yuzu: config: Disable realtime audio by default.

- This causes too many issues with other games.

src/audio_core/CMakeLists.txt

@@ -7,9 +7,12 @@ add_library(audio_core STATIC
     audio_out.h
     audio_renderer.cpp
     audio_renderer.h
+    behavior_info.cpp
+    behavior_info.h
     buffer.h
     codec.cpp
     codec.h
+    common.h
     null_sink.h
     sink.h
     sink_details.cpp

src/audio_core/audio_renderer.cpp

@@ -6,6 +6,7 @@
 #include "audio_core/audio_out.h"
 #include "audio_core/audio_renderer.h"
 #include "audio_core/codec.h"
+#include "audio_core/common.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
@@ -79,7 +80,7 @@ AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory
                              std::size_t instance_number)
     : worker_params{params}, buffer_event{buffer_event}, voices(params.voice_count),
       effects(params.effect_count), memory{memory_} {
-
+    behavior_info.SetUserRevision(params.revision);
     audio_out = std::make_unique<AudioCore::AudioOut>();
     stream = audio_out->OpenStream(core_timing, STREAM_SAMPLE_RATE, STREAM_NUM_CHANNELS,
                                    fmt::format("AudioRenderer-Instance{}", instance_number),
@@ -109,17 +110,17 @@ Stream::State AudioRenderer::GetStreamState() const {
     return stream->GetState();
 }
 
-static constexpr u32 VersionFromRevision(u32_le rev) {
-    // "REV7" -> 7
-    return ((rev >> 24) & 0xff) - 0x30;
-}
-
-std::vector<u8> AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_params) {
+ResultVal<std::vector<u8>> AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_params) {
     // Copy UpdateDataHeader struct
     UpdateDataHeader config{};
     std::memcpy(&config, input_params.data(), sizeof(UpdateDataHeader));
     u32 memory_pool_count = worker_params.effect_count + (worker_params.voice_count * 4);
 
+    if (!behavior_info.UpdateInput(input_params, sizeof(UpdateDataHeader))) {
+        LOG_ERROR(Audio, "Failed to update behavior info input parameters");
+        return Audren::ERR_INVALID_PARAMETERS;
+    }
+
     // Copy MemoryPoolInfo structs
     std::vector<MemoryPoolInfo> mem_pool_info(memory_pool_count);
     std::memcpy(mem_pool_info.data(),
@@ -173,8 +174,7 @@ std::vector<u8> AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_
     // Copy output header
     UpdateDataHeader response_data{worker_params};
     std::vector<u8> output_params(response_data.total_size);
-    const auto audren_revision = VersionFromRevision(config.revision);
-    if (audren_revision >= 5) {
+    if (behavior_info.IsElapsedFrameCountSupported()) {
         response_data.frame_count = 0x10;
         response_data.total_size += 0x10;
     }
@@ -200,7 +200,19 @@ std::vector<u8> AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_
                     sizeof(EffectOutStatus));
         effect_out_status_offset += sizeof(EffectOutStatus);
     }
-    return output_params;
+
+    // Update behavior info output
+    const std::size_t behavior_out_status_offset{
+        sizeof(UpdateDataHeader) + response_data.memory_pools_size + response_data.voices_size +
+        response_data.effects_size + response_data.sinks_size +
+        response_data.performance_manager_size};
+
+    if (!behavior_info.UpdateOutput(output_params, behavior_out_status_offset)) {
+        LOG_ERROR(Audio, "Failed to update behavior info output parameters");
+        return Audren::ERR_INVALID_PARAMETERS;
+    }
+
+    return MakeResult(output_params);
 }
 
 void AudioRenderer::VoiceState::SetWaveIndex(std::size_t index) {

src/audio_core/audio_renderer.h

@@ -8,11 +8,13 @@
 #include <memory>
 #include <vector>
 
+#include "audio_core/behavior_info.h"
 #include "audio_core/stream.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/swap.h"
 #include "core/hle/kernel/object.h"
+#include "core/hle/result.h"
 
 namespace Core::Timing {
 class CoreTiming;
@@ -226,7 +228,7 @@ public:
                   std::shared_ptr<Kernel::WritableEvent> buffer_event, std::size_t instance_number);
     ~AudioRenderer();
 
-    std::vector<u8> UpdateAudioRenderer(const std::vector<u8>& input_params);
+    ResultVal<std::vector<u8>> UpdateAudioRenderer(const std::vector<u8>& input_params);
     void QueueMixedBuffer(Buffer::Tag tag);
     void ReleaseAndQueueBuffers();
     u32 GetSampleRate() const;
@@ -237,6 +239,7 @@ public:
 private:
     class EffectState;
     class VoiceState;
+    BehaviorInfo behavior_info{};
 
     AudioRendererParameter worker_params;
     std::shared_ptr<Kernel::WritableEvent> buffer_event;

src/audio_core/behavior_info.cpp

@@ -0,0 +1,100 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include "audio_core/behavior_info.h"
+#include "audio_core/common.h"
+#include "common/logging/log.h"
+
+namespace AudioCore {
+
+BehaviorInfo::BehaviorInfo() : process_revision(CURRENT_PROCESS_REVISION) {}
+BehaviorInfo::~BehaviorInfo() = default;
+
+bool BehaviorInfo::UpdateInput(const std::vector<u8>& buffer, std::size_t offset) {
+    if (!CanConsumeBuffer(buffer.size(), offset, sizeof(InParams))) {
+        LOG_ERROR(Audio, "Buffer is an invalid size!");
+        return false;
+    }
+    InParams params{};
+    std::memcpy(&params, buffer.data() + offset, sizeof(InParams));
+
+    if (!IsValidRevision(params.revision)) {
+        LOG_ERROR(Audio, "Invalid input revision, revision=0x{:08X}", params.revision);
+        return false;
+    }
+
+    if (user_revision != params.revision) {
+        LOG_ERROR(Audio,
+                  "User revision differs from input revision, expecting 0x{:08X} but got 0x{:08X}",
+                  user_revision, params.revision);
+        return false;
+    }
+
+    ClearError();
+    UpdateFlags(params.flags);
+
+    // TODO(ogniK): Check input params size when InfoUpdater is used
+
+    return true;
+}
+
+bool BehaviorInfo::UpdateOutput(std::vector<u8>& buffer, std::size_t offset) {
+    if (!CanConsumeBuffer(buffer.size(), offset, sizeof(OutParams))) {
+        LOG_ERROR(Audio, "Buffer is an invalid size!");
+        return false;
+    }
+
+    OutParams params{};
+    std::memcpy(params.errors.data(), errors.data(), sizeof(ErrorInfo) * errors.size());
+    params.error_count = static_cast<u32_le>(error_count);
+    std::memcpy(buffer.data() + offset, &params, sizeof(OutParams));
+    return true;
+}
+
+void BehaviorInfo::ClearError() {
+    error_count = 0;
+}
+
+void BehaviorInfo::UpdateFlags(u64_le dest_flags) {
+    flags = dest_flags;
+}
+
+void BehaviorInfo::SetUserRevision(u32_le revision) {
+    user_revision = revision;
+}
+
+bool BehaviorInfo::IsAdpcmLoopContextBugFixed() const {
+    return IsRevisionSupported(2, user_revision);
+}
+
+bool BehaviorInfo::IsSplitterSupported() const {
+    return IsRevisionSupported(2, user_revision);
+}
+
+bool BehaviorInfo::IsLongSizePreDelaySupported() const {
+    return IsRevisionSupported(3, user_revision);
+}
+
+bool BehaviorInfo::IsAudioRenererProcessingTimeLimit80PercentSupported() const {
+    return IsRevisionSupported(5, user_revision);
+}
+
+bool BehaviorInfo::IsAudioRenererProcessingTimeLimit75PercentSupported() const {
+    return IsRevisionSupported(4, user_revision);
+}
+
+bool BehaviorInfo::IsAudioRenererProcessingTimeLimit70PercentSupported() const {
+    return IsRevisionSupported(1, user_revision);
+}
+
+bool BehaviorInfo::IsElapsedFrameCountSupported() const {
+    return IsRevisionSupported(5, user_revision);
+}
+
+bool BehaviorInfo::IsMemoryPoolForceMappingEnabled() const {
+    return (flags & 1) != 0;
+}
+
+} // namespace AudioCore

src/audio_core/behavior_info.h

@@ -0,0 +1,66 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+#include <vector>
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "common/swap.h"
+
+namespace AudioCore {
+class BehaviorInfo {
+public:
+    explicit BehaviorInfo();
+    ~BehaviorInfo();
+
+    bool UpdateInput(const std::vector<u8>& buffer, std::size_t offset);
+    bool UpdateOutput(std::vector<u8>& buffer, std::size_t offset);
+
+    void ClearError();
+    void UpdateFlags(u64_le dest_flags);
+    void SetUserRevision(u32_le revision);
+
+    bool IsAdpcmLoopContextBugFixed() const;
+    bool IsSplitterSupported() const;
+    bool IsLongSizePreDelaySupported() const;
+    bool IsAudioRenererProcessingTimeLimit80PercentSupported() const;
+    bool IsAudioRenererProcessingTimeLimit75PercentSupported() const;
+    bool IsAudioRenererProcessingTimeLimit70PercentSupported() const;
+    bool IsElapsedFrameCountSupported() const;
+    bool IsMemoryPoolForceMappingEnabled() const;
+
+private:
+    u32_le process_revision{};
+    u32_le user_revision{};
+    u64_le flags{};
+
+    struct ErrorInfo {
+        u32_le result{};
+        INSERT_PADDING_WORDS(1);
+        u64_le result_info{};
+    };
+    static_assert(sizeof(ErrorInfo) == 0x10, "ErrorInfo is an invalid size");
+
+    std::array<ErrorInfo, 10> errors{};
+    std::size_t error_count{};
+
+    struct InParams {
+        u32_le revision{};
+        u32_le padding{};
+        u64_le flags{};
+    };
+    static_assert(sizeof(InParams) == 0x10, "InParams is an invalid size");
+
+    struct OutParams {
+        std::array<ErrorInfo, 10> errors{};
+        u32_le error_count{};
+        INSERT_PADDING_BYTES(12);
+    };
+    static_assert(sizeof(OutParams) == 0xb0, "OutParams is an invalid size");
+};
+
+} // namespace AudioCore

src/audio_core/common.h

@@ -0,0 +1,47 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "common/swap.h"
+#include "core/hle/result.h"
+
+namespace AudioCore {
+namespace Audren {
+constexpr ResultCode ERR_INVALID_PARAMETERS{ErrorModule::Audio, 41};
+}
+
+constexpr u32_le CURRENT_PROCESS_REVISION = Common::MakeMagic('R', 'E', 'V', '8');
+
+static constexpr u32 VersionFromRevision(u32_le rev) {
+    // "REV7" -> 7
+    return ((rev >> 24) & 0xff) - 0x30;
+}
+
+static constexpr bool IsRevisionSupported(u32 required, u32_le user_revision) {
+    const auto base = VersionFromRevision(user_revision);
+    return required <= base;
+}
+
+static constexpr bool IsValidRevision(u32_le revision) {
+    const auto base = VersionFromRevision(revision);
+    constexpr auto max_rev = VersionFromRevision(CURRENT_PROCESS_REVISION);
+    return base <= max_rev;
+}
+
+static constexpr bool CanConsumeBuffer(std::size_t size, std::size_t offset, std::size_t required) {
+    if (offset > size) {
+        return false;
+    }
+    if (size < required) {
+        return false;
+    }
+    if ((size - offset) < required) {
+        return false;
+    }
+    return true;
+}
+
+} // namespace AudioCore

src/audio_core/stream.cpp

@@ -11,8 +11,10 @@
 #include "audio_core/stream.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
+#include "core/perf_stats.h"
 #include "core/settings.h"
 
 namespace AudioCore {
@@ -61,8 +63,11 @@ Stream::State Stream::GetState() const {
 
 s64 Stream::GetBufferReleaseCycles(const Buffer& buffer) const {
     const std::size_t num_samples{buffer.GetSamples().size() / GetNumChannels()};
-    const auto us =
-        std::chrono::microseconds((static_cast<u64>(num_samples) * 1000000) / sample_rate);
+    const double time_scale{Settings::values.enable_realtime_audio
+                                ? Core::System::GetInstance().GetPerfStats().GetLastFrameTimeScale()
+                                : 1.0f};
+    const auto us{std::chrono::microseconds(
+        (static_cast<u64>(num_samples) * static_cast<u64>(1000000 / time_scale)) / sample_rate)};
     return Core::Timing::usToCycles(us);
 }
 

src/core/arm/unicorn/arm_unicorn.cpp

@@ -11,6 +11,7 @@
 #include "core/core_timing.h"
 #include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/svc.h"
+#include "core/memory.h"
 
 namespace Core {
 
@@ -171,7 +172,17 @@ MICROPROFILE_DEFINE(ARM_Jit_Unicorn, "ARM JIT", "Unicorn", MP_RGB(255, 64, 64));
 
 void ARM_Unicorn::ExecuteInstructions(std::size_t num_instructions) {
     MICROPROFILE_SCOPE(ARM_Jit_Unicorn);
+
+    // Temporarily map the code page for Unicorn
+    u64 map_addr{GetPC() & ~Memory::PAGE_MASK};
+    std::vector<u8> page_buffer(Memory::PAGE_SIZE);
+    system.Memory().ReadBlock(map_addr, page_buffer.data(), page_buffer.size());
+
+    CHECKED(uc_mem_map_ptr(uc, map_addr, page_buffer.size(),
+                           UC_PROT_READ | UC_PROT_WRITE | UC_PROT_EXEC, page_buffer.data()));
     CHECKED(uc_emu_start(uc, GetPC(), 1ULL << 63, 0, num_instructions));
+    CHECKED(uc_mem_unmap(uc, map_addr, page_buffer.size()));
+
     system.CoreTiming().AddTicks(num_instructions);
     if (GDBStub::IsServerEnabled()) {
         if (last_bkpt_hit && last_bkpt.type == GDBStub::BreakpointType::Execute) {

src/core/crypto/key_manager.cpp

@@ -1202,7 +1202,8 @@ const boost::container::flat_map<std::string, KeyIndex<S128KeyType>> KeyManager:
      {S128KeyType::Source, static_cast<u64>(SourceKeyType::KeyAreaKey),
       static_cast<u64>(KeyAreaKeyType::System)}},
     {"titlekek_source", {S128KeyType::Source, static_cast<u64>(SourceKeyType::Titlekek), 0}},
-    {"keyblob_mac_key_source", {S128KeyType::Source, static_cast<u64>(SourceKeyType::KeyblobMAC)}},
+    {"keyblob_mac_key_source",
+     {S128KeyType::Source, static_cast<u64>(SourceKeyType::KeyblobMAC), 0}},
     {"tsec_key", {S128KeyType::TSEC, 0, 0}},
     {"secure_boot_key", {S128KeyType::SecureBoot, 0, 0}},
     {"sd_seed", {S128KeyType::SDSeed, 0, 0}},

src/core/hle/service/audio/audren_u.cpp

@@ -94,9 +94,14 @@ private:
     void RequestUpdateImpl(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_Audio, "(STUBBED) called");
 
-        ctx.WriteBuffer(renderer->UpdateAudioRenderer(ctx.ReadBuffer()));
+        auto result = renderer->UpdateAudioRenderer(ctx.ReadBuffer());
+
+        if (result.Succeeded()) {
+            ctx.WriteBuffer(result.Unwrap());
+        }
+
         IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(RESULT_SUCCESS);
+        rb.Push(result.Code());
     }
 
     void Start(Kernel::HLERequestContext& ctx) {

src/core/hle/service/time/time_zone_manager.cpp

@@ -518,8 +518,8 @@ static bool ParseTimeZoneBinary(TimeZoneRule& time_zone_rule, FileSys::VirtualFi
     constexpr s32 time_zone_max_leaps{50};
     constexpr s32 time_zone_max_chars{50};
     if (!(0 <= header.leap_count && header.leap_count < time_zone_max_leaps &&
-          0 < header.type_count && header.type_count < time_zone_rule.ttis.size() &&
-          0 <= header.time_count && header.time_count < time_zone_rule.ats.size() &&
+          0 < header.type_count && header.type_count < s32(time_zone_rule.ttis.size()) &&
+          0 <= header.time_count && header.time_count < s32(time_zone_rule.ats.size()) &&
           0 <= header.char_count && header.char_count < time_zone_max_chars &&
           (header.ttis_std_count == header.type_count || header.ttis_std_count == 0) &&
           (header.ttis_gmt_count == header.type_count || header.ttis_gmt_count == 0))) {

src/core/perf_stats.cpp

@@ -108,13 +108,15 @@ PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us
     system_frames = 0;
     game_frames = 0;
 
+    target_fps = static_cast<u32>(results.game_fps / results.emulation_speed);
+
     return results;
 }
 
 double PerfStats::GetLastFrameTimeScale() {
     std::lock_guard lock{object_mutex};
 
-    constexpr double FRAME_LENGTH = 1.0 / 60;
+    double FRAME_LENGTH = 1.0 / target_fps;
     return duration_cast<DoubleSecs>(previous_frame_length).count() / FRAME_LENGTH;
 }
 

src/core/perf_stats.h

@@ -81,6 +81,8 @@ private:
     Clock::time_point frame_begin = reset_point;
     /// Total visible duration (including frame-limiting, etc.) of the previous system frame
     Clock::duration previous_frame_length = Clock::duration::zero();
+
+    u32 target_fps{60};
 };
 
 class FrameLimiter {

src/core/settings.cpp

@@ -98,6 +98,7 @@ void LogSettings() {
     LogSetting("Renderer_UseVsync", Settings::values.use_vsync);
     LogSetting("Audio_OutputEngine", Settings::values.sink_id);
     LogSetting("Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);
+    LogSetting("Audio_EnableRealTime", Settings::values.enable_realtime_audio);
     LogSetting("Audio_OutputDevice", Settings::values.audio_device_id);
     LogSetting("DataStorage_UseVirtualSd", Settings::values.use_virtual_sd);
     LogSetting("DataStorage_NandDir", FileUtil::GetUserPath(FileUtil::UserPath::NANDDir));

src/core/settings.h

@@ -452,6 +452,7 @@ struct Values {
     // Audio
     std::string sink_id;
     bool enable_audio_stretching;
+    bool enable_realtime_audio;
     std::string audio_device_id;
     float volume;
 

src/core/telemetry_session.cpp

@@ -178,6 +178,7 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
     constexpr auto field_type = Telemetry::FieldType::UserConfig;
     AddField(field_type, "Audio_SinkId", Settings::values.sink_id);
     AddField(field_type, "Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);
+    AddField(field_type, "Audio_EnableRealTime", Settings::values.enable_realtime_audio);
     AddField(field_type, "Core_UseMultiCore", Settings::values.use_multi_core);
     AddField(field_type, "Renderer_Backend", TranslateRenderer(Settings::values.renderer_backend));
     AddField(field_type, "Renderer_ResolutionFactor", Settings::values.resolution_factor);

src/video_core/engines/maxwell_3d.cpp

@@ -92,6 +92,10 @@ void Maxwell3D::InitializeRegisterDefaults() {
         color_mask.A.Assign(1);
     }
 
+    for (auto& format : regs.vertex_attrib_format) {
+        format.constant.Assign(1);
+    }
+
     // NVN games expect these values to be enabled at boot
     regs.rasterize_enable = 1;
     regs.rt_separate_frag_data = 1;

src/video_core/engines/maxwell_3d.h

@@ -1149,7 +1149,7 @@ public:
 
                     /// Returns whether the vertex array specified by index is supposed to be
                     /// accessed per instance or not.
-                    bool IsInstancingEnabled(u32 index) const {
+                    bool IsInstancingEnabled(std::size_t index) const {
                         return is_instanced[index];
                     }
                 } instanced_arrays;

src/video_core/renderer_vulkan/fixed_pipeline_state.cpp

@@ -2,10 +2,12 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <cstring>
 #include <tuple>
 
 #include <boost/functional/hash.hpp>
 
+#include "common/cityhash.h"
 #include "common/common_types.h"
 #include "video_core/renderer_vulkan/fixed_pipeline_state.h"
 
@@ -13,289 +15,352 @@ namespace Vulkan {
 
 namespace {
 
-constexpr FixedPipelineState::DepthStencil GetDepthStencilState(const Maxwell& regs) {
-    const FixedPipelineState::StencilFace front_stencil(
-        regs.stencil_front_op_fail, regs.stencil_front_op_zfail, regs.stencil_front_op_zpass,
-        regs.stencil_front_func_func);
-    const FixedPipelineState::StencilFace back_stencil =
-        regs.stencil_two_side_enable
-            ? FixedPipelineState::StencilFace(regs.stencil_back_op_fail, regs.stencil_back_op_zfail,
-                                              regs.stencil_back_op_zpass,
-                                              regs.stencil_back_func_func)
-            : front_stencil;
-    return FixedPipelineState::DepthStencil(
-        regs.depth_test_enable == 1, regs.depth_write_enabled == 1, regs.depth_bounds_enable == 1,
-        regs.stencil_enable == 1, regs.depth_test_func, front_stencil, back_stencil);
-}
-
-constexpr FixedPipelineState::InputAssembly GetInputAssemblyState(const Maxwell& regs) {
-    return FixedPipelineState::InputAssembly(
-        regs.draw.topology, regs.primitive_restart.enabled,
-        regs.draw.topology == Maxwell::PrimitiveTopology::Points ? regs.point_size : 0.0f);
-}
-
-constexpr FixedPipelineState::BlendingAttachment GetBlendingAttachmentState(
-    const Maxwell& regs, std::size_t render_target) {
-    const auto& mask = regs.color_mask[regs.color_mask_common ? 0 : render_target];
-    const std::array components = {mask.R != 0, mask.G != 0, mask.B != 0, mask.A != 0};
-
-    const FixedPipelineState::BlendingAttachment default_blending(
-        false, Maxwell::Blend::Equation::Add, Maxwell::Blend::Factor::One,
-        Maxwell::Blend::Factor::Zero, Maxwell::Blend::Equation::Add, Maxwell::Blend::Factor::One,
-        Maxwell::Blend::Factor::Zero, components);
-    if (render_target >= regs.rt_control.count) {
-        return default_blending;
-    }
-
-    if (!regs.independent_blend_enable) {
-        const auto& src = regs.blend;
-        if (!src.enable[render_target]) {
-            return default_blending;
-        }
-        return FixedPipelineState::BlendingAttachment(
-            true, src.equation_rgb, src.factor_source_rgb, src.factor_dest_rgb, src.equation_a,
-            src.factor_source_a, src.factor_dest_a, components);
-    }
-
-    if (!regs.blend.enable[render_target]) {
-        return default_blending;
-    }
-    const auto& src = regs.independent_blend[render_target];
-    return FixedPipelineState::BlendingAttachment(
-        true, src.equation_rgb, src.factor_source_rgb, src.factor_dest_rgb, src.equation_a,
-        src.factor_source_a, src.factor_dest_a, components);
-}
-
-constexpr FixedPipelineState::ColorBlending GetColorBlendingState(const Maxwell& regs) {
-    return FixedPipelineState::ColorBlending(
-        {regs.blend_color.r, regs.blend_color.g, regs.blend_color.b, regs.blend_color.a},
-        regs.rt_control.count,
-        {GetBlendingAttachmentState(regs, 0), GetBlendingAttachmentState(regs, 1),
-         GetBlendingAttachmentState(regs, 2), GetBlendingAttachmentState(regs, 3),
-         GetBlendingAttachmentState(regs, 4), GetBlendingAttachmentState(regs, 5),
-         GetBlendingAttachmentState(regs, 6), GetBlendingAttachmentState(regs, 7)});
-}
-
-constexpr FixedPipelineState::Tessellation GetTessellationState(const Maxwell& regs) {
-    return FixedPipelineState::Tessellation(regs.patch_vertices, regs.tess_mode.prim,
-                                            regs.tess_mode.spacing, regs.tess_mode.cw != 0);
-}
-
-constexpr std::size_t Point = 0;
-constexpr std::size_t Line = 1;
-constexpr std::size_t Polygon = 2;
-constexpr std::array PolygonOffsetEnableLUT = {
-    Point,   // Points
-    Line,    // Lines
-    Line,    // LineLoop
-    Line,    // LineStrip
-    Polygon, // Triangles
-    Polygon, // TriangleStrip
-    Polygon, // TriangleFan
-    Polygon, // Quads
-    Polygon, // QuadStrip
-    Polygon, // Polygon
-    Line,    // LinesAdjacency
-    Line,    // LineStripAdjacency
-    Polygon, // TrianglesAdjacency
-    Polygon, // TriangleStripAdjacency
-    Polygon, // Patches
+constexpr std::size_t POINT = 0;
+constexpr std::size_t LINE = 1;
+constexpr std::size_t POLYGON = 2;
+constexpr std::array POLYGON_OFFSET_ENABLE_LUT = {
+    POINT,   // Points
+    LINE,    // Lines
+    LINE,    // LineLoop
+    LINE,    // LineStrip
+    POLYGON, // Triangles
+    POLYGON, // TriangleStrip
+    POLYGON, // TriangleFan
+    POLYGON, // Quads
+    POLYGON, // QuadStrip
+    POLYGON, // Polygon
+    LINE,    // LinesAdjacency
+    LINE,    // LineStripAdjacency
+    POLYGON, // TrianglesAdjacency
+    POLYGON, // TriangleStripAdjacency
+    POLYGON, // Patches
 };
 
-constexpr FixedPipelineState::Rasterizer GetRasterizerState(const Maxwell& regs) {
-    const std::array enabled_lut = {regs.polygon_offset_point_enable,
-                                    regs.polygon_offset_line_enable,
-                                    regs.polygon_offset_fill_enable};
-    const auto topology = static_cast<std::size_t>(regs.draw.topology.Value());
-    const bool depth_bias_enabled = enabled_lut[PolygonOffsetEnableLUT[topology]];
-
-    const auto& clip = regs.view_volume_clip_control;
-    const bool depth_clamp_enabled = clip.depth_clamp_near == 1 || clip.depth_clamp_far == 1;
-
-    Maxwell::FrontFace front_face = regs.front_face;
-    if (regs.screen_y_control.triangle_rast_flip != 0 &&
-        regs.viewport_transform[0].scale_y > 0.0f) {
-        if (front_face == Maxwell::FrontFace::CounterClockWise)
-            front_face = Maxwell::FrontFace::ClockWise;
-        else if (front_face == Maxwell::FrontFace::ClockWise)
-            front_face = Maxwell::FrontFace::CounterClockWise;
-    }
-
-    const bool gl_ndc = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne;
-    return FixedPipelineState::Rasterizer(regs.cull_test_enabled, depth_bias_enabled,
-                                          depth_clamp_enabled, gl_ndc, regs.cull_face, front_face);
-}
-
 } // Anonymous namespace
 
-std::size_t FixedPipelineState::VertexBinding::Hash() const noexcept {
-    return (index << stride) ^ divisor;
-}
-
-bool FixedPipelineState::VertexBinding::operator==(const VertexBinding& rhs) const noexcept {
-    return std::tie(index, stride, divisor) == std::tie(rhs.index, rhs.stride, rhs.divisor);
-}
-
-std::size_t FixedPipelineState::VertexAttribute::Hash() const noexcept {
-    return static_cast<std::size_t>(index) ^ (static_cast<std::size_t>(buffer) << 13) ^
-           (static_cast<std::size_t>(type) << 22) ^ (static_cast<std::size_t>(size) << 31) ^
-           (static_cast<std::size_t>(offset) << 36);
-}
-
-bool FixedPipelineState::VertexAttribute::operator==(const VertexAttribute& rhs) const noexcept {
-    return std::tie(index, buffer, type, size, offset) ==
-           std::tie(rhs.index, rhs.buffer, rhs.type, rhs.size, rhs.offset);
-}
-
-std::size_t FixedPipelineState::StencilFace::Hash() const noexcept {
-    return static_cast<std::size_t>(action_stencil_fail) ^
-           (static_cast<std::size_t>(action_depth_fail) << 4) ^
-           (static_cast<std::size_t>(action_depth_fail) << 20) ^
-           (static_cast<std::size_t>(action_depth_pass) << 36);
-}
-
-bool FixedPipelineState::StencilFace::operator==(const StencilFace& rhs) const noexcept {
-    return std::tie(action_stencil_fail, action_depth_fail, action_depth_pass, test_func) ==
-           std::tie(rhs.action_stencil_fail, rhs.action_depth_fail, rhs.action_depth_pass,
-                    rhs.test_func);
-}
-
-std::size_t FixedPipelineState::BlendingAttachment::Hash() const noexcept {
-    return static_cast<std::size_t>(enable) ^ (static_cast<std::size_t>(rgb_equation) << 5) ^
-           (static_cast<std::size_t>(src_rgb_func) << 10) ^
-           (static_cast<std::size_t>(dst_rgb_func) << 15) ^
-           (static_cast<std::size_t>(a_equation) << 20) ^
-           (static_cast<std::size_t>(src_a_func) << 25) ^
-           (static_cast<std::size_t>(dst_a_func) << 30) ^
-           (static_cast<std::size_t>(components[0]) << 35) ^
-           (static_cast<std::size_t>(components[1]) << 36) ^
-           (static_cast<std::size_t>(components[2]) << 37) ^
-           (static_cast<std::size_t>(components[3]) << 38);
-}
-
-bool FixedPipelineState::BlendingAttachment::operator==(const BlendingAttachment& rhs) const
-    noexcept {
-    return std::tie(enable, rgb_equation, src_rgb_func, dst_rgb_func, a_equation, src_a_func,
-                    dst_a_func, components) ==
-           std::tie(rhs.enable, rhs.rgb_equation, rhs.src_rgb_func, rhs.dst_rgb_func,
-                    rhs.a_equation, rhs.src_a_func, rhs.dst_a_func, rhs.components);
-}
-
-std::size_t FixedPipelineState::VertexInput::Hash() const noexcept {
-    std::size_t hash = num_bindings ^ (num_attributes << 32);
-    for (std::size_t i = 0; i < num_bindings; ++i) {
-        boost::hash_combine(hash, bindings[i].Hash());
+void FixedPipelineState::DepthStencil::Fill(const Maxwell& regs) noexcept {
+    raw = 0;
+    front.action_stencil_fail.Assign(PackStencilOp(regs.stencil_front_op_fail));
+    front.action_depth_fail.Assign(PackStencilOp(regs.stencil_front_op_zfail));
+    front.action_depth_pass.Assign(PackStencilOp(regs.stencil_front_op_zpass));
+    front.test_func.Assign(PackComparisonOp(regs.stencil_front_func_func));
+    if (regs.stencil_two_side_enable) {
+        back.action_stencil_fail.Assign(PackStencilOp(regs.stencil_back_op_fail));
+        back.action_depth_fail.Assign(PackStencilOp(regs.stencil_back_op_zfail));
+        back.action_depth_pass.Assign(PackStencilOp(regs.stencil_back_op_zpass));
+        back.test_func.Assign(PackComparisonOp(regs.stencil_back_func_func));
+    } else {
+        back.action_stencil_fail.Assign(front.action_stencil_fail);
+        back.action_depth_fail.Assign(front.action_depth_fail);
+        back.action_depth_pass.Assign(front.action_depth_pass);
+        back.test_func.Assign(front.test_func);
     }
-    for (std::size_t i = 0; i < num_attributes; ++i) {
-        boost::hash_combine(hash, attributes[i].Hash());
+    depth_test_enable.Assign(regs.depth_test_enable);
+    depth_write_enable.Assign(regs.depth_write_enabled);
+    depth_bounds_enable.Assign(regs.depth_bounds_enable);
+    stencil_enable.Assign(regs.stencil_enable);
+    depth_test_func.Assign(PackComparisonOp(regs.depth_test_func));
+}
+
+void FixedPipelineState::Rasterizer::Fill(const Maxwell& regs) noexcept {
+    const auto& clip = regs.view_volume_clip_control;
+    const std::array enabled_lut = {regs.polygon_offset_point_enable,
+                                    regs.polygon_offset_line_enable,
+                                    regs.polygon_offset_fill_enable};
+    const u32 topology_index = static_cast<u32>(regs.draw.topology.Value());
+
+    u32 packed_front_face = PackFrontFace(regs.front_face);
+    if (regs.screen_y_control.triangle_rast_flip != 0 &&
+        regs.viewport_transform[0].scale_y > 0.0f) {
+        // Flip front face
+        packed_front_face = 1 - packed_front_face;
     }
-    return hash;
+
+    raw = 0;
+    topology.Assign(topology_index);
+    primitive_restart_enable.Assign(regs.primitive_restart.enabled != 0 ? 1 : 0);
+    cull_enable.Assign(regs.cull_test_enabled != 0 ? 1 : 0);
+    depth_bias_enable.Assign(enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]] != 0 ? 1 : 0);
+    depth_clamp_enable.Assign(clip.depth_clamp_near == 1 || clip.depth_clamp_far == 1 ? 1 : 0);
+    ndc_minus_one_to_one.Assign(regs.depth_mode == Maxwell::DepthMode::MinusOneToOne ? 1 : 0);
+    cull_face.Assign(PackCullFace(regs.cull_face));
+    front_face.Assign(packed_front_face);
+    polygon_mode.Assign(PackPolygonMode(regs.polygon_mode_front));
+    patch_control_points_minus_one.Assign(regs.patch_vertices - 1);
+    tessellation_primitive.Assign(static_cast<u32>(regs.tess_mode.prim.Value()));
+    tessellation_spacing.Assign(static_cast<u32>(regs.tess_mode.spacing.Value()));
+    tessellation_clockwise.Assign(regs.tess_mode.cw.Value());
+    logic_op_enable.Assign(regs.logic_op.enable != 0 ? 1 : 0);
+    logic_op.Assign(PackLogicOp(regs.logic_op.operation));
+    std::memcpy(&point_size, &regs.point_size, sizeof(point_size)); // TODO: C++20 std::bit_cast
 }
 
-bool FixedPipelineState::VertexInput::operator==(const VertexInput& rhs) const noexcept {
-    return std::equal(bindings.begin(), bindings.begin() + num_bindings, rhs.bindings.begin(),
-                      rhs.bindings.begin() + rhs.num_bindings) &&
-           std::equal(attributes.begin(), attributes.begin() + num_attributes,
-                      rhs.attributes.begin(), rhs.attributes.begin() + rhs.num_attributes);
-}
-
-std::size_t FixedPipelineState::InputAssembly::Hash() const noexcept {
-    std::size_t point_size_int = 0;
-    std::memcpy(&point_size_int, &point_size, sizeof(point_size));
-    return (static_cast<std::size_t>(topology) << 24) ^ (point_size_int << 32) ^
-           static_cast<std::size_t>(primitive_restart_enable);
-}
-
-bool FixedPipelineState::InputAssembly::operator==(const InputAssembly& rhs) const noexcept {
-    return std::tie(topology, primitive_restart_enable, point_size) ==
-           std::tie(rhs.topology, rhs.primitive_restart_enable, rhs.point_size);
-}
-
-std::size_t FixedPipelineState::Tessellation::Hash() const noexcept {
-    return static_cast<std::size_t>(patch_control_points) ^
-           (static_cast<std::size_t>(primitive) << 6) ^ (static_cast<std::size_t>(spacing) << 8) ^
-           (static_cast<std::size_t>(clockwise) << 10);
-}
-
-bool FixedPipelineState::Tessellation::operator==(const Tessellation& rhs) const noexcept {
-    return std::tie(patch_control_points, primitive, spacing, clockwise) ==
-           std::tie(rhs.patch_control_points, rhs.primitive, rhs.spacing, rhs.clockwise);
-}
-
-std::size_t FixedPipelineState::Rasterizer::Hash() const noexcept {
-    return static_cast<std::size_t>(cull_enable) ^
-           (static_cast<std::size_t>(depth_bias_enable) << 1) ^
-           (static_cast<std::size_t>(depth_clamp_enable) << 2) ^
-           (static_cast<std::size_t>(ndc_minus_one_to_one) << 3) ^
-           (static_cast<std::size_t>(cull_face) << 24) ^
-           (static_cast<std::size_t>(front_face) << 48);
-}
-
-bool FixedPipelineState::Rasterizer::operator==(const Rasterizer& rhs) const noexcept {
-    return std::tie(cull_enable, depth_bias_enable, depth_clamp_enable, ndc_minus_one_to_one,
-                    cull_face, front_face) ==
-           std::tie(rhs.cull_enable, rhs.depth_bias_enable, rhs.depth_clamp_enable,
-                    rhs.ndc_minus_one_to_one, rhs.cull_face, rhs.front_face);
-}
-
-std::size_t FixedPipelineState::DepthStencil::Hash() const noexcept {
-    std::size_t hash = static_cast<std::size_t>(depth_test_enable) ^
-                       (static_cast<std::size_t>(depth_write_enable) << 1) ^
-                       (static_cast<std::size_t>(depth_bounds_enable) << 2) ^
-                       (static_cast<std::size_t>(stencil_enable) << 3) ^
-                       (static_cast<std::size_t>(depth_test_function) << 4);
-    boost::hash_combine(hash, front_stencil.Hash());
-    boost::hash_combine(hash, back_stencil.Hash());
-    return hash;
-}
-
-bool FixedPipelineState::DepthStencil::operator==(const DepthStencil& rhs) const noexcept {
-    return std::tie(depth_test_enable, depth_write_enable, depth_bounds_enable, depth_test_function,
-                    stencil_enable, front_stencil, back_stencil) ==
-           std::tie(rhs.depth_test_enable, rhs.depth_write_enable, rhs.depth_bounds_enable,
-                    rhs.depth_test_function, rhs.stencil_enable, rhs.front_stencil,
-                    rhs.back_stencil);
-}
-
-std::size_t FixedPipelineState::ColorBlending::Hash() const noexcept {
-    std::size_t hash = attachments_count << 13;
-    for (std::size_t rt = 0; rt < static_cast<std::size_t>(attachments_count); ++rt) {
-        boost::hash_combine(hash, attachments[rt].Hash());
+void FixedPipelineState::ColorBlending::Fill(const Maxwell& regs) noexcept {
+    for (std::size_t index = 0; index < std::size(attachments); ++index) {
+        attachments[index].Fill(regs, index);
     }
-    return hash;
 }
 
-bool FixedPipelineState::ColorBlending::operator==(const ColorBlending& rhs) const noexcept {
-    return std::equal(attachments.begin(), attachments.begin() + attachments_count,
-                      rhs.attachments.begin(), rhs.attachments.begin() + rhs.attachments_count);
+void FixedPipelineState::BlendingAttachment::Fill(const Maxwell& regs, std::size_t index) {
+    const auto& mask = regs.color_mask[regs.color_mask_common ? 0 : index];
+
+    raw = 0;
+    mask_r.Assign(mask.R);
+    mask_g.Assign(mask.G);
+    mask_b.Assign(mask.B);
+    mask_a.Assign(mask.A);
+
+    // TODO: C++20 Use templated lambda to deduplicate code
+
+    if (!regs.independent_blend_enable) {
+        const auto& src = regs.blend;
+        if (!src.enable[index]) {
+            return;
+        }
+        equation_rgb.Assign(PackBlendEquation(src.equation_rgb));
+        equation_a.Assign(PackBlendEquation(src.equation_a));
+        factor_source_rgb.Assign(PackBlendFactor(src.factor_source_rgb));
+        factor_dest_rgb.Assign(PackBlendFactor(src.factor_dest_rgb));
+        factor_source_a.Assign(PackBlendFactor(src.factor_source_a));
+        factor_dest_a.Assign(PackBlendFactor(src.factor_dest_a));
+        enable.Assign(1);
+        return;
+    }
+
+    if (!regs.blend.enable[index]) {
+        return;
+    }
+    const auto& src = regs.independent_blend[index];
+    equation_rgb.Assign(PackBlendEquation(src.equation_rgb));
+    equation_a.Assign(PackBlendEquation(src.equation_a));
+    factor_source_rgb.Assign(PackBlendFactor(src.factor_source_rgb));
+    factor_dest_rgb.Assign(PackBlendFactor(src.factor_dest_rgb));
+    factor_source_a.Assign(PackBlendFactor(src.factor_source_a));
+    factor_dest_a.Assign(PackBlendFactor(src.factor_dest_a));
+    enable.Assign(1);
 }
 
 std::size_t FixedPipelineState::Hash() const noexcept {
-    std::size_t hash = 0;
-    boost::hash_combine(hash, vertex_input.Hash());
-    boost::hash_combine(hash, input_assembly.Hash());
-    boost::hash_combine(hash, tessellation.Hash());
-    boost::hash_combine(hash, rasterizer.Hash());
-    boost::hash_combine(hash, depth_stencil.Hash());
-    boost::hash_combine(hash, color_blending.Hash());
-    return hash;
+    const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), sizeof *this);
+    return static_cast<std::size_t>(hash);
 }
 
 bool FixedPipelineState::operator==(const FixedPipelineState& rhs) const noexcept {
-    return std::tie(vertex_input, input_assembly, tessellation, rasterizer, depth_stencil,
-                    color_blending) == std::tie(rhs.vertex_input, rhs.input_assembly,
-                                                rhs.tessellation, rhs.rasterizer, rhs.depth_stencil,
-                                                rhs.color_blending);
+    return std::memcmp(this, &rhs, sizeof *this) == 0;
 }
 
 FixedPipelineState GetFixedPipelineState(const Maxwell& regs) {
     FixedPipelineState fixed_state;
-    fixed_state.input_assembly = GetInputAssemblyState(regs);
-    fixed_state.tessellation = GetTessellationState(regs);
-    fixed_state.rasterizer = GetRasterizerState(regs);
-    fixed_state.depth_stencil = GetDepthStencilState(regs);
-    fixed_state.color_blending = GetColorBlendingState(regs);
+    fixed_state.rasterizer.Fill(regs);
+    fixed_state.depth_stencil.Fill(regs);
+    fixed_state.color_blending.Fill(regs);
+    fixed_state.padding = {};
     return fixed_state;
 }
 
+u32 FixedPipelineState::PackComparisonOp(Maxwell::ComparisonOp op) noexcept {
+    // OpenGL enums go from 0x200 to 0x207 and the others from 1 to 8
+    // If we substract 0x200 to OpenGL enums and 1 to the others we get a 0-7 range.
+    // Perfect for a hash.
+    const u32 value = static_cast<u32>(op);
+    return value - (value >= 0x200 ? 0x200 : 1);
+}
+
+Maxwell::ComparisonOp FixedPipelineState::UnpackComparisonOp(u32 packed) noexcept {
+    // Read PackComparisonOp for the logic behind this.
+    return static_cast<Maxwell::ComparisonOp>(packed + 1);
+}
+
+u32 FixedPipelineState::PackStencilOp(Maxwell::StencilOp op) noexcept {
+    switch (op) {
+    case Maxwell::StencilOp::Keep:
+    case Maxwell::StencilOp::KeepOGL:
+        return 0;
+    case Maxwell::StencilOp::Zero:
+    case Maxwell::StencilOp::ZeroOGL:
+        return 1;
+    case Maxwell::StencilOp::Replace:
+    case Maxwell::StencilOp::ReplaceOGL:
+        return 2;
+    case Maxwell::StencilOp::Incr:
+    case Maxwell::StencilOp::IncrOGL:
+        return 3;
+    case Maxwell::StencilOp::Decr:
+    case Maxwell::StencilOp::DecrOGL:
+        return 4;
+    case Maxwell::StencilOp::Invert:
+    case Maxwell::StencilOp::InvertOGL:
+        return 5;
+    case Maxwell::StencilOp::IncrWrap:
+    case Maxwell::StencilOp::IncrWrapOGL:
+        return 6;
+    case Maxwell::StencilOp::DecrWrap:
+    case Maxwell::StencilOp::DecrWrapOGL:
+        return 7;
+    }
+    return 0;
+}
+
+Maxwell::StencilOp FixedPipelineState::UnpackStencilOp(u32 packed) noexcept {
+    static constexpr std::array LUT = {Maxwell::StencilOp::Keep,     Maxwell::StencilOp::Zero,
+                                       Maxwell::StencilOp::Replace,  Maxwell::StencilOp::Incr,
+                                       Maxwell::StencilOp::Decr,     Maxwell::StencilOp::Invert,
+                                       Maxwell::StencilOp::IncrWrap, Maxwell::StencilOp::DecrWrap};
+    return LUT[packed];
+}
+
+u32 FixedPipelineState::PackCullFace(Maxwell::CullFace cull) noexcept {
+    // FrontAndBack is 0x408, by substracting 0x406 in it we get 2.
+    // Individual cull faces are in 0x404 and 0x405, substracting 0x404 we get 0 and 1.
+    const u32 value = static_cast<u32>(cull);
+    return value - (value == 0x408 ? 0x406 : 0x404);
+}
+
+Maxwell::CullFace FixedPipelineState::UnpackCullFace(u32 packed) noexcept {
+    static constexpr std::array LUT = {Maxwell::CullFace::Front, Maxwell::CullFace::Back,
+                                       Maxwell::CullFace::FrontAndBack};
+    return LUT[packed];
+}
+
+u32 FixedPipelineState::PackFrontFace(Maxwell::FrontFace face) noexcept {
+    return static_cast<u32>(face) - 0x900;
+}
+
+Maxwell::FrontFace FixedPipelineState::UnpackFrontFace(u32 packed) noexcept {
+    return static_cast<Maxwell::FrontFace>(packed + 0x900);
+}
+
+u32 FixedPipelineState::PackPolygonMode(Maxwell::PolygonMode mode) noexcept {
+    return static_cast<u32>(mode) - 0x1B00;
+}
+
+Maxwell::PolygonMode FixedPipelineState::UnpackPolygonMode(u32 packed) noexcept {
+    return static_cast<Maxwell::PolygonMode>(packed + 0x1B00);
+}
+
+u32 FixedPipelineState::PackLogicOp(Maxwell::LogicOperation op) noexcept {
+    return static_cast<u32>(op) - 0x1500;
+}
+
+Maxwell::LogicOperation FixedPipelineState::UnpackLogicOp(u32 packed) noexcept {
+    return static_cast<Maxwell::LogicOperation>(packed + 0x1500);
+}
+
+u32 FixedPipelineState::PackBlendEquation(Maxwell::Blend::Equation equation) noexcept {
+    switch (equation) {
+    case Maxwell::Blend::Equation::Add:
+    case Maxwell::Blend::Equation::AddGL:
+        return 0;
+    case Maxwell::Blend::Equation::Subtract:
+    case Maxwell::Blend::Equation::SubtractGL:
+        return 1;
+    case Maxwell::Blend::Equation::ReverseSubtract:
+    case Maxwell::Blend::Equation::ReverseSubtractGL:
+        return 2;
+    case Maxwell::Blend::Equation::Min:
+    case Maxwell::Blend::Equation::MinGL:
+        return 3;
+    case Maxwell::Blend::Equation::Max:
+    case Maxwell::Blend::Equation::MaxGL:
+        return 4;
+    }
+    return 0;
+}
+
+Maxwell::Blend::Equation FixedPipelineState::UnpackBlendEquation(u32 packed) noexcept {
+    static constexpr std::array LUT = {
+        Maxwell::Blend::Equation::Add, Maxwell::Blend::Equation::Subtract,
+        Maxwell::Blend::Equation::ReverseSubtract, Maxwell::Blend::Equation::Min,
+        Maxwell::Blend::Equation::Max};
+    return LUT[packed];
+}
+
+u32 FixedPipelineState::PackBlendFactor(Maxwell::Blend::Factor factor) noexcept {
+    switch (factor) {
+    case Maxwell::Blend::Factor::Zero:
+    case Maxwell::Blend::Factor::ZeroGL:
+        return 0;
+    case Maxwell::Blend::Factor::One:
+    case Maxwell::Blend::Factor::OneGL:
+        return 1;
+    case Maxwell::Blend::Factor::SourceColor:
+    case Maxwell::Blend::Factor::SourceColorGL:
+        return 2;
+    case Maxwell::Blend::Factor::OneMinusSourceColor:
+    case Maxwell::Blend::Factor::OneMinusSourceColorGL:
+        return 3;
+    case Maxwell::Blend::Factor::SourceAlpha:
+    case Maxwell::Blend::Factor::SourceAlphaGL:
+        return 4;
+    case Maxwell::Blend::Factor::OneMinusSourceAlpha:
+    case Maxwell::Blend::Factor::OneMinusSourceAlphaGL:
+        return 5;
+    case Maxwell::Blend::Factor::DestAlpha:
+    case Maxwell::Blend::Factor::DestAlphaGL:
+        return 6;
+    case Maxwell::Blend::Factor::OneMinusDestAlpha:
+    case Maxwell::Blend::Factor::OneMinusDestAlphaGL:
+        return 7;
+    case Maxwell::Blend::Factor::DestColor:
+    case Maxwell::Blend::Factor::DestColorGL:
+        return 8;
+    case Maxwell::Blend::Factor::OneMinusDestColor:
+    case Maxwell::Blend::Factor::OneMinusDestColorGL:
+        return 9;
+    case Maxwell::Blend::Factor::SourceAlphaSaturate:
+    case Maxwell::Blend::Factor::SourceAlphaSaturateGL:
+        return 10;
+    case Maxwell::Blend::Factor::Source1Color:
+    case Maxwell::Blend::Factor::Source1ColorGL:
+        return 11;
+    case Maxwell::Blend::Factor::OneMinusSource1Color:
+    case Maxwell::Blend::Factor::OneMinusSource1ColorGL:
+        return 12;
+    case Maxwell::Blend::Factor::Source1Alpha:
+    case Maxwell::Blend::Factor::Source1AlphaGL:
+        return 13;
+    case Maxwell::Blend::Factor::OneMinusSource1Alpha:
+    case Maxwell::Blend::Factor::OneMinusSource1AlphaGL:
+        return 14;
+    case Maxwell::Blend::Factor::ConstantColor:
+    case Maxwell::Blend::Factor::ConstantColorGL:
+        return 15;
+    case Maxwell::Blend::Factor::OneMinusConstantColor:
+    case Maxwell::Blend::Factor::OneMinusConstantColorGL:
+        return 16;
+    case Maxwell::Blend::Factor::ConstantAlpha:
+    case Maxwell::Blend::Factor::ConstantAlphaGL:
+        return 17;
+    case Maxwell::Blend::Factor::OneMinusConstantAlpha:
+    case Maxwell::Blend::Factor::OneMinusConstantAlphaGL:
+        return 18;
+    }
+    return 0;
+}
+
+Maxwell::Blend::Factor FixedPipelineState::UnpackBlendFactor(u32 packed) noexcept {
+    static constexpr std::array LUT = {
+        Maxwell::Blend::Factor::Zero,
+        Maxwell::Blend::Factor::One,
+        Maxwell::Blend::Factor::SourceColor,
+        Maxwell::Blend::Factor::OneMinusSourceColor,
+        Maxwell::Blend::Factor::SourceAlpha,
+        Maxwell::Blend::Factor::OneMinusSourceAlpha,
+        Maxwell::Blend::Factor::DestAlpha,
+        Maxwell::Blend::Factor::OneMinusDestAlpha,
+        Maxwell::Blend::Factor::DestColor,
+        Maxwell::Blend::Factor::OneMinusDestColor,
+        Maxwell::Blend::Factor::SourceAlphaSaturate,
+        Maxwell::Blend::Factor::Source1Color,
+        Maxwell::Blend::Factor::OneMinusSource1Color,
+        Maxwell::Blend::Factor::Source1Alpha,
+        Maxwell::Blend::Factor::OneMinusSource1Alpha,
+        Maxwell::Blend::Factor::ConstantColor,
+        Maxwell::Blend::Factor::OneMinusConstantColor,
+        Maxwell::Blend::Factor::ConstantAlpha,
+        Maxwell::Blend::Factor::OneMinusConstantAlpha,
+    };
+    return LUT[packed];
+}
+
 } // namespace Vulkan

src/video_core/renderer_vulkan/fixed_pipeline_state.h

@@ -7,6 +7,7 @@
 #include <array>
 #include <type_traits>
 
+#include "common/bit_field.h"
 #include "common/common_types.h"
 
 #include "video_core/engines/maxwell_3d.h"
@@ -16,93 +17,48 @@ namespace Vulkan {
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
-// TODO(Rodrigo): Optimize this structure.
+struct alignas(32) FixedPipelineState {
+    static u32 PackComparisonOp(Maxwell::ComparisonOp op) noexcept;
+    static Maxwell::ComparisonOp UnpackComparisonOp(u32 packed) noexcept;
 
-struct FixedPipelineState {
-    using PixelFormat = VideoCore::Surface::PixelFormat;
+    static u32 PackStencilOp(Maxwell::StencilOp op) noexcept;
+    static Maxwell::StencilOp UnpackStencilOp(u32 packed) noexcept;
 
-    struct VertexBinding {
-        constexpr VertexBinding(u32 index, u32 stride, u32 divisor)
-            : index{index}, stride{stride}, divisor{divisor} {}
-        VertexBinding() = default;
+    static u32 PackCullFace(Maxwell::CullFace cull) noexcept;
+    static Maxwell::CullFace UnpackCullFace(u32 packed) noexcept;
 
-        u32 index;
-        u32 stride;
-        u32 divisor;
+    static u32 PackFrontFace(Maxwell::FrontFace face) noexcept;
+    static Maxwell::FrontFace UnpackFrontFace(u32 packed) noexcept;
 
-        std::size_t Hash() const noexcept;
+    static u32 PackPolygonMode(Maxwell::PolygonMode mode) noexcept;
+    static Maxwell::PolygonMode UnpackPolygonMode(u32 packed) noexcept;
 
-        bool operator==(const VertexBinding& rhs) const noexcept;
+    static u32 PackLogicOp(Maxwell::LogicOperation op) noexcept;
+    static Maxwell::LogicOperation UnpackLogicOp(u32 packed) noexcept;
 
-        bool operator!=(const VertexBinding& rhs) const noexcept {
-            return !operator==(rhs);
-        }
-    };
+    static u32 PackBlendEquation(Maxwell::Blend::Equation equation) noexcept;
+    static Maxwell::Blend::Equation UnpackBlendEquation(u32 packed) noexcept;
 
-    struct VertexAttribute {
-        constexpr VertexAttribute(u32 index, u32 buffer, Maxwell::VertexAttribute::Type type,
-                                  Maxwell::VertexAttribute::Size size, u32 offset)
-            : index{index}, buffer{buffer}, type{type}, size{size}, offset{offset} {}
-        VertexAttribute() = default;
-
-        u32 index;
-        u32 buffer;
-        Maxwell::VertexAttribute::Type type;
-        Maxwell::VertexAttribute::Size size;
-        u32 offset;
-
-        std::size_t Hash() const noexcept;
-
-        bool operator==(const VertexAttribute& rhs) const noexcept;
-
-        bool operator!=(const VertexAttribute& rhs) const noexcept {
-            return !operator==(rhs);
-        }
-    };
-
-    struct StencilFace {
-        constexpr StencilFace(Maxwell::StencilOp action_stencil_fail,
-                              Maxwell::StencilOp action_depth_fail,
-                              Maxwell::StencilOp action_depth_pass, Maxwell::ComparisonOp test_func)
-            : action_stencil_fail{action_stencil_fail}, action_depth_fail{action_depth_fail},
-              action_depth_pass{action_depth_pass}, test_func{test_func} {}
-        StencilFace() = default;
-
-        Maxwell::StencilOp action_stencil_fail;
-        Maxwell::StencilOp action_depth_fail;
-        Maxwell::StencilOp action_depth_pass;
-        Maxwell::ComparisonOp test_func;
-
-        std::size_t Hash() const noexcept;
-
-        bool operator==(const StencilFace& rhs) const noexcept;
-
-        bool operator!=(const StencilFace& rhs) const noexcept {
-            return !operator==(rhs);
-        }
-    };
+    static u32 PackBlendFactor(Maxwell::Blend::Factor factor) noexcept;
+    static Maxwell::Blend::Factor UnpackBlendFactor(u32 packed) noexcept;
 
     struct BlendingAttachment {
-        constexpr BlendingAttachment(bool enable, Maxwell::Blend::Equation rgb_equation,
-                                     Maxwell::Blend::Factor src_rgb_func,
-                                     Maxwell::Blend::Factor dst_rgb_func,
-                                     Maxwell::Blend::Equation a_equation,
-                                     Maxwell::Blend::Factor src_a_func,
-                                     Maxwell::Blend::Factor dst_a_func,
-                                     std::array<bool, 4> components)
-            : enable{enable}, rgb_equation{rgb_equation}, src_rgb_func{src_rgb_func},
-              dst_rgb_func{dst_rgb_func}, a_equation{a_equation}, src_a_func{src_a_func},
-              dst_a_func{dst_a_func}, components{components} {}
-        BlendingAttachment() = default;
+        union {
+            u32 raw;
+            BitField<0, 1, u32> mask_r;
+            BitField<1, 1, u32> mask_g;
+            BitField<2, 1, u32> mask_b;
+            BitField<3, 1, u32> mask_a;
+            BitField<4, 3, u32> equation_rgb;
+            BitField<7, 3, u32> equation_a;
+            BitField<10, 5, u32> factor_source_rgb;
+            BitField<15, 5, u32> factor_dest_rgb;
+            BitField<20, 5, u32> factor_source_a;
+            BitField<25, 5, u32> factor_dest_a;
+            BitField<30, 1, u32> enable;
+        };
 
-        bool enable;
-        Maxwell::Blend::Equation rgb_equation;
-        Maxwell::Blend::Factor src_rgb_func;
-        Maxwell::Blend::Factor dst_rgb_func;
-        Maxwell::Blend::Equation a_equation;
-        Maxwell::Blend::Factor src_a_func;
-        Maxwell::Blend::Factor dst_a_func;
-        std::array<bool, 4> components;
+        void Fill(const Maxwell& regs, std::size_t index);
 
         std::size_t Hash() const noexcept;
 
@@ -111,136 +67,178 @@ struct FixedPipelineState {
         bool operator!=(const BlendingAttachment& rhs) const noexcept {
             return !operator==(rhs);
         }
+
+        constexpr std::array<bool, 4> Mask() const noexcept {
+            return {mask_r != 0, mask_g != 0, mask_b != 0, mask_a != 0};
+        }
+
+        Maxwell::Blend::Equation EquationRGB() const noexcept {
+            return UnpackBlendEquation(equation_rgb.Value());
+        }
+
+        Maxwell::Blend::Equation EquationAlpha() const noexcept {
+            return UnpackBlendEquation(equation_a.Value());
+        }
+
+        Maxwell::Blend::Factor SourceRGBFactor() const noexcept {
+            return UnpackBlendFactor(factor_source_rgb.Value());
+        }
+
+        Maxwell::Blend::Factor DestRGBFactor() const noexcept {
+            return UnpackBlendFactor(factor_dest_rgb.Value());
+        }
+
+        Maxwell::Blend::Factor SourceAlphaFactor() const noexcept {
+            return UnpackBlendFactor(factor_source_a.Value());
+        }
+
+        Maxwell::Blend::Factor DestAlphaFactor() const noexcept {
+            return UnpackBlendFactor(factor_dest_a.Value());
+        }
     };
 
     struct VertexInput {
-        std::size_t num_bindings = 0;
-        std::size_t num_attributes = 0;
-        std::array<VertexBinding, Maxwell::NumVertexArrays> bindings;
-        std::array<VertexAttribute, Maxwell::NumVertexAttributes> attributes;
+        union Binding {
+            u16 raw;
+            BitField<0, 1, u16> enabled;
+            BitField<1, 12, u16> stride;
+        };
 
-        std::size_t Hash() const noexcept;
+        union Attribute {
+            u32 raw;
+            BitField<0, 1, u32> enabled;
+            BitField<1, 5, u32> buffer;
+            BitField<6, 14, u32> offset;
+            BitField<20, 3, u32> type;
+            BitField<23, 6, u32> size;
 
-        bool operator==(const VertexInput& rhs) const noexcept;
+            constexpr Maxwell::VertexAttribute::Type Type() const noexcept {
+                return static_cast<Maxwell::VertexAttribute::Type>(type.Value());
+            }
 
-        bool operator!=(const VertexInput& rhs) const noexcept {
-            return !operator==(rhs);
+            constexpr Maxwell::VertexAttribute::Size Size() const noexcept {
+                return static_cast<Maxwell::VertexAttribute::Size>(size.Value());
+            }
+        };
+
+        std::array<Binding, Maxwell::NumVertexArrays> bindings;
+        std::array<u32, Maxwell::NumVertexArrays> binding_divisors;
+        std::array<Attribute, Maxwell::NumVertexAttributes> attributes;
+
+        void SetBinding(std::size_t index, bool enabled, u32 stride, u32 divisor) noexcept {
+            auto& binding = bindings[index];
+            binding.raw = 0;
+            binding.enabled.Assign(enabled ? 1 : 0);
+            binding.stride.Assign(stride);
+            binding_divisors[index] = divisor;
         }
-    };
 
-    struct InputAssembly {
-        constexpr InputAssembly(Maxwell::PrimitiveTopology topology, bool primitive_restart_enable,
-                                float point_size)
-            : topology{topology}, primitive_restart_enable{primitive_restart_enable},
-              point_size{point_size} {}
-        InputAssembly() = default;
-
-        Maxwell::PrimitiveTopology topology;
-        bool primitive_restart_enable;
-        float point_size;
-
-        std::size_t Hash() const noexcept;
-
-        bool operator==(const InputAssembly& rhs) const noexcept;
-
-        bool operator!=(const InputAssembly& rhs) const noexcept {
-            return !operator==(rhs);
-        }
-    };
-
-    struct Tessellation {
-        constexpr Tessellation(u32 patch_control_points, Maxwell::TessellationPrimitive primitive,
-                               Maxwell::TessellationSpacing spacing, bool clockwise)
-            : patch_control_points{patch_control_points}, primitive{primitive}, spacing{spacing},
-              clockwise{clockwise} {}
-        Tessellation() = default;
-
-        u32 patch_control_points;
-        Maxwell::TessellationPrimitive primitive;
-        Maxwell::TessellationSpacing spacing;
-        bool clockwise;
-
-        std::size_t Hash() const noexcept;
-
-        bool operator==(const Tessellation& rhs) const noexcept;
-
-        bool operator!=(const Tessellation& rhs) const noexcept {
-            return !operator==(rhs);
+        void SetAttribute(std::size_t index, bool enabled, u32 buffer, u32 offset,
+                          Maxwell::VertexAttribute::Type type,
+                          Maxwell::VertexAttribute::Size size) noexcept {
+            auto& attribute = attributes[index];
+            attribute.raw = 0;
+            attribute.enabled.Assign(enabled ? 1 : 0);
+            attribute.buffer.Assign(buffer);
+            attribute.offset.Assign(offset);
+            attribute.type.Assign(static_cast<u32>(type));
+            attribute.size.Assign(static_cast<u32>(size));
         }
     };
 
     struct Rasterizer {
-        constexpr Rasterizer(bool cull_enable, bool depth_bias_enable, bool depth_clamp_enable,
-                             bool ndc_minus_one_to_one, Maxwell::CullFace cull_face,
-                             Maxwell::FrontFace front_face)
-            : cull_enable{cull_enable}, depth_bias_enable{depth_bias_enable},
-              depth_clamp_enable{depth_clamp_enable}, ndc_minus_one_to_one{ndc_minus_one_to_one},
-              cull_face{cull_face}, front_face{front_face} {}
-        Rasterizer() = default;
+        union {
+            u32 raw;
+            BitField<0, 4, u32> topology;
+            BitField<4, 1, u32> primitive_restart_enable;
+            BitField<5, 1, u32> cull_enable;
+            BitField<6, 1, u32> depth_bias_enable;
+            BitField<7, 1, u32> depth_clamp_enable;
+            BitField<8, 1, u32> ndc_minus_one_to_one;
+            BitField<9, 2, u32> cull_face;
+            BitField<11, 1, u32> front_face;
+            BitField<12, 2, u32> polygon_mode;
+            BitField<14, 5, u32> patch_control_points_minus_one;
+            BitField<19, 2, u32> tessellation_primitive;
+            BitField<21, 2, u32> tessellation_spacing;
+            BitField<23, 1, u32> tessellation_clockwise;
+            BitField<24, 1, u32> logic_op_enable;
+            BitField<25, 4, u32> logic_op;
+        };
 
-        bool cull_enable;
-        bool depth_bias_enable;
-        bool depth_clamp_enable;
-        bool ndc_minus_one_to_one;
-        Maxwell::CullFace cull_face;
-        Maxwell::FrontFace front_face;
+        // TODO(Rodrigo): Move this to push constants
+        u32 point_size;
 
-        std::size_t Hash() const noexcept;
+        void Fill(const Maxwell& regs) noexcept;
 
-        bool operator==(const Rasterizer& rhs) const noexcept;
+        constexpr Maxwell::PrimitiveTopology Topology() const noexcept {
+            return static_cast<Maxwell::PrimitiveTopology>(topology.Value());
+        }
 
-        bool operator!=(const Rasterizer& rhs) const noexcept {
-            return !operator==(rhs);
+        Maxwell::CullFace CullFace() const noexcept {
+            return UnpackCullFace(cull_face.Value());
+        }
+
+        Maxwell::FrontFace FrontFace() const noexcept {
+            return UnpackFrontFace(front_face.Value());
         }
     };
 
     struct DepthStencil {
-        constexpr DepthStencil(bool depth_test_enable, bool depth_write_enable,
-                               bool depth_bounds_enable, bool stencil_enable,
-                               Maxwell::ComparisonOp depth_test_function, StencilFace front_stencil,
-                               StencilFace back_stencil)
-            : depth_test_enable{depth_test_enable}, depth_write_enable{depth_write_enable},
-              depth_bounds_enable{depth_bounds_enable}, stencil_enable{stencil_enable},
-              depth_test_function{depth_test_function}, front_stencil{front_stencil},
-              back_stencil{back_stencil} {}
-        DepthStencil() = default;
+        template <std::size_t Position>
+        union StencilFace {
+            BitField<Position + 0, 3, u32> action_stencil_fail;
+            BitField<Position + 3, 3, u32> action_depth_fail;
+            BitField<Position + 6, 3, u32> action_depth_pass;
+            BitField<Position + 9, 3, u32> test_func;
 
-        bool depth_test_enable;
-        bool depth_write_enable;
-        bool depth_bounds_enable;
-        bool stencil_enable;
-        Maxwell::ComparisonOp depth_test_function;
-        StencilFace front_stencil;
-        StencilFace back_stencil;
+            Maxwell::StencilOp ActionStencilFail() const noexcept {
+                return UnpackStencilOp(action_stencil_fail);
+            }
 
-        std::size_t Hash() const noexcept;
+            Maxwell::StencilOp ActionDepthFail() const noexcept {
+                return UnpackStencilOp(action_depth_fail);
+            }
 
-        bool operator==(const DepthStencil& rhs) const noexcept;
+            Maxwell::StencilOp ActionDepthPass() const noexcept {
+                return UnpackStencilOp(action_depth_pass);
+            }
 
-        bool operator!=(const DepthStencil& rhs) const noexcept {
-            return !operator==(rhs);
+            Maxwell::ComparisonOp TestFunc() const noexcept {
+                return UnpackComparisonOp(test_func);
+            }
+        };
+
+        union {
+            u32 raw;
+            StencilFace<0> front;
+            StencilFace<12> back;
+            BitField<24, 1, u32> depth_test_enable;
+            BitField<25, 1, u32> depth_write_enable;
+            BitField<26, 1, u32> depth_bounds_enable;
+            BitField<27, 1, u32> stencil_enable;
+            BitField<28, 3, u32> depth_test_func;
+        };
+
+        void Fill(const Maxwell& regs) noexcept;
+
+        Maxwell::ComparisonOp DepthTestFunc() const noexcept {
+            return UnpackComparisonOp(depth_test_func);
         }
     };
 
     struct ColorBlending {
-        constexpr ColorBlending(
-            std::array<float, 4> blend_constants, std::size_t attachments_count,
-            std::array<BlendingAttachment, Maxwell::NumRenderTargets> attachments)
-            : attachments_count{attachments_count}, attachments{attachments} {}
-        ColorBlending() = default;
-
-        std::size_t attachments_count;
         std::array<BlendingAttachment, Maxwell::NumRenderTargets> attachments;
 
-        std::size_t Hash() const noexcept;
-
-        bool operator==(const ColorBlending& rhs) const noexcept;
-
-        bool operator!=(const ColorBlending& rhs) const noexcept {
-            return !operator==(rhs);
-        }
+        void Fill(const Maxwell& regs) noexcept;
     };
 
+    VertexInput vertex_input;
+    Rasterizer rasterizer;
+    DepthStencil depth_stencil;
+    ColorBlending color_blending;
+    std::array<u8, 20> padding;
+
     std::size_t Hash() const noexcept;
 
     bool operator==(const FixedPipelineState& rhs) const noexcept;
@@ -248,25 +246,11 @@ struct FixedPipelineState {
     bool operator!=(const FixedPipelineState& rhs) const noexcept {
         return !operator==(rhs);
     }
-
-    VertexInput vertex_input;
-    InputAssembly input_assembly;
-    Tessellation tessellation;
-    Rasterizer rasterizer;
-    DepthStencil depth_stencil;
-    ColorBlending color_blending;
 };
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::VertexBinding>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::VertexAttribute>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::StencilFace>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::BlendingAttachment>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::VertexInput>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::InputAssembly>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::Tessellation>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::Rasterizer>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::DepthStencil>);
-static_assert(std::is_trivially_copyable_v<FixedPipelineState::ColorBlending>);
+static_assert(std::has_unique_object_representations_v<FixedPipelineState>);
 static_assert(std::is_trivially_copyable_v<FixedPipelineState>);
+static_assert(std::is_trivially_constructible_v<FixedPipelineState>);
+static_assert(sizeof(FixedPipelineState) % 32 == 0, "Size is not aligned");
 
 FixedPipelineState GetFixedPipelineState(const Maxwell& regs);
 

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -26,12 +26,13 @@ MICROPROFILE_DECLARE(Vulkan_PipelineCache);
 
 namespace {
 
-VkStencilOpState GetStencilFaceState(const FixedPipelineState::StencilFace& face) {
+template <class StencilFace>
+VkStencilOpState GetStencilFaceState(const StencilFace& face) {
     VkStencilOpState state;
-    state.failOp = MaxwellToVK::StencilOp(face.action_stencil_fail);
-    state.passOp = MaxwellToVK::StencilOp(face.action_depth_pass);
-    state.depthFailOp = MaxwellToVK::StencilOp(face.action_depth_fail);
-    state.compareOp = MaxwellToVK::ComparisonOp(face.test_func);
+    state.failOp = MaxwellToVK::StencilOp(face.ActionStencilFail());
+    state.passOp = MaxwellToVK::StencilOp(face.ActionDepthPass());
+    state.depthFailOp = MaxwellToVK::StencilOp(face.ActionDepthFail());
+    state.compareOp = MaxwellToVK::ComparisonOp(face.TestFunc());
     state.compareMask = 0;
     state.writeMask = 0;
     state.reference = 0;
@@ -157,43 +158,47 @@ std::vector<vk::ShaderModule> VKGraphicsPipeline::CreateShaderModules(
 vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpass_params,
                                                 const SPIRVProgram& program) const {
     const auto& vi = fixed_state.vertex_input;
-    const auto& ia = fixed_state.input_assembly;
     const auto& ds = fixed_state.depth_stencil;
     const auto& cd = fixed_state.color_blending;
-    const auto& ts = fixed_state.tessellation;
     const auto& rs = fixed_state.rasterizer;
 
     std::vector<VkVertexInputBindingDescription> vertex_bindings;
     std::vector<VkVertexInputBindingDivisorDescriptionEXT> vertex_binding_divisors;
-    for (std::size_t i = 0; i < vi.num_bindings; ++i) {
-        const auto& binding = vi.bindings[i];
-        const bool instanced = binding.divisor != 0;
+    for (std::size_t index = 0; index < std::size(vi.bindings); ++index) {
+        const auto& binding = vi.bindings[index];
+        if (!binding.enabled) {
+            continue;
+        }
+        const bool instanced = vi.binding_divisors[index] != 0;
         const auto rate = instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX;
 
         auto& vertex_binding = vertex_bindings.emplace_back();
-        vertex_binding.binding = binding.index;
+        vertex_binding.binding = static_cast<u32>(index);
         vertex_binding.stride = binding.stride;
         vertex_binding.inputRate = rate;
 
         if (instanced) {
             auto& binding_divisor = vertex_binding_divisors.emplace_back();
-            binding_divisor.binding = binding.index;
-            binding_divisor.divisor = binding.divisor;
+            binding_divisor.binding = static_cast<u32>(index);
+            binding_divisor.divisor = vi.binding_divisors[index];
         }
     }
 
     std::vector<VkVertexInputAttributeDescription> vertex_attributes;
     const auto& input_attributes = program[0]->entries.attributes;
-    for (std::size_t i = 0; i < vi.num_attributes; ++i) {
-        const auto& attribute = vi.attributes[i];
-        if (input_attributes.find(attribute.index) == input_attributes.end()) {
+    for (std::size_t index = 0; index < std::size(vi.attributes); ++index) {
+        const auto& attribute = vi.attributes[index];
+        if (!attribute.enabled) {
+            continue;
+        }
+        if (input_attributes.find(static_cast<u32>(index)) == input_attributes.end()) {
             // Skip attributes not used by the vertex shaders.
             continue;
         }
         auto& vertex_attribute = vertex_attributes.emplace_back();
-        vertex_attribute.location = attribute.index;
+        vertex_attribute.location = static_cast<u32>(index);
         vertex_attribute.binding = attribute.buffer;
-        vertex_attribute.format = MaxwellToVK::VertexFormat(attribute.type, attribute.size);
+        vertex_attribute.format = MaxwellToVK::VertexFormat(attribute.Type(), attribute.Size());
         vertex_attribute.offset = attribute.offset;
     }
 
@@ -219,15 +224,15 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
     input_assembly_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO;
     input_assembly_ci.pNext = nullptr;
     input_assembly_ci.flags = 0;
-    input_assembly_ci.topology = MaxwellToVK::PrimitiveTopology(device, ia.topology);
+    input_assembly_ci.topology = MaxwellToVK::PrimitiveTopology(device, rs.Topology());
     input_assembly_ci.primitiveRestartEnable =
-        ia.primitive_restart_enable && SupportsPrimitiveRestart(input_assembly_ci.topology);
+        rs.primitive_restart_enable != 0 && SupportsPrimitiveRestart(input_assembly_ci.topology);
 
     VkPipelineTessellationStateCreateInfo tessellation_ci;
     tessellation_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO;
     tessellation_ci.pNext = nullptr;
     tessellation_ci.flags = 0;
-    tessellation_ci.patchControlPoints = ts.patch_control_points;
+    tessellation_ci.patchControlPoints = rs.patch_control_points_minus_one.Value() + 1;
 
     VkPipelineViewportStateCreateInfo viewport_ci;
     viewport_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO;
@@ -246,8 +251,8 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
     rasterization_ci.rasterizerDiscardEnable = VK_FALSE;
     rasterization_ci.polygonMode = VK_POLYGON_MODE_FILL;
     rasterization_ci.cullMode =
-        rs.cull_enable ? MaxwellToVK::CullFace(rs.cull_face) : VK_CULL_MODE_NONE;
-    rasterization_ci.frontFace = MaxwellToVK::FrontFace(rs.front_face);
+        rs.cull_enable ? MaxwellToVK::CullFace(rs.CullFace()) : VK_CULL_MODE_NONE;
+    rasterization_ci.frontFace = MaxwellToVK::FrontFace(rs.FrontFace());
     rasterization_ci.depthBiasEnable = rs.depth_bias_enable;
     rasterization_ci.depthBiasConstantFactor = 0.0f;
     rasterization_ci.depthBiasClamp = 0.0f;
@@ -271,40 +276,38 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
     depth_stencil_ci.flags = 0;
     depth_stencil_ci.depthTestEnable = ds.depth_test_enable;
     depth_stencil_ci.depthWriteEnable = ds.depth_write_enable;
-    depth_stencil_ci.depthCompareOp = ds.depth_test_enable
-                                          ? MaxwellToVK::ComparisonOp(ds.depth_test_function)
-                                          : VK_COMPARE_OP_ALWAYS;
+    depth_stencil_ci.depthCompareOp =
+        ds.depth_test_enable ? MaxwellToVK::ComparisonOp(ds.DepthTestFunc()) : VK_COMPARE_OP_ALWAYS;
     depth_stencil_ci.depthBoundsTestEnable = ds.depth_bounds_enable;
     depth_stencil_ci.stencilTestEnable = ds.stencil_enable;
-    depth_stencil_ci.front = GetStencilFaceState(ds.front_stencil);
-    depth_stencil_ci.back = GetStencilFaceState(ds.back_stencil);
+    depth_stencil_ci.front = GetStencilFaceState(ds.front);
+    depth_stencil_ci.back = GetStencilFaceState(ds.back);
     depth_stencil_ci.minDepthBounds = 0.0f;
     depth_stencil_ci.maxDepthBounds = 0.0f;
 
     std::array<VkPipelineColorBlendAttachmentState, Maxwell::NumRenderTargets> cb_attachments;
-    const std::size_t num_attachments =
-        std::min(cd.attachments_count, renderpass_params.color_attachments.size());
-    for (std::size_t i = 0; i < num_attachments; ++i) {
-        static constexpr std::array component_table = {
+    const std::size_t num_attachments = renderpass_params.color_attachments.size();
+    for (std::size_t index = 0; index < num_attachments; ++index) {
+        static constexpr std::array COMPONENT_TABLE = {
             VK_COLOR_COMPONENT_R_BIT, VK_COLOR_COMPONENT_G_BIT, VK_COLOR_COMPONENT_B_BIT,
             VK_COLOR_COMPONENT_A_BIT};
-        const auto& blend = cd.attachments[i];
+        const auto& blend = cd.attachments[index];
 
         VkColorComponentFlags color_components = 0;
-        for (std::size_t j = 0; j < component_table.size(); ++j) {
-            if (blend.components[j]) {
-                color_components |= component_table[j];
+        for (std::size_t i = 0; i < COMPONENT_TABLE.size(); ++i) {
+            if (blend.Mask()[i]) {
+                color_components |= COMPONENT_TABLE[i];
             }
         }
 
-        VkPipelineColorBlendAttachmentState& attachment = cb_attachments[i];
-        attachment.blendEnable = blend.enable;
-        attachment.srcColorBlendFactor = MaxwellToVK::BlendFactor(blend.src_rgb_func);
-        attachment.dstColorBlendFactor = MaxwellToVK::BlendFactor(blend.dst_rgb_func);
-        attachment.colorBlendOp = MaxwellToVK::BlendEquation(blend.rgb_equation);
-        attachment.srcAlphaBlendFactor = MaxwellToVK::BlendFactor(blend.src_a_func);
-        attachment.dstAlphaBlendFactor = MaxwellToVK::BlendFactor(blend.dst_a_func);
-        attachment.alphaBlendOp = MaxwellToVK::BlendEquation(blend.a_equation);
+        VkPipelineColorBlendAttachmentState& attachment = cb_attachments[index];
+        attachment.blendEnable = blend.enable != 0;
+        attachment.srcColorBlendFactor = MaxwellToVK::BlendFactor(blend.SourceRGBFactor());
+        attachment.dstColorBlendFactor = MaxwellToVK::BlendFactor(blend.DestRGBFactor());
+        attachment.colorBlendOp = MaxwellToVK::BlendEquation(blend.EquationRGB());
+        attachment.srcAlphaBlendFactor = MaxwellToVK::BlendFactor(blend.SourceAlphaFactor());
+        attachment.dstAlphaBlendFactor = MaxwellToVK::BlendFactor(blend.DestAlphaFactor());
+        attachment.alphaBlendOp = MaxwellToVK::BlendEquation(blend.EquationAlpha());
         attachment.colorWriteMask = color_components;
     }
 

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -329,12 +329,12 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
     const auto& gpu = system.GPU().Maxwell3D();
 
     Specialization specialization;
-    if (fixed_state.input_assembly.topology == Maxwell::PrimitiveTopology::Points) {
-        ASSERT(fixed_state.input_assembly.point_size != 0.0f);
-        specialization.point_size = fixed_state.input_assembly.point_size;
+    if (fixed_state.rasterizer.Topology() == Maxwell::PrimitiveTopology::Points) {
+        ASSERT(fixed_state.rasterizer.point_size != 0);
+        std::memcpy(&specialization.point_size, &fixed_state.rasterizer.point_size, sizeof(u32));
     }
     for (std::size_t i = 0; i < Maxwell::NumVertexAttributes; ++i) {
-        specialization.attribute_types[i] = fixed_state.vertex_input.attributes[i].type;
+        specialization.attribute_types[i] = fixed_state.vertex_input.attributes[i].Type();
     }
     specialization.ndc_minus_one_to_one = fixed_state.rasterizer.ndc_minus_one_to_one;
 

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -292,8 +292,8 @@ RasterizerVulkan::RasterizerVulkan(Core::System& system, Core::Frontend::EmuWind
       staging_pool(device, memory_manager, scheduler), descriptor_pool(device),
       update_descriptor_queue(device, scheduler), renderpass_cache(device),
       quad_array_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
-      uint8_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
       quad_indexed_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
+      uint8_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
       texture_cache(system, *this, device, resource_manager, memory_manager, scheduler,
                     staging_pool),
       pipeline_cache(system, *this, device, scheduler, descriptor_pool, update_descriptor_queue,
@@ -807,25 +807,29 @@ void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex
                                          BufferBindings& buffer_bindings) {
     const auto& regs = system.GPU().Maxwell3D().regs;
 
-    for (u32 index = 0; index < static_cast<u32>(Maxwell::NumVertexAttributes); ++index) {
+    for (std::size_t index = 0; index < Maxwell::NumVertexAttributes; ++index) {
         const auto& attrib = regs.vertex_attrib_format[index];
         if (!attrib.IsValid()) {
+            vertex_input.SetAttribute(index, false, 0, 0, {}, {});
             continue;
         }
 
-        const auto& buffer = regs.vertex_array[attrib.buffer];
+        [[maybe_unused]] const auto& buffer = regs.vertex_array[attrib.buffer];
         ASSERT(buffer.IsEnabled());
 
-        vertex_input.attributes[vertex_input.num_attributes++] =
-            FixedPipelineState::VertexAttribute(index, attrib.buffer, attrib.type, attrib.size,
-                                                attrib.offset);
+        vertex_input.SetAttribute(index, true, attrib.buffer, attrib.offset, attrib.type.Value(),
+                                  attrib.size.Value());
     }
 
-    for (u32 index = 0; index < static_cast<u32>(Maxwell::NumVertexArrays); ++index) {
+    for (std::size_t index = 0; index < Maxwell::NumVertexArrays; ++index) {
         const auto& vertex_array = regs.vertex_array[index];
         if (!vertex_array.IsEnabled()) {
+            vertex_input.SetBinding(index, false, 0, 0);
             continue;
         }
+        vertex_input.SetBinding(
+            index, true, vertex_array.stride,
+            regs.instanced_arrays.IsInstancingEnabled(index) ? vertex_array.divisor : 0);
 
         const GPUVAddr start{vertex_array.StartAddress()};
         const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
@@ -833,10 +837,6 @@ void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex
         ASSERT(end > start);
         const std::size_t size{end - start + 1};
         const auto [buffer, offset] = buffer_cache.UploadMemory(start, size);
-
-        vertex_input.bindings[vertex_input.num_bindings++] = FixedPipelineState::VertexBinding(
-            index, vertex_array.stride,
-            regs.instanced_arrays.IsInstancingEnabled(index) ? vertex_array.divisor : 0);
         buffer_bindings.AddVertexBinding(buffer, offset);
     }
 }

src/video_core/shader/decode/memory.cpp

@@ -479,7 +479,7 @@ std::tuple<Node, Node, GlobalMemoryBase> ShaderIR::TrackGlobalMemory(NodeBlock&
     bb.push_back(Comment(fmt::format("Base address is c[0x{:x}][0x{:x}]", index, offset)));
 
     const GlobalMemoryBase descriptor{index, offset};
-    const auto& [entry, is_new] = used_global_memory.try_emplace(descriptor);
+    const auto& entry = used_global_memory.try_emplace(descriptor).first;
     auto& usage = entry->second;
     usage.is_written |= is_write;
     usage.is_read |= is_read;

src/video_core/shader/decode/texture.cpp

@@ -139,7 +139,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         }
         const Node component = Immediate(static_cast<u32>(instr.tld4s.component));
 
-        const SamplerInfo info{TextureType::Texture2D, false, is_depth_compare};
+        const SamplerInfo info{TextureType::Texture2D, false, is_depth_compare, false};
         const Sampler& sampler = *GetSampler(instr.sampler, info);
 
         Node4 values;
@@ -171,13 +171,12 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         const auto coord_count = GetCoordCount(texture_type);
         Node index_var{};
         const Sampler* sampler =
-            is_bindless ? GetBindlessSampler(base_reg, index_var, {{texture_type, is_array, false}})
-                        : GetSampler(instr.sampler, {{texture_type, is_array, false}});
+            is_bindless
+                ? GetBindlessSampler(base_reg, index_var, {{texture_type, is_array, false, false}})
+                : GetSampler(instr.sampler, {{texture_type, is_array, false, false}});
         Node4 values;
         if (sampler == nullptr) {
-            for (u32 element = 0; element < values.size(); ++element) {
-                values[element] = Immediate(0);
-            }
+            std::generate(values.begin(), values.end(), [] { return Immediate(0); });
             WriteTexInstructionFloat(bb, instr, values);
             break;
         }
@@ -269,7 +268,6 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
                              "NDV is not implemented");
 
         auto texture_type = instr.tmml.texture_type.Value();
-        const bool is_array = instr.tmml.array != 0;
         Node index_var{};
         const Sampler* sampler =
             is_bindless ? GetBindlessSampler(instr.gpr20, index_var) : GetSampler(instr.sampler);
@@ -593,8 +591,9 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
         ++parameter_register;
     }
 
-    const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
-        texture_type, depth_compare, is_array, lod_bias_enabled, 4, 5);
+    const auto coord_counts = ValidateAndGetCoordinateElement(texture_type, depth_compare, is_array,
+                                                              lod_bias_enabled, 4, 5);
+    const auto coord_count = std::get<0>(coord_counts);
     // If enabled arrays index is always stored in the gpr8 field
     const u64 array_register = instr.gpr8.Value();
     // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
@@ -632,8 +631,10 @@ Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
     const bool lod_bias_enabled =
         (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
 
-    const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
-        texture_type, depth_compare, is_array, lod_bias_enabled, 4, 4);
+    const auto coord_counts = ValidateAndGetCoordinateElement(texture_type, depth_compare, is_array,
+                                                              lod_bias_enabled, 4, 4);
+    const auto coord_count = std::get<0>(coord_counts);
+
     // If enabled arrays index is always stored in the gpr8 field
     const u64 array_register = instr.gpr8.Value();
     // First coordinate index is stored in gpr8 field or (gpr8 + 1) when arrays are used

src/yuzu/configuration/config.cpp

@@ -407,6 +407,8 @@ void Config::ReadAudioValues() {
                                    .toStdString();
     Settings::values.enable_audio_stretching =
         ReadSetting(QStringLiteral("enable_audio_stretching"), true).toBool();
+    Settings::values.enable_realtime_audio =
+        ReadSetting(QStringLiteral("enable_realtime_audio"), false).toBool();
     Settings::values.audio_device_id =
         ReadSetting(QStringLiteral("output_device"), QStringLiteral("auto"))
             .toString()
@@ -915,6 +917,8 @@ void Config::SaveAudioValues() {
                  QStringLiteral("auto"));
     WriteSetting(QStringLiteral("enable_audio_stretching"),
                  Settings::values.enable_audio_stretching, true);
+    WriteSetting(QStringLiteral("enable_realtime_audio"), Settings::values.enable_realtime_audio,
+                 false);
     WriteSetting(QStringLiteral("output_device"),
                  QString::fromStdString(Settings::values.audio_device_id), QStringLiteral("auto"));
     WriteSetting(QStringLiteral("volume"), Settings::values.volume, 1.0f);

src/yuzu/configuration/configure_audio.cpp

@@ -42,6 +42,7 @@ void ConfigureAudio::SetConfiguration() {
     SetAudioDeviceFromDeviceID();
 
     ui->toggle_audio_stretching->setChecked(Settings::values.enable_audio_stretching);
+    ui->toggle_realtime_audio->setChecked(Settings::values.enable_realtime_audio);
     ui->volume_slider->setValue(Settings::values.volume * ui->volume_slider->maximum());
     SetVolumeIndicatorText(ui->volume_slider->sliderPosition());
 }
@@ -84,6 +85,7 @@ void ConfigureAudio::ApplyConfiguration() {
         ui->output_sink_combo_box->itemText(ui->output_sink_combo_box->currentIndex())
             .toStdString();
     Settings::values.enable_audio_stretching = ui->toggle_audio_stretching->isChecked();
+    Settings::values.enable_realtime_audio = ui->toggle_realtime_audio->isChecked();
     Settings::values.audio_device_id =
         ui->audio_device_combo_box->itemText(ui->audio_device_combo_box->currentIndex())
             .toStdString();

src/yuzu/configuration/configure_audio.ui

@@ -31,16 +31,26 @@
         </item>
        </layout>
       </item>
-       <item>
-         <widget class="QCheckBox" name="toggle_audio_stretching">
-           <property name="toolTip">
-             <string>This post-processing effect adjusts audio speed to match emulation speed and helps prevent audio stutter. This however increases audio latency.</string>
-           </property>
-           <property name="text">
-             <string>Enable audio stretching</string>
-           </property>
-         </widget>
-       </item>
+      <item>
+        <widget class="QCheckBox" name="toggle_audio_stretching">
+          <property name="toolTip">
+            <string>This post-processing effect adjusts audio speed to match emulation speed and helps prevent audio stutter. This however increases audio latency.</string>
+          </property>
+          <property name="text">
+            <string>Enable audio stretching</string>
+          </property>
+        </widget>
+      </item>
+      <item>
+        <widget class="QCheckBox" name="toggle_realtime_audio">
+          <property name="toolTip">
+            <string>This adjusts audio timing to match real-time speed and helps prevent audio stutter.</string>
+          </property>
+          <property name="text">
+            <string>Enable real-time audio</string>
+          </property>
+        </widget>
+      </item>
       <item>
        <layout class="QHBoxLayout">
         <item>

src/yuzu_cmd/config.cpp

@@ -404,6 +404,8 @@ void Config::ReadValues() {
     Settings::values.sink_id = sdl2_config->Get("Audio", "output_engine", "auto");
     Settings::values.enable_audio_stretching =
         sdl2_config->GetBoolean("Audio", "enable_audio_stretching", true);
+    Settings::values.enable_realtime_audio =
+        sdl2_config->GetBoolean("Audio", "enable_realtime_audio", false);
     Settings::values.audio_device_id = sdl2_config->Get("Audio", "output_device", "auto");
     Settings::values.volume = static_cast<float>(sdl2_config->GetReal("Audio", "volume", 1));
 

src/yuzu_cmd/default_ini.h

@@ -201,6 +201,11 @@ output_engine =
 # 0: No, 1 (default): Yes
 enable_audio_stretching =
 
+# Whether or not to enable the real-time audio processing.
+# This effect adjusts audio speed to match real-time speed and helps prevent audio stutter.
+# 0: No, 1 (default): Yes
+enable_realtime_audio =
+
 # Which audio device to use.
 # auto (default): Auto-select
 output_device =