Clang format and other fixes

Implement flushing in the rasterizer cache

src/core/file_sys/content_archive.cpp

@@ -97,288 +97,11 @@ union NCASectionHeader {
 };
 static_assert(sizeof(NCASectionHeader) == 0x200, "NCASectionHeader has incorrect size.");
 
-static bool IsValidNCA(const NCAHeader& header) {
+bool IsValidNCA(const NCAHeader& header) {
     // TODO(DarkLordZach): Add NCA2/NCA0 support.
     return header.magic == Common::MakeMagic('N', 'C', 'A', '3');
 }
 
-NCA::NCA(VirtualFile file_, VirtualFile bktr_base_romfs_, u64 bktr_base_ivfc_offset)
-    : file(std::move(file_)), bktr_base_romfs(std::move(bktr_base_romfs_)) {
-    if (file == nullptr) {
-        status = Loader::ResultStatus::ErrorNullFile;
-        return;
-    }
-
-    if (sizeof(NCAHeader) != file->ReadObject(&header)) {
-        LOG_ERROR(Loader, "File reader errored out during header read.");
-        status = Loader::ResultStatus::ErrorBadNCAHeader;
-        return;
-    }
-
-    if (!HandlePotentialHeaderDecryption()) {
-        return;
-    }
-
-    has_rights_id = std::any_of(header.rights_id.begin(), header.rights_id.end(),
-                                [](char c) { return c != '\0'; });
-
-    const std::vector<NCASectionHeader> sections = ReadSectionHeaders();
-    is_update = std::any_of(sections.begin(), sections.end(), [](const NCASectionHeader& header) {
-        return header.raw.header.crypto_type == NCASectionCryptoType::BKTR;
-    });
-
-    if (!ReadSections(sections, bktr_base_ivfc_offset)) {
-        return;
-    }
-
-    status = Loader::ResultStatus::Success;
-}
-
-NCA::~NCA() = default;
-
-bool NCA::CheckSupportedNCA(const NCAHeader& nca_header) {
-    if (nca_header.magic == Common::MakeMagic('N', 'C', 'A', '2')) {
-        status = Loader::ResultStatus::ErrorNCA2;
-        return false;
-    }
-
-    if (nca_header.magic == Common::MakeMagic('N', 'C', 'A', '0')) {
-        status = Loader::ResultStatus::ErrorNCA0;
-        return false;
-    }
-
-    return true;
-}
-
-bool NCA::HandlePotentialHeaderDecryption() {
-    if (IsValidNCA(header)) {
-        return true;
-    }
-
-    if (!CheckSupportedNCA(header)) {
-        return false;
-    }
-
-    NCAHeader dec_header{};
-    Core::Crypto::AESCipher<Core::Crypto::Key256> cipher(
-        keys.GetKey(Core::Crypto::S256KeyType::Header), Core::Crypto::Mode::XTS);
-    cipher.XTSTranscode(&header, sizeof(NCAHeader), &dec_header, 0, 0x200,
-                        Core::Crypto::Op::Decrypt);
-    if (IsValidNCA(dec_header)) {
-        header = dec_header;
-        encrypted = true;
-    } else {
-        if (!CheckSupportedNCA(dec_header)) {
-            return false;
-        }
-
-        if (keys.HasKey(Core::Crypto::S256KeyType::Header)) {
-            status = Loader::ResultStatus::ErrorIncorrectHeaderKey;
-        } else {
-            status = Loader::ResultStatus::ErrorMissingHeaderKey;
-        }
-        return false;
-    }
-
-    return true;
-}
-
-std::vector<NCASectionHeader> NCA::ReadSectionHeaders() const {
-    const std::ptrdiff_t number_sections =
-        std::count_if(std::begin(header.section_tables), std::end(header.section_tables),
-                      [](NCASectionTableEntry entry) { return entry.media_offset > 0; });
-
-    std::vector<NCASectionHeader> sections(number_sections);
-    const auto length_sections = SECTION_HEADER_SIZE * number_sections;
-
-    if (encrypted) {
-        auto raw = file->ReadBytes(length_sections, SECTION_HEADER_OFFSET);
-        Core::Crypto::AESCipher<Core::Crypto::Key256> cipher(
-            keys.GetKey(Core::Crypto::S256KeyType::Header), Core::Crypto::Mode::XTS);
-        cipher.XTSTranscode(raw.data(), length_sections, sections.data(), 2, SECTION_HEADER_SIZE,
-                            Core::Crypto::Op::Decrypt);
-    } else {
-        file->ReadBytes(sections.data(), length_sections, SECTION_HEADER_OFFSET);
-    }
-
-    return sections;
-}
-
-bool NCA::ReadSections(const std::vector<NCASectionHeader>& sections, u64 bktr_base_ivfc_offset) {
-    for (std::size_t i = 0; i < sections.size(); ++i) {
-        const auto& section = sections[i];
-
-        if (section.raw.header.filesystem_type == NCASectionFilesystemType::ROMFS) {
-            if (!ReadRomFSSection(section, header.section_tables[i], bktr_base_ivfc_offset)) {
-                return false;
-            }
-        } else if (section.raw.header.filesystem_type == NCASectionFilesystemType::PFS0) {
-            if (!ReadPFS0Section(section, header.section_tables[i])) {
-                return false;
-            }
-        }
-    }
-
-    return true;
-}
-
-bool NCA::ReadRomFSSection(const NCASectionHeader& section, const NCASectionTableEntry& entry,
-                           u64 bktr_base_ivfc_offset) {
-    const std::size_t base_offset = entry.media_offset * MEDIA_OFFSET_MULTIPLIER;
-    ivfc_offset = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset;
-    const std::size_t romfs_offset = base_offset + ivfc_offset;
-    const std::size_t romfs_size = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].size;
-    auto raw = std::make_shared<OffsetVfsFile>(file, romfs_size, romfs_offset);
-    auto dec = Decrypt(section, raw, romfs_offset);
-
-    if (dec == nullptr) {
-        if (status != Loader::ResultStatus::Success)
-            return false;
-        if (has_rights_id)
-            status = Loader::ResultStatus::ErrorIncorrectTitlekeyOrTitlekek;
-        else
-            status = Loader::ResultStatus::ErrorIncorrectKeyAreaKey;
-        return false;
-    }
-
-    if (section.raw.header.crypto_type == NCASectionCryptoType::BKTR) {
-        if (section.bktr.relocation.magic != Common::MakeMagic('B', 'K', 'T', 'R') ||
-            section.bktr.subsection.magic != Common::MakeMagic('B', 'K', 'T', 'R')) {
-            status = Loader::ResultStatus::ErrorBadBKTRHeader;
-            return false;
-        }
-
-        if (section.bktr.relocation.offset + section.bktr.relocation.size !=
-            section.bktr.subsection.offset) {
-            status = Loader::ResultStatus::ErrorBKTRSubsectionNotAfterRelocation;
-            return false;
-        }
-
-        const u64 size = MEDIA_OFFSET_MULTIPLIER * (entry.media_end_offset - entry.media_offset);
-        if (section.bktr.subsection.offset + section.bktr.subsection.size != size) {
-            status = Loader::ResultStatus::ErrorBKTRSubsectionNotAtEnd;
-            return false;
-        }
-
-        const u64 offset = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset;
-        RelocationBlock relocation_block{};
-        if (dec->ReadObject(&relocation_block, section.bktr.relocation.offset - offset) !=
-            sizeof(RelocationBlock)) {
-            status = Loader::ResultStatus::ErrorBadRelocationBlock;
-            return false;
-        }
-        SubsectionBlock subsection_block{};
-        if (dec->ReadObject(&subsection_block, section.bktr.subsection.offset - offset) !=
-            sizeof(RelocationBlock)) {
-            status = Loader::ResultStatus::ErrorBadSubsectionBlock;
-            return false;
-        }
-
-        std::vector<RelocationBucketRaw> relocation_buckets_raw(
-            (section.bktr.relocation.size - sizeof(RelocationBlock)) / sizeof(RelocationBucketRaw));
-        if (dec->ReadBytes(relocation_buckets_raw.data(),
-                           section.bktr.relocation.size - sizeof(RelocationBlock),
-                           section.bktr.relocation.offset + sizeof(RelocationBlock) - offset) !=
-            section.bktr.relocation.size - sizeof(RelocationBlock)) {
-            status = Loader::ResultStatus::ErrorBadRelocationBuckets;
-            return false;
-        }
-
-        std::vector<SubsectionBucketRaw> subsection_buckets_raw(
-            (section.bktr.subsection.size - sizeof(SubsectionBlock)) / sizeof(SubsectionBucketRaw));
-        if (dec->ReadBytes(subsection_buckets_raw.data(),
-                           section.bktr.subsection.size - sizeof(SubsectionBlock),
-                           section.bktr.subsection.offset + sizeof(SubsectionBlock) - offset) !=
-            section.bktr.subsection.size - sizeof(SubsectionBlock)) {
-            status = Loader::ResultStatus::ErrorBadSubsectionBuckets;
-            return false;
-        }
-
-        std::vector<RelocationBucket> relocation_buckets(relocation_buckets_raw.size());
-        std::transform(relocation_buckets_raw.begin(), relocation_buckets_raw.end(),
-                       relocation_buckets.begin(), &ConvertRelocationBucketRaw);
-        std::vector<SubsectionBucket> subsection_buckets(subsection_buckets_raw.size());
-        std::transform(subsection_buckets_raw.begin(), subsection_buckets_raw.end(),
-                       subsection_buckets.begin(), &ConvertSubsectionBucketRaw);
-
-        u32 ctr_low;
-        std::memcpy(&ctr_low, section.raw.section_ctr.data(), sizeof(ctr_low));
-        subsection_buckets.back().entries.push_back({section.bktr.relocation.offset, {0}, ctr_low});
-        subsection_buckets.back().entries.push_back({size, {0}, 0});
-
-        boost::optional<Core::Crypto::Key128> key = boost::none;
-        if (encrypted) {
-            if (has_rights_id) {
-                status = Loader::ResultStatus::Success;
-                key = GetTitlekey();
-                if (key == boost::none) {
-                    status = Loader::ResultStatus::ErrorMissingTitlekey;
-                    return false;
-                }
-            } else {
-                key = GetKeyAreaKey(NCASectionCryptoType::BKTR);
-                if (key == boost::none) {
-                    status = Loader::ResultStatus::ErrorMissingKeyAreaKey;
-                    return false;
-                }
-            }
-        }
-
-        if (bktr_base_romfs == nullptr) {
-            status = Loader::ResultStatus::ErrorMissingBKTRBaseRomFS;
-            return false;
-        }
-
-        auto bktr = std::make_shared<BKTR>(
-            bktr_base_romfs, std::make_shared<OffsetVfsFile>(file, romfs_size, base_offset),
-            relocation_block, relocation_buckets, subsection_block, subsection_buckets, encrypted,
-            encrypted ? key.get() : Core::Crypto::Key128{}, base_offset, bktr_base_ivfc_offset,
-            section.raw.section_ctr);
-
-        // BKTR applies to entire IVFC, so make an offset version to level 6
-        files.push_back(std::make_shared<OffsetVfsFile>(
-            bktr, romfs_size, section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset));
-    } else {
-        files.push_back(std::move(dec));
-    }
-
-    romfs = files.back();
-    return true;
-}
-
-bool NCA::ReadPFS0Section(const NCASectionHeader& section, const NCASectionTableEntry& entry) {
-    const u64 offset = (static_cast<u64>(entry.media_offset) * MEDIA_OFFSET_MULTIPLIER) +
-                       section.pfs0.pfs0_header_offset;
-    const u64 size = MEDIA_OFFSET_MULTIPLIER * (entry.media_end_offset - entry.media_offset);
-
-    auto dec = Decrypt(section, std::make_shared<OffsetVfsFile>(file, size, offset), offset);
-    if (dec != nullptr) {
-        auto npfs = std::make_shared<PartitionFilesystem>(std::move(dec));
-
-        if (npfs->GetStatus() == Loader::ResultStatus::Success) {
-            dirs.push_back(std::move(npfs));
-            if (IsDirectoryExeFS(dirs.back()))
-                exefs = dirs.back();
-        } else {
-            if (has_rights_id)
-                status = Loader::ResultStatus::ErrorIncorrectTitlekeyOrTitlekek;
-            else
-                status = Loader::ResultStatus::ErrorIncorrectKeyAreaKey;
-            return false;
-        }
-    } else {
-        if (status != Loader::ResultStatus::Success)
-            return false;
-        if (has_rights_id)
-            status = Loader::ResultStatus::ErrorIncorrectTitlekeyOrTitlekek;
-        else
-            status = Loader::ResultStatus::ErrorIncorrectKeyAreaKey;
-        return false;
-    }
-
-    return true;
-}
-
 u8 NCA::GetCryptoRevision() const {
     u8 master_key_id = header.crypto_type;
     if (header.crypto_type_2 > master_key_id)
@@ -444,7 +167,7 @@ boost::optional<Core::Crypto::Key128> NCA::GetTitlekey() {
     return titlekey;
 }
 
-VirtualFile NCA::Decrypt(const NCASectionHeader& s_header, VirtualFile in, u64 starting_offset) {
+VirtualFile NCA::Decrypt(NCASectionHeader s_header, VirtualFile in, u64 starting_offset) {
     if (!encrypted)
         return in;
 
@@ -492,6 +215,256 @@ VirtualFile NCA::Decrypt(const NCASectionHeader& s_header, VirtualFile in, u64 s
     }
 }
 
+NCA::NCA(VirtualFile file_, VirtualFile bktr_base_romfs_, u64 bktr_base_ivfc_offset)
+    : file(std::move(file_)),
+      bktr_base_romfs(bktr_base_romfs_ ? std::move(bktr_base_romfs_) : nullptr) {
+    status = Loader::ResultStatus::Success;
+
+    if (file == nullptr) {
+        status = Loader::ResultStatus::ErrorNullFile;
+        return;
+    }
+
+    if (sizeof(NCAHeader) != file->ReadObject(&header)) {
+        LOG_ERROR(Loader, "File reader errored out during header read.");
+        status = Loader::ResultStatus::ErrorBadNCAHeader;
+        return;
+    }
+
+    encrypted = false;
+
+    if (!IsValidNCA(header)) {
+        if (header.magic == Common::MakeMagic('N', 'C', 'A', '2')) {
+            status = Loader::ResultStatus::ErrorNCA2;
+            return;
+        }
+        if (header.magic == Common::MakeMagic('N', 'C', 'A', '0')) {
+            status = Loader::ResultStatus::ErrorNCA0;
+            return;
+        }
+
+        NCAHeader dec_header{};
+        Core::Crypto::AESCipher<Core::Crypto::Key256> cipher(
+            keys.GetKey(Core::Crypto::S256KeyType::Header), Core::Crypto::Mode::XTS);
+        cipher.XTSTranscode(&header, sizeof(NCAHeader), &dec_header, 0, 0x200,
+                            Core::Crypto::Op::Decrypt);
+        if (IsValidNCA(dec_header)) {
+            header = dec_header;
+            encrypted = true;
+        } else {
+            if (dec_header.magic == Common::MakeMagic('N', 'C', 'A', '2')) {
+                status = Loader::ResultStatus::ErrorNCA2;
+                return;
+            }
+            if (dec_header.magic == Common::MakeMagic('N', 'C', 'A', '0')) {
+                status = Loader::ResultStatus::ErrorNCA0;
+                return;
+            }
+
+            if (!keys.HasKey(Core::Crypto::S256KeyType::Header))
+                status = Loader::ResultStatus::ErrorMissingHeaderKey;
+            else
+                status = Loader::ResultStatus::ErrorIncorrectHeaderKey;
+            return;
+        }
+    }
+
+    has_rights_id = std::find_if_not(header.rights_id.begin(), header.rights_id.end(),
+                                     [](char c) { return c == '\0'; }) != header.rights_id.end();
+
+    const std::ptrdiff_t number_sections =
+        std::count_if(std::begin(header.section_tables), std::end(header.section_tables),
+                      [](NCASectionTableEntry entry) { return entry.media_offset > 0; });
+
+    std::vector<NCASectionHeader> sections(number_sections);
+    const auto length_sections = SECTION_HEADER_SIZE * number_sections;
+
+    if (encrypted) {
+        auto raw = file->ReadBytes(length_sections, SECTION_HEADER_OFFSET);
+        Core::Crypto::AESCipher<Core::Crypto::Key256> cipher(
+            keys.GetKey(Core::Crypto::S256KeyType::Header), Core::Crypto::Mode::XTS);
+        cipher.XTSTranscode(raw.data(), length_sections, sections.data(), 2, SECTION_HEADER_SIZE,
+                            Core::Crypto::Op::Decrypt);
+    } else {
+        file->ReadBytes(sections.data(), length_sections, SECTION_HEADER_OFFSET);
+    }
+
+    is_update = std::find_if(sections.begin(), sections.end(), [](const NCASectionHeader& header) {
+                    return header.raw.header.crypto_type == NCASectionCryptoType::BKTR;
+                }) != sections.end();
+    ivfc_offset = 0;
+
+    for (std::ptrdiff_t i = 0; i < number_sections; ++i) {
+        auto section = sections[i];
+
+        if (section.raw.header.filesystem_type == NCASectionFilesystemType::ROMFS) {
+            const std::size_t base_offset =
+                header.section_tables[i].media_offset * MEDIA_OFFSET_MULTIPLIER;
+            ivfc_offset = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset;
+            const std::size_t romfs_offset = base_offset + ivfc_offset;
+            const std::size_t romfs_size = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].size;
+            auto raw = std::make_shared<OffsetVfsFile>(file, romfs_size, romfs_offset);
+            auto dec = Decrypt(section, raw, romfs_offset);
+
+            if (dec == nullptr) {
+                if (status != Loader::ResultStatus::Success)
+                    return;
+                if (has_rights_id)
+                    status = Loader::ResultStatus::ErrorIncorrectTitlekeyOrTitlekek;
+                else
+                    status = Loader::ResultStatus::ErrorIncorrectKeyAreaKey;
+                return;
+            }
+
+            if (section.raw.header.crypto_type == NCASectionCryptoType::BKTR) {
+                if (section.bktr.relocation.magic != Common::MakeMagic('B', 'K', 'T', 'R') ||
+                    section.bktr.subsection.magic != Common::MakeMagic('B', 'K', 'T', 'R')) {
+                    status = Loader::ResultStatus::ErrorBadBKTRHeader;
+                    return;
+                }
+
+                if (section.bktr.relocation.offset + section.bktr.relocation.size !=
+                    section.bktr.subsection.offset) {
+                    status = Loader::ResultStatus::ErrorBKTRSubsectionNotAfterRelocation;
+                    return;
+                }
+
+                const u64 size =
+                    MEDIA_OFFSET_MULTIPLIER * (header.section_tables[i].media_end_offset -
+                                               header.section_tables[i].media_offset);
+                if (section.bktr.subsection.offset + section.bktr.subsection.size != size) {
+                    status = Loader::ResultStatus::ErrorBKTRSubsectionNotAtEnd;
+                    return;
+                }
+
+                const u64 offset = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset;
+                RelocationBlock relocation_block{};
+                if (dec->ReadObject(&relocation_block, section.bktr.relocation.offset - offset) !=
+                    sizeof(RelocationBlock)) {
+                    status = Loader::ResultStatus::ErrorBadRelocationBlock;
+                    return;
+                }
+                SubsectionBlock subsection_block{};
+                if (dec->ReadObject(&subsection_block, section.bktr.subsection.offset - offset) !=
+                    sizeof(RelocationBlock)) {
+                    status = Loader::ResultStatus::ErrorBadSubsectionBlock;
+                    return;
+                }
+
+                std::vector<RelocationBucketRaw> relocation_buckets_raw(
+                    (section.bktr.relocation.size - sizeof(RelocationBlock)) /
+                    sizeof(RelocationBucketRaw));
+                if (dec->ReadBytes(relocation_buckets_raw.data(),
+                                   section.bktr.relocation.size - sizeof(RelocationBlock),
+                                   section.bktr.relocation.offset + sizeof(RelocationBlock) -
+                                       offset) !=
+                    section.bktr.relocation.size - sizeof(RelocationBlock)) {
+                    status = Loader::ResultStatus::ErrorBadRelocationBuckets;
+                    return;
+                }
+
+                std::vector<SubsectionBucketRaw> subsection_buckets_raw(
+                    (section.bktr.subsection.size - sizeof(SubsectionBlock)) /
+                    sizeof(SubsectionBucketRaw));
+                if (dec->ReadBytes(subsection_buckets_raw.data(),
+                                   section.bktr.subsection.size - sizeof(SubsectionBlock),
+                                   section.bktr.subsection.offset + sizeof(SubsectionBlock) -
+                                       offset) !=
+                    section.bktr.subsection.size - sizeof(SubsectionBlock)) {
+                    status = Loader::ResultStatus::ErrorBadSubsectionBuckets;
+                    return;
+                }
+
+                std::vector<RelocationBucket> relocation_buckets(relocation_buckets_raw.size());
+                std::transform(relocation_buckets_raw.begin(), relocation_buckets_raw.end(),
+                               relocation_buckets.begin(), &ConvertRelocationBucketRaw);
+                std::vector<SubsectionBucket> subsection_buckets(subsection_buckets_raw.size());
+                std::transform(subsection_buckets_raw.begin(), subsection_buckets_raw.end(),
+                               subsection_buckets.begin(), &ConvertSubsectionBucketRaw);
+
+                u32 ctr_low;
+                std::memcpy(&ctr_low, section.raw.section_ctr.data(), sizeof(ctr_low));
+                subsection_buckets.back().entries.push_back(
+                    {section.bktr.relocation.offset, {0}, ctr_low});
+                subsection_buckets.back().entries.push_back({size, {0}, 0});
+
+                boost::optional<Core::Crypto::Key128> key = boost::none;
+                if (encrypted) {
+                    if (has_rights_id) {
+                        status = Loader::ResultStatus::Success;
+                        key = GetTitlekey();
+                        if (key == boost::none) {
+                            status = Loader::ResultStatus::ErrorMissingTitlekey;
+                            return;
+                        }
+                    } else {
+                        key = GetKeyAreaKey(NCASectionCryptoType::BKTR);
+                        if (key == boost::none) {
+                            status = Loader::ResultStatus::ErrorMissingKeyAreaKey;
+                            return;
+                        }
+                    }
+                }
+
+                if (bktr_base_romfs == nullptr) {
+                    status = Loader::ResultStatus::ErrorMissingBKTRBaseRomFS;
+                    return;
+                }
+
+                auto bktr = std::make_shared<BKTR>(
+                    bktr_base_romfs, std::make_shared<OffsetVfsFile>(file, romfs_size, base_offset),
+                    relocation_block, relocation_buckets, subsection_block, subsection_buckets,
+                    encrypted, encrypted ? key.get() : Core::Crypto::Key128{}, base_offset,
+                    bktr_base_ivfc_offset, section.raw.section_ctr);
+
+                // BKTR applies to entire IVFC, so make an offset version to level 6
+
+                files.push_back(std::make_shared<OffsetVfsFile>(
+                    bktr, romfs_size, section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset));
+                romfs = files.back();
+            } else {
+                files.push_back(std::move(dec));
+                romfs = files.back();
+            }
+        } else if (section.raw.header.filesystem_type == NCASectionFilesystemType::PFS0) {
+            u64 offset = (static_cast<u64>(header.section_tables[i].media_offset) *
+                          MEDIA_OFFSET_MULTIPLIER) +
+                         section.pfs0.pfs0_header_offset;
+            u64 size = MEDIA_OFFSET_MULTIPLIER * (header.section_tables[i].media_end_offset -
+                                                  header.section_tables[i].media_offset);
+            auto dec =
+                Decrypt(section, std::make_shared<OffsetVfsFile>(file, size, offset), offset);
+            if (dec != nullptr) {
+                auto npfs = std::make_shared<PartitionFilesystem>(std::move(dec));
+
+                if (npfs->GetStatus() == Loader::ResultStatus::Success) {
+                    dirs.push_back(std::move(npfs));
+                    if (IsDirectoryExeFS(dirs.back()))
+                        exefs = dirs.back();
+                } else {
+                    if (has_rights_id)
+                        status = Loader::ResultStatus::ErrorIncorrectTitlekeyOrTitlekek;
+                    else
+                        status = Loader::ResultStatus::ErrorIncorrectKeyAreaKey;
+                    return;
+                }
+            } else {
+                if (status != Loader::ResultStatus::Success)
+                    return;
+                if (has_rights_id)
+                    status = Loader::ResultStatus::ErrorIncorrectTitlekeyOrTitlekek;
+                else
+                    status = Loader::ResultStatus::ErrorIncorrectKeyAreaKey;
+                return;
+            }
+        }
+    }
+
+    status = Loader::ResultStatus::Success;
+}
+
+NCA::~NCA() = default;
+
 Loader::ResultStatus NCA::GetStatus() const {
     return status;
 }

src/core/file_sys/content_archive.h

@@ -73,6 +73,8 @@ inline bool IsDirectoryExeFS(const std::shared_ptr<VfsDirectory>& pfs) {
     return pfs->GetFile("main") != nullptr && pfs->GetFile("main.npdm") != nullptr;
 }
 
+bool IsValidNCA(const NCAHeader& header);
+
 // An implementation of VfsDirectory that represents a Nintendo Content Archive (NCA) conatiner.
 // After construction, use GetStatus to determine if the file is valid and ready to be used.
 class NCA : public ReadOnlyVfsDirectory {
@@ -104,19 +106,10 @@ protected:
     bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;
 
 private:
-    bool CheckSupportedNCA(const NCAHeader& header);
-    bool HandlePotentialHeaderDecryption();
-
-    std::vector<NCASectionHeader> ReadSectionHeaders() const;
-    bool ReadSections(const std::vector<NCASectionHeader>& sections, u64 bktr_base_ivfc_offset);
-    bool ReadRomFSSection(const NCASectionHeader& section, const NCASectionTableEntry& entry,
-                          u64 bktr_base_ivfc_offset);
-    bool ReadPFS0Section(const NCASectionHeader& section, const NCASectionTableEntry& entry);
-
     u8 GetCryptoRevision() const;
     boost::optional<Core::Crypto::Key128> GetKeyAreaKey(NCASectionCryptoType type) const;
     boost::optional<Core::Crypto::Key128> GetTitlekey();
-    VirtualFile Decrypt(const NCASectionHeader& header, VirtualFile in, u64 starting_offset);
+    VirtualFile Decrypt(NCASectionHeader header, VirtualFile in, u64 starting_offset);
 
     std::vector<VirtualDir> dirs;
     std::vector<VirtualFile> files;
@@ -125,15 +118,15 @@ private:
     VirtualDir exefs = nullptr;
     VirtualFile file;
     VirtualFile bktr_base_romfs;
-    u64 ivfc_offset = 0;
+    u64 ivfc_offset;
 
     NCAHeader header{};
     bool has_rights_id{};
 
     Loader::ResultStatus status{};
 
-    bool encrypted = false;
-    bool is_update = false;
+    bool encrypted;
+    bool is_update;
 
     Core::Crypto::KeyManager keys;
 };

src/core/file_sys/savedata_factory.cpp

@@ -51,6 +51,13 @@ ResultVal<VirtualDir> SaveDataFactory::Open(SaveDataSpaceId space, SaveDataDescr
                     meta.title_id);
     }
 
+    if (meta.type == SaveDataType::DeviceSaveData && meta.user_id != u128{0, 0}) {
+        LOG_WARNING(Service_FS,
+                    "Possibly incorrect SaveDataDescriptor, type is DeviceSaveData but user_id is "
+                    "non-zero ({:016X}{:016X})",
+                    meta.user_id[1], meta.user_id[0]);
+    }
+
     std::string save_directory =
         GetFullPath(space, meta.type, meta.title_id, meta.user_id, meta.save_id);
 
@@ -92,6 +99,9 @@ std::string SaveDataFactory::GetFullPath(SaveDataSpaceId space, SaveDataType typ
     case SaveDataSpaceId::NandUser:
         out = "/user/";
         break;
+    case SaveDataSpaceId::TemporaryStorage:
+        out = "/temp/";
+        break;
     default:
         ASSERT_MSG(false, "Unrecognized SaveDataSpaceId: {:02X}", static_cast<u8>(space));
     }
@@ -100,10 +110,11 @@ std::string SaveDataFactory::GetFullPath(SaveDataSpaceId space, SaveDataType typ
     case SaveDataType::SystemSaveData:
         return fmt::format("{}save/{:016X}/{:016X}{:016X}", out, save_id, user_id[1], user_id[0]);
     case SaveDataType::SaveData:
+    case SaveDataType::DeviceSaveData:
         return fmt::format("{}save/{:016X}/{:016X}{:016X}/{:016X}", out, 0, user_id[1], user_id[0],
                            title_id);
     case SaveDataType::TemporaryStorage:
-        return fmt::format("{}temp/{:016X}/{:016X}{:016X}/{:016X}", out, 0, user_id[1], user_id[0],
+        return fmt::format("{}{:016X}/{:016X}{:016X}/{:016X}", out, 0, user_id[1], user_id[0],
                            title_id);
     default:
         ASSERT_MSG(false, "Unrecognized SaveDataType: {:02X}", static_cast<u8>(type));

src/core/hle/kernel/svc.cpp

@@ -448,25 +448,12 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
     case GetInfoType::RandomEntropy:
         *result = 0;
         break;
-    case GetInfoType::AddressSpaceBaseAddr:
-        *result = vm_manager.GetCodeRegionBaseAddress();
+    case GetInfoType::ASLRRegionBaseAddr:
+        *result = vm_manager.GetASLRRegionBaseAddress();
         break;
-    case GetInfoType::AddressSpaceSize: {
-        const u64 width = vm_manager.GetAddressSpaceWidth();
-
-        switch (width) {
-        case 32:
-            *result = 0xFFE00000;
-            break;
-        case 36:
-            *result = 0xFF8000000;
-            break;
-        case 39:
-            *result = 0x7FF8000000;
-            break;
-        }
+    case GetInfoType::ASLRRegionSize:
+        *result = vm_manager.GetASLRRegionSize();
         break;
-    }
     case GetInfoType::NewMapRegionBaseAddr:
         *result = vm_manager.GetNewMapRegionBaseAddress();
         break;

src/core/hle/kernel/svc.h

@@ -41,8 +41,8 @@ enum class GetInfoType : u64 {
     RandomEntropy = 11,
     PerformanceCounter = 0xF0000002,
     // 2.0.0+
-    AddressSpaceBaseAddr = 12,
-    AddressSpaceSize = 13,
+    ASLRRegionBaseAddr = 12,
+    ASLRRegionSize = 13,
     NewMapRegionBaseAddr = 14,
     NewMapRegionSize = 15,
     // 3.0.0+

src/core/hle/kernel/vm_manager.cpp

@@ -393,30 +393,35 @@ void VMManager::InitializeMemoryRegionRanges(FileSys::ProgramAddressSpaceType ty
 
     switch (type) {
     case FileSys::ProgramAddressSpaceType::Is32Bit:
+    case FileSys::ProgramAddressSpaceType::Is32BitNoMap:
         address_space_width = 32;
         code_region_base = 0x200000;
         code_region_end = code_region_base + 0x3FE00000;
-        map_region_size = 0x40000000;
-        heap_region_size = 0x40000000;
+        aslr_region_base = 0x200000;
+        aslr_region_end = aslr_region_base + 0xFFE00000;
+        if (type == FileSys::ProgramAddressSpaceType::Is32Bit) {
+            map_region_size = 0x40000000;
+            heap_region_size = 0x40000000;
+        } else {
+            map_region_size = 0;
+            heap_region_size = 0x80000000;
+        }
         break;
     case FileSys::ProgramAddressSpaceType::Is36Bit:
         address_space_width = 36;
         code_region_base = 0x8000000;
         code_region_end = code_region_base + 0x78000000;
+        aslr_region_base = 0x8000000;
+        aslr_region_end = aslr_region_base + 0xFF8000000;
         map_region_size = 0x180000000;
         heap_region_size = 0x180000000;
         break;
-    case FileSys::ProgramAddressSpaceType::Is32BitNoMap:
-        address_space_width = 32;
-        code_region_base = 0x200000;
-        code_region_end = code_region_base + 0x3FE00000;
-        map_region_size = 0;
-        heap_region_size = 0x80000000;
-        break;
     case FileSys::ProgramAddressSpaceType::Is39Bit:
         address_space_width = 39;
         code_region_base = 0x8000000;
         code_region_end = code_region_base + 0x80000000;
+        aslr_region_base = 0x8000000;
+        aslr_region_end = aslr_region_base + 0x7FF8000000;
         map_region_size = 0x1000000000;
         heap_region_size = 0x180000000;
         new_map_region_size = 0x80000000;
@@ -490,6 +495,18 @@ u64 VMManager::GetAddressSpaceWidth() const {
     return address_space_width;
 }
 
+VAddr VMManager::GetASLRRegionBaseAddress() const {
+    return aslr_region_base;
+}
+
+VAddr VMManager::GetASLRRegionEndAddress() const {
+    return aslr_region_end;
+}
+
+u64 VMManager::GetASLRRegionSize() const {
+    return aslr_region_end - aslr_region_base;
+}
+
 VAddr VMManager::GetCodeRegionBaseAddress() const {
     return code_region_base;
 }

src/core/hle/kernel/vm_manager.h

@@ -205,6 +205,15 @@ public:
     /// Gets the address space width in bits.
     u64 GetAddressSpaceWidth() const;
 
+    /// Gets the base address of the ASLR region.
+    VAddr GetASLRRegionBaseAddress() const;
+
+    /// Gets the end address of the ASLR region.
+    VAddr GetASLRRegionEndAddress() const;
+
+    /// Gets the size of the ASLR region
+    u64 GetASLRRegionSize() const;
+
     /// Gets the base address of the code region.
     VAddr GetCodeRegionBaseAddress() const;
 
@@ -306,6 +315,9 @@ private:
     VAddr address_space_base = 0;
     VAddr address_space_end = 0;
 
+    VAddr aslr_region_base = 0;
+    VAddr aslr_region_end = 0;
+
     VAddr code_region_base = 0;
     VAddr code_region_end = 0;
 

src/core/settings.h

@@ -136,7 +136,7 @@ struct Values {
     float resolution_factor;
     bool use_frame_limit;
     u16 frame_limit;
-    bool use_accurate_framebuffers;
+    bool use_accurate_gpu_emulation;
 
     float bg_red;
     float bg_green;

src/core/telemetry_session.cpp

@@ -163,8 +163,8 @@ TelemetrySession::TelemetrySession() {
     AddField(Telemetry::FieldType::UserConfig, "Renderer_UseFrameLimit",
              Settings::values.use_frame_limit);
     AddField(Telemetry::FieldType::UserConfig, "Renderer_FrameLimit", Settings::values.frame_limit);
-    AddField(Telemetry::FieldType::UserConfig, "Renderer_UseAccurateFramebuffers",
-             Settings::values.use_accurate_framebuffers);
+    AddField(Telemetry::FieldType::UserConfig, "Renderer_UseAccurateGpuEmulation",
+             Settings::values.use_accurate_gpu_emulation);
     AddField(Telemetry::FieldType::UserConfig, "System_UseDockedMode",
              Settings::values.use_docked_mode);
 }

src/video_core/engines/maxwell_3d.h

@@ -448,7 +448,10 @@ public:
                 BitField<8, 3, u32> block_depth;
                 BitField<12, 1, InvMemoryLayout> type;
             } memory_layout;
-            u32 array_mode;
+            union {
+                BitField<0, 16, u32> array_mode;
+                BitField<16, 1, u32> volume;
+            };
             u32 layer_stride;
             u32 base_layer;
             INSERT_PADDING_WORDS(7);

src/video_core/memory_manager.cpp

@@ -87,6 +87,16 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) {
     return gpu_addr;
 }
 
+GPUVAddr MemoryManager::GetRegionEnd(GPUVAddr region_start) const {
+    for (const auto& region : mapped_regions) {
+        const GPUVAddr region_end{region.gpu_addr + region.size};
+        if (region_start >= region.gpu_addr && region_start < region_end) {
+            return region_end;
+        }
+    }
+    return {};
+}
+
 boost::optional<GPUVAddr> MemoryManager::FindFreeBlock(u64 size, u64 align) {
     GPUVAddr gpu_addr = 0;
     u64 free_space = 0;

src/video_core/memory_manager.h

@@ -26,6 +26,7 @@ public:
     GPUVAddr MapBufferEx(VAddr cpu_addr, u64 size);
     GPUVAddr MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size);
     GPUVAddr UnmapBuffer(GPUVAddr gpu_addr, u64 size);
+    GPUVAddr GetRegionEnd(GPUVAddr region_start) const;
     boost::optional<VAddr> GpuToCpuAddress(GPUVAddr gpu_addr);
     std::vector<GPUVAddr> CpuToGpuAddress(VAddr cpu_addr) const;
 

src/video_core/rasterizer_cache.h

@@ -11,32 +11,77 @@
 
 #include "common/common_types.h"
 #include "core/core.h"
+#include "core/settings.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
 
+class RasterizerCacheObject {
+public:
+    /// Gets the address of the shader in guest memory, required for cache management
+    virtual VAddr GetAddr() const = 0;
+
+    /// Gets the size of the shader in guest memory, required for cache management
+    virtual std::size_t GetSizeInBytes() const = 0;
+
+    /// Wriets any cached resources back to memory
+    virtual void Flush() = 0;
+
+    /// Sets whether the cached object should be considered registered
+    void SetIsRegistered(bool registered) {
+        is_registered = registered;
+    }
+
+    /// Returns true if the cached object is registered
+    bool IsRegistered() const {
+        return is_registered;
+    }
+
+    /// Returns true if the cached object is dirty
+    bool IsDirty() const {
+        return is_dirty;
+    }
+
+    /// Returns ticks from when this cached object was last modified
+    u64 GetLastModifiedTicks() const {
+        return last_modified_ticks;
+    }
+
+    /// Marks an object as recently modified, used to specify whether it is clean or dirty
+    template <class T>
+    void MarkAsModified(bool dirty, T& cache) {
+        is_dirty = dirty;
+        last_modified_ticks = cache.GetModifiedTicks();
+    }
+
+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)
+    u64 last_modified_ticks{}; ///< When the object was last modified, used for in-order flushing
+};
+
 template <class T>
 class RasterizerCache : NonCopyable {
+    friend class RasterizerCacheObject;
+
 public:
+    /// Write any cached resources overlapping the specified region back to memory
+    void FlushRegion(Tegra::GPUVAddr addr, size_t size) {
+        const auto& objects{GetSortedObjectsFromRegion(addr, size)};
+        for (auto& object : objects) {
+            FlushObject(object);
+        }
+    }
+
     /// Mark the specified region as being invalidated
     void InvalidateRegion(VAddr addr, u64 size) {
-        if (size == 0)
-            return;
-
-        const ObjectInterval interval{addr, addr + size};
-        for (auto& pair : boost::make_iterator_range(object_cache.equal_range(interval))) {
-            for (auto& cached_object : pair.second) {
-                if (!cached_object)
-                    continue;
-
-                remove_objects.emplace(cached_object);
+        const auto& objects{GetSortedObjectsFromRegion(addr, size)};
+        for (auto& object : objects) {
+            if (!object->IsRegistered()) {
+                // Skip duplicates
+                continue;
             }
+            Unregister(object);
         }
-
-        for (auto& remove_object : remove_objects) {
-            Unregister(remove_object);
-        }
-
-        remove_objects.clear();
     }
 
     /// Invalidates everything in the cache
@@ -62,6 +107,7 @@ protected:
 
     /// Register an object into the cache
     void Register(const T& object) {
+        object->SetIsRegistered(true);
         object_cache.add({GetInterval(object), ObjectSet{object}});
         auto& rasterizer = Core::System::GetInstance().Renderer().Rasterizer();
         rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), 1);
@@ -69,12 +115,57 @@ protected:
 
     /// Unregisters an object from the cache
     void Unregister(const T& object) {
+        object->SetIsRegistered(false);
         auto& rasterizer = Core::System::GetInstance().Renderer().Rasterizer();
         rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), -1);
+
+        // Only flush if use_accurate_gpu_emulation is enabled, as it incurs a performance hit
+        if (Settings::values.use_accurate_gpu_emulation) {
+            FlushObject(object);
+        }
+
         object_cache.subtract({GetInterval(object), ObjectSet{object}});
     }
 
+    /// Returns a ticks counter used for tracking when cached objects were last modified
+    u64 GetModifiedTicks() {
+        return ++modified_ticks;
+    }
+
 private:
+    /// Returns a list of cached objects from the specified memory region, ordered by access time
+    std::vector<T> GetSortedObjectsFromRegion(VAddr addr, u64 size) {
+        if (size == 0) {
+            return {};
+        }
+
+        std::vector<T> objects;
+        const ObjectInterval interval{addr, addr + size};
+        for (auto& pair : boost::make_iterator_range(object_cache.equal_range(interval))) {
+            for (auto& cached_object : pair.second) {
+                if (!cached_object) {
+                    continue;
+                }
+                objects.push_back(cached_object);
+            }
+        }
+
+        std::sort(objects.begin(), objects.end(), [](const T& a, const T& b) -> bool {
+            return a->GetLastModifiedTicks() < b->GetLastModifiedTicks();
+        });
+
+        return objects;
+    }
+
+    /// Flushes the specified object, updating appropriate cache state as needed
+    void FlushObject(const T& object) {
+        if (!object->IsDirty()) {
+            return;
+        }
+        object->Flush();
+        object->MarkAsModified(false, *this);
+    }
+
     using ObjectSet = std::set<T>;
     using ObjectCache = boost::icl::interval_map<VAddr, ObjectSet>;
     using ObjectInterval = typename ObjectCache::interval_type;
@@ -84,6 +175,6 @@ private:
                                           object->GetAddr() + object->GetSizeInBytes());
     }
 
-    ObjectCache object_cache;
-    ObjectSet remove_objects;
+    ObjectCache object_cache; ///< Cache of objects
+    u64 modified_ticks{};     ///< Counter of cache state ticks, used for in-order flushing
 };

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -15,15 +15,18 @@
 
 namespace OpenGL {
 
-struct CachedBufferEntry final {
-    VAddr GetAddr() const {
+struct CachedBufferEntry final : public RasterizerCacheObject {
+    VAddr GetAddr() const override {
         return addr;
     }
 
-    std::size_t GetSizeInBytes() const {
+    std::size_t GetSizeInBytes() const override {
         return size;
     }
 
+    // We do not have to flush this cache as things in it are never modified by us.
+    void Flush() override {}
+
     VAddr addr;
     std::size_t size;
     GLintptr offset;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -424,6 +424,13 @@ void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_dep
             // Used when just a single color attachment is enabled, e.g. for clearing a color buffer
             Surface color_surface =
                 res_cache.GetColorBufferSurface(*single_color_target, preserve_contents);
+
+            if (color_surface) {
+                // Assume that a surface will be written to if it is used as a framebuffer, even if
+                // the shader doesn't actually write to it.
+                color_surface->MarkAsModified(true, res_cache);
+            }
+
             glFramebufferTexture2D(
                 GL_DRAW_FRAMEBUFFER,
                 GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(*single_color_target), GL_TEXTURE_2D,
@@ -434,6 +441,13 @@ void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_dep
             std::array<GLenum, Maxwell::NumRenderTargets> buffers;
             for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) {
                 Surface color_surface = res_cache.GetColorBufferSurface(index, preserve_contents);
+
+                if (color_surface) {
+                    // Assume that a surface will be written to if it is used as a framebuffer, even
+                    // if the shader doesn't actually write to it.
+                    color_surface->MarkAsModified(true, res_cache);
+                }
+
                 buffers[index] = GL_COLOR_ATTACHMENT0 + regs.rt_control.GetMap(index);
                 glFramebufferTexture2D(
                     GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(index),
@@ -453,6 +467,10 @@ void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_dep
     }
 
     if (depth_surface) {
+        // Assume that a surface will be written to if it is used as a framebuffer, even if
+        // the shader doesn't actually write to it.
+        depth_surface->MarkAsModified(true, res_cache);
+
         if (regs.stencil_enable) {
             // Attach both depth and stencil
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
@@ -617,7 +635,14 @@ void RasterizerOpenGL::DrawArrays() {
 
 void RasterizerOpenGL::FlushAll() {}
 
-void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {}
+void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+
+    if (Settings::values.use_accurate_gpu_emulation) {
+        // Only flush if use_accurate_gpu_emulation is enabled, as it incurs a performance hit
+        res_cache.FlushRegion(addr, size);
+    }
+}
 
 void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
@@ -627,6 +652,7 @@ void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
 }
 
 void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+    FlushRegion(addr, size);
     InvalidateRegion(addr, size);
 }
 

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -34,16 +34,53 @@ struct FormatTuple {
     bool compressed;
 };
 
-static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
-    auto& gpu{Core::System::GetInstance().GPU()};
-    const auto cpu_addr{gpu.MemoryManager().GpuToCpuAddress(gpu_addr)};
-    return cpu_addr ? *cpu_addr : 0;
+static bool IsPixelFormatASTC(PixelFormat format) {
+    switch (format) {
+    case PixelFormat::ASTC_2D_4X4:
+    case PixelFormat::ASTC_2D_5X4:
+    case PixelFormat::ASTC_2D_8X8:
+    case PixelFormat::ASTC_2D_8X5:
+        return true;
+    default:
+        return false;
+    }
+}
+
+static std::pair<u32, u32> GetASTCBlockSize(PixelFormat format) {
+    switch (format) {
+    case PixelFormat::ASTC_2D_4X4:
+        return {4, 4};
+    case PixelFormat::ASTC_2D_5X4:
+        return {5, 4};
+    case PixelFormat::ASTC_2D_8X8:
+        return {8, 8};
+    case PixelFormat::ASTC_2D_8X5:
+        return {8, 5};
+    default:
+        LOG_CRITICAL(HW_GPU, "Unhandled format: {}", static_cast<u32>(format));
+        UNREACHABLE();
+    }
+}
+
+void SurfaceParams::InitCacheParameters(Tegra::GPUVAddr gpu_addr_) {
+    auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
+    const auto cpu_addr{memory_manager.GpuToCpuAddress(gpu_addr_)};
+
+    addr = cpu_addr ? *cpu_addr : 0;
+    gpu_addr = gpu_addr_;
+    size_in_bytes = SizeInBytesRaw();
+
+    if (IsPixelFormatASTC(pixel_format)) {
+        // ASTC is uncompressed in software, in emulated as RGBA8
+        size_in_bytes_gl = width * height * depth * 4;
+    } else {
+        size_in_bytes_gl = SizeInBytesGL();
+    }
 }
 
 /*static*/ SurfaceParams SurfaceParams::CreateForTexture(
     const Tegra::Texture::FullTextureInfo& config, const GLShader::SamplerEntry& entry) {
     SurfaceParams params{};
-    params.addr = TryGetCpuAddr(config.tic.Address());
     params.is_tiled = config.tic.IsTiled();
     params.block_width = params.is_tiled ? config.tic.BlockWidth() : 0,
     params.block_height = params.is_tiled ? config.tic.BlockHeight() : 0,
@@ -87,18 +124,18 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
         break;
     }
 
-    params.size_in_bytes_total = params.SizeInBytesTotal();
-    params.size_in_bytes_2d = params.SizeInBytes2D();
     params.max_mip_level = config.tic.max_mip_level + 1;
     params.rt = {};
 
+    params.InitCacheParameters(config.tic.Address());
+
     return params;
 }
 
 /*static*/ SurfaceParams SurfaceParams::CreateForFramebuffer(std::size_t index) {
     const auto& config{Core::System::GetInstance().GPU().Maxwell3D().regs.rt[index]};
     SurfaceParams params{};
-    params.addr = TryGetCpuAddr(config.Address());
+
     params.is_tiled =
         config.memory_layout.type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
     params.block_width = 1 << config.memory_layout.block_width;
@@ -112,16 +149,17 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     params.unaligned_height = config.height;
     params.target = SurfaceTarget::Texture2D;
     params.depth = 1;
-    params.size_in_bytes_total = params.SizeInBytesTotal();
-    params.size_in_bytes_2d = params.SizeInBytes2D();
     params.max_mip_level = 0;
 
     // Render target specific parameters, not used for caching
     params.rt.index = static_cast<u32>(index);
     params.rt.array_mode = config.array_mode;
     params.rt.layer_stride = config.layer_stride;
+    params.rt.volume = config.volume;
     params.rt.base_layer = config.base_layer;
 
+    params.InitCacheParameters(config.Address());
+
     return params;
 }
 
@@ -130,7 +168,7 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     u32 block_width, u32 block_height, u32 block_depth,
     Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type) {
     SurfaceParams params{};
-    params.addr = TryGetCpuAddr(zeta_address);
+
     params.is_tiled = type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
     params.block_width = 1 << std::min(block_width, 5U);
     params.block_height = 1 << std::min(block_height, 5U);
@@ -143,18 +181,18 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     params.unaligned_height = zeta_height;
     params.target = SurfaceTarget::Texture2D;
     params.depth = 1;
-    params.size_in_bytes_total = params.SizeInBytesTotal();
-    params.size_in_bytes_2d = params.SizeInBytes2D();
     params.max_mip_level = 0;
     params.rt = {};
 
+    params.InitCacheParameters(zeta_address);
+
     return params;
 }
 
 /*static*/ SurfaceParams SurfaceParams::CreateForFermiCopySurface(
     const Tegra::Engines::Fermi2D::Regs::Surface& config) {
     SurfaceParams params{};
-    params.addr = TryGetCpuAddr(config.Address());
+
     params.is_tiled = !config.linear;
     params.block_width = params.is_tiled ? std::min(config.BlockWidth(), 32U) : 0,
     params.block_height = params.is_tiled ? std::min(config.BlockHeight(), 32U) : 0,
@@ -167,11 +205,11 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     params.unaligned_height = config.height;
     params.target = SurfaceTarget::Texture2D;
     params.depth = 1;
-    params.size_in_bytes_total = params.SizeInBytesTotal();
-    params.size_in_bytes_2d = params.SizeInBytes2D();
     params.max_mip_level = 0;
     params.rt = {};
 
+    params.InitCacheParameters(config.Address());
+
     return params;
 }
 
@@ -276,34 +314,6 @@ static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType
     return format;
 }
 
-static bool IsPixelFormatASTC(PixelFormat format) {
-    switch (format) {
-    case PixelFormat::ASTC_2D_4X4:
-    case PixelFormat::ASTC_2D_5X4:
-    case PixelFormat::ASTC_2D_8X8:
-    case PixelFormat::ASTC_2D_8X5:
-        return true;
-    default:
-        return false;
-    }
-}
-
-static std::pair<u32, u32> GetASTCBlockSize(PixelFormat format) {
-    switch (format) {
-    case PixelFormat::ASTC_2D_4X4:
-        return {4, 4};
-    case PixelFormat::ASTC_2D_5X4:
-        return {5, 4};
-    case PixelFormat::ASTC_2D_8X8:
-        return {8, 8};
-    case PixelFormat::ASTC_2D_8X5:
-        return {8, 5};
-    default:
-        LOG_CRITICAL(HW_GPU, "Unhandled format: {}", static_cast<u32>(format));
-        UNREACHABLE();
-    }
-}
-
 MathUtil::Rectangle<u32> SurfaceParams::GetRect() const {
     u32 actual_height{unaligned_height};
     if (IsPixelFormatASTC(pixel_format)) {
@@ -333,23 +343,21 @@ static bool IsFormatBCn(PixelFormat format) {
 template <bool morton_to_gl, PixelFormat format>
 void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth, u32 depth, u8* gl_buffer,
                 std::size_t gl_buffer_size, VAddr addr) {
-    constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
-    constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format);
+    constexpr u32 bytes_per_pixel = SurfaceParams::GetBytesPerPixel(format);
+
+    // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
+    // pixel values.
+    const u32 tile_size{IsFormatBCn(format) ? 4U : 1U};
 
     if (morton_to_gl) {
-        // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
-        // pixel values.
-        const u32 tile_size{IsFormatBCn(format) ? 4U : 1U};
         const std::vector<u8> data = Tegra::Texture::UnswizzleTexture(
             addr, tile_size, bytes_per_pixel, stride, height, depth, block_height, block_depth);
         const std::size_t size_to_copy{std::min(gl_buffer_size, data.size())};
         memcpy(gl_buffer, data.data(), size_to_copy);
     } else {
-        // TODO(bunnei): Assumes the default rendering GOB size of 16 (128 lines). We should
-        // check the configuration for this and perform more generic un/swizzle
-        LOG_WARNING(Render_OpenGL, "need to use correct swizzle/GOB parameters!");
-        VideoCore::MortonCopyPixels128(stride, height, bytes_per_pixel, gl_bytes_per_pixel,
-                                       Memory::GetPointer(addr), gl_buffer, morton_to_gl);
+        Tegra::Texture::CopySwizzledData(stride / tile_size, height / tile_size, depth,
+                                         bytes_per_pixel, bytes_per_pixel, Memory::GetPointer(addr),
+                                         gl_buffer, false, block_height, block_depth);
     }
 }
 
@@ -430,17 +438,16 @@ static constexpr std::array<void (*)(u32, u32, u32, u32, u32, u8*, std::size_t,
         MortonCopy<false, PixelFormat::RGBA16UI>,
         MortonCopy<false, PixelFormat::R11FG11FB10F>,
         MortonCopy<false, PixelFormat::RGBA32UI>,
-        // TODO(Subv): Swizzling DXT1/DXT23/DXT45/DXN1/DXN2/BC7U/BC6H_UF16/BC6H_SF16/ASTC_2D_4X4
-        // formats are not supported
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
+        MortonCopy<false, PixelFormat::DXT1>,
+        MortonCopy<false, PixelFormat::DXT23>,
+        MortonCopy<false, PixelFormat::DXT45>,
+        MortonCopy<false, PixelFormat::DXN1>,
+        MortonCopy<false, PixelFormat::DXN2UNORM>,
+        MortonCopy<false, PixelFormat::DXN2SNORM>,
+        MortonCopy<false, PixelFormat::BC7U>,
+        MortonCopy<false, PixelFormat::BC6H_UF16>,
+        MortonCopy<false, PixelFormat::BC6H_SF16>,
+        // TODO(Subv): Swizzling ASTC formats are not supported
         nullptr,
         MortonCopy<false, PixelFormat::G8R8U>,
         MortonCopy<false, PixelFormat::G8R8S>,
@@ -626,22 +633,21 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
     auto source_format = GetFormatTuple(src_params.pixel_format, src_params.component_type);
     auto dest_format = GetFormatTuple(dst_params.pixel_format, dst_params.component_type);
 
-    std::size_t buffer_size =
-        std::max(src_params.size_in_bytes_total, dst_params.size_in_bytes_total);
+    std::size_t buffer_size = std::max(src_params.size_in_bytes, dst_params.size_in_bytes);
 
     glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle);
     glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW_ARB);
     if (source_format.compressed) {
         glGetCompressedTextureImage(src_surface->Texture().handle, src_attachment,
-                                    static_cast<GLsizei>(src_params.size_in_bytes_total), nullptr);
+                                    static_cast<GLsizei>(src_params.size_in_bytes), nullptr);
     } else {
         glGetTextureImage(src_surface->Texture().handle, src_attachment, source_format.format,
-                          source_format.type, static_cast<GLsizei>(src_params.size_in_bytes_total),
+                          source_format.type, static_cast<GLsizei>(src_params.size_in_bytes),
                           nullptr);
     }
     // If the new texture is bigger than the previous one, we need to fill in the rest with data
     // from the CPU.
-    if (src_params.size_in_bytes_total < dst_params.size_in_bytes_total) {
+    if (src_params.size_in_bytes < dst_params.size_in_bytes) {
         // Upload the rest of the memory.
         if (dst_params.is_tiled) {
             // TODO(Subv): We might have to de-tile the subtexture and re-tile it with the rest
@@ -651,12 +657,12 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
             LOG_DEBUG(HW_GPU, "Trying to upload extra texture data from the CPU during "
                               "reinterpretation but the texture is tiled.");
         }
-        std::size_t remaining_size =
-            dst_params.size_in_bytes_total - src_params.size_in_bytes_total;
+        std::size_t remaining_size = dst_params.size_in_bytes - src_params.size_in_bytes;
         std::vector<u8> data(remaining_size);
-        Memory::ReadBlock(dst_params.addr + src_params.size_in_bytes_total, data.data(),
-                          data.size());
-        glBufferSubData(GL_PIXEL_PACK_BUFFER, src_params.size_in_bytes_total, remaining_size,
+        std::memcpy(data.data(), Memory::GetPointer(dst_params.addr + src_params.size_in_bytes),
+                    data.size());
+
+        glBufferSubData(GL_PIXEL_PACK_BUFFER, src_params.size_in_bytes, remaining_size,
                         data.data());
     }
 
@@ -702,7 +708,8 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
 }
 
 CachedSurface::CachedSurface(const SurfaceParams& params)
-    : params(params), gl_target(SurfaceTargetToGL(params.target)) {
+    : params(params), gl_target(SurfaceTargetToGL(params.target)),
+      cached_size_in_bytes(params.size_in_bytes) {
     texture.Create();
     const auto& rect{params.GetRect()};
 
@@ -752,9 +759,21 @@ CachedSurface::CachedSurface(const SurfaceParams& params)
 
     VideoCore::LabelGLObject(GL_TEXTURE, texture.handle, params.addr,
                              SurfaceParams::SurfaceTargetName(params.target));
+
+    // Clamp size to mapped GPU memory region
+    // TODO(bunnei): Super Mario Odyssey maps a 0x40000 byte region and then uses it for a 0x80000
+    // R32F render buffer. We do not yet know if this is a game bug or something else, but this
+    // check is necessary to prevent flushing from overwriting unmapped memory.
+
+    auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
+    const u64 max_size{memory_manager.GetRegionEnd(params.gpu_addr) - params.gpu_addr};
+    if (cached_size_in_bytes > max_size) {
+        LOG_ERROR(HW_GPU, "Surface size {} exceeds region size {}", params.size_in_bytes, max_size);
+        cached_size_in_bytes = max_size;
+    }
 }
 
-static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height) {
+static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height, bool reverse) {
     union S8Z24 {
         BitField<0, 24, u32> z24;
         BitField<24, 8, u32> s8;
@@ -767,22 +786,29 @@ static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height) {
     };
     static_assert(sizeof(Z24S8) == 4, "Z24S8 is incorrect size");
 
-    S8Z24 input_pixel{};
-    Z24S8 output_pixel{};
-    constexpr auto bpp{CachedSurface::GetGLBytesPerPixel(PixelFormat::S8Z24)};
+    S8Z24 s8z24_pixel{};
+    Z24S8 z24s8_pixel{};
+    constexpr auto bpp{SurfaceParams::GetBytesPerPixel(PixelFormat::S8Z24)};
     for (std::size_t y = 0; y < height; ++y) {
         for (std::size_t x = 0; x < width; ++x) {
             const std::size_t offset{bpp * (y * width + x)};
-            std::memcpy(&input_pixel, &data[offset], sizeof(S8Z24));
-            output_pixel.s8.Assign(input_pixel.s8);
-            output_pixel.z24.Assign(input_pixel.z24);
-            std::memcpy(&data[offset], &output_pixel, sizeof(Z24S8));
+            if (reverse) {
+                std::memcpy(&z24s8_pixel, &data[offset], sizeof(Z24S8));
+                s8z24_pixel.s8.Assign(z24s8_pixel.s8);
+                s8z24_pixel.z24.Assign(z24s8_pixel.z24);
+                std::memcpy(&data[offset], &s8z24_pixel, sizeof(S8Z24));
+            } else {
+                std::memcpy(&s8z24_pixel, &data[offset], sizeof(S8Z24));
+                z24s8_pixel.s8.Assign(s8z24_pixel.s8);
+                z24s8_pixel.z24.Assign(s8z24_pixel.z24);
+                std::memcpy(&data[offset], &z24s8_pixel, sizeof(Z24S8));
+            }
         }
     }
 }
 
 static void ConvertG8R8ToR8G8(std::vector<u8>& data, u32 width, u32 height) {
-    constexpr auto bpp{CachedSurface::GetGLBytesPerPixel(PixelFormat::G8R8U)};
+    constexpr auto bpp{SurfaceParams::GetBytesPerPixel(PixelFormat::G8R8U)};
     for (std::size_t y = 0; y < height; ++y) {
         for (std::size_t x = 0; x < width; ++x) {
             const std::size_t offset{bpp * (y * width + x)};
@@ -814,7 +840,7 @@ static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelForma
     }
     case PixelFormat::S8Z24:
         // Convert the S8Z24 depth format to Z24S8, as OpenGL does not support S8Z24.
-        ConvertS8Z24ToZ24S8(data, width, height);
+        ConvertS8Z24ToZ24S8(data, width, height, false);
         break;
 
     case PixelFormat::G8R8U:
@@ -825,22 +851,36 @@ static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelForma
     }
 }
 
+/**
+ * Helper function to perform software conversion (as needed) when flushing a buffer from OpenGL to
+ * Switch memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or
+ * with typical desktop GPUs.
+ */
+static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
+                                                u32 width, u32 height) {
+    switch (pixel_format) {
+    case PixelFormat::G8R8U:
+    case PixelFormat::G8R8S:
+    case PixelFormat::ASTC_2D_4X4:
+    case PixelFormat::ASTC_2D_8X8: {
+        LOG_CRITICAL(HW_GPU, "Conversion of format {} after texture flushing is not implemented",
+                     static_cast<u32>(pixel_format));
+        UNREACHABLE();
+        break;
+    }
+    case PixelFormat::S8Z24:
+        // Convert the Z24S8 depth format to S8Z24, as OpenGL does not support S8Z24.
+        ConvertS8Z24ToZ24S8(data, width, height, true);
+        break;
+    }
+}
+
 MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 64, 192));
 void CachedSurface::LoadGLBuffer() {
-    ASSERT(params.type != SurfaceType::Fill);
-
-    const u8* const texture_src_data = Memory::GetPointer(params.addr);
-
-    ASSERT(texture_src_data);
-
-    const u32 bytes_per_pixel = GetGLBytesPerPixel(params.pixel_format);
-    const u32 copy_size = params.width * params.height * bytes_per_pixel;
-    const std::size_t total_size = copy_size * params.depth;
-
     MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
 
+    gl_buffer.resize(params.size_in_bytes_gl);
     if (params.is_tiled) {
-        gl_buffer.resize(total_size);
         u32 depth = params.depth;
         u32 block_depth = params.block_depth;
 
@@ -853,13 +893,12 @@ void CachedSurface::LoadGLBuffer() {
             block_depth = 1U;
         }
 
-        const std::size_t size = copy_size * depth;
-
         morton_to_gl_fns[static_cast<std::size_t>(params.pixel_format)](
             params.width, params.block_height, params.height, block_depth, depth, gl_buffer.data(),
-            size, params.addr);
+            gl_buffer.size(), params.addr);
     } else {
-        const u8* const texture_src_data_end{texture_src_data + total_size};
+        const auto texture_src_data{Memory::GetPointer(params.addr)};
+        const auto texture_src_data_end{texture_src_data + params.size_in_bytes_gl};
         gl_buffer.assign(texture_src_data, texture_src_data_end);
     }
 
@@ -868,7 +907,44 @@ void CachedSurface::LoadGLBuffer() {
 
 MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
 void CachedSurface::FlushGLBuffer() {
-    ASSERT_MSG(false, "Unimplemented");
+    MICROPROFILE_SCOPE(OpenGL_SurfaceFlush);
+
+    ASSERT_MSG(!IsPixelFormatASTC(params.pixel_format), "Unimplemented");
+
+    // OpenGL temporary buffer needs to be big enough to store raw texture size
+    gl_buffer.resize(GetSizeInBytes());
+
+    const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
+    // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
+    ASSERT(params.width * SurfaceParams::GetBytesPerPixel(params.pixel_format) % 4 == 0);
+    glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.width));
+    ASSERT(!tuple.compressed);
+    glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
+    glGetTextureImage(texture.handle, 0, tuple.format, tuple.type, gl_buffer.size(),
+                      gl_buffer.data());
+    glPixelStorei(GL_PACK_ROW_LENGTH, 0);
+    ConvertFormatAsNeeded_FlushGLBuffer(gl_buffer, params.pixel_format, params.width,
+                                        params.height);
+    ASSERT(params.type != SurfaceType::Fill);
+    const u8* const texture_src_data = Memory::GetPointer(params.addr);
+    ASSERT(texture_src_data);
+    if (params.is_tiled) {
+        u32 depth = params.depth;
+        u32 block_depth = params.block_depth;
+
+        ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
+                   params.block_width, static_cast<u32>(params.target));
+
+        if (params.target == SurfaceParams::SurfaceTarget::Texture2D) {
+            // TODO(Blinkhawk): Eliminate this condition once all texture types are implemented.
+            depth = 1U;
+        }
+        gl_to_morton_fns[static_cast<size_t>(params.pixel_format)](
+            params.width, params.block_height, params.height, block_depth, depth, gl_buffer.data(),
+            gl_buffer.size(), GetAddr());
+    } else {
+        std::memcpy(Memory::GetPointer(GetAddr()), gl_buffer.data(), GetSizeInBytes());
+    }
 }
 
 MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 64, 192));
@@ -878,9 +954,6 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
 
     MICROPROFILE_SCOPE(OpenGL_TextureUL);
 
-    ASSERT(gl_buffer.size() == static_cast<std::size_t>(params.width) * params.height *
-                                   GetGLBytesPerPixel(params.pixel_format) * params.depth);
-
     const auto& rect{params.GetRect()};
 
     // Load data from memory to the surface
@@ -889,7 +962,7 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
     std::size_t buffer_offset =
         static_cast<std::size_t>(static_cast<std::size_t>(y0) * params.width +
                                  static_cast<std::size_t>(x0)) *
-        GetGLBytesPerPixel(params.pixel_format);
+        SurfaceParams::GetBytesPerPixel(params.pixel_format);
 
     const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
     const GLuint target_tex = texture.handle;
@@ -905,7 +978,7 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
     cur_state.Apply();
 
     // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
-    ASSERT(params.width * GetGLBytesPerPixel(params.pixel_format) % 4 == 0);
+    ASSERT(params.width * SurfaceParams::GetBytesPerPixel(params.pixel_format) % 4 == 0);
     glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.width));
 
     glActiveTexture(GL_TEXTURE0);
@@ -915,7 +988,7 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
             glCompressedTexImage2D(
                 SurfaceTargetToGL(params.target), 0, tuple.internal_format,
                 static_cast<GLsizei>(params.width), static_cast<GLsizei>(params.height), 0,
-                static_cast<GLsizei>(params.size_in_bytes_2d), &gl_buffer[buffer_offset]);
+                static_cast<GLsizei>(params.size_in_bytes_gl), &gl_buffer[buffer_offset]);
             break;
         case SurfaceParams::SurfaceTarget::Texture3D:
         case SurfaceParams::SurfaceTarget::Texture2DArray:
@@ -923,16 +996,16 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
                 SurfaceTargetToGL(params.target), 0, tuple.internal_format,
                 static_cast<GLsizei>(params.width), static_cast<GLsizei>(params.height),
                 static_cast<GLsizei>(params.depth), 0,
-                static_cast<GLsizei>(params.size_in_bytes_total), &gl_buffer[buffer_offset]);
+                static_cast<GLsizei>(params.size_in_bytes_gl), &gl_buffer[buffer_offset]);
             break;
         case SurfaceParams::SurfaceTarget::TextureCubemap:
             for (std::size_t face = 0; face < params.depth; ++face) {
                 glCompressedTexImage2D(static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face),
                                        0, tuple.internal_format, static_cast<GLsizei>(params.width),
                                        static_cast<GLsizei>(params.height), 0,
-                                       static_cast<GLsizei>(params.size_in_bytes_2d),
+                                       static_cast<GLsizei>(params.SizeInBytesCubeFaceGL()),
                                        &gl_buffer[buffer_offset]);
-                buffer_offset += params.size_in_bytes_2d;
+                buffer_offset += params.SizeInBytesCubeFace();
             }
             break;
         default:
@@ -942,7 +1015,7 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
             glCompressedTexImage2D(
                 GL_TEXTURE_2D, 0, tuple.internal_format, static_cast<GLsizei>(params.width),
                 static_cast<GLsizei>(params.height), 0,
-                static_cast<GLsizei>(params.size_in_bytes_2d), &gl_buffer[buffer_offset]);
+                static_cast<GLsizei>(params.size_in_bytes_gl), &gl_buffer[buffer_offset]);
         }
     } else {
 
@@ -971,7 +1044,7 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
                                 y0, static_cast<GLsizei>(rect.GetWidth()),
                                 static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
                                 &gl_buffer[buffer_offset]);
-                buffer_offset += params.size_in_bytes_2d;
+                buffer_offset += params.SizeInBytesCubeFace();
             }
             break;
         default:
@@ -1033,10 +1106,7 @@ Surface RasterizerCacheOpenGL::GetColorBufferSurface(std::size_t index, bool pre
 void RasterizerCacheOpenGL::LoadSurface(const Surface& surface) {
     surface->LoadGLBuffer();
     surface->UploadGLTexture(read_framebuffer.handle, draw_framebuffer.handle);
-}
-
-void RasterizerCacheOpenGL::FlushSurface(const Surface& surface) {
-    surface->FlushGLBuffer();
+    surface->MarkAsModified(false, *this);
 }
 
 Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool preserve_contents) {
@@ -1053,8 +1123,8 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool pres
         } else if (preserve_contents) {
             // If surface parameters changed and we care about keeping the previous data, recreate
             // the surface from the old one
-            Unregister(surface);
             Surface new_surface{RecreateSurface(surface, params)};
+            Unregister(surface);
             Register(new_surface);
             return new_surface;
         } else {
@@ -1105,6 +1175,14 @@ void RasterizerCacheOpenGL::FermiCopySurface(
     FastCopySurface(GetSurface(src_params, true), GetSurface(dst_params, false));
 }
 
+void RasterizerCacheOpenGL::AccurateCopySurface(const Surface& src_surface,
+                                                const Surface& dst_surface) {
+    const auto& src_params{src_surface->GetSurfaceParams()};
+    const auto& dst_params{dst_surface->GetSurfaceParams()};
+    FlushRegion(src_params.addr, dst_params.size_in_bytes);
+    LoadSurface(dst_surface);
+}
+
 Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
                                                const SurfaceParams& new_params) {
     // Verify surface is compatible for blitting
@@ -1113,6 +1191,12 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
     // Get a new surface with the new parameters, and blit the previous surface to it
     Surface new_surface{GetUncachedSurface(new_params)};
 
+    // With use_accurate_gpu_emulation enabled, do an accurate surface copy
+    if (Settings::values.use_accurate_gpu_emulation) {
+        AccurateCopySurface(old_surface, new_surface);
+        return new_surface;
+    }
+
     // For compatible surfaces, we can just do fast glCopyImageSubData based copy
     if (old_params.target == new_params.target && old_params.type == new_params.type &&
         old_params.depth == new_params.depth && old_params.depth == 1 &&
@@ -1124,11 +1208,10 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
 
     // If the format is the same, just do a framebuffer blit. This is significantly faster than
     // using PBOs. The is also likely less accurate, as textures will be converted rather than
-    // reinterpreted. When use_accurate_framebuffers setting is enabled, perform a more accurate
+    // reinterpreted. When use_accurate_gpu_emulation setting is enabled, perform a more accurate
     // surface copy, where pixels are reinterpreted as a new format (without conversion). This
     // code path uses OpenGL PBOs and is quite slow.
-    const bool is_blit{old_params.pixel_format == new_params.pixel_format ||
-                       !Settings::values.use_accurate_framebuffers};
+    const bool is_blit{old_params.pixel_format == new_params.pixel_format};
 
     switch (new_params.target) {
     case SurfaceParams::SurfaceTarget::Texture2D:
@@ -1138,6 +1221,9 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
             CopySurface(old_surface, new_surface, copy_pbo.handle);
         }
         break;
+    case SurfaceParams::SurfaceTarget::Texture3D:
+        AccurateCopySurface(old_surface, new_surface);
+        break;
     case SurfaceParams::SurfaceTarget::TextureCubemap: {
         if (old_params.rt.array_mode != 1) {
             // TODO(bunnei): This is used by Breath of the Wild, I'm not sure how to implement this

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -18,6 +18,7 @@
 #include "video_core/rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
+#include "video_core/textures/decoders.h"
 #include "video_core/textures/texture.h"
 
 namespace OpenGL {
@@ -131,6 +132,8 @@ struct SurfaceParams {
         case Tegra::Texture::TextureType::Texture2D:
         case Tegra::Texture::TextureType::Texture2DNoMipmap:
             return SurfaceTarget::Texture2D;
+        case Tegra::Texture::TextureType::Texture3D:
+            return SurfaceTarget::Texture3D;
         case Tegra::Texture::TextureType::TextureCubemap:
             return SurfaceTarget::TextureCubemap;
         case Tegra::Texture::TextureType::Texture1DArray:
@@ -701,21 +704,42 @@ struct SurfaceParams {
         return SurfaceType::Invalid;
     }
 
+    /// Returns the sizer in bytes of the specified pixel format
+    static constexpr u32 GetBytesPerPixel(PixelFormat pixel_format) {
+        if (pixel_format == SurfaceParams::PixelFormat::Invalid) {
+            return 0;
+        }
+        return GetFormatBpp(pixel_format) / CHAR_BIT;
+    }
+
     /// Returns the rectangle corresponding to this surface
     MathUtil::Rectangle<u32> GetRect() const;
 
-    /// Returns the size of this surface as a 2D texture in bytes, adjusted for compression
-    std::size_t SizeInBytes2D() const {
+    /// Returns the total size of this surface in bytes, adjusted for compression
+    std::size_t SizeInBytesRaw(bool ignore_tiled = false) const {
         const u32 compression_factor{GetCompressionFactor(pixel_format)};
-        ASSERT(width % compression_factor == 0);
-        ASSERT(height % compression_factor == 0);
-        return (width / compression_factor) * (height / compression_factor) *
-               GetFormatBpp(pixel_format) / CHAR_BIT;
+        const u32 bytes_per_pixel{GetBytesPerPixel(pixel_format)};
+        const size_t uncompressed_size{
+            Tegra::Texture::CalculateSize((ignore_tiled ? false : is_tiled), bytes_per_pixel, width,
+                                          height, depth, block_height, block_depth)};
+
+        // Divide by compression_factor^2, as height and width are factored by this
+        return uncompressed_size / (compression_factor * compression_factor);
     }
 
-    /// Returns the total size of this surface in bytes, adjusted for compression
-    std::size_t SizeInBytesTotal() const {
-        return SizeInBytes2D() * depth;
+    /// Returns the size of this surface as an OpenGL texture in bytes
+    std::size_t SizeInBytesGL() const {
+        return SizeInBytesRaw(true);
+    }
+
+    /// Returns the size of this surface as a cube face in bytes
+    std::size_t SizeInBytesCubeFace() const {
+        return size_in_bytes / 6;
+    }
+
+    /// Returns the size of this surface as an OpenGL cube face in bytes
+    std::size_t SizeInBytesCubeFaceGL() const {
+        return size_in_bytes_gl / 6;
     }
 
     /// Creates SurfaceParams from a texture configuration
@@ -742,7 +766,9 @@ struct SurfaceParams {
                         other.depth);
     }
 
-    VAddr addr;
+    /// Initializes parameters for caching, should be called after everything has been initialized
+    void InitCacheParameters(Tegra::GPUVAddr gpu_addr);
+
     bool is_tiled;
     u32 block_width;
     u32 block_height;
@@ -754,15 +780,20 @@ struct SurfaceParams {
     u32 height;
     u32 depth;
     u32 unaligned_height;
-    std::size_t size_in_bytes_total;
-    std::size_t size_in_bytes_2d;
     SurfaceTarget target;
     u32 max_mip_level;
 
+    // Parameters used for caching
+    VAddr addr;
+    Tegra::GPUVAddr gpu_addr;
+    std::size_t size_in_bytes;
+    std::size_t size_in_bytes_gl;
+
     // Render target specific parameters, not used in caching
     struct {
         u32 index;
         u32 array_mode;
+        u32 volume;
         u32 layer_stride;
         u32 base_layer;
     } rt;
@@ -775,7 +806,8 @@ struct SurfaceReserveKey : Common::HashableStruct<OpenGL::SurfaceParams> {
     static SurfaceReserveKey Create(const OpenGL::SurfaceParams& params) {
         SurfaceReserveKey res;
         res.state = params;
-        res.state.rt = {}; // Ignore rt config in caching
+        res.state.gpu_addr = {}; // Ignore GPU vaddr in caching
+        res.state.rt = {};       // Ignore rt config in caching
         return res;
     }
 };
@@ -790,16 +822,20 @@ struct hash<SurfaceReserveKey> {
 
 namespace OpenGL {
 
-class CachedSurface final {
+class CachedSurface final : public RasterizerCacheObject {
 public:
     CachedSurface(const SurfaceParams& params);
 
-    VAddr GetAddr() const {
+    VAddr GetAddr() const override {
         return params.addr;
     }
 
-    std::size_t GetSizeInBytes() const {
-        return params.size_in_bytes_total;
+    std::size_t GetSizeInBytes() const override {
+        return cached_size_in_bytes;
+    }
+
+    void Flush() override {
+        FlushGLBuffer();
     }
 
     const OGLTexture& Texture() const {
@@ -810,13 +846,6 @@ public:
         return gl_target;
     }
 
-    static constexpr unsigned int GetGLBytesPerPixel(SurfaceParams::PixelFormat format) {
-        if (format == SurfaceParams::PixelFormat::Invalid)
-            return 0;
-
-        return SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
-    }
-
     const SurfaceParams& GetSurfaceParams() const {
         return params;
     }
@@ -833,6 +862,7 @@ private:
     std::vector<u8> gl_buffer;
     SurfaceParams params;
     GLenum gl_target;
+    std::size_t cached_size_in_bytes;
 };
 
 class RasterizerCacheOpenGL final : public RasterizerCache<Surface> {
@@ -849,9 +879,6 @@ public:
     /// Get the color surface based on the framebuffer configuration and the specified render target
     Surface GetColorBufferSurface(std::size_t index, bool preserve_contents);
 
-    /// Flushes the surface to Switch memory
-    void FlushSurface(const Surface& surface);
-
     /// Tries to find a framebuffer using on the provided CPU address
     Surface TryFindFramebufferSurface(VAddr addr) const;
 
@@ -875,6 +902,9 @@ private:
     /// Tries to get a reserved surface for the specified parameters
     Surface TryGetReservedSurface(const SurfaceParams& params);
 
+    /// Performs a slow but accurate surface copy, flushing to RAM and reinterpreting the data
+    void AccurateCopySurface(const Surface& src_surface, const Surface& dst_surface);
+
     /// The surface reserve is a "backup" cache, this is where we put unique surfaces that have
     /// previously been used. This is to prevent surfaces from being constantly created and
     /// destroyed when used with different surface parameters.

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -19,20 +19,21 @@ class CachedShader;
 using Shader = std::shared_ptr<CachedShader>;
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
-class CachedShader final {
+class CachedShader final : public RasterizerCacheObject {
 public:
     CachedShader(VAddr addr, Maxwell::ShaderProgram program_type);
 
-    /// Gets the address of the shader in guest memory, required for cache management
-    VAddr GetAddr() const {
+    VAddr GetAddr() const override {
         return addr;
     }
 
-    /// Gets the size of the shader in guest memory, required for cache management
-    std::size_t GetSizeInBytes() const {
+    std::size_t GetSizeInBytes() const override {
         return GLShader::MAX_PROGRAM_CODE_LENGTH * sizeof(u64);
     }
 
+    // We do not have to flush this cache as things in it are never modified by us.
+    void Flush() override {}
+
     /// Gets the shader entries for the shader
     const GLShader::ShaderEntries& GetShaderEntries() const {
         return entries;

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -1142,6 +1142,7 @@ private:
         case Tegra::Shader::TextureType::Texture2D: {
             return 2;
         }
+        case Tegra::Shader::TextureType::Texture3D:
         case Tegra::Shader::TextureType::TextureCube: {
             return 3;
         }
@@ -2036,9 +2037,9 @@ private:
                 break;
             }
             case OpCode::Id::TEX: {
-                ASSERT_MSG(instr.tex.array == 0, "TEX arrays unimplemented");
                 Tegra::Shader::TextureType texture_type{instr.tex.texture_type};
                 std::string coord;
+                const bool is_array = instr.tex.array != 0;
 
                 ASSERT_MSG(!instr.tex.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
                            "NODEP is not implemented");
@@ -2053,21 +2054,59 @@ private:
 
                 switch (num_coordinates) {
                 case 1: {
-                    const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
-                    coord = "float coords = " + x + ';';
+                    if (is_array) {
+                        const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
+                        const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                        coord = "vec2 coords = vec2(" + x + ", " + index + ");";
+                    } else {
+                        const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                        coord = "float coords = " + x + ';';
+                    }
                     break;
                 }
                 case 2: {
-                    const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
-                    const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
-                    coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+                    if (is_array) {
+                        const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
+                        const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                        const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2);
+                        coord = "vec3 coords = vec3(" + x + ", " + y + ", " + index + ");";
+                    } else {
+                        const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                        const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                        coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+                    }
                     break;
                 }
                 case 3: {
-                    const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
-                    const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
-                    const std::string z = regs.GetRegisterAsFloat(instr.gpr20);
-                    coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+                    if (depth_compare) {
+                        if (is_array) {
+                            const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
+                            const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                            const std::string y = regs.GetRegisterAsFloat(instr.gpr20);
+                            const std::string z = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
+                            coord = "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " + index +
+                                    ");";
+                        } else {
+                            const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                            const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                            const std::string z = regs.GetRegisterAsFloat(instr.gpr20);
+                            coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+                        }
+                    } else {
+                        if (is_array) {
+                            const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
+                            const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                            const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2);
+                            const std::string z = regs.GetRegisterAsFloat(instr.gpr8.Value() + 3);
+                            coord = "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " + index +
+                                    ");";
+                        } else {
+                            const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                            const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                            const std::string z = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2);
+                            coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+                        }
+                    }
                     break;
                 }
                 default:
@@ -2086,7 +2125,7 @@ private:
                 std::string op_c;
 
                 const std::string sampler =
-                    GetSampler(instr.sampler, texture_type, false, depth_compare);
+                    GetSampler(instr.sampler, texture_type, is_array, depth_compare);
                 // Add an extra scope and declare the texture coords inside to prevent
                 // overwriting them in case they are used as outputs of the texs instruction.
 
@@ -2106,10 +2145,13 @@ private:
                 }
                 case Tegra::Shader::TextureProcessMode::LB:
                 case Tegra::Shader::TextureProcessMode::LBA: {
-                    if (num_coordinates <= 2) {
-                        op_c = regs.GetRegisterAsFloat(instr.gpr20);
+                    if (depth_compare) {
+                        if (is_array)
+                            op_c = regs.GetRegisterAsFloat(instr.gpr20.Value() + 2);
+                        else
+                            op_c = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
                     } else {
-                        op_c = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
+                        op_c = regs.GetRegisterAsFloat(instr.gpr20);
                     }
                     // TODO: Figure if A suffix changes the equation at all.
                     texture = "texture(" + sampler + ", coords, " + op_c + ')';

src/yuzu/configuration/config.cpp

@@ -85,8 +85,8 @@ void Config::ReadValues() {
     Settings::values.resolution_factor = qt_config->value("resolution_factor", 1.0).toFloat();
     Settings::values.use_frame_limit = qt_config->value("use_frame_limit", true).toBool();
     Settings::values.frame_limit = qt_config->value("frame_limit", 100).toInt();
-    Settings::values.use_accurate_framebuffers =
-        qt_config->value("use_accurate_framebuffers", false).toBool();
+    Settings::values.use_accurate_gpu_emulation =
+        qt_config->value("use_accurate_gpu_emulation", false).toBool();
 
     Settings::values.bg_red = qt_config->value("bg_red", 0.0).toFloat();
     Settings::values.bg_green = qt_config->value("bg_green", 0.0).toFloat();
@@ -233,7 +233,7 @@ void Config::SaveValues() {
     qt_config->setValue("resolution_factor", (double)Settings::values.resolution_factor);
     qt_config->setValue("use_frame_limit", Settings::values.use_frame_limit);
     qt_config->setValue("frame_limit", Settings::values.frame_limit);
-    qt_config->setValue("use_accurate_framebuffers", Settings::values.use_accurate_framebuffers);
+    qt_config->setValue("use_accurate_gpu_emulation", Settings::values.use_accurate_gpu_emulation);
 
     // Cast to double because Qt's written float values are not human-readable
     qt_config->setValue("bg_red", (double)Settings::values.bg_red);

src/yuzu/configuration/configure_graphics.cpp

@@ -75,7 +75,7 @@ void ConfigureGraphics::setConfiguration() {
         static_cast<int>(FromResolutionFactor(Settings::values.resolution_factor)));
     ui->toggle_frame_limit->setChecked(Settings::values.use_frame_limit);
     ui->frame_limit->setValue(Settings::values.frame_limit);
-    ui->use_accurate_framebuffers->setChecked(Settings::values.use_accurate_framebuffers);
+    ui->use_accurate_gpu_emulation->setChecked(Settings::values.use_accurate_gpu_emulation);
     bg_color = QColor::fromRgbF(Settings::values.bg_red, Settings::values.bg_green,
                                 Settings::values.bg_blue);
     ui->bg_button->setStyleSheet(
@@ -87,7 +87,7 @@ void ConfigureGraphics::applyConfiguration() {
         ToResolutionFactor(static_cast<Resolution>(ui->resolution_factor_combobox->currentIndex()));
     Settings::values.use_frame_limit = ui->toggle_frame_limit->isChecked();
     Settings::values.frame_limit = ui->frame_limit->value();
-    Settings::values.use_accurate_framebuffers = ui->use_accurate_framebuffers->isChecked();
+    Settings::values.use_accurate_gpu_emulation = ui->use_accurate_gpu_emulation->isChecked();
     Settings::values.bg_red = static_cast<float>(bg_color.redF());
     Settings::values.bg_green = static_cast<float>(bg_color.greenF());
     Settings::values.bg_blue = static_cast<float>(bg_color.blueF());

src/yuzu/configuration/configure_graphics.ui

@@ -50,9 +50,9 @@
           </layout>
         </item>
         <item>
-         <widget class="QCheckBox" name="use_accurate_framebuffers">
+         <widget class="QCheckBox" name="use_accurate_gpu_emulation">
           <property name="text">
-           <string>Use accurate framebuffers (slow)</string>
+           <string>Use accurate GPU emulation (slow)</string>
           </property>
          </widget>
         </item>

src/yuzu_cmd/config.cpp

@@ -99,8 +99,8 @@ void Config::ReadValues() {
     Settings::values.use_frame_limit = sdl2_config->GetBoolean("Renderer", "use_frame_limit", true);
     Settings::values.frame_limit =
         static_cast<u16>(sdl2_config->GetInteger("Renderer", "frame_limit", 100));
-    Settings::values.use_accurate_framebuffers =
-        sdl2_config->GetBoolean("Renderer", "use_accurate_framebuffers", false);
+    Settings::values.use_accurate_gpu_emulation =
+        sdl2_config->GetBoolean("Renderer", "use_accurate_gpu_emulation", false);
 
     Settings::values.bg_red = (float)sdl2_config->GetReal("Renderer", "bg_red", 0.0);
     Settings::values.bg_green = (float)sdl2_config->GetReal("Renderer", "bg_green", 0.0);

src/yuzu_cmd/default_ini.h

@@ -110,9 +110,9 @@ use_frame_limit =
 # 1 - 9999: Speed limit as a percentage of target game speed. 100 (default)
 frame_limit =
 
-# Whether to use accurate framebuffers
+# Whether to use accurate GPU emulation
 # 0 (default): Off (fast), 1 : On (slow)
-use_accurate_framebuffers =
+use_accurate_gpu_emulation =
 
 # The clear color for the renderer. What shows up on the sides of the bottom screen.
 # Must be in range of 0.0-1.0. Defaults to 1.0 for all.