add UUID validation check

Co-authored-by: VolcaEM <63682805+VolcaEM@users.noreply.github.com>

.gitmodules

@@ -28,6 +28,3 @@
 [submodule "libzip"]
     path = externals/libzip/libzip
     url = https://github.com/nih-at/libzip.git
-[submodule "xbyak"]
-    path = externals/xbyak
-    url = https://github.com/herumi/xbyak.git

externals/CMakeLists.txt

@@ -75,11 +75,3 @@ if (ENABLE_WEB_SERVICE)
     target_compile_definitions(httplib INTERFACE -DCPPHTTPLIB_OPENSSL_SUPPORT)
     target_link_libraries(httplib INTERFACE OpenSSL::SSL OpenSSL::Crypto)
 endif()
-
-if (NOT TARGET xbyak)
-    if (ARCHITECTURE_x86 OR ARCHITECTURE_x86_64)
-        add_library(xbyak INTERFACE)
-        target_include_directories(xbyak SYSTEM INTERFACE ./xbyak/xbyak)
-        target_compile_definitions(xbyak INTERFACE XBYAK_NO_OP_NAMES)
-    endif()
-endif()

src/common/CMakeLists.txt

@@ -123,8 +123,6 @@ add_library(common STATIC
     lz4_compression.cpp
     lz4_compression.h
     math_util.h
-    memory_detect.cpp
-    memory_detect.h
     memory_hook.cpp
     memory_hook.h
     microprofile.cpp
@@ -171,12 +169,10 @@ if(ARCHITECTURE_x86_64)
         PRIVATE
             x64/cpu_detect.cpp
             x64/cpu_detect.h
-            x64/xbyak_abi.h
-            x64/xbyak_util.h
     )
 endif()
 
 create_target_directory_groups(common)
 
 target_link_libraries(common PUBLIC Boost::boost fmt::fmt microprofile)
-target_link_libraries(common PRIVATE lz4::lz4 zstd::zstd xbyak)
+target_link_libraries(common PRIVATE lz4::lz4 zstd::zstd)

src/common/memory_detect.cpp

@@ -1,60 +0,0 @@
-// Copyright 2020 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#ifdef _WIN32
-// clang-format off
-#include <windows.h>
-#include <sysinfoapi.h>
-// clang-format on
-#else
-#include <sys/types.h>
-#ifdef __APPLE__
-#include <sys/sysctl.h>
-#else
-#include <sys/sysinfo.h>
-#endif
-#endif
-
-#include "common/memory_detect.h"
-
-namespace Common {
-
-// Detects the RAM and Swapfile sizes
-static MemoryInfo Detect() {
-    MemoryInfo mem_info{};
-
-#ifdef _WIN32
-    MEMORYSTATUSEX memorystatus;
-    memorystatus.dwLength = sizeof(memorystatus);
-    GlobalMemoryStatusEx(&memorystatus);
-    mem_info.TotalPhysicalMemory = memorystatus.ullTotalPhys;
-    mem_info.TotalSwapMemory = memorystatus.ullTotalPageFile - mem_info.TotalPhysicalMemory;
-#elif defined(__APPLE__)
-    u64 ramsize;
-    struct xsw_usage vmusage;
-    std::size_t sizeof_ramsize = sizeof(ramsize);
-    std::size_t sizeof_vmusage = sizeof(vmusage);
-    // hw and vm are defined in sysctl.h
-    // https://github.com/apple/darwin-xnu/blob/master/bsd/sys/sysctl.h#L471
-    // sysctlbyname(const char *, void *, size_t *, void *, size_t);
-    sysctlbyname("hw.memsize", &ramsize, &sizeof_ramsize, NULL, 0);
-    sysctlbyname("vm.swapusage", &vmusage, &sizeof_vmusage, NULL, 0);
-    mem_info.TotalPhysicalMemory = ramsize;
-    mem_info.TotalSwapMemory = vmusage.xsu_total;
-#else
-    struct sysinfo meminfo;
-    sysinfo(&meminfo);
-    mem_info.TotalPhysicalMemory = meminfo.totalram;
-    mem_info.TotalSwapMemory = meminfo.totalswap;
-#endif
-
-    return mem_info;
-}
-
-const MemoryInfo& GetMemInfo() {
-    static MemoryInfo mem_info = Detect();
-    return mem_info;
-}
-
-} // namespace Common

src/common/memory_detect.h

@@ -1,22 +0,0 @@
-// Copyright 2020 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include "common/common_types.h"
-
-namespace Common {
-
-struct MemoryInfo {
-    u64 TotalPhysicalMemory{};
-    u64 TotalSwapMemory{};
-};
-
-/**
- * Gets the memory info of the host system
- * @return Reference to a MemoryInfo struct with the physical and swap memory sizes in bytes
- */
-const MemoryInfo& GetMemInfo();
-
-} // namespace Common

src/common/x64/xbyak_abi.h

@@ -1,266 +0,0 @@
-// Copyright 2016 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <bitset>
-#include <initializer_list>
-#include <xbyak.h>
-#include "common/assert.h"
-
-namespace Common::X64 {
-
-inline int RegToIndex(const Xbyak::Reg& reg) {
-    using Kind = Xbyak::Reg::Kind;
-    ASSERT_MSG((reg.getKind() & (Kind::REG | Kind::XMM)) != 0,
-               "RegSet only support GPRs and XMM registers.");
-    ASSERT_MSG(reg.getIdx() < 16, "RegSet only supports XXM0-15.");
-    return reg.getIdx() + (reg.getKind() == Kind::REG ? 0 : 16);
-}
-
-inline Xbyak::Reg64 IndexToReg64(int reg_index) {
-    ASSERT(reg_index < 16);
-    return Xbyak::Reg64(reg_index);
-}
-
-inline Xbyak::Xmm IndexToXmm(int reg_index) {
-    ASSERT(reg_index >= 16 && reg_index < 32);
-    return Xbyak::Xmm(reg_index - 16);
-}
-
-inline Xbyak::Reg IndexToReg(int reg_index) {
-    if (reg_index < 16) {
-        return IndexToReg64(reg_index);
-    } else {
-        return IndexToXmm(reg_index);
-    }
-}
-
-inline std::bitset<32> BuildRegSet(std::initializer_list<Xbyak::Reg> regs) {
-    std::bitset<32> bits;
-    for (const Xbyak::Reg& reg : regs) {
-        bits[RegToIndex(reg)] = true;
-    }
-    return bits;
-}
-
-const std::bitset<32> ABI_ALL_GPRS(0x0000FFFF);
-const std::bitset<32> ABI_ALL_XMMS(0xFFFF0000);
-
-#ifdef _WIN32
-
-// Microsoft x64 ABI
-const Xbyak::Reg ABI_RETURN = Xbyak::util::rax;
-const Xbyak::Reg ABI_PARAM1 = Xbyak::util::rcx;
-const Xbyak::Reg ABI_PARAM2 = Xbyak::util::rdx;
-const Xbyak::Reg ABI_PARAM3 = Xbyak::util::r8;
-const Xbyak::Reg ABI_PARAM4 = Xbyak::util::r9;
-
-const std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
-    // GPRs
-    Xbyak::util::rcx,
-    Xbyak::util::rdx,
-    Xbyak::util::r8,
-    Xbyak::util::r9,
-    Xbyak::util::r10,
-    Xbyak::util::r11,
-    // XMMs
-    Xbyak::util::xmm0,
-    Xbyak::util::xmm1,
-    Xbyak::util::xmm2,
-    Xbyak::util::xmm3,
-    Xbyak::util::xmm4,
-    Xbyak::util::xmm5,
-});
-
-const std::bitset<32> ABI_ALL_CALLEE_SAVED = BuildRegSet({
-    // GPRs
-    Xbyak::util::rbx,
-    Xbyak::util::rsi,
-    Xbyak::util::rdi,
-    Xbyak::util::rbp,
-    Xbyak::util::r12,
-    Xbyak::util::r13,
-    Xbyak::util::r14,
-    Xbyak::util::r15,
-    // XMMs
-    Xbyak::util::xmm6,
-    Xbyak::util::xmm7,
-    Xbyak::util::xmm8,
-    Xbyak::util::xmm9,
-    Xbyak::util::xmm10,
-    Xbyak::util::xmm11,
-    Xbyak::util::xmm12,
-    Xbyak::util::xmm13,
-    Xbyak::util::xmm14,
-    Xbyak::util::xmm15,
-});
-
-constexpr size_t ABI_SHADOW_SPACE = 0x20;
-
-#else
-
-// System V x86-64 ABI
-const Xbyak::Reg ABI_RETURN = Xbyak::util::rax;
-const Xbyak::Reg ABI_PARAM1 = Xbyak::util::rdi;
-const Xbyak::Reg ABI_PARAM2 = Xbyak::util::rsi;
-const Xbyak::Reg ABI_PARAM3 = Xbyak::util::rdx;
-const Xbyak::Reg ABI_PARAM4 = Xbyak::util::rcx;
-
-const std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
-    // GPRs
-    Xbyak::util::rcx,
-    Xbyak::util::rdx,
-    Xbyak::util::rdi,
-    Xbyak::util::rsi,
-    Xbyak::util::r8,
-    Xbyak::util::r9,
-    Xbyak::util::r10,
-    Xbyak::util::r11,
-    // XMMs
-    Xbyak::util::xmm0,
-    Xbyak::util::xmm1,
-    Xbyak::util::xmm2,
-    Xbyak::util::xmm3,
-    Xbyak::util::xmm4,
-    Xbyak::util::xmm5,
-    Xbyak::util::xmm6,
-    Xbyak::util::xmm7,
-    Xbyak::util::xmm8,
-    Xbyak::util::xmm9,
-    Xbyak::util::xmm10,
-    Xbyak::util::xmm11,
-    Xbyak::util::xmm12,
-    Xbyak::util::xmm13,
-    Xbyak::util::xmm14,
-    Xbyak::util::xmm15,
-});
-
-const std::bitset<32> ABI_ALL_CALLEE_SAVED = BuildRegSet({
-    // GPRs
-    Xbyak::util::rbx,
-    Xbyak::util::rbp,
-    Xbyak::util::r12,
-    Xbyak::util::r13,
-    Xbyak::util::r14,
-    Xbyak::util::r15,
-});
-
-constexpr size_t ABI_SHADOW_SPACE = 0;
-
-#endif
-
-inline void ABI_CalculateFrameSize(std::bitset<32> regs, size_t rsp_alignment,
-                                   size_t needed_frame_size, s32* out_subtraction,
-                                   s32* out_xmm_offset) {
-    const auto count = (regs & ABI_ALL_GPRS).count();
-    rsp_alignment -= count * 8;
-    size_t subtraction = 0;
-    const auto xmm_count = (regs & ABI_ALL_XMMS).count();
-    if (xmm_count) {
-        // If we have any XMMs to save, we must align the stack here.
-        subtraction = rsp_alignment & 0xF;
-    }
-    subtraction += 0x10 * xmm_count;
-    size_t xmm_base_subtraction = subtraction;
-    subtraction += needed_frame_size;
-    subtraction += ABI_SHADOW_SPACE;
-    // Final alignment.
-    rsp_alignment -= subtraction;
-    subtraction += rsp_alignment & 0xF;
-
-    *out_subtraction = (s32)subtraction;
-    *out_xmm_offset = (s32)(subtraction - xmm_base_subtraction);
-}
-
-inline size_t ABI_PushRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bitset<32> regs,
-                                              size_t rsp_alignment, size_t needed_frame_size = 0) {
-    s32 subtraction, xmm_offset;
-    ABI_CalculateFrameSize(regs, rsp_alignment, needed_frame_size, &subtraction, &xmm_offset);
-    for (std::size_t i = 0; i < regs.size(); ++i) {
-        if (regs[i] && ABI_ALL_GPRS[i]) {
-            code.push(IndexToReg64(static_cast<int>(i)));
-        }
-    }
-    if (subtraction != 0) {
-        code.sub(code.rsp, subtraction);
-    }
-
-    for (int i = 0; i < regs.count(); i++) {
-        if (regs.test(i) & ABI_ALL_GPRS.test(i)) {
-            code.push(IndexToReg64(i));
-        }
-    }
-
-    for (std::size_t i = 0; i < regs.size(); ++i) {
-        if (regs[i] && ABI_ALL_XMMS[i]) {
-            code.movaps(code.xword[code.rsp + xmm_offset], IndexToXmm(static_cast<int>(i)));
-            xmm_offset += 0x10;
-        }
-    }
-
-    return ABI_SHADOW_SPACE;
-}
-
-inline void ABI_PopRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bitset<32> regs,
-                                           size_t rsp_alignment, size_t needed_frame_size = 0) {
-    s32 subtraction, xmm_offset;
-    ABI_CalculateFrameSize(regs, rsp_alignment, needed_frame_size, &subtraction, &xmm_offset);
-
-    for (std::size_t i = 0; i < regs.size(); ++i) {
-        if (regs[i] && ABI_ALL_XMMS[i]) {
-            code.movaps(IndexToXmm(static_cast<int>(i)), code.xword[code.rsp + xmm_offset]);
-            xmm_offset += 0x10;
-        }
-    }
-
-    if (subtraction != 0) {
-        code.add(code.rsp, subtraction);
-    }
-
-    // GPRs need to be popped in reverse order
-    for (int i = 15; i >= 0; i--) {
-        if (regs[i]) {
-            code.pop(IndexToReg64(i));
-        }
-    }
-}
-
-inline size_t ABI_PushRegistersAndAdjustStackGPS(Xbyak::CodeGenerator& code, std::bitset<32> regs,
-                                                 size_t rsp_alignment,
-                                                 size_t needed_frame_size = 0) {
-    s32 subtraction, xmm_offset;
-    ABI_CalculateFrameSize(regs, rsp_alignment, needed_frame_size, &subtraction, &xmm_offset);
-
-    for (std::size_t i = 0; i < regs.size(); ++i) {
-        if (regs[i] && ABI_ALL_GPRS[i]) {
-            code.push(IndexToReg64(static_cast<int>(i)));
-        }
-    }
-
-    if (subtraction != 0) {
-        code.sub(code.rsp, subtraction);
-    }
-
-    return ABI_SHADOW_SPACE;
-}
-
-inline void ABI_PopRegistersAndAdjustStackGPS(Xbyak::CodeGenerator& code, std::bitset<32> regs,
-                                              size_t rsp_alignment, size_t needed_frame_size = 0) {
-    s32 subtraction, xmm_offset;
-    ABI_CalculateFrameSize(regs, rsp_alignment, needed_frame_size, &subtraction, &xmm_offset);
-
-    if (subtraction != 0) {
-        code.add(code.rsp, subtraction);
-    }
-
-    // GPRs need to be popped in reverse order
-    for (int i = 15; i >= 0; i--) {
-        if (regs[i]) {
-            code.pop(IndexToReg64(i));
-        }
-    }
-}
-
-} // namespace Common::X64

src/common/x64/xbyak_util.h

@@ -1,47 +0,0 @@
-// Copyright 2016 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <type_traits>
-#include <xbyak.h>
-#include "common/x64/xbyak_abi.h"
-
-namespace Common::X64 {
-
-// Constants for use with cmpps/cmpss
-enum {
-    CMP_EQ = 0,
-    CMP_LT = 1,
-    CMP_LE = 2,
-    CMP_UNORD = 3,
-    CMP_NEQ = 4,
-    CMP_NLT = 5,
-    CMP_NLE = 6,
-    CMP_ORD = 7,
-};
-
-constexpr bool IsWithin2G(uintptr_t ref, uintptr_t target) {
-    const u64 distance = target - (ref + 5);
-    return !(distance >= 0x8000'0000ULL && distance <= ~0x8000'0000ULL);
-}
-
-inline bool IsWithin2G(const Xbyak::CodeGenerator& code, uintptr_t target) {
-    return IsWithin2G(reinterpret_cast<uintptr_t>(code.getCurr()), target);
-}
-
-template <typename T>
-inline void CallFarFunction(Xbyak::CodeGenerator& code, const T f) {
-    static_assert(std::is_pointer_v<T>, "Argument must be a (function) pointer.");
-    size_t addr = reinterpret_cast<size_t>(f);
-    if (IsWithin2G(code, addr)) {
-        code.call(f);
-    } else {
-        // ABI_RETURN is a safe temp register to use before a call
-        code.mov(ABI_RETURN, addr);
-        code.call(ABI_RETURN);
-    }
-}
-
-} // namespace Common::X64

src/core/file_sys/patch_manager.cpp

@@ -10,7 +10,6 @@
 #include "common/file_util.h"
 #include "common/hex_util.h"
 #include "common/logging/log.h"
-#include "common/string_util.h"
 #include "core/core.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/control_metadata.h"
@@ -49,23 +48,6 @@ std::string FormatTitleVersion(u32 version, TitleVersionFormat format) {
     return fmt::format("v{}.{}.{}", bytes[3], bytes[2], bytes[1]);
 }
 
-std::shared_ptr<VfsDirectory> FindSubdirectoryCaseless(const std::shared_ptr<VfsDirectory> dir,
-                                                       std::string_view name) {
-#ifdef _WIN32
-    return dir->GetSubdirectory(name);
-#else
-    const auto subdirs = dir->GetSubdirectories();
-    for (const auto& subdir : subdirs) {
-        std::string dir_name = Common::ToLower(subdir->GetName());
-        if (dir_name == name) {
-            return subdir;
-        }
-    }
-
-    return nullptr;
-#endif
-}
-
 PatchManager::PatchManager(u64 title_id) : title_id(title_id) {}
 
 PatchManager::~PatchManager() = default;
@@ -122,7 +104,7 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
             if (std::find(disabled.begin(), disabled.end(), subdir->GetName()) != disabled.end())
                 continue;
 
-            auto exefs_dir = FindSubdirectoryCaseless(subdir, "exefs");
+            auto exefs_dir = subdir->GetSubdirectory("exefs");
             if (exefs_dir != nullptr)
                 layers.push_back(std::move(exefs_dir));
         }
@@ -148,7 +130,7 @@ std::vector<VirtualFile> PatchManager::CollectPatches(const std::vector<VirtualD
         if (std::find(disabled.cbegin(), disabled.cend(), subdir->GetName()) != disabled.cend())
             continue;
 
-        auto exefs_dir = FindSubdirectoryCaseless(subdir, "exefs");
+        auto exefs_dir = subdir->GetSubdirectory("exefs");
         if (exefs_dir != nullptr) {
             for (const auto& file : exefs_dir->GetFiles()) {
                 if (file->GetExtension() == "ips") {
@@ -313,7 +295,7 @@ std::vector<Core::Memory::CheatEntry> PatchManager::CreateCheatList(
             continue;
         }
 
-        auto cheats_dir = FindSubdirectoryCaseless(subdir, "cheats");
+        auto cheats_dir = subdir->GetSubdirectory("cheats");
         if (cheats_dir != nullptr) {
             auto res = ReadCheatFileFromFolder(system, title_id, build_id_, cheats_dir, true);
             if (res.has_value()) {
@@ -358,11 +340,11 @@ static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType t
             continue;
         }
 
-        auto romfs_dir = FindSubdirectoryCaseless(subdir, "romfs");
+        auto romfs_dir = subdir->GetSubdirectory("romfs");
         if (romfs_dir != nullptr)
             layers.push_back(std::move(romfs_dir));
 
-        auto ext_dir = FindSubdirectoryCaseless(subdir, "romfs_ext");
+        auto ext_dir = subdir->GetSubdirectory("romfs_ext");
         if (ext_dir != nullptr)
             layers_ext.push_back(std::move(ext_dir));
     }
@@ -488,7 +470,7 @@ std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNam
         for (const auto& mod : mod_dir->GetSubdirectories()) {
             std::string types;
 
-            const auto exefs_dir = FindSubdirectoryCaseless(mod, "exefs");
+            const auto exefs_dir = mod->GetSubdirectory("exefs");
             if (IsDirValidAndNonEmpty(exefs_dir)) {
                 bool ips = false;
                 bool ipswitch = false;
@@ -512,9 +494,9 @@ std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNam
                 if (layeredfs)
                     AppendCommaIfNotEmpty(types, "LayeredExeFS");
             }
-            if (IsDirValidAndNonEmpty(FindSubdirectoryCaseless(mod, "romfs")))
+            if (IsDirValidAndNonEmpty(mod->GetSubdirectory("romfs")))
                 AppendCommaIfNotEmpty(types, "LayeredFS");
-            if (IsDirValidAndNonEmpty(FindSubdirectoryCaseless(mod, "cheats")))
+            if (IsDirValidAndNonEmpty(mod->GetSubdirectory("cheats")))
                 AppendCommaIfNotEmpty(types, "Cheats");
 
             if (types.empty())

src/core/file_sys/patch_manager.h

@@ -29,11 +29,6 @@ enum class TitleVersionFormat : u8 {
 std::string FormatTitleVersion(u32 version,
                                TitleVersionFormat format = TitleVersionFormat::ThreeElements);
 
-// Returns a directory with name matching name case-insensitive. Returns nullptr if directory
-// doesn't have a directory with name.
-std::shared_ptr<VfsDirectory> FindSubdirectoryCaseless(const std::shared_ptr<VfsDirectory> dir,
-                                                       std::string_view name);
-
 // A centralized class to manage patches to games.
 class PatchManager {
 public:

src/core/file_sys/system_archive/system_version.cpp

@@ -12,17 +12,17 @@ namespace SystemVersionData {
 // This section should reflect the best system version to describe yuzu's HLE api.
 // TODO(DarkLordZach): Update when HLE gets better.
 
-constexpr u8 VERSION_MAJOR = 10;
-constexpr u8 VERSION_MINOR = 0;
-constexpr u8 VERSION_MICRO = 2;
+constexpr u8 VERSION_MAJOR = 5;
+constexpr u8 VERSION_MINOR = 1;
+constexpr u8 VERSION_MICRO = 0;
 
-constexpr u8 REVISION_MAJOR = 1;
+constexpr u8 REVISION_MAJOR = 3;
 constexpr u8 REVISION_MINOR = 0;
 
 constexpr char PLATFORM_STRING[] = "NX";
-constexpr char VERSION_HASH[] = "f90143fa8bbc061d4f68c35f95f04f8080c0ecdc";
-constexpr char DISPLAY_VERSION[] = "10.0.2";
-constexpr char DISPLAY_TITLE[] = "NintendoSDK Firmware for NX 10.0.2-1.0";
+constexpr char VERSION_HASH[] = "23f9df53e25709d756e0c76effcb2473bd3447dd";
+constexpr char DISPLAY_VERSION[] = "5.1.0";
+constexpr char DISPLAY_TITLE[] = "NintendoSDK Firmware for NX 5.1.0-3.0";
 
 } // namespace SystemVersionData
 

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

@@ -38,11 +38,10 @@ void Controller_Keyboard::OnUpdate(const Core::Timing::CoreTiming& core_timing,
     cur_entry.sampling_number = last_entry.sampling_number + 1;
     cur_entry.sampling_number2 = cur_entry.sampling_number;
 
-    cur_entry.key.fill(0);
-    cur_entry.modifier = 0;
-
     for (std::size_t i = 0; i < keyboard_keys.size(); ++i) {
-        cur_entry.key[i / KEYS_PER_BYTE] |= (keyboard_keys[i]->GetStatus() << (i % KEYS_PER_BYTE));
+        for (std::size_t k = 0; k < KEYS_PER_BYTE; ++k) {
+            cur_entry.key[i / KEYS_PER_BYTE] |= (keyboard_keys[i]->GetStatus() << k);
+        }
     }
 
     for (std::size_t i = 0; i < keyboard_mods.size(); ++i) {

src/core/hle/service/ldr/ldr.cpp

@@ -39,61 +39,42 @@ constexpr ResultCode ERROR_NOT_INITIALIZED{ErrorModule::Loader, 87};
 constexpr std::size_t MAXIMUM_LOADED_RO{0x40};
 constexpr std::size_t MAXIMUM_MAP_RETRIES{0x200};
 
-constexpr std::size_t TEXT_INDEX{0};
-constexpr std::size_t RO_INDEX{1};
-constexpr std::size_t DATA_INDEX{2};
-
-struct NRRCertification {
-    u64_le application_id_mask;
-    u64_le application_id_pattern;
-    std::array<u8, 0x10> reserved;
-    std::array<u8, 0x100> public_key; // Also known as modulus
-    std::array<u8, 0x100> signature;
-};
-static_assert(sizeof(NRRCertification) == 0x220, "NRRCertification has invalid size.");
-
 struct NRRHeader {
     u32_le magic;
-    u32_le certification_signature_key_generation; // 9.0.0+
-    u64_le reserved;
-    NRRCertification certification;
-    std::array<u8, 0x100> signature;
-    u64_le application_id;
+    INSERT_PADDING_BYTES(12);
+    u64_le title_id_mask;
+    u64_le title_id_pattern;
+    INSERT_PADDING_BYTES(16);
+    std::array<u8, 0x100> modulus;
+    std::array<u8, 0x100> signature_1;
+    std::array<u8, 0x100> signature_2;
+    u64_le title_id;
     u32_le size;
-    u8 nrr_kind; // 7.0.0+
-    std::array<u8, 3> reserved_2;
+    INSERT_PADDING_BYTES(4);
     u32_le hash_offset;
     u32_le hash_count;
-    u64_le reserved_3;
+    INSERT_PADDING_BYTES(8);
 };
-static_assert(sizeof(NRRHeader) == 0x350, "NRRHeader has invalid size.");
-
-struct SegmentHeader {
-    u32_le memory_offset;
-    u32_le memory_size;
-};
-static_assert(sizeof(SegmentHeader) == 0x8, "SegmentHeader has invalid size.");
+static_assert(sizeof(NRRHeader) == 0x350, "NRRHeader has incorrect size.");
 
 struct NROHeader {
-    // Switchbrew calls this "Start" (0x10)
-    u32_le unused;
+    INSERT_PADDING_WORDS(1);
     u32_le mod_offset;
-    u64_le padding;
-
-    // Switchbrew calls this "Header" (0x70)
+    INSERT_PADDING_WORDS(2);
     u32_le magic;
     u32_le version;
     u32_le nro_size;
     u32_le flags;
-    // .text, .ro, .data
-    std::array<SegmentHeader, 3> segment_headers;
+    u32_le text_offset;
+    u32_le text_size;
+    u32_le ro_offset;
+    u32_le ro_size;
+    u32_le rw_offset;
+    u32_le rw_size;
     u32_le bss_size;
-    u32_le reserved;
+    INSERT_PADDING_WORDS(1);
     std::array<u8, 0x20> build_id;
-    u32_le dso_handle_offset;
-    u32_le unused_2;
-    // .apiInfo, .dynstr, .dynsym
-    std::array<SegmentHeader, 3> segment_headers_2;
+    INSERT_PADDING_BYTES(0x20);
 };
 static_assert(sizeof(NROHeader) == 0x80, "NROHeader has invalid size.");
 
@@ -110,7 +91,6 @@ struct NROInfo {
     std::size_t data_size{};
     VAddr src_addr{};
 };
-static_assert(sizeof(NROInfo) == 0x60, "NROInfo has invalid size.");
 
 class DebugMonitor final : public ServiceFramework<DebugMonitor> {
 public:
@@ -246,11 +226,11 @@ public:
             return;
         }
 
-        if (system.CurrentProcess()->GetTitleID() != header.application_id) {
+        if (system.CurrentProcess()->GetTitleID() != header.title_id) {
             LOG_ERROR(Service_LDR,
                       "Attempting to load NRR with title ID other than current process. (actual "
                       "{:016X})!",
-                      header.application_id);
+                      header.title_id);
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(ERROR_INVALID_NRR);
             return;
@@ -368,10 +348,10 @@ public:
 
     ResultCode LoadNro(Kernel::Process* process, const NROHeader& nro_header, VAddr nro_addr,
                        VAddr start) const {
-        const VAddr text_start{start + nro_header.segment_headers[TEXT_INDEX].memory_offset};
-        const VAddr ro_start{start + nro_header.segment_headers[RO_INDEX].memory_offset};
-        const VAddr data_start{start + nro_header.segment_headers[DATA_INDEX].memory_offset};
-        const VAddr bss_start{data_start + nro_header.segment_headers[DATA_INDEX].memory_size};
+        const VAddr text_start{start + nro_header.text_offset};
+        const VAddr ro_start{start + nro_header.ro_offset};
+        const VAddr data_start{start + nro_header.rw_offset};
+        const VAddr bss_start{data_start + nro_header.rw_size};
         const VAddr bss_end_addr{
             Common::AlignUp(bss_start + nro_header.bss_size, Kernel::Memory::PageSize)};
 
@@ -380,12 +360,9 @@ public:
             system.Memory().ReadBlock(src_addr, source_data.data(), source_data.size());
             system.Memory().WriteBlock(dst_addr, source_data.data(), source_data.size());
         }};
-        CopyCode(nro_addr + nro_header.segment_headers[TEXT_INDEX].memory_offset, text_start,
-                 nro_header.segment_headers[TEXT_INDEX].memory_size);
-        CopyCode(nro_addr + nro_header.segment_headers[RO_INDEX].memory_offset, ro_start,
-                 nro_header.segment_headers[RO_INDEX].memory_size);
-        CopyCode(nro_addr + nro_header.segment_headers[DATA_INDEX].memory_offset, data_start,
-                 nro_header.segment_headers[DATA_INDEX].memory_size);
+        CopyCode(nro_addr + nro_header.text_offset, text_start, nro_header.text_size);
+        CopyCode(nro_addr + nro_header.ro_offset, ro_start, nro_header.ro_size);
+        CopyCode(nro_addr + nro_header.rw_offset, data_start, nro_header.rw_size);
 
         CASCADE_CODE(process->PageTable().SetCodeMemoryPermission(
             text_start, ro_start - text_start, Kernel::Memory::MemoryPermission::ReadAndExecute));
@@ -507,11 +484,9 @@ public:
         }
 
         // Track the loaded NRO
-        nro.insert_or_assign(*map_result,
-                             NROInfo{hash, *map_result, nro_size, bss_address, bss_size,
-                                     header.segment_headers[TEXT_INDEX].memory_size,
-                                     header.segment_headers[RO_INDEX].memory_size,
-                                     header.segment_headers[DATA_INDEX].memory_size, nro_address});
+        nro.insert_or_assign(*map_result, NROInfo{hash, *map_result, nro_size, bss_address,
+                                                  bss_size, header.text_size, header.ro_size,
+                                                  header.rw_size, nro_address});
 
         // Invalidate JIT caches for the newly mapped process code
         system.InvalidateCpuInstructionCaches();
@@ -609,21 +584,11 @@ private:
     static bool IsValidNRO(const NROHeader& header, u64 nro_size, u64 bss_size) {
         return header.magic == Common::MakeMagic('N', 'R', 'O', '0') &&
                header.nro_size == nro_size && header.bss_size == bss_size &&
-
-               header.segment_headers[RO_INDEX].memory_offset ==
-                   header.segment_headers[TEXT_INDEX].memory_offset +
-                       header.segment_headers[TEXT_INDEX].memory_size &&
-
-               header.segment_headers[DATA_INDEX].memory_offset ==
-                   header.segment_headers[RO_INDEX].memory_offset +
-                       header.segment_headers[RO_INDEX].memory_size &&
-
-               nro_size == header.segment_headers[DATA_INDEX].memory_offset +
-                               header.segment_headers[DATA_INDEX].memory_size &&
-
-               Common::Is4KBAligned(header.segment_headers[TEXT_INDEX].memory_size) &&
-               Common::Is4KBAligned(header.segment_headers[RO_INDEX].memory_size) &&
-               Common::Is4KBAligned(header.segment_headers[DATA_INDEX].memory_size);
+               header.ro_offset == header.text_offset + header.text_size &&
+               header.rw_offset == header.ro_offset + header.ro_size &&
+               nro_size == header.rw_offset + header.rw_size &&
+               Common::Is4KBAligned(header.text_size) && Common::Is4KBAligned(header.ro_size) &&
+               Common::Is4KBAligned(header.rw_size);
     }
     Core::System& system;
 };

src/core/hle/service/nifm/nifm.cpp

@@ -177,8 +177,7 @@ private:
     void CreateTemporaryNetworkProfile(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_NIFM, "called");
 
-        ASSERT_MSG(ctx.GetReadBufferSize() == 0x17c,
-                   "SfNetworkProfileData is not the correct size");
+        ASSERT_MSG(ctx.GetReadBufferSize() == 0x17c, "NetworkProfileData is not the correct size");
         u128 uuid{};
         auto buffer = ctx.ReadBuffer();
         std::memcpy(&uuid, buffer.data() + 8, sizeof(u128));

src/core/settings.cpp

@@ -112,7 +112,6 @@ void LogSettings() {
     LogSetting("Renderer_UseAsynchronousGpuEmulation",
                Settings::values.use_asynchronous_gpu_emulation);
     LogSetting("Renderer_UseVsync", Settings::values.use_vsync);
-    LogSetting("Renderer_UseAssemblyShaders", Settings::values.use_assembly_shaders);
     LogSetting("Renderer_AnisotropicFilteringLevel", Settings::values.max_anisotropy);
     LogSetting("Audio_OutputEngine", Settings::values.sink_id);
     LogSetting("Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);

src/core/settings.h

@@ -446,7 +446,6 @@ struct Values {
     GPUAccuracy gpu_accuracy;
     bool use_asynchronous_gpu_emulation;
     bool use_vsync;
-    bool use_assembly_shaders;
     bool force_30fps_mode;
     bool use_fast_gpu_time;
 

src/core/telemetry_session.cpp

@@ -201,7 +201,6 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
     AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
              Settings::values.use_asynchronous_gpu_emulation);
     AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync);
-    AddField(field_type, "Renderer_UseAssemblyShaders", Settings::values.use_assembly_shaders);
     AddField(field_type, "System_UseDockedMode", Settings::values.use_docked_mode);
 }
 

src/video_core/CMakeLists.txt

@@ -1,7 +1,6 @@
 add_library(video_core STATIC
     buffer_cache/buffer_block.h
     buffer_cache/buffer_cache.h
-    buffer_cache/map_interval.cpp
     buffer_cache/map_interval.h
     dirty_flags.cpp
     dirty_flags.h
@@ -229,7 +228,7 @@ endif()
 create_target_directory_groups(video_core)
 
 target_link_libraries(video_core PUBLIC common core)
-target_link_libraries(video_core PRIVATE glad xbyak)
+target_link_libraries(video_core PRIVATE glad)
 
 if (ENABLE_VULKAN)
     target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)

src/video_core/buffer_cache/buffer_cache.h

@@ -12,12 +12,11 @@
 #include <utility>
 #include <vector>
 
-#include <boost/container/small_vector.hpp>
+#include <boost/icl/interval_map.hpp>
 #include <boost/icl/interval_set.hpp>
-#include <boost/intrusive/set.hpp>
+#include <boost/range/iterator_range.hpp>
 
 #include "common/alignment.h"
-#include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
@@ -30,12 +29,10 @@
 
 namespace VideoCommon {
 
+using MapInterval = std::shared_ptr<MapIntervalBase>;
+
 template <typename OwnerBuffer, typename BufferType, typename StreamBuffer>
 class BufferCache {
-    using IntervalSet = boost::icl::interval_set<VAddr>;
-    using IntervalType = typename IntervalSet::interval_type;
-    using VectorMapInterval = boost::container::small_vector<MapInterval*, 1>;
-
 public:
     using BufferInfo = std::pair<BufferType, u64>;
 
@@ -43,12 +40,14 @@ public:
                             bool is_written = false, bool use_fast_cbuf = false) {
         std::lock_guard lock{mutex};
 
-        const auto& memory_manager = system.GPU().MemoryManager();
-        const std::optional<VAddr> cpu_addr_opt = memory_manager.GpuToCpuAddress(gpu_addr);
+        const std::optional<VAddr> cpu_addr_opt =
+            system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
+
         if (!cpu_addr_opt) {
             return {GetEmptyBuffer(size), 0};
         }
-        const VAddr cpu_addr = *cpu_addr_opt;
+
+        VAddr cpu_addr = *cpu_addr_opt;
 
         // Cache management is a big overhead, so only cache entries with a given size.
         // TODO: Figure out which size is the best for given games.
@@ -78,19 +77,16 @@ public:
             }
         }
 
-        OwnerBuffer block = GetBlock(cpu_addr, size);
-        MapInterval* const map = MapAddress(block, gpu_addr, cpu_addr, size);
-        if (!map) {
-            return {GetEmptyBuffer(size), 0};
-        }
+        auto block = GetBlock(cpu_addr, size);
+        auto map = MapAddress(block, gpu_addr, cpu_addr, size);
         if (is_written) {
             map->MarkAsModified(true, GetModifiedTicks());
             if (Settings::IsGPULevelHigh() && Settings::values.use_asynchronous_gpu_emulation) {
                 MarkForAsyncFlush(map);
             }
-            if (!map->is_written) {
-                map->is_written = true;
-                MarkRegionAsWritten(map->start, map->end - 1);
+            if (!map->IsWritten()) {
+                map->MarkAsWritten(true);
+                MarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
             }
         }
 
@@ -136,11 +132,12 @@ public:
     void FlushRegion(VAddr addr, std::size_t size) {
         std::lock_guard lock{mutex};
 
-        VectorMapInterval objects = GetMapsInRange(addr, size);
-        std::sort(objects.begin(), objects.end(),
-                  [](MapInterval* lhs, MapInterval* rhs) { return lhs->ticks < rhs->ticks; });
-        for (MapInterval* object : objects) {
-            if (object->is_modified && object->is_registered) {
+        std::vector<MapInterval> objects = GetMapsInRange(addr, size);
+        std::sort(objects.begin(), objects.end(), [](const MapInterval& a, const MapInterval& b) {
+            return a->GetModificationTick() < b->GetModificationTick();
+        });
+        for (auto& object : objects) {
+            if (object->IsModified() && object->IsRegistered()) {
                 mutex.unlock();
                 FlushMap(object);
                 mutex.lock();
@@ -151,9 +148,9 @@ public:
     bool MustFlushRegion(VAddr addr, std::size_t size) {
         std::lock_guard lock{mutex};
 
-        const VectorMapInterval objects = GetMapsInRange(addr, size);
-        return std::any_of(objects.cbegin(), objects.cend(), [](const MapInterval* map) {
-            return map->is_modified && map->is_registered;
+        const std::vector<MapInterval> objects = GetMapsInRange(addr, size);
+        return std::any_of(objects.cbegin(), objects.cend(), [](const MapInterval& map) {
+            return map->IsModified() && map->IsRegistered();
         });
     }
 
@@ -161,8 +158,9 @@ public:
     void InvalidateRegion(VAddr addr, u64 size) {
         std::lock_guard lock{mutex};
 
-        for (auto& object : GetMapsInRange(addr, size)) {
-            if (object->is_registered) {
+        std::vector<MapInterval> objects = GetMapsInRange(addr, size);
+        for (auto& object : objects) {
+            if (object->IsRegistered()) {
                 Unregister(object);
             }
         }
@@ -171,10 +169,10 @@ public:
     void OnCPUWrite(VAddr addr, std::size_t size) {
         std::lock_guard lock{mutex};
 
-        for (MapInterval* object : GetMapsInRange(addr, size)) {
-            if (object->is_memory_marked && object->is_registered) {
+        for (const auto& object : GetMapsInRange(addr, size)) {
+            if (object->IsMemoryMarked() && object->IsRegistered()) {
                 UnmarkMemory(object);
-                object->is_sync_pending = true;
+                object->SetSyncPending(true);
                 marked_for_unregister.emplace_back(object);
             }
         }
@@ -183,9 +181,9 @@ public:
     void SyncGuestHost() {
         std::lock_guard lock{mutex};
 
-        for (auto& object : marked_for_unregister) {
-            if (object->is_registered) {
-                object->is_sync_pending = false;
+        for (const auto& object : marked_for_unregister) {
+            if (object->IsRegistered()) {
+                object->SetSyncPending(false);
                 Unregister(object);
             }
         }
@@ -194,9 +192,9 @@ public:
 
     void CommitAsyncFlushes() {
         if (uncommitted_flushes) {
-            auto commit_list = std::make_shared<std::list<MapInterval*>>();
-            for (MapInterval* map : *uncommitted_flushes) {
-                if (map->is_registered && map->is_modified) {
+            auto commit_list = std::make_shared<std::list<MapInterval>>();
+            for (auto& map : *uncommitted_flushes) {
+                if (map->IsRegistered() && map->IsModified()) {
                     // TODO(Blinkhawk): Implement backend asynchronous flushing
                     // AsyncFlushMap(map)
                     commit_list->push_back(map);
@@ -230,8 +228,8 @@ public:
             committed_flushes.pop_front();
             return;
         }
-        for (MapInterval* map : *flush_list) {
-            if (map->is_registered) {
+        for (MapInterval& map : *flush_list) {
+            if (map->IsRegistered()) {
                 // TODO(Blinkhawk): Replace this for reading the asynchronous flush
                 FlushMap(map);
             }
@@ -267,60 +265,61 @@ protected:
     }
 
     /// Register an object into the cache
-    MapInterval* Register(MapInterval new_map, bool inherit_written = false) {
-        const VAddr cpu_addr = new_map.start;
+    void Register(const MapInterval& new_map, bool inherit_written = false) {
+        const VAddr cpu_addr = new_map->GetStart();
         if (!cpu_addr) {
             LOG_CRITICAL(HW_GPU, "Failed to register buffer with unmapped gpu_address 0x{:016x}",
-                         new_map.gpu_addr);
-            return nullptr;
-        }
-        const std::size_t size = new_map.end - new_map.start;
-        new_map.is_registered = true;
-        rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
-        new_map.is_memory_marked = true;
-        if (inherit_written) {
-            MarkRegionAsWritten(new_map.start, new_map.end - 1);
-            new_map.is_written = true;
-        }
-        MapInterval* const storage = mapped_addresses_allocator.Allocate();
-        *storage = new_map;
-        mapped_addresses.insert(*storage);
-        return storage;
-    }
-
-    void UnmarkMemory(MapInterval* map) {
-        if (!map->is_memory_marked) {
+                         new_map->GetGpuAddress());
             return;
         }
-        const std::size_t size = map->end - map->start;
-        rasterizer.UpdatePagesCachedCount(map->start, size, -1);
-        map->is_memory_marked = false;
+        const std::size_t size = new_map->GetEnd() - new_map->GetStart();
+        new_map->MarkAsRegistered(true);
+        const IntervalType interval{new_map->GetStart(), new_map->GetEnd()};
+        mapped_addresses.insert({interval, new_map});
+        rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
+        new_map->SetMemoryMarked(true);
+        if (inherit_written) {
+            MarkRegionAsWritten(new_map->GetStart(), new_map->GetEnd() - 1);
+            new_map->MarkAsWritten(true);
+        }
+    }
+
+    void UnmarkMemory(const MapInterval& map) {
+        if (!map->IsMemoryMarked()) {
+            return;
+        }
+        const std::size_t size = map->GetEnd() - map->GetStart();
+        rasterizer.UpdatePagesCachedCount(map->GetStart(), size, -1);
+        map->SetMemoryMarked(false);
     }
 
     /// Unregisters an object from the cache
-    void Unregister(MapInterval* map) {
+    void Unregister(const MapInterval& map) {
         UnmarkMemory(map);
-        map->is_registered = false;
-        if (map->is_sync_pending) {
-            map->is_sync_pending = false;
+        map->MarkAsRegistered(false);
+        if (map->IsSyncPending()) {
             marked_for_unregister.remove(map);
+            map->SetSyncPending(false);
         }
-        if (map->is_written) {
-            UnmarkRegionAsWritten(map->start, map->end - 1);
+        if (map->IsWritten()) {
+            UnmarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
         }
-        const auto it = mapped_addresses.find(*map);
-        ASSERT(it != mapped_addresses.end());
-        mapped_addresses.erase(it);
-        mapped_addresses_allocator.Release(map);
+        const IntervalType delete_interval{map->GetStart(), map->GetEnd()};
+        mapped_addresses.erase(delete_interval);
     }
 
 private:
-    MapInterval* MapAddress(const OwnerBuffer& block, GPUVAddr gpu_addr, VAddr cpu_addr,
-                            std::size_t size) {
-        const VectorMapInterval overlaps = GetMapsInRange(cpu_addr, size);
+    MapInterval CreateMap(const VAddr start, const VAddr end, const GPUVAddr gpu_addr) {
+        return std::make_shared<MapIntervalBase>(start, end, gpu_addr);
+    }
+
+    MapInterval MapAddress(const OwnerBuffer& block, const GPUVAddr gpu_addr, const VAddr cpu_addr,
+                           const std::size_t size) {
+        std::vector<MapInterval> overlaps = GetMapsInRange(cpu_addr, size);
         if (overlaps.empty()) {
             auto& memory_manager = system.GPU().MemoryManager();
             const VAddr cpu_addr_end = cpu_addr + size;
+            MapInterval new_map = CreateMap(cpu_addr, cpu_addr_end, gpu_addr);
             if (memory_manager.IsGranularRange(gpu_addr, size)) {
                 u8* host_ptr = memory_manager.GetPointer(gpu_addr);
                 UploadBlockData(block, block->GetOffset(cpu_addr), size, host_ptr);
@@ -329,12 +328,13 @@ private:
                 memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
                 UploadBlockData(block, block->GetOffset(cpu_addr), size, staging_buffer.data());
             }
-            return Register(MapInterval(cpu_addr, cpu_addr_end, gpu_addr));
+            Register(new_map);
+            return new_map;
         }
 
         const VAddr cpu_addr_end = cpu_addr + size;
         if (overlaps.size() == 1) {
-            MapInterval* const current_map = overlaps[0];
+            MapInterval& current_map = overlaps[0];
             if (current_map->IsInside(cpu_addr, cpu_addr_end)) {
                 return current_map;
             }
@@ -344,39 +344,35 @@ private:
         bool write_inheritance = false;
         bool modified_inheritance = false;
         // Calculate new buffer parameters
-        for (MapInterval* overlap : overlaps) {
-            new_start = std::min(overlap->start, new_start);
-            new_end = std::max(overlap->end, new_end);
-            write_inheritance |= overlap->is_written;
-            modified_inheritance |= overlap->is_modified;
+        for (auto& overlap : overlaps) {
+            new_start = std::min(overlap->GetStart(), new_start);
+            new_end = std::max(overlap->GetEnd(), new_end);
+            write_inheritance |= overlap->IsWritten();
+            modified_inheritance |= overlap->IsModified();
         }
         GPUVAddr new_gpu_addr = gpu_addr + new_start - cpu_addr;
         for (auto& overlap : overlaps) {
             Unregister(overlap);
         }
         UpdateBlock(block, new_start, new_end, overlaps);
-
-        const MapInterval new_map{new_start, new_end, new_gpu_addr};
-        MapInterval* const map = Register(new_map, write_inheritance);
-        if (!map) {
-            return nullptr;
-        }
+        MapInterval new_map = CreateMap(new_start, new_end, new_gpu_addr);
         if (modified_inheritance) {
-            map->MarkAsModified(true, GetModifiedTicks());
+            new_map->MarkAsModified(true, GetModifiedTicks());
             if (Settings::IsGPULevelHigh() && Settings::values.use_asynchronous_gpu_emulation) {
-                MarkForAsyncFlush(map);
+                MarkForAsyncFlush(new_map);
             }
         }
-        return map;
+        Register(new_map, write_inheritance);
+        return new_map;
     }
 
     void UpdateBlock(const OwnerBuffer& block, VAddr start, VAddr end,
-                     const VectorMapInterval& overlaps) {
+                     std::vector<MapInterval>& overlaps) {
         const IntervalType base_interval{start, end};
         IntervalSet interval_set{};
         interval_set.add(base_interval);
         for (auto& overlap : overlaps) {
-            const IntervalType subtract{overlap->start, overlap->end};
+            const IntervalType subtract{overlap->GetStart(), overlap->GetEnd()};
             interval_set.subtract(subtract);
         }
         for (auto& interval : interval_set) {
@@ -390,24 +386,18 @@ private:
         }
     }
 
-    VectorMapInterval GetMapsInRange(VAddr addr, std::size_t size) {
-        VectorMapInterval result;
+    std::vector<MapInterval> GetMapsInRange(VAddr addr, std::size_t size) {
         if (size == 0) {
-            return result;
+            return {};
         }
 
-        const VAddr addr_end = addr + size;
-        auto it = mapped_addresses.lower_bound(addr);
-        if (it != mapped_addresses.begin()) {
-            --it;
+        std::vector<MapInterval> objects{};
+        const IntervalType interval{addr, addr + size};
+        for (auto& pair : boost::make_iterator_range(mapped_addresses.equal_range(interval))) {
+            objects.push_back(pair.second);
         }
-        while (it != mapped_addresses.end() && it->start < addr_end) {
-            if (it->Overlaps(addr, addr_end)) {
-                result.push_back(&*it);
-            }
-            ++it;
-        }
-        return result;
+
+        return objects;
     }
 
     /// Returns a ticks counter used for tracking when cached objects were last modified
@@ -415,12 +405,12 @@ private:
         return ++modified_ticks;
     }
 
-    void FlushMap(MapInterval* map) {
-        const std::size_t size = map->end - map->start;
-        OwnerBuffer block = blocks[map->start >> block_page_bits];
+    void FlushMap(MapInterval map) {
+        std::size_t size = map->GetEnd() - map->GetStart();
+        OwnerBuffer block = blocks[map->GetStart() >> block_page_bits];
         staging_buffer.resize(size);
-        DownloadBlockData(block, block->GetOffset(map->start), size, staging_buffer.data());
-        system.Memory().WriteBlockUnsafe(map->start, staging_buffer.data(), size);
+        DownloadBlockData(block, block->GetOffset(map->GetStart()), size, staging_buffer.data());
+        system.Memory().WriteBlockUnsafe(map->GetStart(), staging_buffer.data(), size);
         map->MarkAsModified(false, 0);
     }
 
@@ -525,7 +515,7 @@ private:
             } else {
                 written_pages[page_start] = 1;
             }
-            ++page_start;
+            page_start++;
         }
     }
 
@@ -541,7 +531,7 @@ private:
                     written_pages.erase(it);
                 }
             }
-            ++page_start;
+            page_start++;
         }
     }
 
@@ -552,14 +542,14 @@ private:
             if (written_pages.count(page_start) > 0) {
                 return true;
             }
-            ++page_start;
+            page_start++;
         }
         return false;
     }
 
-    void MarkForAsyncFlush(MapInterval* map) {
+    void MarkForAsyncFlush(MapInterval& map) {
         if (!uncommitted_flushes) {
-            uncommitted_flushes = std::make_shared<std::unordered_set<MapInterval*>>();
+            uncommitted_flushes = std::make_shared<std::unordered_set<MapInterval>>();
         }
         uncommitted_flushes->insert(map);
     }
@@ -576,9 +566,10 @@ private:
     u64 buffer_offset = 0;
     u64 buffer_offset_base = 0;
 
-    MapIntervalAllocator mapped_addresses_allocator;
-    boost::intrusive::set<MapInterval, boost::intrusive::compare<MapIntervalCompare>>
-        mapped_addresses;
+    using IntervalSet = boost::icl::interval_set<VAddr>;
+    using IntervalCache = boost::icl::interval_map<VAddr, MapInterval>;
+    using IntervalType = typename IntervalCache::interval_type;
+    IntervalCache mapped_addresses;
 
     static constexpr u64 write_page_bit = 11;
     std::unordered_map<u64, u32> written_pages;
@@ -592,10 +583,10 @@ private:
     u64 modified_ticks = 0;
 
     std::vector<u8> staging_buffer;
-    std::list<MapInterval*> marked_for_unregister;
+    std::list<MapInterval> marked_for_unregister;
 
-    std::shared_ptr<std::unordered_set<MapInterval*>> uncommitted_flushes;
-    std::list<std::shared_ptr<std::list<MapInterval*>>> committed_flushes;
+    std::shared_ptr<std::unordered_set<MapInterval>> uncommitted_flushes{};
+    std::list<std::shared_ptr<std::list<MapInterval>>> committed_flushes;
 
     std::recursive_mutex mutex;
 };

src/video_core/buffer_cache/map_interval.cpp

@@ -1,33 +0,0 @@
-// Copyright 2020 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <array>
-#include <cstddef>
-#include <memory>
-
-#include "video_core/buffer_cache/map_interval.h"
-
-namespace VideoCommon {
-
-MapIntervalAllocator::MapIntervalAllocator() {
-    FillFreeList(first_chunk);
-}
-
-MapIntervalAllocator::~MapIntervalAllocator() = default;
-
-void MapIntervalAllocator::AllocateNewChunk() {
-    *new_chunk = std::make_unique<Chunk>();
-    FillFreeList(**new_chunk);
-    new_chunk = &(*new_chunk)->next;
-}
-
-void MapIntervalAllocator::FillFreeList(Chunk& chunk) {
-    const std::size_t old_size = free_list.size();
-    free_list.resize(old_size + chunk.data.size());
-    std::transform(chunk.data.rbegin(), chunk.data.rend(), free_list.begin() + old_size,
-                   [](MapInterval& interval) { return &interval; });
-}
-
-} // namespace VideoCommon

src/video_core/buffer_cache/map_interval.h

@@ -4,89 +4,104 @@
 
 #pragma once
 
-#include <array>
-#include <cstddef>
-#include <memory>
-#include <vector>
-
-#include <boost/intrusive/set_hook.hpp>
-
 #include "common/common_types.h"
 #include "video_core/gpu.h"
 
 namespace VideoCommon {
 
-struct MapInterval : public boost::intrusive::set_base_hook<boost::intrusive::optimize_size<true>> {
-    MapInterval() = default;
-
-    /*implicit*/ MapInterval(VAddr start_) noexcept : start{start_} {}
-
-    explicit MapInterval(VAddr start_, VAddr end_, GPUVAddr gpu_addr_) noexcept
-        : start{start_}, end{end_}, gpu_addr{gpu_addr_} {}
-
-    bool IsInside(VAddr other_start, VAddr other_end) const noexcept {
-        return start <= other_start && other_end <= end;
-    }
-
-    bool Overlaps(VAddr other_start, VAddr other_end) const noexcept {
-        return start < other_end && other_start < end;
-    }
-
-    void MarkAsModified(bool is_modified_, u64 ticks_) noexcept {
-        is_modified = is_modified_;
-        ticks = ticks_;
-    }
-
-    boost::intrusive::set_member_hook<> member_hook_;
-    VAddr start = 0;
-    VAddr end = 0;
-    GPUVAddr gpu_addr = 0;
-    u64 ticks = 0;
-    bool is_written = false;
-    bool is_modified = false;
-    bool is_registered = false;
-    bool is_memory_marked = false;
-    bool is_sync_pending = false;
-};
-
-struct MapIntervalCompare {
-    constexpr bool operator()(const MapInterval& lhs, const MapInterval& rhs) const noexcept {
-        return lhs.start < rhs.start;
-    }
-};
-
-class MapIntervalAllocator {
+class MapIntervalBase {
 public:
-    MapIntervalAllocator();
-    ~MapIntervalAllocator();
+    MapIntervalBase(const VAddr start, const VAddr end, const GPUVAddr gpu_addr)
+        : start{start}, end{end}, gpu_addr{gpu_addr} {}
 
-    MapInterval* Allocate() {
-        if (free_list.empty()) {
-            AllocateNewChunk();
-        }
-        MapInterval* const interval = free_list.back();
-        free_list.pop_back();
-        return interval;
+    void SetCpuAddress(VAddr new_cpu_addr) {
+        cpu_addr = new_cpu_addr;
     }
 
-    void Release(MapInterval* interval) {
-        free_list.push_back(interval);
+    VAddr GetCpuAddress() const {
+        return cpu_addr;
+    }
+
+    GPUVAddr GetGpuAddress() const {
+        return gpu_addr;
+    }
+
+    bool IsInside(const VAddr other_start, const VAddr other_end) const {
+        return (start <= other_start && other_end <= end);
+    }
+
+    bool operator==(const MapIntervalBase& rhs) const {
+        return std::tie(start, end) == std::tie(rhs.start, rhs.end);
+    }
+
+    bool operator!=(const MapIntervalBase& rhs) const {
+        return !operator==(rhs);
+    }
+
+    void MarkAsRegistered(const bool registered) {
+        is_registered = registered;
+    }
+
+    bool IsRegistered() const {
+        return is_registered;
+    }
+
+    void SetMemoryMarked(bool is_memory_marked_) {
+        is_memory_marked = is_memory_marked_;
+    }
+
+    bool IsMemoryMarked() const {
+        return is_memory_marked;
+    }
+
+    void SetSyncPending(bool is_sync_pending_) {
+        is_sync_pending = is_sync_pending_;
+    }
+
+    bool IsSyncPending() const {
+        return is_sync_pending;
+    }
+
+    VAddr GetStart() const {
+        return start;
+    }
+
+    VAddr GetEnd() const {
+        return end;
+    }
+
+    void MarkAsModified(const bool is_modified_, const u64 tick) {
+        is_modified = is_modified_;
+        ticks = tick;
+    }
+
+    bool IsModified() const {
+        return is_modified;
+    }
+
+    u64 GetModificationTick() const {
+        return ticks;
+    }
+
+    void MarkAsWritten(const bool is_written_) {
+        is_written = is_written_;
+    }
+
+    bool IsWritten() const {
+        return is_written;
     }
 
 private:
-    struct Chunk {
-        std::unique_ptr<Chunk> next;
-        std::array<MapInterval, 0x8000> data;
-    };
-
-    void AllocateNewChunk();
-
-    void FillFreeList(Chunk& chunk);
-
-    std::vector<MapInterval*> free_list;
-    std::unique_ptr<Chunk>* new_chunk = &first_chunk.next;
-
-    Chunk first_chunk;
+    VAddr start;
+    VAddr end;
+    GPUVAddr gpu_addr;
+    VAddr cpu_addr{};
+    bool is_written{};
+    bool is_modified{};
+    bool is_registered{};
+    bool is_memory_marked{};
+    bool is_sync_pending{};
+    u64 ticks{};
 };
 
 } // namespace VideoCommon

src/video_core/engines/maxwell_3d.cpp

@@ -457,9 +457,8 @@ void Maxwell3D::StampQueryResult(u64 payload, bool long_query) {
 
 void Maxwell3D::ProcessQueryGet() {
     // TODO(Subv): Support the other query units.
-    if (regs.query.query_get.unit != Regs::QueryUnit::Crop) {
-        LOG_DEBUG(HW_GPU, "Units other than CROP are unimplemented");
-    }
+    ASSERT_MSG(regs.query.query_get.unit == Regs::QueryUnit::Crop,
+               "Units other than CROP are unimplemented");
 
     switch (regs.query.query_get.operation) {
     case Regs::QueryOperation::Release:
@@ -535,8 +534,8 @@ void Maxwell3D::ProcessCounterReset() {
         rasterizer.ResetCounter(QueryType::SamplesPassed);
         break;
     default:
-        LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}",
-                  static_cast<int>(regs.counter_reset));
+        LOG_WARNING(Render_OpenGL, "Unimplemented counter reset={}",
+                    static_cast<int>(regs.counter_reset));
         break;
     }
 }
@@ -593,8 +592,8 @@ std::optional<u64> Maxwell3D::GetQueryResult() {
                          system.GPU().GetTicks());
         return {};
     default:
-        LOG_DEBUG(HW_GPU, "Unimplemented query select type {}",
-                  static_cast<u32>(regs.query.query_get.select.Value()));
+        UNIMPLEMENTED_MSG("Unimplemented query select type {}",
+                          static_cast<u32>(regs.query.query_get.select.Value()));
         return 1;
     }
 }

src/video_core/rasterizer_cache.h

@@ -56,27 +56,9 @@ public:
         last_modified_ticks = cache.GetModifiedTicks();
     }
 
-    void SetMemoryMarked(bool is_memory_marked_) {
-        is_memory_marked = is_memory_marked_;
-    }
-
-    bool IsMemoryMarked() const {
-        return is_memory_marked;
-    }
-
-    void SetSyncPending(bool is_sync_pending_) {
-        is_sync_pending = is_sync_pending_;
-    }
-
-    bool IsSyncPending() const {
-        return is_sync_pending;
-    }
-
 private:
     bool is_registered{};      ///< Whether the object is currently registered with the cache
     bool is_dirty{};           ///< Whether the object is dirty (out of sync with guest memory)
-    bool is_memory_marked{};   ///< Whether the object is marking rasterizer memory.
-    bool is_sync_pending{};    ///< Whether the object is pending deletion.
     u64 last_modified_ticks{}; ///< When the object was last modified, used for in-order flushing
     VAddr cpu_addr{};          ///< Cpu address memory, unique from emulated virtual address space
 };
@@ -112,30 +94,6 @@ public:
         }
     }
 
-    void OnCPUWrite(VAddr addr, std::size_t size) {
-        std::lock_guard lock{mutex};
-
-        for (const auto& object : GetSortedObjectsFromRegion(addr, size)) {
-            if (object->IsRegistered()) {
-                UnmarkMemory(object);
-                object->SetSyncPending(true);
-                marked_for_unregister.emplace_back(object);
-            }
-        }
-    }
-
-    void SyncGuestHost() {
-        std::lock_guard lock{mutex};
-
-        for (const auto& object : marked_for_unregister) {
-            if (object->IsRegistered()) {
-                object->SetSyncPending(false);
-                Unregister(object);
-            }
-        }
-        marked_for_unregister.clear();
-    }
-
     /// Invalidates everything in the cache
     void InvalidateAll() {
         std::lock_guard lock{mutex};
@@ -162,32 +120,19 @@ protected:
         interval_cache.add({GetInterval(object), ObjectSet{object}});
         map_cache.insert({object->GetCpuAddr(), object});
         rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), 1);
-        object->SetMemoryMarked(true);
     }
 
     /// Unregisters an object from the cache
     virtual void Unregister(const T& object) {
         std::lock_guard lock{mutex};
 
-        UnmarkMemory(object);
         object->SetIsRegistered(false);
-        if (object->IsSyncPending()) {
-            marked_for_unregister.remove(object);
-            object->SetSyncPending(false);
-        }
+        rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), -1);
         const VAddr addr = object->GetCpuAddr();
         interval_cache.subtract({GetInterval(object), ObjectSet{object}});
         map_cache.erase(addr);
     }
 
-    void UnmarkMemory(const T& object) {
-        if (!object->IsMemoryMarked()) {
-            return;
-        }
-        rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), -1);
-        object->SetMemoryMarked(false);
-    }
-
     /// Returns a ticks counter used for tracking when cached objects were last modified
     u64 GetModifiedTicks() {
         std::lock_guard lock{mutex};
@@ -249,5 +194,4 @@ private:
     IntervalCache interval_cache; ///< Cache of objects
     u64 modified_ticks{};         ///< Counter of cache state ticks, used for in-order flushing
     VideoCore::RasterizerInterface& rasterizer;
-    std::list<T> marked_for_unregister;
 };

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -8,7 +8,6 @@
 
 #include "common/assert.h"
 #include "common/microprofile.h"
-#include "video_core/buffer_cache/buffer_cache.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"

src/video_core/renderer_opengl/gl_device.cpp

@@ -13,7 +13,6 @@
 
 #include "common/logging/log.h"
 #include "common/scope_exit.h"
-#include "core/settings.h"
 #include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 
@@ -184,16 +183,10 @@ Device::Device() : base_bindings{BuildBaseBindings()} {
     has_precise_bug = TestPreciseBug();
     has_broken_compute = is_intel_proprietary;
     has_fast_buffer_sub_data = is_nvidia;
-    use_assembly_shaders = Settings::values.use_assembly_shaders && GLAD_GL_NV_gpu_program5 &&
-                           GLAD_GL_NV_compute_program5;
 
     LOG_INFO(Render_OpenGL, "Renderer_VariableAOFFI: {}", has_variable_aoffi);
     LOG_INFO(Render_OpenGL, "Renderer_ComponentIndexingBug: {}", has_component_indexing_bug);
     LOG_INFO(Render_OpenGL, "Renderer_PreciseBug: {}", has_precise_bug);
-
-    if (Settings::values.use_assembly_shaders && !use_assembly_shaders) {
-        LOG_ERROR(Render_OpenGL, "Assembly shaders enabled but not supported");
-    }
 }
 
 Device::Device(std::nullptr_t) {

src/video_core/renderer_opengl/gl_device.h

@@ -88,10 +88,6 @@ public:
         return has_fast_buffer_sub_data;
     }
 
-    bool UseAssemblyShaders() const {
-        return use_assembly_shaders;
-    }
-
 private:
     static bool TestVariableAoffi();
     static bool TestPreciseBug();
@@ -111,7 +107,6 @@ private:
     bool has_precise_bug{};
     bool has_broken_compute{};
     bool has_fast_buffer_sub_data{};
-    bool use_assembly_shaders{};
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_fence_manager.cpp

@@ -4,7 +4,6 @@
 
 #include "common/assert.h"
 
-#include "video_core/renderer_opengl/gl_buffer_cache.h"
 #include "video_core/renderer_opengl/gl_fence_manager.h"
 
 namespace OpenGL {

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -94,30 +94,17 @@ void oglEnable(GLenum cap, bool state) {
 } // Anonymous namespace
 
 RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
-                                   const Device& device, ScreenInfo& info,
-                                   ProgramManager& program_manager, StateTracker& state_tracker)
-    : RasterizerAccelerated{system.Memory()}, device{device}, texture_cache{system, *this, device,
-                                                                            state_tracker},
+                                   ScreenInfo& info, GLShader::ProgramManager& program_manager,
+                                   StateTracker& state_tracker)
+    : RasterizerAccelerated{system.Memory()}, texture_cache{system, *this, device, state_tracker},
       shader_cache{*this, system, emu_window, device}, query_cache{system, *this},
       buffer_cache{*this, system, device, STREAM_BUFFER_SIZE},
       fence_manager{system, *this, texture_cache, buffer_cache, query_cache}, system{system},
       screen_info{info}, program_manager{program_manager}, state_tracker{state_tracker} {
     CheckExtensions();
-
-    if (device.UseAssemblyShaders()) {
-        glCreateBuffers(static_cast<GLsizei>(staging_cbufs.size()), staging_cbufs.data());
-        for (const GLuint cbuf : staging_cbufs) {
-            glNamedBufferStorage(cbuf, static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize),
-                                 nullptr, 0);
-        }
-    }
 }
 
-RasterizerOpenGL::~RasterizerOpenGL() {
-    if (device.UseAssemblyShaders()) {
-        glDeleteBuffers(static_cast<GLsizei>(staging_cbufs.size()), staging_cbufs.data());
-    }
-}
+RasterizerOpenGL::~RasterizerOpenGL() {}
 
 void RasterizerOpenGL::CheckExtensions() {
     if (!GLAD_GL_ARB_texture_filter_anisotropic && !GLAD_GL_EXT_texture_filter_anisotropic) {
@@ -243,7 +230,6 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() {
 void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     MICROPROFILE_SCOPE(OpenGL_Shader);
     auto& gpu = system.GPU().Maxwell3D();
-    std::size_t num_ssbos = 0;
     u32 clip_distances = 0;
 
     for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
@@ -275,14 +261,6 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
 
         Shader shader{shader_cache.GetStageProgram(program)};
 
-        if (device.UseAssemblyShaders()) {
-            // Check for ARB limitation. We only have 16 SSBOs per context state. To workaround this
-            // all stages share the same bindings.
-            const std::size_t num_stage_ssbos = shader->GetEntries().global_memory_entries.size();
-            ASSERT_MSG(num_stage_ssbos == 0 || num_ssbos == 0, "SSBOs on more than one stage");
-            num_ssbos += num_stage_ssbos;
-        }
-
         // Stage indices are 0 - 5
         const std::size_t stage = index == 0 ? 0 : index - 1;
         SetupDrawConstBuffers(stage, shader);
@@ -548,7 +526,6 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     SyncFramebufferSRGB();
 
     buffer_cache.Acquire();
-    current_cbuf = 0;
 
     std::size_t buffer_size = CalculateVertexArraysSize();
 
@@ -558,9 +535,9 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     }
 
     // Uniform space for the 5 shader stages
-    buffer_size =
-        Common::AlignUp<std::size_t>(buffer_size, 4) +
-        (sizeof(MaxwellUniformData) + device.GetUniformBufferAlignment()) * Maxwell::MaxShaderStage;
+    buffer_size = Common::AlignUp<std::size_t>(buffer_size, 4) +
+                  (sizeof(GLShader::MaxwellUniformData) + device.GetUniformBufferAlignment()) *
+                      Maxwell::MaxShaderStage;
 
     // Add space for at least 18 constant buffers
     buffer_size += Maxwell::MaxConstBuffers *
@@ -581,14 +558,12 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     }
 
     // Setup emulation uniform buffer.
-    if (!device.UseAssemblyShaders()) {
-        MaxwellUniformData ubo;
-        ubo.SetFromRegs(gpu);
-        const auto [buffer, offset] =
-            buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
-        glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, buffer, offset,
-                          static_cast<GLsizeiptr>(sizeof(ubo)));
-    }
+    GLShader::MaxwellUniformData ubo;
+    ubo.SetFromRegs(gpu);
+    const auto [buffer, offset] =
+        buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
+    glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, buffer, offset,
+                      static_cast<GLsizeiptr>(sizeof(ubo)));
 
     // Setup shaders and their used resources.
     texture_cache.GuardSamplers(true);
@@ -660,11 +635,11 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
     }
 
     buffer_cache.Acquire();
-    current_cbuf = 0;
 
     auto kernel = shader_cache.GetComputeKernel(code_addr);
     SetupComputeTextures(kernel);
     SetupComputeImages(kernel);
+    program_manager.BindComputeShader(kernel->GetHandle());
 
     const std::size_t buffer_size =
         Tegra::Engines::KeplerCompute::NumConstBuffers *
@@ -677,7 +652,6 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
     buffer_cache.Unmap();
 
     const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
-    program_manager.BindCompute(kernel->GetHandle());
     glDispatchCompute(launch_desc.grid_dim_x, launch_desc.grid_dim_y, launch_desc.grid_dim_z);
     ++num_queued_commands;
 }
@@ -727,15 +701,15 @@ void RasterizerOpenGL::OnCPUWrite(VAddr addr, u64 size) {
         return;
     }
     texture_cache.OnCPUWrite(addr, size);
-    shader_cache.OnCPUWrite(addr, size);
+    shader_cache.InvalidateRegion(addr, size);
     buffer_cache.OnCPUWrite(addr, size);
+    query_cache.InvalidateRegion(addr, size);
 }
 
 void RasterizerOpenGL::SyncGuestHost() {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     texture_cache.SyncGuestHost();
     buffer_cache.SyncGuestHost();
-    shader_cache.SyncGuestHost();
 }
 
 void RasterizerOpenGL::SignalSemaphore(GPUVAddr addr, u32 value) {
@@ -838,20 +812,14 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
 }
 
 void RasterizerOpenGL::SetupDrawConstBuffers(std::size_t stage_index, const Shader& shader) {
-    static constexpr std::array PARAMETER_LUT = {
-        GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV, GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,
-        GL_TESS_EVALUATION_PROGRAM_PARAMETER_BUFFER_NV, GL_GEOMETRY_PROGRAM_PARAMETER_BUFFER_NV,
-        GL_FRAGMENT_PROGRAM_PARAMETER_BUFFER_NV};
-
     MICROPROFILE_SCOPE(OpenGL_UBO);
     const auto& stages = system.GPU().Maxwell3D().state.shader_stages;
     const auto& shader_stage = stages[stage_index];
 
-    u32 binding =
-        device.UseAssemblyShaders() ? 0 : device.GetBaseBindings(stage_index).uniform_buffer;
+    u32 binding = device.GetBaseBindings(stage_index).uniform_buffer;
     for (const auto& entry : shader->GetEntries().const_buffers) {
         const auto& buffer = shader_stage.const_buffers[entry.GetIndex()];
-        SetupConstBuffer(PARAMETER_LUT[stage_index], binding++, buffer, entry);
+        SetupConstBuffer(binding++, buffer, entry);
     }
 }
 
@@ -867,21 +835,16 @@ void RasterizerOpenGL::SetupComputeConstBuffers(const Shader& kernel) {
         buffer.address = config.Address();
         buffer.size = config.size;
         buffer.enabled = mask[entry.GetIndex()];
-        SetupConstBuffer(GL_COMPUTE_PROGRAM_PARAMETER_BUFFER_NV, binding++, buffer, entry);
+        SetupConstBuffer(binding++, buffer, entry);
     }
 }
 
-void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
-                                        const Tegra::Engines::ConstBufferInfo& buffer,
+void RasterizerOpenGL::SetupConstBuffer(u32 binding, const Tegra::Engines::ConstBufferInfo& buffer,
                                         const ConstBufferEntry& entry) {
     if (!buffer.enabled) {
         // Set values to zero to unbind buffers
-        if (device.UseAssemblyShaders()) {
-            glBindBufferRangeNV(stage, entry.GetIndex(), 0, 0, 0);
-        } else {
-            glBindBufferRange(GL_UNIFORM_BUFFER, binding,
-                              buffer_cache.GetEmptyBuffer(sizeof(float)), 0, sizeof(float));
-        }
+        glBindBufferRange(GL_UNIFORM_BUFFER, binding, buffer_cache.GetEmptyBuffer(sizeof(float)), 0,
+                          sizeof(float));
         return;
     }
 
@@ -890,19 +853,9 @@ void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
     const std::size_t size = Common::AlignUp(GetConstBufferSize(buffer, entry), sizeof(GLvec4));
 
     const auto alignment = device.GetUniformBufferAlignment();
-    auto [cbuf, offset] = buffer_cache.UploadMemory(buffer.address, size, alignment, false,
-                                                    device.HasFastBufferSubData());
-    if (!device.UseAssemblyShaders()) {
-        glBindBufferRange(GL_UNIFORM_BUFFER, binding, cbuf, offset, size);
-        return;
-    }
-    if (offset != 0) {
-        const GLuint staging_cbuf = staging_cbufs[current_cbuf++];
-        glCopyNamedBufferSubData(cbuf, staging_cbuf, offset, 0, size);
-        cbuf = staging_cbuf;
-        offset = 0;
-    }
-    glBindBufferRangeNV(stage, binding, cbuf, offset, size);
+    const auto [cbuf, offset] = buffer_cache.UploadMemory(buffer.address, size, alignment, false,
+                                                          device.HasFastBufferSubData());
+    glBindBufferRange(GL_UNIFORM_BUFFER, binding, cbuf, offset, size);
 }
 
 void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, const Shader& shader) {
@@ -910,8 +863,7 @@ void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, const Shad
     auto& memory_manager{gpu.MemoryManager()};
     const auto cbufs{gpu.Maxwell3D().state.shader_stages[stage_index]};
 
-    u32 binding =
-        device.UseAssemblyShaders() ? 0 : device.GetBaseBindings(stage_index).shader_storage_buffer;
+    u32 binding = device.GetBaseBindings(stage_index).shader_storage_buffer;
     for (const auto& entry : shader->GetEntries().global_memory_entries) {
         const GPUVAddr addr{cbufs.const_buffers[entry.cbuf_index].address + entry.cbuf_offset};
         const GPUVAddr gpu_addr{memory_manager.Read<u64>(addr)};
@@ -977,12 +929,16 @@ void RasterizerOpenGL::SetupTexture(u32 binding, const Tegra::Texture::FullTextu
         glBindTextureUnit(binding, 0);
         return;
     }
-    const GLuint handle = view->GetTexture(texture.tic.x_source, texture.tic.y_source,
-                                           texture.tic.z_source, texture.tic.w_source);
-    glBindTextureUnit(binding, handle);
-    if (!view->GetSurfaceParams().IsBuffer()) {
-        glBindSampler(binding, sampler_cache.GetSampler(texture.tsc));
+    glBindTextureUnit(binding, view->GetTexture());
+
+    if (view->GetSurfaceParams().IsBuffer()) {
+        return;
     }
+    // Apply swizzle to textures that are not buffers.
+    view->ApplySwizzle(texture.tic.x_source, texture.tic.y_source, texture.tic.z_source,
+                       texture.tic.w_source);
+
+    glBindSampler(binding, sampler_cache.GetSampler(texture.tsc));
 }
 
 void RasterizerOpenGL::SetupDrawImages(std::size_t stage_index, const Shader& shader) {
@@ -1011,11 +967,14 @@ void RasterizerOpenGL::SetupImage(u32 binding, const Tegra::Texture::TICEntry& t
         glBindImageTexture(binding, 0, 0, GL_FALSE, 0, GL_READ_ONLY, GL_R8);
         return;
     }
+    if (!tic.IsBuffer()) {
+        view->ApplySwizzle(tic.x_source, tic.y_source, tic.z_source, tic.w_source);
+    }
     if (entry.is_written) {
         view->MarkAsModified(texture_cache.Tick());
     }
-    const GLuint handle = view->GetTexture(tic.x_source, tic.y_source, tic.z_source, tic.w_source);
-    glBindImageTexture(binding, handle, 0, GL_TRUE, 0, GL_READ_WRITE, view->GetFormat());
+    glBindImageTexture(binding, view->GetTexture(), 0, GL_TRUE, 0, GL_READ_WRITE,
+                       view->GetFormat());
 }
 
 void RasterizerOpenGL::SyncViewport() {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -56,8 +56,8 @@ struct DrawParameters;
 class RasterizerOpenGL : public VideoCore::RasterizerAccelerated {
 public:
     explicit RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
-                              const Device& device, ScreenInfo& info,
-                              ProgramManager& program_manager, StateTracker& state_tracker);
+                              ScreenInfo& info, GLShader::ProgramManager& program_manager,
+                              StateTracker& state_tracker);
     ~RasterizerOpenGL() override;
 
     void Draw(bool is_indexed, bool is_instanced) override;
@@ -106,7 +106,7 @@ private:
     void SetupComputeConstBuffers(const Shader& kernel);
 
     /// Configures a constant buffer.
-    void SetupConstBuffer(GLenum stage, u32 binding, const Tegra::Engines::ConstBufferInfo& buffer,
+    void SetupConstBuffer(u32 binding, const Tegra::Engines::ConstBufferInfo& buffer,
                           const ConstBufferEntry& entry);
 
     /// Configures the current global memory entries to use for the draw command.
@@ -224,7 +224,7 @@ private:
 
     void SetupShaders(GLenum primitive_mode);
 
-    const Device& device;
+    const Device device;
 
     TextureCacheOpenGL texture_cache;
     ShaderCacheOpenGL shader_cache;
@@ -236,7 +236,7 @@ private:
 
     Core::System& system;
     ScreenInfo& screen_info;
-    ProgramManager& program_manager;
+    GLShader::ProgramManager& program_manager;
     StateTracker& state_tracker;
 
     static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
@@ -248,12 +248,6 @@ private:
     std::bitset<Tegra::Engines::Maxwell3D::Regs::NumTransformFeedbackBuffers>
         enabled_transform_feedback_buffers;
 
-    static constexpr std::size_t NUM_CONSTANT_BUFFERS =
-        Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
-        Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram;
-    std::array<GLuint, NUM_CONSTANT_BUFFERS> staging_cbufs{};
-    std::size_t current_cbuf = 0;
-
     /// Number of commands queued to the OpenGL driver. Reseted on flush.
     std::size_t num_queued_commands = 0;
 

src/video_core/renderer_opengl/gl_resource_manager.cpp

@@ -125,15 +125,6 @@ void OGLProgram::Release() {
     handle = 0;
 }
 
-void OGLAssemblyProgram::Release() {
-    if (handle == 0) {
-        return;
-    }
-    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
-    glDeleteProgramsARB(1, &handle);
-    handle = 0;
-}
-
 void OGLPipeline::Create() {
     if (handle != 0)
         return;

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -167,22 +167,6 @@ public:
     GLuint handle = 0;
 };
 
-class OGLAssemblyProgram : private NonCopyable {
-public:
-    OGLAssemblyProgram() = default;
-
-    OGLAssemblyProgram(OGLAssemblyProgram&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
-
-    ~OGLAssemblyProgram() {
-        Release();
-    }
-
-    /// Deletes the internal OpenGL resource
-    void Release();
-
-    GLuint handle = 0;
-};
-
 class OGLPipeline : private NonCopyable {
 public:
     OGLPipeline() = default;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -97,24 +97,6 @@ constexpr ShaderType GetShaderType(Maxwell::ShaderProgram program_type) {
     return {};
 }
 
-constexpr GLenum AssemblyEnum(ShaderType shader_type) {
-    switch (shader_type) {
-    case ShaderType::Vertex:
-        return GL_VERTEX_PROGRAM_NV;
-    case ShaderType::TesselationControl:
-        return GL_TESS_CONTROL_PROGRAM_NV;
-    case ShaderType::TesselationEval:
-        return GL_TESS_EVALUATION_PROGRAM_NV;
-    case ShaderType::Geometry:
-        return GL_GEOMETRY_PROGRAM_NV;
-    case ShaderType::Fragment:
-        return GL_FRAGMENT_PROGRAM_NV;
-    case ShaderType::Compute:
-        return GL_COMPUTE_PROGRAM_NV;
-    }
-    return {};
-}
-
 std::string MakeShaderID(u64 unique_identifier, ShaderType shader_type) {
     return fmt::format("{}{:016X}", GetShaderTypeName(shader_type), unique_identifier);
 }
@@ -138,43 +120,18 @@ std::shared_ptr<Registry> MakeRegistry(const ShaderDiskCacheEntry& entry) {
     return registry;
 }
 
-ProgramSharedPtr BuildShader(const Device& device, ShaderType shader_type, u64 unique_identifier,
-                             const ShaderIR& ir, const Registry& registry,
-                             bool hint_retrievable = false) {
+std::shared_ptr<OGLProgram> BuildShader(const Device& device, ShaderType shader_type,
+                                        u64 unique_identifier, const ShaderIR& ir,
+                                        const Registry& registry, bool hint_retrievable = false) {
     const std::string shader_id = MakeShaderID(unique_identifier, shader_type);
     LOG_INFO(Render_OpenGL, "{}", shader_id);
 
-    auto program = std::make_shared<ProgramHandle>();
-
-    if (device.UseAssemblyShaders()) {
-        const std::string arb = "Not implemented";
-
-        GLuint& arb_prog = program->assembly_program.handle;
-
-// Commented out functions signal OpenGL errors but are compatible with apitrace.
-// Use them only to capture and replay on apitrace.
-#if 0
-        glGenProgramsNV(1, &arb_prog);
-        glLoadProgramNV(AssemblyEnum(shader_type), arb_prog, static_cast<GLsizei>(arb.size()),
-                        reinterpret_cast<const GLubyte*>(arb.data()));
-#else
-        glGenProgramsARB(1, &arb_prog);
-        glNamedProgramStringEXT(arb_prog, AssemblyEnum(shader_type), GL_PROGRAM_FORMAT_ASCII_ARB,
-                                static_cast<GLsizei>(arb.size()), arb.data());
-#endif
-        const auto err = reinterpret_cast<const char*>(glGetString(GL_PROGRAM_ERROR_STRING_NV));
-        if (err && *err) {
-            LOG_CRITICAL(Render_OpenGL, "{}", err);
-            LOG_INFO(Render_OpenGL, "\n{}", arb);
-        }
-    } else {
-        const std::string glsl = DecompileShader(device, ir, registry, shader_type, shader_id);
-        OGLShader shader;
-        shader.Create(glsl.c_str(), GetGLShaderType(shader_type));
-
-        program->source_program.Create(true, hint_retrievable, shader.handle);
-    }
+    const std::string glsl = DecompileShader(device, ir, registry, shader_type, shader_id);
+    OGLShader shader;
+    shader.Create(glsl.c_str(), GetGLShaderType(shader_type));
 
+    auto program = std::make_shared<OGLProgram>();
+    program->Create(true, hint_retrievable, shader.handle);
     return program;
 }
 
@@ -196,22 +153,15 @@ std::unordered_set<GLenum> GetSupportedFormats() {
 
 CachedShader::CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
                            std::shared_ptr<VideoCommon::Shader::Registry> registry,
-                           ShaderEntries entries, ProgramSharedPtr program_)
+                           ShaderEntries entries, std::shared_ptr<OGLProgram> program)
     : RasterizerCacheObject{cpu_addr}, registry{std::move(registry)}, entries{std::move(entries)},
-      size_in_bytes{size_in_bytes}, program{std::move(program_)} {
-    // Assign either the assembly program or source program. We can't have both.
-    handle = program->assembly_program.handle;
-    if (handle == 0) {
-        handle = program->source_program.handle;
-    }
-    ASSERT(handle != 0);
-}
+      size_in_bytes{size_in_bytes}, program{std::move(program)} {}
 
 CachedShader::~CachedShader() = default;
 
 GLuint CachedShader::GetHandle() const {
     DEBUG_ASSERT(registry->IsConsistent());
-    return handle;
+    return program->handle;
 }
 
 Shader CachedShader::CreateStageFromMemory(const ShaderParameters& params,
@@ -289,11 +239,7 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
         return;
     }
 
-    std::vector<ShaderDiskCachePrecompiled> gl_cache;
-    if (!device.UseAssemblyShaders()) {
-        // Only load precompiled cache when we are not using assembly shaders
-        gl_cache = disk_cache.LoadPrecompiled();
-    }
+    const std::vector gl_cache = disk_cache.LoadPrecompiled();
     const auto supported_formats = GetSupportedFormats();
 
     // Track if precompiled cache was altered during loading to know if we have to
@@ -332,7 +278,7 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
             auto registry = MakeRegistry(entry);
             const ShaderIR ir(entry.code, main_offset, COMPILER_SETTINGS, *registry);
 
-            ProgramSharedPtr program;
+            std::shared_ptr<OGLProgram> program;
             if (precompiled_entry) {
                 // If the shader is precompiled, attempt to load it with
                 program = GeneratePrecompiledProgram(entry, *precompiled_entry, supported_formats);
@@ -386,11 +332,6 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
         return;
     }
 
-    if (device.UseAssemblyShaders()) {
-        // Don't store precompiled binaries for assembly shaders.
-        return;
-    }
-
     // TODO(Rodrigo): Do state tracking for transferable shaders and do a dummy draw
     // before precompiling them
 
@@ -398,7 +339,7 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
         const u64 id = (*transferable)[i].unique_identifier;
         const auto it = find_precompiled(id);
         if (it == gl_cache.end()) {
-            const GLuint program = runtime_cache.at(id).program->source_program.handle;
+            const GLuint program = runtime_cache.at(id).program->handle;
             disk_cache.SavePrecompiled(id, program);
             precompiled_cache_altered = true;
         }
@@ -409,7 +350,7 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
     }
 }
 
-ProgramSharedPtr ShaderCacheOpenGL::GeneratePrecompiledProgram(
+std::shared_ptr<OGLProgram> ShaderCacheOpenGL::GeneratePrecompiledProgram(
     const ShaderDiskCacheEntry& entry, const ShaderDiskCachePrecompiled& precompiled_entry,
     const std::unordered_set<GLenum>& supported_formats) {
     if (supported_formats.find(precompiled_entry.binary_format) == supported_formats.end()) {
@@ -417,15 +358,15 @@ ProgramSharedPtr ShaderCacheOpenGL::GeneratePrecompiledProgram(
         return {};
     }
 
-    auto program = std::make_shared<ProgramHandle>();
-    GLuint& handle = program->source_program.handle;
-    handle = glCreateProgram();
-    glProgramParameteri(handle, GL_PROGRAM_SEPARABLE, GL_TRUE);
-    glProgramBinary(handle, precompiled_entry.binary_format, precompiled_entry.binary.data(),
+    auto program = std::make_shared<OGLProgram>();
+    program->handle = glCreateProgram();
+    glProgramParameteri(program->handle, GL_PROGRAM_SEPARABLE, GL_TRUE);
+    glProgramBinary(program->handle, precompiled_entry.binary_format,
+                    precompiled_entry.binary.data(),
                     static_cast<GLsizei>(precompiled_entry.binary.size()));
 
     GLint link_status;
-    glGetProgramiv(handle, GL_LINK_STATUS, &link_status);
+    glGetProgramiv(program->handle, GL_LINK_STATUS, &link_status);
     if (link_status == GL_FALSE) {
         LOG_INFO(Render_OpenGL, "Precompiled cache rejected by the driver, removing");
         return {};

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -43,14 +43,8 @@ struct UnspecializedShader;
 using Shader = std::shared_ptr<CachedShader>;
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
-struct ProgramHandle {
-    OGLProgram source_program;
-    OGLAssemblyProgram assembly_program;
-};
-using ProgramSharedPtr = std::shared_ptr<ProgramHandle>;
-
 struct PrecompiledShader {
-    ProgramSharedPtr program;
+    std::shared_ptr<OGLProgram> program;
     std::shared_ptr<VideoCommon::Shader::Registry> registry;
     ShaderEntries entries;
 };
@@ -93,13 +87,12 @@ public:
 private:
     explicit CachedShader(VAddr cpu_addr, std::size_t size_in_bytes,
                           std::shared_ptr<VideoCommon::Shader::Registry> registry,
-                          ShaderEntries entries, ProgramSharedPtr program);
+                          ShaderEntries entries, std::shared_ptr<OGLProgram> program);
 
     std::shared_ptr<VideoCommon::Shader::Registry> registry;
     ShaderEntries entries;
     std::size_t size_in_bytes = 0;
-    ProgramSharedPtr program;
-    GLuint handle = 0;
+    std::shared_ptr<OGLProgram> program;
 };
 
 class ShaderCacheOpenGL final : public RasterizerCache<Shader> {
@@ -122,7 +115,7 @@ protected:
     void FlushObjectInner(const Shader& object) override {}
 
 private:
-    ProgramSharedPtr GeneratePrecompiledProgram(
+    std::shared_ptr<OGLProgram> GeneratePrecompiledProgram(
         const ShaderDiskCacheEntry& entry, const ShaderDiskCachePrecompiled& precompiled_entry,
         const std::unordered_set<GLenum>& supported_formats);
 

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -1538,9 +1538,7 @@ private:
         Expression target;
         if (const auto gpr = std::get_if<GprNode>(&*dest)) {
             if (gpr->GetIndex() == Register::ZeroIndex) {
-                // Writing to Register::ZeroIndex is a no op but we still have to visit the source
-                // as it might have side effects.
-                code.AddLine("{};", Visit(src).GetCode());
+                // Writing to Register::ZeroIndex is a no op
                 return {};
             }
             target = {GetRegister(gpr->GetIndex()), Type::Float};
@@ -2311,18 +2309,6 @@ private:
         return {"gl_SubGroupInvocationARB", Type::Uint};
     }
 
-    template <const std::string_view& comparison>
-    Expression ThreadMask(Operation) {
-        if (device.HasWarpIntrinsics()) {
-            return {fmt::format("gl_Thread{}MaskNV", comparison), Type::Uint};
-        }
-        if (device.HasShaderBallot()) {
-            return {fmt::format("uint(gl_SubGroup{}MaskARB)", comparison), Type::Uint};
-        }
-        LOG_ERROR(Render_OpenGL, "Thread mask intrinsics are required by the shader");
-        return {"0U", Type::Uint};
-    }
-
     Expression ShuffleIndexed(Operation operation) {
         std::string value = VisitOperand(operation, 0).AsFloat();
 
@@ -2335,21 +2321,7 @@ private:
         return {fmt::format("readInvocationARB({}, {})", value, index), Type::Float};
     }
 
-    Expression Barrier(Operation) {
-        if (!ir.IsDecompiled()) {
-            LOG_ERROR(Render_OpenGL, "barrier() used but shader is not decompiled");
-            return {};
-        }
-        code.AddLine("barrier();");
-        return {};
-    }
-
-    Expression MemoryBarrierGroup(Operation) {
-        code.AddLine("groupMemoryBarrier();");
-        return {};
-    }
-
-    Expression MemoryBarrierGlobal(Operation) {
+    Expression MemoryBarrierGL(Operation) {
         code.AddLine("memoryBarrier();");
         return {};
     }
@@ -2365,12 +2337,6 @@ private:
         static constexpr std::string_view NotEqual = "!=";
         static constexpr std::string_view GreaterEqual = ">=";
 
-        static constexpr std::string_view Eq = "Eq";
-        static constexpr std::string_view Ge = "Ge";
-        static constexpr std::string_view Gt = "Gt";
-        static constexpr std::string_view Le = "Le";
-        static constexpr std::string_view Lt = "Lt";
-
         static constexpr std::string_view Add = "Add";
         static constexpr std::string_view Min = "Min";
         static constexpr std::string_view Max = "Max";
@@ -2588,16 +2554,9 @@ private:
         &GLSLDecompiler::VoteEqual,
 
         &GLSLDecompiler::ThreadId,
-        &GLSLDecompiler::ThreadMask<Func::Eq>,
-        &GLSLDecompiler::ThreadMask<Func::Ge>,
-        &GLSLDecompiler::ThreadMask<Func::Gt>,
-        &GLSLDecompiler::ThreadMask<Func::Le>,
-        &GLSLDecompiler::ThreadMask<Func::Lt>,
         &GLSLDecompiler::ShuffleIndexed,
 
-        &GLSLDecompiler::Barrier,
-        &GLSLDecompiler::MemoryBarrierGroup,
-        &GLSLDecompiler::MemoryBarrierGlobal,
+        &GLSLDecompiler::MemoryBarrierGL,
     };
     static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
 

src/video_core/renderer_opengl/gl_shader_manager.cpp

@@ -6,111 +6,47 @@
 
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
-#include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 
-namespace OpenGL {
+namespace OpenGL::GLShader {
 
-ProgramManager::ProgramManager(const Device& device) {
-    use_assembly_programs = device.UseAssemblyShaders();
-    if (use_assembly_programs) {
-        glEnable(GL_COMPUTE_PROGRAM_NV);
-    } else {
-        graphics_pipeline.Create();
-        glBindProgramPipeline(graphics_pipeline.handle);
-    }
-}
+ProgramManager::ProgramManager() = default;
 
 ProgramManager::~ProgramManager() = default;
 
-void ProgramManager::BindCompute(GLuint program) {
-    if (use_assembly_programs) {
-        glBindProgramARB(GL_COMPUTE_PROGRAM_NV, program);
-    } else {
-        is_graphics_bound = false;
-        glUseProgram(program);
-    }
+void ProgramManager::Create() {
+    graphics_pipeline.Create();
+    glBindProgramPipeline(graphics_pipeline.handle);
 }
 
 void ProgramManager::BindGraphicsPipeline() {
-    if (use_assembly_programs) {
-        UpdateAssemblyPrograms();
-    } else {
-        UpdateSourcePrograms();
-    }
-}
-
-void ProgramManager::BindHostPipeline(GLuint pipeline) {
-    if (use_assembly_programs) {
-        if (geometry_enabled) {
-            geometry_enabled = false;
-            old_state.geometry = 0;
-            glDisable(GL_GEOMETRY_PROGRAM_NV);
-        }
-    } else {
-        if (!is_graphics_bound) {
-            glUseProgram(0);
-        }
-    }
-    glBindProgramPipeline(pipeline);
-}
-
-void ProgramManager::RestoreGuestPipeline() {
-    if (use_assembly_programs) {
-        glBindProgramPipeline(0);
-    } else {
-        glBindProgramPipeline(graphics_pipeline.handle);
-    }
-}
-
-void ProgramManager::UpdateAssemblyPrograms() {
-    const auto update_state = [](GLenum stage, bool& enabled, GLuint current, GLuint old) {
-        if (current == old) {
-            return;
-        }
-        if (current == 0) {
-            if (enabled) {
-                enabled = false;
-                glDisable(stage);
-            }
-            return;
-        }
-        if (!enabled) {
-            enabled = true;
-            glEnable(stage);
-        }
-        glBindProgramARB(stage, current);
-    };
-
-    update_state(GL_VERTEX_PROGRAM_NV, vertex_enabled, current_state.vertex, old_state.vertex);
-    update_state(GL_GEOMETRY_PROGRAM_NV, geometry_enabled, current_state.geometry,
-                 old_state.geometry);
-    update_state(GL_FRAGMENT_PROGRAM_NV, fragment_enabled, current_state.fragment,
-                 old_state.fragment);
-
-    old_state = current_state;
-}
-
-void ProgramManager::UpdateSourcePrograms() {
     if (!is_graphics_bound) {
         is_graphics_bound = true;
         glUseProgram(0);
     }
 
+    // Avoid updating the pipeline when values have no changed
+    if (old_state == current_state) {
+        return;
+    }
+
+    // Workaround for AMD bug
+    static constexpr GLenum all_used_stages{GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT |
+                                            GL_FRAGMENT_SHADER_BIT};
     const GLuint handle = graphics_pipeline.handle;
-    const auto update_state = [handle](GLenum stage, GLuint current, GLuint old) {
-        if (current == old) {
-            return;
-        }
-        glUseProgramStages(handle, stage, current);
-    };
-    update_state(GL_VERTEX_SHADER_BIT, current_state.vertex, old_state.vertex);
-    update_state(GL_GEOMETRY_SHADER_BIT, current_state.geometry, old_state.geometry);
-    update_state(GL_FRAGMENT_SHADER_BIT, current_state.fragment, old_state.fragment);
+    glUseProgramStages(handle, all_used_stages, 0);
+    glUseProgramStages(handle, GL_VERTEX_SHADER_BIT, current_state.vertex_shader);
+    glUseProgramStages(handle, GL_GEOMETRY_SHADER_BIT, current_state.geometry_shader);
+    glUseProgramStages(handle, GL_FRAGMENT_SHADER_BIT, current_state.fragment_shader);
 
     old_state = current_state;
 }
 
+void ProgramManager::BindComputeShader(GLuint program) {
+    is_graphics_bound = false;
+    glUseProgram(program);
+}
+
 void MaxwellUniformData::SetFromRegs(const Tegra::Engines::Maxwell3D& maxwell) {
     const auto& regs = maxwell.regs;
 
@@ -118,4 +54,4 @@ void MaxwellUniformData::SetFromRegs(const Tegra::Engines::Maxwell3D& maxwell) {
     y_direction = regs.screen_y_control.y_negate == 0 ? 1.0f : -1.0f;
 }
 
-} // namespace OpenGL
+} // namespace OpenGL::GLShader

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -11,9 +11,7 @@
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/maxwell_to_gl.h"
 
-namespace OpenGL {
-
-class Device;
+namespace OpenGL::GLShader {
 
 /// Uniform structure for the Uniform Buffer Object, all vectors must be 16-byte aligned
 /// @note Always keep a vec4 at the end. The GL spec is not clear whether the alignment at
@@ -30,58 +28,50 @@ static_assert(sizeof(MaxwellUniformData) < 16384,
 
 class ProgramManager {
 public:
-    explicit ProgramManager(const Device& device);
+    explicit ProgramManager();
     ~ProgramManager();
 
-    /// Binds a compute program
-    void BindCompute(GLuint program);
+    void Create();
 
-    /// Updates bound programs.
+    /// Updates the graphics pipeline and binds it.
     void BindGraphicsPipeline();
 
-    /// Binds an OpenGL pipeline object unsynchronized with the guest state.
-    void BindHostPipeline(GLuint pipeline);
-
-    /// Rewinds BindHostPipeline state changes.
-    void RestoreGuestPipeline();
+    /// Binds a compute shader.
+    void BindComputeShader(GLuint program);
 
     void UseVertexShader(GLuint program) {
-        current_state.vertex = program;
+        current_state.vertex_shader = program;
     }
 
     void UseGeometryShader(GLuint program) {
-        current_state.geometry = program;
+        current_state.geometry_shader = program;
     }
 
     void UseFragmentShader(GLuint program) {
-        current_state.fragment = program;
+        current_state.fragment_shader = program;
     }
 
 private:
     struct PipelineState {
-        GLuint vertex = 0;
-        GLuint geometry = 0;
-        GLuint fragment = 0;
+        bool operator==(const PipelineState& rhs) const noexcept {
+            return vertex_shader == rhs.vertex_shader && fragment_shader == rhs.fragment_shader &&
+                   geometry_shader == rhs.geometry_shader;
+        }
+
+        bool operator!=(const PipelineState& rhs) const noexcept {
+            return !operator==(rhs);
+        }
+
+        GLuint vertex_shader = 0;
+        GLuint fragment_shader = 0;
+        GLuint geometry_shader = 0;
     };
 
-    /// Update NV_gpu_program5 programs.
-    void UpdateAssemblyPrograms();
-
-    /// Update GLSL programs.
-    void UpdateSourcePrograms();
-
     OGLPipeline graphics_pipeline;
-
+    OGLPipeline compute_pipeline;
     PipelineState current_state;
     PipelineState old_state;
-
-    bool use_assembly_programs = false;
-
     bool is_graphics_bound = true;
-
-    bool vertex_enabled = false;
-    bool geometry_enabled = false;
-    bool fragment_enabled = false;
 };
 
-} // namespace OpenGL
+} // namespace OpenGL::GLShader

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -35,7 +35,7 @@ MICROPROFILE_DEFINE(OpenGL_Texture_Buffer_Copy, "OpenGL", "Texture Buffer Copy",
 namespace {
 
 struct FormatTuple {
-    GLenum internal_format;
+    GLint internal_format;
     GLenum format = GL_NONE;
     GLenum type = GL_NONE;
 };
@@ -238,12 +238,6 @@ OGLTexture CreateTexture(const SurfaceParams& params, GLenum target, GLenum inte
     return texture;
 }
 
-constexpr u32 EncodeSwizzle(SwizzleSource x_source, SwizzleSource y_source, SwizzleSource z_source,
-                            SwizzleSource w_source) {
-    return (static_cast<u32>(x_source) << 24) | (static_cast<u32>(y_source) << 16) |
-           (static_cast<u32>(z_source) << 8) | static_cast<u32>(w_source);
-}
-
 } // Anonymous namespace
 
 CachedSurface::CachedSurface(const GPUVAddr gpu_addr, const SurfaceParams& params,
@@ -387,7 +381,7 @@ void CachedSurface::DecorateSurfaceName() {
 }
 
 void CachedSurfaceView::DecorateViewName(GPUVAddr gpu_addr, std::string prefix) {
-    LabelGLObject(GL_TEXTURE, main_view.handle, gpu_addr, prefix);
+    LabelGLObject(GL_TEXTURE, texture_view.handle, gpu_addr, prefix);
 }
 
 View CachedSurface::CreateView(const ViewParams& view_key) {
@@ -403,13 +397,14 @@ View CachedSurface::CreateViewInner(const ViewParams& view_key, const bool is_pr
 }
 
 CachedSurfaceView::CachedSurfaceView(CachedSurface& surface, const ViewParams& params,
-                                     bool is_proxy)
-    : VideoCommon::ViewBase(params), surface{surface},
-      format{GetFormatTuple(surface.GetSurfaceParams().pixel_format).internal_format},
-      target{GetTextureTarget(params.target)}, is_proxy{is_proxy} {
+                                     const bool is_proxy)
+    : VideoCommon::ViewBase(params), surface{surface}, is_proxy{is_proxy} {
+    target = GetTextureTarget(params.target);
+    format = GetFormatTuple(surface.GetSurfaceParams().pixel_format).internal_format;
     if (!is_proxy) {
-        main_view = CreateTextureView();
+        texture_view = CreateTextureView();
     }
+    swizzle = EncodeSwizzle(SwizzleSource::R, SwizzleSource::G, SwizzleSource::B, SwizzleSource::A);
 }
 
 CachedSurfaceView::~CachedSurfaceView() = default;
@@ -452,49 +447,27 @@ void CachedSurfaceView::Attach(GLenum attachment, GLenum target) const {
     }
 }
 
-GLuint CachedSurfaceView::GetTexture(SwizzleSource x_source, SwizzleSource y_source,
+void CachedSurfaceView::ApplySwizzle(SwizzleSource x_source, SwizzleSource y_source,
                                      SwizzleSource z_source, SwizzleSource w_source) {
-    if (GetSurfaceParams().IsBuffer()) {
-        return GetTexture();
-    }
-    const u32 new_swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source);
-    if (current_swizzle == new_swizzle) {
-        return current_view;
-    }
-    current_swizzle = new_swizzle;
-
-    const auto [entry, is_cache_miss] = view_cache.try_emplace(new_swizzle);
-    OGLTextureView& view = entry->second;
-    if (!is_cache_miss) {
-        current_view = view.handle;
-        return view.handle;
-    }
-    view = CreateTextureView();
-    current_view = view.handle;
-
-    std::array swizzle{x_source, y_source, z_source, w_source};
-
-    switch (const PixelFormat format = GetSurfaceParams().pixel_format) {
+    u32 new_swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source);
+    if (new_swizzle == swizzle)
+        return;
+    swizzle = new_swizzle;
+    const std::array gl_swizzle = {GetSwizzleSource(x_source), GetSwizzleSource(y_source),
+                                   GetSwizzleSource(z_source), GetSwizzleSource(w_source)};
+    const GLuint handle = GetTexture();
+    const PixelFormat format = surface.GetSurfaceParams().pixel_format;
+    switch (format) {
     case PixelFormat::Z24S8:
     case PixelFormat::Z32FS8:
     case PixelFormat::S8Z24:
-        UNIMPLEMENTED_IF(x_source != SwizzleSource::R && x_source != SwizzleSource::G);
-        glTextureParameteri(view.handle, GL_DEPTH_STENCIL_TEXTURE_MODE,
+        glTextureParameteri(handle, GL_DEPTH_STENCIL_TEXTURE_MODE,
                             GetComponent(format, x_source == SwizzleSource::R));
-
-        // Make sure we sample the first component
-        std::transform(swizzle.begin(), swizzle.end(), swizzle.begin(), [](SwizzleSource value) {
-            return value == SwizzleSource::G ? SwizzleSource::R : value;
-        });
-        [[fallthrough]];
-    default: {
-        const std::array gl_swizzle = {GetSwizzleSource(swizzle[0]), GetSwizzleSource(swizzle[1]),
-                                       GetSwizzleSource(swizzle[2]), GetSwizzleSource(swizzle[3])};
-        glTextureParameteriv(view.handle, GL_TEXTURE_SWIZZLE_RGBA, gl_swizzle.data());
+        break;
+    default:
+        glTextureParameteriv(handle, GL_TEXTURE_SWIZZLE_RGBA, gl_swizzle.data());
         break;
     }
-    }
-    return view.handle;
 }
 
 OGLTextureView CachedSurfaceView::CreateTextureView() const {

src/video_core/renderer_opengl/gl_texture_cache.h

@@ -83,7 +83,7 @@ public:
     /// Attaches this texture view to the current bound GL_DRAW_FRAMEBUFFER
     void Attach(GLenum attachment, GLenum target) const;
 
-    GLuint GetTexture(Tegra::Texture::SwizzleSource x_source,
+    void ApplySwizzle(Tegra::Texture::SwizzleSource x_source,
                       Tegra::Texture::SwizzleSource y_source,
                       Tegra::Texture::SwizzleSource z_source,
                       Tegra::Texture::SwizzleSource w_source);
@@ -98,7 +98,7 @@ public:
         if (is_proxy) {
             return surface.GetTexture();
         }
-        return main_view.handle;
+        return texture_view.handle;
     }
 
     GLenum GetFormat() const {
@@ -110,19 +110,23 @@ public:
     }
 
 private:
+    u32 EncodeSwizzle(Tegra::Texture::SwizzleSource x_source,
+                      Tegra::Texture::SwizzleSource y_source,
+                      Tegra::Texture::SwizzleSource z_source,
+                      Tegra::Texture::SwizzleSource w_source) const {
+        return (static_cast<u32>(x_source) << 24) | (static_cast<u32>(y_source) << 16) |
+               (static_cast<u32>(z_source) << 8) | static_cast<u32>(w_source);
+    }
+
     OGLTextureView CreateTextureView() const;
 
     CachedSurface& surface;
-    const GLenum format;
-    const GLenum target;
-    const bool is_proxy;
+    GLenum target{};
+    GLenum format{};
 
-    std::unordered_map<u32, OGLTextureView> view_cache;
-    OGLTextureView main_view;
-
-    // Use an invalid default so it always fails the comparison test
-    u32 current_swizzle = 0xffffffff;
-    GLuint current_view = 0;
+    OGLTextureView texture_view;
+    u32 swizzle{};
+    bool is_proxy{};
 };
 
 class TextureCacheOpenGL final : public TextureCacheBase {

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -316,7 +316,7 @@ public:
 RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system,
                                Core::Frontend::GraphicsContext& context)
     : RendererBase{emu_window}, emu_window{emu_window}, system{system}, context{context},
-      program_manager{device}, has_debug_tool{HasDebugTool()} {}
+      has_debug_tool{HasDebugTool()} {}
 
 RendererOpenGL::~RendererOpenGL() = default;
 
@@ -468,9 +468,8 @@ void RendererOpenGL::InitOpenGLObjects() {
     vertex_program.Create(true, false, vertex_shader.handle);
     fragment_program.Create(true, false, fragment_shader.handle);
 
-    pipeline.Create();
-    glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, vertex_program.handle);
-    glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fragment_program.handle);
+    // Create program pipeline
+    program_manager.Create();
 
     // Generate VBO handle for drawing
     vertex_buffer.Create();
@@ -509,7 +508,7 @@ void RendererOpenGL::CreateRasterizer() {
     if (rasterizer) {
         return;
     }
-    rasterizer = std::make_unique<RasterizerOpenGL>(system, emu_window, device, screen_info,
+    rasterizer = std::make_unique<RasterizerOpenGL>(system, emu_window, screen_info,
                                                     program_manager, state_tracker);
 }
 
@@ -621,7 +620,10 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
     state_tracker.NotifyClipControl();
     state_tracker.NotifyAlphaTest();
 
-    program_manager.BindHostPipeline(pipeline.handle);
+    program_manager.UseVertexShader(vertex_program.handle);
+    program_manager.UseGeometryShader(0);
+    program_manager.UseFragmentShader(fragment_program.handle);
+    program_manager.BindGraphicsPipeline();
 
     glEnable(GL_CULL_FACE);
     if (screen_info.display_srgb) {
@@ -663,8 +665,6 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
 
     glClear(GL_COLOR_BUFFER_BIT);
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
-
-    program_manager.RestoreGuestPipeline();
 }
 
 bool RendererOpenGL::TryPresent(int timeout_ms) {

src/video_core/renderer_opengl/renderer_opengl.h

@@ -9,7 +9,6 @@
 #include "common/common_types.h"
 #include "common/math_util.h"
 #include "video_core/renderer_base.h"
-#include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state_tracker.h"
@@ -96,7 +95,6 @@ private:
     Core::Frontend::EmuWindow& emu_window;
     Core::System& system;
     Core::Frontend::GraphicsContext& context;
-    const Device device;
 
     StateTracker state_tracker{system};
 
@@ -104,14 +102,13 @@ private:
     OGLBuffer vertex_buffer;
     OGLProgram vertex_program;
     OGLProgram fragment_program;
-    OGLPipeline pipeline;
     OGLFramebuffer screenshot_framebuffer;
 
     /// Display information for Switch screen
     ScreenInfo screen_info;
 
     /// Global dummy shader pipeline
-    ProgramManager program_manager;
+    GLShader::ProgramManager program_manager;
 
     /// OpenGL framebuffer data
     std::vector<u8> gl_framebuffer_data;

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -142,7 +142,7 @@ struct FormatTuple {
     {VK_FORMAT_BC6H_UFLOAT_BLOCK},                              // BC6H_UF16
     {VK_FORMAT_BC6H_SFLOAT_BLOCK},                              // BC6H_SF16
     {VK_FORMAT_ASTC_4x4_UNORM_BLOCK},                           // ASTC_2D_4X4
-    {VK_FORMAT_B8G8R8A8_UNORM, Attachable},                     // BGRA8
+    {VK_FORMAT_B8G8R8A8_UNORM},                                 // BGRA8
     {VK_FORMAT_R32G32B32A32_SFLOAT, Attachable | Storage},      // RGBA32F
     {VK_FORMAT_R32G32_SFLOAT, Attachable | Storage},            // RG32F
     {VK_FORMAT_R32_SFLOAT, Attachable | Storage},               // R32F
@@ -168,7 +168,7 @@ struct FormatTuple {
     {VK_FORMAT_ASTC_8x8_UNORM_BLOCK},                           // ASTC_2D_8X8
     {VK_FORMAT_UNDEFINED},                                      // ASTC_2D_8X5
     {VK_FORMAT_UNDEFINED},                                      // ASTC_2D_5X4
-    {VK_FORMAT_B8G8R8A8_SRGB, Attachable},                      // BGRA8_SRGB
+    {VK_FORMAT_UNDEFINED},                                      // BGRA8_SRGB
     {VK_FORMAT_BC1_RGBA_SRGB_BLOCK},                            // DXT1_SRGB
     {VK_FORMAT_BC2_SRGB_BLOCK},                                 // DXT23_SRGB
     {VK_FORMAT_BC3_SRGB_BLOCK},                                 // DXT45_SRGB

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -7,7 +7,6 @@
 #include <memory>
 
 #include "core/core.h"
-#include "video_core/buffer_cache/buffer_cache.h"
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"
 #include "video_core/renderer_vulkan/vk_device.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"

src/video_core/renderer_vulkan/vk_device.cpp

@@ -104,7 +104,6 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(
                                         VK_FORMAT_R16_SFLOAT,
                                         VK_FORMAT_R16G16B16A16_SFLOAT,
                                         VK_FORMAT_B8G8R8A8_UNORM,
-                                        VK_FORMAT_B8G8R8A8_SRGB,
                                         VK_FORMAT_R4G4B4A4_UNORM_PACK16,
                                         VK_FORMAT_D32_SFLOAT,
                                         VK_FORMAT_D16_UNORM,

src/video_core/renderer_vulkan/vk_fence_manager.h

@@ -7,7 +7,6 @@
 #include <memory>
 
 #include "video_core/fence_manager.h"
-#include "video_core/renderer_vulkan/vk_buffer_cache.h"
 #include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Core {

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -329,7 +329,8 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
 
         const GPUVAddr gpu_addr = GetShaderAddress(system, program_enum);
         const auto cpu_addr = memory_manager.GpuToCpuAddress(gpu_addr);
-        const auto shader = cpu_addr ? TryGet(*cpu_addr) : null_shader;
+        ASSERT(cpu_addr);
+        const auto shader = TryGet(*cpu_addr);
         ASSERT(shader);
 
         const std::size_t stage = index == 0 ? 0 : index - 1; // Stage indices are 0 - 5

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -532,14 +532,14 @@ void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
         return;
     }
     texture_cache.OnCPUWrite(addr, size);
-    pipeline_cache.OnCPUWrite(addr, size);
+    pipeline_cache.InvalidateRegion(addr, size);
     buffer_cache.OnCPUWrite(addr, size);
+    query_cache.InvalidateRegion(addr, size);
 }
 
 void RasterizerVulkan::SyncGuestHost() {
     texture_cache.SyncGuestHost();
     buffer_cache.SyncGuestHost();
-    pipeline_cache.SyncGuestHost();
 }
 
 void RasterizerVulkan::SignalSemaphore(GPUVAddr addr, u32 value) {

src/video_core/renderer_vulkan/vk_shader_decompiler.cpp

@@ -515,16 +515,6 @@ private:
     void DeclareCommon() {
         thread_id =
             DeclareInputBuiltIn(spv::BuiltIn::SubgroupLocalInvocationId, t_in_uint, "thread_id");
-        thread_masks[0] =
-            DeclareInputBuiltIn(spv::BuiltIn::SubgroupEqMask, t_in_uint4, "thread_eq_mask");
-        thread_masks[1] =
-            DeclareInputBuiltIn(spv::BuiltIn::SubgroupGeMask, t_in_uint4, "thread_ge_mask");
-        thread_masks[2] =
-            DeclareInputBuiltIn(spv::BuiltIn::SubgroupGtMask, t_in_uint4, "thread_gt_mask");
-        thread_masks[3] =
-            DeclareInputBuiltIn(spv::BuiltIn::SubgroupLeMask, t_in_uint4, "thread_le_mask");
-        thread_masks[4] =
-            DeclareInputBuiltIn(spv::BuiltIn::SubgroupLtMask, t_in_uint4, "thread_lt_mask");
     }
 
     void DeclareVertex() {
@@ -1081,7 +1071,8 @@ private:
 
     void VisitBasicBlock(const NodeBlock& bb) {
         for (const auto& node : bb) {
-            Visit(node);
+            [[maybe_unused]] const Type type = Visit(node).type;
+            ASSERT(type == Type::Void);
         }
     }
 
@@ -1371,9 +1362,7 @@ private:
         Expression target{};
         if (const auto gpr = std::get_if<GprNode>(&*dest)) {
             if (gpr->GetIndex() == Register::ZeroIndex) {
-                // Writing to Register::ZeroIndex is a no op but we still have to visit its source
-                // because it might have side effects.
-                Visit(src);
+                // Writing to Register::ZeroIndex is a no op
                 return {};
             }
             target = {registers.at(gpr->GetIndex()), Type::Float};
@@ -2186,37 +2175,14 @@ private:
         return {OpLoad(t_uint, thread_id), Type::Uint};
     }
 
-    template <std::size_t index>
-    Expression ThreadMask(Operation) {
-        // TODO(Rodrigo): Handle devices with different warp sizes
-        const Id mask = thread_masks[index];
-        return {OpLoad(t_uint, AccessElement(t_in_uint, mask, 0)), Type::Uint};
-    }
-
     Expression ShuffleIndexed(Operation operation) {
         const Id value = AsFloat(Visit(operation[0]));
         const Id index = AsUint(Visit(operation[1]));
         return {OpSubgroupReadInvocationKHR(t_float, value, index), Type::Float};
     }
 
-    Expression Barrier(Operation) {
-        if (!ir.IsDecompiled()) {
-            LOG_ERROR(Render_Vulkan, "OpBarrier used by shader is not decompiled");
-            return {};
-        }
-
-        const auto scope = spv::Scope::Workgroup;
-        const auto memory = spv::Scope::Workgroup;
-        const auto semantics =
-            spv::MemorySemanticsMask::WorkgroupMemory | spv::MemorySemanticsMask::AcquireRelease;
-        OpControlBarrier(Constant(t_uint, static_cast<u32>(scope)),
-                         Constant(t_uint, static_cast<u32>(memory)),
-                         Constant(t_uint, static_cast<u32>(semantics)));
-        return {};
-    }
-
-    template <spv::Scope scope>
-    Expression MemoryBarrier(Operation) {
+    Expression MemoryBarrierGL(Operation) {
+        const auto scope = spv::Scope::Device;
         const auto semantics =
             spv::MemorySemanticsMask::AcquireRelease | spv::MemorySemanticsMask::UniformMemory |
             spv::MemorySemanticsMask::WorkgroupMemory |
@@ -2673,16 +2639,9 @@ private:
         &SPIRVDecompiler::Vote<&Module::OpSubgroupAllEqualKHR>,
 
         &SPIRVDecompiler::ThreadId,
-        &SPIRVDecompiler::ThreadMask<0>, // Eq
-        &SPIRVDecompiler::ThreadMask<1>, // Ge
-        &SPIRVDecompiler::ThreadMask<2>, // Gt
-        &SPIRVDecompiler::ThreadMask<3>, // Le
-        &SPIRVDecompiler::ThreadMask<4>, // Lt
         &SPIRVDecompiler::ShuffleIndexed,
 
-        &SPIRVDecompiler::Barrier,
-        &SPIRVDecompiler::MemoryBarrier<spv::Scope::Workgroup>,
-        &SPIRVDecompiler::MemoryBarrier<spv::Scope::Device>,
+        &SPIRVDecompiler::MemoryBarrierGL,
     };
     static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
 
@@ -2804,7 +2763,6 @@ private:
     Id workgroup_id{};
     Id local_invocation_id{};
     Id thread_id{};
-    std::array<Id, 5> thread_masks{}; // eq, ge, gt, le, lt
 
     VertexIndices in_indices;
     VertexIndices out_indices;

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -354,23 +354,26 @@ CachedSurfaceView::~CachedSurfaceView() = default;
 
 VkImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource y_source,
                                          SwizzleSource z_source, SwizzleSource w_source) {
-    const u32 new_swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source);
-    if (last_image_view && last_swizzle == new_swizzle) {
+    const u32 swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source);
+    if (last_image_view && last_swizzle == swizzle) {
         return last_image_view;
     }
-    last_swizzle = new_swizzle;
+    last_swizzle = swizzle;
 
-    const auto [entry, is_cache_miss] = view_cache.try_emplace(new_swizzle);
+    const auto [entry, is_cache_miss] = view_cache.try_emplace(swizzle);
     auto& image_view = entry->second;
     if (!is_cache_miss) {
         return last_image_view = *image_view;
     }
 
-    std::array swizzle{MaxwellToVK::SwizzleSource(x_source), MaxwellToVK::SwizzleSource(y_source),
-                       MaxwellToVK::SwizzleSource(z_source), MaxwellToVK::SwizzleSource(w_source)};
+    auto swizzle_x = MaxwellToVK::SwizzleSource(x_source);
+    auto swizzle_y = MaxwellToVK::SwizzleSource(y_source);
+    auto swizzle_z = MaxwellToVK::SwizzleSource(z_source);
+    auto swizzle_w = MaxwellToVK::SwizzleSource(w_source);
+
     if (params.pixel_format == VideoCore::Surface::PixelFormat::A1B5G5R5U) {
         // A1B5G5R5 is implemented as A1R5G5B5, we have to change the swizzle here.
-        std::swap(swizzle[0], swizzle[2]);
+        std::swap(swizzle_x, swizzle_z);
     }
 
     // Games can sample depth or stencil values on textures. This is decided by the swizzle value on
@@ -392,11 +395,11 @@ VkImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource y
             UNIMPLEMENTED();
         }
 
-        // Make sure we sample the first component
-        std::transform(
-            swizzle.begin(), swizzle.end(), swizzle.begin(), [](VkComponentSwizzle component) {
-                return component == VK_COMPONENT_SWIZZLE_G ? VK_COMPONENT_SWIZZLE_R : component;
-            });
+        // Vulkan doesn't seem to understand swizzling of a depth stencil image, use identity
+        swizzle_x = VK_COMPONENT_SWIZZLE_R;
+        swizzle_y = VK_COMPONENT_SWIZZLE_G;
+        swizzle_z = VK_COMPONENT_SWIZZLE_B;
+        swizzle_w = VK_COMPONENT_SWIZZLE_A;
     }
 
     VkImageViewCreateInfo ci;
@@ -406,7 +409,7 @@ VkImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource y
     ci.image = surface.GetImageHandle();
     ci.viewType = image_view_type;
     ci.format = surface.GetImage().GetFormat();
-    ci.components = {swizzle[0], swizzle[1], swizzle[2], swizzle[3]};
+    ci.components = {swizzle_x, swizzle_y, swizzle_z, swizzle_w};
     ci.subresourceRange.aspectMask = aspect;
     ci.subresourceRange.baseMipLevel = base_level;
     ci.subresourceRange.levelCount = num_levels;

src/video_core/shader/decode/memory.cpp

@@ -387,6 +387,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
     }
     case OpCode::Id::RED: {
         UNIMPLEMENTED_IF_MSG(instr.red.type != GlobalAtomicType::U32);
+        UNIMPLEMENTED_IF_MSG(instr.red.operation != AtomicOp::Add);
         const auto [real_address, base_address, descriptor] =
             TrackGlobalMemory(bb, instr, true, true);
         if (!real_address || !base_address) {
@@ -395,7 +396,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         }
         Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor);
         Node value = GetRegister(instr.gpr0);
-        bb.push_back(Operation(GetAtomOperation(instr.red.operation), move(gmem), move(value)));
+        bb.push_back(Operation(OperationCode::ReduceIAdd, move(gmem), move(value)));
         break;
     }
     case OpCode::Id::ATOM: {

src/video_core/shader/decode/other.cpp

@@ -109,27 +109,6 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
                 return Operation(OperationCode::WorkGroupIdY);
             case SystemVariable::CtaIdZ:
                 return Operation(OperationCode::WorkGroupIdZ);
-            case SystemVariable::EqMask:
-            case SystemVariable::LtMask:
-            case SystemVariable::LeMask:
-            case SystemVariable::GtMask:
-            case SystemVariable::GeMask:
-                uses_warps = true;
-                switch (instr.sys20) {
-                case SystemVariable::EqMask:
-                    return Operation(OperationCode::ThreadEqMask);
-                case SystemVariable::LtMask:
-                    return Operation(OperationCode::ThreadLtMask);
-                case SystemVariable::LeMask:
-                    return Operation(OperationCode::ThreadLeMask);
-                case SystemVariable::GtMask:
-                    return Operation(OperationCode::ThreadGtMask);
-                case SystemVariable::GeMask:
-                    return Operation(OperationCode::ThreadGeMask);
-                default:
-                    UNREACHABLE();
-                    return Immediate(0u);
-                }
             default:
                 UNIMPLEMENTED_MSG("Unhandled system move: {}",
                                   static_cast<u32>(instr.sys20.Value()));
@@ -293,25 +272,10 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
         SetRegister(bb, instr.gpr0, GetRegister(instr.gpr8));
         break;
     }
-    case OpCode::Id::BAR: {
-        UNIMPLEMENTED_IF_MSG(instr.value != 0xF0A81B8000070000ULL, "BAR is not BAR.SYNC 0x0");
-        bb.push_back(Operation(OperationCode::Barrier));
-        break;
-    }
     case OpCode::Id::MEMBAR: {
+        UNIMPLEMENTED_IF(instr.membar.type != Tegra::Shader::MembarType::GL);
         UNIMPLEMENTED_IF(instr.membar.unknown != Tegra::Shader::MembarUnknown::Default);
-        const OperationCode type = [instr] {
-            switch (instr.membar.type) {
-            case Tegra::Shader::MembarType::CTA:
-                return OperationCode::MemoryBarrierGroup;
-            case Tegra::Shader::MembarType::GL:
-                return OperationCode::MemoryBarrierGlobal;
-            default:
-                UNIMPLEMENTED_MSG("MEMBAR type={}", static_cast<int>(instr.membar.type.Value()));
-                return OperationCode::MemoryBarrierGlobal;
-            }
-        }();
-        bb.push_back(Operation(type));
+        bb.push_back(Operation(OperationCode::MemoryBarrierGL));
         break;
     }
     case OpCode::Id::DEPBAR: {

src/video_core/shader/node.h

@@ -226,16 +226,9 @@ enum class OperationCode {
     VoteEqual,    /// (bool) -> bool
 
     ThreadId,       /// () -> uint
-    ThreadEqMask,   /// () -> uint
-    ThreadGeMask,   /// () -> uint
-    ThreadGtMask,   /// () -> uint
-    ThreadLeMask,   /// () -> uint
-    ThreadLtMask,   /// () -> uint
     ShuffleIndexed, /// (uint value, uint index) -> uint
 
-    Barrier,             /// () -> void
-    MemoryBarrierGroup,  /// () -> void
-    MemoryBarrierGlobal, /// () -> void
+    MemoryBarrierGL, /// () -> void
 
     Amount,
 };

src/video_core/texture_cache/texture_cache.h

@@ -991,9 +991,7 @@ private:
         params.target = target;
         params.is_tiled = false;
         params.srgb_conversion = false;
-        params.is_layered =
-            target == SurfaceTarget::Texture1DArray || target == SurfaceTarget::Texture2DArray ||
-            target == SurfaceTarget::TextureCubemap || target == SurfaceTarget::TextureCubeArray;
+        params.is_layered = false;
         params.block_width = 0;
         params.block_height = 0;
         params.block_depth = 0;

src/yuzu/configuration/config.cpp

@@ -643,8 +643,6 @@ void Config::ReadRendererValues() {
     Settings::values.use_asynchronous_gpu_emulation =
         ReadSetting(QStringLiteral("use_asynchronous_gpu_emulation"), false).toBool();
     Settings::values.use_vsync = ReadSetting(QStringLiteral("use_vsync"), true).toBool();
-    Settings::values.use_assembly_shaders =
-        ReadSetting(QStringLiteral("use_assembly_shaders"), false).toBool();
     Settings::values.use_fast_gpu_time =
         ReadSetting(QStringLiteral("use_fast_gpu_time"), true).toBool();
     Settings::values.force_30fps_mode =
@@ -1092,8 +1090,6 @@ void Config::SaveRendererValues() {
     WriteSetting(QStringLiteral("use_asynchronous_gpu_emulation"),
                  Settings::values.use_asynchronous_gpu_emulation, false);
     WriteSetting(QStringLiteral("use_vsync"), Settings::values.use_vsync, true);
-    WriteSetting(QStringLiteral("use_assembly_shaders"), Settings::values.use_assembly_shaders,
-                 false);
     WriteSetting(QStringLiteral("use_fast_gpu_time"), Settings::values.use_fast_gpu_time, true);
     WriteSetting(QStringLiteral("force_30fps_mode"), Settings::values.force_30fps_mode, false);
 

src/yuzu/configuration/configure_graphics_advanced.cpp

@@ -12,9 +12,6 @@ ConfigureGraphicsAdvanced::ConfigureGraphicsAdvanced(QWidget* parent)
 
     ui->setupUi(this);
 
-    // TODO: Remove this after assembly shaders are fully integrated
-    ui->use_assembly_shaders->setVisible(false);
-
     SetConfiguration();
 }
 
@@ -25,8 +22,6 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
     ui->gpu_accuracy->setCurrentIndex(static_cast<int>(Settings::values.gpu_accuracy));
     ui->use_vsync->setEnabled(runtime_lock);
     ui->use_vsync->setChecked(Settings::values.use_vsync);
-    ui->use_assembly_shaders->setEnabled(runtime_lock);
-    ui->use_assembly_shaders->setChecked(Settings::values.use_assembly_shaders);
     ui->use_fast_gpu_time->setChecked(Settings::values.use_fast_gpu_time);
     ui->force_30fps_mode->setEnabled(runtime_lock);
     ui->force_30fps_mode->setChecked(Settings::values.force_30fps_mode);
@@ -38,7 +33,6 @@ void ConfigureGraphicsAdvanced::ApplyConfiguration() {
     auto gpu_accuracy = static_cast<Settings::GPUAccuracy>(ui->gpu_accuracy->currentIndex());
     Settings::values.gpu_accuracy = gpu_accuracy;
     Settings::values.use_vsync = ui->use_vsync->isChecked();
-    Settings::values.use_assembly_shaders = ui->use_assembly_shaders->isChecked();
     Settings::values.use_fast_gpu_time = ui->use_fast_gpu_time->isChecked();
     Settings::values.force_30fps_mode = ui->force_30fps_mode->isChecked();
     Settings::values.max_anisotropy = ui->anisotropic_filtering_combobox->currentIndex();

src/yuzu/configuration/configure_graphics_advanced.ui

@@ -62,16 +62,6 @@
           </property>
          </widget>
         </item>
-        <item>
-         <widget class="QCheckBox" name="use_assembly_shaders">
-          <property name="toolTip">
-           <string>Enabling this reduces shader stutter. Enables OpenGL assembly shaders on supported Nvidia devices (NV_gpu_program5 is required). This feature is experimental.</string>
-          </property>
-          <property name="text">
-           <string>Use assembly shaders (experimental, Nvidia OpenGL only)</string>
-          </property>
-         </widget>
-        </item>
         <item>
          <widget class="QCheckBox" name="force_30fps_mode">
           <property name="text">

src/yuzu/discord_impl.cpp

@@ -18,7 +18,7 @@ DiscordImpl::DiscordImpl() {
 
     // The number is the client ID for yuzu, it's used for images and the
     // application name
-    Discord_Initialize("712465656758665259", &handlers, 1, nullptr);
+    Discord_Initialize("471872241299226636", &handlers, 1, nullptr);
 }
 
 DiscordImpl::~DiscordImpl() {

src/yuzu/main.cpp

@@ -65,7 +65,6 @@ static FileSys::VirtualFile VfsDirectoryCreateFileWrapper(const FileSys::Virtual
 #include "common/logging/backend.h"
 #include "common/logging/filter.h"
 #include "common/logging/log.h"
-#include "common/memory_detect.h"
 #include "common/microprofile.h"
 #include "common/scm_rev.h"
 #include "common/scope_exit.h"
@@ -136,6 +135,28 @@ __declspec(dllexport) int AmdPowerXpressRequestHighPerformance = 1;
 }
 #endif
 
+namespace {
+
+QString GetAccountUsername() {
+    const QString nouser = QString::fromStdString("No User");
+    Service::Account::ProfileManager manager;
+    const auto current_user = manager.GetUser(Settings::values.current_user);
+    if (!current_user.has_value() || (current_user == Common::UUID{})) {
+        return nouser;
+    }
+    Service::Account::ProfileBase profile;
+    if (!manager.GetProfileBase(*current_user, profile)) {
+        return nouser;
+    }
+
+    const auto text = Common::StringFromFixedZeroTerminatedBuffer(
+        reinterpret_cast<const char*>(profile.username.data()), profile.username.size());
+
+    return text.empty() ? nouser : QString::fromStdString(text);
+}
+
+} // Anonymous namespace
+
 constexpr int default_mouse_timeout = 2500;
 
 constexpr u64 DLC_BASE_TITLE_ID_MASK = 0xFFFFFFFFFFFFE000;
@@ -220,10 +241,6 @@ GMainWindow::GMainWindow()
     LOG_INFO(Frontend, "Host CPU: {}", Common::GetCPUCaps().cpu_string);
 #endif
     LOG_INFO(Frontend, "Host OS: {}", QSysInfo::prettyProductName().toStdString());
-    LOG_INFO(Frontend, "Host RAM: {:.2f} GB",
-             Common::GetMemInfo().TotalPhysicalMemory / 1024.0f / 1024 / 1024);
-    LOG_INFO(Frontend, "Host Swap: {:.2f} GB",
-             Common::GetMemInfo().TotalSwapMemory / 1024.0f / 1024 / 1024);
     UpdateWindowTitle();
 
     show();
@@ -498,6 +515,48 @@ void GMainWindow::InitializeWidgets() {
         statusBar()->addPermanentWidget(label);
     }
 
+    // Setup Profile button
+    profile_status_button = new QPushButton();
+    profile_status_button->setObjectName(QStringLiteral("TogglableStatusBarButton"));
+    profile_status_button->setCheckable(true);
+    profile_status_button->setChecked(true);
+    profile_status_button->setFocusPolicy(Qt::NoFocus);
+    const auto username = GetAccountUsername();
+    profile_status_button->setText(username);
+    connect(profile_status_button, &QPushButton::clicked, [=] {
+        profile_status_button->setChecked(true);
+
+        if (emulation_running) {
+            return;
+        }
+
+        // User save data
+        const auto select_profile = [this] {
+            QtProfileSelectionDialog dialog(this);
+            dialog.setWindowFlags(Qt::Dialog | Qt::CustomizeWindowHint | Qt::WindowTitleHint |
+                                  Qt::WindowSystemMenuHint | Qt::WindowCloseButtonHint);
+            dialog.setWindowModality(Qt::WindowModal);
+
+            if (dialog.exec() == QDialog::Rejected) {
+                return -1;
+            }
+
+            return dialog.GetIndex();
+        };
+
+        const auto index = select_profile();
+        if (index == -1) {
+            return;
+        }
+
+        Settings::values.current_user = index;
+        Settings::Apply();
+
+        const auto username = GetAccountUsername();
+        profile_status_button->setText(username);
+    });
+    statusBar()->insertPermanentWidget(0, profile_status_button);
+
     // Setup Dock button
     dock_status_button = new QPushButton();
     dock_status_button->setObjectName(QStringLiteral("TogglableStatusBarButton"));
@@ -1907,6 +1966,8 @@ void GMainWindow::OnConfigure() {
         ui.centralwidget->setMouseTracking(false);
     }
 
+    const auto username = GetAccountUsername();
+    profile_status_button->setText(username);
     dock_status_button->setChecked(Settings::values.use_docked_mode);
     async_status_button->setChecked(Settings::values.use_asynchronous_gpu_emulation);
 #ifdef HAS_VULKAN

src/yuzu/main.h

@@ -233,6 +233,7 @@ private:
     QLabel* emu_speed_label = nullptr;
     QLabel* game_fps_label = nullptr;
     QLabel* emu_frametime_label = nullptr;
+    QPushButton* profile_status_button = nullptr;
     QPushButton* async_status_button = nullptr;
     QPushButton* renderer_status_button = nullptr;
     QPushButton* dock_status_button = nullptr;

src/yuzu_cmd/config.cpp

@@ -397,8 +397,6 @@ void Config::ReadValues() {
         sdl2_config->GetBoolean("Renderer", "use_asynchronous_gpu_emulation", false);
     Settings::values.use_vsync =
         static_cast<u16>(sdl2_config->GetInteger("Renderer", "use_vsync", 1));
-    Settings::values.use_assembly_shaders =
-        sdl2_config->GetBoolean("Renderer", "use_assembly_shaders", false);
     Settings::values.use_fast_gpu_time =
         sdl2_config->GetBoolean("Renderer", "use_fast_gpu_time", true);
 

src/yuzu_cmd/default_ini.h

@@ -134,10 +134,6 @@ max_anisotropy =
 # 0 (default): Off, 1: On
 use_vsync =
 
-# Whether to use OpenGL assembly shaders or not. NV_gpu_program5 is required.
-# 0 (default): Off, 1: On
-use_assembly_shaders =
-
 # Turns on the frame limiter, which will limit frames output to the target game speed
 # 0: Off, 1: On (default)
 use_frame_limit =