vibrate{true};
 
-    void InitNewlyAddedControler(std::size_t controller_idx);
-    bool IsControllerSupported(NPadControllerType controller) const;
-    NPadControllerType DecideBestController(NPadControllerType priority) const;
-    void RequestPadStateUpdate(u32 npad_id);
     std::array<ControllerPad, 10> npad_pad_states{};
-    bool IsControllerSupported(NPadControllerType controller);
     bool is_in_lr_assignment_mode{false};
     Core::System& system;
 };

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

@@ -6,13 +6,6 @@
 #include <cstring>
 #include <vector>
 
-#include <FontChineseSimplified.h>
-#include <FontChineseTraditional.h>
-#include <FontExtendedChineseSimplified.h>
-#include <FontKorean.h>
-#include <FontNintendoExtended.h>
-#include <FontStandard.h>
-
 #include "common/assert.h"
 #include "common/common_paths.h"
 #include "common/common_types.h"
@@ -24,7 +17,9 @@
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/registered_cache.h"
 #include "core/file_sys/romfs.h"
+#include "core/file_sys/system_archive/system_archive.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/physical_memory.h"
 #include "core/hle/kernel/shared_memory.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/ns/pl_u.h"
@@ -94,15 +89,20 @@ static void DecryptSharedFont(const std::vector<u32>& input, Kernel::PhysicalMem
     offset += transformed_font.size() * sizeof(u32);
 }
 
-static void EncryptSharedFont(const std::vector<u8>& input, Kernel::PhysicalMemory& output,
-                              std::size_t& offset) {
-    ASSERT_MSG(offset + input.size() + 8 < SHARED_FONT_MEM_SIZE, "Shared fonts exceeds 17mb!");
-    const u32 KEY = EXPECTED_MAGIC ^ EXPECTED_RESULT;
-    std::memcpy(output.data() + offset, &EXPECTED_RESULT, sizeof(u32)); // Magic header
-    const u32 ENC_SIZE = static_cast<u32>(input.size()) ^ KEY;
-    std::memcpy(output.data() + offset + sizeof(u32), &ENC_SIZE, sizeof(u32));
-    std::memcpy(output.data() + offset + (sizeof(u32) * 2), input.data(), input.size());
-    offset += input.size() + (sizeof(u32) * 2);
+void EncryptSharedFont(const std::vector<u32>& input, std::vector<u8>& output,
+                       std::size_t& offset) {
+    ASSERT_MSG(offset + (input.size() * sizeof(u32)) < SHARED_FONT_MEM_SIZE,
+               "Shared fonts exceeds 17mb!");
+
+    const auto key = Common::swap32(EXPECTED_RESULT ^ EXPECTED_MAGIC);
+    std::vector<u32> transformed_font(input.size() + 2);
+    transformed_font[0] = Common::swap32(EXPECTED_MAGIC);
+    transformed_font[1] = Common::swap32(input.size() * sizeof(u32)) ^ key;
+    std::transform(input.begin(), input.end(), transformed_font.begin() + 2,
+                   [key](u32 in) { return in ^ key; });
+    std::memcpy(output.data() + offset, transformed_font.data(),
+                transformed_font.size() * sizeof(u32));
+    offset += transformed_font.size() * sizeof(u32);
 }
 
 // Helper function to make BuildSharedFontsRawRegions a bit nicer
@@ -168,114 +168,49 @@ PL_U::PL_U(Core::System& system)
     // Attempt to load shared font data from disk
     const auto* nand = fsc.GetSystemNANDContents();
     std::size_t offset = 0;
-    // Rebuild shared fonts from data ncas
-    if (nand->HasEntry(static_cast<u64>(FontArchives::Standard),
-                       FileSys::ContentRecordType::Data)) {
-        impl->shared_font = std::make_shared<Kernel::PhysicalMemory>(SHARED_FONT_MEM_SIZE);
-        for (auto font : SHARED_FONTS) {
-            const auto nca =
-                nand->GetEntry(static_cast<u64>(font.first), FileSys::ContentRecordType::Data);
-            if (!nca) {
-                LOG_ERROR(Service_NS, "Failed to find {:016X}! Skipping",
-                          static_cast<u64>(font.first));
-                continue;
-            }
-            const auto romfs = nca->GetRomFS();
-            if (!romfs) {
-                LOG_ERROR(Service_NS, "{:016X} has no RomFS! Skipping",
-                          static_cast<u64>(font.first));
-                continue;
-            }
-            const auto extracted_romfs = FileSys::ExtractRomFS(romfs);
-            if (!extracted_romfs) {
-                LOG_ERROR(Service_NS, "Failed to extract RomFS for {:016X}! Skipping",
-                          static_cast<u64>(font.first));
-                continue;
-            }
-            const auto font_fp = extracted_romfs->GetFile(font.second);
-            if (!font_fp) {
-                LOG_ERROR(Service_NS, "{:016X} has no file \"{}\"! Skipping",
-                          static_cast<u64>(font.first), font.second);
-                continue;
-            }
-            std::vector<u32> font_data_u32(font_fp->GetSize() / sizeof(u32));
-            font_fp->ReadBytes<u32>(font_data_u32.data(), font_fp->GetSize());
-            // We need to be BigEndian as u32s for the xor encryption
-            std::transform(font_data_u32.begin(), font_data_u32.end(), font_data_u32.begin(),
-                           Common::swap32);
-            FontRegion region{
-                static_cast<u32>(offset + 8),
-                static_cast<u32>((font_data_u32.size() * sizeof(u32)) -
-                                 8)}; // Font offset and size do not account for the header
-            DecryptSharedFont(font_data_u32, *impl->shared_font, offset);
-            impl->shared_font_regions.push_back(region);
+    // Rebuild shared fonts from data ncas or synthesize
+
+    impl->shared_font = std::make_shared<Kernel::PhysicalMemory>(SHARED_FONT_MEM_SIZE);
+    for (auto font : SHARED_FONTS) {
+        FileSys::VirtualFile romfs;
+        const auto nca =
+            nand->GetEntry(static_cast<u64>(font.first), FileSys::ContentRecordType::Data);
+        if (nca) {
+            romfs = nca->GetRomFS();
         }
 
-    } else {
-        impl->shared_font = std::make_shared<Kernel::PhysicalMemory>(
-            SHARED_FONT_MEM_SIZE); // Shared memory needs to always be allocated and a fixed size
-
-        const std::string user_path = FileUtil::GetUserPath(FileUtil::UserPath::SysDataDir);
-        const std::string filepath{user_path + SHARED_FONT};
-
-        // Create path if not already created
-        if (!FileUtil::CreateFullPath(filepath)) {
-            LOG_ERROR(Service_NS, "Failed to create sharedfonts path \"{}\"!", filepath);
-            return;
+        if (!romfs) {
+            romfs = FileSys::SystemArchive::SynthesizeSystemArchive(static_cast<u64>(font.first));
         }
 
-        bool using_ttf = false;
-        for (const char* font_ttf : SHARED_FONTS_TTF) {
-            if (FileUtil::Exists(user_path + font_ttf)) {
-                using_ttf = true;
-                FileUtil::IOFile file(user_path + font_ttf, "rb");
-                if (file.IsOpen()) {
-                    std::vector<u8> ttf_bytes(file.GetSize());
-                    file.ReadBytes<u8>(ttf_bytes.data(), ttf_bytes.size());
-                    FontRegion region{
-                        static_cast<u32>(offset + 8),
-                        static_cast<u32>(ttf_bytes.size())}; // Font offset and size do not account
-                                                             // for the header
-                    EncryptSharedFont(ttf_bytes, *impl->shared_font, offset);
-                    impl->shared_font_regions.push_back(region);
-                } else {
-                    LOG_WARNING(Service_NS, "Unable to load font: {}", font_ttf);
-                }
-            } else if (using_ttf) {
-                LOG_WARNING(Service_NS, "Unable to find font: {}", font_ttf);
-            }
+        if (!romfs) {
+            LOG_ERROR(Service_NS, "Failed to find or synthesize {:016X}! Skipping",
+                      static_cast<u64>(font.first));
+            continue;
         }
-        if (using_ttf)
-            return;
-        FileUtil::IOFile file(filepath, "rb");
 
-        if (file.IsOpen()) {
-            // Read shared font data
-            ASSERT(file.GetSize() == SHARED_FONT_MEM_SIZE);
-            file.ReadBytes(impl->shared_font->data(), impl->shared_font->size());
-            impl->BuildSharedFontsRawRegions(*impl->shared_font);
-        } else {
-            LOG_WARNING(Service_NS,
-                        "Shared Font file missing. Loading open source replacement from memory");
-
-            // clang-format off
-            const std::vector<std::vector<u8>> open_source_shared_fonts_ttf = {
-                {std::begin(FontChineseSimplified), std::end(FontChineseSimplified)},
-                {std::begin(FontChineseTraditional), std::end(FontChineseTraditional)},
-                {std::begin(FontExtendedChineseSimplified), std::end(FontExtendedChineseSimplified)},
-                {std::begin(FontKorean), std::end(FontKorean)},
-                {std::begin(FontNintendoExtended), std::end(FontNintendoExtended)},
-                {std::begin(FontStandard), std::end(FontStandard)},
-            };
-            // clang-format on
-
-            for (const std::vector<u8>& font_ttf : open_source_shared_fonts_ttf) {
-                const FontRegion region{static_cast<u32>(offset + 8),
-                                        static_cast<u32>(font_ttf.size())};
-                EncryptSharedFont(font_ttf, *impl->shared_font, offset);
-                impl->shared_font_regions.push_back(region);
-            }
+        const auto extracted_romfs = FileSys::ExtractRomFS(romfs);
+        if (!extracted_romfs) {
+            LOG_ERROR(Service_NS, "Failed to extract RomFS for {:016X}! Skipping",
+                      static_cast<u64>(font.first));
+            continue;
         }
+        const auto font_fp = extracted_romfs->GetFile(font.second);
+        if (!font_fp) {
+            LOG_ERROR(Service_NS, "{:016X} has no file \"{}\"! Skipping",
+                      static_cast<u64>(font.first), font.second);
+            continue;
+        }
+        std::vector<u32> font_data_u32(font_fp->GetSize() / sizeof(u32));
+        font_fp->ReadBytes<u32>(font_data_u32.data(), font_fp->GetSize());
+        // We need to be BigEndian as u32s for the xor encryption
+        std::transform(font_data_u32.begin(), font_data_u32.end(), font_data_u32.begin(),
+                       Common::swap32);
+        // Font offset and size do not account for the header
+        const FontRegion region{static_cast<u32>(offset + 8),
+                                static_cast<u32>((font_data_u32.size() * sizeof(u32)) - 8)};
+        DecryptSharedFont(font_data_u32, *impl->shared_font, offset);
+        impl->shared_font_regions.push_back(region);
     }
 }
 

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

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <memory>
+#include <vector>
 #include "core/hle/service/service.h"
 
 namespace Service {
@@ -15,6 +16,8 @@ class FileSystemController;
 
 namespace NS {
 
+void EncryptSharedFont(const std::vector<u32>& input, std::vector<u8>& output, std::size_t& offset);
+
 class PL_U final : public ServiceFramework<PL_U> {
 public:
     explicit PL_U(Core::System& system);

src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp

@@ -5,6 +5,7 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
+#include "core/core_timing.h"
 #include "core/hle/service/nvdrv/devices/nvdisp_disp0.h"
 #include "core/hle/service/nvdrv/devices/nvmap.h"
 #include "core/perf_stats.h"
@@ -38,7 +39,10 @@ void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u3
         transform, crop_rect};
 
     system.GetPerfStats().EndGameFrame();
+    system.GetPerfStats().EndSystemFrame();
     system.GPU().SwapBuffers(&framebuffer);
+    system.FrameLimiter().DoFrameLimiting(system.CoreTiming().GetGlobalTimeUs());
+    system.GetPerfStats().BeginSystemFrame();
 }
 
 } // namespace Service::Nvidia::Devices

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

@@ -63,16 +63,26 @@ u32 nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>&
         return NvResult::BadParameter;
     }
 
+    u32 event_id = params.value & 0x00FF;
+
+    if (event_id >= MaxNvEvents) {
+        std::memcpy(output.data(), &params, sizeof(params));
+        return NvResult::BadParameter;
+    }
+
+    auto event = events_interface.events[event_id];
     auto& gpu = system.GPU();
     // This is mostly to take into account unimplemented features. As synced
     // gpu is always synced.
     if (!gpu.IsAsync()) {
+        event.writable->Signal();
         return NvResult::Success;
     }
     auto lock = gpu.LockSync();
     const u32 current_syncpoint_value = gpu.GetSyncpointValue(params.syncpt_id);
     const s32 diff = current_syncpoint_value - params.threshold;
     if (diff >= 0) {
+        event.writable->Signal();
         params.value = current_syncpoint_value;
         std::memcpy(output.data(), &params, sizeof(params));
         return NvResult::Success;
@@ -88,27 +98,6 @@ u32 nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>&
         return NvResult::Timeout;
     }
 
-    u32 event_id;
-    if (is_async) {
-        event_id = params.value & 0x00FF;
-        if (event_id >= MaxNvEvents) {
-            std::memcpy(output.data(), &params, sizeof(params));
-            return NvResult::BadParameter;
-        }
-    } else {
-        if (ctrl.fresh_call) {
-            const auto result = events_interface.GetFreeEvent();
-            if (result) {
-                event_id = *result;
-            } else {
-                LOG_CRITICAL(Service_NVDRV, "No Free Events available!");
-                event_id = params.value & 0x00FF;
-            }
-        } else {
-            event_id = ctrl.event_id;
-        }
-    }
-
     EventState status = events_interface.status[event_id];
     if (event_id < MaxNvEvents || status == EventState::Free || status == EventState::Registered) {
         events_interface.SetEventStatus(event_id, EventState::Waiting);
@@ -120,7 +109,7 @@ u32 nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>&
             params.value = ((params.syncpt_id & 0xfff) << 16) | 0x10000000;
         }
         params.value |= event_id;
-        events_interface.events[event_id].writable->Clear();
+        event.writable->Clear();
         gpu.RegisterSyncptInterrupt(params.syncpt_id, target_value);
         if (!is_async && ctrl.fresh_call) {
             ctrl.must_delay = true;

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

@@ -134,7 +134,9 @@ void NVDRV::QueryEvent(Kernel::HLERequestContext& ctx) {
     IPC::ResponseBuilder rb{ctx, 3, 1};
     rb.Push(RESULT_SUCCESS);
     if (event_id < MaxNvEvents) {
-        rb.PushCopyObjects(nvdrv->GetEvent(event_id));
+        auto event = nvdrv->GetEvent(event_id);
+        event->Clear();
+        rb.PushCopyObjects(event);
         rb.Push<u32>(NvResult::Success);
     } else {
         rb.Push<u32>(0);

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

@@ -40,8 +40,8 @@ Module::Module(Core::System& system) {
     auto& kernel = system.Kernel();
     for (u32 i = 0; i < MaxNvEvents; i++) {
         std::string event_label = fmt::format("NVDRV::NvEvent_{}", i);
-        events_interface.events[i] = Kernel::WritableEvent::CreateEventPair(
-            kernel, Kernel::ResetType::Automatic, event_label);
+        events_interface.events[i] =
+            Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Manual, event_label);
         events_interface.status[i] = EventState::Free;
         events_interface.registered[i] = false;
     }

src/core/hle/service/nvflinger/buffer_queue.cpp

@@ -14,8 +14,8 @@
 
 namespace Service::NVFlinger {
 
-BufferQueue::BufferQueue(u32 id, u64 layer_id) : id(id), layer_id(layer_id) {
-    auto& kernel = Core::System::GetInstance().Kernel();
+BufferQueue::BufferQueue(Kernel::KernelCore& kernel, u32 id, u64 layer_id)
+    : id(id), layer_id(layer_id) {
     buffer_wait_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Manual,
                                                                "BufferQueue NativeHandle");
 }

src/core/hle/service/nvflinger/buffer_queue.h

@@ -15,6 +15,10 @@
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/nvdrv/nvdata.h"
 
+namespace Kernel {
+class KernelCore;
+}
+
 namespace Service::NVFlinger {
 
 struct IGBPBuffer {
@@ -44,7 +48,7 @@ public:
         NativeWindowFormat = 2,
     };
 
-    BufferQueue(u32 id, u64 layer_id);
+    explicit BufferQueue(Kernel::KernelCore& kernel, u32 id, u64 layer_id);
     ~BufferQueue();
 
     enum class BufferTransformFlags : u32 {

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -83,7 +83,7 @@ std::optional<u64> NVFlinger::CreateLayer(u64 display_id) {
 
     const u64 layer_id = next_layer_id++;
     const u32 buffer_queue_id = next_buffer_queue_id++;
-    buffer_queues.emplace_back(buffer_queue_id, layer_id);
+    buffer_queues.emplace_back(system.Kernel(), buffer_queue_id, layer_id);
     display->CreateLayer(layer_id, buffer_queues.back());
     return layer_id;
 }
@@ -187,14 +187,18 @@ void NVFlinger::Compose() {
         MicroProfileFlip();
 
         if (!buffer) {
-            // There was no queued buffer to draw, render previous frame
-            system.GetPerfStats().EndGameFrame();
-            system.GPU().SwapBuffers({});
             continue;
         }
 
         const auto& igbp_buffer = buffer->get().igbp_buffer;
 
+        const auto& gpu = system.GPU();
+        const auto& multi_fence = buffer->get().multi_fence;
+        for (u32 fence_id = 0; fence_id < multi_fence.num_fences; fence_id++) {
+            const auto& fence = multi_fence.fences[fence_id];
+            gpu.WaitFence(fence.id, fence.value);
+        }
+
         // Now send the buffer to the GPU for drawing.
         // TODO(Subv): Support more than just disp0. The display device selection is probably based
         // on which display we're drawing (Default, Internal, External, etc)

src/core/memory/cheat_engine.cpp

@@ -22,7 +22,7 @@ constexpr u32 KEYPAD_BITMASK = 0x3FFFFFF;
 
 StandardVmCallbacks::StandardVmCallbacks(const Core::System& system,
                                          const CheatProcessMetadata& metadata)
-    : system(system), metadata(metadata) {}
+    : metadata(metadata), system(system) {}
 
 StandardVmCallbacks::~StandardVmCallbacks() = default;
 
@@ -176,9 +176,8 @@ std::vector<CheatEntry> TextCheatParser::Parse(const Core::System& system,
 
 CheatEngine::CheatEngine(Core::System& system, std::vector<CheatEntry> cheats,
                          const std::array<u8, 0x20>& build_id)
-    : system{system}, core_timing{system.CoreTiming()}, vm{std::make_unique<StandardVmCallbacks>(
-                                                            system, metadata)},
-      cheats(std::move(cheats)) {
+    : vm{std::make_unique<StandardVmCallbacks>(system, metadata)},
+      cheats(std::move(cheats)), core_timing{system.CoreTiming()}, system{system} {
     metadata.main_nso_build_id = build_id;
 }
 

src/core/memory/dmnt_cheat_vm.cpp

@@ -1133,8 +1133,8 @@ void DmntCheatVm::Execute(const CheatProcessMetadata& metadata) {
             case SaveRestoreRegisterOpType::ClearRegs:
             case SaveRestoreRegisterOpType::Restore:
             default:
-                src = registers.data();
-                dst = saved_values.data();
+                src = saved_values.data();
+                dst = registers.data();
                 break;
             }
             for (std::size_t i = 0; i < NumRegisters; i++) {

src/tests/core/core_timing.cpp

@@ -6,6 +6,7 @@
 
 #include <array>
 #include <bitset>
+#include <cstdlib>
 #include <string>
 #include "common/file_util.h"
 #include "core/core.h"
@@ -13,7 +14,7 @@
 
 // Numbers are chosen randomly to make sure the correct one is given.
 static constexpr std::array<u64, 5> CB_IDS{{42, 144, 93, 1026, UINT64_C(0xFFFF7FFFF7FFFF)}};
-static constexpr int MAX_SLICE_LENGTH = 20000; // Copied from CoreTiming internals
+static constexpr int MAX_SLICE_LENGTH = 10000; // Copied from CoreTiming internals
 
 static std::bitset<CB_IDS.size()> callbacks_ran_flags;
 static u64 expected_callback = 0;
@@ -28,6 +29,12 @@ void CallbackTemplate(u64 userdata, s64 cycles_late) {
     REQUIRE(lateness == cycles_late);
 }
 
+static u64 callbacks_done = 0;
+
+void EmptyCallback(u64 userdata, s64 cycles_late) {
+    ++callbacks_done;
+}
+
 struct ScopeInit final {
     ScopeInit() {
         core_timing.Initialize();
@@ -39,18 +46,19 @@ struct ScopeInit final {
     Core::Timing::CoreTiming core_timing;
 };
 
-static void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, int downcount,
+static void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, u32 context = 0,
                             int expected_lateness = 0, int cpu_downcount = 0) {
     callbacks_ran_flags = 0;
     expected_callback = CB_IDS[idx];
     lateness = expected_lateness;
 
     // Pretend we executed X cycles of instructions.
+    core_timing.SwitchContext(context);
     core_timing.AddTicks(core_timing.GetDowncount() - cpu_downcount);
     core_timing.Advance();
+    core_timing.SwitchContext((context + 1) % 4);
 
     REQUIRE(decltype(callbacks_ran_flags)().set(idx) == callbacks_ran_flags);
-    REQUIRE(downcount == core_timing.GetDowncount());
 }
 
 TEST_CASE("CoreTiming[BasicOrder]", "[core]") {
@@ -64,9 +72,10 @@ TEST_CASE("CoreTiming[BasicOrder]", "[core]") {
     Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", CallbackTemplate<4>);
 
     // Enter slice 0
-    core_timing.Advance();
+    core_timing.ResetRun();
 
     // D -> B -> C -> A -> E
+    core_timing.SwitchContext(0);
     core_timing.ScheduleEvent(1000, cb_a, CB_IDS[0]);
     REQUIRE(1000 == core_timing.GetDowncount());
     core_timing.ScheduleEvent(500, cb_b, CB_IDS[1]);
@@ -78,98 +87,46 @@ TEST_CASE("CoreTiming[BasicOrder]", "[core]") {
     core_timing.ScheduleEvent(1200, cb_e, CB_IDS[4]);
     REQUIRE(100 == core_timing.GetDowncount());
 
-    AdvanceAndCheck(core_timing, 3, 400);
-    AdvanceAndCheck(core_timing, 1, 300);
-    AdvanceAndCheck(core_timing, 2, 200);
-    AdvanceAndCheck(core_timing, 0, 200);
-    AdvanceAndCheck(core_timing, 4, MAX_SLICE_LENGTH);
+    AdvanceAndCheck(core_timing, 3, 0);
+    AdvanceAndCheck(core_timing, 1, 1);
+    AdvanceAndCheck(core_timing, 2, 2);
+    AdvanceAndCheck(core_timing, 0, 3);
+    AdvanceAndCheck(core_timing, 4, 0);
 }
 
-TEST_CASE("CoreTiming[Threadsave]", "[core]") {
-    ScopeInit guard;
-    auto& core_timing = guard.core_timing;
-
-    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
-    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
-    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", CallbackTemplate<2>);
-    Core::Timing::EventType* cb_d = core_timing.RegisterEvent("callbackD", CallbackTemplate<3>);
-    Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", CallbackTemplate<4>);
-
-    // Enter slice 0
-    core_timing.Advance();
-
-    // D -> B -> C -> A -> E
-    core_timing.ScheduleEvent(1000, cb_a, CB_IDS[0]);
-    // Manually force since ScheduleEvent doesn't call it
-    core_timing.ForceExceptionCheck(1000);
-    REQUIRE(1000 == core_timing.GetDowncount());
-    core_timing.ScheduleEvent(500, cb_b, CB_IDS[1]);
-    // Manually force since ScheduleEvent doesn't call it
-    core_timing.ForceExceptionCheck(500);
-    REQUIRE(500 == core_timing.GetDowncount());
-    core_timing.ScheduleEvent(800, cb_c, CB_IDS[2]);
-    // Manually force since ScheduleEvent doesn't call it
-    core_timing.ForceExceptionCheck(800);
-    REQUIRE(500 == core_timing.GetDowncount());
-    core_timing.ScheduleEvent(100, cb_d, CB_IDS[3]);
-    // Manually force since ScheduleEvent doesn't call it
-    core_timing.ForceExceptionCheck(100);
-    REQUIRE(100 == core_timing.GetDowncount());
-    core_timing.ScheduleEvent(1200, cb_e, CB_IDS[4]);
-    // Manually force since ScheduleEvent doesn't call it
-    core_timing.ForceExceptionCheck(1200);
-    REQUIRE(100 == core_timing.GetDowncount());
-
-    AdvanceAndCheck(core_timing, 3, 400);
-    AdvanceAndCheck(core_timing, 1, 300);
-    AdvanceAndCheck(core_timing, 2, 200);
-    AdvanceAndCheck(core_timing, 0, 200);
-    AdvanceAndCheck(core_timing, 4, MAX_SLICE_LENGTH);
-}
-
-namespace SharedSlotTest {
-static unsigned int counter = 0;
-
-template <unsigned int ID>
-void FifoCallback(u64 userdata, s64 cycles_late) {
-    static_assert(ID < CB_IDS.size(), "ID out of range");
-    callbacks_ran_flags.set(ID);
-    REQUIRE(CB_IDS[ID] == userdata);
-    REQUIRE(ID == counter);
-    REQUIRE(lateness == cycles_late);
-    ++counter;
-}
-} // namespace SharedSlotTest
-
-TEST_CASE("CoreTiming[SharedSlot]", "[core]") {
-    using namespace SharedSlotTest;
+TEST_CASE("CoreTiming[FairSharing]", "[core]") {
 
     ScopeInit guard;
     auto& core_timing = guard.core_timing;
 
-    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", FifoCallback<0>);
-    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", FifoCallback<1>);
-    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", FifoCallback<2>);
-    Core::Timing::EventType* cb_d = core_timing.RegisterEvent("callbackD", FifoCallback<3>);
-    Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", FifoCallback<4>);
+    Core::Timing::EventType* empty_callback =
+        core_timing.RegisterEvent("empty_callback", EmptyCallback);
 
-    core_timing.ScheduleEvent(1000, cb_a, CB_IDS[0]);
-    core_timing.ScheduleEvent(1000, cb_b, CB_IDS[1]);
-    core_timing.ScheduleEvent(1000, cb_c, CB_IDS[2]);
-    core_timing.ScheduleEvent(1000, cb_d, CB_IDS[3]);
-    core_timing.ScheduleEvent(1000, cb_e, CB_IDS[4]);
+    callbacks_done = 0;
+    u64 MAX_CALLBACKS = 10;
+    for (std::size_t i = 0; i < 10; i++) {
+        core_timing.ScheduleEvent(i * 3333U, empty_callback, 0);
+    }
 
-    // Enter slice 0
-    core_timing.Advance();
-    REQUIRE(1000 == core_timing.GetDowncount());
+    const s64 advances = MAX_SLICE_LENGTH / 10;
+    core_timing.ResetRun();
+    u64 current_time = core_timing.GetTicks();
+    bool keep_running{};
+    do {
+        keep_running = false;
+        for (u32 active_core = 0; active_core < 4; ++active_core) {
+            core_timing.SwitchContext(active_core);
+            if (core_timing.CanCurrentContextRun()) {
+                core_timing.AddTicks(std::min<s64>(advances, core_timing.GetDowncount()));
+                core_timing.Advance();
+            }
+            keep_running |= core_timing.CanCurrentContextRun();
+        }
+    } while (keep_running);
+    u64 current_time_2 = core_timing.GetTicks();
 
-    callbacks_ran_flags = 0;
-    counter = 0;
-    lateness = 0;
-    core_timing.AddTicks(core_timing.GetDowncount());
-    core_timing.Advance();
-    REQUIRE(MAX_SLICE_LENGTH == core_timing.GetDowncount());
-    REQUIRE(0x1FULL == callbacks_ran_flags.to_ullong());
+    REQUIRE(MAX_CALLBACKS == callbacks_done);
+    REQUIRE(current_time_2 == current_time + MAX_SLICE_LENGTH * 4);
 }
 
 TEST_CASE("Core::Timing[PredictableLateness]", "[core]") {
@@ -180,13 +137,13 @@ TEST_CASE("Core::Timing[PredictableLateness]", "[core]") {
     Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
 
     // Enter slice 0
-    core_timing.Advance();
+    core_timing.ResetRun();
 
     core_timing.ScheduleEvent(100, cb_a, CB_IDS[0]);
     core_timing.ScheduleEvent(200, cb_b, CB_IDS[1]);
 
-    AdvanceAndCheck(core_timing, 0, 90, 10, -10); // (100 - 10)
-    AdvanceAndCheck(core_timing, 1, MAX_SLICE_LENGTH, 50, -50);
+    AdvanceAndCheck(core_timing, 0, 0, 10, -10); // (100 - 10)
+    AdvanceAndCheck(core_timing, 1, 1, 50, -50);
 }
 
 namespace ChainSchedulingTest {
@@ -220,7 +177,7 @@ TEST_CASE("CoreTiming[ChainScheduling]", "[core]") {
         });
 
     // Enter slice 0
-    core_timing.Advance();
+    core_timing.ResetRun();
 
     core_timing.ScheduleEvent(800, cb_a, CB_IDS[0]);
     core_timing.ScheduleEvent(1000, cb_b, CB_IDS[1]);
@@ -229,19 +186,19 @@ TEST_CASE("CoreTiming[ChainScheduling]", "[core]") {
     REQUIRE(800 == core_timing.GetDowncount());
 
     reschedules = 3;
-    AdvanceAndCheck(core_timing, 0, 200);  // cb_a
-    AdvanceAndCheck(core_timing, 1, 1000); // cb_b, cb_rs
+    AdvanceAndCheck(core_timing, 0, 0); // cb_a
+    AdvanceAndCheck(core_timing, 1, 1); // cb_b, cb_rs
     REQUIRE(2 == reschedules);
 
     core_timing.AddTicks(core_timing.GetDowncount());
     core_timing.Advance(); // cb_rs
+    core_timing.SwitchContext(3);
     REQUIRE(1 == reschedules);
     REQUIRE(200 == core_timing.GetDowncount());
 
-    AdvanceAndCheck(core_timing, 2, 800); // cb_c
+    AdvanceAndCheck(core_timing, 2, 3); // cb_c
 
     core_timing.AddTicks(core_timing.GetDowncount());
     core_timing.Advance(); // cb_rs
     REQUIRE(0 == reschedules);
-    REQUIRE(MAX_SLICE_LENGTH == core_timing.GetDowncount());
 }

src/video_core/CMakeLists.txt

@@ -6,6 +6,7 @@ add_library(video_core STATIC
     dma_pusher.h
     debug_utils/debug_utils.cpp
     debug_utils/debug_utils.h
+    engines/const_buffer_engine_interface.h
     engines/const_buffer_info.h
     engines/engine_upload.cpp
     engines/engine_upload.h
@@ -35,6 +36,8 @@ add_library(video_core STATIC
     memory_manager.h
     morton.cpp
     morton.h
+    rasterizer_accelerated.cpp
+    rasterizer_accelerated.h
     rasterizer_cache.cpp
     rasterizer_cache.h
     rasterizer_interface.h
@@ -107,10 +110,12 @@ add_library(video_core STATIC
     shader/decode/other.cpp
     shader/ast.cpp
     shader/ast.h
-    shader/control_flow.cpp
-    shader/control_flow.h
     shader/compiler_settings.cpp
     shader/compiler_settings.h
+    shader/const_buffer_locker.cpp
+    shader/const_buffer_locker.h
+    shader/control_flow.cpp
+    shader/control_flow.h
     shader/decode.cpp
     shader/expr.cpp
     shader/expr.h

src/video_core/engines/const_buffer_engine_interface.h

@@ -0,0 +1,119 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <type_traits>
+#include "common/bit_field.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/textures/texture.h"
+
+namespace Tegra::Engines {
+
+enum class ShaderType : u32 {
+    Vertex = 0,
+    TesselationControl = 1,
+    TesselationEval = 2,
+    Geometry = 3,
+    Fragment = 4,
+    Compute = 5,
+};
+
+struct SamplerDescriptor {
+    union {
+        BitField<0, 20, Tegra::Shader::TextureType> texture_type;
+        BitField<20, 1, u32> is_array;
+        BitField<21, 1, u32> is_buffer;
+        BitField<22, 1, u32> is_shadow;
+        u32 raw{};
+    };
+
+    bool operator==(const SamplerDescriptor& rhs) const noexcept {
+        return raw == rhs.raw;
+    }
+
+    bool operator!=(const SamplerDescriptor& rhs) const noexcept {
+        return !operator==(rhs);
+    }
+
+    static SamplerDescriptor FromTicTexture(Tegra::Texture::TextureType tic_texture_type) {
+        SamplerDescriptor result;
+        switch (tic_texture_type) {
+        case Tegra::Texture::TextureType::Texture1D:
+            result.texture_type.Assign(Tegra::Shader::TextureType::Texture1D);
+            result.is_array.Assign(0);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        case Tegra::Texture::TextureType::Texture2D:
+            result.texture_type.Assign(Tegra::Shader::TextureType::Texture2D);
+            result.is_array.Assign(0);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        case Tegra::Texture::TextureType::Texture3D:
+            result.texture_type.Assign(Tegra::Shader::TextureType::Texture3D);
+            result.is_array.Assign(0);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        case Tegra::Texture::TextureType::TextureCubemap:
+            result.texture_type.Assign(Tegra::Shader::TextureType::TextureCube);
+            result.is_array.Assign(0);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        case Tegra::Texture::TextureType::Texture1DArray:
+            result.texture_type.Assign(Tegra::Shader::TextureType::Texture1D);
+            result.is_array.Assign(1);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        case Tegra::Texture::TextureType::Texture2DArray:
+            result.texture_type.Assign(Tegra::Shader::TextureType::Texture2D);
+            result.is_array.Assign(1);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        case Tegra::Texture::TextureType::Texture1DBuffer:
+            result.texture_type.Assign(Tegra::Shader::TextureType::Texture1D);
+            result.is_array.Assign(0);
+            result.is_buffer.Assign(1);
+            result.is_shadow.Assign(0);
+            return result;
+        case Tegra::Texture::TextureType::Texture2DNoMipmap:
+            result.texture_type.Assign(Tegra::Shader::TextureType::Texture2D);
+            result.is_array.Assign(0);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        case Tegra::Texture::TextureType::TextureCubeArray:
+            result.texture_type.Assign(Tegra::Shader::TextureType::TextureCube);
+            result.is_array.Assign(1);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        default:
+            result.texture_type.Assign(Tegra::Shader::TextureType::Texture2D);
+            result.is_array.Assign(0);
+            result.is_buffer.Assign(0);
+            result.is_shadow.Assign(0);
+            return result;
+        }
+    }
+};
+static_assert(std::is_trivially_copyable_v<SamplerDescriptor>);
+
+class ConstBufferEngineInterface {
+public:
+    virtual ~ConstBufferEngineInterface() = default;
+    virtual u32 AccessConstBuffer32(ShaderType stage, u64 const_buffer, u64 offset) const = 0;
+    virtual SamplerDescriptor AccessBoundSampler(ShaderType stage, u64 offset) const = 0;
+    virtual SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer,
+                                                    u64 offset) const = 0;
+    virtual u32 GetBoundBuffer() const = 0;
+};
+
+} // namespace Tegra::Engines

src/video_core/engines/kepler_compute.cpp

@@ -50,7 +50,7 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
-Tegra::Texture::FullTextureInfo KeplerCompute::GetTexture(std::size_t offset) const {
+Texture::FullTextureInfo KeplerCompute::GetTexture(std::size_t offset) const {
     const std::bitset<8> cbuf_mask = launch_description.const_buffer_enable_mask.Value();
     ASSERT(cbuf_mask[regs.tex_cb_index]);
 
@@ -61,22 +61,38 @@ Tegra::Texture::FullTextureInfo KeplerCompute::GetTexture(std::size_t offset) co
     ASSERT(address < texinfo.Address() + texinfo.size);
 
     const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(address)};
-    return GetTextureInfo(tex_handle, offset);
+    return GetTextureInfo(tex_handle);
 }
 
-Texture::FullTextureInfo KeplerCompute::GetTextureInfo(const Texture::TextureHandle tex_handle,
-                                                       std::size_t offset) const {
-    return Texture::FullTextureInfo{static_cast<u32>(offset), GetTICEntry(tex_handle.tic_id),
-                                    GetTSCEntry(tex_handle.tsc_id)};
+Texture::FullTextureInfo KeplerCompute::GetTextureInfo(Texture::TextureHandle tex_handle) const {
+    return Texture::FullTextureInfo{GetTICEntry(tex_handle.tic_id), GetTSCEntry(tex_handle.tsc_id)};
 }
 
-u32 KeplerCompute::AccessConstBuffer32(u64 const_buffer, u64 offset) const {
+u32 KeplerCompute::AccessConstBuffer32(ShaderType stage, u64 const_buffer, u64 offset) const {
+    ASSERT(stage == ShaderType::Compute);
     const auto& buffer = launch_description.const_buffer_config[const_buffer];
     u32 result;
     std::memcpy(&result, memory_manager.GetPointer(buffer.Address() + offset), sizeof(u32));
     return result;
 }
 
+SamplerDescriptor KeplerCompute::AccessBoundSampler(ShaderType stage, u64 offset) const {
+    return AccessBindlessSampler(stage, regs.tex_cb_index, offset * sizeof(Texture::TextureHandle));
+}
+
+SamplerDescriptor KeplerCompute::AccessBindlessSampler(ShaderType stage, u64 const_buffer,
+                                                       u64 offset) const {
+    ASSERT(stage == ShaderType::Compute);
+    const auto& tex_info_buffer = launch_description.const_buffer_config[const_buffer];
+    const GPUVAddr tex_info_address = tex_info_buffer.Address() + offset;
+
+    const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
+    const Texture::FullTextureInfo tex_info = GetTextureInfo(tex_handle);
+    SamplerDescriptor result = SamplerDescriptor::FromTicTexture(tex_info.tic.texture_type.Value());
+    result.is_shadow.Assign(tex_info.tsc.depth_compare_enabled.Value());
+    return result;
+}
+
 void KeplerCompute::ProcessLaunch() {
     const GPUVAddr launch_desc_loc = regs.launch_desc_loc.Address();
     memory_manager.ReadBlockUnsafe(launch_desc_loc, &launch_description,

src/video_core/engines/kepler_compute.h

@@ -10,6 +10,7 @@
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
+#include "video_core/engines/const_buffer_engine_interface.h"
 #include "video_core/engines/engine_upload.h"
 #include "video_core/gpu.h"
 #include "video_core/textures/texture.h"
@@ -37,7 +38,7 @@ namespace Tegra::Engines {
 #define KEPLER_COMPUTE_REG_INDEX(field_name)                                                       \
     (offsetof(Tegra::Engines::KeplerCompute::Regs, field_name) / sizeof(u32))
 
-class KeplerCompute final {
+class KeplerCompute final : public ConstBufferEngineInterface {
 public:
     explicit KeplerCompute(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                            MemoryManager& memory_manager);
@@ -195,13 +196,21 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
-    Tegra::Texture::FullTextureInfo GetTexture(std::size_t offset) const;
+    Texture::FullTextureInfo GetTexture(std::size_t offset) const;
 
-    /// Given a Texture Handle, returns the TSC and TIC entries.
-    Texture::FullTextureInfo GetTextureInfo(const Texture::TextureHandle tex_handle,
-                                            std::size_t offset) const;
+    /// Given a texture handle, returns the TSC and TIC entries.
+    Texture::FullTextureInfo GetTextureInfo(Texture::TextureHandle tex_handle) const;
 
-    u32 AccessConstBuffer32(u64 const_buffer, u64 offset) const;
+    u32 AccessConstBuffer32(ShaderType stage, u64 const_buffer, u64 offset) const override;
+
+    SamplerDescriptor AccessBoundSampler(ShaderType stage, u64 offset) const override;
+
+    SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer,
+                                            u64 offset) const override;
+
+    u32 GetBoundBuffer() const override {
+        return regs.tex_cb_index;
+    }
 
 private:
     Core::System& system;

src/video_core/engines/maxwell_3d.cpp

@@ -98,10 +98,10 @@ void Maxwell3D::InitializeRegisterDefaults() {
     mme_inline[MAXWELL3D_REG_INDEX(index_array.count)] = true;
 }
 
-#define DIRTY_REGS_POS(field_name) (offsetof(Maxwell3D::DirtyRegs, field_name))
+#define DIRTY_REGS_POS(field_name) static_cast<u8>(offsetof(Maxwell3D::DirtyRegs, field_name))
 
 void Maxwell3D::InitDirtySettings() {
-    const auto set_block = [this](const u32 start, const u32 range, const u8 position) {
+    const auto set_block = [this](std::size_t start, std::size_t range, u8 position) {
         const auto start_itr = dirty_pointers.begin() + start;
         const auto end_itr = start_itr + range;
         std::fill(start_itr, end_itr, position);
@@ -112,10 +112,10 @@ void Maxwell3D::InitDirtySettings() {
     constexpr u32 registers_per_rt = sizeof(regs.rt[0]) / sizeof(u32);
     constexpr u32 rt_start_reg = MAXWELL3D_REG_INDEX(rt);
     constexpr u32 rt_end_reg = rt_start_reg + registers_per_rt * 8;
-    u32 rt_dirty_reg = DIRTY_REGS_POS(render_target);
+    u8 rt_dirty_reg = DIRTY_REGS_POS(render_target);
     for (u32 rt_reg = rt_start_reg; rt_reg < rt_end_reg; rt_reg += registers_per_rt) {
         set_block(rt_reg, registers_per_rt, rt_dirty_reg);
-        rt_dirty_reg++;
+        ++rt_dirty_reg;
     }
     constexpr u32 depth_buffer_flag = DIRTY_REGS_POS(depth_buffer);
     dirty_pointers[MAXWELL3D_REG_INDEX(zeta_enable)] = depth_buffer_flag;
@@ -129,35 +129,35 @@ void Maxwell3D::InitDirtySettings() {
     constexpr u32 vertex_array_start = MAXWELL3D_REG_INDEX(vertex_array);
     constexpr u32 vertex_array_size = sizeof(regs.vertex_array[0]) / sizeof(u32);
     constexpr u32 vertex_array_end = vertex_array_start + vertex_array_size * Regs::NumVertexArrays;
-    u32 va_reg = DIRTY_REGS_POS(vertex_array);
-    u32 vi_reg = DIRTY_REGS_POS(vertex_instance);
+    u8 va_dirty_reg = DIRTY_REGS_POS(vertex_array);
+    u8 vi_dirty_reg = DIRTY_REGS_POS(vertex_instance);
     for (u32 vertex_reg = vertex_array_start; vertex_reg < vertex_array_end;
          vertex_reg += vertex_array_size) {
-        set_block(vertex_reg, 3, va_reg);
+        set_block(vertex_reg, 3, va_dirty_reg);
         // The divisor concerns vertex array instances
-        dirty_pointers[vertex_reg + 3] = vi_reg;
-        va_reg++;
-        vi_reg++;
+        dirty_pointers[static_cast<std::size_t>(vertex_reg) + 3] = vi_dirty_reg;
+        ++va_dirty_reg;
+        ++vi_dirty_reg;
     }
     constexpr u32 vertex_limit_start = MAXWELL3D_REG_INDEX(vertex_array_limit);
     constexpr u32 vertex_limit_size = sizeof(regs.vertex_array_limit[0]) / sizeof(u32);
     constexpr u32 vertex_limit_end = vertex_limit_start + vertex_limit_size * Regs::NumVertexArrays;
-    va_reg = DIRTY_REGS_POS(vertex_array);
+    va_dirty_reg = DIRTY_REGS_POS(vertex_array);
     for (u32 vertex_reg = vertex_limit_start; vertex_reg < vertex_limit_end;
          vertex_reg += vertex_limit_size) {
-        set_block(vertex_reg, vertex_limit_size, va_reg);
-        va_reg++;
+        set_block(vertex_reg, vertex_limit_size, va_dirty_reg);
+        va_dirty_reg++;
     }
     constexpr u32 vertex_instance_start = MAXWELL3D_REG_INDEX(instanced_arrays);
     constexpr u32 vertex_instance_size =
         sizeof(regs.instanced_arrays.is_instanced[0]) / sizeof(u32);
     constexpr u32 vertex_instance_end =
         vertex_instance_start + vertex_instance_size * Regs::NumVertexArrays;
-    vi_reg = DIRTY_REGS_POS(vertex_instance);
+    vi_dirty_reg = DIRTY_REGS_POS(vertex_instance);
     for (u32 vertex_reg = vertex_instance_start; vertex_reg < vertex_instance_end;
          vertex_reg += vertex_instance_size) {
-        set_block(vertex_reg, vertex_instance_size, vi_reg);
-        vi_reg++;
+        set_block(vertex_reg, vertex_instance_size, vi_dirty_reg);
+        vi_dirty_reg++;
     }
     set_block(MAXWELL3D_REG_INDEX(vertex_attrib_format), regs.vertex_attrib_format.size(),
               DIRTY_REGS_POS(vertex_attrib_format));
@@ -171,7 +171,7 @@ void Maxwell3D::InitDirtySettings() {
     // State
 
     // Viewport
-    constexpr u32 viewport_dirty_reg = DIRTY_REGS_POS(viewport);
+    constexpr u8 viewport_dirty_reg = DIRTY_REGS_POS(viewport);
     constexpr u32 viewport_start = MAXWELL3D_REG_INDEX(viewports);
     constexpr u32 viewport_size = sizeof(regs.viewports) / sizeof(u32);
     set_block(viewport_start, viewport_size, viewport_dirty_reg);
@@ -198,7 +198,7 @@ void Maxwell3D::InitDirtySettings() {
     set_block(primitive_restart_start, primitive_restart_size, DIRTY_REGS_POS(primitive_restart));
 
     // Depth Test
-    constexpr u32 depth_test_dirty_reg = DIRTY_REGS_POS(depth_test);
+    constexpr u8 depth_test_dirty_reg = DIRTY_REGS_POS(depth_test);
     dirty_pointers[MAXWELL3D_REG_INDEX(depth_test_enable)] = depth_test_dirty_reg;
     dirty_pointers[MAXWELL3D_REG_INDEX(depth_write_enabled)] = depth_test_dirty_reg;
     dirty_pointers[MAXWELL3D_REG_INDEX(depth_test_func)] = depth_test_dirty_reg;
@@ -223,12 +223,12 @@ void Maxwell3D::InitDirtySettings() {
     dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_mask)] = stencil_test_dirty_reg;
 
     // Color Mask
-    constexpr u32 color_mask_dirty_reg = DIRTY_REGS_POS(color_mask);
+    constexpr u8 color_mask_dirty_reg = DIRTY_REGS_POS(color_mask);
     dirty_pointers[MAXWELL3D_REG_INDEX(color_mask_common)] = color_mask_dirty_reg;
     set_block(MAXWELL3D_REG_INDEX(color_mask), sizeof(regs.color_mask) / sizeof(u32),
               color_mask_dirty_reg);
     // Blend State
-    constexpr u32 blend_state_dirty_reg = DIRTY_REGS_POS(blend_state);
+    constexpr u8 blend_state_dirty_reg = DIRTY_REGS_POS(blend_state);
     set_block(MAXWELL3D_REG_INDEX(blend_color), sizeof(regs.blend_color) / sizeof(u32),
               blend_state_dirty_reg);
     dirty_pointers[MAXWELL3D_REG_INDEX(independent_blend_enable)] = blend_state_dirty_reg;
@@ -237,18 +237,23 @@ void Maxwell3D::InitDirtySettings() {
               blend_state_dirty_reg);
 
     // Scissor State
-    constexpr u32 scissor_test_dirty_reg = DIRTY_REGS_POS(scissor_test);
+    constexpr u8 scissor_test_dirty_reg = DIRTY_REGS_POS(scissor_test);
     set_block(MAXWELL3D_REG_INDEX(scissor_test), sizeof(regs.scissor_test) / sizeof(u32),
               scissor_test_dirty_reg);
 
     // Polygon Offset
-    constexpr u32 polygon_offset_dirty_reg = DIRTY_REGS_POS(polygon_offset);
+    constexpr u8 polygon_offset_dirty_reg = DIRTY_REGS_POS(polygon_offset);
     dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_fill_enable)] = polygon_offset_dirty_reg;
     dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_line_enable)] = polygon_offset_dirty_reg;
     dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_point_enable)] = polygon_offset_dirty_reg;
     dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_units)] = polygon_offset_dirty_reg;
     dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_factor)] = polygon_offset_dirty_reg;
     dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_clamp)] = polygon_offset_dirty_reg;
+
+    // Depth bounds
+    constexpr u8 depth_bounds_values_dirty_reg = DIRTY_REGS_POS(depth_bounds_values);
+    dirty_pointers[MAXWELL3D_REG_INDEX(depth_bounds[0])] = depth_bounds_values_dirty_reg;
+    dirty_pointers[MAXWELL3D_REG_INDEX(depth_bounds[1])] = depth_bounds_values_dirty_reg;
 }
 
 void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters) {
@@ -473,7 +478,7 @@ void Maxwell3D::CallMethodFromMME(const GPU::MethodCall& method_call) {
 }
 
 void Maxwell3D::FlushMMEInlineDraw() {
-    LOG_DEBUG(HW_GPU, "called, topology={}, count={}", static_cast<u32>(regs.draw.topology.Value()),
+    LOG_TRACE(HW_GPU, "called, topology={}, count={}", static_cast<u32>(regs.draw.topology.Value()),
               regs.vertex_buffer.count);
     ASSERT_MSG(!(regs.index_array.count && regs.vertex_buffer.count), "Both indexed and direct?");
     ASSERT(mme_draw.instance_count == mme_draw.gl_end_count);
@@ -755,61 +760,8 @@ Texture::TSCEntry Maxwell3D::GetTSCEntry(u32 tsc_index) const {
     return tsc_entry;
 }
 
-std::vector<Texture::FullTextureInfo> Maxwell3D::GetStageTextures(Regs::ShaderStage stage) const {
-    std::vector<Texture::FullTextureInfo> textures;
-
-    auto& fragment_shader = state.shader_stages[static_cast<std::size_t>(stage)];
-    auto& tex_info_buffer = fragment_shader.const_buffers[regs.tex_cb_index];
-    ASSERT(tex_info_buffer.enabled && tex_info_buffer.address != 0);
-
-    GPUVAddr tex_info_buffer_end = tex_info_buffer.address + tex_info_buffer.size;
-
-    // Offset into the texture constbuffer where the texture info begins.
-    static constexpr std::size_t TextureInfoOffset = 0x20;
-
-    for (GPUVAddr current_texture = tex_info_buffer.address + TextureInfoOffset;
-         current_texture < tex_info_buffer_end; current_texture += sizeof(Texture::TextureHandle)) {
-
-        const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(current_texture)};
-
-        Texture::FullTextureInfo tex_info{};
-        // TODO(Subv): Use the shader to determine which textures are actually accessed.
-        tex_info.index =
-            static_cast<u32>(current_texture - tex_info_buffer.address - TextureInfoOffset) /
-            sizeof(Texture::TextureHandle);
-
-        // Load the TIC data.
-        auto tic_entry = GetTICEntry(tex_handle.tic_id);
-        // TODO(Subv): Workaround for BitField's move constructor being deleted.
-        std::memcpy(&tex_info.tic, &tic_entry, sizeof(tic_entry));
-
-        // Load the TSC data
-        auto tsc_entry = GetTSCEntry(tex_handle.tsc_id);
-        // TODO(Subv): Workaround for BitField's move constructor being deleted.
-        std::memcpy(&tex_info.tsc, &tsc_entry, sizeof(tsc_entry));
-
-        textures.push_back(tex_info);
-    }
-
-    return textures;
-}
-
-Texture::FullTextureInfo Maxwell3D::GetTextureInfo(const Texture::TextureHandle tex_handle,
-                                                   std::size_t offset) const {
-    Texture::FullTextureInfo tex_info{};
-    tex_info.index = static_cast<u32>(offset);
-
-    // Load the TIC data.
-    auto tic_entry = GetTICEntry(tex_handle.tic_id);
-    // TODO(Subv): Workaround for BitField's move constructor being deleted.
-    std::memcpy(&tex_info.tic, &tic_entry, sizeof(tic_entry));
-
-    // Load the TSC data
-    auto tsc_entry = GetTSCEntry(tex_handle.tsc_id);
-    // TODO(Subv): Workaround for BitField's move constructor being deleted.
-    std::memcpy(&tex_info.tsc, &tsc_entry, sizeof(tsc_entry));
-
-    return tex_info;
+Texture::FullTextureInfo Maxwell3D::GetTextureInfo(Texture::TextureHandle tex_handle) const {
+    return Texture::FullTextureInfo{GetTICEntry(tex_handle.tic_id), GetTSCEntry(tex_handle.tsc_id)};
 }
 
 Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage,
@@ -825,7 +777,7 @@ Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage,
 
     const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
 
-    return GetTextureInfo(tex_handle, offset);
+    return GetTextureInfo(tex_handle);
 }
 
 u32 Maxwell3D::GetRegisterValue(u32 method) const {
@@ -841,7 +793,8 @@ void Maxwell3D::ProcessClearBuffers() {
     rasterizer.Clear();
 }
 
-u32 Maxwell3D::AccessConstBuffer32(Regs::ShaderStage stage, u64 const_buffer, u64 offset) const {
+u32 Maxwell3D::AccessConstBuffer32(ShaderType stage, u64 const_buffer, u64 offset) const {
+    ASSERT(stage != ShaderType::Compute);
     const auto& shader_stage = state.shader_stages[static_cast<std::size_t>(stage)];
     const auto& buffer = shader_stage.const_buffers[const_buffer];
     u32 result;
@@ -849,4 +802,22 @@ u32 Maxwell3D::AccessConstBuffer32(Regs::ShaderStage stage, u64 const_buffer, u6
     return result;
 }
 
+SamplerDescriptor Maxwell3D::AccessBoundSampler(ShaderType stage, u64 offset) const {
+    return AccessBindlessSampler(stage, regs.tex_cb_index, offset * sizeof(Texture::TextureHandle));
+}
+
+SamplerDescriptor Maxwell3D::AccessBindlessSampler(ShaderType stage, u64 const_buffer,
+                                                   u64 offset) const {
+    ASSERT(stage != ShaderType::Compute);
+    const auto& shader = state.shader_stages[static_cast<std::size_t>(stage)];
+    const auto& tex_info_buffer = shader.const_buffers[const_buffer];
+    const GPUVAddr tex_info_address = tex_info_buffer.address + offset;
+
+    const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
+    const Texture::FullTextureInfo tex_info = GetTextureInfo(tex_handle);
+    SamplerDescriptor result = SamplerDescriptor::FromTicTexture(tex_info.tic.texture_type.Value());
+    result.is_shadow.Assign(tex_info.tsc.depth_compare_enabled.Value());
+    return result;
+}
+
 } // namespace Tegra::Engines

src/video_core/engines/maxwell_3d.h

@@ -15,6 +15,7 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/math_util.h"
+#include "video_core/engines/const_buffer_engine_interface.h"
 #include "video_core/engines/const_buffer_info.h"
 #include "video_core/engines/engine_upload.h"
 #include "video_core/gpu.h"
@@ -44,7 +45,7 @@ namespace Tegra::Engines {
 #define MAXWELL3D_REG_INDEX(field_name)                                                            \
     (offsetof(Tegra::Engines::Maxwell3D::Regs, field_name) / sizeof(u32))
 
-class Maxwell3D final {
+class Maxwell3D final : public ConstBufferEngineInterface {
 public:
     explicit Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                        MemoryManager& memory_manager);
@@ -687,7 +688,9 @@ public:
 
                 u32 rt_separate_frag_data;
 
-                INSERT_PADDING_WORDS(0xC);
+                f32 depth_bounds[2];
+
+                INSERT_PADDING_WORDS(0xA);
 
                 struct {
                     u32 address_high;
@@ -1163,6 +1166,8 @@ public:
 
     struct DirtyRegs {
         static constexpr std::size_t NUM_REGS = 256;
+        static_assert(NUM_REGS - 1 <= std::numeric_limits<u8>::max());
+
         union {
             struct {
                 bool null_dirty;
@@ -1201,6 +1206,7 @@ public:
                 bool transform_feedback;
                 bool color_mask;
                 bool polygon_offset;
+                bool depth_bounds_values;
 
                 // Complementary
                 bool viewport_transform;
@@ -1244,17 +1250,22 @@ public:
 
     void FlushMMEInlineDraw();
 
-    /// Given a Texture Handle, returns the TSC and TIC entries.
-    Texture::FullTextureInfo GetTextureInfo(const Texture::TextureHandle tex_handle,
-                                            std::size_t offset) const;
-
-    /// Returns a list of enabled textures for the specified shader stage.
-    std::vector<Texture::FullTextureInfo> GetStageTextures(Regs::ShaderStage stage) const;
+    /// Given a texture handle, returns the TSC and TIC entries.
+    Texture::FullTextureInfo GetTextureInfo(Texture::TextureHandle tex_handle) const;
 
     /// Returns the texture information for a specific texture in a specific shader stage.
     Texture::FullTextureInfo GetStageTexture(Regs::ShaderStage stage, std::size_t offset) const;
 
-    u32 AccessConstBuffer32(Regs::ShaderStage stage, u64 const_buffer, u64 offset) const;
+    u32 AccessConstBuffer32(ShaderType stage, u64 const_buffer, u64 offset) const override;
+
+    SamplerDescriptor AccessBoundSampler(ShaderType stage, u64 offset) const override;
+
+    SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer,
+                                            u64 offset) const override;
+
+    u32 GetBoundBuffer() const override {
+        return regs.tex_cb_index;
+    }
 
     /// Memory for macro code - it's undetermined how big this is, however 1MB is much larger than
     /// we've seen used.
@@ -1400,6 +1411,7 @@ ASSERT_REG_POSITION(stencil_back_mask, 0x3D6);
 ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7);
 ASSERT_REG_POSITION(color_mask_common, 0x3E4);
 ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
+ASSERT_REG_POSITION(depth_bounds, 0x3EC);
 ASSERT_REG_POSITION(zeta, 0x3F8);
 ASSERT_REG_POSITION(clear_flags, 0x43E);
 ASSERT_REG_POSITION(vertex_attrib_format, 0x458);

src/video_core/engines/shader_bytecode.h

@@ -574,7 +574,7 @@ enum class ShuffleOperation : u64 {
 };
 
 union Instruction {
-    Instruction& operator=(const Instruction& instr) {
+    constexpr Instruction& operator=(const Instruction& instr) {
         value = instr.value;
         return *this;
     }
@@ -1237,6 +1237,32 @@ union Instruction {
         }
     } tld4;
 
+    union {
+        BitField<35, 1, u64> ndv_flag;
+        BitField<49, 1, u64> nodep_flag;
+        BitField<50, 1, u64> dc_flag;
+        BitField<33, 2, u64> info;
+        BitField<37, 2, u64> component;
+
+        bool UsesMiscMode(TextureMiscMode mode) const {
+            switch (mode) {
+            case TextureMiscMode::NDV:
+                return ndv_flag != 0;
+            case TextureMiscMode::NODEP:
+                return nodep_flag != 0;
+            case TextureMiscMode::DC:
+                return dc_flag != 0;
+            case TextureMiscMode::AOFFI:
+                return info == 1;
+            case TextureMiscMode::PTP:
+                return info == 2;
+            default:
+                break;
+            }
+            return false;
+        }
+    } tld4_b;
+
     union {
         BitField<49, 1, u64> nodep_flag;
         BitField<50, 1, u64> dc_flag;
@@ -1590,7 +1616,8 @@ public:
         TEXS,   // Texture Fetch with scalar/non-vec4 source/destinations
         TLD,    // Texture Load
         TLDS,   // Texture Load with scalar/non-vec4 source/destinations
-        TLD4,   // Texture Load 4
+        TLD4,   // Texture Gather 4
+        TLD4_B, // Texture Gather 4 Bindless
         TLD4S,  // Texture Load 4 with scalar / non - vec4 source / destinations
         TMML_B, // Texture Mip Map Level
         TMML,   // Texture Mip Map Level
@@ -1760,22 +1787,22 @@ public:
 
     class Matcher {
     public:
-        Matcher(const char* const name, u16 mask, u16 expected, OpCode::Id id, OpCode::Type type)
+        constexpr Matcher(const char* const name, u16 mask, u16 expected, Id id, Type type)
             : name{name}, mask{mask}, expected{expected}, id{id}, type{type} {}
 
-        const char* GetName() const {
+        constexpr const char* GetName() const {
             return name;
         }
 
-        u16 GetMask() const {
+        constexpr u16 GetMask() const {
             return mask;
         }
 
-        Id GetId() const {
+        constexpr Id GetId() const {
             return id;
         }
 
-        Type GetType() const {
+        constexpr Type GetType() const {
             return type;
         }
 
@@ -1784,7 +1811,7 @@ public:
          * @param instruction The instruction to test
          * @returns true if the given instruction matches.
          */
-        bool Matches(u16 instruction) const {
+        constexpr bool Matches(u16 instruction) const {
             return (instruction & mask) == expected;
         }
 
@@ -1818,7 +1845,7 @@ private:
          * A '0' in a bitstring indicates that a zero must be present at that bit position.
          * A '1' in a bitstring indicates that a one must be present at that bit position.
          */
-        static auto GetMaskAndExpect(const char* const bitstring) {
+        static constexpr auto GetMaskAndExpect(const char* const bitstring) {
             u16 mask = 0, expect = 0;
             for (std::size_t i = 0; i < opcode_bitsize; i++) {
                 const std::size_t bit_position = opcode_bitsize - i - 1;
@@ -1835,15 +1862,15 @@ private:
                     break;
                 }
             }
-            return std::make_tuple(mask, expect);
+            return std::make_pair(mask, expect);
         }
 
     public:
         /// Creates a matcher that can match and parse instructions based on bitstring.
-        static auto GetMatcher(const char* const bitstring, OpCode::Id op, OpCode::Type type,
-                               const char* const name) {
-            const auto mask_expect = GetMaskAndExpect(bitstring);
-            return Matcher(name, std::get<0>(mask_expect), std::get<1>(mask_expect), op, type);
+        static constexpr auto GetMatcher(const char* const bitstring, Id op, Type type,
+                                         const char* const name) {
+            const auto [mask, expected] = GetMaskAndExpect(bitstring);
+            return Matcher(name, mask, expected, op, type);
         }
     };
 
@@ -1881,6 +1908,7 @@ private:
             INST("11011100--11----", Id::TLD, Type::Texture, "TLD"),
             INST("1101-01---------", Id::TLDS, Type::Texture, "TLDS"),
             INST("110010----111---", Id::TLD4, Type::Texture, "TLD4"),
+            INST("1101111011111---", Id::TLD4_B, Type::Texture, "TLD4_B"),
             INST("1101111100------", Id::TLD4S, Type::Texture, "TLD4S"),
             INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"),
             INST("1101111101011---", Id::TMML, Type::Texture, "TMML"),

src/video_core/gpu.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include "common/assert.h"
+#include "common/microprofile.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/memory.h"
@@ -17,6 +18,8 @@
 
 namespace Tegra {
 
+MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192));
+
 GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async)
     : system{system}, renderer{renderer}, is_async{is_async} {
     auto& rasterizer{renderer.Rasterizer()};
@@ -63,6 +66,16 @@ const DmaPusher& GPU::DmaPusher() const {
     return *dma_pusher;
 }
 
+void GPU::WaitFence(u32 syncpoint_id, u32 value) const {
+    // Synced GPU, is always in sync
+    if (!is_async) {
+        return;
+    }
+    MICROPROFILE_SCOPE(GPU_wait);
+    while (syncpoints[syncpoint_id].load(std::memory_order_relaxed) < value) {
+    }
+}
+
 void GPU::IncrementSyncPoint(const u32 syncpoint_id) {
     syncpoints[syncpoint_id]++;
     std::lock_guard lock{sync_mutex};
@@ -326,7 +339,7 @@ void GPU::ProcessSemaphoreTriggerMethod() {
         block.sequence = regs.semaphore_sequence;
         // TODO(Kmather73): Generate a real GPU timestamp and write it here instead of
         // CoreTiming
-        block.timestamp = Core::System::GetInstance().CoreTiming().GetTicks();
+        block.timestamp = system.CoreTiming().GetTicks();
         memory_manager->WriteBlock(regs.semaphore_address.SemaphoreAddress(), &block,
                                    sizeof(block));
     } else {

src/video_core/gpu.h

@@ -177,6 +177,12 @@ public:
     /// Returns a reference to the GPU DMA pusher.
     Tegra::DmaPusher& DmaPusher();
 
+    // Waits for the GPU to finish working
+    virtual void WaitIdle() const = 0;
+
+    /// Allows the CPU/NvFlinger to wait on the GPU before presenting a frame.
+    void WaitFence(u32 syncpoint_id, u32 value) const;
+
     void IncrementSyncPoint(u32 syncpoint_id);
 
     u32 GetSyncpointValue(u32 syncpoint_id) const;