[TEST] Prune on passthrough stores

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt

@@ -18,7 +18,6 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
     SKIP_CPU_INNER_INVALIDATION("skip_cpu_inner_invalidation"),
     FIX_BLOOM_EFFECTS("fix_bloom_effects"),
     EMULATE_BGR565("emulate_bgr565"),
-    RESCALE_HACK("rescale_hack"),
     CPUOPT_UNSAFE_HOST_MMU("cpuopt_unsafe_host_mmu"),
     USE_DOCKED_MODE("use_docked_mode"),
     USE_AUTO_STUB("use_auto_stub"),

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt

@@ -624,6 +624,7 @@ abstract class SettingsItem(
                     IntSetting.FSR_SHARPENING_SLIDER,
                     titleId = R.string.fsr_sharpness,
                     descriptionId = R.string.fsr_sharpness_description,
+                    max = 200,
                     units = "%"
                 )
             )
@@ -876,13 +877,6 @@ abstract class SettingsItem(
                     descriptionId = R.string.fix_bloom_effects_description
                 )
             )
-            put(
-                SwitchSetting(
-                    BooleanSetting.RESCALE_HACK,
-                    titleId = R.string.rescale_hack,
-                    descriptionId = R.string.rescale_hack_description
-                )
-            )
             put(
                 SwitchSetting(
                     BooleanSetting.EMULATE_BGR565,

src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/EmulationFragment.kt

@@ -1182,7 +1182,7 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
                     container,
                     IntSetting.FSR_SHARPENING_SLIDER,
                     minValue = 0,
-                    maxValue = 100,
+                    maxValue = 200,
                     units = "%"
                 )
             }

src/android/app/src/main/jni/android_config.cpp

@@ -33,6 +33,7 @@ void AndroidConfig::SaveAllValues() {
 }
 
 void AndroidConfig::ReadAndroidValues() {
+    Settings::values.rescale_hack.SetValue(false);
     ReadAndroidUIValues();
     if (global) {
         ReadUIValues();

src/android/app/src/main/res/values-ar/strings.xml

@@ -1088,7 +1088,6 @@
     <string name="theme_mode_light">فاتح</string>
     <string name="theme_mode_dark">داكن</string>
 
-    <string name="multiplier_none">لا شيء</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">خلفيات سوداء</string>

src/android/app/src/main/res/values-de/strings.xml

@@ -1004,7 +1004,6 @@ Wirklich fortfahren?</string>
     <string name="theme_mode_light">Hell</string>
     <string name="theme_mode_dark">Dunkel</string>
 
-    <string name="multiplier_none">Keine</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">Schwarze Hintergründe</string>

src/android/app/src/main/res/values-es/strings.xml

@@ -1080,7 +1080,6 @@
     <string name="theme_mode_light">Claro</string>
     <string name="theme_mode_dark">Oscuro</string>
 
-    <string name="multiplier_none">Nada</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">Fondos oscuros</string>

src/android/app/src/main/res/values-fa/strings.xml

@@ -808,9 +808,6 @@
     <string name="multiplier_x4">x4</string>
     <string name="multiplier_x8">x8</string>
     <string name="multiplier_x16">x16</string>
-    <string name="multiplier_x32">x32</string>
-    <string name="multiplier_x64">x64</string>
-    <string name="multiplier_none">None</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">پس‌زمینه مشکی</string>

src/android/app/src/main/res/values-fr/strings.xml

@@ -1016,7 +1016,6 @@
     <string name="theme_mode_light">Lumineux</string>
     <string name="theme_mode_dark">Sombre</string>
 
-    <string name="multiplier_none">Aucun</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">Arrière-plan noir</string>

src/android/app/src/main/res/values-pl/strings.xml

@@ -948,7 +948,6 @@
     <string name="theme_mode_light">Jasny</string>
     <string name="theme_mode_dark">Ciemny</string>
 
-    <string name="multiplier_none">Brak</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">Czarne tła</string>

src/android/app/src/main/res/values-pt-rBR/strings.xml

@@ -897,7 +897,6 @@
     <string name="theme_mode_light">Claro</string>
     <string name="theme_mode_dark">Escuro</string>
 
-    <string name="multiplier_none">Nenhum</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">Planos de fundo pretos</string>

src/android/app/src/main/res/values-ru/strings.xml

@@ -1084,7 +1084,6 @@
     <string name="theme_mode_light">Светлая</string>
     <string name="theme_mode_dark">Темная</string>
 
-    <string name="multiplier_none">Отключено</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">Чёрный фон</string>

src/android/app/src/main/res/values-uk/strings.xml

@@ -1066,7 +1066,6 @@
     <string name="theme_mode_light">Світла</string>
     <string name="theme_mode_dark">Темна</string>
 
-    <string name="multiplier_none">Жодного</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">Чорний фон</string>

src/android/app/src/main/res/values-zh-rCN/strings.xml

@@ -1078,7 +1078,6 @@
     <string name="theme_mode_light">浅色</string>
     <string name="theme_mode_dark">深色</string>
 
-    <string name="multiplier_none">无</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">使用黑色背景</string>

src/android/app/src/main/res/values-zh-rTW/strings.xml

@@ -926,7 +926,6 @@
     <string name="theme_mode_light">淺色</string>
     <string name="theme_mode_dark">深色</string>
 
-    <string name="multiplier_none">無</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">黑色背景</string>

src/android/app/src/main/res/values/arrays.xml

@@ -513,9 +513,6 @@
         <item>@string/multiplier_x4</item>
         <item>@string/multiplier_x8</item>
         <item>@string/multiplier_x16</item>
-        <item>@string/multiplier_x32</item>
-        <item>@string/multiplier_x64</item>
-        <item>@string/multiplier_none</item>
     </string-array>
     <integer-array name="anisoValues">
         <item>0</item>
@@ -524,9 +521,6 @@
         <item>3</item>
         <item>4</item>
         <item>5</item>
-        <item>6</item>
-        <item>7</item>
-        <item>8</item>
     </integer-array>
 
     <string-array name="verticalAlignmentEntries">

src/android/app/src/main/res/values/strings.xml

@@ -1246,9 +1246,6 @@
     <string name="multiplier_x4" translatable="false">x4</string>
     <string name="multiplier_x8" translatable="false">x8</string>
     <string name="multiplier_x16" translatable="false">x16</string>
-    <string name="multiplier_x32" translatable="false">x32</string>
-    <string name="multiplier_x64" translatable="false">x64</string>
-    <string name="multiplier_none">None</string>
 
     <!-- Black backgrounds theme -->
     <string name="use_black_backgrounds">Black backgrounds</string>

src/common/settings_enums.h

@@ -128,7 +128,7 @@ ENUM(TimeZone, Auto, Default, Cet, Cst6Cdt, Cuba, Eet, Egypt, Eire, Est, Est5Edt
     GmtPlusZero, GmtMinusZero, GmtZero, Greenwich, Hongkong, Hst, Iceland, Iran, Israel, Jamaica,
     Japan, Kwajalein, Libya, Met, Mst, Mst7Mdt, Navajo, Nz, NzChat, Poland, Portugal, Prc, Pst8Pdt,
     Roc, Rok, Singapore, Turkey, Uct, Universal, Utc, WSu, Wet, Zulu);
-ENUM(AnisotropyMode, Automatic, Default, X2, X4, X8, X16, X32, X64, None);
+ENUM(AnisotropyMode, Automatic, Default, X2, X4, X8, X16);
 ENUM(AstcDecodeMode, Cpu, Gpu, CpuAsynchronous);
 ENUM(AstcRecompression, Uncompressed, Bc1, Bc3);
 ENUM(FramePacingMode, Target_Auto, Target_30, Target_60, Target_90, Target_120);

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

@@ -4,6 +4,7 @@
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include <algorithm>
 #include <cstring>
 #include "common/assert.h"
 #include "common/logging.h"
@@ -264,7 +265,7 @@ NvResult nvhost_ctrl_gpu::ZCullGetInfo(IoctlNvgpuGpuZcullGetInfoArgs& params) {
 }
 
 NvResult nvhost_ctrl_gpu::ZBCSetTable(IoctlZbcSetTable& params) {
-    if (params.type > supported_types) {
+    if (params.type == 0 || params.type > supported_types) {
         LOG_ERROR(Service_NVDRV, "ZBCSetTable: invalid type {:#x}", params.type);
         return NvResult::BadParameter;
     }
@@ -279,42 +280,61 @@ NvResult nvhost_ctrl_gpu::ZBCSetTable(IoctlZbcSetTable& params) {
             color_entry.format = params.format;
             color_entry.ref_cnt = 1u;
 
-            auto color_it = std::ranges::find_if(zbc_colors,
-                                                 [&](const ZbcColorEntry& color_in_question) {
-                                                     return color_entry.format == color_in_question.format &&
-                                                            color_entry.color_ds == color_in_question.color_ds &&
-                                                            color_entry.color_l2 == color_in_question.color_l2;
-                                                 });
+            const auto color_end = zbc_colors.begin() + zbc_used_color_entries;
+            auto color_it = std::find_if(zbc_colors.begin(), color_end,
+                                         [&](const ZbcColorEntry& color_in_question) {
+                                             return color_entry.format == color_in_question.format &&
+                                                    color_entry.color_ds == color_in_question.color_ds &&
+                                                    color_entry.color_l2 == color_in_question.color_l2;
+                                         });
 
-            if (color_it != zbc_colors.end()) {
+            if (color_it != color_end) {
                 ++color_it->ref_cnt;
                 LOG_DEBUG(Service_NVDRV, "ZBCSetTable: reused color entry fmt={:#x}, ref_cnt={:#x}",
                           params.format, color_it->ref_cnt);
-            } else {
-                zbc_colors.push_back(color_entry);
-                LOG_DEBUG(Service_NVDRV, "ZBCSetTable: added color entry fmt={:#x}, index={:#x}",
-                          params.format, zbc_colors.size() - 1);
+                break;
             }
+
+            if (zbc_used_color_entries >= zbc_table_size) {
+                LOG_WARNING(Service_NVDRV, "ZBCSetTable: color table is full, fmt={:#x}",
+                            params.format);
+                return NvResult::InsufficientMemory;
+            }
+
+            zbc_colors[zbc_used_color_entries] = color_entry;
+            LOG_DEBUG(Service_NVDRV, "ZBCSetTable: added color entry fmt={:#x}, index={:#x}",
+                      params.format, zbc_used_color_entries);
+            ++zbc_used_color_entries;
             break;
         }
         case ZBCTypes::depth: {
             ZbcDepthEntry depth_entry{params.depth, params.format, 1u};
 
-            auto depth_it = std::ranges::find_if(zbc_depths,
-                                                 [&](const ZbcDepthEntry& depth_entry_in_question) {
-                                                     return depth_entry.format == depth_entry_in_question.format &&
-                                                            depth_entry.depth == depth_entry_in_question.depth;
-                                                 });
+            const auto depth_end = zbc_depths.begin() + zbc_used_depth_entries;
+            auto depth_it = std::find_if(zbc_depths.begin(), depth_end,
+                                         [&](const ZbcDepthEntry& depth_entry_in_question) {
+                                             return depth_entry.format == depth_entry_in_question.format &&
+                                                    depth_entry.depth == depth_entry_in_question.depth;
+                                         });
 
-            if (depth_it != zbc_depths.end()) {
+            if (depth_it != depth_end) {
                 ++depth_it->ref_cnt;
                 LOG_DEBUG(Service_NVDRV, "ZBCSetTable: reused depth entry fmt={:#x}, ref_cnt={:#x}",
                           depth_entry.format, depth_it->ref_cnt);
-            } else {
-                zbc_depths.push_back(depth_entry);
-                LOG_DEBUG(Service_NVDRV, "ZBCSetTable: added depth entry fmt={:#x}, index={:#x}",
-                          depth_entry.format, zbc_depths.size() - 1);
+                break;
             }
+
+            if (zbc_used_depth_entries >= zbc_table_size) {
+                LOG_WARNING(Service_NVDRV, "ZBCSetTable: depth table is full, fmt={:#x}",
+                            depth_entry.format);
+                return NvResult::InsufficientMemory;
+            }
+
+            zbc_depths[zbc_used_depth_entries] = depth_entry;
+            LOG_DEBUG(Service_NVDRV, "ZBCSetTable: added depth entry fmt={:#x}, index={:#x}",
+                      depth_entry.format, zbc_used_depth_entries);
+            ++zbc_used_depth_entries;
+            break;
         }
     }
 
@@ -329,35 +349,34 @@ NvResult nvhost_ctrl_gpu::ZBCQueryTable(IoctlZbcQueryTable& params) {
 
     std::scoped_lock lk(zbc_mutex);
 
+    if (params.type == 0) {
+        params.index_size = zbc_table_size;
+        return NvResult::Success;
+    }
+
+    if (params.index_size >= zbc_table_size) {
+        LOG_ERROR(Service_NVDRV, "ZBCQueryTable: invalid index {:#x}", params.index_size);
+        return NvResult::BadParameter;
+    }
+
     switch (static_cast<ZBCTypes>(params.type)) {
         case ZBCTypes::color: {
-            if (params.index_size >= zbc_colors.size()) {
-                LOG_ERROR(Service_NVDRV, "ZBCQueryTable: invalid color index {:#x}", params.index_size);
-                return NvResult::BadParameter;
-            }
-
             const auto& colors = zbc_colors[params.index_size];
             std::copy_n(colors.color_ds.begin(), colors.color_ds.size(), std::begin(params.color_ds));
             std::copy_n(colors.color_l2.begin(), colors.color_l2.size(), std::begin(params.color_l2));
             params.depth = 0;
             params.ref_cnt = colors.ref_cnt;
             params.format = colors.format;
-            params.index_size = static_cast<u32>(zbc_colors.size());
             break;
         }
         case ZBCTypes::depth: {
-            if (params.index_size >= zbc_depths.size()) {
-                LOG_ERROR(Service_NVDRV, "ZBCQueryTable: invalid depth index {:#x}", params.index_size);
-                return NvResult::BadParameter;
-            }
-
             const auto& depth_entry = zbc_depths[params.index_size];
             std::fill(std::begin(params.color_ds), std::end(params.color_ds), 0);
             std::fill(std::begin(params.color_l2), std::end(params.color_l2), 0);
             params.depth = depth_entry.depth;
             params.ref_cnt = depth_entry.ref_cnt;
             params.format = depth_entry.format;
-            params.index_size = static_cast<u32>(zbc_depths.size());
+            break;
         }
     }
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h

@@ -6,7 +6,7 @@
 
 #pragma once
 
-#include <vector>
+#include <array>
 
 #include "common/common_funcs.h"
 #include "common/common_types.h"
@@ -212,10 +212,14 @@ private:
     Kernel::KEvent* unknown_event;
 
     // ZBC Tables
+    static constexpr u32 zbc_table_size = 15u;
+    static constexpr u32 supported_types = 2u;
+
     std::mutex zbc_mutex{};
-    std::vector<ZbcColorEntry> zbc_colors{};
-    std::vector<ZbcDepthEntry> zbc_depths{};
-    const u32 supported_types = 2u;
+    std::array<ZbcColorEntry, zbc_table_size> zbc_colors{};
+    std::array<ZbcDepthEntry, zbc_table_size> zbc_depths{};
+    u32 zbc_used_color_entries{};
+    u32 zbc_used_depth_entries{};
 };
 
 } // namespace Service::Nvidia::Devices

src/core/hle/service/vi/application_display_service.cpp

@@ -88,25 +88,10 @@ Result IApplicationDisplayService::GetIndirectDisplayTransactionService(
 }
 
 Result IApplicationDisplayService::OpenDisplay(Out<u64> out_display_id, DisplayName display_name) {
-    LOG_DEBUG(Service_VI, "called with display_name={}", display_name.data());
-
     // Ensure the display name is null-terminated
     display_name[display_name.size() - 1] = '\0';
 
-    // According to switchbrew, only "Default", "External", "Edid", "Internal" and "Null" are valid
-    const std::array<std::string_view, 5> valid_names = {
-        "Default", "External", "Edid", "Internal", "Null"
-    };
-
-    bool valid_name = false;
-    for (const auto& name : valid_names) {
-        if (name == display_name.data()) {
-            valid_name = true;
-            break;
-        }
-    }
-
-    R_UNLESS(valid_name, ResultOperationFailed);
+    LOG_DEBUG(Service_VI, "called with display_name={}", display_name.data());
 
     R_RETURN(m_container->OpenDisplay(out_display_id, display_name));
 }

src/qt_common/config/shared_translation.cpp

@@ -522,9 +522,6 @@ std::unique_ptr<ComboboxTranslationMap> ComboboxEnumeration(QObject* parent) {
                               PAIR(AnisotropyMode, X4, tr("4x")),
                               PAIR(AnisotropyMode, X8, tr("8x")),
                               PAIR(AnisotropyMode, X16, tr("16x")),
-                              PAIR(AnisotropyMode, X32, tr("32x")),
-                              PAIR(AnisotropyMode, X64, tr("64x")),
-                              PAIR(AnisotropyMode, None, tr("None")),
                           }});
     translations->insert(
         {Settings::EnumMetadata<Settings::Language>::Index(),

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

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -50,12 +53,6 @@ constexpr std::array RGBA_LUT{
     R | G | B | A, //
 };
 
-void CheckAlignment(IR::Reg reg, size_t alignment) {
-    if (!IR::IsAligned(reg, alignment)) {
-        throw NotImplementedException("Unaligned source register {}", reg);
-    }
-}
-
 template <typename... Args>
 IR::Value Composite(TranslatorVisitor& v, Args... regs) {
     return v.ir.CompositeConstruct(v.F(regs)...);
@@ -86,67 +83,53 @@ IR::Value Sample(TranslatorVisitor& v, u64 insn) {
         info.type.Assign(TextureType::Color2D);
         return v.ir.ImageSampleExplicitLod(handle, Composite(v, reg_a, reg_b), zero, {}, info);
     case 3: // 2D.LL
-        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color2D);
         return v.ir.ImageSampleExplicitLod(handle, Composite(v, reg_a, reg_a + 1), v.F(reg_b), {},
                                            info);
     case 4: // 2D.DC
-        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color2D);
         info.is_depth.Assign(1);
         return v.ir.ImageSampleDrefImplicitLod(handle, Composite(v, reg_a, reg_a + 1), v.F(reg_b),
                                                {}, {}, {}, info);
     case 5: // 2D.LL.DC
-        CheckAlignment(reg_a, 2);
-        CheckAlignment(reg_b, 2);
         info.type.Assign(TextureType::Color2D);
         info.is_depth.Assign(1);
         return v.ir.ImageSampleDrefExplicitLod(handle, Composite(v, reg_a, reg_a + 1),
                                                v.F(reg_b + 1), v.F(reg_b), {}, info);
     case 6: // 2D.LZ.DC
-        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color2D);
         info.is_depth.Assign(1);
         return v.ir.ImageSampleDrefExplicitLod(handle, Composite(v, reg_a, reg_a + 1), v.F(reg_b),
                                                zero, {}, info);
     case 7: // ARRAY_2D
-        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::ColorArray2D);
         return v.ir.ImageSampleImplicitLod(
             handle, v.ir.CompositeConstruct(v.F(reg_a + 1), v.F(reg_b), ReadArray(v, v.X(reg_a))),
             {}, {}, {}, info);
     case 8: // ARRAY_2D.LZ
-        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::ColorArray2D);
         return v.ir.ImageSampleExplicitLod(
             handle, v.ir.CompositeConstruct(v.F(reg_a + 1), v.F(reg_b), ReadArray(v, v.X(reg_a))),
             zero, {}, info);
     case 9: // ARRAY_2D.LZ.DC
-        CheckAlignment(reg_a, 2);
-        CheckAlignment(reg_b, 2);
         info.type.Assign(TextureType::ColorArray2D);
         info.is_depth.Assign(1);
         return v.ir.ImageSampleDrefExplicitLod(
             handle, v.ir.CompositeConstruct(v.F(reg_a + 1), v.F(reg_b), ReadArray(v, v.X(reg_a))),
             v.F(reg_b + 1), zero, {}, info);
     case 10: // 3D
-        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color3D);
         return v.ir.ImageSampleImplicitLod(handle, Composite(v, reg_a, reg_a + 1, reg_b), {}, {},
                                            {}, info);
     case 11: // 3D.LZ
-        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color3D);
         return v.ir.ImageSampleExplicitLod(handle, Composite(v, reg_a, reg_a + 1, reg_b), zero, {},
                                            info);
     case 12: // CUBE
-        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::ColorCube);
         return v.ir.ImageSampleImplicitLod(handle, Composite(v, reg_a, reg_a + 1, reg_b), {}, {},
                                            {}, info);
     case 13: // CUBE.LL
-        CheckAlignment(reg_a, 2);
-        CheckAlignment(reg_b, 2);
         info.type.Assign(TextureType::ColorCube);
         return v.ir.ImageSampleExplicitLod(handle, Composite(v, reg_a, reg_a + 1, reg_b),
                                            v.F(reg_b + 1), {}, info);
@@ -187,12 +170,10 @@ IR::Reg RegStoreComponent32(u64 insn, unsigned index) {
     case 0:
         return texs.dest_reg_a;
     case 1:
-        CheckAlignment(texs.dest_reg_a, 2);
         return texs.dest_reg_a + 1;
     case 2:
         return texs.dest_reg_b;
     case 3:
-        CheckAlignment(texs.dest_reg_b, 2);
         return texs.dest_reg_b + 1;
     }
     throw LogicError("Invalid store index {}", index);

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

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -34,12 +37,6 @@ union Encoding {
     BitField<36, 13, u64> cbuf_offset;
 };
 
-void CheckAlignment(IR::Reg reg, size_t alignment) {
-    if (!IR::IsAligned(reg, alignment)) {
-        throw NotImplementedException("Unaligned source register {}", reg);
-    }
-}
-
 IR::Value MakeOffset(TranslatorVisitor& v, IR::Reg reg) {
     const IR::U32 value{v.X(reg)};
     return v.ir.CompositeConstruct(v.ir.BitFieldExtract(value, v.ir.Imm32(0), v.ir.Imm32(6), true),
@@ -60,18 +57,15 @@ IR::Value Sample(TranslatorVisitor& v, u64 insn) {
     info.is_depth.Assign(tld4s.dc != 0 ? 1 : 0);
     IR::Value coords;
     if (tld4s.aoffi != 0) {
-        CheckAlignment(reg_a, 2);
         coords = v.ir.CompositeConstruct(v.F(reg_a), v.F(reg_a + 1));
         IR::Value offset = MakeOffset(v, reg_b);
         if (tld4s.dc != 0) {
-            CheckAlignment(reg_b, 2);
             IR::F32 dref = v.F(reg_b + 1);
             return v.ir.ImageGatherDref(handle, coords, offset, {}, dref, info);
         }
         return v.ir.ImageGather(handle, coords, offset, {}, info);
     }
     if (tld4s.dc != 0) {
-        CheckAlignment(reg_a, 2);
         coords = v.ir.CompositeConstruct(v.F(reg_a), v.F(reg_a + 1));
         IR::F32 dref = v.F(reg_b);
         return v.ir.ImageGatherDref(handle, coords, {}, {}, dref, info);
@@ -86,12 +80,10 @@ IR::Reg RegStoreComponent32(u64 insn, size_t index) {
     case 0:
         return tlds4.dest_reg_a;
     case 1:
-        CheckAlignment(tlds4.dest_reg_a, 2);
         return tlds4.dest_reg_a + 1;
     case 2:
         return tlds4.dest_reg_b;
     case 3:
-        CheckAlignment(tlds4.dest_reg_b, 2);
         return tlds4.dest_reg_b + 1;
     }
     throw LogicError("Invalid store index {}", index);

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

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -55,12 +58,6 @@ union Encoding {
     BitField<53, 4, u64> encoding;
 };
 
-void CheckAlignment(IR::Reg reg, size_t alignment) {
-    if (!IR::IsAligned(reg, alignment)) {
-        throw NotImplementedException("Unaligned source register {}", reg);
-    }
-}
-
 IR::Value MakeOffset(TranslatorVisitor& v, IR::Reg reg) {
     const IR::U32 value{v.X(reg)};
     return v.ir.CompositeConstruct(v.ir.BitFieldExtract(value, v.ir.Imm32(0), v.ir.Imm32(4), true),
@@ -92,38 +89,31 @@ IR::Value Sample(TranslatorVisitor& v, u64 insn) {
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_b));
         break;
     case 4:
-        CheckAlignment(reg_a, 2);
         texture_type = Shader::TextureType::Color2D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1));
         offsets = MakeOffset(v, reg_b);
         break;
     case 5:
-        CheckAlignment(reg_a, 2);
         texture_type = Shader::TextureType::Color2D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1));
         lod = v.X(reg_b);
         break;
     case 6:
-        CheckAlignment(reg_a, 2);
         texture_type = Shader::TextureType::Color2D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1));
         multisample = v.X(reg_b);
         break;
     case 7:
-        CheckAlignment(reg_a, 2);
         texture_type = Shader::TextureType::Color3D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1), v.X(reg_b));
         break;
     case 8: {
-        CheckAlignment(reg_b, 2);
         const IR::U32 array{v.ir.BitFieldExtract(v.X(reg_a), v.ir.Imm32(0), v.ir.Imm32(16))};
         texture_type = Shader::TextureType::ColorArray2D;
         coords = v.ir.CompositeConstruct(v.X(reg_b), v.X(reg_b + 1), array);
         break;
     }
     case 12:
-        CheckAlignment(reg_a, 2);
-        CheckAlignment(reg_b, 2);
         texture_type = Shader::TextureType::Color2D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1));
         lod = v.X(reg_b);
@@ -166,12 +156,10 @@ IR::Reg RegStoreComponent32(u64 insn, unsigned index) {
     case 0:
         return tlds.dest_reg_a;
     case 1:
-        CheckAlignment(tlds.dest_reg_a, 2);
         return tlds.dest_reg_a + 1;
     case 2:
         return tlds.dest_reg_b;
     case 3:
-        CheckAlignment(tlds.dest_reg_b, 2);
         return tlds.dest_reg_b + 1;
     }
     throw LogicError("Invalid store index {}", index);

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -144,12 +144,15 @@ std::map<IR::Attribute, IR::Attribute> GenerateLegacyToGenericMappings(
             for (size_t i = 0; i < count; ++i) {
                 mapping.insert({attr + i, previous_stage_mapping.at(attr + i)});
             }
-        } else {
-            for (size_t i = 0; i < count; ++i) {
-                mapping.insert({attr + i, unused_generics.front() + i});
-            }
-            unused_generics.pop();
+            return;
         }
+        if (unused_generics.empty()) {
+            return;
+        }
+        for (size_t i = 0; i < count; ++i) {
+            mapping.insert({attr + i, unused_generics.front() + i});
+        }
+        unused_generics.pop();
     };
     for (size_t index = 0; index < 4; ++index) {
         auto attr = IR::Attribute::ColorFrontDiffuseR + index * 4;
@@ -169,31 +172,76 @@ std::map<IR::Attribute, IR::Attribute> GenerateLegacyToGenericMappings(
     return mapping;
 }
 
+struct PassthroughVertices {
+    u32 count;
+    u32 first;
+    u32 stride;
+};
+
+PassthroughVertices GetPassthroughVertices(InputTopology input_topology) {
+    switch (input_topology) {
+    case InputTopology::Points:
+        return {1, 0, 1};
+    case InputTopology::Lines:
+        return {2, 0, 1};
+    case InputTopology::LinesAdjacency:
+        return {2, 1, 1};
+    case InputTopology::Triangles:
+        return {3, 0, 1};
+    case InputTopology::TrianglesAdjacency:
+        return {3, 0, 2};
+    }
+    return {3, 0, 1};
+}
+
 void EmitGeometryPassthrough(IR::IREmitter& ir, const IR::Program& program,
                              const Shader::VaryingState& passthrough_mask,
                              bool passthrough_position,
-                             std::optional<IR::Attribute> passthrough_layer_attr) {
-    for (u32 i = 0; i < program.output_vertices; i++) {
-        // Assign generics from input
-        for (u32 j = 0; j < 32; j++) {
-            if (!passthrough_mask.Generic(j)) {
-                continue;
-            }
-
-            const IR::Attribute attr = IR::Attribute::Generic0X + (j * 4);
+                             std::optional<IR::Attribute> passthrough_layer_attr,
+                             InputTopology input_topology) {
+    const PassthroughVertices vertices{GetPassthroughVertices(input_topology)};
+    for (u32 vertex = 0; vertex < vertices.count; vertex++) {
+        const u32 i = vertices.first + vertex * vertices.stride;
+        const auto copy_vec4{[&ir, i](IR::Attribute attr) {
             ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
             ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
             ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
             ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
+        }};
+
+        // Assign generics from input
+        for (u32 j = 0; j < IR::NUM_GENERICS; j++) {
+            if (!passthrough_mask.Generic(j)) {
+                continue;
+            }
+            copy_vec4(IR::Attribute::Generic0X + (j * 4));
+        }
+
+        for (u32 j = 0; j < 4; j++) {
+            const IR::Attribute attr = IR::Attribute::ColorFrontDiffuseR + (j * 4);
+            if (!passthrough_mask.AnyComponent(attr)) {
+                continue;
+            }
+            copy_vec4(attr);
+        }
+
+        if (passthrough_mask[IR::Attribute::FogCoordinate]) {
+            ir.SetAttribute(IR::Attribute::FogCoordinate,
+                            ir.GetAttribute(IR::Attribute::FogCoordinate, ir.Imm32(i)),
+                            ir.Imm32(0));
+        }
+
+        for (u32 j = 0; j < IR::NUM_FIXEDFNCTEXTURE; j++) {
+            const IR::Attribute attr = IR::Attribute::FixedFncTexture0S + (j * 4);
+            if (!passthrough_mask.AnyComponent(attr)) {
+                continue;
+            }
+            copy_vec4(attr);
         }
 
         if (passthrough_position) {
             // Assign position from input
-            const IR::Attribute attr = IR::Attribute::PositionX;
-            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
+            copy_vec4(IR::Attribute::PositionX);
         }
 
         if (passthrough_layer_attr) {
@@ -208,25 +256,16 @@ void EmitGeometryPassthrough(IR::IREmitter& ir, const IR::Program& program,
     ir.EndPrimitive(ir.Imm32(0));
 }
 
-u32 GetOutputTopologyVertices(OutputTopology output_topology) {
-    switch (output_topology) {
-    case OutputTopology::PointList:
-        return 1;
-    case OutputTopology::LineStrip:
-        return 2;
-    default:
-        return 3;
-    }
-}
-
-void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInfo& host_info) {
+void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInfo& host_info,
+                              InputTopology input_topology) {
     for (IR::Block* const block : program.blocks) {
         for (IR::Inst& inst : block->Instructions()) {
             if (inst.GetOpcode() == IR::Opcode::Epilogue) {
                 IR::IREmitter ir{*block, IR::Block::InstructionList::s_iterator_to(inst)};
                 EmitGeometryPassthrough(
                     ir, program, program.info.passthrough,
-                    program.info.passthrough.AnyComponent(IR::Attribute::PositionX), {});
+                    program.info.passthrough.AnyComponent(IR::Attribute::PositionX), {},
+                    input_topology);
             }
         }
     }
@@ -235,7 +274,8 @@ void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInf
 } // Anonymous namespace
 
 IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
-                             Environment& env, Flow::CFG& cfg, const HostTranslateInfo& host_info) {
+                             Environment& env, Flow::CFG& cfg, const HostTranslateInfo& host_info,
+                             InputTopology input_topology) {
     HostTranslateInfo normalized_host_info{host_info};
     normalized_host_info.ApplyDescriptorLimitPolicy();
 
@@ -264,8 +304,9 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
             }
 
             if (!normalized_host_info.support_geometry_shader_passthrough) {
-                program.output_vertices = GetOutputTopologyVertices(program.output_topology);
-                LowerGeometryPassthrough(program, normalized_host_info);
+                program.output_vertices = GetPassthroughVertices(input_topology).count;
+                LowerGeometryPassthrough(program, normalized_host_info, input_topology);
+                program.is_geometry_passthrough = false;
             }
         }
         break;
@@ -350,6 +391,52 @@ IR::Program MergeDualVertexPrograms(IR::Program& vertex_a, IR::Program& vertex_b
     return result;
 }
 
+void PrunePassthroughStores(IR::Program& program, const VaryingState& previous_stage_stores) {
+    if (program.stage != Stage::Geometry || program.is_geometry_passthrough) {
+        return;
+    }
+    VaryingState pruned;
+    for (size_t index = 0; index < program.info.passthrough.mask.size(); ++index) {
+        if (!program.info.passthrough.mask[index] || previous_stage_stores.mask[index]) {
+            continue;
+        }
+        const IR::Attribute attr{static_cast<IR::Attribute>(index)};
+        if (attr >= IR::Attribute::PositionX && attr <= IR::Attribute::PositionW) {
+            continue;
+        }
+        pruned.mask[index] = true;
+    }
+    if (pruned.mask.none()) {
+        return;
+    }
+    const auto erase_matching{[&pruned](IR::Block* block, IR::Opcode opcode, bool skip_used) {
+        auto it{block->begin()};
+        while (it != block->end()) {
+            IR::Inst& inst{*it};
+            if (inst.GetOpcode() != opcode ||
+                !pruned.mask[static_cast<size_t>(inst.Arg(0).Attribute())]) {
+                ++it;
+                continue;
+            }
+            if (skip_used && inst.HasUses()) {
+                ++it;
+                continue;
+            }
+            inst.Invalidate();
+            it = block->Instructions().erase(it);
+        }
+    }};
+    for (IR::Block* const block : program.post_order_blocks) {
+        erase_matching(block, IR::Opcode::SetAttribute, false);
+    }
+    for (IR::Block* const block : program.post_order_blocks) {
+        erase_matching(block, IR::Opcode::GetAttribute, true);
+    }
+    program.info.stores.mask &= ~pruned.mask;
+    program.info.loads.mask &= ~pruned.mask;
+    program.info.passthrough.mask &= ~pruned.mask;
+}
+
 void ConvertLegacyToGeneric(IR::Program& program, const Shader::RuntimeInfo& runtime_info) {
     auto& stores = program.info.stores;
     if (stores.Legacy()) {
@@ -362,19 +449,24 @@ void ConvertLegacyToGeneric(IR::Program& program, const Shader::RuntimeInfo& run
         program.info.legacy_stores_mapping =
             GenerateLegacyToGenericMappings(stores, unused_output_generics, {});
         for (IR::Block* const block : program.post_order_blocks) {
-            for (IR::Inst& inst : block->Instructions()) {
-                switch (inst.GetOpcode()) {
-                case IR::Opcode::SetAttribute: {
-                    const auto attr = inst.Arg(0).Attribute();
-                    if (IsLegacyAttribute(attr)) {
-                        stores.Set(program.info.legacy_stores_mapping[attr], true);
-                        inst.SetArg(0, Shader::IR::Value(program.info.legacy_stores_mapping[attr]));
-                    }
-                    break;
+            auto it{block->begin()};
+            while (it != block->end()) {
+                IR::Inst& inst{*it};
+                if (inst.GetOpcode() != IR::Opcode::SetAttribute ||
+                    !IsLegacyAttribute(inst.Arg(0).Attribute())) {
+                    ++it;
+                    continue;
                 }
-                default:
-                    break;
+                const auto& mapping{program.info.legacy_stores_mapping};
+                const auto mapped{mapping.find(inst.Arg(0).Attribute())};
+                if (mapped == mapping.end()) {
+                    inst.Invalidate();
+                    it = block->Instructions().erase(it);
+                    continue;
                 }
+                stores.Set(mapped->second, true);
+                inst.SetArg(0, Shader::IR::Value(mapped->second));
+                ++it;
             }
         }
     }
@@ -396,10 +488,16 @@ void ConvertLegacyToGeneric(IR::Program& program, const Shader::RuntimeInfo& run
                 switch (inst.GetOpcode()) {
                 case IR::Opcode::GetAttribute: {
                     const auto attr = inst.Arg(0).Attribute();
-                    if (IsLegacyAttribute(attr)) {
-                        loads.Set(mappings[attr], true);
-                        inst.SetArg(0, Shader::IR::Value(mappings[attr]));
+                    if (!IsLegacyAttribute(attr)) {
+                        break;
                     }
+                    const auto mapped{mappings.find(attr)};
+                    if (mapped == mappings.end()) {
+                        inst.ReplaceUsesWith(IR::Value{0.0f});
+                        break;
+                    }
+                    loads.Set(mapped->second, true);
+                    inst.SetArg(0, Shader::IR::Value(mapped->second));
                     break;
                 }
                 default:
@@ -414,11 +512,12 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
                                         ObjectPool<IR::Block>& block_pool,
                                         const HostTranslateInfo& host_info,
                                         IR::Program& source_program,
-                                        Shader::OutputTopology output_topology) {
+                                        Shader::OutputTopology output_topology,
+                                        InputTopology input_topology) {
     IR::Program program;
     program.stage = Stage::Geometry;
     program.output_topology = output_topology;
-    program.output_vertices = GetOutputTopologyVertices(output_topology);
+    program.output_vertices = GetPassthroughVertices(input_topology).count;
 
     program.is_geometry_passthrough = false;
     program.info.loads.mask = source_program.info.stores.mask;
@@ -433,7 +532,7 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
 
     IR::IREmitter ir{*current_block};
     EmitGeometryPassthrough(ir, program, program.info.stores, true,
-                            source_program.info.emulated_layer);
+                            source_program.info.emulated_layer, input_topology);
 
     IR::Block* return_block{block_pool.Create(inst_pool)};
     IR::IREmitter{*return_block}.Epilogue();

src/shader_recompiler/frontend/maxwell/translate_program.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -18,13 +21,16 @@ namespace Shader::Maxwell {
 
 [[nodiscard]] IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool,
                                            ObjectPool<IR::Block>& block_pool, Environment& env,
-                                           Flow::CFG& cfg, const HostTranslateInfo& host_info);
+                                           Flow::CFG& cfg, const HostTranslateInfo& host_info,
+                                           InputTopology input_topology);
 
 [[nodiscard]] IR::Program MergeDualVertexPrograms(IR::Program& vertex_a, IR::Program& vertex_b,
                                                   Environment& env_vertex_b);
 
 void ConvertLegacyToGeneric(IR::Program& program, const RuntimeInfo& runtime_info);
 
+void PrunePassthroughStores(IR::Program& program, const VaryingState& previous_stage_stores);
+
 // Maxwell v1 and older Nvidia cards don't support setting gl_Layer from non-geometry stages.
 // This creates a workaround by setting the layer as a generic output and creating a
 // passthrough geometry shader that reads the generic and sets the layer.
@@ -32,6 +38,7 @@ void ConvertLegacyToGeneric(IR::Program& program, const RuntimeInfo& runtime_inf
                                                       ObjectPool<IR::Block>& block_pool,
                                                       const HostTranslateInfo& host_info,
                                                       IR::Program& source_program,
-                                                      Shader::OutputTopology output_topology);
+                                                      Shader::OutputTopology output_topology,
+                                                      InputTopology input_topology);
 
 } // namespace Shader::Maxwell

src/video_core/buffer_cache/buffer_cache.h

@@ -7,6 +7,7 @@
 #pragma once
 
 #include <algorithm>
+#include <bit>
 #include <memory>
 #include <numeric>
 
@@ -809,46 +810,46 @@ void BufferCache<P>::BindHostVertexBuffers() {
 
     if (use_optimized_vertex_buffers) {
         auto& flags = maxwell3d->dirty.flags;
-        u32 enabled_mask = enabled_vertex_buffers_mask;
-        HostBindings<Buffer> bindings{};
-        u32 last_index = (std::numeric_limits<u32>::max)();
-        const auto flush_bindings = [&]() {
-            if (bindings.buffers.empty()) {
-                return;
-            }
-            bindings.max_index = bindings.min_index + static_cast<u32>(bindings.buffers.size());
-            runtime.BindVertexBuffers(bindings);
-            bindings = HostBindings<Buffer>{};
-            last_index = (std::numeric_limits<u32>::max)();
-        };
-        while (enabled_mask != 0) {
-            const u32 index = std::countr_zero(enabled_mask);
-            enabled_mask &= (enabled_mask - 1);
+        const u32 enabled_mask = enabled_vertex_buffers_mask;
+        bool any_dirty = false;
+        u32 pending_mask = enabled_mask;
+        while (pending_mask != 0) {
+            const u32 index = std::countr_zero(pending_mask);
+            pending_mask &= (pending_mask - 1);
             const Binding& binding = VertexBufferSlot(index);
             Buffer& buffer = slot_buffers[binding.buffer_id];
             TouchBuffer(buffer, binding.buffer_id);
             SynchronizeBuffer(buffer, binding.device_addr, binding.size);
-            if (!flags[Dirty::VertexBuffer0 + index]) {
-                flush_bindings();
-                continue;
-            }
+            any_dirty |= flags[Dirty::VertexBuffer0 + index];
+        }
+        if (enabled_mask == 0 || !any_dirty) {
+            return;
+        }
+        const u32 min_index = static_cast<u32>(std::countr_zero(enabled_mask));
+        const u32 max_index = 32u - static_cast<u32>(std::countl_zero(enabled_mask));
+        HostBindings<Buffer> bindings{};
+        bindings.min_index = min_index;
+        bindings.max_index = max_index;
+        for (u32 index = min_index; index < max_index; ++index) {
             flags[Dirty::VertexBuffer0 + index] = false;
             const u32 stride = maxwell3d->regs.vertex_streams[index].stride;
+            if ((enabled_mask & (1u << index)) == 0) {
+                bindings.buffers.push_back(&slot_buffers[NULL_BUFFER_ID]);
+                bindings.offsets.push_back(0);
+                bindings.sizes.push_back(0);
+                bindings.strides.push_back(stride);
+                continue;
+            }
+            const Binding& binding = VertexBufferSlot(index);
+            Buffer& buffer = slot_buffers[binding.buffer_id];
             const u32 offset = buffer.Offset(binding.device_addr);
             buffer.MarkUsage(offset, binding.size);
-            if (!bindings.buffers.empty() && index != last_index + 1) {
-                flush_bindings();
-            }
-            if (bindings.buffers.empty()) {
-                bindings.min_index = index;
-            }
             bindings.buffers.push_back(&buffer);
             bindings.offsets.push_back(offset);
             bindings.sizes.push_back(binding.size);
             bindings.strides.push_back(stride);
-            last_index = index;
         }
-        flush_bindings();
+        runtime.BindVertexBuffers(bindings);
     } else {
         HostBindings<typename P::Buffer> host_bindings;
         bool any_valid{false};

src/video_core/host_shaders/CMakeLists.txt

@@ -17,11 +17,14 @@ set(SHADER_FILES
     ${CMAKE_CURRENT_SOURCE_DIR}/astc_decoder.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_color_float.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_2d.comp
+    ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_2d_buffer.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_color_msaa.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/blit_depth.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_depth_msaa.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_depth_stencil_msaa.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_3d.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_3d_bcn.comp
+    ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_3d_buffer.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_abgr8_to_d24s8.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_abgr8_to_d32f.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_d32f_to_abgr8.frag
@@ -30,8 +33,16 @@ set(SHADER_FILES
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_float_to_depth.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa_depth.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa_depth_stencil.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa_sint.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa_uint.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa_sint.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa_uint.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa_depth.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa_depth_stencil.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_s8d24_to_abgr8.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/full_screen_triangle.vert
     ${CMAKE_CURRENT_SOURCE_DIR}/fxaa.frag

src/video_core/host_shaders/blit_depth.frag

@@ -0,0 +1,12 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+
+layout(binding = 0) uniform sampler2D depth_tex;
+
+layout(location = 0) in vec2 texcoord;
+
+void main() {
+    gl_FragDepth = textureLod(depth_tex, texcoord, 0).r;
+}

src/video_core/host_shaders/blit_depth_msaa.frag

@@ -8,5 +8,5 @@ layout(binding = 0) uniform sampler2DMS depth_tex;
 layout(location = 0) in vec2 texcoord;
 
 void main() {
-    gl_FragDepth = texelFetch(depth_tex, ivec2(texcoord), 0).r;
+    gl_FragDepth = texelFetch(depth_tex, ivec2(texcoord), gl_SampleID).r;
 }

src/video_core/host_shaders/blit_depth_stencil_msaa.frag

@@ -10,6 +10,6 @@ layout(binding = 1) uniform usampler2DMS stencil_tex;
 layout(location = 0) in vec2 texcoord;
 
 void main() {
-    gl_FragDepth = texelFetch(depth_tex, ivec2(texcoord), 0).r;
-    gl_FragStencilRefARB = int(texelFetch(stencil_tex, ivec2(texcoord), 0).r);
+    gl_FragDepth = texelFetch(depth_tex, ivec2(texcoord), gl_SampleID).r;
+    gl_FragStencilRefARB = int(texelFetch(stencil_tex, ivec2(texcoord), gl_SampleID).r);
 }

src/video_core/host_shaders/block_linear_unswizzle_2d_buffer.comp

@@ -0,0 +1,104 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 430
+
+#extension GL_EXT_shader_16bit_storage : require
+#extension GL_EXT_shader_8bit_storage : require
+
+#define BINDING_INPUT_BUFFER 0
+#define BINDING_OUTPUT_BUFFER 1
+
+layout(push_constant) uniform PushConstants {
+    uvec3 dim;
+    uint bytes_per_block_log2;
+
+    uvec3 origin;
+    uint layer_stride;
+
+    uint block_size;
+    uint x_shift;
+    uint block_height;
+    uint block_height_mask;
+} pc;
+
+layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU32 { uint u32data[]; };
+layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU64 { uvec2 u64data[]; };
+layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU128 { uvec4 u128data[]; };
+
+layout(binding = BINDING_OUTPUT_BUFFER, std430) writeonly buffer OutputBuffer {
+    uint out_u32[];
+};
+
+layout(local_size_x = 16, local_size_y = 8, local_size_z = 1) in;
+
+const uint GOB_SIZE_X = 64;
+const uint GOB_SIZE_Y = 8;
+
+const uint GOB_SIZE_X_SHIFT = 6;
+const uint GOB_SIZE_Y_SHIFT = 3;
+const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT;
+
+const uvec2 SWIZZLE_MASK = uvec2(GOB_SIZE_X - 1u, GOB_SIZE_Y - 1u);
+
+uint SwizzleTable(uint pos) {
+    const uint t[8] = uint[](
+        0x12100200, 0x13110301, 0x16140604, 0x17150705,
+        0x1a180a08, 0x1b190b09, 0x1e1c0e0c, 0x1f1d0f0d
+    );
+    const uint i = pos >> 4;
+    const uint h = (t[i / 4] >> ((i % 4) * 8)) & 0xff;
+    return (h << 4) | (pos & 0xf);
+}
+
+uint SwizzleOffset(uvec2 pos) {
+    pos = pos & SWIZZLE_MASK;
+    return SwizzleTable(pos.y * 64u + pos.x);
+}
+
+uvec4 ReadTexel(uint offset) {
+    switch (pc.bytes_per_block_log2) {
+    case 2u:
+        return uvec4(u32data[offset / 4u], 0u, 0u, 0u);
+    case 3u:
+        return uvec4(u64data[offset / 8u], 0u, 0u);
+    case 4u:
+        return u128data[offset / 16u];
+    }
+    return uvec4(0u);
+}
+
+void main() {
+    uvec3 coord = gl_GlobalInvocationID;
+    if (coord.x >= pc.dim.x || coord.y >= pc.dim.y || coord.z >= pc.dim.z) {
+        return;
+    }
+
+    uvec3 pos = coord + pc.origin;
+    pos.x <<= pc.bytes_per_block_log2;
+
+    uint swizzle = SwizzleOffset(pos.xy);
+    uint block_y = pos.y >> GOB_SIZE_Y_SHIFT;
+
+    uint offset = 0u;
+    offset += pos.z * pc.layer_stride;
+    offset += (block_y >> pc.block_height) * pc.block_size;
+    offset += (block_y & pc.block_height_mask) << GOB_SIZE_SHIFT;
+    offset += (pos.x >> GOB_SIZE_X_SHIFT) << pc.x_shift;
+    offset += swizzle;
+
+    uvec4 texel = ReadTexel(offset);
+
+    uint words = 1u << (pc.bytes_per_block_log2 - 2u);
+    uint linear_index = coord.x + coord.y * pc.dim.x + coord.z * pc.dim.x * pc.dim.y;
+    uint out_idx = linear_index * words;
+
+    out_u32[out_idx] = texel.x;
+    if (words > 1u) {
+        out_u32[out_idx + 1u] = texel.y;
+    }
+    if (words > 2u) {
+        out_u32[out_idx + 2u] = texel.z;
+        out_u32[out_idx + 3u] = texel.w;
+    }
+}

src/video_core/host_shaders/block_linear_unswizzle_3d_buffer.comp

@@ -0,0 +1,105 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 430
+
+#define BINDING_INPUT_BUFFER 0
+#define BINDING_OUTPUT_BUFFER 1
+
+layout(push_constant) uniform PushConstants {
+    uvec3 dim;
+    uint bytes_per_block_log2;
+
+    uvec3 origin;
+    uint slice_size;
+
+    uint block_size;
+    uint x_shift;
+    uint block_height;
+    uint block_height_mask;
+
+    uint block_depth;
+    uint block_depth_mask;
+} pc;
+
+layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU32 { uint u32data[]; };
+layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU64 { uvec2 u64data[]; };
+layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU128 { uvec4 u128data[]; };
+
+layout(binding = BINDING_OUTPUT_BUFFER, std430) writeonly buffer OutputBuffer {
+    uint out_u32[];
+};
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 4) in;
+
+const uint GOB_SIZE_X = 64;
+const uint GOB_SIZE_Y = 8;
+
+const uint GOB_SIZE_X_SHIFT = 6;
+const uint GOB_SIZE_Y_SHIFT = 3;
+const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT;
+
+const uvec2 SWIZZLE_MASK = uvec2(GOB_SIZE_X - 1u, GOB_SIZE_Y - 1u);
+
+uint SwizzleTable(uint pos) {
+    const uint t[8] = uint[](
+        0x12100200, 0x13110301, 0x16140604, 0x17150705,
+        0x1a180a08, 0x1b190b09, 0x1e1c0e0c, 0x1f1d0f0d
+    );
+    const uint i = pos >> 4;
+    const uint h = (t[i / 4] >> ((i % 4) * 8)) & 0xff;
+    return (h << 4) | (pos & 0xf);
+}
+
+uint SwizzleOffset(uvec2 pos) {
+    pos = pos & SWIZZLE_MASK;
+    return SwizzleTable(pos.y * 64u + pos.x);
+}
+
+uvec4 ReadTexel(uint offset) {
+    switch (pc.bytes_per_block_log2) {
+    case 2u:
+        return uvec4(u32data[offset / 4u], 0u, 0u, 0u);
+    case 3u:
+        return uvec4(u64data[offset / 8u], 0u, 0u);
+    case 4u:
+        return u128data[offset / 16u];
+    }
+    return uvec4(0u);
+}
+
+void main() {
+    uvec3 coord = gl_GlobalInvocationID;
+    if (coord.x >= pc.dim.x || coord.y >= pc.dim.y || coord.z >= pc.dim.z) {
+        return;
+    }
+
+    uvec3 pos = coord + pc.origin;
+    pos.x <<= pc.bytes_per_block_log2;
+
+    uint swizzle = SwizzleOffset(pos.xy);
+    uint block_y = pos.y >> GOB_SIZE_Y_SHIFT;
+
+    uint offset = 0u;
+    offset += (pos.z >> pc.block_depth) * pc.slice_size;
+    offset += (pos.z & pc.block_depth_mask) << (GOB_SIZE_SHIFT + pc.block_height);
+    offset += (block_y >> pc.block_height) * pc.block_size;
+    offset += (block_y & pc.block_height_mask) << GOB_SIZE_SHIFT;
+    offset += (pos.x >> GOB_SIZE_X_SHIFT) << pc.x_shift;
+    offset += swizzle;
+
+    uvec4 texel = ReadTexel(offset);
+
+    uint words = 1u << (pc.bytes_per_block_log2 - 2u);
+    uint linear_index = coord.x + coord.y * pc.dim.x + coord.z * pc.dim.x * pc.dim.y;
+    uint out_idx = linear_index * words;
+
+    out_u32[out_idx] = texel.x;
+    if (words > 1u) {
+        out_u32[out_idx + 1u] = texel.y;
+    }
+    if (words > 2u) {
+        out_u32[out_idx + 2u] = texel.z;
+        out_u32[out_idx + 3u] = texel.w;
+    }
+}

src/video_core/host_shaders/convert_msaa_to_non_msaa_depth.frag

@@ -0,0 +1,20 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+
+layout(binding = 0) uniform sampler2DMS msaa_in;
+
+layout(push_constant) uniform PushConstants {
+    ivec2 dst_offset;
+    ivec2 src_offset;
+    ivec2 scale;
+};
+
+void main() {
+    const ivec2 coord = ivec2(gl_FragCoord.xy) - dst_offset + src_offset;
+    const ivec2 msaa_coord = coord / scale;
+    const ivec2 sample_offset = coord % scale;
+    const int sample_id = sample_offset.x + scale.x * sample_offset.y;
+    gl_FragDepth = texelFetch(msaa_in, msaa_coord, sample_id).r;
+}

src/video_core/host_shaders/convert_msaa_to_non_msaa_depth_stencil.frag

@@ -0,0 +1,23 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+#extension GL_ARB_shader_stencil_export : require
+
+layout(binding = 0) uniform sampler2DMS depth_tex;
+layout(binding = 1) uniform usampler2DMS stencil_tex;
+
+layout(push_constant) uniform PushConstants {
+    ivec2 dst_offset;
+    ivec2 src_offset;
+    ivec2 scale;
+};
+
+void main() {
+    const ivec2 coord = ivec2(gl_FragCoord.xy) - dst_offset + src_offset;
+    const ivec2 msaa_coord = coord / scale;
+    const ivec2 sample_offset = coord % scale;
+    const int sample_id = sample_offset.x + scale.x * sample_offset.y;
+    gl_FragDepth = texelFetch(depth_tex, msaa_coord, sample_id).r;
+    gl_FragStencilRefARB = int(texelFetch(stencil_tex, msaa_coord, sample_id).r);
+}

src/video_core/host_shaders/convert_msaa_to_non_msaa_sint.frag

@@ -0,0 +1,22 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+
+layout(binding = 0) uniform isampler2DMS msaa_in;
+
+layout(push_constant) uniform PushConstants {
+    ivec2 dst_offset;
+    ivec2 src_offset;
+    ivec2 scale;
+};
+
+layout(location = 0) out ivec4 frag_color;
+
+void main() {
+    const ivec2 coord = ivec2(gl_FragCoord.xy) - dst_offset + src_offset;
+    const ivec2 msaa_coord = coord / scale;
+    const ivec2 sample_offset = coord % scale;
+    const int sample_id = sample_offset.x + scale.x * sample_offset.y;
+    frag_color = texelFetch(msaa_in, msaa_coord, sample_id);
+}

src/video_core/host_shaders/convert_msaa_to_non_msaa_uint.frag

@@ -0,0 +1,22 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+
+layout(binding = 0) uniform usampler2DMS msaa_in;
+
+layout(push_constant) uniform PushConstants {
+    ivec2 dst_offset;
+    ivec2 src_offset;
+    ivec2 scale;
+};
+
+layout(location = 0) out uvec4 frag_color;
+
+void main() {
+    const ivec2 coord = ivec2(gl_FragCoord.xy) - dst_offset + src_offset;
+    const ivec2 msaa_coord = coord / scale;
+    const ivec2 sample_offset = coord % scale;
+    const int sample_id = sample_offset.x + scale.x * sample_offset.y;
+    frag_color = texelFetch(msaa_in, msaa_coord, sample_id);
+}

src/video_core/host_shaders/convert_non_msaa_to_msaa_depth.frag

@@ -0,0 +1,19 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+
+layout(binding = 0) uniform sampler2D img_in;
+
+layout(push_constant) uniform PushConstants {
+    ivec2 dst_offset;
+    ivec2 src_offset;
+    ivec2 scale;
+};
+
+void main() {
+    const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
+    const ivec2 sample_offset = ivec2(gl_SampleID % scale.x, gl_SampleID / scale.x);
+    const ivec2 coord = msaa_coord * scale + sample_offset + src_offset;
+    gl_FragDepth = texelFetch(img_in, coord, 0).r;
+}

src/video_core/host_shaders/convert_non_msaa_to_msaa_depth_stencil.frag

@@ -0,0 +1,22 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+#extension GL_ARB_shader_stencil_export : require
+
+layout(binding = 0) uniform sampler2D depth_tex;
+layout(binding = 1) uniform usampler2D stencil_tex;
+
+layout(push_constant) uniform PushConstants {
+    ivec2 dst_offset;
+    ivec2 src_offset;
+    ivec2 scale;
+};
+
+void main() {
+    const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
+    const ivec2 sample_offset = ivec2(gl_SampleID % scale.x, gl_SampleID / scale.x);
+    const ivec2 coord = msaa_coord * scale + sample_offset + src_offset;
+    gl_FragDepth = texelFetch(depth_tex, coord, 0).r;
+    gl_FragStencilRefARB = int(texelFetch(stencil_tex, coord, 0).r);
+}

src/video_core/host_shaders/convert_non_msaa_to_msaa_sint.frag

@@ -0,0 +1,21 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+
+layout(binding = 0) uniform isampler2D img_in;
+
+layout(push_constant) uniform PushConstants {
+    ivec2 dst_offset;
+    ivec2 src_offset;
+    ivec2 scale;
+};
+
+layout(location = 0) out ivec4 frag_color;
+
+void main() {
+    const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
+    const ivec2 sample_offset = ivec2(gl_SampleID % scale.x, gl_SampleID / scale.x);
+    const ivec2 coord = msaa_coord * scale + sample_offset + src_offset;
+    frag_color = texelFetch(img_in, coord, 0);
+}

src/video_core/host_shaders/convert_non_msaa_to_msaa_uint.frag

@@ -0,0 +1,21 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450 core
+
+layout(binding = 0) uniform usampler2D img_in;
+
+layout(push_constant) uniform PushConstants {
+    ivec2 dst_offset;
+    ivec2 src_offset;
+    ivec2 scale;
+};
+
+layout(location = 0) out uvec4 frag_color;
+
+void main() {
+    const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
+    const ivec2 sample_offset = ivec2(gl_SampleID % scale.x, gl_SampleID / scale.x);
+    const ivec2 coord = msaa_coord * scale + sample_offset + src_offset;
+    frag_color = texelFetch(img_in, coord, 0);
+}

src/video_core/host_shaders/sgsr1_shader.vert

@@ -7,6 +7,7 @@ layout(push_constant) uniform constants {
     vec2 scale;
     vec2 size;
     vec2 resize_factor;
+    vec2 crop_offset;
     float edge_sharpness;
 };
 layout(location = 0) out highp vec2 texcoord;
@@ -15,5 +16,5 @@ void main() {
     float x = float((gl_VertexIndex & 1) << 2);
     float y = float((gl_VertexIndex & 2) << 1);
     gl_Position = vec4(x - 1.0f, y - 1.0f, 0.0, 1.0f) * vec4(sign(resize_factor), 1.f, 1.f);
-    texcoord = vec2(x, y) * abs(resize_factor) * 0.5;
+    texcoord = crop_offset + vec2(x, y) * abs(resize_factor) * 0.5;
 }

src/video_core/host_shaders/sgsr1_shader_mobile.frag

@@ -14,6 +14,7 @@ layout(push_constant) uniform constants {
     vec2 scale;
     vec2 size;
     vec2 resize_factor;
+    vec2 crop_offset;
     float edge_sharpness;
 };
 layout(set = 0, binding = 0) uniform sampler2D sampler0;

src/video_core/host_shaders/sgsr1_shader_mobile_edge_direction.frag

@@ -13,6 +13,7 @@
 layout( push_constant ) uniform constants {
     vec4 ViewportInfo[1];
     vec2 ResizeFactor;
+    vec2 CropOffset;
     float EdgeSharpness;
 };
 layout(set = 0, binding = 0) uniform sampler2D ps0;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -45,6 +45,7 @@ using Shader::Backend::SPIRV::EmitSPIRV;
 using Shader::Maxwell::ConvertLegacyToGeneric;
 using Shader::Maxwell::GenerateGeometryPassthrough;
 using Shader::Maxwell::MergeDualVertexPrograms;
+using Shader::Maxwell::PrunePassthroughStores;
 using Shader::Maxwell::TranslateProgram;
 using VideoCommon::ComputeEnvironment;
 using VideoCommon::FileEnvironment;
@@ -54,7 +55,7 @@ using VideoCommon::LoadPipelines;
 using VideoCommon::SerializePipeline;
 using Context = ShaderContext::Context;
 
-constexpr u32 CACHE_VERSION = 15;
+constexpr u32 CACHE_VERSION = 18;
 
 template <typename Container>
 auto MakeSpan(Container& container) {
@@ -72,6 +73,25 @@ Shader::OutputTopology MaxwellToOutputTopology(Maxwell::PrimitiveTopology topolo
     }
 }
 
+Shader::InputTopology MaxwellToInputTopology(Maxwell::PrimitiveTopology topology) {
+    switch (topology) {
+    case Maxwell::PrimitiveTopology::Points:
+        return Shader::InputTopology::Points;
+    case Maxwell::PrimitiveTopology::Lines:
+    case Maxwell::PrimitiveTopology::LineLoop:
+    case Maxwell::PrimitiveTopology::LineStrip:
+        return Shader::InputTopology::Lines;
+    case Maxwell::PrimitiveTopology::LinesAdjacency:
+    case Maxwell::PrimitiveTopology::LineStripAdjacency:
+        return Shader::InputTopology::LinesAdjacency;
+    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
+    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
+        return Shader::InputTopology::TrianglesAdjacency;
+    default:
+        return Shader::InputTopology::Triangles;
+    }
+}
+
 Shader::RuntimeInfo MakeRuntimeInfo(const GraphicsPipelineKey& key,
                                     const Shader::IR::Program& program,
                                     const Shader::IR::Program* previous_program,
@@ -127,33 +147,7 @@ Shader::RuntimeInfo MakeRuntimeInfo(const GraphicsPipelineKey& key,
     default:
         break;
     }
-    switch (key.gs_input_topology) {
-    case Maxwell::PrimitiveTopology::Points:
-        info.input_topology = Shader::InputTopology::Points;
-        break;
-    case Maxwell::PrimitiveTopology::Lines:
-    case Maxwell::PrimitiveTopology::LineLoop:
-    case Maxwell::PrimitiveTopology::LineStrip:
-        info.input_topology = Shader::InputTopology::Lines;
-        break;
-    case Maxwell::PrimitiveTopology::Triangles:
-    case Maxwell::PrimitiveTopology::TriangleStrip:
-    case Maxwell::PrimitiveTopology::TriangleFan:
-    case Maxwell::PrimitiveTopology::Quads:
-    case Maxwell::PrimitiveTopology::QuadStrip:
-    case Maxwell::PrimitiveTopology::Polygon:
-    case Maxwell::PrimitiveTopology::Patches:
-        info.input_topology = Shader::InputTopology::Triangles;
-        break;
-    case Maxwell::PrimitiveTopology::LinesAdjacency:
-    case Maxwell::PrimitiveTopology::LineStripAdjacency:
-        info.input_topology = Shader::InputTopology::LinesAdjacency;
-        break;
-    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
-    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
-        info.input_topology = Shader::InputTopology::TrianglesAdjacency;
-        break;
-    }
+    info.input_topology = MaxwellToInputTopology(key.gs_input_topology);
     info.glasm_use_storage_buffers = glasm_use_storage_buffers;
     return info;
 }
@@ -483,8 +477,10 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
             && index == u32(Maxwell::ShaderType::Geometry);
         if (key.unique_hashes[index] == 0 && is_emulated_stage) {
             auto topology = MaxwellToOutputTopology(key.gs_input_topology);
-            programs[index] = GenerateGeometryPassthrough(pools.inst, pools.block, host_info,
-                                                          *layer_source_program, topology);
+            programs[index] =
+                GenerateGeometryPassthrough(pools.inst, pools.block, host_info,
+                                            *layer_source_program, topology,
+                                            MaxwellToInputTopology(key.gs_input_topology));
             continue;
         }
         if (key.unique_hashes[index] == 0) {
@@ -502,13 +498,15 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
 
         if (!uses_vertex_a || index != 1) {
             // Normal path
-            programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg, host_info);
+            programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
+                                              MaxwellToInputTopology(key.gs_input_topology));
 
             total_storage_buffers += Shader::NumDescriptors(programs[index].info.storage_buffers_descriptors);
         } else {
             // VertexB path when VertexA is present.
             auto& program_va{programs[0]};
-            auto program_vb{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
+            auto program_vb{TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
+                                             MaxwellToInputTopology(key.gs_input_topology))};
             total_storage_buffers += Shader::NumDescriptors(program_vb.info.storage_buffers_descriptors);
             programs[index] = MergeDualVertexPrograms(program_va, program_vb, env);
         }
@@ -540,6 +538,9 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
         const size_t stage_index{index - 1};
         infos[stage_index] = &program.info;
 
+        if (previous_program) {
+            PrunePassthroughStores(program, previous_program->info.stores);
+        }
         const auto runtime_info = MakeRuntimeInfo(key, program, previous_program, glasm_use_storage_buffers, use_glasm);
         switch (::Settings::values.renderer_backend.GetValue()) {
         case Settings::RendererBackend::OpenGL_GLSL:
@@ -597,7 +598,8 @@ std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline(
         env.Dump(hash, key.unique_hash);
     }
 
-    auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
+    auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
+                                  Shader::InputTopology::Points)};
     const u32 num_storage_buffers{Shader::NumDescriptors(program.info.storage_buffers_descriptors)};
     Shader::RuntimeInfo info;
     info.glasm_use_storage_buffers = num_storage_buffers <= device.GetMaxGLASMStorageBufferBlocks();

src/video_core/renderer_opengl/gl_texture_cache.h

@@ -370,6 +370,7 @@ struct TextureCacheParams {
     static constexpr bool HAS_EMULATED_COPIES = true;
     static constexpr bool HAS_DEVICE_MEMORY_INFO = true;
     static constexpr bool IMPLEMENTS_ASYNC_DOWNLOADS = true;
+    static constexpr bool HAS_MSAA_DOWNLOADS = false;
 
     using Runtime = OpenGL::TextureCacheRuntime;
     using Image = OpenGL::Image;

src/video_core/renderer_vulkan/blit_image.cpp

@@ -12,6 +12,7 @@
 #include "common/settings.h"
 #include "video_core/host_shaders/blit_color_float_frag_spv.h"
 #include "video_core/host_shaders/blit_color_msaa_frag_spv.h"
+#include "video_core/host_shaders/blit_depth_frag_spv.h"
 #include "video_core/host_shaders/blit_depth_msaa_frag_spv.h"
 #include "video_core/host_shaders/blit_depth_stencil_msaa_frag_spv.h"
 #include "video_core/host_shaders/convert_abgr8_to_d24s8_frag_spv.h"
@@ -21,7 +22,15 @@
 #include "video_core/host_shaders/convert_depth_to_float_frag_spv.h"
 #include "video_core/host_shaders/convert_float_to_depth_frag_spv.h"
 #include "video_core/host_shaders/convert_msaa_to_non_msaa_frag_spv.h"
+#include "video_core/host_shaders/convert_msaa_to_non_msaa_depth_frag_spv.h"
+#include "video_core/host_shaders/convert_msaa_to_non_msaa_depth_stencil_frag_spv.h"
+#include "video_core/host_shaders/convert_msaa_to_non_msaa_sint_frag_spv.h"
+#include "video_core/host_shaders/convert_msaa_to_non_msaa_uint_frag_spv.h"
 #include "video_core/host_shaders/convert_non_msaa_to_msaa_frag_spv.h"
+#include "video_core/host_shaders/convert_non_msaa_to_msaa_sint_frag_spv.h"
+#include "video_core/host_shaders/convert_non_msaa_to_msaa_uint_frag_spv.h"
+#include "video_core/host_shaders/convert_non_msaa_to_msaa_depth_frag_spv.h"
+#include "video_core/host_shaders/convert_non_msaa_to_msaa_depth_stencil_frag_spv.h"
 #include "video_core/host_shaders/convert_s8d24_to_abgr8_frag_spv.h"
 #include "video_core/host_shaders/full_screen_triangle_vert_spv.h"
 #include "video_core/host_shaders/vulkan_blit_depth_stencil_frag_spv.h"
@@ -518,8 +527,19 @@ void RecordShaderReadBarrier(Scheduler& scheduler, const ImageView& image_view)
     }
 }
 
+[[nodiscard]] MSAACopyFormatClass FormatClass(VideoCore::Surface::PixelFormat format) {
+    if (!VideoCore::Surface::IsPixelFormatInteger(format)) {
+        return MSAACopyFormatClass::Float;
+    }
+    if (VideoCore::Surface::IsPixelFormatSignedInteger(format)) {
+        return MSAACopyFormatClass::SignedInteger;
+    }
+    return MSAACopyFormatClass::UnsignedInteger;
+}
+
 [[nodiscard]] vk::ImageView MakeMSAACopyView(const vk::Device& device, VkImage image,
-                                             VkFormat format, u32 base_level) {
+                                             VkFormat format, u32 base_level, u32 base_layer,
+                                             VkImageAspectFlags aspect_mask) {
     return device.CreateImageView(VkImageViewCreateInfo{
         .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
         .pNext = nullptr,
@@ -534,10 +554,10 @@ void RecordShaderReadBarrier(Scheduler& scheduler, const ImageView& image_view)
             .a = VK_COMPONENT_SWIZZLE_IDENTITY,
         },
         .subresourceRange{
-            .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+            .aspectMask = aspect_mask,
             .baseMipLevel = base_level,
             .levelCount = 1,
-            .baseArrayLayer = 0,
+            .baseArrayLayer = base_layer,
             .layerCount = 1,
         },
     });
@@ -586,12 +606,17 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
       msaa_copy_pipeline_layout(device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo(
           one_texture_set_layout.address(),
           PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(MSAACopyPushConstants)>))),
+      msaa_copy_depth_stencil_pipeline_layout(
+          device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo(
+              two_textures_set_layout.address(),
+              PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(MSAACopyPushConstants)>))),
       full_screen_vert(BuildShader(device, FULL_SCREEN_TRIANGLE_VERT_SPV)),
       blit_color_to_color_frag(BuildShader(device, BLIT_COLOR_FLOAT_FRAG_SPV)),
       blit_color_msaa_frag(BuildShader(device, BLIT_COLOR_MSAA_FRAG_SPV)),
       blit_depth_stencil_frag(device.IsExtShaderStencilExportSupported()
                              ? BuildShader(device, VULKAN_BLIT_DEPTH_STENCIL_FRAG_SPV)
                              : vk::ShaderModule{}),
+      blit_depth_frag(BuildShader(device, BLIT_DEPTH_FRAG_SPV)),
       blit_depth_msaa_frag(BuildShader(device, BLIT_DEPTH_MSAA_FRAG_SPV)),
       blit_depth_stencil_msaa_frag(device.IsExtShaderStencilExportSupported()
                              ? BuildShader(device, BLIT_DEPTH_STENCIL_MSAA_FRAG_SPV)
@@ -609,7 +634,25 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
       convert_d24s8_to_abgr8_frag(BuildShader(device, CONVERT_D24S8_TO_ABGR8_FRAG_SPV)),
       convert_s8d24_to_abgr8_frag(BuildShader(device, CONVERT_S8D24_TO_ABGR8_FRAG_SPV)),
       convert_msaa_to_non_msaa_frag(BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_FRAG_SPV)),
+      convert_msaa_to_non_msaa_sint_frag(
+          BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_SINT_FRAG_SPV)),
+      convert_msaa_to_non_msaa_uint_frag(
+          BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_UINT_FRAG_SPV)),
+      convert_msaa_to_non_msaa_depth_frag(
+          BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_DEPTH_FRAG_SPV)),
+      convert_msaa_to_non_msaa_depth_stencil_frag(
+          BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_DEPTH_STENCIL_FRAG_SPV)),
       convert_non_msaa_to_msaa_frag(BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_FRAG_SPV)),
+      convert_non_msaa_to_msaa_sint_frag(
+          BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_SINT_FRAG_SPV)),
+      convert_non_msaa_to_msaa_uint_frag(
+          BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_UINT_FRAG_SPV)),
+      convert_non_msaa_to_msaa_depth_frag(
+          BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_DEPTH_FRAG_SPV)),
+      convert_non_msaa_to_msaa_depth_stencil_frag(
+          device.IsExtShaderStencilExportSupported()
+              ? BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_DEPTH_STENCIL_FRAG_SPV)
+              : vk::ShaderModule{}),
       linear_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_LINEAR>)),
       nearest_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_NEAREST>)) {}
 
@@ -697,6 +740,68 @@ void BlitImageHelper::BlitColorMSAA(const Framebuffer* dst_framebuffer,
     scheduler.InvalidateState();
 }
 
+void BlitImageHelper::BlitDepthStencilMSAA(const Framebuffer* dst_framebuffer,
+                                           ImageView& src_image_view, const Region2D& dst_region,
+                                           const Region2D& src_region) {
+    const bool blit_stencil =
+        dst_framebuffer->HasAspectStencilBit() && device.IsExtShaderStencilExportSupported();
+    const BlitMSAAPipelineKey key{
+        .renderpass = dst_framebuffer->RenderPass(),
+        .samples = dst_framebuffer->Samples(),
+    };
+    const VkPipeline pipeline = FindOrEmplaceBlitDepthStencilMSAAPipeline(key, blit_stencil);
+    const VkPipelineLayout layout =
+        blit_stencil ? *two_textures_pipeline_layout : *one_texture_pipeline_layout;
+    const VkSampler sampler = *nearest_sampler;
+    const VkImageView src_depth_view = src_image_view.DepthView();
+    const VkImageView src_stencil_view =
+        blit_stencil ? src_image_view.StencilView() : VK_NULL_HANDLE;
+
+    RecordShaderReadBarrier(scheduler, src_image_view);
+    scheduler.RequestRenderpass(dst_framebuffer);
+    scheduler.Record([this, dst_region, src_region, pipeline, layout, sampler, src_depth_view,
+                      src_stencil_view, blit_stencil](vk::CommandBuffer cmdbuf) {
+        if (blit_stencil) {
+            const VkDescriptorSet descriptor_set = two_textures_descriptor_allocator.Commit();
+            UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_depth_view,
+                                           src_stencil_view);
+            cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
+                                      nullptr);
+        } else {
+            const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
+            UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_depth_view);
+            cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
+                                      nullptr);
+        }
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+        BindBlitState(cmdbuf, layout, dst_region, src_region);
+        cmdbuf.Draw(3, 1, 0, 0);
+    });
+    scheduler.InvalidateState();
+}
+
+void BlitImageHelper::BlitDepth(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
+                                const Region2D& dst_region, const Region2D& src_region) {
+    const VkPipeline pipeline = FindOrEmplaceBlitDepthPipeline(dst_framebuffer->RenderPass());
+    const VkPipelineLayout layout = *one_texture_pipeline_layout;
+    const VkSampler sampler = *nearest_sampler;
+    const VkImageView src_depth_view = src_image_view.DepthView();
+
+    RecordShaderReadBarrier(scheduler, src_image_view);
+    scheduler.RequestRenderpass(dst_framebuffer);
+    scheduler.Record([this, dst_region, src_region, pipeline, layout, sampler,
+                      src_depth_view](vk::CommandBuffer cmdbuf) {
+        const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
+        UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_depth_view);
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
+                                  nullptr);
+        BindBlitState(cmdbuf, layout, dst_region, src_region);
+        cmdbuf.Draw(3, 1, 0, 0);
+    });
+    scheduler.InvalidateState();
+}
+
 void BlitImageHelper::ResolveDepthStencil(const Framebuffer* dst_framebuffer,
                                           ImageView& src_image_view, const Region2D& dst_region,
                                           const Region2D& src_region) {
@@ -905,6 +1010,7 @@ void BlitImageHelper::CopyMSAA(RenderPassCache& render_pass_cache, VkImage dst_i
         .renderpass = renderpass,
         .samples = samples,
         .msaa_to_non_msaa = msaa_to_non_msaa,
+        .format_class = FormatClass(dst_format),
     };
     const VkPipeline pipeline = FindOrEmplaceMSAACopyPipeline(key);
     const VkPipelineLayout layout = *msaa_copy_pipeline_layout;
@@ -914,138 +1020,142 @@ void BlitImageHelper::CopyMSAA(RenderPassCache& render_pass_cache, VkImage dst_i
     const VkFormat dst_vk_format =
         MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, dst_format).format;
     for (const VideoCommon::ImageCopy& copy : copies) {
-        ASSERT(copy.src_subresource.base_layer == 0);
-        ASSERT(copy.src_subresource.num_layers == 1);
-        ASSERT(copy.dst_subresource.base_layer == 0);
-        ASSERT(copy.dst_subresource.num_layers == 1);
-        vk::ImageView src_view =
-            MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format,
-                             static_cast<u32>(copy.src_subresource.base_level));
-        vk::ImageView dst_view =
-            MakeMSAACopyView(device.GetLogical(), dst_image, dst_vk_format,
-                             static_cast<u32>(copy.dst_subresource.base_level));
-        const VkOffset2D dst_offset{copy.dst_offset.x, copy.dst_offset.y};
-        const VkExtent2D dst_extent{copy.extent.width, copy.extent.height};
-        const VkRect2D render_area{
-            .offset = dst_offset,
-            .extent = dst_extent,
-        };
-        vk::Framebuffer framebuffer = device.GetLogical().CreateFramebuffer(VkFramebufferCreateInfo{
-            .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
-            .pNext = nullptr,
-            .flags = 0,
-            .renderPass = renderpass,
-            .attachmentCount = 1,
-            .pAttachments = dst_view.address(),
-            .width = static_cast<u32>(dst_offset.x) + dst_extent.width,
-            .height = static_cast<u32>(dst_offset.y) + dst_extent.height,
-            .layers = 1,
-        });
-        const MSAACopyPushConstants push_constants{
-            .dst_offset = {dst_offset.x, dst_offset.y},
-            .src_offset = {copy.src_offset.x, copy.src_offset.y},
-            .scale = {scale_x, scale_y},
-        };
-        scheduler.RequestOutsideRenderPassOperationContext();
-        scheduler.Record([this, pipeline, layout, sampler, renderpass,
-                          framebuffer_handle = *framebuffer, src_view_handle = *src_view,
-                          src = src_image, dst = dst_image, render_area,
-                          push_constants](vk::CommandBuffer cmdbuf) {
-            constexpr VkImageSubresourceRange color_range{
-                .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
-                .baseMipLevel = 0,
-                .levelCount = VK_REMAINING_MIP_LEVELS,
-                .baseArrayLayer = 0,
-                .layerCount = VK_REMAINING_ARRAY_LAYERS,
+        const s32 num_layers = (std::min)(copy.src_subresource.num_layers,
+                                          copy.dst_subresource.num_layers);
+        for (s32 layer = 0; layer < num_layers; ++layer) {
+            vk::ImageView src_view =
+                MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format,
+                                 static_cast<u32>(copy.src_subresource.base_level),
+                                 static_cast<u32>(copy.src_subresource.base_layer + layer),
+                                 VK_IMAGE_ASPECT_COLOR_BIT);
+            vk::ImageView dst_view =
+                MakeMSAACopyView(device.GetLogical(), dst_image, dst_vk_format,
+                                 static_cast<u32>(copy.dst_subresource.base_level),
+                                 static_cast<u32>(copy.dst_subresource.base_layer + layer),
+                                 VK_IMAGE_ASPECT_COLOR_BIT);
+            const VkOffset2D dst_offset{copy.dst_offset.x, copy.dst_offset.y};
+            const VkExtent2D dst_extent{copy.extent.width, copy.extent.height};
+            const VkRect2D render_area{
+                .offset = dst_offset,
+                .extent = dst_extent,
             };
-            const std::array pre_barriers{
-                VkImageMemoryBarrier{
+            vk::Framebuffer framebuffer = device.GetLogical().CreateFramebuffer(VkFramebufferCreateInfo{
+                .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
+                .pNext = nullptr,
+                .flags = 0,
+                .renderPass = renderpass,
+                .attachmentCount = 1,
+                .pAttachments = dst_view.address(),
+                .width = static_cast<u32>(dst_offset.x) + dst_extent.width,
+                .height = static_cast<u32>(dst_offset.y) + dst_extent.height,
+                .layers = 1,
+            });
+            const MSAACopyPushConstants push_constants{
+                .dst_offset = {dst_offset.x, dst_offset.y},
+                .src_offset = {copy.src_offset.x, copy.src_offset.y},
+                .scale = {scale_x, scale_y},
+            };
+            scheduler.RequestOutsideRenderPassOperationContext();
+            scheduler.Record([this, pipeline, layout, sampler, renderpass,
+                              framebuffer_handle = *framebuffer, src_view_handle = *src_view,
+                              src = src_image, dst = dst_image, render_area,
+                              push_constants](vk::CommandBuffer cmdbuf) {
+                constexpr VkImageSubresourceRange color_range{
+                    .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+                    .baseMipLevel = 0,
+                    .levelCount = VK_REMAINING_MIP_LEVELS,
+                    .baseArrayLayer = 0,
+                    .layerCount = VK_REMAINING_ARRAY_LAYERS,
+                };
+                const std::array pre_barriers{
+                    VkImageMemoryBarrier{
+                        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+                        .pNext = nullptr,
+                        .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+                                         VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
+                        .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
+                        .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+                        .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+                        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                        .image = src,
+                        .subresourceRange = color_range,
+                    },
+                    VkImageMemoryBarrier{
+                        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+                        .pNext = nullptr,
+                        .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+                                         VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
+                        .dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
+                                         VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
+                        .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+                        .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+                        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                        .image = dst,
+                        .subresourceRange = color_range,
+                    },
+                };
+                cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
+                                           VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
+                                           VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+                                           VK_PIPELINE_STAGE_TRANSFER_BIT,
+                                       VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+                                           VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
+                                       0, nullptr, nullptr, pre_barriers);
+                const VkRenderPassBeginInfo renderpass_bi{
+                    .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+                    .pNext = nullptr,
+                    .renderPass = renderpass,
+                    .framebuffer = framebuffer_handle,
+                    .renderArea = render_area,
+                    .clearValueCount = 0,
+                    .pClearValues = nullptr,
+                };
+                cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
+                const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
+                UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view_handle);
+                cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+                cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
+                                          nullptr);
+                const VkViewport viewport{
+                    .x = static_cast<float>(render_area.offset.x),
+                    .y = static_cast<float>(render_area.offset.y),
+                    .width = static_cast<float>(render_area.extent.width),
+                    .height = static_cast<float>(render_area.extent.height),
+                    .minDepth = 0.0f,
+                    .maxDepth = 1.0f,
+                };
+                cmdbuf.SetViewport(0, viewport);
+                cmdbuf.SetScissor(0, render_area);
+                cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, push_constants);
+                cmdbuf.Draw(3, 1, 0, 0);
+                cmdbuf.EndRenderPass();
+                const VkImageMemoryBarrier post_barrier{
                     .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
                     .pNext = nullptr,
-                    .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
-                                     VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
-                    .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
-                    .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
-                    .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-                    .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                    .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                    .image = src,
-                    .subresourceRange = color_range,
-                },
-                VkImageMemoryBarrier{
-                    .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-                    .pNext = nullptr,
-                    .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
-                                     VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
-                    .dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
-                                     VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
+                    .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
+                    .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT,
                     .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
                     .newLayout = VK_IMAGE_LAYOUT_GENERAL,
                     .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                     .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                     .image = dst,
                     .subresourceRange = color_range,
-                },
-            };
-            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
-                                       VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
+                };
+                cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
                                        VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
-                                       VK_PIPELINE_STAGE_TRANSFER_BIT,
-                                   VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
-                                       VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
-                                   0, nullptr, nullptr, pre_barriers);
-            const VkRenderPassBeginInfo renderpass_bi{
-                .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
-                .pNext = nullptr,
-                .renderPass = renderpass,
-                .framebuffer = framebuffer_handle,
-                .renderArea = render_area,
-                .clearValueCount = 0,
-                .pClearValues = nullptr,
-            };
-            cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
-            const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
-            UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view_handle);
-            cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
-            cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
-                                      nullptr);
-            const VkViewport viewport{
-                .x = static_cast<float>(render_area.offset.x),
-                .y = static_cast<float>(render_area.offset.y),
-                .width = static_cast<float>(render_area.extent.width),
-                .height = static_cast<float>(render_area.extent.height),
-                .minDepth = 0.0f,
-                .maxDepth = 1.0f,
-            };
-            cmdbuf.SetViewport(0, viewport);
-            cmdbuf.SetScissor(0, render_area);
-            cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, push_constants);
-            cmdbuf.Draw(3, 1, 0, 0);
-            cmdbuf.EndRenderPass();
-            const VkImageMemoryBarrier post_barrier{
-                .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-                .pNext = nullptr,
-                .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
-                .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT,
-                .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
-                .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-                .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                .image = dst,
-                .subresourceRange = color_range,
-            };
-            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
-                                   VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
-                                       VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
-                                       VK_PIPELINE_STAGE_TRANSFER_BIT,
-                                   0, post_barrier);
-        });
-        msaa_copy_resources.push_back(MSAACopyResources{
-            .tick = scheduler.CurrentTick(),
-            .src_view = std::move(src_view),
-            .dst_view = std::move(dst_view),
-            .framebuffer = std::move(framebuffer),
-        });
+                                           VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
+                                           VK_PIPELINE_STAGE_TRANSFER_BIT,
+                                       0, post_barrier);
+            });
+            msaa_copy_resources.push_back(MSAACopyResources{
+                .tick = scheduler.CurrentTick(),
+                .src_view = std::move(src_view),
+                .dst_view = std::move(dst_view),
+                .framebuffer = std::move(framebuffer),
+            });
+        }
     }
     scheduler.InvalidateState();
 }
@@ -1191,7 +1301,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceColorPipeline(const BlitImagePipelineKe
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }));
+    }, device.StaticPipelineCache()));
     return *blit_color_pipelines.back();
 }
 
@@ -1223,7 +1333,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceDepthStencilPipeline(const BlitImagePip
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }));
+    }, device.StaticPipelineCache()));
     return *blit_depth_stencil_pipelines.back();
 }
 
@@ -1276,7 +1386,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceClearColorPipeline(const BlitImagePipel
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }));
+    }, device.StaticPipelineCache()));
     return *clear_color_pipelines.back();
 }
 
@@ -1332,7 +1442,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceClearStencilPipeline(
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }));
+    }, device.StaticPipelineCache()));
     return *clear_stencil_pipelines.back();
 }
 
@@ -1375,10 +1485,91 @@ VkPipeline BlitImageHelper::FindOrEmplaceBlitColorMSAAPipeline(const BlitMSAAPip
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }));
+    }, device.StaticPipelineCache()));
     return *blit_msaa_color_pipelines.back();
 }
 
+VkPipeline BlitImageHelper::FindOrEmplaceBlitDepthStencilMSAAPipeline(
+    const BlitMSAAPipelineKey& key, bool blit_stencil) {
+    auto& keys = blit_stencil ? blit_msaa_depth_stencil_keys : blit_msaa_depth_keys;
+    auto& pipelines = blit_stencil ? blit_msaa_depth_stencil_pipelines : blit_msaa_depth_pipelines;
+    const auto it = std::ranges::find(keys, key);
+    if (it != keys.end()) {
+        return *pipelines[std::distance(keys.begin(), it)];
+    }
+    keys.push_back(key);
+    const std::array stages =
+        MakeStages(*full_screen_vert,
+                   blit_stencil ? *blit_depth_stencil_msaa_frag : *blit_depth_msaa_frag);
+    const VkPipelineMultisampleStateCreateInfo multisample_ci{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .rasterizationSamples = key.samples,
+        .sampleShadingEnable = VK_TRUE,
+        .minSampleShading = 1.0f,
+        .pSampleMask = nullptr,
+        .alphaToCoverageEnable = VK_FALSE,
+        .alphaToOneEnable = VK_FALSE,
+    };
+    const VkPipelineInputAssemblyStateCreateInfo input_assembly_ci = GetPipelineInputAssemblyStateCreateInfo(device);
+    pipelines.push_back(device.GetLogical().CreateGraphicsPipeline({
+        .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .stageCount = static_cast<u32>(stages.size()),
+        .pStages = stages.data(),
+        .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+        .pInputAssemblyState = &input_assembly_ci,
+        .pTessellationState = nullptr,
+        .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+        .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+        .pMultisampleState = &multisample_ci,
+        .pDepthStencilState = blit_stencil ? &PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO
+                                           : &PIPELINE_DEPTH_ONLY_STATE_CREATE_INFO,
+        .pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
+        .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+        .layout = blit_stencil ? *two_textures_pipeline_layout : *one_texture_pipeline_layout,
+        .renderPass = key.renderpass,
+        .subpass = 0,
+        .basePipelineHandle = VK_NULL_HANDLE,
+        .basePipelineIndex = 0,
+    }, device.StaticPipelineCache()));
+    return *pipelines.back();
+}
+
+VkPipeline BlitImageHelper::FindOrEmplaceBlitDepthPipeline(VkRenderPass renderpass) {
+    const auto it = std::ranges::find(blit_depth_keys, renderpass);
+    if (it != blit_depth_keys.end()) {
+        return *blit_depth_pipelines[std::distance(blit_depth_keys.begin(), it)];
+    }
+    blit_depth_keys.push_back(renderpass);
+    const std::array stages = MakeStages(*full_screen_vert, *blit_depth_frag);
+    const VkPipelineInputAssemblyStateCreateInfo input_assembly_ci = GetPipelineInputAssemblyStateCreateInfo(device);
+    blit_depth_pipelines.push_back(device.GetLogical().CreateGraphicsPipeline({
+        .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .stageCount = static_cast<u32>(stages.size()),
+        .pStages = stages.data(),
+        .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+        .pInputAssemblyState = &input_assembly_ci,
+        .pTessellationState = nullptr,
+        .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+        .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+        .pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+        .pDepthStencilState = &PIPELINE_DEPTH_ONLY_STATE_CREATE_INFO,
+        .pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
+        .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+        .layout = *one_texture_pipeline_layout,
+        .renderPass = renderpass,
+        .subpass = 0,
+        .basePipelineHandle = VK_NULL_HANDLE,
+        .basePipelineIndex = 0,
+    }, device.StaticPipelineCache()));
+    return *blit_depth_pipelines.back();
+}
+
 VkPipeline BlitImageHelper::FindOrEmplaceResolveDepthStencilPipeline(VkRenderPass renderpass,
                                                                      bool resolve_stencil) {
     auto& keys = resolve_stencil ? resolve_depth_stencil_keys : resolve_depth_keys;
@@ -1413,19 +1604,223 @@ VkPipeline BlitImageHelper::FindOrEmplaceResolveDepthStencilPipeline(VkRenderPas
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }));
+    }, device.StaticPipelineCache()));
     return *pipelines.back();
 }
 
+void BlitImageHelper::CopyMSAADepth(RenderPassCache& render_pass_cache, VkImage dst_image,
+                                    VideoCore::Surface::PixelFormat dst_format, VkImage src_image,
+                                    VideoCore::Surface::PixelFormat src_format, u32 num_samples,
+                                    std::span<const VideoCommon::ImageCopy> copies,
+                                    bool copy_stencil, bool msaa_to_non_msaa) {
+    while (!msaa_copy_resources.empty() && scheduler.IsFree(msaa_copy_resources.front().tick)) {
+        msaa_copy_resources.pop_front();
+    }
+    const auto [samples_x, samples_y] = VideoCommon::SamplesLog2(static_cast<int>(num_samples));
+    const s32 scale_x = 1 << samples_x;
+    const s32 scale_y = 1 << samples_y;
+    const VkSampleCountFlagBits samples =
+        msaa_to_non_msaa ? VK_SAMPLE_COUNT_1_BIT : SampleCountFlag(num_samples);
+    RenderPassKey renderpass_key{};
+    renderpass_key.color_formats.fill(VideoCore::Surface::PixelFormat::Invalid);
+    renderpass_key.depth_format = dst_format;
+    renderpass_key.samples = samples;
+    const VkRenderPass renderpass = render_pass_cache.Get(renderpass_key);
+    const MSAACopyPipelineKey key{
+        .renderpass = renderpass,
+        .samples = samples,
+        .msaa_to_non_msaa = msaa_to_non_msaa,
+        .format_class = MSAACopyFormatClass::Float,
+    };
+    const VkPipeline pipeline = FindOrEmplaceMSAACopyDepthPipeline(key, copy_stencil);
+    const VkPipelineLayout layout = copy_stencil ? *msaa_copy_depth_stencil_pipeline_layout
+                                                 : *msaa_copy_pipeline_layout;
+    const VkSampler sampler = *nearest_sampler;
+    const VkFormat src_vk_format =
+        MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, src_format).format;
+    const VkFormat dst_vk_format =
+        MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, dst_format).format;
+    VkImageAspectFlags attachment_aspect = VK_IMAGE_ASPECT_DEPTH_BIT;
+    if (VideoCore::Surface::GetFormatType(dst_format) ==
+        VideoCore::Surface::SurfaceType::DepthStencil) {
+        attachment_aspect |= VK_IMAGE_ASPECT_STENCIL_BIT;
+    }
+    for (const VideoCommon::ImageCopy& copy : copies) {
+        const s32 num_layers = (std::min)(copy.src_subresource.num_layers,
+                                          copy.dst_subresource.num_layers);
+        for (s32 layer = 0; layer < num_layers; ++layer) {
+            vk::ImageView src_view =
+                MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format,
+                                 static_cast<u32>(copy.src_subresource.base_level),
+                                 static_cast<u32>(copy.src_subresource.base_layer + layer),
+                                 VK_IMAGE_ASPECT_DEPTH_BIT);
+            vk::ImageView src_stencil_view =
+                copy_stencil ? MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format,
+                                                static_cast<u32>(copy.src_subresource.base_level),
+                                                static_cast<u32>(copy.src_subresource.base_layer + layer),
+                                                VK_IMAGE_ASPECT_STENCIL_BIT)
+                             : vk::ImageView{};
+            vk::ImageView dst_view =
+                MakeMSAACopyView(device.GetLogical(), dst_image, dst_vk_format,
+                                 static_cast<u32>(copy.dst_subresource.base_level),
+                                 static_cast<u32>(copy.dst_subresource.base_layer + layer),
+                                 attachment_aspect);
+            const VkOffset2D dst_offset{copy.dst_offset.x, copy.dst_offset.y};
+            const VkExtent2D dst_extent{copy.extent.width, copy.extent.height};
+            const VkRect2D render_area{
+                .offset = dst_offset,
+                .extent = dst_extent,
+            };
+            vk::Framebuffer framebuffer = device.GetLogical().CreateFramebuffer(VkFramebufferCreateInfo{
+                .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
+                .pNext = nullptr,
+                .flags = 0,
+                .renderPass = renderpass,
+                .attachmentCount = 1,
+                .pAttachments = dst_view.address(),
+                .width = static_cast<u32>(dst_offset.x) + dst_extent.width,
+                .height = static_cast<u32>(dst_offset.y) + dst_extent.height,
+                .layers = 1,
+            });
+            const MSAACopyPushConstants push_constants{
+                .dst_offset = {dst_offset.x, dst_offset.y},
+                .src_offset = {copy.src_offset.x, copy.src_offset.y},
+                .scale = {scale_x, scale_y},
+            };
+            scheduler.RequestOutsideRenderPassOperationContext();
+            const VkImageView src_stencil_handle = copy_stencil ? *src_stencil_view : VK_NULL_HANDLE;
+            scheduler.Record([this, pipeline, layout, sampler, renderpass,
+                              framebuffer_handle = *framebuffer, src_view_handle = *src_view,
+                              src_stencil_handle, src = src_image, dst = dst_image, render_area,
+                              attachment_aspect, push_constants](vk::CommandBuffer cmdbuf) {
+                const VkImageSubresourceRange src_range{
+                    .aspectMask = attachment_aspect,
+                    .baseMipLevel = 0,
+                    .levelCount = VK_REMAINING_MIP_LEVELS,
+                    .baseArrayLayer = 0,
+                    .layerCount = VK_REMAINING_ARRAY_LAYERS,
+                };
+                const std::array pre_barriers{
+                    VkImageMemoryBarrier{
+                        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+                        .pNext = nullptr,
+                        .srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
+                                         VK_ACCESS_TRANSFER_WRITE_BIT,
+                        .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
+                        .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+                        .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+                        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                        .image = src,
+                        .subresourceRange = src_range,
+                    },
+                    VkImageMemoryBarrier{
+                        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+                        .pNext = nullptr,
+                        .srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
+                                         VK_ACCESS_TRANSFER_WRITE_BIT,
+                        .dstAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
+                                         VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
+                        .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+                        .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+                        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                        .image = dst,
+                        .subresourceRange = src_range,
+                    },
+                };
+                cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
+                                           VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
+                                           VK_PIPELINE_STAGE_TRANSFER_BIT,
+                                       VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+                                           VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT,
+                                       0, nullptr, nullptr, pre_barriers);
+                const VkRenderPassBeginInfo renderpass_bi{
+                    .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+                    .pNext = nullptr,
+                    .renderPass = renderpass,
+                    .framebuffer = framebuffer_handle,
+                    .renderArea = render_area,
+                    .clearValueCount = 0,
+                    .pClearValues = nullptr,
+                };
+                cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
+                const VkDescriptorSet descriptor_set =
+                    src_stencil_handle != VK_NULL_HANDLE
+                        ? two_textures_descriptor_allocator.Commit()
+                        : one_texture_descriptor_allocator.Commit();
+                if (src_stencil_handle != VK_NULL_HANDLE) {
+                    UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_view_handle,
+                                                   src_stencil_handle);
+                } else {
+                    UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view_handle);
+                }
+                cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+                cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
+                                          nullptr);
+                const VkViewport viewport{
+                    .x = static_cast<float>(render_area.offset.x),
+                    .y = static_cast<float>(render_area.offset.y),
+                    .width = static_cast<float>(render_area.extent.width),
+                    .height = static_cast<float>(render_area.extent.height),
+                    .minDepth = 0.0f,
+                    .maxDepth = 1.0f,
+                };
+                cmdbuf.SetViewport(0, viewport);
+                cmdbuf.SetScissor(0, render_area);
+                cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, push_constants);
+                cmdbuf.Draw(3, 1, 0, 0);
+                cmdbuf.EndRenderPass();
+                const VkImageMemoryBarrier post_barrier{
+                    .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+                    .pNext = nullptr,
+                    .srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
+                    .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT |
+                                     VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT,
+                    .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+                    .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+                    .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                    .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+                    .image = dst,
+                    .subresourceRange = src_range,
+                };
+                cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
+                                       vk::PIPELINE_STAGE_GRAPHICS_COMPUTE_TRANSFER, 0, post_barrier);
+            });
+            msaa_copy_resources.push_back(MSAACopyResources{
+                .tick = scheduler.CurrentTick(),
+                .src_view = std::move(src_view),
+                .dst_view = std::move(dst_view),
+                .framebuffer = std::move(framebuffer),
+            });
+            if (copy_stencil) {
+                msaa_copy_resources.push_back(MSAACopyResources{
+                    .tick = scheduler.CurrentTick(),
+                    .src_view = std::move(src_stencil_view),
+                    .dst_view = vk::ImageView{},
+                    .framebuffer = vk::Framebuffer{},
+                });
+            }
+        }
+    }
+}
+
 VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyPipeline(const MSAACopyPipelineKey& key) {
     const auto it = std::ranges::find(msaa_copy_keys, key);
     if (it != msaa_copy_keys.end()) {
         return *msaa_copy_pipelines[std::distance(msaa_copy_keys.begin(), it)];
     }
     msaa_copy_keys.push_back(key);
-    const std::array stages = MakeStages(*clear_color_vert, key.msaa_to_non_msaa
-                                                                ? *convert_msaa_to_non_msaa_frag
-                                                                : *convert_non_msaa_to_msaa_frag);
+    VkShaderModule frag_module = key.msaa_to_non_msaa ? *convert_msaa_to_non_msaa_frag
+                                                     : *convert_non_msaa_to_msaa_frag;
+    if (key.format_class == MSAACopyFormatClass::SignedInteger) {
+        frag_module = key.msaa_to_non_msaa ? *convert_msaa_to_non_msaa_sint_frag
+                                           : *convert_non_msaa_to_msaa_sint_frag;
+    } else if (key.format_class == MSAACopyFormatClass::UnsignedInteger) {
+        frag_module = key.msaa_to_non_msaa ? *convert_msaa_to_non_msaa_uint_frag
+                                           : *convert_non_msaa_to_msaa_uint_frag;
+    }
+    const std::array stages = MakeStages(*clear_color_vert, frag_module);
     const VkPipelineMultisampleStateCreateInfo multisample_ci{
         .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
         .pNext = nullptr,
@@ -1458,10 +1853,89 @@ VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyPipeline(const MSAACopyPipeline
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }));
+    }, device.StaticPipelineCache()));
     return *msaa_copy_pipelines.back();
 }
 
+VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyDepthPipeline(const MSAACopyPipelineKey& key,
+                                                               bool copy_stencil) {
+    auto& keys = copy_stencil ? msaa_copy_depth_stencil_keys : msaa_copy_depth_keys;
+    auto& pipelines = copy_stencil ? msaa_copy_depth_stencil_pipelines : msaa_copy_depth_pipelines;
+    const auto it = std::ranges::find(keys, key);
+    if (it != keys.end()) {
+        return *pipelines[std::distance(keys.begin(), it)];
+    }
+    keys.push_back(key);
+    VkShaderModule frag_module;
+    if (key.msaa_to_non_msaa) {
+        frag_module = copy_stencil ? *convert_msaa_to_non_msaa_depth_stencil_frag
+                                   : *convert_msaa_to_non_msaa_depth_frag;
+    } else {
+        frag_module = copy_stencil ? *convert_non_msaa_to_msaa_depth_stencil_frag
+                                   : *convert_non_msaa_to_msaa_depth_frag;
+    }
+    const std::array stages = MakeStages(*clear_color_vert, frag_module);
+    const VkPipelineMultisampleStateCreateInfo multisample_ci{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .rasterizationSamples = key.samples,
+        .sampleShadingEnable = key.msaa_to_non_msaa ? VK_FALSE : VK_TRUE,
+        .minSampleShading = key.msaa_to_non_msaa ? 0.0f : 1.0f,
+        .pSampleMask = nullptr,
+        .alphaToCoverageEnable = VK_FALSE,
+        .alphaToOneEnable = VK_FALSE,
+    };
+    static constexpr VkStencilOpState REPLACE_STENCIL_OP{
+        .failOp = VK_STENCIL_OP_REPLACE,
+        .passOp = VK_STENCIL_OP_REPLACE,
+        .depthFailOp = VK_STENCIL_OP_REPLACE,
+        .compareOp = VK_COMPARE_OP_ALWAYS,
+        .compareMask = 0xFF,
+        .writeMask = 0xFF,
+        .reference = 0,
+    };
+    const VkPipelineDepthStencilStateCreateInfo depth_stencil_ci{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .depthTestEnable = VK_TRUE,
+        .depthWriteEnable = VK_TRUE,
+        .depthCompareOp = VK_COMPARE_OP_ALWAYS,
+        .depthBoundsTestEnable = VK_FALSE,
+        .stencilTestEnable = copy_stencil ? VK_TRUE : VK_FALSE,
+        .front = copy_stencil ? REPLACE_STENCIL_OP : VkStencilOpState{},
+        .back = copy_stencil ? REPLACE_STENCIL_OP : VkStencilOpState{},
+        .minDepthBounds = 0.0f,
+        .maxDepthBounds = 0.0f,
+    };
+    const VkPipelineInputAssemblyStateCreateInfo input_assembly_ci =
+        GetPipelineInputAssemblyStateCreateInfo(device);
+    pipelines.push_back(device.GetLogical().CreateGraphicsPipeline({
+        .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .stageCount = static_cast<u32>(stages.size()),
+        .pStages = stages.data(),
+        .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+        .pInputAssemblyState = &input_assembly_ci,
+        .pTessellationState = nullptr,
+        .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+        .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+        .pMultisampleState = &multisample_ci,
+        .pDepthStencilState = &depth_stencil_ci,
+        .pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
+        .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+        .layout = copy_stencil ? *msaa_copy_depth_stencil_pipeline_layout
+                               : *msaa_copy_pipeline_layout,
+        .renderPass = key.renderpass,
+        .subpass = 0,
+        .basePipelineHandle = VK_NULL_HANDLE,
+        .basePipelineIndex = 0,
+    }, device.StaticPipelineCache()));
+    return *pipelines.back();
+}
+
 void BlitImageHelper::ConvertDepthToColorPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass) {
     ConvertPipeline(pipeline, renderpass, false);
 }
@@ -1499,7 +1973,7 @@ void BlitImageHelper::ConvertPipelineEx(vk::Pipeline& pipeline, VkRenderPass ren
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    });
+    }, device.StaticPipelineCache());
 }
 
 void BlitImageHelper::ConvertPipelineColorTargetEx(vk::Pipeline& pipeline, VkRenderPass renderpass,
@@ -1542,7 +2016,7 @@ void BlitImageHelper::ConvertPipeline(vk::Pipeline& pipeline, VkRenderPass rende
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    });
+    }, device.StaticPipelineCache());
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/blit_image.h

@@ -45,12 +45,19 @@ struct BlitDepthStencilPipelineKey {
     u32 stencil_ref;
 };
 
+enum class MSAACopyFormatClass : u32 {
+    Float,
+    SignedInteger,
+    UnsignedInteger,
+};
+
 struct MSAACopyPipelineKey {
     constexpr auto operator<=>(const MSAACopyPipelineKey&) const noexcept = default;
 
     VkRenderPass renderpass;
     VkSampleCountFlagBits samples;
     bool msaa_to_non_msaa;
+    MSAACopyFormatClass format_class;
 };
 
 struct BlitMSAAPipelineKey {
@@ -78,6 +85,12 @@ public:
     void BlitColorMSAA(const Framebuffer* dst_framebuffer, const ImageView& src_image_view,
                        const Region2D& dst_region, const Region2D& src_region);
 
+    void BlitDepthStencilMSAA(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
+                              const Region2D& dst_region, const Region2D& src_region);
+
+    void BlitDepth(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
+                   const Region2D& dst_region, const Region2D& src_region);
+
     void ResolveDepthStencil(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
                              const Region2D& dst_region, const Region2D& src_region);
 
@@ -116,6 +129,12 @@ public:
                   VideoCore::Surface::PixelFormat src_format, u32 num_samples,
                   std::span<const VideoCommon::ImageCopy> copies, bool msaa_to_non_msaa);
 
+    void CopyMSAADepth(RenderPassCache& render_pass_cache, VkImage dst_image,
+                       VideoCore::Surface::PixelFormat dst_format, VkImage src_image,
+                       VideoCore::Surface::PixelFormat src_format, u32 num_samples,
+                       std::span<const VideoCommon::ImageCopy> copies, bool copy_stencil,
+                       bool msaa_to_non_msaa);
+
 private:
     void Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
                  const ImageView& src_image_view);
@@ -131,7 +150,13 @@ private:
     [[nodiscard]] VkPipeline FindOrEmplaceClearStencilPipeline(
         const BlitDepthStencilPipelineKey& key);
     [[nodiscard]] VkPipeline FindOrEmplaceMSAACopyPipeline(const MSAACopyPipelineKey& key);
+
+    [[nodiscard]] VkPipeline FindOrEmplaceMSAACopyDepthPipeline(const MSAACopyPipelineKey& key,
+                                                                bool copy_stencil);
     [[nodiscard]] VkPipeline FindOrEmplaceBlitColorMSAAPipeline(const BlitMSAAPipelineKey& key);
+    [[nodiscard]] VkPipeline FindOrEmplaceBlitDepthStencilMSAAPipeline(
+        const BlitMSAAPipelineKey& key, bool blit_stencil);
+    [[nodiscard]] VkPipeline FindOrEmplaceBlitDepthPipeline(VkRenderPass renderpass);
     [[nodiscard]] VkPipeline FindOrEmplaceResolveDepthStencilPipeline(VkRenderPass renderpass,
                                                                       bool resolve_stencil);
 
@@ -162,10 +187,12 @@ private:
     vk::PipelineLayout two_textures_pipeline_layout;
     vk::PipelineLayout clear_color_pipeline_layout;
     vk::PipelineLayout msaa_copy_pipeline_layout;
+    vk::PipelineLayout msaa_copy_depth_stencil_pipeline_layout;
     vk::ShaderModule full_screen_vert;
     vk::ShaderModule blit_color_to_color_frag;
     vk::ShaderModule blit_color_msaa_frag;
     vk::ShaderModule blit_depth_stencil_frag;
+    vk::ShaderModule blit_depth_frag;
     vk::ShaderModule blit_depth_msaa_frag;
     vk::ShaderModule blit_depth_stencil_msaa_frag;
     vk::ShaderModule clear_color_vert;
@@ -179,7 +206,15 @@ private:
     vk::ShaderModule convert_d24s8_to_abgr8_frag;
     vk::ShaderModule convert_s8d24_to_abgr8_frag;
     vk::ShaderModule convert_msaa_to_non_msaa_frag;
+    vk::ShaderModule convert_msaa_to_non_msaa_sint_frag;
+    vk::ShaderModule convert_msaa_to_non_msaa_uint_frag;
+    vk::ShaderModule convert_msaa_to_non_msaa_depth_frag;
+    vk::ShaderModule convert_msaa_to_non_msaa_depth_stencil_frag;
     vk::ShaderModule convert_non_msaa_to_msaa_frag;
+    vk::ShaderModule convert_non_msaa_to_msaa_sint_frag;
+    vk::ShaderModule convert_non_msaa_to_msaa_uint_frag;
+    vk::ShaderModule convert_non_msaa_to_msaa_depth_frag;
+    vk::ShaderModule convert_non_msaa_to_msaa_depth_stencil_frag;
     vk::Sampler linear_sampler;
     vk::Sampler nearest_sampler;
 
@@ -193,8 +228,18 @@ private:
     std::vector<vk::Pipeline> clear_stencil_pipelines;
     std::vector<MSAACopyPipelineKey> msaa_copy_keys;
     std::vector<vk::Pipeline> msaa_copy_pipelines;
+    std::vector<MSAACopyPipelineKey> msaa_copy_depth_keys;
+    std::vector<vk::Pipeline> msaa_copy_depth_pipelines;
+    std::vector<MSAACopyPipelineKey> msaa_copy_depth_stencil_keys;
+    std::vector<vk::Pipeline> msaa_copy_depth_stencil_pipelines;
     std::vector<BlitMSAAPipelineKey> blit_msaa_color_keys;
     std::vector<vk::Pipeline> blit_msaa_color_pipelines;
+    std::vector<VkRenderPass> blit_depth_keys;
+    std::vector<vk::Pipeline> blit_depth_pipelines;
+    std::vector<BlitMSAAPipelineKey> blit_msaa_depth_keys;
+    std::vector<vk::Pipeline> blit_msaa_depth_pipelines;
+    std::vector<BlitMSAAPipelineKey> blit_msaa_depth_stencil_keys;
+    std::vector<vk::Pipeline> blit_msaa_depth_stencil_pipelines;
     std::vector<VkRenderPass> resolve_depth_keys;
     std::vector<vk::Pipeline> resolve_depth_pipelines;
     std::vector<VkRenderPass> resolve_depth_stencil_keys;

src/video_core/renderer_vulkan/pipeline_helper.h

@@ -370,7 +370,7 @@ inline void PushImageDescriptors(TextureCache& texture_cache,
                 const VkImageView null_image_view{texture_cache.GetImageView(VideoCommon::NULL_IMAGE_VIEW_ID).Handle(desc.type)};
                 if (null_image_view != VK_NULL_HANDLE) vk_image_view = null_image_view;
             }
-            const Sampler& sampler{texture_cache.GetSampler(sampler_id)};
+            Sampler& sampler{texture_cache.GetSampler(sampler_id)};
             const bool use_fallback_sampler{sampler.HasAddedAnisotropy() &&
                                             !image_view.SupportsAnisotropy()};
             VkSampler vk_sampler{use_fallback_sampler ? sampler.HandleWithDefaultAnisotropy()
@@ -383,6 +383,19 @@ inline void PushImageDescriptors(TextureCache& texture_cache,
                 !image_view.SupportsDepthComparison()) {
                 vk_sampler = sampler.HandleWithoutDepthComparison();
             }
+            const bool srgb_border{sampler.HasSrgbBorderColor() &&
+                                   VideoCore::Surface::IsPixelFormatSRGB(image_view.format)};
+            if (sampler.NeedsSwizzleMapping() && !image_view.HasIdentitySwizzle()) {
+                vk_sampler = sampler.HandleWithSwizzle(image_view.Swizzle(), srgb_border);
+            } else if (srgb_border) {
+                vk_sampler = sampler.HandleWithSrgbBorderColor();
+            }
+            if (sampler.HasMinmaxReduction() && !image_view.SupportsMinmaxFilter()) {
+                vk_sampler = sampler.HandleWithDefaultReduction();
+            }
+            if (sampler.HasCustomBorderColor() && image_view.RequiresBorderColorFormat()) {
+                vk_sampler = sampler.HandleWithDefaultBorderColor();
+            }
             guest_descriptor_queue.AddSampledImage(vk_image_view, vk_sampler);
             const bool element_rescaled{texture_cache.IsRescaling(image_view)};
             is_rescaled |= element_rescaled;

src/video_core/renderer_vulkan/present/sgsr.cpp

@@ -1,6 +1,8 @@
 // SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
+#include <algorithm>
+
 #include "common/common_types.h"
 #include "common/div_ceil.h"
 #include "common/settings.h"
@@ -17,7 +19,7 @@
 
 namespace Vulkan {
 
-using PushConstants = std::array<u32, 4 + 2 + 1>;
+using PushConstants = std::array<u32, 4 + 2 + 2 + 1>;
 
 SGSR::SGSR(const Device& device, MemoryAllocator& memory_allocator, size_t image_count, VkExtent2D extent, bool edge_dir)
     : m_memory_allocator{memory_allocator}
@@ -100,26 +102,28 @@ VkImageView SGSR::Draw(const Device& device, Scheduler& scheduler, size_t image_
 
     const f32 input_image_width = f32(input_image_extent.width);
     const f32 input_image_height = f32(input_image_extent.height);
-    const f32 viewport_width = (crop_rect.right - crop_rect.left) * input_image_width;
-    const f32 viewport_height = (crop_rect.bottom - crop_rect.top) * input_image_height;
-    // expected [0, 2]
-    const f32 sharpening = f32(Settings::values.fsr_sharpening_slider.GetValue()) / 100.0f;
+    const f32 crop_width = (crop_rect.right - crop_rect.left) * input_image_width;
+    const f32 crop_height = (crop_rect.bottom - crop_rect.top) * input_image_height;
+    static constexpr f32 EDGE_SHARPNESS_MAX = 2.0f;
+    const f32 edge_sharpness =
+        EDGE_SHARPNESS_MAX - f32(Settings::values.fsr_sharpening_slider.GetValue()) / 200.0f;
 
-    // p = (tex * viewport) / input = [0,n] (normalized texcoords)
-    // p * input = [0,1024], [0,768]
     // layout( push_constant ) uniform constants {
     //     highp vec4 ViewportInfo[1];
     //     highp vec2 ResizeFactor;
+    //     highp vec2 CropOffset;
     //     highp float EdgeSharpness;
     // };
     PushConstants viewport_con{};
-    viewport_con[0] = std::bit_cast<u32>(std::abs(1.f / viewport_width));
-    viewport_con[1] = std::bit_cast<u32>(std::abs(1.f / viewport_height));
-    viewport_con[2] = std::bit_cast<u32>(std::abs(viewport_width));
-    viewport_con[3] = std::bit_cast<u32>(std::abs(viewport_height));
-    viewport_con[4] = std::bit_cast<u32>(viewport_width / input_image_width);
-    viewport_con[5] = std::bit_cast<u32>(viewport_height / input_image_height);
-    viewport_con[6] = std::bit_cast<u32>(sharpening);
+    viewport_con[0] = std::bit_cast<u32>(1.f / input_image_width);
+    viewport_con[1] = std::bit_cast<u32>(1.f / input_image_height);
+    viewport_con[2] = std::bit_cast<u32>(input_image_width);
+    viewport_con[3] = std::bit_cast<u32>(input_image_height);
+    viewport_con[4] = std::bit_cast<u32>(crop_width / input_image_width);
+    viewport_con[5] = std::bit_cast<u32>(crop_height / input_image_height);
+    viewport_con[6] = std::bit_cast<u32>((std::min)(crop_rect.left, crop_rect.right));
+    viewport_con[7] = std::bit_cast<u32>((std::min)(crop_rect.top, crop_rect.bottom));
+    viewport_con[8] = std::bit_cast<u32>(edge_sharpness);
 
     UploadImages(device, scheduler);
     UpdateDescriptorSets(device, source_image_view, image_index);

src/video_core/renderer_vulkan/present/util.cpp

@@ -521,7 +521,7 @@ static vk::Pipeline CreateWrappedPipelineImpl(
         .subpass = 0,
         .basePipelineHandle = 0,
         .basePipelineIndex = 0,
-    });
+    }, device.StaticPipelineCache());
 }
 
 vk::Pipeline CreateWrappedPipeline(const Device& device, vk::RenderPass& renderpass,

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -22,7 +22,9 @@
 #include "video_core/host_shaders/resolve_conditional_render_comp_spv.h"
 #include "video_core/host_shaders/vulkan_quad_indexed_comp_spv.h"
 #include "video_core/host_shaders/vulkan_uint8_comp_spv.h"
+#include "video_core/host_shaders/block_linear_unswizzle_2d_buffer_comp_spv.h"
 #include "video_core/host_shaders/block_linear_unswizzle_3d_bcn_comp_spv.h"
+#include "video_core/host_shaders/block_linear_unswizzle_3d_buffer_comp_spv.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
 #include "video_core/surface.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
@@ -268,7 +270,7 @@ ComputePass::ComputePass(const Device& device_, Scheduler& scheduler, Descriptor
         .layout = *layout,
         .basePipelineHandle = {},
         .basePipelineIndex = 0,
-    });
+    }, device.StaticPipelineCache());
 }
 
 ComputePass::~ComputePass() = default;
@@ -872,4 +874,493 @@ void BlockLinearUnswizzle3DPass::UnswizzleChunk(
     });
 }
 
+namespace {
+constexpr u32 BL2D_BINDING_INPUT_BUFFER = 0;
+constexpr u32 BL2D_BINDING_OUTPUT_BUFFER = 1;
+
+struct alignas(16) BlockLinearUnswizzle2DPushConstants {
+    std::array<u32, 3> dim;
+    u32 bytes_per_block_log2;
+    std::array<u32, 3> origin;
+    u32 layer_stride;
+    u32 block_size;
+    u32 x_shift;
+    u32 block_height;
+    u32 block_height_mask;
+};
+static_assert(sizeof(BlockLinearUnswizzle2DPushConstants) <= 128);
+
+constexpr std::array<VkDescriptorSetLayoutBinding, 2> BL2D_BINDINGS{{
+    {
+        .binding = BL2D_BINDING_INPUT_BUFFER,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .descriptorCount = 1,
+        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
+        .pImmutableSamplers = nullptr,
+    },
+    {
+        .binding = BL2D_BINDING_OUTPUT_BUFFER,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .descriptorCount = 1,
+        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
+        .pImmutableSamplers = nullptr,
+    },
+}};
+
+constexpr std::array<VkDescriptorUpdateTemplateEntry, 2> BL2D_TEMPLATE{{
+    {
+        .dstBinding = BL2D_BINDING_INPUT_BUFFER,
+        .dstArrayElement = 0,
+        .descriptorCount = 1,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .offset = BL2D_BINDING_INPUT_BUFFER * sizeof(DescriptorUpdateEntry),
+        .stride = sizeof(DescriptorUpdateEntry),
+    },
+    {
+        .dstBinding = BL2D_BINDING_OUTPUT_BUFFER,
+        .dstArrayElement = 0,
+        .descriptorCount = 1,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .offset = BL2D_BINDING_OUTPUT_BUFFER * sizeof(DescriptorUpdateEntry),
+        .stride = sizeof(DescriptorUpdateEntry),
+    },
+}};
+
+constexpr DescriptorBankInfo BL2D_BANK_INFO{
+    .uniform_buffers = 0,
+    .storage_buffers = 2,
+    .texture_buffers = 0,
+    .image_buffers = 0,
+    .textures = 0,
+    .images = 0,
+    .score = 2,
+};
+} // Anonymous namespace
+
+BlockLinearUnswizzle2DPass::BlockLinearUnswizzle2DPass(
+    const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
+    StagingBufferPool& staging_buffer_pool_,
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
+    : ComputePass(device_, scheduler_, descriptor_pool_, BL2D_BINDINGS, BL2D_TEMPLATE,
+                  BL2D_BANK_INFO,
+                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(BlockLinearUnswizzle2DPushConstants)>,
+                  BLOCK_LINEAR_UNSWIZZLE_2D_BUFFER_COMP_SPV),
+      scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
+      compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
+
+BlockLinearUnswizzle2DPass::~BlockLinearUnswizzle2DPass() = default;
+
+bool BlockLinearUnswizzle2DPass::IsSupported(const VideoCommon::ImageInfo& info) {
+    if (info.type != VideoCommon::ImageType::e2D) {
+        return false;
+    }
+    if (info.resources.levels != 1 || info.resources.layers != 1) {
+        return false;
+    }
+    if (info.num_samples > 1) {
+        return false;
+    }
+    if (VideoCore::Surface::IsPixelFormatASTC(info.format) ||
+        VideoCore::Surface::IsPixelFormatBCn(info.format)) {
+        return false;
+    }
+    if (info.format >= VideoCore::Surface::PixelFormat::MaxColorFormat) {
+        return false;
+    }
+    const u32 bytes_per_block = VideoCore::Surface::BytesPerBlock(info.format);
+    if (bytes_per_block != 4 && bytes_per_block != 8 && bytes_per_block != 16) {
+        return false;
+    }
+    return VideoCore::Surface::DefaultBlockWidth(info.format) == 1 &&
+           VideoCore::Surface::DefaultBlockHeight(info.format) == 1;
+}
+
+void BlockLinearUnswizzle2DPass::Unswizzle(
+    Image& image, const StagingBufferRef& swizzled,
+    std::span<const VideoCommon::SwizzleParameters> swizzles) {
+    if (swizzles.empty()) {
+        return;
+    }
+    const VideoCommon::SwizzleParameters& sw = swizzles.front();
+    const auto params = VideoCommon::Accelerated::MakeBlockLinearSwizzle2DParams(sw, image.info);
+
+    const u32 width = sw.num_tiles.width;
+    const u32 height = sw.num_tiles.height;
+    const u32 depth = image.info.resources.layers;
+    const u32 bytes_per_block = 1u << params.bytes_per_block_log2;
+    const VkDeviceSize output_size =
+        static_cast<VkDeviceSize>(width) * height * depth * bytes_per_block;
+
+    const StagingBufferRef output =
+        staging_buffer_pool.Request(static_cast<size_t>(output_size), MemoryUsage::DeviceLocal);
+
+    BlockLinearUnswizzle2DPushConstants pc{};
+    pc.dim = {width, height, depth};
+    pc.bytes_per_block_log2 = params.bytes_per_block_log2;
+    pc.origin = params.origin;
+    pc.layer_stride = params.layer_stride;
+    pc.block_size = params.block_size;
+    pc.x_shift = params.x_shift;
+    pc.block_height = params.block_height;
+    pc.block_height_mask = params.block_height_mask;
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+
+    compute_pass_descriptor_queue.Acquire(scheduler, 2);
+    compute_pass_descriptor_queue.AddBuffer(swizzled.buffer, sw.buffer_offset + swizzled.offset,
+                                            image.guest_size_bytes - sw.buffer_offset);
+    compute_pass_descriptor_queue.AddBuffer(output.buffer, output.offset, output_size);
+
+    const void* descriptor_data = compute_pass_descriptor_queue.UpdateData();
+    const VkDescriptorSet set = descriptor_allocator.Commit();
+
+    const u32 gx = Common::DivCeil(width, 16u);
+    const u32 gy = Common::DivCeil(height, 8u);
+    const bool is_initialized = image.ExchangeInitialization();
+
+    const VkBuffer out_buffer = output.buffer;
+    const VkDeviceSize out_offset = output.offset;
+    const VkImage dst_image = image.Handle();
+    const VkImageAspectFlags aspect = image.AspectMask();
+
+    scheduler.Record([this, set, descriptor_data, pc, gx, gy, depth, output_size, out_buffer,
+                      out_offset, dst_image, aspect, width, height,
+                      is_initialized](vk::CommandBuffer cmdbuf) {
+        if (dst_image == VK_NULL_HANDLE || out_buffer == VK_NULL_HANDLE) {
+            return;
+        }
+        device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
+        cmdbuf.PushConstants(*layout, VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(pc), &pc);
+        cmdbuf.Dispatch(gx, gy, depth);
+
+        const VkBufferMemoryBarrier buffer_barrier{
+            .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
+            .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .buffer = out_buffer,
+            .offset = out_offset,
+            .size = output_size,
+        };
+        const VkImageMemoryBarrier pre_copy{
+            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = static_cast<VkAccessFlags>(is_initialized ? VK_ACCESS_SHADER_READ_BIT
+                                                                      : VK_ACCESS_NONE),
+            .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+            .oldLayout = is_initialized ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_UNDEFINED,
+            .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .image = dst_image,
+            .subresourceRange{
+                .aspectMask = aspect,
+                .baseMipLevel = 0,
+                .levelCount = VK_REMAINING_MIP_LEVELS,
+                .baseArrayLayer = 0,
+                .layerCount = VK_REMAINING_ARRAY_LAYERS,
+            },
+        };
+        VkPipelineStageFlags pre_copy_src_stages = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
+        if (is_initialized) {
+            pre_copy_src_stages |= vk::PIPELINE_STAGE_GRAPHICS_COMPUTE;
+        } else {
+            pre_copy_src_stages |= VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
+        }
+        cmdbuf.PipelineBarrier(pre_copy_src_stages, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {},
+                               buffer_barrier, pre_copy);
+
+        const VkBufferImageCopy copy{
+            .bufferOffset = out_offset,
+            .bufferRowLength = 0,
+            .bufferImageHeight = 0,
+            .imageSubresource{
+                .aspectMask = aspect,
+                .mipLevel = 0,
+                .baseArrayLayer = 0,
+                .layerCount = depth,
+            },
+            .imageOffset = {0, 0, 0},
+            .imageExtent = {width, height, 1},
+        };
+        cmdbuf.CopyBufferToImage(out_buffer, dst_image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy);
+
+        const VkImageMemoryBarrier post_copy{
+            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
+            .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+            .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .image = dst_image,
+            .subresourceRange{
+                .aspectMask = aspect,
+                .baseMipLevel = 0,
+                .levelCount = VK_REMAINING_MIP_LEVELS,
+                .baseArrayLayer = 0,
+                .layerCount = VK_REMAINING_ARRAY_LAYERS,
+            },
+        };
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE,
+                               0, {}, {}, post_copy);
+    });
+}
+
+namespace {
+constexpr u32 BL3DB_BINDING_INPUT_BUFFER = 0;
+constexpr u32 BL3DB_BINDING_OUTPUT_BUFFER = 1;
+
+struct alignas(16) BlockLinearUnswizzle3DBufferPushConstants {
+    std::array<u32, 3> dim;
+    u32 bytes_per_block_log2;
+    std::array<u32, 3> origin;
+    u32 slice_size;
+    u32 block_size;
+    u32 x_shift;
+    u32 block_height;
+    u32 block_height_mask;
+    u32 block_depth;
+    u32 block_depth_mask;
+};
+static_assert(sizeof(BlockLinearUnswizzle3DBufferPushConstants) <= 128);
+
+constexpr std::array<VkDescriptorSetLayoutBinding, 2> BL3DB_BINDINGS{{
+    {
+        .binding = BL3DB_BINDING_INPUT_BUFFER,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .descriptorCount = 1,
+        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
+        .pImmutableSamplers = nullptr,
+    },
+    {
+        .binding = BL3DB_BINDING_OUTPUT_BUFFER,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .descriptorCount = 1,
+        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
+        .pImmutableSamplers = nullptr,
+    },
+}};
+
+constexpr std::array<VkDescriptorUpdateTemplateEntry, 2> BL3DB_TEMPLATE{{
+    {
+        .dstBinding = BL3DB_BINDING_INPUT_BUFFER,
+        .dstArrayElement = 0,
+        .descriptorCount = 1,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .offset = BL3DB_BINDING_INPUT_BUFFER * sizeof(DescriptorUpdateEntry),
+        .stride = sizeof(DescriptorUpdateEntry),
+    },
+    {
+        .dstBinding = BL3DB_BINDING_OUTPUT_BUFFER,
+        .dstArrayElement = 0,
+        .descriptorCount = 1,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .offset = BL3DB_BINDING_OUTPUT_BUFFER * sizeof(DescriptorUpdateEntry),
+        .stride = sizeof(DescriptorUpdateEntry),
+    },
+}};
+
+constexpr DescriptorBankInfo BL3DB_BANK_INFO{
+    .uniform_buffers = 0,
+    .storage_buffers = 2,
+    .texture_buffers = 0,
+    .image_buffers = 0,
+    .textures = 0,
+    .images = 0,
+    .score = 2,
+};
+} // Anonymous namespace
+
+BlockLinearUnswizzle3DBufferPass::BlockLinearUnswizzle3DBufferPass(
+    const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
+    StagingBufferPool& staging_buffer_pool_,
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
+    : ComputePass(device_, scheduler_, descriptor_pool_, BL3DB_BINDINGS, BL3DB_TEMPLATE,
+                  BL3DB_BANK_INFO,
+                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(BlockLinearUnswizzle3DBufferPushConstants)>,
+                  BLOCK_LINEAR_UNSWIZZLE_3D_BUFFER_COMP_SPV),
+      scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
+      compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
+
+BlockLinearUnswizzle3DBufferPass::~BlockLinearUnswizzle3DBufferPass() = default;
+
+bool BlockLinearUnswizzle3DBufferPass::IsSupported(const Device& device,
+                                                   const VideoCommon::ImageInfo& info) {
+    if (info.type != VideoCommon::ImageType::e3D) {
+        return false;
+    }
+    if (info.resources.levels != 1 || info.resources.layers != 1) {
+        return false;
+    }
+    if (info.num_samples > 1) {
+        return false;
+    }
+    if (info.size.depth <= 1) {
+        return false;
+    }
+    if (VideoCore::Surface::IsPixelFormatASTC(info.format)) {
+        return false;
+    }
+    if (VideoCore::Surface::IsPixelFormatBCn(info.format) && !device.IsOptimalBcnSupported()) {
+        return false;
+    }
+    if (info.format >= VideoCore::Surface::PixelFormat::MaxColorFormat) {
+        return false;
+    }
+    const u32 bytes_per_block = VideoCore::Surface::BytesPerBlock(info.format);
+    return bytes_per_block == 4 || bytes_per_block == 8 || bytes_per_block == 16;
+}
+
+void BlockLinearUnswizzle3DBufferPass::Unswizzle(
+    Image& image, const StagingBufferRef& swizzled,
+    std::span<const VideoCommon::SwizzleParameters> swizzles) {
+    if (swizzles.empty()) {
+        return;
+    }
+
+    const VideoCommon::SwizzleParameters& sw = swizzles.front();
+    const auto params = VideoCommon::Accelerated::MakeBlockLinearSwizzle3DParams(sw, image.info);
+
+    const u32 blocks_x = sw.num_tiles.width;
+    const u32 blocks_y = sw.num_tiles.height;
+    const u32 blocks_z = sw.num_tiles.depth;
+    const u32 bytes_per_block = 1u << params.bytes_per_block_log2;
+    const VkDeviceSize output_size =
+        static_cast<VkDeviceSize>(blocks_x) * blocks_y * blocks_z * bytes_per_block;
+
+    const StagingBufferRef output =
+        staging_buffer_pool.Request(static_cast<size_t>(output_size), MemoryUsage::DeviceLocal);
+
+    BlockLinearUnswizzle3DBufferPushConstants pc{};
+    pc.dim = {blocks_x, blocks_y, blocks_z};
+    pc.bytes_per_block_log2 = params.bytes_per_block_log2;
+    pc.origin = params.origin;
+    pc.slice_size = params.slice_size;
+    pc.block_size = params.block_size;
+    pc.x_shift = params.x_shift;
+    pc.block_height = params.block_height;
+    pc.block_height_mask = params.block_height_mask;
+    pc.block_depth = params.block_depth;
+    pc.block_depth_mask = params.block_depth_mask;
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+
+    compute_pass_descriptor_queue.Acquire(scheduler, 2);
+    compute_pass_descriptor_queue.AddBuffer(swizzled.buffer, sw.buffer_offset + swizzled.offset,
+                                            image.guest_size_bytes - sw.buffer_offset);
+    compute_pass_descriptor_queue.AddBuffer(output.buffer, output.offset, output_size);
+
+    const void* descriptor_data = compute_pass_descriptor_queue.UpdateData();
+    const VkDescriptorSet set = descriptor_allocator.Commit();
+
+    const u32 gx = Common::DivCeil(blocks_x, 8u);
+    const u32 gy = Common::DivCeil(blocks_y, 8u);
+    const u32 gz = Common::DivCeil(blocks_z, 4u);
+    const bool is_initialized = image.ExchangeInitialization();
+
+    const VkBuffer out_buffer = output.buffer;
+    const VkDeviceSize out_offset = output.offset;
+    const VkImage dst_image = image.Handle();
+    const VkImageAspectFlags aspect = image.AspectMask();
+    const VkExtent3D extent{
+        .width = image.info.size.width,
+        .height = image.info.size.height,
+        .depth = image.info.size.depth,
+    };
+
+    scheduler.Record([this, set, descriptor_data, pc, gx, gy, gz, output_size, out_buffer,
+                      out_offset, dst_image, aspect, extent,
+                      is_initialized](vk::CommandBuffer cmdbuf) {
+        if (dst_image == VK_NULL_HANDLE || out_buffer == VK_NULL_HANDLE) {
+            return;
+        }
+        device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
+        cmdbuf.PushConstants(*layout, VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(pc), &pc);
+        cmdbuf.Dispatch(gx, gy, gz);
+
+        const VkBufferMemoryBarrier buffer_barrier{
+            .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
+            .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .buffer = out_buffer,
+            .offset = out_offset,
+            .size = output_size,
+        };
+        const VkImageMemoryBarrier pre_copy{
+            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = static_cast<VkAccessFlags>(is_initialized ? VK_ACCESS_SHADER_READ_BIT
+                                                                      : VK_ACCESS_NONE),
+            .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+            .oldLayout = is_initialized ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_UNDEFINED,
+            .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .image = dst_image,
+            .subresourceRange{
+                .aspectMask = aspect,
+                .baseMipLevel = 0,
+                .levelCount = VK_REMAINING_MIP_LEVELS,
+                .baseArrayLayer = 0,
+                .layerCount = VK_REMAINING_ARRAY_LAYERS,
+            },
+        };
+        VkPipelineStageFlags pre_copy_src_stages = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
+        if (is_initialized) {
+            pre_copy_src_stages |= vk::PIPELINE_STAGE_GRAPHICS_COMPUTE;
+        } else {
+            pre_copy_src_stages |= VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
+        }
+        cmdbuf.PipelineBarrier(pre_copy_src_stages, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {},
+                               buffer_barrier, pre_copy);
+
+        const VkBufferImageCopy copy{
+            .bufferOffset = out_offset,
+            .bufferRowLength = 0,
+            .bufferImageHeight = 0,
+            .imageSubresource{
+                .aspectMask = aspect,
+                .mipLevel = 0,
+                .baseArrayLayer = 0,
+                .layerCount = 1,
+            },
+            .imageOffset = {0, 0, 0},
+            .imageExtent = extent,
+        };
+        cmdbuf.CopyBufferToImage(out_buffer, dst_image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy);
+
+        const VkImageMemoryBarrier post_copy{
+            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
+            .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+            .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .image = dst_image,
+            .subresourceRange{
+                .aspectMask = aspect,
+                .baseMipLevel = 0,
+                .levelCount = VK_REMAINING_MIP_LEVELS,
+                .baseArrayLayer = 0,
+                .layerCount = VK_REMAINING_ARRAY_LAYERS,
+            },
+        };
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE,
+                               0, {}, {}, post_copy);
+    });
+}
+
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pass.h

@@ -164,4 +164,42 @@ private:
     ComputePassDescriptorQueue& compute_pass_descriptor_queue;
 };
 
+class BlockLinearUnswizzle2DPass final : public ComputePass {
+public:
+    explicit BlockLinearUnswizzle2DPass(const Device& device_, Scheduler& scheduler_,
+                                        DescriptorPool& descriptor_pool_,
+                                        StagingBufferPool& staging_buffer_pool_,
+                                        ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
+    ~BlockLinearUnswizzle2DPass();
+
+    [[nodiscard]] static bool IsSupported(const VideoCommon::ImageInfo& info);
+
+    void Unswizzle(Image& image, const StagingBufferRef& swizzled,
+                   std::span<const VideoCommon::SwizzleParameters> swizzles);
+
+private:
+    Scheduler& scheduler;
+    StagingBufferPool& staging_buffer_pool;
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
+};
+
+class BlockLinearUnswizzle3DBufferPass final : public ComputePass {
+public:
+    explicit BlockLinearUnswizzle3DBufferPass(
+        const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
+        StagingBufferPool& staging_buffer_pool_,
+        ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
+    ~BlockLinearUnswizzle3DBufferPass();
+
+    [[nodiscard]] static bool IsSupported(const Device& device, const VideoCommon::ImageInfo& info);
+
+    void Unswizzle(Image& image, const StagingBufferRef& swizzled,
+                   std::span<const VideoCommon::SwizzleParameters> swizzles);
+
+private:
+    Scheduler& scheduler;
+    StagingBufferPool& staging_buffer_pool;
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
+};
+
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -140,6 +140,8 @@ RenderPassKey MakeRenderPassKey(const FixedPipelineState& state, const Device& d
     });
     key.resolve_color =
         key.samples != VK_SAMPLE_COUNT_1_BIT && has_color && device.IsTiler();
+    key.resolve_depth_stencil = key.samples != VK_SAMPLE_COUNT_1_BIT && device.IsTiler() &&
+                                SupportsDepthStencilResolve(device, key.depth_format);
     return key;
 }
 
@@ -693,23 +695,30 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
     if (!key.state.dynamic_vertex_input) {
         const size_t num_vertex_arrays = (std::min)(
             Maxwell::NumVertexArrays, static_cast<size_t>(device.GetMaxVertexInputBindings()));
+        const u32 max_stride = device.GetMaxVertexInputBindingStride();
+        const u32 max_divisor = device.GetMaxVertexAttribDivisor();
         for (size_t index = 0; index < num_vertex_arrays; ++index) {
             const bool instanced = key.state.binding_divisors[index] != 0;
             const auto rate =
                 instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX;
             vertex_bindings.push_back({
                 .binding = static_cast<u32>(index),
-                .stride = key.state.vertex_strides[index],
+                .stride = (std::min)(u32{key.state.vertex_strides[index]}, max_stride),
                 .inputRate = rate,
             });
             if (instanced) {
                 vertex_binding_divisors.push_back({
                     .binding = static_cast<u32>(index),
-                    .divisor = key.state.binding_divisors[index],
+                    .divisor = (std::min)(key.state.binding_divisors[index], max_divisor),
                 });
             }
         }
+        const size_t max_attributes = static_cast<size_t>(device.GetMaxVertexInputAttributes());
+        const u32 max_offset = device.GetMaxVertexInputAttributeOffset();
         for (size_t index = 0; index < key.state.attributes.size(); ++index) {
+            if (vertex_attributes.size() >= max_attributes) {
+                break;
+            }
             const auto& attribute = key.state.attributes[index];
             if (!attribute.enabled || !stage_infos[0].loads.Generic(index)) {
                 continue;
@@ -718,11 +727,10 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
                 .location = static_cast<u32>(index),
                 .binding = attribute.buffer,
                 .format = MaxwellToVK::VertexFormat(device, attribute.Type(), attribute.Size()),
-                .offset = attribute.offset,
+                .offset = (std::min)(attribute.offset.Value(), max_offset),
             });
         }
     }
-    ASSERT(vertex_attributes.size() <= device.GetMaxVertexInputAttributes());
 
     VkPipelineVertexInputStateCreateInfo vertex_input_ci{
         .sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -56,13 +56,16 @@ using Shader::Backend::SPIRV::EmitSPIRV;
 using Shader::Maxwell::ConvertLegacyToGeneric;
 using Shader::Maxwell::GenerateGeometryPassthrough;
 using Shader::Maxwell::MergeDualVertexPrograms;
+using Shader::Maxwell::PrunePassthroughStores;
 using Shader::Maxwell::TranslateProgram;
 using VideoCommon::ComputeEnvironment;
 using VideoCommon::FileEnvironment;
 using VideoCommon::GenericEnvironment;
 using VideoCommon::GraphicsEnvironment;
 
-constexpr u32 CACHE_VERSION = 18;
+constexpr u32 CACHE_VERSION = 21;
+constexpr size_t VULKAN_CACHE_FLUSH_PIPELINES = 128;
+constexpr size_t VULKAN_CACHE_FLUSH_MIN_SECONDS = 30;
 constexpr std::array<char, 8> VULKAN_CACHE_MAGIC_NUMBER{'y', 'u', 'z', 'u', 'v', 'k', 'c', 'h'};
 
 template <typename Container>
@@ -150,6 +153,25 @@ Shader::AttributeType AttributeType(const FixedPipelineState& state, size_t inde
     return Shader::AttributeType::Disabled;
 }
 
+Shader::InputTopology MaxwellToInputTopology(Maxwell::PrimitiveTopology topology) {
+    switch (topology) {
+    case Maxwell::PrimitiveTopology::Points:
+        return Shader::InputTopology::Points;
+    case Maxwell::PrimitiveTopology::Lines:
+    case Maxwell::PrimitiveTopology::LineLoop:
+    case Maxwell::PrimitiveTopology::LineStrip:
+        return Shader::InputTopology::Lines;
+    case Maxwell::PrimitiveTopology::LinesAdjacency:
+    case Maxwell::PrimitiveTopology::LineStripAdjacency:
+        return Shader::InputTopology::LinesAdjacency;
+    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
+    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
+        return Shader::InputTopology::TrianglesAdjacency;
+    default:
+        return Shader::InputTopology::Triangles;
+    }
+}
+
 Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> programs,
                                     const GraphicsPipelineCacheKey& key,
                                     const Shader::IR::Program& program,
@@ -268,33 +290,7 @@ Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> program
     default:
         break;
     }
-    switch (key.state.topology) {
-    case Maxwell::PrimitiveTopology::Points:
-        info.input_topology = Shader::InputTopology::Points;
-        break;
-    case Maxwell::PrimitiveTopology::Lines:
-    case Maxwell::PrimitiveTopology::LineLoop:
-    case Maxwell::PrimitiveTopology::LineStrip:
-        info.input_topology = Shader::InputTopology::Lines;
-        break;
-    case Maxwell::PrimitiveTopology::Triangles:
-    case Maxwell::PrimitiveTopology::TriangleStrip:
-    case Maxwell::PrimitiveTopology::TriangleFan:
-    case Maxwell::PrimitiveTopology::Quads:
-    case Maxwell::PrimitiveTopology::QuadStrip:
-    case Maxwell::PrimitiveTopology::Polygon:
-    case Maxwell::PrimitiveTopology::Patches:
-        info.input_topology = Shader::InputTopology::Triangles;
-        break;
-    case Maxwell::PrimitiveTopology::LinesAdjacency:
-    case Maxwell::PrimitiveTopology::LineStripAdjacency:
-        info.input_topology = Shader::InputTopology::LinesAdjacency;
-        break;
-    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
-    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
-        info.input_topology = Shader::InputTopology::TrianglesAdjacency;
-        break;
-    }
+    info.input_topology = MaxwellToInputTopology(key.state.topology);
     info.force_early_z = key.state.early_z != 0;
     info.y_negate = key.state.y_negate != 0;
     return info;
@@ -699,6 +695,10 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
     if (use_vulkan_pipeline_cache) {
         SerializeVulkanPipelineCache(vulkan_pipeline_cache_filename, vulkan_pipeline_cache,
                                      CACHE_VERSION);
+        size_t size = 0;
+        vulkan_pipeline_cache.Read(&size, nullptr);
+        last_cache_size.store(size, std::memory_order_relaxed);
+        last_flush = std::chrono::steady_clock::now();
     }
 
     if (state.statistics) {
@@ -706,6 +706,35 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
     }
 }
 
+void PipelineCache::QueueVulkanPipelineCacheFlush() {
+    if (!use_vulkan_pipeline_cache || vulkan_pipeline_cache_filename.empty()) {
+        return;
+    }
+    if (++pipelines_since_flush < VULKAN_CACHE_FLUSH_PIPELINES) {
+        return;
+    }
+    const auto now = std::chrono::steady_clock::now();
+    const auto megabytes = last_cache_size.load(std::memory_order_relaxed) / (1024 * 1024);
+    const std::chrono::seconds interval{
+        std::max<size_t>(VULKAN_CACHE_FLUSH_MIN_SECONDS, megabytes)};
+    if (last_flush.time_since_epoch().count() != 0 && now - last_flush < interval) {
+        return;
+    }
+    if (flush_in_flight.exchange(true, std::memory_order_acq_rel)) {
+        return;
+    }
+    pipelines_since_flush = 0;
+    last_flush = now;
+    serialization_thread.QueueWork([this] {
+        SerializeVulkanPipelineCache(vulkan_pipeline_cache_filename, vulkan_pipeline_cache,
+                                     CACHE_VERSION);
+        size_t size = 0;
+        vulkan_pipeline_cache.Read(&size, nullptr);
+        last_cache_size.store(size, std::memory_order_relaxed);
+        flush_in_flight.store(false, std::memory_order_release);
+    });
+}
+
 GraphicsPipeline* PipelineCache::CurrentGraphicsPipelineSlowPath() {
     const auto [pair, is_new]{graphics_cache.try_emplace(graphics_key)};
     auto& pipeline{pair->second};
@@ -744,7 +773,7 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
     std::span<Shader::Environment* const> envs, PipelineStatistics* statistics,
     bool build_in_parallel) try {
     auto hash = key.Hash();
-    LOG_INFO(Render_Vulkan, "{:#016x}", hash);
+    LOG_DEBUG(Render_Vulkan, "{:#016x}", hash);
     size_t env_index{0};
     std::array<Shader::IR::Program, Maxwell::MaxShaderProgram> programs;
     const bool uses_vertex_a{key.unique_hashes[0] != 0};
@@ -758,8 +787,10 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
                                        index == static_cast<u32>(Maxwell::ShaderType::Geometry);
         if (key.unique_hashes[index] == 0 && is_emulated_stage) {
             auto topology = MaxwellToOutputTopology(key.state.topology);
-            programs[index] = GenerateGeometryPassthrough(pools.inst, pools.block, host_info,
-                                                          *layer_source_program, topology);
+            programs[index] =
+                GenerateGeometryPassthrough(pools.inst, pools.block, host_info,
+                                            *layer_source_program, topology,
+                                            MaxwellToInputTopology(key.state.topology));
             continue;
         }
         if (key.unique_hashes[index] == 0) {
@@ -772,11 +803,13 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
         Shader::Maxwell::Flow::CFG cfg(env, pools.flow_block, cfg_offset, index == 0);
         if (!uses_vertex_a || index != 1) {
             // Normal path
-            programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg, host_info);
+            programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
+                                              MaxwellToInputTopology(key.state.topology));
         } else {
             // VertexB path when VertexA is present.
             auto& program_va{programs[0]};
-            auto program_vb{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
+            auto program_vb{TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
+                                             MaxwellToInputTopology(key.state.topology))};
             programs[index] = MergeDualVertexPrograms(program_va, program_vb, env);
         }
 
@@ -806,6 +839,9 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
         const size_t stage_index{index - 1};
         infos[stage_index] = &program.info;
 
+        if (previous_stage) {
+            PrunePassthroughStores(program, previous_stage->info.stores);
+        }
         const auto runtime_info{MakeRuntimeInfo(programs, key, program, previous_stage, device)};
         ConvertLegacyToGeneric(program, runtime_info);
         const std::vector<u32> code{EmitSPIRV(profile, runtime_info, program, binding)};
@@ -880,6 +916,7 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline() {
         }
         SerializePipeline(key, env_ptrs, pipeline_cache_filename, CACHE_VERSION);
     });
+    QueueVulkanPipelineCacheFlush();
     return pipeline;
 }
 
@@ -899,6 +936,7 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
         SerializePipeline(key, std::array<const GenericEnvironment*, 1>{&env_},
                           pipeline_cache_filename, CACHE_VERSION);
     });
+    QueueVulkanPipelineCacheFlush();
     return pipeline;
 }
 
@@ -911,7 +949,7 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
         return nullptr;
     }
 
-    LOG_INFO(Render_Vulkan, "{:#016x}", hash);
+    LOG_DEBUG(Render_Vulkan, "{:#016x}", hash);
 
     Shader::Maxwell::Flow::CFG cfg{env, pools.flow_block, env.StartAddress()};
 
@@ -920,7 +958,8 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
         env.Dump(hash, key.unique_hash);
     }
 
-    auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
+    auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
+                                  Shader::InputTopology::Points)};
     const VkDriverIdKHR driver_id = device.GetDriverID();
     const bool needs_shared_mem_clamp =
         driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY ||

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -7,6 +7,8 @@
 #pragma once
 
 #include <array>
+#include <atomic>
+#include <chrono>
 #include <cstddef>
 #include <filesystem>
 #include <memory>
@@ -144,6 +146,8 @@ private:
     vk::PipelineCache LoadVulkanPipelineCache(const std::filesystem::path& filename,
                                               u32 expected_cache_version);
 
+    void QueueVulkanPipelineCacheFlush();
+
     const Device& device;
     Scheduler& scheduler;
     DescriptorPool& descriptor_pool;
@@ -171,6 +175,10 @@ private:
 
     std::filesystem::path vulkan_pipeline_cache_filename;
     vk::PipelineCache vulkan_pipeline_cache;
+    size_t pipelines_since_flush{};
+    std::chrono::steady_clock::time_point last_flush{};
+    std::atomic<size_t> last_cache_size{};
+    std::atomic_bool flush_in_flight{};
 
     Common::ThreadWorker workers;
     Common::ThreadWorker serialization_thread;

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -725,6 +725,9 @@ public:
         if (!device.IsExtTransformFeedbackSupported()) {
             return;
         }
+        if (!scheduler.IsRenderPassActive()) {
+            return;
+        }
         FlushBeginTFB();
         has_started = true;
     }
@@ -741,6 +744,9 @@ public:
         if (has_flushed_end_pending) {
             if (scheduler.IsRenderPassActive()) {
                 FlushEndTFB();
+            } else {
+                has_flushed_end_pending = false;
+                has_started = false;
             }
         }
         runtime.View3DRegs([this](Maxwell3D& maxwell3d) {

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -424,24 +424,12 @@ void RasterizerVulkan::Clear(u32 layer_count) {
     const bool ds_deferrable =
         !ds_used || ((!framebuffer->HasAspectDepthBit() || use_depth) &&
                      (!framebuffer->HasAspectStencilBit() || use_stencil) && !stencil_partial);
-    const bool can_defer_clear = ENABLE_DEFERRED_CLEAR && !regs.clear_control.use_scissor &&
-                                 regs.clear_surface.layer == 0 &&
-                                 !scheduler.IsRenderPassActive() &&
-                                 (!use_color || color_full_channels) && ds_deferrable;
-    if (!can_defer_clear) {
-        scheduler.RequestRenderpass(framebuffer);
-    }
-
-    query_cache.NotifySegment(true);
-    query_cache.CounterEnable(VideoCommon::QueryType::ZPassPixelCount64, maxwell3d->regs.zpass_pixel_count_enable);
     u32 up_scale = 1;
     u32 down_shift = 0;
     if (texture_cache.IsRescaling()) {
         up_scale = Settings::values.resolution_info.up_scale;
         down_shift = Settings::values.resolution_info.down_shift;
     }
-    UpdateViewportsState(regs);
-
     VkRect2D default_scissor{};
     default_scissor.offset.x = 0;
     default_scissor.offset.y = 0;
@@ -497,6 +485,22 @@ void RasterizerVulkan::Clear(u32 layer_count) {
         return;
     }
 
+    const bool clear_covers_render_area =
+        clear_rect.rect.offset.x == 0 && clear_rect.rect.offset.y == 0 &&
+        clear_rect.rect.extent.width >= render_area.width &&
+        clear_rect.rect.extent.height >= render_area.height;
+    const bool can_defer_clear = ENABLE_DEFERRED_CLEAR && (!regs.clear_control.use_scissor || clear_covers_render_area) &&
+                                 regs.clear_surface.layer == 0 &&
+                                 !scheduler.IsRenderPassActive() &&
+                                 (!use_color || color_full_channels) && ds_deferrable;
+    if (!can_defer_clear) {
+        scheduler.RequestRenderpass(framebuffer);
+    }
+
+    query_cache.NotifySegment(true);
+    query_cache.CounterEnable(VideoCommon::QueryType::ZPassPixelCount64, maxwell3d->regs.zpass_pixel_count_enable);
+    UpdateViewportsState(regs);
+
     const u32 color_attachment = regs.clear_surface.RT;
     if (use_color && framebuffer->HasAspectColorBit(color_attachment)) {
         const auto format = VideoCore::Surface::PixelFormatFromRenderTargetFormat(regs.rt[color_attachment].format);
@@ -852,8 +856,11 @@ void RasterizerVulkan::WaitForIdle() {
         VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT | VK_PIPELINE_STAGE_VERTEX_INPUT_BIT |
         VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT |
         VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT |
-        VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
-        VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_TRANSFER_BIT;
+        VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
+        VK_PIPELINE_STAGE_TRANSFER_BIT;
+    if (device.SupportsGeometryShader()) {
+        flags |= VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT;
+    }
     if (device.IsExtTransformFeedbackSupported()) {
         flags |= VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT;
     }
@@ -1286,33 +1293,42 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {
                         regs.zeta.format == Tegra::DepthFormat::X8Z24_UNORM ||
                         regs.zeta.format == Tegra::DepthFormat::S8Z24_UNORM ||
                         regs.zeta.format == Tegra::DepthFormat::V8Z24_UNORM;
+    const bool forces_unorm_representation = device.IsExtDepthBiasControlSupported();
 
     if (is_d24 && !device.SupportsD24DepthBuffer()) {
-        static constexpr const size_t length = sizeof(NEEDS_D24) / sizeof(NEEDS_D24[0]);
+        static constexpr double GUEST_TO_HOST_UNORM_BITS =
+            static_cast<double>(1ULL << (32 - 24));
 
-        static constexpr const u64* start = NEEDS_D24;
-        static constexpr const u64* end = NEEDS_D24 + length;
+        if (forces_unorm_representation) {
+            units = static_cast<float>(static_cast<double>(units) * GUEST_TO_HOST_UNORM_BITS);
+        } else {
+            static constexpr const size_t length = sizeof(NEEDS_D24) / sizeof(NEEDS_D24[0]);
 
-        const u64* it = std::find(start, end, program_id);
+            static constexpr const u64* start = NEEDS_D24;
+            static constexpr const u64* end = NEEDS_D24 + length;
 
-        if (it != end) {
-            // the base formulas can be obtained from here:
-            //   https://docs.microsoft.com/en-us/windows/win32/direct3d11/d3d10-graphics-programming-guide-output-merger-stage-depth-bias
-            const double rescale_factor =
-                static_cast<double>(1ULL << (32 - 24)) / (static_cast<double>(0x1.ep+127));
-            units = static_cast<float>(static_cast<double>(units) * rescale_factor);
+            const u64* it = std::find(start, end, program_id);
+
+            if (it != end) {
+                // the base formulas can be obtained from here:
+                //   https://docs.microsoft.com/en-us/windows/win32/direct3d11/d3d10-graphics-programming-guide-output-merger-stage-depth-bias
+                const double rescale_factor =
+                    GUEST_TO_HOST_UNORM_BITS / (static_cast<double>(0x1.ep+127));
+                units = static_cast<float>(static_cast<double>(units) * rescale_factor);
+            }
         }
     }
 
     scheduler.Record([constant = units, clamp = regs.depth_bias_clamp,
-                      factor = regs.slope_scale_depth_bias, this](vk::CommandBuffer cmdbuf) {
-        if (device.IsExtDepthBiasControlSupported()) {
-            static VkDepthBiasRepresentationInfoEXT bias_info{
+                      factor = regs.slope_scale_depth_bias,
+                      forces_unorm_representation, this](vk::CommandBuffer cmdbuf) {
+        if (forces_unorm_representation) {
+            const VkDepthBiasRepresentationInfoEXT bias_info{
                 .sType = VK_STRUCTURE_TYPE_DEPTH_BIAS_REPRESENTATION_INFO_EXT,
                 .pNext = nullptr,
                 .depthBiasRepresentation =
                     VK_DEPTH_BIAS_REPRESENTATION_LEAST_REPRESENTABLE_VALUE_FORCE_UNORM_EXT,
-                .depthBiasExact = VK_FALSE,
+                .depthBiasExact = static_cast<VkBool32>(device.HasExactDepthBiasControl()),
             };
 
             cmdbuf.SetDepthBias(constant, clamp, factor, &bias_info);
@@ -1336,8 +1352,14 @@ void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs)
     if (!state_tracker.TouchDepthBounds()) {
         return;
     }
-    scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]](
-                         vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBounds(min, max); });
+    float min = regs.depth_bounds[0];
+    float max = regs.depth_bounds[1];
+    if (!device.IsExtDepthRangeUnrestrictedSupported()) {
+        min = std::clamp(min, 0.0f, 1.0f);
+        max = std::clamp(max, 0.0f, 1.0f);
+    }
+    scheduler.Record(
+        [min, max](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBounds(min, max); });
 }
 
 void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1892,7 +1914,9 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs)
     const u32 max_bindings =
         static_cast<u32>(std::min<size_t>(Maxwell::NumVertexArrays,
                                           device.GetMaxVertexInputBindings()));
-
+    const u32 max_offset = device.GetMaxVertexInputAttributeOffset();
+    const u32 max_stride = device.GetMaxVertexInputBindingStride();
+    const u32 max_divisor = device.GetMaxVertexAttribDivisor();
 
     for (u32 index = 0; index < max_attributes; ++index) {
         const Maxwell::VertexAttribute attribute{regs.vertex_attrib_format[index]};
@@ -1906,20 +1930,24 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs)
             .location = index,
             .binding = binding,
             .format = MaxwellToVK::VertexFormat(device, attribute.type, attribute.size),
-            .offset = attribute.offset,
+            .offset = (std::min)(attribute.offset.Value(), max_offset),
         });
     }
 
     for (u32 binding = 0; binding < max_bindings; ++binding) {
         const auto& input_binding{regs.vertex_streams[binding]};
         const bool is_instanced{regs.vertex_stream_instances.IsInstancingEnabled(binding)};
+        u32 divisor = 1;
+        if (is_instanced) {
+            divisor = (std::min)(input_binding.frequency, max_divisor);
+        }
         bindings.push_back({
             .sType = VK_STRUCTURE_TYPE_VERTEX_INPUT_BINDING_DESCRIPTION_2_EXT,
             .pNext = nullptr,
             .binding = binding,
-            .stride = input_binding.stride,
+            .stride = (std::min)(input_binding.stride.Value(), max_stride),
             .inputRate = is_instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX,
-            .divisor = is_instanced ? input_binding.frequency : 1,
+            .divisor = divisor,
         });
     }
 

src/video_core/renderer_vulkan/vk_render_pass_cache.cpp

@@ -65,8 +65,72 @@ using VideoCore::Surface::SurfaceType;
                 .finalLayout = VK_IMAGE_LAYOUT_GENERAL,
             };
         }
+
+        struct ResolveAspects {
+            bool depth;
+            bool stencil;
+        };
+
+        struct ResolveModes {
+            VkResolveModeFlagBits depth;
+            VkResolveModeFlagBits stencil;
+        };
+
+        constexpr ResolveAspects GetResolveAspects(PixelFormat format) {
+            const SurfaceType surface_type = GetSurfaceType(format);
+            return ResolveAspects{
+                .depth = surface_type == SurfaceType::Depth ||
+                         surface_type == SurfaceType::DepthStencil,
+                .stencil = surface_type == SurfaceType::Stencil ||
+                           surface_type == SurfaceType::DepthStencil,
+            };
+        }
+
+        ResolveModes PickResolveModes(const Device& device, PixelFormat format) {
+            constexpr VkResolveModeFlagBits mode = VK_RESOLVE_MODE_SAMPLE_ZERO_BIT;
+
+            const ResolveAspects aspects = GetResolveAspects(format);
+            const bool depth_mode_supported = (device.GetDepthResolveModes() & mode) != 0;
+            const bool stencil_mode_supported = (device.GetStencilResolveModes() & mode) != 0;
+
+            ResolveModes modes{
+                .depth = VK_RESOLVE_MODE_NONE,
+                .stencil = VK_RESOLVE_MODE_NONE,
+            };
+            if (aspects.depth && depth_mode_supported) {
+                modes.depth = mode;
+            }
+            if (aspects.stencil && stencil_mode_supported) {
+                modes.stencil = mode;
+            }
+            if (modes.depth == modes.stencil || device.SupportsIndependentResolveNone()) {
+                return modes;
+            }
+            if (modes.depth != VK_RESOLVE_MODE_NONE && stencil_mode_supported) {
+                modes.stencil = mode;
+            } else if (modes.stencil != VK_RESOLVE_MODE_NONE && depth_mode_supported) {
+                modes.depth = mode;
+            }
+            return modes;
+        }
     } // Anonymous namespace
 
+bool SupportsDepthStencilResolve(const Device& device, PixelFormat depth_format) {
+    if (depth_format == PixelFormat::Invalid || !device.IsKhrDepthStencilResolveSupported()) {
+        return false;
+    }
+    const ResolveAspects aspects = GetResolveAspects(depth_format);
+    if (!aspects.depth && !aspects.stencil) {
+        return false;
+    }
+    const ResolveModes modes = PickResolveModes(device, depth_format);
+    if ((aspects.depth && modes.depth == VK_RESOLVE_MODE_NONE) ||
+        (aspects.stencil && modes.stencil == VK_RESOLVE_MODE_NONE)) {
+        return false;
+    }
+    return modes.depth == modes.stencil || device.SupportsIndependentResolveNone();
+}
+
 RenderPassCache::RenderPassCache(const Device& device_) : device{&device_} {}
 
 VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
@@ -75,7 +139,9 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
     if (!is_new) {
         return *pair->second;
     }
-    boost::container::static_vector<VkAttachmentDescription, 9> descriptions;
+    static constexpr size_t MAX_ATTACHMENTS =
+        2 * std::tuple_size_v<decltype(RenderPassKey::color_formats)> + 2;
+    boost::container::static_vector<VkAttachmentDescription, MAX_ATTACHMENTS> descriptions;
     std::array<VkAttachmentReference, 8> references{};
     u32 num_attachments{};
     u32 num_colors{};
@@ -109,8 +175,11 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
         const VkAttachmentLoadOp depth_load_op = key.depth_stencil_clear
                                                      ? VK_ATTACHMENT_LOAD_OP_CLEAR
                                                      : VK_ATTACHMENT_LOAD_OP_LOAD;
+        const VkAttachmentStoreOp depth_store_op = key.depth_stencil_discard
+                                                       ? VK_ATTACHMENT_STORE_OP_DONT_CARE
+                                                       : VK_ATTACHMENT_STORE_OP_STORE;
         descriptions.push_back(AttachmentDescription(*device, key.depth_format, key.samples,
-                                                     depth_load_op, VK_ATTACHMENT_STORE_OP_STORE));
+                                                     depth_load_op, depth_store_op));
     }
     std::array<VkAttachmentReference, 8> resolve_references{};
     const bool do_resolve_color =
@@ -133,6 +202,21 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
             }
         }
     }
+    const bool do_resolve_depth_stencil = key.resolve_depth_stencil && has_depth &&
+                                          key.samples != VK_SAMPLE_COUNT_1_BIT &&
+                                          SupportsDepthStencilResolve(*device, key.depth_format);
+    VkAttachmentReference depth_resolve_reference{};
+    if (do_resolve_depth_stencil) {
+        depth_resolve_reference = VkAttachmentReference{
+            .attachment = static_cast<u32>(descriptions.size()),
+            .layout = VK_IMAGE_LAYOUT_GENERAL,
+        };
+        VkAttachmentDescription resolve_desc =
+            AttachmentDescription(*device, key.depth_format, VK_SAMPLE_COUNT_1_BIT,
+                                  VK_ATTACHMENT_LOAD_OP_DONT_CARE, VK_ATTACHMENT_STORE_OP_STORE);
+        resolve_desc.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
+        descriptions.push_back(resolve_desc);
+    }
     const VkSubpassDescription subpass{
         .flags = 0,
         .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
@@ -145,18 +229,117 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
         .preserveAttachmentCount = 0,
         .pPreserveAttachments = nullptr,
     };
-    const VkSubpassDependency dependency{
-            .srcSubpass = 0,  // Current subpass
-            .dstSubpass = 0,  // Same subpass (self-dependency)
-            .srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
-                            VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
-                            VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
-            .dstStageMask = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT,
-            .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
-                             VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
-            .dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT
+    const VkSubpassDependency counter_resume_dependency{
+        .srcSubpass = 0,
+        .dstSubpass = 0,
+        .srcStageMask = VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT,
+        .dstStageMask = VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT,
+        .srcAccessMask = VK_ACCESS_TRANSFORM_FEEDBACK_COUNTER_WRITE_BIT_EXT,
+        .dstAccessMask = VK_ACCESS_TRANSFORM_FEEDBACK_COUNTER_READ_BIT_EXT,
+        .dependencyFlags = 0,
     };
+    const bool can_resume_transform_feedback = device->IsExtTransformFeedbackSupported();
+
+    if (device->IsKhrCreateRenderPass2Supported()) {
+        boost::container::static_vector<VkAttachmentDescription2, MAX_ATTACHMENTS> descriptions2;
+        for (const VkAttachmentDescription& description : descriptions) {
+            descriptions2.push_back(VkAttachmentDescription2{
+                .sType = VK_STRUCTURE_TYPE_ATTACHMENT_DESCRIPTION_2,
+                .pNext = nullptr,
+                .flags = description.flags,
+                .format = description.format,
+                .samples = description.samples,
+                .loadOp = description.loadOp,
+                .storeOp = description.storeOp,
+                .stencilLoadOp = description.stencilLoadOp,
+                .stencilStoreOp = description.stencilStoreOp,
+                .initialLayout = description.initialLayout,
+                .finalLayout = description.finalLayout,
+            });
+        }
+        const auto promote = [](const VkAttachmentReference& reference) {
+            return VkAttachmentReference2{
+                .sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2,
+                .pNext = nullptr,
+                .attachment = reference.attachment,
+                .layout = reference.layout,
+                .aspectMask = 0,
+            };
+        };
+        std::array<VkAttachmentReference2, 8> references2{};
+        std::array<VkAttachmentReference2, 8> resolve_references2{};
+        for (size_t index = 0; index < references.size(); ++index) {
+            references2[index] = promote(references[index]);
+            resolve_references2[index] = promote(resolve_references[index]);
+        }
+        const VkAttachmentReference2 depth_reference2 = promote(depth_reference);
+        const VkAttachmentReference2 depth_resolve_reference2 = promote(depth_resolve_reference);
+        const ResolveModes resolve_modes = PickResolveModes(*device, key.depth_format);
+        const VkSubpassDescriptionDepthStencilResolve depth_stencil_resolve{
+            .sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_DEPTH_STENCIL_RESOLVE,
+            .pNext = nullptr,
+            .depthResolveMode = resolve_modes.depth,
+            .stencilResolveMode = resolve_modes.stencil,
+            .pDepthStencilResolveAttachment = &depth_resolve_reference2,
+        };
+        const VkSubpassDescription2 subpass2{
+            .sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2,
+            .pNext = do_resolve_depth_stencil ? &depth_stencil_resolve : nullptr,
+            .flags = 0,
+            .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
+            .viewMask = 0,
+            .inputAttachmentCount = 0,
+            .pInputAttachments = nullptr,
+            .colorAttachmentCount = num_attachments,
+            .pColorAttachments = references2.data(),
+            .pResolveAttachments = do_resolve_color ? resolve_references2.data() : nullptr,
+            .pDepthStencilAttachment = has_depth ? &depth_reference2 : nullptr,
+            .preserveAttachmentCount = 0,
+            .pPreserveAttachments = nullptr,
+        };
+        const VkMemoryBarrier2 counter_resume_barrier{
+            .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER_2,
+            .pNext = nullptr,
+            .srcStageMask = VK_PIPELINE_STAGE_2_TRANSFORM_FEEDBACK_BIT_EXT,
+            .srcAccessMask = VK_ACCESS_2_TRANSFORM_FEEDBACK_COUNTER_WRITE_BIT_EXT,
+            .dstStageMask = VK_PIPELINE_STAGE_2_TRANSFORM_FEEDBACK_BIT_EXT,
+            .dstAccessMask = VK_ACCESS_2_TRANSFORM_FEEDBACK_COUNTER_READ_BIT_EXT,
+        };
+        VkSubpassDependency2 counter_resume_dependency2{
+            .sType = VK_STRUCTURE_TYPE_SUBPASS_DEPENDENCY_2,
+            .pNext = nullptr,
+            .srcSubpass = counter_resume_dependency.srcSubpass,
+            .dstSubpass = counter_resume_dependency.dstSubpass,
+            .srcStageMask = counter_resume_dependency.srcStageMask,
+            .dstStageMask = counter_resume_dependency.dstStageMask,
+            .srcAccessMask = counter_resume_dependency.srcAccessMask,
+            .dstAccessMask = counter_resume_dependency.dstAccessMask,
+            .dependencyFlags = counter_resume_dependency.dependencyFlags,
+            .viewOffset = 0,
+        };
+        if (device->HasSynchronization2()) {
+            counter_resume_dependency2.pNext = &counter_resume_barrier;
+            counter_resume_dependency2.srcStageMask = 0;
+            counter_resume_dependency2.dstStageMask = 0;
+            counter_resume_dependency2.srcAccessMask = 0;
+            counter_resume_dependency2.dstAccessMask = 0;
+        }
+        pair->second = device->GetLogical().CreateRenderPass2({
+            .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2,
+            .pNext = nullptr,
+            .flags = 0,
+            .attachmentCount = static_cast<u32>(descriptions2.size()),
+            .pAttachments = descriptions2.empty() ? nullptr : descriptions2.data(),
+            .subpassCount = 1,
+            .pSubpasses = &subpass2,
+            .dependencyCount = can_resume_transform_feedback ? 1u : 0u,
+            .pDependencies = can_resume_transform_feedback ? &counter_resume_dependency2 : nullptr,
+            .correlatedViewMaskCount = 0,
+            .pCorrelatedViewMasks = nullptr,
+        });
+        return *pair->second;
+    }
+
     pair->second = device->GetLogical().CreateRenderPass({
         .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
         .pNext = nullptr,
@@ -165,8 +348,8 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
         .pAttachments = descriptions.empty() ? nullptr : descriptions.data(),
         .subpassCount = 1,
         .pSubpasses = &subpass,
-        .dependencyCount = 1,
-        .pDependencies = &dependency,
+        .dependencyCount = can_resume_transform_feedback ? 1u : 0u,
+        .pDependencies = can_resume_transform_feedback ? &counter_resume_dependency : nullptr,
     });
     return *pair->second;
 }

src/video_core/renderer_vulkan/vk_render_pass_cache.h

@@ -9,6 +9,7 @@
 #include <mutex>
 #include <ankerl/unordered_dense.h>
 
+#include "common/container_hash.h"
 #include "video_core/surface.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
@@ -21,9 +22,11 @@ struct RenderPassKey {
     VideoCore::Surface::PixelFormat depth_format;
     VkSampleCountFlagBits samples;
     bool resolve_color;
+    bool resolve_depth_stencil;
     u32 color_clear_mask;
     bool depth_stencil_clear;
     u32 color_discard_mask;
+    bool depth_stencil_discard;
 };
 
 } // namespace Vulkan
@@ -31,17 +34,28 @@ struct RenderPassKey {
 namespace std {
 template <>
 struct hash<Vulkan::RenderPassKey> {
+    static_assert(std::tuple_size_v<decltype(Vulkan::RenderPassKey::color_formats)> <= 8);
+    static_assert(static_cast<u32>(VideoCore::Surface::PixelFormat::Invalid) <= 0xFF);
+    static_assert(static_cast<u32>(VideoCore::Surface::PixelFormat::Max) <= 0xFF);
+    static_assert(VK_SAMPLE_COUNT_64_BIT <= 0xFF);
+
     [[nodiscard]] size_t operator()(const Vulkan::RenderPassKey& key) const noexcept {
-        size_t value = static_cast<size_t>(key.depth_format) << 48;
-        value ^= static_cast<size_t>(key.samples) << 52;
-        value ^= static_cast<size_t>(key.resolve_color) << 63;
-        value ^= static_cast<size_t>(key.color_clear_mask) << 54;
-        value ^= static_cast<size_t>(key.depth_stencil_clear) << 62;
-        value ^= static_cast<size_t>(key.color_discard_mask) << 24;
-        for (size_t i = 0; i < key.color_formats.size(); ++i) {
-            value ^= static_cast<size_t>(key.color_formats[i]) << (i * 6);
+        u64 formats = 0;
+        for (size_t index = 0; index < key.color_formats.size(); ++index) {
+            formats |= static_cast<u64>(key.color_formats[index]) << (index * 8);
         }
-        return value;
+        const u64 state = static_cast<u64>(key.depth_format) |
+                          (static_cast<u64>(key.samples) << 8) |
+                          (static_cast<u64>(key.color_clear_mask) << 16) |
+                          (static_cast<u64>(key.color_discard_mask) << 24) |
+                          (static_cast<u64>(key.resolve_color) << 32) |
+                          (static_cast<u64>(key.depth_stencil_clear) << 33) |
+                          (static_cast<u64>(key.resolve_depth_stencil) << 34) |
+                          (static_cast<u64>(key.depth_stencil_discard) << 35);
+        size_t seed = 0;
+        Common::HashCombine(seed, formats);
+        Common::HashCombine(seed, state);
+        return seed;
     }
 };
 } // namespace std
@@ -50,6 +64,9 @@ namespace Vulkan {
 
 class Device;
 
+[[nodiscard]] bool SupportsDepthStencilResolve(const Device& device,
+                                               VideoCore::Surface::PixelFormat depth_format);
+
 class RenderPassCache {
 public:
     explicit RenderPassCache(const Device& device_);

src/video_core/renderer_vulkan/vk_scheduler.cpp

@@ -155,8 +155,10 @@ void Scheduler::RealizeDeferredClear() {
     }
     const u32 color_discard_mask =
         dc.framebuffer->DiscardsMsaaColor() ? dc.color_clear_mask : 0u;
+    const bool depth_stencil_discard =
+        dc.depth_stencil && dc.framebuffer->DiscardsMsaaDepthStencil();
     const VkRenderPass renderpass = dc.framebuffer->RenderPassVariant(
-        dc.color_clear_mask, dc.depth_stencil, color_discard_mask);
+        dc.color_clear_mask, dc.depth_stencil, color_discard_mask, depth_stencil_discard);
     EndRenderPass();
     BeginRenderPassImpl(dc.framebuffer, renderpass, clear_values.data(), count);
 }
@@ -410,8 +412,17 @@ void Scheduler::EndRenderPass()
         Record([num_images = num_renderpass_images,
                        images = renderpass_images,
                        ranges = renderpass_image_ranges,
+                       consumer_stages = device.AttachmentConsumerStages(),
                        has_transform_feedback = device.IsExtTransformFeedbackSupported()](
                           vk::CommandBuffer cmdbuf) {
+            static constexpr VkAccessFlags SHADER_ACCESS =
+                VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
+            static constexpr VkAccessFlags COLOR_ACCESS =
+                VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
+            static constexpr VkAccessFlags DEPTH_STENCIL_ACCESS =
+                VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
+                VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+
             std::array<VkImageMemoryBarrier, 9> barriers;
             for (size_t i = 0; i < num_images; ++i) {
                 const VkImageSubresourceRange& range = ranges[i];
@@ -421,24 +432,25 @@ void Scheduler::EndRenderPass()
                                                  | VK_IMAGE_ASPECT_STENCIL_BIT)) !=0;
 
                 VkAccessFlags src_access = 0;
+                VkAccessFlags dst_access = SHADER_ACCESS;
 
-                if (is_color)
+                if (is_color) {
                     src_access |= VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-                else if (is_depth_stencil)
+                    dst_access |= COLOR_ACCESS;
+                } else if (is_depth_stencil) {
                     src_access |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
-                else
+                    dst_access |= DEPTH_STENCIL_ACCESS;
+                } else {
                     src_access |= VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT
                                   | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+                    dst_access |= COLOR_ACCESS | DEPTH_STENCIL_ACCESS;
+                }
 
                 barriers[i] = VkImageMemoryBarrier{
                         .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
                         .pNext = nullptr,
                         .srcAccessMask = src_access,
-                        .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT
-                                         | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT
-                                         | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT
-                                         | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT
-                                         | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
+                        .dstAccessMask = dst_access,
                         .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
                         .newLayout = VK_IMAGE_LAYOUT_GENERAL,
                         .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
@@ -449,7 +461,7 @@ void Scheduler::EndRenderPass()
             }
             cmdbuf.EndRenderPass();
             cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
-                                   VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE,
+                                   VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, consumer_stages,
                                    0, nullptr, nullptr, vk::Span(barriers.data(), num_images));
             if (has_transform_feedback) {
                 static constexpr VkMemoryBarrier XFB_OUTPUT_BARRIER{

src/video_core/renderer_vulkan/vk_texture_cache.h

@@ -66,7 +66,7 @@ public:
 
     bool CanReportMemoryUsage() const;
 
-    std::optional<size_t> GetSamplerHeapBudget() const;
+    bool CanDownloadMsaa(const VideoCommon::ImageInfo& info) const;
 
     void BlitImage(Framebuffer* dst_framebuffer, ImageView& dst, ImageView& src,
                    const Region2D& dst_region, const Region2D& src_region,
@@ -117,11 +117,13 @@ public:
         VkFormat format = VK_FORMAT_UNDEFINED;
         VkExtent2D extent{};
         u32 layers = 0;
+        VkImageAspectFlags aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
         bool up_to_date = false;
     };
 
     [[nodiscard]] VkImageView GetOrCreateResolveShadow(VkImage msaa_image, VkFormat format,
-                                                       VkExtent2D extent, u32 layers);
+                                                       VkExtent2D extent, u32 layers,
+                                                       VkImageAspectFlags aspect_mask);
 
     [[nodiscard]] const ResolveShadow* GetValidResolveShadow(VkImage msaa_image) const;
 
@@ -149,6 +151,8 @@ public:
     std::optional<ASTCDecoderPass> astc_decoder_pass;
 
     std::optional<BlockLinearUnswizzle3DPass> bl3d_unswizzle_pass;
+    std::optional<BlockLinearUnswizzle2DPass> bl2d_unswizzle_pass;
+    std::optional<BlockLinearUnswizzle3DBufferPass> bl3db_unswizzle_pass;
     const Settings::ResolutionScalingInfo& resolution;
     std::array<std::vector<VkFormat>, VideoCore::Surface::MaxPixelFormat> view_formats;
 
@@ -156,6 +160,7 @@ public:
     std::array<vk::Buffer, indexing_slots> buffers{};
     std::vector<std::pair<u64, vk::Image>> pending_msaa_images;
     ankerl::unordered_dense::map<VkImage, ResolveShadow> resolve_shadows;
+    std::vector<std::pair<u64, ResolveShadow>> pending_resolve_shadows;
 };
 
 class Framebuffer {
@@ -191,7 +196,8 @@ public:
     }
 
     [[nodiscard]] VkRenderPass RenderPassVariant(u32 color_clear_mask, bool depth_stencil_clear,
-                                                 u32 color_discard_mask) const;
+                                                 u32 color_discard_mask,
+                                                 bool depth_stencil_discard) const;
 
     [[nodiscard]] VkExtent2D RenderArea() const noexcept {
         return render_area;
@@ -245,7 +251,13 @@ public:
         return discard_msaa_color;
     }
 
+    [[nodiscard]] bool DiscardsMsaaDepthStencil() const noexcept {
+        return discard_msaa_depth_stencil;
+    }
+
 private:
+    static constexpr size_t NUM_MEMOIZED_RENDER_PASS_VARIANTS = 8;
+
     vk::Framebuffer framebuffer;
     VkRenderPass renderpass{};
     VkExtent2D render_area{};
@@ -263,6 +275,10 @@ private:
     RenderPassKey render_pass_key{};
     RenderPassCache* render_pass_cache{nullptr};
     bool discard_msaa_color{};
+    bool discard_msaa_depth_stencil{};
+    mutable std::array<u32, NUM_MEMOIZED_RENDER_PASS_VARIANTS> variant_keys{};
+    mutable std::array<VkRenderPass, NUM_MEMOIZED_RENDER_PASS_VARIANTS> variant_render_passes{};
+    mutable u32 num_memoized_variants{};
 };
 
 class Image : public VideoCommon::ImageBase {
@@ -404,6 +420,22 @@ public:
         return supports_depth_comparison;
     }
 
+    [[nodiscard]] bool RequiresBorderColorFormat() const noexcept {
+        return requires_border_color_format;
+    }
+
+    [[nodiscard]] bool SupportsMinmaxFilter() const noexcept {
+        return supports_minmax_filter;
+    }
+
+    [[nodiscard]] const VkComponentMapping& Swizzle() const noexcept {
+        return swizzle_mapping;
+    }
+
+    [[nodiscard]] bool HasIdentitySwizzle() const noexcept {
+        return has_identity_swizzle;
+    }
+
     [[nodiscard]] GPUVAddr GpuAddr() const noexcept {
         return gpu_addr;
     }
@@ -436,11 +468,36 @@ private:
     VkSampleCountFlagBits samples = VK_SAMPLE_COUNT_1_BIT;
     u32 buffer_size = 0;
 
+    VkComponentMapping swizzle_mapping{};
+
     bool supports_depth_comparison = false;
+    bool requires_border_color_format = false;
+    bool supports_minmax_filter = false;
+    bool has_identity_swizzle = true;
 };
 
 class ImageAlloc : public VideoCommon::ImageAllocBase {};
 
+class CustomBorderColorBudget {
+public:
+    CustomBorderColorBudget() = default;
+    ~CustomBorderColorBudget();
+
+    CustomBorderColorBudget(const CustomBorderColorBudget&) = delete;
+    CustomBorderColorBudget& operator=(const CustomBorderColorBudget&) = delete;
+
+    CustomBorderColorBudget(CustomBorderColorBudget&& rhs) noexcept;
+    CustomBorderColorBudget& operator=(CustomBorderColorBudget&& rhs) noexcept;
+
+    bool TryAcquire(const Device& device, size_t count);
+
+private:
+    void Release() noexcept;
+
+    const Device* device_ptr = nullptr;
+    size_t held = 0;
+};
+
 class Sampler {
 public:
     explicit Sampler(TextureCacheRuntime&, const Tegra::Texture::TSCEntry&);
@@ -473,11 +530,61 @@ public:
         return static_cast<bool>(sampler_noncompare);
     }
 
+    [[nodiscard]] VkSampler HandleWithDefaultBorderColor() const noexcept {
+        return *sampler_default_border;
+    }
+
+    [[nodiscard]] bool HasCustomBorderColor() const noexcept {
+        return static_cast<bool>(sampler_default_border);
+    }
+
+    [[nodiscard]] VkSampler HandleWithDefaultReduction() const noexcept {
+        return *sampler_default_reduction;
+    }
+
+    [[nodiscard]] bool HasMinmaxReduction() const noexcept {
+        return static_cast<bool>(sampler_default_reduction);
+    }
+
+    [[nodiscard]] VkSampler HandleWithSrgbBorderColor() const noexcept {
+        return *sampler_srgb_border;
+    }
+
+    [[nodiscard]] bool HasSrgbBorderColor() const noexcept {
+        return static_cast<bool>(sampler_srgb_border);
+    }
+
+    [[nodiscard]] bool NeedsSwizzleMapping() const noexcept {
+        return needs_swizzle_mapping;
+    }
+
+    [[nodiscard]] VkSampler HandleWithSwizzle(const VkComponentMapping& mapping, bool srgb);
+
 private:
+    struct SwizzleVariant {
+        VkComponentMapping mapping;
+        bool srgb;
+        vk::Sampler sampler;
+    };
+
+    CustomBorderColorBudget custom_border_color_budget;
+
     vk::Sampler sampler;
     vk::Sampler sampler_default_anisotropy;
     vk::Sampler sampler_nearest;
     vk::Sampler sampler_noncompare;
+    vk::Sampler sampler_default_border;
+    vk::Sampler sampler_default_reduction;
+    vk::Sampler sampler_srgb_border;
+
+    std::vector<SwizzleVariant> swizzle_variants;
+    const Device* device_ptr = nullptr;
+    VkSamplerCreateInfo swizzle_base_ci{};
+    VkSamplerReductionModeEXT swizzle_reduction_mode = VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE_EXT;
+    std::array<float, 4> border_color_value{};
+    std::array<float, 4> srgb_border_color_value{};
+    bool swizzle_uses_reduction = false;
+    bool needs_swizzle_mapping = false;
 };
 
 struct TextureCacheParams {
@@ -486,6 +593,7 @@ struct TextureCacheParams {
     static constexpr bool HAS_EMULATED_COPIES = false;
     static constexpr bool HAS_DEVICE_MEMORY_INFO = true;
     static constexpr bool IMPLEMENTS_ASYNC_DOWNLOADS = true;
+    static constexpr bool HAS_MSAA_DOWNLOADS = true;
 
     using Runtime = Vulkan::TextureCacheRuntime;
     using Image = Vulkan::Image;

src/video_core/texture_cache/image_base.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
@@ -121,7 +121,7 @@ void ImageBase::InsertView(const ImageViewInfo& view_info, ImageViewId image_vie
     image_view_ids.push_back(image_view_id);
 }
 
-bool ImageBase::IsSafeDownload() const noexcept {
+bool ImageBase::IsSafeGpuCopy() const noexcept {
     // Skip images that were not modified from the GPU
     if (False(flags & ImageFlagBits::GpuModified)) {
         return false;
@@ -131,10 +131,6 @@ bool ImageBase::IsSafeDownload() const noexcept {
     if (True(flags & ImageFlagBits::CpuModified)) {
         return false;
     }
-    if (info.num_samples > 1) {
-        LOG_WARNING(HW_GPU, "MSAA image downloads are not implemented");
-        return false;
-    }
     return true;
 }
 

src/video_core/texture_cache/image_base.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -64,7 +67,7 @@ struct ImageBase {
 
     void InsertView(const ImageViewInfo& view_info, ImageViewId image_view_id);
 
-    [[nodiscard]] bool IsSafeDownload() const noexcept;
+    [[nodiscard]] bool IsSafeGpuCopy() const noexcept;
 
     [[nodiscard]] bool Overlaps(VAddr overlap_cpu_addr, size_t overlap_size) const noexcept {
         const VAddr overlap_end = overlap_cpu_addr + overlap_size;

src/video_core/texture_cache/texture_cache.h

@@ -134,7 +134,7 @@ void TextureCache<P>::RunGarbageCollector() {
         if (True(image.flags & ImageFlagBits::IsDecoding)) {
             return false;
         }
-        const bool must_download = image.IsSafeDownload() && False(image.flags & ImageFlagBits::BadOverlap);
+        const bool must_download = IsDownloadable(image) && False(image.flags & ImageFlagBits::BadOverlap);
         if ((!aggressive_mode && True(image.flags & ImageFlagBits::CostlyLoad)) || (!high_priority_mode && must_download)) {
             return false;
         }
@@ -594,11 +594,26 @@ void TextureCache<P>::WriteMemory(DAddr cpu_addr, size_t size) {
     });
 }
 
+template <class P>
+bool TextureCache<P>::IsDownloadable(const ImageBase& image) const noexcept {
+    if (!image.IsSafeGpuCopy()) {
+        return false;
+    }
+    if (image.info.num_samples == 1) {
+        return true;
+    }
+    if constexpr (P::HAS_MSAA_DOWNLOADS) {
+        return runtime.CanDownloadMsaa(image.info);
+    } else {
+        return false;
+    }
+}
+
 template <class P>
 void TextureCache<P>::DownloadMemory(DAddr cpu_addr, size_t size) {
     boost::container::small_vector<ImageId, 16> images;
-    ForEachImageInRegion(cpu_addr, size, [&images](ImageId image_id, ImageBase& image) {
-        if (!image.IsSafeDownload()) {
+    ForEachImageInRegion(cpu_addr, size, [this, &images](ImageId image_id, ImageBase& image) {
+        if (!IsDownloadable(image)) {
             return;
         }
         image.flags &= ~ImageFlagBits::GpuModified;
@@ -1476,12 +1491,12 @@ template <class P>
 bool TextureCache<P>::ScaleUp(Image& image) {
     const bool has_copy = image.HasScaled();
     const bool rescaled = image.ScaleUp();
+    if (!has_copy && image.HasScaled()) {
+        total_used_memory += GetScaledImageSizeBytes(image);
+    }
     if (!rescaled) {
         return false;
     }
-    if (!has_copy) {
-        total_used_memory += GetScaledImageSizeBytes(image);
-    }
     InvalidateScale(image);
     return true;
 }
@@ -1691,7 +1706,7 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, DA
     for (const auto& copy_object : join_copies_to_do) {
         Image& overlap = slot_images[copy_object.id];
         if (copy_object.is_alias) {
-            if (!overlap.IsSafeDownload()) {
+            if (!overlap.IsSafeGpuCopy()) {
                 continue;
             }
             const auto alias_pointer = join_alias_indices.find(copy_object.id);
@@ -1877,67 +1892,10 @@ SamplerId TextureCache<P>::FindSampler(const TSCEntry& config, bool compute) {
     const auto [pair, is_new] = channel_state->samplers.try_emplace(config);
     if (is_new) {
         pair->second = slot_samplers.insert(runtime, config);
-        EnforceSamplerBudget();
     }
     return pair->second;
 }
 
-template <class P>
-std::optional<size_t> TextureCache<P>::QuerySamplerBudget() const {
-    if constexpr (requires { runtime.GetSamplerHeapBudget(); }) {
-        return runtime.GetSamplerHeapBudget();
-    } else {
-        return std::nullopt;
-    }
-}
-
-template <class P>
-void TextureCache<P>::EnforceSamplerBudget() {
-    if (auto const budget = QuerySamplerBudget(); budget) {
-        if (slot_samplers.size() < *budget) {
-            return;
-        }
-        if (!channel_state) {
-            return;
-        }
-        if (last_sampler_gc_frame == frame_tick) {
-            return;
-        }
-        last_sampler_gc_frame = frame_tick;
-        TrimInactiveSamplers(*budget);
-    }
-}
-
-template <class P>
-void TextureCache<P>::TrimInactiveSamplers(size_t budget) {
-    if (channel_state->samplers.size() > 0) {
-        constexpr size_t SAMPLER_GC_SLACK = 1024;
-        ankerl::unordered_dense::set<SamplerId> active_sampler_ids;
-        for (auto const& e : channel_state->sampler_ids)
-            active_sampler_ids.insert(e.second);
-        // Elements in the map must be necesarily valid
-        size_t removed = 0;
-        for (auto it = channel_state->samplers.begin(); it != channel_state->samplers.end();) {
-            const SamplerId sampler_id = it->second;
-            if (!sampler_id || sampler_id == CORRUPT_ID) {
-                it = channel_state->samplers.erase(it);
-            } else if (std::ranges::find(active_sampler_ids, sampler_id) != active_sampler_ids.end()) {
-                ++it;
-            } else {
-                slot_samplers.erase(sampler_id);
-                it = channel_state->samplers.erase(it);
-                ++removed;
-                if (slot_samplers.size() + SAMPLER_GC_SLACK <= budget) {
-                    break;
-                }
-            }
-        }
-        if (removed != 0) {
-            LOG_WARNING(HW_GPU, "Sampler cache exceeded {} entries on this driver; reclaimed {} inactive samplers", budget, removed);
-        }
-    }
-}
-
 template <class P>
 ImageViewId TextureCache<P>::FindColorBuffer(size_t index) {
     const auto& regs = maxwell3d->regs;

src/video_core/texture_cache/texture_cache_base.h

@@ -303,6 +303,8 @@ private:
     FramebufferId GetFramebufferId(const RenderTargets& key);
 
     /// Refresh the contents (pixel data) of an image
+    [[nodiscard]] bool IsDownloadable(const ImageBase& image) const noexcept;
+
     void RefreshContents(Image& image, ImageId image_id);
 
     /// Upload data from guest to an image
@@ -418,9 +420,6 @@ private:
 
     void QueueAsyncDecode(Image& image, ImageId image_id);
     void TickAsyncDecode();
-    void EnforceSamplerBudget();
-    void TrimInactiveSamplers(size_t budget);
-    std::optional<size_t> QuerySamplerBudget() const;
 
     void QueueAsyncUnswizzle(Image& image, ImageId image_id);
     void TickAsyncUnswizzle();
@@ -507,7 +506,6 @@ private:
 
     u64 modification_tick = 0;
     u64 frame_tick = 0;
-    u64 last_sampler_gc_frame = (std::numeric_limits<u64>::max)();
 
     Common::ThreadWorker texture_decode_worker{1, "TextureDecoder", {},
                                                Common::ThreadPlacement::Efficiency};

src/video_core/textures/decoders.cpp

@@ -1,9 +1,10 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include <algorithm>
 #include <array>
 #include <cmath>
 #include <cstring>
@@ -30,12 +31,90 @@ constexpr u32 pdep(u32 value) {
     return result;
 }
 
+constexpr u32 SWIZZLE_RUN_BYTES = 16;
+constexpr u32 SWIZZLE_RUN_SHIFT = 4;
+constexpr u32 SWIZZLE_RUN_MASK = SWIZZLE_RUN_BYTES - 1;
+constexpr u32 SWIZZLE_RUN_INDEX_MASK = GOB_SIZE_X / SWIZZLE_RUN_BYTES - 1;
+
+static_assert((SWIZZLE_X_BITS & SWIZZLE_RUN_MASK) == SWIZZLE_RUN_MASK);
+
+constexpr std::array<u32, GOB_SIZE_X / SWIZZLE_RUN_BYTES> SWIZZLE_X_RUN_TABLE = [] {
+    std::array<u32, GOB_SIZE_X / SWIZZLE_RUN_BYTES> table{};
+    for (u32 index = 0; index < static_cast<u32>(table.size()); ++index) {
+        table[index] = pdep<SWIZZLE_X_BITS>(index << SWIZZLE_RUN_SHIFT);
+    }
+    return table;
+}();
+
 template <u32 mask, u32 incr_amount>
 void incrpdep(u32& value) {
     static constexpr u32 swizzled_incr = pdep<mask>(incr_amount);
     value = ((value | ~mask) + swizzled_incr) & mask;
 }
 
+template <bool TO_LINEAR>
+void SwizzleRow(std::span<u8> output, std::span<const u8> input, u32 offset_zy, u32 swizzled_y,
+                u32 x_shift, u32 x, u32 num_bytes, u32 linear) {
+    const auto copy = [&](u32 swizzled_x, u32 count) {
+        const u32 swizzled =
+            offset_zy + ((x >> GOB_SIZE_X_SHIFT) << x_shift) + (swizzled_x | swizzled_y);
+        u8* const dst = &output[TO_LINEAR ? swizzled : linear];
+        const u8* const src = &input[TO_LINEAR ? linear : swizzled];
+        std::memcpy(dst, src, count);
+        x += count;
+        linear += count;
+    };
+
+    u32 swizzled_run = SWIZZLE_X_RUN_TABLE[(x >> SWIZZLE_RUN_SHIFT) & SWIZZLE_RUN_INDEX_MASK];
+    u32 remaining = num_bytes;
+
+    const u32 head =
+        (std::min)(SWIZZLE_RUN_BYTES - (x & SWIZZLE_RUN_MASK), remaining) & SWIZZLE_RUN_MASK;
+    if (head != 0) {
+        copy(swizzled_run | (x & SWIZZLE_RUN_MASK), head);
+        remaining -= head;
+        incrpdep<SWIZZLE_X_BITS, SWIZZLE_RUN_BYTES>(swizzled_run);
+    }
+
+    while (remaining >= SWIZZLE_RUN_BYTES) {
+        copy(swizzled_run, SWIZZLE_RUN_BYTES);
+        remaining -= SWIZZLE_RUN_BYTES;
+        incrpdep<SWIZZLE_X_BITS, SWIZZLE_RUN_BYTES>(swizzled_run);
+    }
+
+    if (remaining != 0) {
+        copy(swizzled_run, remaining);
+    }
+}
+
+template <bool TO_LINEAR, u32 BYTES_PER_PIXEL>
+void SwizzleRowPerPixel(std::span<u8> output, std::span<const u8> input, u32 offset_zy,
+                        u32 swizzled_y, u32 x_shift, u32 x, u32 num_bytes, u32 linear) {
+    u32 swizzled_x = pdep<SWIZZLE_X_BITS>(x);
+    for (u32 byte = 0; byte < num_bytes;
+         byte += BYTES_PER_PIXEL, incrpdep<SWIZZLE_X_BITS, BYTES_PER_PIXEL>(swizzled_x)) {
+        const u32 offset_x = ((x + byte) >> GOB_SIZE_X_SHIFT) << x_shift;
+        const u32 swizzled = offset_zy + offset_x + (swizzled_x | swizzled_y);
+        const u32 unswizzled = linear + byte;
+
+        u8* const dst = &output[TO_LINEAR ? swizzled : unswizzled];
+        const u8* const src = &input[TO_LINEAR ? unswizzled : swizzled];
+
+        std::memcpy(dst, src, BYTES_PER_PIXEL);
+    }
+}
+
+template <bool TO_LINEAR, u32 BYTES_PER_PIXEL>
+void SwizzleRowDispatch(std::span<u8> output, std::span<const u8> input, u32 offset_zy,
+                        u32 swizzled_y, u32 x_shift, u32 x, u32 num_bytes, u32 linear) {
+    if constexpr (SWIZZLE_RUN_BYTES % BYTES_PER_PIXEL == 0) {
+        SwizzleRow<TO_LINEAR>(output, input, offset_zy, swizzled_y, x_shift, x, num_bytes, linear);
+    } else {
+        SwizzleRowPerPixel<TO_LINEAR, BYTES_PER_PIXEL>(output, input, offset_zy, swizzled_y, x_shift,
+                                                       x, num_bytes, linear);
+    }
+}
+
 template <bool TO_LINEAR, u32 BYTES_PER_PIXEL>
 void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 height, u32 depth,
                  u32 block_height, u32 block_depth, u32 stride) {
@@ -70,23 +149,10 @@ void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32
             const u32 offset_y = (block_y >> block_height) * block_size +
                                  ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
 
-            u32 swizzled_x = pdep<SWIZZLE_X_BITS>(origin_x * BYTES_PER_PIXEL);
-            for (u32 column = 0; column < width;
-                 ++column, incrpdep<SWIZZLE_X_BITS, BYTES_PER_PIXEL>(swizzled_x)) {
-                const u32 x = (column + origin_x) * BYTES_PER_PIXEL;
-                const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift;
-
-                const u32 base_swizzled_offset = offset_z + offset_y + offset_x;
-                const u32 swizzled_offset = base_swizzled_offset + (swizzled_x | swizzled_y);
-
-                const u32 unswizzled_offset =
-                    slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL;
-
-                u8* const dst = &output[TO_LINEAR ? swizzled_offset : unswizzled_offset];
-                const u8* const src = &input[TO_LINEAR ? unswizzled_offset : swizzled_offset];
-
-                std::memcpy(dst, src, BYTES_PER_PIXEL);
-            }
+            SwizzleRowDispatch<TO_LINEAR, BYTES_PER_PIXEL>(
+                output, input, offset_z + offset_y, swizzled_y, x_shift,
+                origin_x * BYTES_PER_PIXEL, width * BYTES_PER_PIXEL,
+                slice * pitch * height + line * pitch);
         }
     }
 }
@@ -129,23 +195,10 @@ void SwizzleSubrectImpl(std::span<u8> output, std::span<const u8> input, u32 wid
             const u32 offset_y = (block_y >> block_height) * block_size +
                                  ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
 
-            u32 swizzled_x = pdep<SWIZZLE_X_BITS>(origin_x * BYTES_PER_PIXEL);
-            for (u32 column = 0; column < extent_x;
-                 ++column, incrpdep<SWIZZLE_X_BITS, BYTES_PER_PIXEL>(swizzled_x)) {
-                const u32 x = (column + origin_x) * BYTES_PER_PIXEL;
-                const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift;
-
-                const u32 base_swizzled_offset = offset_z + offset_y + offset_x;
-                const u32 swizzled_offset = base_swizzled_offset + (swizzled_x | swizzled_y);
-
-                const u32 unswizzled_offset =
-                    slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL;
-
-                u8* const dst = &output[TO_LINEAR ? swizzled_offset : unswizzled_offset];
-                const u8* const src = &input[TO_LINEAR ? unswizzled_offset : swizzled_offset];
-
-                std::memcpy(dst, src, BYTES_PER_PIXEL);
-            }
+            SwizzleRowDispatch<TO_LINEAR, BYTES_PER_PIXEL>(
+                output, input, offset_z + offset_y, swizzled_y, x_shift,
+                origin_x * BYTES_PER_PIXEL, extent_x * BYTES_PER_PIXEL,
+                slice * pitch * height + line * pitch);
         }
         unprocessed_lines -= lines_in_y;
         if (unprocessed_lines == 0) {

src/video_core/textures/texture.cpp

@@ -1,10 +1,11 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <array>
+#include <cmath>
 
 #include "common/cityhash.h"
 #include "common/settings.h"
@@ -17,53 +18,25 @@ namespace Tegra::Texture {
 
 namespace {
 
-[[maybe_unused]] constexpr std::array<float, 256> SRGB_CONVERSION_LUT = {
-    0.000000f, 0.000000f, 0.000000f, 0.000012f, 0.000021f, 0.000033f, 0.000046f, 0.000062f,
-    0.000081f, 0.000102f, 0.000125f, 0.000151f, 0.000181f, 0.000214f, 0.000251f, 0.000293f,
-    0.000338f, 0.000388f, 0.000443f, 0.000503f, 0.000568f, 0.000639f, 0.000715f, 0.000798f,
-    0.000887f, 0.000983f, 0.001085f, 0.001195f, 0.001312f, 0.001437f, 0.001569f, 0.001710f,
-    0.001860f, 0.002019f, 0.002186f, 0.002364f, 0.002551f, 0.002748f, 0.002955f, 0.003174f,
-    0.003403f, 0.003643f, 0.003896f, 0.004160f, 0.004436f, 0.004725f, 0.005028f, 0.005343f,
-    0.005672f, 0.006015f, 0.006372f, 0.006744f, 0.007130f, 0.007533f, 0.007950f, 0.008384f,
-    0.008834f, 0.009301f, 0.009785f, 0.010286f, 0.010805f, 0.011342f, 0.011898f, 0.012472f,
-    0.013066f, 0.013680f, 0.014313f, 0.014967f, 0.015641f, 0.016337f, 0.017054f, 0.017793f,
-    0.018554f, 0.019337f, 0.020144f, 0.020974f, 0.021828f, 0.022706f, 0.023609f, 0.024536f,
-    0.025489f, 0.026468f, 0.027473f, 0.028504f, 0.029563f, 0.030649f, 0.031762f, 0.032904f,
-    0.034074f, 0.035274f, 0.036503f, 0.037762f, 0.039050f, 0.040370f, 0.041721f, 0.043103f,
-    0.044518f, 0.045964f, 0.047444f, 0.048956f, 0.050503f, 0.052083f, 0.053699f, 0.055349f,
-    0.057034f, 0.058755f, 0.060513f, 0.062307f, 0.064139f, 0.066008f, 0.067915f, 0.069861f,
-    0.071845f, 0.073869f, 0.075933f, 0.078037f, 0.080182f, 0.082369f, 0.084597f, 0.086867f,
-    0.089180f, 0.091535f, 0.093935f, 0.096378f, 0.098866f, 0.101398f, 0.103977f, 0.106601f,
-    0.109271f, 0.111988f, 0.114753f, 0.117565f, 0.120426f, 0.123335f, 0.126293f, 0.129301f,
-    0.132360f, 0.135469f, 0.138629f, 0.141841f, 0.145105f, 0.148421f, 0.151791f, 0.155214f,
-    0.158691f, 0.162224f, 0.165810f, 0.169453f, 0.173152f, 0.176907f, 0.180720f, 0.184589f,
-    0.188517f, 0.192504f, 0.196549f, 0.200655f, 0.204820f, 0.209046f, 0.213334f, 0.217682f,
-    0.222093f, 0.226567f, 0.231104f, 0.235704f, 0.240369f, 0.245099f, 0.249894f, 0.254754f,
-    0.259681f, 0.264674f, 0.269736f, 0.274864f, 0.280062f, 0.285328f, 0.290664f, 0.296070f,
-    0.301546f, 0.307094f, 0.312713f, 0.318404f, 0.324168f, 0.330006f, 0.335916f, 0.341902f,
-    0.347962f, 0.354097f, 0.360309f, 0.366597f, 0.372961f, 0.379403f, 0.385924f, 0.392524f,
-    0.399202f, 0.405960f, 0.412798f, 0.419718f, 0.426719f, 0.433802f, 0.440967f, 0.448216f,
-    0.455548f, 0.462965f, 0.470465f, 0.478052f, 0.485725f, 0.493484f, 0.501329f, 0.509263f,
-    0.517285f, 0.525396f, 0.533595f, 0.541885f, 0.550265f, 0.558736f, 0.567299f, 0.575954f,
-    0.584702f, 0.593542f, 0.602477f, 0.611507f, 0.620632f, 0.629852f, 0.639168f, 0.648581f,
-    0.658092f, 0.667700f, 0.677408f, 0.687214f, 0.697120f, 0.707127f, 0.717234f, 0.727443f,
-    0.737753f, 0.748167f, 0.758685f, 0.769305f, 0.780031f, 0.790861f, 0.801798f, 0.812839f,
-    0.823989f, 0.835246f, 0.846611f, 0.858085f, 0.869668f, 0.881360f, 0.893164f, 0.905078f,
-    0.917104f, 0.929242f, 0.941493f, 0.953859f, 0.966338f, 1.000000f, 1.000000f, 1.000000f,
-};
+float SrgbToLinear(u32 value) {
+    const float encoded = static_cast<float>(value) / 255.0f;
+    if (encoded <= 0.04045f) {
+        return encoded / 12.92f;
+    }
+    return std::pow((encoded + 0.055f) / 1.055f, 2.4f);
+}
 
 } // Anonymous namespace
 
 std::array<float, 4> TSCEntry::BorderColor() const noexcept {
-    // TODO: Handle SRGB correctly. Using this breaks shadows in some games (Xenoblade).
-    // if (!srgb_conversion) {
-    //    return border_color;
-    //}
-    // return {SRGB_CONVERSION_LUT[srgb_border_color_r], SRGB_CONVERSION_LUT[srgb_border_color_g],
-    //        SRGB_CONVERSION_LUT[srgb_border_color_b], border_color[3]};
     return border_color;
 }
 
+std::array<float, 4> TSCEntry::SrgbBorderColor() const noexcept {
+    return {SrgbToLinear(srgb_border_color_r), SrgbToLinear(srgb_border_color_g),
+            SrgbToLinear(srgb_border_color_b), border_color[3]};
+}
+
 float TSCEntry::MaxAnisotropy() const noexcept {
     const bool is_suitable_mipmap_filter = mipmap_filter != TextureMipmapFilter::None;
     const bool has_regular_lods = min_lod_clamp == 0 && max_lod_clamp >= 256;
@@ -80,16 +53,16 @@ float TSCEntry::MaxAnisotropy() const noexcept {
     case Settings::AnisotropyMode::X4:
     case Settings::AnisotropyMode::X8:
     case Settings::AnisotropyMode::X16:
-    case Settings::AnisotropyMode::X32:
-    case Settings::AnisotropyMode::X64:
-        added_anisotropic = u32(anisotropic_settings) - 1U;
+        added_anisotropic = s32(anisotropic_settings) - 1;
         break;
-    case Settings::AnisotropyMode::Automatic:
-        added_anisotropic = Settings::values.resolution_info.up_scale >> Settings::values.resolution_info.down_shift;
-        added_anisotropic = (std::max)(added_anisotropic - 1U, 0U);
+    case Settings::AnisotropyMode::Automatic: {
+        const u32 resolution_scale = Settings::values.resolution_info.up_scale >>
+                                     Settings::values.resolution_info.down_shift;
+        if (resolution_scale > 1U) {
+            added_anisotropic = static_cast<s32>(resolution_scale - 1U);
+        }
         break;
-    case Settings::AnisotropyMode::None:
-        return 1.0f; //No use of anisotropy
+    }
     }
     return float(1U << (max_anisotropy + added_anisotropic));
 }

src/video_core/textures/texture.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -378,6 +381,8 @@ struct TSCEntry {
 
     std::array<float, 4> BorderColor() const noexcept;
 
+    std::array<float, 4> SrgbBorderColor() const noexcept;
+
     float MaxAnisotropy() const noexcept;
 
     float MinLod() const {

src/video_core/vulkan_common/vulkan_device.cpp

@@ -7,6 +7,8 @@
 #include <algorithm>
 #include <bitset>
 #include <chrono>
+#include <filesystem>
+#include <fstream>
 #include <optional>
 #include <thread>
 #include <ankerl/unordered_dense.h>
@@ -16,6 +18,8 @@
 #include <fmt/format.h>
 
 #include "common/assert.h"
+#include "common/fs/fs.h"
+#include "common/fs/path_util.h"
 #include "common/literals.h"
 #include <ranges>
 #include "common/settings.h"
@@ -393,6 +397,17 @@ std::vector<const char*> ExtensionListForVulkan(
     return output;
 }
 
+constexpr std::array<char, 8> STATIC_CACHE_MAGIC_NUMBER{'e', 'd', 'e', 'n', 's', 't', 'p', 'c'};
+constexpr u32 STATIC_CACHE_VERSION = 1;
+
+std::filesystem::path StaticPipelineCacheFilename() {
+    const auto shader_dir = Common::FS::GetEdenPath(Common::FS::EdenPath::ShaderDir);
+    if (!Common::FS::CreateDir(shader_dir)) {
+        return {};
+    }
+    return shader_dir / "vulkan_static_pipelines.bin";
+}
+
 } // Anonymous namespace
 
 void Device::RemoveExtension(bool& extension, const std::string& extension_name) {
@@ -508,12 +523,6 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
         LOG_WARNING(Render_Vulkan, "Qualcomm drivers require scaled vertex format emulation.");
         has_broken_descriptor_aliasing = true;
         LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken descriptor aliasing.");
-        LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken custom border color.");
-        RemoveExtensionFeature(extensions.custom_border_color, features.custom_border_color,
-                               VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME);
-        LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken border color swizzle.");
-        RemoveExtensionFeature(extensions.border_color_swizzle, features.border_color_swizzle,
-                               VK_EXT_BORDER_COLOR_SWIZZLE_EXTENSION_NAME);
         LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken color write enable.");
         RemoveExtensionFeature(extensions.color_write_enable, features.color_write_enable,
                                VK_EXT_COLOR_WRITE_ENABLE_EXTENSION_NAME);
@@ -619,21 +628,6 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
         }
     }
 
-    if (is_qualcomm) {
-        const size_t sampler_limit = properties.properties.limits.maxSamplerAllocationCount;
-        if (sampler_limit > 0) {
-            constexpr size_t MIN_SAMPLER_BUDGET = 1024U;
-            const size_t reserved = sampler_limit / 4U;
-            const size_t derived_budget =
-                (std::max)(MIN_SAMPLER_BUDGET, sampler_limit - reserved);
-            sampler_heap_budget = derived_budget;
-            LOG_WARNING(Render_Vulkan,
-                        "Qualcomm driver reports max {} samplers; reserving {} (25%) and "
-                        "allowing Eden to use {} (75%) to avoid heap exhaustion",
-                        sampler_limit, reserved, sampler_heap_budget);
-        }
-    }
-
     if (extensions.sampler_filter_minmax && is_amd) {
         // Disable ext_sampler_filter_minmax on AMD GCN4 and lower as it is broken.
         if (!features.shader_float16_int8.shaderFloat16) {
@@ -650,13 +644,6 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
         features.shader_float16_int8.shaderFloat16 = false;
     }
 
-    if (is_intel_windows) {
-        LOG_WARNING(Render_Vulkan,
-                    "Intel proprietary drivers do not support MSAA->MSAA image blits. "
-                    "MSAA scaling will use 3D helpers. MSAA resolves work normally.");
-        cant_blit_msaa = true;
-    }
-
     has_broken_compute =
         CheckBrokenCompute(properties.driver.driverID, properties.properties.driverVersion) &&
         !Settings::values.enable_compute_pipelines.GetValue();
@@ -780,15 +767,100 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
 
     vk::Check(vmaCreateAllocator(&allocator_info, &allocator));
 
+    owns_static_pipeline_cache = surface != VkSurfaceKHR{};
+    LoadStaticPipelineCache();
+
     // Initialize GPU logging if enabled
     InitializeGPULogging();
 }
 
 Device::~Device() {
+    SaveStaticPipelineCache();
     ShutdownGPULogging();
     vmaDestroyAllocator(allocator);
 }
 
+void Device::LoadStaticPipelineCache() {
+    const auto create = [this](size_t size, const void* data) {
+        static_pipeline_cache = logical.CreatePipelineCache({
+            .sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO,
+            .pNext = nullptr,
+            .flags = 0,
+            .initialDataSize = size,
+            .pInitialData = data,
+        });
+    };
+    if (!owns_static_pipeline_cache) {
+        create(0, nullptr);
+        return;
+    }
+    const auto filename = StaticPipelineCacheFilename();
+    if (filename.empty()) {
+        create(0, nullptr);
+        return;
+    }
+    std::vector<char> data;
+    try {
+        std::ifstream file(filename, std::ios::binary | std::ios::ate);
+        if (!file.is_open()) {
+            create(0, nullptr);
+            return;
+        }
+        file.exceptions(std::ifstream::failbit | std::ifstream::badbit);
+        const size_t total = static_cast<size_t>(file.tellg());
+        file.seekg(0, std::ios::beg);
+        std::array<char, 8> magic{};
+        u32 version{};
+        if (total < magic.size() + sizeof(version)) {
+            create(0, nullptr);
+            return;
+        }
+        file.read(magic.data(), magic.size())
+            .read(reinterpret_cast<char*>(&version), sizeof(version));
+        if (magic != STATIC_CACHE_MAGIC_NUMBER || version != STATIC_CACHE_VERSION) {
+            create(0, nullptr);
+            return;
+        }
+        data.resize(total - magic.size() - sizeof(version));
+        file.read(data.data(), static_cast<std::streamsize>(data.size()));
+    } catch (const std::ios_base::failure& e) {
+        create(0, nullptr);
+        return;
+    }
+    create(data.size(), data.empty() ? nullptr : data.data());
+}
+
+void Device::SaveStaticPipelineCache() const {
+    if (!owns_static_pipeline_cache || !static_pipeline_cache) {
+        return;
+    }
+    const auto filename = StaticPipelineCacheFilename();
+    if (filename.empty()) {
+        return;
+    }
+    size_t size = 0;
+    std::vector<char> data;
+    static_pipeline_cache.Read(&size, nullptr);
+    if (size == 0) {
+        return;
+    }
+    data.resize(size);
+    static_pipeline_cache.Read(&size, data.data());
+    try {
+        std::ofstream file(filename, std::ios::binary | std::ios::trunc);
+        file.exceptions(std::ofstream::failbit);
+        if (!file.is_open()) {
+            return;
+        }
+        file.write(STATIC_CACHE_MAGIC_NUMBER.data(), STATIC_CACHE_MAGIC_NUMBER.size())
+            .write(reinterpret_cast<const char*>(&STATIC_CACHE_VERSION),
+                   sizeof(STATIC_CACHE_VERSION))
+            .write(data.data(), static_cast<std::streamsize>(size));
+    } catch (const std::ios_base::failure& e) {
+        Common::FS::RemoveFile(filename);
+    }
+}
+
 VkFormat Device::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage,
                                     FormatType format_type) const {
     if (IsFormatSupported(wanted_format, wanted_usage, format_type)) {
@@ -975,6 +1047,12 @@ bool Device::GetSuitability(bool requires_swapchain) {
     FOR_EACH_VK_FEATURE_EXT(FEATURE_EXTENSION);
     FOR_EACH_VK_EXTENSION(EXTENSION);
 
+    extensions.depth_stencil_resolve =
+        extensions.depth_stencil_resolve &&
+        (instance_version >= VK_API_VERSION_1_2 || extensions.create_renderpass2);
+    RemoveExtensionIfUnsuitable(extensions.depth_stencil_resolve,
+                                VK_KHR_DEPTH_STENCIL_RESOLVE_EXTENSION_NAME);
+
     if (supported_extensions.contains(VK_KHR_ROBUSTNESS_2_EXTENSION_NAME)) {
         loaded_extensions.erase(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME);
         loaded_extensions.insert(VK_KHR_ROBUSTNESS_2_EXTENSION_NAME);
@@ -1122,6 +1200,11 @@ bool Device::GetSuitability(bool requires_swapchain) {
             VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PUSH_DESCRIPTOR_PROPERTIES_KHR;
         SetNext(next, properties.push_descriptor);
     }
+    if (extensions.depth_stencil_resolve || instance_version >= VK_API_VERSION_1_2) {
+        properties.depth_stencil_resolve.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DEPTH_STENCIL_RESOLVE_PROPERTIES;
+        SetNext(next, properties.depth_stencil_resolve);
+    }
     if (extensions.descriptor_buffer) {
         properties.descriptor_buffer.sType =
             VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DESCRIPTOR_BUFFER_PROPERTIES_EXT;
@@ -1142,6 +1225,16 @@ bool Device::GetSuitability(bool requires_swapchain) {
             VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MAINTENANCE_5_PROPERTIES_KHR;
         SetNext(next, properties.maintenance5);
     }
+    if (extensions.custom_border_color) {
+        properties.custom_border_color.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CUSTOM_BORDER_COLOR_PROPERTIES_EXT;
+        SetNext(next, properties.custom_border_color);
+    }
+    if (extensions.vertex_attribute_divisor) {
+        properties.vertex_attribute_divisor.sType =
+            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_ATTRIBUTE_DIVISOR_PROPERTIES_EXT;
+        SetNext(next, properties.vertex_attribute_divisor);
+    }
 
     // Perform the property fetch.
     physical.GetProperties2(properties2);
@@ -1238,9 +1331,7 @@ void Device::RemoveUnsuitableExtensions() {
     // VK_EXT_border_color_swizzle
     if (extensions.border_color_swizzle) {
         extensions.border_color_swizzle =
-            extensions.custom_border_color &&
-            features.border_color_swizzle.borderColorSwizzle &&
-            features.border_color_swizzle.borderColorSwizzleFromImage;
+            extensions.custom_border_color && features.border_color_swizzle.borderColorSwizzle;
     }
     RemoveExtensionFeatureIfUnsuitable(extensions.border_color_swizzle,
                                        features.border_color_swizzle,
@@ -1486,11 +1577,67 @@ void Device::SetupFamilies(VkSurfaceKHR surface) {
     }
 }
 
-std::optional<size_t> Device::GetSamplerHeapBudget() const {
-    if (sampler_heap_budget == 0) {
-        return std::nullopt;
+VkSampleCountFlags Device::GetSupportedSampleCounts(VkImageUsageFlags usage,
+                                                    VkImageAspectFlags aspect,
+                                                    bool is_integer) const {
+    const VkPhysicalDeviceLimits& limits = properties.properties.limits;
+    const bool has_color = (aspect & VK_IMAGE_ASPECT_COLOR_BIT) != 0;
+    const bool has_depth = (aspect & VK_IMAGE_ASPECT_DEPTH_BIT) != 0;
+    const bool has_stencil = (aspect & VK_IMAGE_ASPECT_STENCIL_BIT) != 0;
+
+    VkSampleCountFlags counts = ~VkSampleCountFlags{0};
+    if ((usage & VK_IMAGE_USAGE_SAMPLED_BIT) != 0) {
+        if (has_color) {
+            if (is_integer) {
+                counts &= limits.sampledImageIntegerSampleCounts;
+            } else {
+                counts &= limits.sampledImageColorSampleCounts;
+            }
+        }
+        if (has_depth) {
+            counts &= limits.sampledImageDepthSampleCounts;
+        }
+        if (has_stencil) {
+            counts &= limits.sampledImageStencilSampleCounts;
+        }
     }
-    return sampler_heap_budget;
+    if ((usage & VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT) != 0) {
+        counts &= limits.framebufferColorSampleCounts;
+    }
+    if ((usage & VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) != 0) {
+        if (has_depth) {
+            counts &= limits.framebufferDepthSampleCounts;
+        }
+        if (has_stencil) {
+            counts &= limits.framebufferStencilSampleCounts;
+        }
+    }
+    if ((usage & VK_IMAGE_USAGE_STORAGE_BIT) != 0) {
+        counts &= limits.storageImageSampleCounts;
+    }
+    return counts;
+}
+
+bool Device::TryReserveCustomBorderColorSamplers(size_t count) const {
+    const size_t limit = properties.custom_border_color.maxCustomBorderColorSamplers;
+    if (limit == 0) {
+        return true;
+    }
+    size_t used = custom_border_color_samplers_used.load(std::memory_order_relaxed);
+    while (used + count <= limit) {
+        if (custom_border_color_samplers_used.compare_exchange_weak(
+                used, used + count, std::memory_order_relaxed, std::memory_order_relaxed)) {
+            return true;
+        }
+    }
+    return false;
+}
+
+void Device::ReleaseCustomBorderColorSamplers(size_t count) const {
+    if (count == 0) {
+        return;
+    }
+    custom_border_color_samplers_used.fetch_sub(count, std::memory_order_relaxed);
 }
 
 u64 Device::GetDeviceMemoryUsage() const {

src/video_core/vulkan_common/vulkan_device.h

@@ -6,6 +6,7 @@
 
 #pragma once
 
+#include <atomic>
 #include <optional>
 #include <set>
 #include <span>
@@ -91,6 +92,8 @@ VK_DEFINE_HANDLE(VmaAllocator)
     EXTENSION(EXT, SHADER_VIEWPORT_INDEX_LAYER, shader_viewport_index_layer)                       \
     EXTENSION(EXT, TOOLING_INFO, tooling_info)                                                     \
     EXTENSION(EXT, VERTEX_ATTRIBUTE_DIVISOR, vertex_attribute_divisor)                             \
+    EXTENSION(KHR, CREATE_RENDERPASS_2, create_renderpass2)                                        \
+    EXTENSION(KHR, DEPTH_STENCIL_RESOLVE, depth_stencil_resolve)                                   \
     EXTENSION(KHR, DRAW_INDIRECT_COUNT, draw_indirect_count)                                       \
     EXTENSION(KHR, DRIVER_PROPERTIES, driver_properties)                                           \
     EXTENSION(KHR, PUSH_DESCRIPTOR, push_descriptor)                                               \
@@ -269,6 +272,10 @@ public:
         return physical;
     }
 
+    VkPipelineCache StaticPipelineCache() const noexcept {
+        return *static_pipeline_cache;
+    }
+
     /// Returns the main graphics queue.
     vk::Queue GetGraphicsQueue() const {
         return graphics_queue;
@@ -452,6 +459,10 @@ FN_MAX_LIMIT_LIST
         return features.features.shaderStorageImageMultisample;
     }
 
+    /// Returns the sample counts an image with the given usage and aspect can be created with.
+    VkSampleCountFlags GetSupportedSampleCounts(VkImageUsageFlags usage, VkImageAspectFlags aspect,
+                                                bool is_integer) const;
+
     /// Returns true if the device warp size can potentially be bigger than guest's warp size.
     bool IsWarpSizePotentiallyBiggerThanGuest() const {
         return is_warp_potentially_bigger;
@@ -613,8 +624,38 @@ FN_MAX_LIMIT_LIST
         return extensions.shader_stencil_export;
     }
 
+    /// Returns true if the device supports VK_KHR_create_renderpass2.
+    bool IsKhrCreateRenderPass2Supported() const {
+        return extensions.create_renderpass2 || instance_version >= VK_API_VERSION_1_2;
+    }
+
+    /// Returns true if the device supports VK_KHR_depth_stencil_resolve.
+    bool IsKhrDepthStencilResolveSupported() const {
+        return (extensions.depth_stencil_resolve || instance_version >= VK_API_VERSION_1_2) &&
+               IsKhrCreateRenderPass2Supported();
+    }
+
+    /// Returns the supported resolve modes for the depth aspect.
+    VkResolveModeFlags GetDepthResolveModes() const {
+        return properties.depth_stencil_resolve.supportedDepthResolveModes;
+    }
+
+    /// Returns the supported resolve modes for the stencil aspect.
+    VkResolveModeFlags GetStencilResolveModes() const {
+        return properties.depth_stencil_resolve.supportedStencilResolveModes;
+    }
+
+    /// Returns true if the depth and stencil aspects may resolve with different modes.
+    bool SupportsIndependentResolve() const {
+        return properties.depth_stencil_resolve.independentResolve == VK_TRUE;
+    }
+
+    /// Returns true if only one of the depth and stencil aspects may be resolved.
+    bool SupportsIndependentResolveNone() const {
+        return properties.depth_stencil_resolve.independentResolveNone == VK_TRUE;
+    }
+
     /// Returns true if depth/stencil operations can be performed efficiently.
-    /// Either through shader export or hardware blits.
     bool CanPerformDepthStencilOperations() const {
         return extensions.shader_stencil_export || is_blit_depth24_stencil8_supported ||
                is_blit_depth32_stencil8_supported;
@@ -671,20 +712,23 @@ FN_MAX_LIMIT_LIST
         return features.transform_feedback.geometryStreams;
     }
 
-    /// Returns true if the device supports VK_EXT_custom_border_color.
-    bool IsExtCustomBorderColorSupported() const {
-        return extensions.custom_border_color;
+    /// Returns true if custom border colors can be created without a format.
+    bool IsCustomBorderColorUsable() const {
+        return extensions.custom_border_color &&
+               features.custom_border_color.customBorderColors &&
+               features.custom_border_color.customBorderColorWithoutFormat;
     }
 
-    /// Returns true if customBorderColors feature is available.
-    bool IsCustomBorderColorsSupported() const {
-        return features.custom_border_color.customBorderColors;
+    /// Returns how many live samplers may carry a custom border color, 0 when unknown.
+    u32 GetMaxCustomBorderColorSamplers() const {
+        return properties.custom_border_color.maxCustomBorderColorSamplers;
     }
 
-    /// Returns true if customBorderColorWithoutFormat feature is available.
-    bool IsCustomBorderColorWithoutFormatSupported() const {
-        return features.custom_border_color.customBorderColorWithoutFormat;
-    }
+    /// Takes budget for samplers carrying a custom border color, false when exhausted.
+    bool TryReserveCustomBorderColorSamplers(size_t count) const;
+
+    /// Gives back budget taken by TryReserveCustomBorderColorSamplers.
+    void ReleaseCustomBorderColorSamplers(size_t count) const;
 
     /// Returns true if the device supports VK_EXT_color_write_enable.
     bool IsExtColorWriteEnableSupported() const {
@@ -696,6 +740,12 @@ FN_MAX_LIMIT_LIST
         return extensions.border_color_swizzle;
     }
 
+    /// Returns true if samplers must be carried with border color swizzle mapping.
+    bool NeedsBorderColorSwizzleMapping() const {
+        return extensions.border_color_swizzle &&
+               !features.border_color_swizzle.borderColorSwizzleFromImage;
+    }
+
     /// Returns true if borderColorSwizzleFromImage is available.
     bool IsBorderColorSwizzleFromImageSupported() const {
         return features.border_color_swizzle.borderColorSwizzleFromImage;
@@ -891,8 +941,6 @@ FN_MAX_LIMIT_LIST
         return has_broken_parallel_compiling;
     }
 
-    std::optional<size_t> GetSamplerHeapBudget() const;
-
     /// Returns the vendor name reported from Vulkan.
     std::string_view GetVendorName() const {
         return properties.driver.driverName;
@@ -921,10 +969,6 @@ FN_MAX_LIMIT_LIST
         return supports_d24_depth;
     }
 
-    bool CantBlitMSAA() const {
-        return cant_blit_msaa;
-    }
-
     bool MustEmulateScaledFormats() const {
         return must_emulate_scaled_formats;
     }
@@ -947,6 +991,22 @@ FN_MAX_LIMIT_LIST
         return properties.properties.limits.maxVertexInputBindings;
     }
 
+    u32 GetMaxVertexInputAttributeOffset() const {
+        return properties.properties.limits.maxVertexInputAttributeOffset;
+    }
+
+    u32 GetMaxVertexInputBindingStride() const {
+        return properties.properties.limits.maxVertexInputBindingStride;
+    }
+
+    u32 GetMaxVertexAttribDivisor() const {
+        const u32 limit = properties.vertex_attribute_divisor.maxVertexAttribDivisor;
+        if (!extensions.vertex_attribute_divisor || limit == 0) {
+            return 1U;
+        }
+        return limit;
+    }
+
     u32 GetMaxViewports() const {
         return properties.properties.limits.maxViewports;
     }
@@ -963,6 +1023,25 @@ FN_MAX_LIMIT_LIST
         return features2.features.multiViewport;
     }
 
+    bool SupportsGeometryShader() const {
+        return features2.features.geometryShader;
+    }
+
+    VkPipelineStageFlags AttachmentConsumerStages() const {
+        VkPipelineStageFlags stages = VK_PIPELINE_STAGE_VERTEX_SHADER_BIT |
+                                      VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT |
+                                      VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT |
+                                      VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+                                      VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
+                                      VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
+                                      VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
+                                      VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
+        if (features2.features.geometryShader) {
+            stages |= VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT;
+        }
+        return stages;
+    }
+
     /// Returns true if the device supports VK_KHR_maintenance1.
     bool IsKhrMaintenance1Supported() const {
         return extensions.maintenance1;
@@ -1094,6 +1173,9 @@ private:
     /// Returns true if the device natively supports blitting depth stencil images.
     bool TestDepthStencilBlits(VkFormat format) const;
 
+    void LoadStaticPipelineCache();
+    void SaveStaticPipelineCache() const;
+
 private:
     VkInstance instance;         ///< Vulkan instance.
     VmaAllocator allocator;      ///< VMA allocator.
@@ -1102,6 +1184,8 @@ private:
     vk::Device logical;          ///< Logical device.
     vk::Queue graphics_queue;    ///< Main graphics queue.
     vk::Queue present_queue;     ///< Main present queue.
+    vk::PipelineCache static_pipeline_cache;
+    bool owns_static_pipeline_cache{};
     u32 instance_version{};      ///< Vulkan instance version.
     u32 graphics_family{};       ///< Main graphics queue family index.
     u32 present_family{};        ///< Main present queue family index.
@@ -1148,6 +1232,9 @@ private:
         VkPhysicalDeviceSubgroupSizeControlProperties subgroup_size_control{};
         VkPhysicalDeviceTransformFeedbackPropertiesEXT transform_feedback{};
         VkPhysicalDeviceMaintenance5PropertiesKHR maintenance5{};
+        VkPhysicalDeviceDepthStencilResolveProperties depth_stencil_resolve{};
+        VkPhysicalDeviceCustomBorderColorPropertiesEXT custom_border_color{};
+        VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT vertex_attribute_divisor{};
 
         VkPhysicalDeviceProperties properties{};
     };
@@ -1175,7 +1262,6 @@ private:
     bool has_nsight_graphics{};                ///< Has Nsight Graphics attached
     bool has_radeon_gpu_profiler{};            ///< Has Radeon GPU Profiler attached.
     bool supports_d24_depth{};                 ///< Supports D24 depth buffers.
-    bool cant_blit_msaa{};                     ///< Does not support MSAA<->MSAA blitting.
     bool must_emulate_scaled_formats{};        ///< Requires scaled vertex format emulation
     bool dynamic_state3_blending{};            ///< Has blending features of dynamic_state3.
     bool dynamic_state3_enables{};             ///< Has at least one enable feature of dynamic_state3.
@@ -1187,7 +1273,7 @@ private:
     bool dynamic_state3_alpha_to_coverage{};
     bool dynamic_state3_alpha_to_one{};
     bool supports_conditional_barriers{};      ///< Allows barriers in conditional control flow.
-    size_t sampler_heap_budget{};              ///< Sampler budget for buggy drivers (0 = unlimited).
+    mutable std::atomic<size_t> custom_border_color_samplers_used{};
     u64 device_access_memory{};                ///< Total size of device local memory in bytes.
     u32 sets_per_pool{};                       ///< Sets per Description Pool
     NvidiaArchitecture nvidia_arch{NvidiaArchitecture::Arch_AmpereOrNewer};

src/video_core/vulkan_common/vulkan_wrapper.cpp

@@ -184,6 +184,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkCreatePipelineLayout);
     X(vkCreateQueryPool);
     X(vkCreateRenderPass);
+    X(vkCreateRenderPass2);
     X(vkCreateSampler);
     X(vkCreateSemaphore);
     X(vkCreateShaderModule);
@@ -270,6 +271,11 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     if (!dld.vkQueueSubmit2) {
         Proc(dld.vkQueueSubmit2, dld, "vkQueueSubmit2KHR", device);
     }
+
+    // Render pass creation v2 is core in Vulkan 1.2, otherwise requires VK_KHR_create_renderpass2
+    if (!dld.vkCreateRenderPass2) {
+        Proc(dld.vkCreateRenderPass2, dld, "vkCreateRenderPass2KHR", device);
+    }
 #undef X
 }
 
@@ -725,6 +731,12 @@ RenderPass Device::CreateRenderPass(const VkRenderPassCreateInfo& ci) const {
     return RenderPass(object, handle, *dld);
 }
 
+RenderPass Device::CreateRenderPass2(const VkRenderPassCreateInfo2& ci) const {
+    VkRenderPass object;
+    Check(dld->vkCreateRenderPass2(handle, &ci, nullptr, &object));
+    return RenderPass(object, handle, *dld);
+}
+
 DescriptorSetLayout Device::CreateDescriptorSetLayout(
     const VkDescriptorSetLayoutCreateInfo& ci) const {
     VkDescriptorSetLayout object;

src/video_core/vulkan_common/vulkan_wrapper.h

@@ -300,6 +300,7 @@ struct DeviceDispatch : InstanceDispatch {
     PFN_vkCreatePipelineLayout vkCreatePipelineLayout{};
     PFN_vkCreateQueryPool vkCreateQueryPool{};
     PFN_vkCreateRenderPass vkCreateRenderPass{};
+    PFN_vkCreateRenderPass2 vkCreateRenderPass2{};
     PFN_vkCreateSampler vkCreateSampler{};
     PFN_vkCreateSemaphore vkCreateSemaphore{};
     PFN_vkCreateShaderModule vkCreateShaderModule{};
@@ -1045,6 +1046,8 @@ public:
 
     [[nodiscard]] RenderPass CreateRenderPass(const VkRenderPassCreateInfo& ci) const;
 
+    [[nodiscard]] RenderPass CreateRenderPass2(const VkRenderPassCreateInfo2& ci) const;
+
     [[nodiscard]] DescriptorSetLayout CreateDescriptorSetLayout(
         const VkDescriptorSetLayoutCreateInfo& ci) const;
 
@@ -1526,7 +1529,7 @@ public:
     }
 
     void SetDepthBias(float constant_factor, float clamp, float slope_factor,
-                      VkDepthBiasRepresentationInfoEXT* extra) const noexcept {
+                      const VkDepthBiasRepresentationInfoEXT* extra) const noexcept {
         VkDepthBiasInfoEXT info{
             .sType = VK_STRUCTURE_TYPE_DEPTH_BIAS_INFO_EXT,
             .pNext = extra,