hle: kernel: Invalidate icache in UnmapProcessMemory (fixes #8174)

service: hid: Partially revert #8123

.gitmodules

@@ -7,9 +7,6 @@
 [submodule "dynarmic"]
     path = externals/dynarmic
     url = https://github.com/MerryMage/dynarmic.git
-[submodule "soundtouch"]
-    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

externals/CMakeLists.txt

@@ -68,9 +68,6 @@ if (YUZU_USE_EXTERNAL_SDL2)
     add_library(SDL2 ALIAS SDL2-static)
 endif()
 
-# SoundTouch
-add_subdirectory(soundtouch)
-
 # Cubeb
 if(ENABLE_CUBEB)
     set(BUILD_TESTS OFF CACHE BOOL "")

src/CMakeLists.txt

@@ -103,12 +103,6 @@ else()
         -Wno-unused-parameter
     )
 
-    # TODO: Remove when we update to a GCC compiler that enables this
-    #       by default (i.e. GCC 10 or newer).
-    if (CMAKE_CXX_COMPILER_ID STREQUAL GNU)
-        add_compile_options(-fconcepts)
-    endif()
-
     if (ARCHITECTURE_x86_64)
         add_compile_options("-mcx16")
     endif()

src/audio_core/CMakeLists.txt

@@ -36,8 +36,6 @@ add_library(audio_core STATIC
     splitter_context.h
     stream.cpp
     stream.h
-    time_stretch.cpp
-    time_stretch.h
     voice_context.cpp
     voice_context.h
 
@@ -63,7 +61,6 @@ if (NOT MSVC)
 endif()
 
 target_link_libraries(audio_core PUBLIC common core)
-target_link_libraries(audio_core PRIVATE SoundTouch)
 
 if(ENABLE_CUBEB)
     target_link_libraries(audio_core PRIVATE cubeb)

src/audio_core/cubeb_sink.cpp

@@ -7,7 +7,6 @@
 #include <cstring>
 #include "audio_core/cubeb_sink.h"
 #include "audio_core/stream.h"
-#include "audio_core/time_stretch.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/ring_buffer.h"
@@ -23,8 +22,7 @@ class CubebSinkStream final : public SinkStream {
 public:
     CubebSinkStream(cubeb* ctx_, u32 sample_rate, u32 num_channels_, cubeb_devid output_device,
                     const std::string& name)
-        : ctx{ctx_}, num_channels{std::min(num_channels_, 6u)}, time_stretch{sample_rate,
-                                                                             num_channels} {
+        : ctx{ctx_}, num_channels{std::min(num_channels_, 6u)} {
 
         cubeb_stream_params params{};
         params.rate = sample_rate;
@@ -131,7 +129,6 @@ private:
     Common::RingBuffer<s16, 0x10000> queue;
     std::array<s16, 2> last_frame{};
     std::atomic<bool> should_flush{};
-    TimeStretcher time_stretch;
 
     static long DataCallback(cubeb_stream* stream, void* user_data, const void* input_buffer,
                              void* output_buffer, long num_frames);
@@ -205,25 +202,7 @@ long CubebSinkStream::DataCallback([[maybe_unused]] cubeb_stream* stream, void*
 
     const std::size_t num_channels = impl->GetNumChannels();
     const std::size_t samples_to_write = num_channels * num_frames;
-    std::size_t samples_written;
-
-    /*
-    if (Settings::values.enable_audio_stretching.GetValue()) {
-        const std::vector<s16> in{impl->queue.Pop()};
-        const std::size_t num_in{in.size() / num_channels};
-        s16* const out{reinterpret_cast<s16*>(buffer)};
-        const std::size_t out_frames =
-            impl->time_stretch.Process(in.data(), num_in, out, num_frames);
-        samples_written = out_frames * num_channels;
-
-        if (impl->should_flush) {
-            impl->time_stretch.Flush();
-            impl->should_flush = false;
-        }
-    } else {
-        samples_written = impl->queue.Pop(buffer, samples_to_write);
-    }*/
-    samples_written = impl->queue.Pop(buffer, samples_to_write);
+    const std::size_t samples_written = impl->queue.Pop(buffer, samples_to_write);
 
     if (samples_written >= num_channels) {
         std::memcpy(&impl->last_frame[0], buffer + (samples_written - num_channels) * sizeof(s16),

src/audio_core/sdl2_sink.cpp

@@ -7,7 +7,6 @@
 #include <cstring>
 #include "audio_core/sdl2_sink.h"
 #include "audio_core/stream.h"
-#include "audio_core/time_stretch.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
 //#include "common/settings.h"
@@ -27,7 +26,7 @@ namespace AudioCore {
 class SDLSinkStream final : public SinkStream {
 public:
     SDLSinkStream(u32 sample_rate, u32 num_channels_, const std::string& output_device)
-        : num_channels{std::min(num_channels_, 6u)}, time_stretch{sample_rate, num_channels} {
+        : num_channels{std::min(num_channels_, 6u)} {
 
         SDL_AudioSpec spec;
         spec.freq = sample_rate;
@@ -116,7 +115,6 @@ private:
     SDL_AudioDeviceID dev = 0;
     u32 num_channels{};
     std::atomic<bool> should_flush{};
-    TimeStretcher time_stretch;
 };
 
 SDLSink::SDLSink(std::string_view target_device_name) {

src/audio_core/time_stretch.cpp

@@ -1,68 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <cmath>
-#include <cstddef>
-#include "audio_core/time_stretch.h"
-#include "common/logging/log.h"
-
-namespace AudioCore {
-
-TimeStretcher::TimeStretcher(u32 sample_rate, u32 channel_count) : m_sample_rate{sample_rate} {
-    m_sound_touch.setChannels(channel_count);
-    m_sound_touch.setSampleRate(sample_rate);
-    m_sound_touch.setPitch(1.0);
-    m_sound_touch.setTempo(1.0);
-}
-
-void TimeStretcher::Clear() {
-    m_sound_touch.clear();
-}
-
-void TimeStretcher::Flush() {
-    m_sound_touch.flush();
-}
-
-std::size_t TimeStretcher::Process(const s16* in, std::size_t num_in, s16* out,
-                                   std::size_t num_out) {
-    const double time_delta = static_cast<double>(num_out) / m_sample_rate; // seconds
-
-    // We were given actual_samples number of samples, and num_samples were requested from us.
-    double current_ratio = static_cast<double>(num_in) / static_cast<double>(num_out);
-
-    const double max_latency = 0.25; // seconds
-    const double max_backlog = m_sample_rate * max_latency;
-    const double backlog_fullness = m_sound_touch.numSamples() / max_backlog;
-    if (backlog_fullness > 4.0) {
-        // Too many samples in backlog: Don't push anymore on
-        num_in = 0;
-    }
-
-    // We ideally want the backlog to be about 50% full.
-    // This gives some headroom both ways to prevent underflow and overflow.
-    // We tweak current_ratio to encourage this.
-    constexpr double tweak_time_scale = 0.05; // seconds
-    const double tweak_correction = (backlog_fullness - 0.5) * (time_delta / tweak_time_scale);
-    current_ratio *= std::pow(1.0 + 2.0 * tweak_correction, tweak_correction < 0 ? 3.0 : 1.0);
-
-    // This low-pass filter smoothes out variance in the calculated stretch ratio.
-    // The time-scale determines how responsive this filter is.
-    constexpr double lpf_time_scale = 0.712; // seconds
-    const double lpf_gain = 1.0 - std::exp(-time_delta / lpf_time_scale);
-    m_stretch_ratio += lpf_gain * (current_ratio - m_stretch_ratio);
-
-    // Place a lower limit of 5% speed. When a game boots up, there will be
-    // many silence samples. These do not need to be timestretched.
-    m_stretch_ratio = std::max(m_stretch_ratio, 0.05);
-    m_sound_touch.setTempo(m_stretch_ratio);
-
-    LOG_TRACE(Audio, "{:5}/{:5} ratio:{:0.6f} backlog:{:0.6f}", num_in, num_out, m_stretch_ratio,
-              backlog_fullness);
-
-    m_sound_touch.putSamples(in, static_cast<u32>(num_in));
-    return m_sound_touch.receiveSamples(out, static_cast<u32>(num_out));
-}
-
-} // namespace AudioCore

src/audio_core/time_stretch.h

@@ -1,34 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <cstddef>
-#include <SoundTouch.h>
-#include "common/common_types.h"
-
-namespace AudioCore {
-
-class TimeStretcher {
-public:
-    TimeStretcher(u32 sample_rate, u32 channel_count);
-
-    /// @param in       Input sample buffer
-    /// @param num_in   Number of input frames in `in`
-    /// @param out      Output sample buffer
-    /// @param num_out  Desired number of output frames in `out`
-    /// @returns Actual number of frames written to `out`
-    std::size_t Process(const s16* in, std::size_t num_in, s16* out, std::size_t num_out);
-
-    void Clear();
-
-    void Flush();
-
-private:
-    u32 m_sample_rate;
-    soundtouch::SoundTouch m_sound_touch;
-    double m_stretch_ratio = 1.0;
-};
-
-} // namespace AudioCore

src/common/atomic_ops.h

@@ -46,6 +46,50 @@ namespace Common {
                                           reinterpret_cast<__int64*>(expected.data())) != 0;
 }
 
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u8* pointer, u8 value, u8 expected,
+                                               u8& actual) {
+    actual =
+        _InterlockedCompareExchange8(reinterpret_cast<volatile char*>(pointer), value, expected);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u16* pointer, u16 value, u16 expected,
+                                               u16& actual) {
+    actual =
+        _InterlockedCompareExchange16(reinterpret_cast<volatile short*>(pointer), value, expected);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u32* pointer, u32 value, u32 expected,
+                                               u32& actual) {
+    actual =
+        _InterlockedCompareExchange(reinterpret_cast<volatile long*>(pointer), value, expected);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u64* pointer, u64 value, u64 expected,
+                                               u64& actual) {
+    actual = _InterlockedCompareExchange64(reinterpret_cast<volatile __int64*>(pointer), value,
+                                           expected);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u64* pointer, u128 value, u128 expected,
+                                               u128& actual) {
+    const bool result =
+        _InterlockedCompareExchange128(reinterpret_cast<volatile __int64*>(pointer), value[1],
+                                       value[0], reinterpret_cast<__int64*>(expected.data())) != 0;
+    actual = expected;
+    return result;
+}
+
+[[nodiscard]] inline u128 AtomicLoad128(volatile u64* pointer) {
+    u128 result{};
+    _InterlockedCompareExchange128(reinterpret_cast<volatile __int64*>(pointer), result[1],
+                                   result[0], reinterpret_cast<__int64*>(result.data()));
+    return result;
+}
+
 #else
 
 [[nodiscard]] inline bool AtomicCompareAndSwap(volatile u8* pointer, u8 value, u8 expected) {
@@ -72,6 +116,52 @@ namespace Common {
     return __sync_bool_compare_and_swap((unsigned __int128*)pointer, expected_a, value_a);
 }
 
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u8* pointer, u8 value, u8 expected,
+                                               u8& actual) {
+    actual = __sync_val_compare_and_swap(pointer, expected, value);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u16* pointer, u16 value, u16 expected,
+                                               u16& actual) {
+    actual = __sync_val_compare_and_swap(pointer, expected, value);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u32* pointer, u32 value, u32 expected,
+                                               u32& actual) {
+    actual = __sync_val_compare_and_swap(pointer, expected, value);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u64* pointer, u64 value, u64 expected,
+                                               u64& actual) {
+    actual = __sync_val_compare_and_swap(pointer, expected, value);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u64* pointer, u128 value, u128 expected,
+                                               u128& actual) {
+    unsigned __int128 value_a;
+    unsigned __int128 expected_a;
+    unsigned __int128 actual_a;
+    std::memcpy(&value_a, value.data(), sizeof(u128));
+    std::memcpy(&expected_a, expected.data(), sizeof(u128));
+    actual_a = __sync_val_compare_and_swap((unsigned __int128*)pointer, expected_a, value_a);
+    std::memcpy(actual.data(), &actual_a, sizeof(u128));
+    return actual_a == expected_a;
+}
+
+[[nodiscard]] inline u128 AtomicLoad128(volatile u64* pointer) {
+    unsigned __int128 zeros_a = 0;
+    unsigned __int128 result_a =
+        __sync_val_compare_and_swap((unsigned __int128*)pointer, zeros_a, zeros_a);
+
+    u128 result;
+    std::memcpy(result.data(), &result_a, sizeof(u128));
+    return result;
+}
+
 #endif
 
 } // namespace Common

src/common/settings.h

@@ -38,6 +38,7 @@ enum class CPUAccuracy : u32 {
     Auto = 0,
     Accurate = 1,
     Unsafe = 2,
+    Paranoid = 3,
 };
 
 enum class FullscreenMode : u32 {
@@ -470,7 +471,7 @@ struct Values {
 
     // Cpu
     RangedSetting<CPUAccuracy> cpu_accuracy{CPUAccuracy::Auto, CPUAccuracy::Auto,
-                                            CPUAccuracy::Unsafe, "cpu_accuracy"};
+                                            CPUAccuracy::Paranoid, "cpu_accuracy"};
     // TODO: remove cpu_accuracy_first_time, migration setting added 8 July 2021
     BasicSetting<bool> cpu_accuracy_first_time{true, "cpu_accuracy_first_time"};
     BasicSetting<bool> cpu_debug_mode{false, "cpu_debug_mode"};

src/common/x64/native_clock.cpp

@@ -55,8 +55,9 @@ NativeClock::NativeClock(u64 emulated_cpu_frequency_, u64 emulated_clock_frequen
 u64 NativeClock::GetRTSC() {
     TimePoint new_time_point{};
     TimePoint current_time_point{};
+
+    current_time_point.pack = Common::AtomicLoad128(time_point.pack.data());
     do {
-        current_time_point.pack = time_point.pack;
         _mm_mfence();
         const u64 current_measure = __rdtsc();
         u64 diff = current_measure - current_time_point.inner.last_measure;
@@ -66,7 +67,7 @@ u64 NativeClock::GetRTSC() {
                                                 : current_time_point.inner.last_measure;
         new_time_point.inner.accumulated_ticks = current_time_point.inner.accumulated_ticks + diff;
     } while (!Common::AtomicCompareAndSwap(time_point.pack.data(), new_time_point.pack,
-                                           current_time_point.pack));
+                                           current_time_point.pack, current_time_point.pack));
     /// The clock cannot be more precise than the guest timer, remove the lower bits
     return new_time_point.inner.accumulated_ticks & inaccuracy_mask;
 }
@@ -75,13 +76,14 @@ void NativeClock::Pause(bool is_paused) {
     if (!is_paused) {
         TimePoint current_time_point{};
         TimePoint new_time_point{};
+
+        current_time_point.pack = Common::AtomicLoad128(time_point.pack.data());
         do {
-            current_time_point.pack = time_point.pack;
             new_time_point.pack = current_time_point.pack;
             _mm_mfence();
             new_time_point.inner.last_measure = __rdtsc();
         } while (!Common::AtomicCompareAndSwap(time_point.pack.data(), new_time_point.pack,
-                                               current_time_point.pack));
+                                               current_time_point.pack, current_time_point.pack));
     }
 }
 

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -70,11 +70,13 @@ public:
     }
 
     void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
+        parent.LogBacktrace();
         UNIMPLEMENTED_MSG("This should never happen, pc = {:08X}, code = {:08X}", pc,
                           MemoryReadCode(pc));
     }
 
     void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
+        parent.LogBacktrace();
         LOG_CRITICAL(Core_ARM,
                      "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X}, thumb = {})",
                      exception, pc, MemoryReadCode(pc), parent.IsInThumbMode());
@@ -186,35 +188,41 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
         if (!Settings::values.cpuopt_recompile_exclusives) {
             config.recompile_on_exclusive_fastmem_failure = false;
         }
-    }
+    } else {
+        // Unsafe optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Unsafe) {
+            config.unsafe_optimizations = true;
+            if (Settings::values.cpuopt_unsafe_unfuse_fma) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
+            }
+            if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
+            }
+            if (Settings::values.cpuopt_unsafe_ignore_standard_fpcr) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreStandardFPCRValue;
+            }
+            if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
+            }
+            if (Settings::values.cpuopt_unsafe_ignore_global_monitor) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
+            }
+        }
 
-    // Unsafe optimizations
-    if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Unsafe) {
-        config.unsafe_optimizations = true;
-        if (Settings::values.cpuopt_unsafe_unfuse_fma) {
+        // Curated optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
+            config.unsafe_optimizations = true;
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-        }
-        if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
-            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
-        }
-        if (Settings::values.cpuopt_unsafe_ignore_standard_fpcr) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreStandardFPCRValue;
-        }
-        if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
-        }
-        if (Settings::values.cpuopt_unsafe_ignore_global_monitor) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
         }
-    }
 
-    // Curated optimizations
-    if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
-        config.unsafe_optimizations = true;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreStandardFPCRValue;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
+        // Paranoia mode for debugging optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Paranoid) {
+            config.unsafe_optimizations = false;
+            config.optimizations = Dynarmic::no_optimizations;
+        }
     }
 
     return std::make_unique<Dynarmic::A32::Jit>(config);

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -81,6 +81,7 @@ public:
     }
 
     void InterpreterFallback(u64 pc, std::size_t num_instructions) override {
+        parent.LogBacktrace();
         LOG_ERROR(Core_ARM,
                   "Unimplemented instruction @ 0x{:X} for {} instructions (instr = {:08X})", pc,
                   num_instructions, MemoryReadCode(pc));
@@ -118,6 +119,7 @@ public:
             return;
         case Dynarmic::A64::Exception::Breakpoint:
         default:
+            parent.LogBacktrace();
             ASSERT_MSG(false, "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X})",
                        static_cast<std::size_t>(exception), pc, MemoryReadCode(pc));
         }
@@ -248,35 +250,41 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
         if (!Settings::values.cpuopt_recompile_exclusives) {
             config.recompile_on_exclusive_fastmem_failure = false;
         }
-    }
+    } else {
+        // Unsafe optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Unsafe) {
+            config.unsafe_optimizations = true;
+            if (Settings::values.cpuopt_unsafe_unfuse_fma) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
+            }
+            if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
+            }
+            if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
+            }
+            if (Settings::values.cpuopt_unsafe_fastmem_check) {
+                config.fastmem_address_space_bits = 64;
+            }
+            if (Settings::values.cpuopt_unsafe_ignore_global_monitor) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
+            }
+        }
 
-    // Unsafe optimizations
-    if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Unsafe) {
-        config.unsafe_optimizations = true;
-        if (Settings::values.cpuopt_unsafe_unfuse_fma) {
+        // Curated optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
+            config.unsafe_optimizations = true;
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-        }
-        if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
-            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
-        }
-        if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
-        }
-        if (Settings::values.cpuopt_unsafe_fastmem_check) {
             config.fastmem_address_space_bits = 64;
-        }
-        if (Settings::values.cpuopt_unsafe_ignore_global_monitor) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
         }
-    }
 
-    // Curated optimizations
-    if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
-        config.unsafe_optimizations = true;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
-        config.fastmem_address_space_bits = 64;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
+        // Paranoia mode for debugging optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Paranoid) {
+            config.unsafe_optimizations = false;
+            config.optimizations = Dynarmic::no_optimizations;
+        }
     }
 
     return std::make_shared<Dynarmic::A64::Jit>(config);

src/core/file_sys/registered_cache.cpp

@@ -387,15 +387,17 @@ std::vector<NcaID> RegisteredCache::AccumulateFiles() const {
             continue;
 
         for (const auto& nca_dir : d2_dir->GetSubdirectories()) {
-            if (!FollowsNcaIdFormat(nca_dir->GetName()))
+            if (nca_dir == nullptr || !FollowsNcaIdFormat(nca_dir->GetName())) {
                 continue;
+            }
 
             ids.push_back(Common::HexStringToArray<0x10, true>(nca_dir->GetName().substr(0, 0x20)));
         }
 
         for (const auto& nca_file : d2_dir->GetFiles()) {
-            if (!FollowsNcaIdFormat(nca_file->GetName()))
+            if (nca_file == nullptr || !FollowsNcaIdFormat(nca_file->GetName())) {
                 continue;
+            }
 
             ids.push_back(
                 Common::HexStringToArray<0x10, true>(nca_file->GetName().substr(0, 0x20)));

src/core/hle/kernel/k_page_table.cpp

@@ -563,6 +563,8 @@ ResultCode KPageTable::UnmapProcessMemory(VAddr dst_addr, std::size_t size,
     block_manager->Update(dst_addr, num_pages, KMemoryState::Free, KMemoryPermission::None,
                           KMemoryAttribute::None);
 
+    system.InvalidateCpuInstructionCaches();
+
     return ResultSuccess;
 }
 

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

@@ -446,6 +446,14 @@ void WebBrowser::ExecuteLogin() {
 }
 
 void WebBrowser::ExecuteOffline() {
+    // TODO (Morph): This is a hack for WebSession foreground web applets such as those used by
+    //               Super Mario 3D All-Stars.
+    // TODO (Morph): Implement WebSession.
+    if (applet_mode == LibraryAppletMode::AllForegroundInitiallyHidden) {
+        LOG_WARNING(Service_AM, "WebSession is not implemented");
+        return;
+    }
+
     const auto main_url = GetMainURL(Common::FS::PathToUTF8String(offline_document));
 
     if (!Common::FS::Exists(main_url)) {

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

@@ -174,7 +174,7 @@ ResultCode VfsDirectoryServiceWrapper::RenameFile(const std::string& src_path_,
     ASSERT_MSG(dest != nullptr, "Newly created file with success cannot be found.");
 
     ASSERT_MSG(dest->WriteBytes(src->ReadAllBytes()) == src->GetSize(),
-               "Could not write all of the bytes but everything else has succeded.");
+               "Could not write all of the bytes but everything else has succeeded.");
 
     if (!src->GetContainingDirectory()->DeleteFile(Common::FS::GetFilename(src_path))) {
         // TODO(DarkLordZach): Find a better error code for this

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

@@ -262,11 +262,6 @@ void Controller_NPad::OnInit() {
             service_context.CreateEvent(fmt::format("npad:NpadStyleSetChanged_{}", i));
     }
 
-    if (hid_core.GetSupportedStyleTag().raw == Core::HID::NpadStyleSet::None) {
-        // We want to support all controllers
-        hid_core.SetSupportedStyleTag({Core::HID::NpadStyleSet::All});
-    }
-
     supported_npad_id_types.resize(npad_id_list.size());
     std::memcpy(supported_npad_id_types.data(), npad_id_list.data(),
                 npad_id_list.size() * sizeof(Core::HID::NpadIdType));
@@ -288,14 +283,6 @@ void Controller_NPad::OnInit() {
             WriteEmptyEntry(npad);
         }
     }
-
-    // Connect controllers
-    for (auto& controller : controller_data) {
-        const auto& device = controller.device;
-        if (device->IsConnected()) {
-            AddNewControllerAt(device->GetNpadStyleIndex(), device->GetNpadIdType());
-        }
-    }
 }
 
 void Controller_NPad::WriteEmptyEntry(NpadInternalState& npad) {
@@ -320,6 +307,7 @@ void Controller_NPad::WriteEmptyEntry(NpadInternalState& npad) {
 }
 
 void Controller_NPad::OnRelease() {
+    is_controller_initialized = false;
     for (std::size_t i = 0; i < controller_data.size(); ++i) {
         auto& controller = controller_data[i];
         service_context.CloseEvent(controller.styleset_changed_event);
@@ -651,9 +639,27 @@ void Controller_NPad::OnMotionUpdate(const Core::Timing::CoreTiming& core_timing
 
 void Controller_NPad::SetSupportedStyleSet(Core::HID::NpadStyleTag style_set) {
     hid_core.SetSupportedStyleTag(style_set);
+
+    if (is_controller_initialized) {
+        return;
+    }
+
+    // Once SetSupportedStyleSet is called controllers are fully initialized
+    is_controller_initialized = true;
+
+    // Connect all active controllers
+    for (auto& controller : controller_data) {
+        const auto& device = controller.device;
+        if (device->IsConnected()) {
+            AddNewControllerAt(device->GetNpadStyleIndex(), device->GetNpadIdType());
+        }
+    }
 }
 
 Core::HID::NpadStyleTag Controller_NPad::GetSupportedStyleSet() const {
+    if (!is_controller_initialized) {
+        return {Core::HID::NpadStyleSet::None};
+    }
     return hid_core.GetSupportedStyleTag();
 }
 

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

@@ -191,16 +191,16 @@ private:
 
     // This is nn::hid::detail::NpadFullKeyColorState
     struct NpadFullKeyColorState {
-        ColorAttribute attribute;
-        Core::HID::NpadControllerColor fullkey;
+        ColorAttribute attribute{ColorAttribute::NoController};
+        Core::HID::NpadControllerColor fullkey{};
     };
     static_assert(sizeof(NpadFullKeyColorState) == 0xC, "NpadFullKeyColorState is an invalid size");
 
     // This is nn::hid::detail::NpadJoyColorState
     struct NpadJoyColorState {
-        ColorAttribute attribute;
-        Core::HID::NpadControllerColor left;
-        Core::HID::NpadControllerColor right;
+        ColorAttribute attribute{ColorAttribute::NoController};
+        Core::HID::NpadControllerColor left{};
+        Core::HID::NpadControllerColor right{};
     };
     static_assert(sizeof(NpadJoyColorState) == 0x14, "NpadJoyColorState is an invalid size");
 
@@ -226,11 +226,11 @@ private:
     // This is nn::hid::NpadPalmaState
     // This is nn::hid::NpadSystemExtState
     struct NPadGenericState {
-        s64_le sampling_number;
-        Core::HID::NpadButtonState npad_buttons;
-        Core::HID::AnalogStickState l_stick;
-        Core::HID::AnalogStickState r_stick;
-        NpadAttribute connection_status;
+        s64_le sampling_number{};
+        Core::HID::NpadButtonState npad_buttons{};
+        Core::HID::AnalogStickState l_stick{};
+        Core::HID::AnalogStickState r_stick{};
+        NpadAttribute connection_status{};
         INSERT_PADDING_BYTES(4); // Reserved
     };
     static_assert(sizeof(NPadGenericState) == 0x28, "NPadGenericState is an invalid size");
@@ -253,7 +253,7 @@ private:
         Common::Vec3f gyro{};
         Common::Vec3f rotation{};
         std::array<Common::Vec3f, 3> orientation{};
-        SixAxisSensorAttribute attribute;
+        SixAxisSensorAttribute attribute{};
         INSERT_PADDING_BYTES(4); // Reserved
     };
     static_assert(sizeof(SixAxisSensorState) == 0x60, "SixAxisSensorState is an invalid size");
@@ -325,11 +325,11 @@ private:
 
     // This is nn::hid::detail::NfcXcdDeviceHandleStateImpl
     struct NfcXcdDeviceHandleStateImpl {
-        u64 handle;
-        bool is_available;
-        bool is_activated;
+        u64 handle{};
+        bool is_available{};
+        bool is_activated{};
         INSERT_PADDING_BYTES(0x6); // Reserved
-        u64 sampling_number;
+        u64 sampling_number{};
     };
     static_assert(sizeof(NfcXcdDeviceHandleStateImpl) == 0x18,
                   "NfcXcdDeviceHandleStateImpl is an invalid size");
@@ -366,8 +366,8 @@ private:
     };
 
     struct AppletFooterUi {
-        AppletFooterUiAttributes attributes;
-        AppletFooterUiType type;
+        AppletFooterUiAttributes attributes{};
+        AppletFooterUiType type{AppletFooterUiType::None};
         INSERT_PADDING_BYTES(0x5B); // Reserved
     };
     static_assert(sizeof(AppletFooterUi) == 0x60, "AppletFooterUi is an invalid size");
@@ -404,41 +404,41 @@ private:
 
     // This is nn::hid::detail::NpadInternalState
     struct NpadInternalState {
-        Core::HID::NpadStyleTag style_tag;
-        NpadJoyAssignmentMode assignment_mode;
-        NpadFullKeyColorState fullkey_color;
-        NpadJoyColorState joycon_color;
-        Lifo<NPadGenericState, hid_entry_count> fullkey_lifo;
-        Lifo<NPadGenericState, hid_entry_count> handheld_lifo;
-        Lifo<NPadGenericState, hid_entry_count> joy_dual_lifo;
-        Lifo<NPadGenericState, hid_entry_count> joy_left_lifo;
-        Lifo<NPadGenericState, hid_entry_count> joy_right_lifo;
-        Lifo<NPadGenericState, hid_entry_count> palma_lifo;
-        Lifo<NPadGenericState, hid_entry_count> system_ext_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_fullkey_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_handheld_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_dual_left_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_dual_right_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_left_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_right_lifo;
-        DeviceType device_type;
+        Core::HID::NpadStyleTag style_tag{Core::HID::NpadStyleSet::None};
+        NpadJoyAssignmentMode assignment_mode{NpadJoyAssignmentMode::Dual};
+        NpadFullKeyColorState fullkey_color{};
+        NpadJoyColorState joycon_color{};
+        Lifo<NPadGenericState, hid_entry_count> fullkey_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> handheld_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> joy_dual_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> joy_left_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> joy_right_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> palma_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> system_ext_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_fullkey_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_handheld_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_dual_left_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_dual_right_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_left_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_right_lifo{};
+        DeviceType device_type{};
         INSERT_PADDING_BYTES(0x4); // Reserved
-        NPadSystemProperties system_properties;
-        NpadSystemButtonProperties button_properties;
-        Core::HID::NpadBatteryLevel battery_level_dual;
-        Core::HID::NpadBatteryLevel battery_level_left;
-        Core::HID::NpadBatteryLevel battery_level_right;
+        NPadSystemProperties system_properties{};
+        NpadSystemButtonProperties button_properties{};
+        Core::HID::NpadBatteryLevel battery_level_dual{};
+        Core::HID::NpadBatteryLevel battery_level_left{};
+        Core::HID::NpadBatteryLevel battery_level_right{};
         union {
-            Lifo<NfcXcdDeviceHandleStateImpl, 0x2> nfc_xcd_device_lifo{};
-            AppletFooterUi applet_footer;
+            AppletFooterUi applet_footer{};
+            Lifo<NfcXcdDeviceHandleStateImpl, 0x2> nfc_xcd_device_lifo;
         };
         INSERT_PADDING_BYTES(0x20); // Unknown
-        Lifo<NpadGcTriggerState, hid_entry_count> gc_trigger_lifo;
-        NpadLarkType lark_type_l_and_main;
-        NpadLarkType lark_type_r;
-        NpadLuciaType lucia_type;
-        NpadLagonType lagon_type;
-        NpadLagerType lager_type;
+        Lifo<NpadGcTriggerState, hid_entry_count> gc_trigger_lifo{};
+        NpadLarkType lark_type_l_and_main{};
+        NpadLarkType lark_type_r{};
+        NpadLuciaType lucia_type{};
+        NpadLagonType lagon_type{};
+        NpadLagerType lager_type{};
         // FW 13.x Investigate there is some sort of bitflag related to joycons
         INSERT_PADDING_BYTES(0x4);
         INSERT_PADDING_BYTES(0xc08); // Unknown
@@ -511,7 +511,8 @@ private:
     NpadHandheldActivationMode handheld_activation_mode{NpadHandheldActivationMode::Dual};
     NpadCommunicationMode communication_mode{NpadCommunicationMode::Default};
     bool permit_vibration_session_enabled{false};
-    bool analog_stick_use_center_clamp{};
+    bool analog_stick_use_center_clamp{false};
     bool is_in_lr_assignment_mode{false};
+    bool is_controller_initialized{false};
 };
 } // namespace Service::HID

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

@@ -26,7 +26,7 @@ void NVDRV::Open(Kernel::HLERequestContext& ctx) {
         rb.Push<DeviceFD>(0);
         rb.PushEnum(NvResult::NotInitialized);
 
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -61,7 +61,7 @@ void NVDRV::Ioctl1(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -87,7 +87,7 @@ void NVDRV::Ioctl2(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -114,7 +114,7 @@ void NVDRV::Ioctl3(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -139,7 +139,7 @@ void NVDRV::Close(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -170,7 +170,7 @@ void NVDRV::QueryEvent(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 

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

@@ -18,8 +18,7 @@ BufferQueueConsumer::BufferQueueConsumer(std::shared_ptr<BufferQueueCore> core_)
 BufferQueueConsumer::~BufferQueueConsumer() = default;
 
 Status BufferQueueConsumer::AcquireBuffer(BufferItem* out_buffer,
-                                          std::chrono::nanoseconds expected_present,
-                                          u64 max_frame_number) {
+                                          std::chrono::nanoseconds expected_present) {
     std::scoped_lock lock(core->mutex);
 
     // Check that the consumer doesn't currently have the maximum number of buffers acquired.
@@ -50,12 +49,6 @@ Status BufferQueueConsumer::AcquireBuffer(BufferItem* out_buffer,
         while (core->queue.size() > 1 && !core->queue[0].is_auto_timestamp) {
             const auto& buffer_item{core->queue[1]};
 
-            // If dropping entry[0] would leave us with a buffer that the consumer is not yet ready
-            // for, don't drop it.
-            if (max_frame_number && buffer_item.frame_number > max_frame_number) {
-                break;
-            }
-
             // If entry[1] is timely, drop entry[0] (and repeat).
             const auto desired_present = buffer_item.timestamp;
             if (desired_present < expected_present.count() - MAX_REASONABLE_NSEC ||
@@ -200,4 +193,39 @@ Status BufferQueueConsumer::Connect(std::shared_ptr<IConsumerListener> consumer_
     return Status::NoError;
 }
 
+Status BufferQueueConsumer::GetReleasedBuffers(u64* out_slot_mask) {
+    if (out_slot_mask == nullptr) {
+        LOG_ERROR(Service_NVFlinger, "out_slot_mask may not be nullptr");
+        return Status::BadValue;
+    }
+
+    std::scoped_lock lock(core->mutex);
+
+    if (core->is_abandoned) {
+        LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
+        return Status::NoInit;
+    }
+
+    u64 mask = 0;
+    for (int s = 0; s < BufferQueueDefs::NUM_BUFFER_SLOTS; ++s) {
+        if (!slots[s].acquire_called) {
+            mask |= (1ULL << s);
+        }
+    }
+
+    // Remove from the mask queued buffers for which acquire has been called, since the consumer
+    // will not receive their buffer addresses and so must retain their cached information
+    auto current(core->queue.begin());
+    while (current != core->queue.end()) {
+        if (current->acquire_called) {
+            mask &= ~(1ULL << current->slot);
+        }
+        ++current;
+    }
+
+    LOG_DEBUG(Service_NVFlinger, "returning mask {}", mask);
+    *out_slot_mask = mask;
+    return Status::NoError;
+}
+
 } // namespace Service::android

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

@@ -24,10 +24,10 @@ public:
     explicit BufferQueueConsumer(std::shared_ptr<BufferQueueCore> core_);
     ~BufferQueueConsumer();
 
-    Status AcquireBuffer(BufferItem* out_buffer, std::chrono::nanoseconds expected_present,
-                         u64 max_frame_number = 0);
+    Status AcquireBuffer(BufferItem* out_buffer, std::chrono::nanoseconds expected_present);
     Status ReleaseBuffer(s32 slot, u64 frame_number, const Fence& release_fence);
     Status Connect(std::shared_ptr<IConsumerListener> consumer_listener, bool controlled_by_app);
+    Status GetReleasedBuffers(u64* out_slot_mask);
 
 private:
     std::shared_ptr<BufferQueueCore> core;

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

@@ -95,7 +95,6 @@ void BufferQueueCore::FreeBufferLocked(s32 slot) {
 }
 
 void BufferQueueCore::FreeAllBuffersLocked() {
-    queue.clear();
     buffer_has_been_queued = false;
 
     for (s32 slot = 0; slot < BufferQueueDefs::NUM_BUFFER_SLOTS; ++slot) {

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

@@ -73,8 +73,6 @@ private:
     u32 transform_hint{};
     bool is_allocating{};
     mutable std::condition_variable_any is_allocating_condition;
-    bool allow_allocation{true};
-    u64 buffer_age{};
     bool is_shutting_down{};
 };
 

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

@@ -62,11 +62,12 @@ Status BufferQueueProducer::RequestBuffer(s32 slot, std::shared_ptr<GraphicBuffe
 
 Status BufferQueueProducer::SetBufferCount(s32 buffer_count) {
     LOG_DEBUG(Service_NVFlinger, "count = {}", buffer_count);
-    std::shared_ptr<IConsumerListener> listener;
 
+    std::shared_ptr<IConsumerListener> listener;
     {
         std::scoped_lock lock(core->mutex);
         core->WaitWhileAllocatingLocked();
+
         if (core->is_abandoned) {
             LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
             return Status::NoInit;
@@ -120,7 +121,7 @@ Status BufferQueueProducer::SetBufferCount(s32 buffer_count) {
 }
 
 Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found,
-                                                      Status* returnFlags) const {
+                                                      Status* return_flags) const {
     bool try_again = true;
 
     while (try_again) {
@@ -142,10 +143,12 @@ Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found,
             ASSERT(slots[s].buffer_state == BufferState::Free);
             if (slots[s].graphic_buffer != nullptr) {
                 core->FreeBufferLocked(s);
-                *returnFlags |= Status::ReleaseAllBuffers;
+                *return_flags |= Status::ReleaseAllBuffers;
             }
         }
 
+        // Look for a free buffer to give to the client
+        *found = BufferQueueCore::INVALID_BUFFER_SLOT;
         s32 dequeued_count{};
         s32 acquired_count{};
         for (s32 s{}; s < max_buffer_count; ++s) {
@@ -235,68 +238,50 @@ Status BufferQueueProducer::DequeueBuffer(s32* out_slot, Fence* out_fence, bool
     {
         std::scoped_lock lock(core->mutex);
         core->WaitWhileAllocatingLocked();
+
         if (format == PixelFormat::NoFormat) {
             format = core->default_buffer_format;
         }
 
         // Enable the usage bits the consumer requested
         usage |= core->consumer_usage_bit;
+
+        s32 found{};
+        Status status = WaitForFreeSlotThenRelock(async, &found, &return_flags);
+        if (status != Status::NoError) {
+            return status;
+        }
+
+        // This should not happen
+        if (found == BufferQueueCore::INVALID_BUFFER_SLOT) {
+            LOG_ERROR(Service_NVFlinger, "no available buffer slots");
+            return Status::Busy;
+        }
+
+        *out_slot = found;
+
+        attached_by_consumer = slots[found].attached_by_consumer;
+
         const bool use_default_size = !width && !height;
         if (use_default_size) {
             width = core->default_width;
             height = core->default_height;
         }
 
-        s32 found = BufferItem::INVALID_BUFFER_SLOT;
-        while (found == BufferItem::INVALID_BUFFER_SLOT) {
-            Status status = WaitForFreeSlotThenRelock(async, &found, &return_flags);
-            if (status != Status::NoError) {
-                return status;
-            }
-
-            // This should not happen
-            if (found == BufferQueueCore::INVALID_BUFFER_SLOT) {
-                LOG_DEBUG(Service_NVFlinger, "no available buffer slots");
-                return Status::Busy;
-            }
-
-            const std::shared_ptr<GraphicBuffer>& buffer(slots[found].graphic_buffer);
-
-            // If we are not allowed to allocate new buffers, WaitForFreeSlotThenRelock must have
-            // returned a slot containing a buffer. If this buffer would require reallocation to
-            // meet the requested attributes, we free it and attempt to get another one.
-            if (!core->allow_allocation) {
-                if (buffer->NeedsReallocation(width, height, format, usage)) {
-                    core->FreeBufferLocked(found);
-                    found = BufferItem::INVALID_BUFFER_SLOT;
-                    continue;
-                }
-            }
-        }
-
-        *out_slot = found;
-        attached_by_consumer = slots[found].attached_by_consumer;
         slots[found].buffer_state = BufferState::Dequeued;
 
         const std::shared_ptr<GraphicBuffer>& buffer(slots[found].graphic_buffer);
-
-        if ((buffer == nullptr) || buffer->NeedsReallocation(width, height, format, usage)) {
+        if ((buffer == nullptr) || (buffer->Width() != width) || (buffer->Height() != height) ||
+            (buffer->Format() != format) || ((buffer->Usage() & usage) != usage)) {
             slots[found].acquire_called = false;
             slots[found].graphic_buffer = nullptr;
             slots[found].request_buffer_called = false;
             slots[found].fence = Fence::NoFence();
-            core->buffer_age = 0;
+
             return_flags |= Status::BufferNeedsReallocation;
-        } else {
-            // We add 1 because that will be the frame number when this buffer
-            // is queued
-            core->buffer_age = core->frame_counter + 1 - slots[found].frame_number;
         }
 
-        LOG_DEBUG(Service_NVFlinger, "setting buffer age to {}", core->buffer_age);
-
         *out_fence = slots[found].fence;
-
         slots[found].fence = Fence::NoFence();
     }
 
@@ -311,6 +296,7 @@ Status BufferQueueProducer::DequeueBuffer(s32* out_slot, Fence* out_fence, bool
 
         {
             std::scoped_lock lock(core->mutex);
+
             if (core->is_abandoned) {
                 LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
                 return Status::NoInit;
@@ -327,6 +313,7 @@ Status BufferQueueProducer::DequeueBuffer(s32* out_slot, Fence* out_fence, bool
 
     LOG_DEBUG(Service_NVFlinger, "returning slot={} frame={}, flags={}", *out_slot,
               slots[*out_slot].frame_number, return_flags);
+
     return return_flags;
 }
 
@@ -334,6 +321,7 @@ Status BufferQueueProducer::DetachBuffer(s32 slot) {
     LOG_DEBUG(Service_NVFlinger, "slot {}", slot);
 
     std::scoped_lock lock(core->mutex);
+
     if (core->is_abandoned) {
         LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
         return Status::NoInit;
@@ -369,7 +357,6 @@ Status BufferQueueProducer::DetachNextBuffer(std::shared_ptr<GraphicBuffer>* out
     }
 
     std::scoped_lock lock(core->mutex);
-
     core->WaitWhileAllocatingLocked();
 
     if (core->is_abandoned) {
@@ -423,6 +410,7 @@ Status BufferQueueProducer::AttachBuffer(s32* out_slot,
         return status;
     }
 
+    // This should not happen
     if (found == BufferQueueCore::INVALID_BUFFER_SLOT) {
         LOG_ERROR(Service_NVFlinger, "No available buffer slots");
         return Status::Busy;
@@ -466,8 +454,8 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
         return Status::BadValue;
     }
 
-    std::shared_ptr<IConsumerListener> frameAvailableListener;
-    std::shared_ptr<IConsumerListener> frameReplacedListener;
+    std::shared_ptr<IConsumerListener> frame_available_listener;
+    std::shared_ptr<IConsumerListener> frame_replaced_listener;
     s32 callback_ticket{};
     BufferItem item;
 
@@ -541,12 +529,13 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
         item.fence = fence;
         item.is_droppable = core->dequeue_buffer_cannot_block || async;
         item.swap_interval = swap_interval;
+
         sticky_transform = sticky_transform_;
 
         if (core->queue.empty()) {
             // When the queue is empty, we can simply queue this buffer
             core->queue.push_back(item);
-            frameAvailableListener = core->consumer_listener;
+            frame_available_listener = core->consumer_listener;
         } else {
             // When the queue is not empty, we need to look at the front buffer
             // state to see if we need to replace it
@@ -563,10 +552,10 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
                 }
                 // Overwrite the droppable buffer with the incoming one
                 *front = item;
-                frameReplacedListener = core->consumer_listener;
+                frame_replaced_listener = core->consumer_listener;
             } else {
                 core->queue.push_back(item);
-                frameAvailableListener = core->consumer_listener;
+                frame_available_listener = core->consumer_listener;
             }
         }
 
@@ -592,10 +581,10 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
             callback_condition.wait(callback_mutex);
         }
 
-        if (frameAvailableListener != nullptr) {
-            frameAvailableListener->OnFrameAvailable(item);
-        } else if (frameReplacedListener != nullptr) {
-            frameReplacedListener->OnFrameReplaced(item);
+        if (frame_available_listener != nullptr) {
+            frame_available_listener->OnFrameAvailable(item);
+        } else if (frame_replaced_listener != nullptr) {
+            frame_replaced_listener->OnFrameReplaced(item);
         }
 
         ++current_callback_ticket;
@@ -669,13 +658,6 @@ Status BufferQueueProducer::Query(NativeWindow what, s32* out_value) {
     case NativeWindow::ConsumerUsageBits:
         value = core->consumer_usage_bit;
         break;
-    case NativeWindow::BufferAge:
-        if (core->buffer_age > INT32_MAX) {
-            value = 0;
-        } else {
-            value = static_cast<u32>(core->buffer_age);
-        }
-        break;
     default:
         UNREACHABLE();
         return Status::BadValue;
@@ -737,7 +719,6 @@ Status BufferQueueProducer::Connect(const std::shared_ptr<IProducerListener>& li
     core->buffer_has_been_queued = false;
     core->dequeue_buffer_cannot_block =
         core->consumer_controlled_by_app && producer_controlled_by_app;
-    core->allow_allocation = true;
 
     return status;
 }
@@ -770,7 +751,7 @@ Status BufferQueueProducer::Disconnect(NativeWindowApi api) {
                 core->SignalDequeueCondition();
                 buffer_wait_event->GetWritableEvent().Signal();
                 listener = core->consumer_listener;
-            } else if (core->connected_api != NativeWindowApi::NoConnectedApi) {
+            } else {
                 LOG_ERROR(Service_NVFlinger, "still connected to another api (cur = {} req = {})",
                           core->connected_api, api);
                 status = Status::BadValue;

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

@@ -66,7 +66,7 @@ public:
 private:
     BufferQueueProducer(const BufferQueueProducer&) = delete;
 
-    Status WaitForFreeSlotThenRelock(bool async, s32* found, Status* returnFlags) const;
+    Status WaitForFreeSlotThenRelock(bool async, s32* found, Status* return_flags) const;
 
     Kernel::KEvent* buffer_wait_event{};
     Service::KernelHelpers::ServiceContext& service_context;

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

@@ -36,38 +36,41 @@ void ConsumerBase::FreeBufferLocked(s32 slot_index) {
 }
 
 void ConsumerBase::OnFrameAvailable(const BufferItem& item) {
-    std::scoped_lock lock(mutex);
     LOG_DEBUG(Service_NVFlinger, "called");
 }
 
 void ConsumerBase::OnFrameReplaced(const BufferItem& item) {
-    std::scoped_lock lock(mutex);
     LOG_DEBUG(Service_NVFlinger, "called");
 }
 
 void ConsumerBase::OnBuffersReleased() {
     std::scoped_lock lock(mutex);
+
     LOG_DEBUG(Service_NVFlinger, "called");
+
+    if (is_abandoned) {
+        // Nothing to do if we're already abandoned.
+        return;
+    }
+
+    u64 mask = 0;
+    consumer->GetReleasedBuffers(&mask);
+    for (int i = 0; i < BufferQueueDefs::NUM_BUFFER_SLOTS; i++) {
+        if (mask & (1ULL << i)) {
+            FreeBufferLocked(i);
+        }
+    }
 }
 
 void ConsumerBase::OnSidebandStreamChanged() {}
 
-Status ConsumerBase::AcquireBufferLocked(BufferItem* item, std::chrono::nanoseconds present_when,
-                                         u64 max_frame_number) {
-    if (is_abandoned) {
-        LOG_ERROR(Service_NVFlinger, "consumer is abandoned!");
-        return Status::NoInit;
-    }
-
-    Status err = consumer->AcquireBuffer(item, present_when, max_frame_number);
+Status ConsumerBase::AcquireBufferLocked(BufferItem* item, std::chrono::nanoseconds present_when) {
+    Status err = consumer->AcquireBuffer(item, present_when);
     if (err != Status::NoError) {
         return err;
     }
 
     if (item->graphic_buffer != nullptr) {
-        if (slots[item->slot].graphic_buffer != nullptr) {
-            FreeBufferLocked(item->slot);
-        }
         slots[item->slot].graphic_buffer = item->graphic_buffer;
     }
 

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

@@ -35,8 +35,7 @@ protected:
     virtual void OnSidebandStreamChanged() override;
 
     void FreeBufferLocked(s32 slot_index);
-    Status AcquireBufferLocked(BufferItem* item, std::chrono::nanoseconds present_when,
-                               u64 max_frame_number = 0);
+    Status AcquireBufferLocked(BufferItem* item, std::chrono::nanoseconds present_when);
     Status ReleaseBufferLocked(s32 slot, const std::shared_ptr<GraphicBuffer> graphic_buffer);
     bool StillTracking(s32 slot, const std::shared_ptr<GraphicBuffer> graphic_buffer) const;
     Status AddReleaseFenceLocked(s32 slot, const std::shared_ptr<GraphicBuffer> graphic_buffer,

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

@@ -104,7 +104,7 @@ void NVFlinger::SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance) {
 std::optional<u64> NVFlinger::OpenDisplay(std::string_view name) {
     const auto lock_guard = Lock();
 
-    LOG_DEBUG(Service, "Opening \"{}\" display", name);
+    LOG_DEBUG(Service_NVFlinger, "Opening \"{}\" display", name);
 
     const auto itr =
         std::find_if(displays.begin(), displays.end(),
@@ -219,7 +219,7 @@ VI::Layer* NVFlinger::FindOrCreateLayer(u64 display_id, u64 layer_id) {
     auto* layer = display->FindLayer(layer_id);
 
     if (layer == nullptr) {
-        LOG_DEBUG(Service, "Layer at id {} not found. Trying to create it.", layer_id);
+        LOG_DEBUG(Service_NVFlinger, "Layer at id {} not found. Trying to create it.", layer_id);
         CreateLayerAtId(*display, layer_id);
         return display->FindLayer(layer_id);
     }

src/shader_recompiler/frontend/maxwell/translate/impl/texture_load.cpp

@@ -132,7 +132,7 @@ void Impl(TranslatorVisitor& v, u64 insn, bool is_bindless) {
         multisample = v.X(meta_reg++);
     }
     if (tld.clamp != 0) {
-        throw NotImplementedException("TLD.CL - CLAMP is not implmented");
+        throw NotImplementedException("TLD.CL - CLAMP is not implemented");
     }
     IR::TextureInstInfo info{};
     info.type.Assign(GetType(tld.type));

src/shader_recompiler/frontend/maxwell/translate/impl/texture_mipmap_level.cpp

@@ -81,7 +81,7 @@ void Impl(TranslatorVisitor& v, u64 insn, bool is_bindless) {
     } const tmml{insn};
 
     if ((tmml.mask & 0b1100) != 0) {
-        throw NotImplementedException("TMML BA results are not implmented");
+        throw NotImplementedException("TMML BA results are not implemented");
     }
     const IR::Value coords{MakeCoords(v, tmml.coord_reg, tmml.type)};
 

src/video_core/renderer_vulkan/vk_compute_pipeline.cpp

@@ -200,7 +200,7 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
         });
     }
     const void* const descriptor_data{update_descriptor_queue.UpdateData()};
-    const bool is_rescaling = info.uses_rescaling_uniform;
+    const bool is_rescaling = !info.texture_descriptors.empty() || !info.image_descriptors.empty();
     scheduler.Record([this, descriptor_data, is_rescaling,
                       rescaling_data = rescaling.Data()](vk::CommandBuffer cmdbuf) {
         cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -238,7 +238,6 @@ GraphicsPipeline::GraphicsPipeline(
         enabled_uniform_buffer_masks[stage] = info->constant_buffer_mask;
         std::ranges::copy(info->constant_buffer_used_sizes, uniform_buffer_sizes[stage].begin());
         num_textures += Shader::NumDescriptors(info->texture_descriptors);
-        uses_rescale_unfiorm |= info->uses_rescaling_uniform;
     }
     auto func{[this, shader_notify, &render_pass_cache, &descriptor_pool, pipeline_statistics] {
         DescriptorLayoutBuilder builder{MakeBuilder(device, stage_infos)};
@@ -472,8 +471,7 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling) {
         });
     }
     const bool is_rescaling{texture_cache.IsRescaling()};
-    const bool update_rescaling{uses_rescale_unfiorm ? scheduler.UpdateRescaling(is_rescaling)
-                                                     : false};
+    const bool update_rescaling{scheduler.UpdateRescaling(is_rescaling)};
     const bool bind_pipeline{scheduler.UpdateGraphicsPipeline(this)};
     const void* const descriptor_data{update_descriptor_queue.UpdateData()};
     scheduler.Record([this, descriptor_data, bind_pipeline, rescaling_data = rescaling.Data(),
@@ -481,12 +479,10 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling) {
         if (bind_pipeline) {
             cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline);
         }
-        if (uses_rescale_unfiorm) {
-            cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
-                                 RESCALING_LAYOUT_WORDS_OFFSET, sizeof(rescaling_data),
-                                 rescaling_data.data());
-        }
-        if (uses_rescale_unfiorm && update_rescaling) {
+        cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
+                             RESCALING_LAYOUT_WORDS_OFFSET, sizeof(rescaling_data),
+                             rescaling_data.data());
+        if (update_rescaling) {
             const f32 config_down_factor{Settings::values.resolution_info.down_factor};
             const f32 scale_down_factor{is_rescaling ? config_down_factor : 1.0f};
             cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,

src/video_core/renderer_vulkan/vk_graphics_pipeline.h

@@ -151,7 +151,6 @@ private:
     std::mutex build_mutex;
     std::atomic_bool is_built{false};
     bool uses_push_descriptor{false};
-    bool uses_rescale_unfiorm{false};
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -234,9 +234,12 @@ void RasterizerVulkan::Clear() {
     const VkExtent2D render_area = framebuffer->RenderArea();
     scheduler.RequestRenderpass(framebuffer);
 
-    const bool is_rescaling = texture_cache.IsRescaling();
-    const u32 up_scale = is_rescaling ? Settings::values.resolution_info.up_scale : 1U;
-    const u32 down_shift = is_rescaling ? Settings::values.resolution_info.down_shift : 0U;
+    u32 up_scale = 1;
+    u32 down_shift = 0;
+    if (texture_cache.IsRescaling()) {
+        up_scale = Settings::values.resolution_info.up_scale;
+        down_shift = Settings::values.resolution_info.down_shift;
+    }
     UpdateViewportsState(regs);
 
     VkClearRect clear_rect{
@@ -692,9 +695,12 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs
     if (!state_tracker.TouchScissors()) {
         return;
     }
-    const bool is_rescaling = texture_cache.IsRescaling();
-    const u32 up_scale = is_rescaling ? Settings::values.resolution_info.up_scale : 1U;
-    const u32 down_shift = is_rescaling ? Settings::values.resolution_info.down_shift : 0U;
+    u32 up_scale = 1;
+    u32 down_shift = 0;
+    if (texture_cache.IsRescaling()) {
+        up_scale = Settings::values.resolution_info.up_scale;
+        down_shift = Settings::values.resolution_info.down_shift;
+    }
     const std::array scissors{
         GetScissorState(regs, 0, up_scale, down_shift),
         GetScissorState(regs, 1, up_scale, down_shift),

src/video_core/texture_cache/texture_cache.h

@@ -328,8 +328,7 @@ void TextureCache<P>::UpdateRenderTargets(bool is_clear) {
     }
 
     const bool rescaled = RescaleRenderTargets(is_clear);
-    const auto& resolution_info = Settings::values.resolution_info;
-    if (resolution_info.active && is_rescaling != rescaled) {
+    if (is_rescaling != rescaled) {
         flags[Dirty::RescaleViewports] = true;
         flags[Dirty::RescaleScissors] = true;
         is_rescaling = rescaled;
@@ -346,8 +345,12 @@ void TextureCache<P>::UpdateRenderTargets(bool is_clear) {
     for (size_t index = 0; index < NUM_RT; ++index) {
         render_targets.draw_buffers[index] = static_cast<u8>(maxwell3d.regs.rt_control.Map(index));
     }
-    const u32 up_scale = is_rescaling ? resolution_info.up_scale : 1U;
-    const u32 down_shift = is_rescaling ? resolution_info.down_shift : 0U;
+    u32 up_scale = 1;
+    u32 down_shift = 0;
+    if (is_rescaling) {
+        up_scale = Settings::values.resolution_info.up_scale;
+        down_shift = Settings::values.resolution_info.down_shift;
+    }
     render_targets.size = Extent2D{
         (maxwell3d.regs.render_area.width * up_scale) >> down_shift,
         (maxwell3d.regs.render_area.height * up_scale) >> down_shift,

src/video_core/vulkan_common/vulkan_device.cpp

@@ -736,7 +736,7 @@ VkFormat Device::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags
 }
 
 void Device::ReportLoss() const {
-    LOG_CRITICAL(Render_Vulkan, "Device loss occured!");
+    LOG_CRITICAL(Render_Vulkan, "Device loss occurred!");
 
     // Wait for the log to flush and for Nsight Aftermath to dump the results
     std::this_thread::sleep_for(std::chrono::seconds{15});

src/yuzu/configuration/configure_cpu.ui

@@ -52,6 +52,11 @@
                <string>Unsafe</string>
               </property>
              </item>
+             <item>
+              <property name="text">
+               <string>Paranoid (disables most optimizations)</string>
+              </property>
+             </item>
             </widget>
            </item>
           </layout>

src/yuzu/configuration/configure_hotkeys.cpp

@@ -35,8 +35,9 @@ ConfigureHotkeys::ConfigureHotkeys(Core::HID::HIDCore& hid_core, QWidget* parent
     ui->hotkey_list->setContextMenuPolicy(Qt::CustomContextMenu);
     ui->hotkey_list->setModel(model);
 
-    ui->hotkey_list->setColumnWidth(name_column, 200);
-    ui->hotkey_list->resizeColumnToContents(hotkey_column);
+    ui->hotkey_list->header()->setStretchLastSection(false);
+    ui->hotkey_list->header()->setSectionResizeMode(name_column, QHeaderView::ResizeMode::Stretch);
+    ui->hotkey_list->header()->setMinimumSectionSize(150);
 
     connect(ui->button_restore_defaults, &QPushButton::clicked, this,
             &ConfigureHotkeys::RestoreDefaults);
@@ -76,8 +77,8 @@ void ConfigureHotkeys::Populate(const HotkeyRegistry& registry) {
     }
 
     ui->hotkey_list->expandAll();
-    ui->hotkey_list->resizeColumnToContents(name_column);
     ui->hotkey_list->resizeColumnToContents(hotkey_column);
+    ui->hotkey_list->resizeColumnToContents(controller_column);
 }
 
 void ConfigureHotkeys::changeEvent(QEvent* event) {

src/yuzu/configuration/configure_per_game_addons.cpp

@@ -47,6 +47,10 @@ ConfigurePerGameAddons::ConfigurePerGameAddons(Core::System& system_, QWidget* p
     item_model->setHeaderData(0, Qt::Horizontal, tr("Patch Name"));
     item_model->setHeaderData(1, Qt::Horizontal, tr("Version"));
 
+    tree_view->header()->setStretchLastSection(false);
+    tree_view->header()->setSectionResizeMode(0, QHeaderView::ResizeMode::Stretch);
+    tree_view->header()->setMinimumSectionSize(150);
+
     // We must register all custom types with the Qt Automoc system so that we are able to use it
     // with signals/slots. In this case, QList falls under the umbrella of custom types.
     qRegisterMetaType<QList<QStandardItem*>>("QList<QStandardItem*>");
@@ -138,5 +142,5 @@ void ConfigurePerGameAddons::LoadConfiguration() {
         item_model->appendRow(list_items.back());
     }
 
-    tree_view->setColumnWidth(0, 5 * tree_view->width() / 16);
+    tree_view->resizeColumnToContents(1);
 }

src/yuzu_cmd/default_ini.h

@@ -342,12 +342,6 @@ fps_cap =
 # null: No audio output
 output_engine =
 
-# Whether or not to enable the audio-stretching post-processing effect.
-# This effect adjusts audio speed to match emulation speed and helps prevent audio stutter,
-# at the cost of increasing audio latency.
-# 0: No, 1 (default): Yes
-enable_audio_stretching =
-
 # Which audio device to use.
 # auto (default): Auto-select
 output_device =